okf-wiki 1.1.1
First public release, published as one snapshot. What changed between versions is in CHANGELOG.md. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
This commit is contained in:
commit
0fd86a23a3
227 changed files with 96077 additions and 0 deletions
182
gates/ANSWERSET.md
Normal file
182
gates/ANSWERSET.md
Normal file
|
|
@ -0,0 +1,182 @@
|
|||
# Answer set for collection 1 (wave 2, order 6)
|
||||
|
||||
The frozen question set for `PLAN.md` criterion 4 on collection 1, made from the fresh frozen copy (contract K1) so
|
||||
that it resembles the collection: questions that need two or more documents seen together, a group of version-history
|
||||
questions reported on its own, and 20 questions held out outside the repository. It replaces the E2 set
|
||||
(`MULTIDOC.md`) for the verdict, because 33 of E2's 60 rest on release notes alone. `answerset.py` holds the gate and
|
||||
`answers.py` the harness the answer measurements use (contract K5). None of them calls a model or starts one.
|
||||
|
||||
```
|
||||
python3 gates/answerset.py check --set <set.json> --sources <K1 root> [--per-question] [--min-paths 3 --min-many 20]
|
||||
python3 gates/answerset.py inside --set <set.json> --part-root <K1 subset root> [--require-all]
|
||||
python3 gates/answerset.py split --set <all.json> --sources <K1 root> --seed <n> --held-out-count 20 \
|
||||
--set-out gates-local/answerset/set-60.json --set-label gates-local/answerset/set-60.json \
|
||||
--held-out <file OUTSIDE the repository> --part gates-local/answerset/part-del1.json \
|
||||
--part-sources <part root> --part-label gates-local/answerset/part-del1.json --manifest gates/answerset.manifest.json
|
||||
python3 gates/answerset.py freeze --set <drafts.json> --sources <K1 root> --seed <n> --count 30 --min-paths 3 \
|
||||
--min-many 20 --prefix x --set-out <set.json> --set-label <label> --manifest <manifest.json>
|
||||
python3 gates/answerset.py gate --set <set.json> --manifest gates/answerset.manifest.json --sources <K1 root> \
|
||||
--arm filsok|bundle --answers <answers.json> [--out <results.json>]
|
||||
python3 gates/answerset.py stub --set <set.json> --arm filsok|bundle --kind fasit|no-citations --out <answers.json>
|
||||
```
|
||||
|
||||
- **check** reads K1 with its own code (`okfwiki.sources` is not imported). The set names the root it was made from
|
||||
(`sources.frozen_at` and `sources.manifest_sha256`, the sha256 of the root's `source-manifest.json`), and every cited
|
||||
file must be listed in K1 with its sha256. It prints the denominators from the set itself: spans that stand byte for
|
||||
byte in their file (20-400 characters), questions with every span standing, questions with at least two distinct
|
||||
paths, questions whose `group` is the one K1 and the path count give, questions where every path carries an answer
|
||||
key inside that path's spans, wording keys (`keyform.py`, as in E2c) and negated keys that stand negated (E2d). PASS
|
||||
only when every count is N of N with N > 0.
|
||||
- **The group is set by the sources:** `history` when one path has `main_topic: false` in K1 (a release note or a
|
||||
weekly What's New page), otherwise `docs-2-3` (2-3 distinct paths) or `docs-4` (4 or more).
|
||||
- **inside** counts the questions whose every span stands in its file in a part root (a K1 subset root). On the 60 it is
|
||||
a number for the report; on the part set it is a requirement (`--require-all`).
|
||||
- **split** checks both the set and the part first (PASS required, and every part question inside the part root),
|
||||
then holds out per group in the fixed order `docs-2-3`, `docs-4`, `history`, with quotas proportional to the group
|
||||
sizes (largest remainder) and one `random.Random(seed)` drawing `sample(sorted(ids of the group), quota)`. Both parts
|
||||
are renumbered in the order of the original ids (`d01..`, `h01..`). The held-out file must lie outside the
|
||||
repository and is never overwritten. The manifest holds the seed, counts per group, sha256 values and the two roots'
|
||||
K1 sha256, and no text.
|
||||
- **gate** scores one arm with the gates' one key test (`multidoc.key_holds`): a question is correct when the answer
|
||||
holds every key and the citations quote every needed path verbatim (whitespace collapsed; `path` is the K1 path,
|
||||
and a leading `/` as in a page's `resource` is accepted). The set's sha256 must be one of the manifest's parts, and
|
||||
the root's K1 sha256 must be that part's root (the copy for the set and the held out, the first part for the part
|
||||
set). A missing answers file scores 0 of N, RED, exit 1 (fail-closed). `--out` writes the results file of K5 point 5,
|
||||
sorted by id. GREEN only when every question is correct; exit 0 when every question was answered.
|
||||
- **stub** writes a known arm: `fasit` (every key as its own sentence, "not <key>" when negated, and every span quoted)
|
||||
and `no-citations` (the same answer, no citations).
|
||||
|
||||
## Frozen 09.10.2026 (wave 2, order 6)
|
||||
|
||||
Written from the frozen copy of 08.10.2026 (635 sources, K1 `manifest_sha256` `8c9f87fd…`) by helpers in the user's
|
||||
session (Opus), each question from a pack of byte-exact source lines that code prepared around rare names shared by an
|
||||
anchor page and its candidates, spread over the sections of the page overview in proportion to their text. A separate
|
||||
helper that did not write the question read every question against windows around its spans: the keys (facts a correct
|
||||
answer must hold) and whether every file is necessary, helped by a code pre-check over all 635 files for another file
|
||||
that holds the keys of two of them. Questions it marked FIX were repaired and, when a file was not necessary, rebuilt
|
||||
and read again; the rest were dropped.
|
||||
|
||||
| Part | Questions | docs-2-3 | docs-4 | history | Confirmed by a separate reader |
|
||||
|---|---:|---:|---:|---:|---:|
|
||||
| all, before the split | 80 | 35 | 34 | 11 | 80 of 80 |
|
||||
| `set-60.json` | 60 | 26 | 26 | 8 | |
|
||||
| held out (outside the repository) | 20 | 9 | 8 | 3 | |
|
||||
| `part-del1.json` (first part only, `p01`-`p18`) | 18 | 9 | 9 | 0 | 18 of 18 |
|
||||
|
||||
- **check:** PASS on the 80 (389 of 389 spans, 0 of 391 keys wording, 13 of 13 negated keys negated) and on the part
|
||||
against the first part's root (91 of 91 spans, 0 of 86 wording, 5 of 5 negated); PASS again on `set-60.json`.
|
||||
- **History:** 11 of 80 (13.8 %); the release notes and weekly What's New pages are 12.45 % of the copy's characters
|
||||
(1,149,271 of 9,231,505). The first part holds no history source, so the part set has none.
|
||||
- **split:** seed 20261009; held out per group docs-2-3 9, docs-4 8, history 3.
|
||||
- **inside** the first part: 1 of the 60; 18 of 18 of the part set.
|
||||
- **Proof, no model:** without an answers file 0 of 60, RED, exit 1; the fasit stub 60 of 60, GREEN; right answers
|
||||
without citations 0 of 60 (correct answer 60 of 60). On the part set: 0 of 18 RED exit 1, 18 of 18 GREEN, 0 of 18.
|
||||
|
||||
## The harder set on the first part (wave 4, order 21, frozen 12.10.2026)
|
||||
|
||||
Wave 3 measured the 18 part questions near the ceiling (both arms 17 of 18), so order 21 made a harder set from the
|
||||
first part's 38 files: every question needs at least 3 documents and at least 20 of 30 need 4 or more, every file is
|
||||
necessary, and at least one key is a condition or relation that only two of the files give together. None of the 18
|
||||
is reused.
|
||||
|
||||
- **check** with `--min-paths N` counts the questions with N or more distinct paths and with `--min-many M` requires
|
||||
at least M questions with 4 or more (`questions with >= 4 distinct paths: x of n (required >= M)`). The defaults
|
||||
(2 and 0) print and judge exactly as before, so the order 6 sets keep their verdict.
|
||||
- **freeze** freezes a set that `split` did not make: it refuses an existing set or manifest, a set that fails the
|
||||
check with the thresholds, or a set with another count than `--count`; it shuffles the sorted ids with
|
||||
`random.Random(seed)`, renumbers them `x01..`, and writes a manifest of its own (seed, rules, counts per group, the
|
||||
set's sha256, the root's and its copy's K1 sha256; no text) that names the set the part of its root, so `gate`
|
||||
scores it as it scores the part.
|
||||
- **Made** by helpers in the user's session (Opus) from packs of byte-exact lines that code built around backticked
|
||||
names standing in 2-6 of the 38 files; one draft helper per question; a separate review helper per question that
|
||||
did not write it read the question against windows around its spans, a code pre-check over all 38 files, and
|
||||
confirmed every file necessary and the join; FIX was repaired (a repair that changed files was read again), DROP
|
||||
dropped. Two questions sharing two keys in one file count as overlap, and the higher id goes. A surplus over 30
|
||||
keeps every docs-4 and drops docs-2-3 with the highest id.
|
||||
|
||||
| `gates-local/answerset/part-del1-hard.json` (`x01`-`x30`) | Questions | docs-2-3 | docs-4 | Confirmed by a separate reader |
|
||||
|---|---:|---:|---:|---:|
|
||||
| frozen, seed 20261011 | 30 | 4 | 26 | 30 of 30 |
|
||||
|
||||
- **check** against the first part with `--min-paths 3 --min-many 20`: PASS, 30 of 30 with 3 or more paths, 26 of 30
|
||||
with 4 or more; **inside** the first part 30 of 30.
|
||||
- **Proof, no model, both arms:** without an answers file 0 of 30, RED; the fasit stub 30 of 30, GREEN; right answers
|
||||
without citations 0 of 30 (correct answer 30 of 30), RED.
|
||||
|
||||
```
|
||||
python3 gates/answerset.py gate --set gates-local/answerset/part-del1-hard.json \
|
||||
--manifest gates/answerset-del1-hard.manifest.json --sources <first part root> --arm filsok|bundle --answers <file>
|
||||
```
|
||||
|
||||
## The harness (`answers.py`, K5 points 2-7)
|
||||
|
||||
```
|
||||
python3 gates/answers.py setup --run <dir> --name <run> --arm filsok|bundle --set <set.json> --sources <K1 root> \
|
||||
[--bundle <collection>] [--part-root <the K1 root the set was made from>]
|
||||
python3 gates/answers.py collect --run <dir> --arm <arm>
|
||||
python3 gates/answers.py harvest --run <dir> --name <run> --arm <arm> --transcripts <dir>
|
||||
python3 gates/answers.py score --results <filsok results> --results <bundle results> [--tokens <file>]... [--out <file>]
|
||||
```
|
||||
|
||||
- **setup** copies every K1 source of the root into `<run>/arms/<arm>/kilde/` (checked against K1; for `bundle` the
|
||||
collection goes to `<run>/arms/bundle/wiki/` beside it) and writes `<run>/tasks/<arm>/<id>.md`: the question, the
|
||||
basis and the answer path, never the keys, the spans or the number of documents. It prints one line per task,
|
||||
`S1-<run>-<arm>-<id>` and the task path: the subagent's description and the file it is given.
|
||||
The set and `--sources` must have the same K1, unless `--part-root` is given (wave 5, order 31 A3): then the set
|
||||
names the K1 of the part root, the part's `subset_of.manifest_sha256` is the K1 of `--sources`, and every source of
|
||||
the part stands in `--sources` with the same path and sha256; otherwise exit 1 with the reason. The arm then searches
|
||||
the whole of `--sources`, while `answerset.py gate` still judges the set against the part root.
|
||||
- **collect** gathers `<run>/answers/<arm>/<id>.json` into `<run>/answers/<arm>.json` (K5 point 4; the id comes from the
|
||||
file name).
|
||||
- **harvest** reads every `*.jsonl` under the transcripts directory: a subagent is matched by the description in its
|
||||
`.meta.json`, or, with no description, by the task path in its first prompt. Per question it sums the four usage
|
||||
fields over all runs (each message id once, at its largest value), adds `output_tokens_derived` (a lower bound for
|
||||
an output the transcript never recorded, as in `okfwiki/cost.py`), and counts isolation breaches as in
|
||||
`evals/direction_probe.py`: a tool call that reads or lists a path outside the arm's basis and the task file, a tool
|
||||
that is not allowed, or a write to another file than the answer. A task without a transcript exits 1.
|
||||
**`echo` (wave 5, order 31 A2):** in a Bash chain that starts with `cd <basis> &&`, `echo` with literal words is not
|
||||
a breach (order 13's helpers printed `---` between searches: 21 of its 26 counted breaches, classified call by call
|
||||
before the change). Its arguments are checked as the search commands' are, so a path outside the basis, `~`, `..`,
|
||||
a redirection, `$` or a backtick is still a breach. The task text does not offer `echo`. No other rule was made:
|
||||
order 13 had no backtick or `$` inside single quotes and no `Read` of the helper's own answer file (its three checks
|
||||
of the answer file were `python3` and `jq` in Bash, which stay breaches).
|
||||
**Four more kinds (wave 5, order 31 A2b, operator 19.10.2026):** the price measurement B1 counted 9 breaches in 4 of
|
||||
4 questions, none of which read a source outside the basis. Each kind has a test that failed first, with the
|
||||
opposite cases that stay breaches:
|
||||
1. `2>/dev/null` right before a separator (`;`, `|`, `&&`, `||`) is not a breach, as before a space or at the end.
|
||||
Every other redirection (`> file`, `>/dev/null`, `2>&1`, `2>` anywhere else) still is.
|
||||
2. The search word of `grep`/`egrep`/`fgrep`/`rg` (the value of `-e`/`--regexp`, or else the first argument that is
|
||||
not an option) and the script of `sed` (the value of `-e`/`--expression`, or else the first such argument) are not
|
||||
paths. With `-f`/`--file` the search words come from a file, none is positional, and the file is checked as a path.
|
||||
A file path after the search word, `~` and `..` are still breaches. **Order 39:** the value of every other option
|
||||
that takes one (`grep -C`, `--include`, `--exclude-from`, `rg --glob`, `--ignore-file` and the rest, GNU and BSD)
|
||||
is consumed and checked as a path, so a file option's value outside the basis is a breach, and `rg --pre` or
|
||||
`--hostname-bin` (which run a program) is one. An option missing from the table in `SEARCH_WORDS` makes its value
|
||||
the search word, and a path there goes unchecked, so the lists must be whole. `grep --context` with a separate
|
||||
value is left out, because BSD grep takes its value only after `=` (`--context 1 /x f` searches for 1 in `/x`).
|
||||
A `sed` script is read command by command instead of checked as a path: `w`, `W` and `e` (also as flags of `s`)
|
||||
are breaches, the file after `r` or `R` is checked as a path, and a script that cannot be parsed is a breach. A
|
||||
flag cluster with `i` or `I` (`-ni`, `-Ei`, `-I`) is `sed -i`. An unquoted newline separates commands as `;`
|
||||
does, and `#` is never read as a comment (bash reads it so only at the start of a word; a real comment is then
|
||||
checked as arguments, which is stricter).
|
||||
3. `$` or a backtick inside single quotes is not an expansion. Unquoted or inside double quotes it still is, and so is
|
||||
an escaped backtick inside double quotes (order 31 A2 classed order 13's one as real).
|
||||
4. `Read` of the helper's own answer file is not a breach, and neither is a Bash command of the read-only commands
|
||||
plus `jq` whose only path outside the basis is the answer file: after `cd <basis> &&`, after `cd` into the answer's
|
||||
folder, or with no `cd` and the answer's absolute path. Outside the basis every operand must be the answer file, so
|
||||
another file in the answers folder, a glob, `ls` of the folder, a relative path without `cd` and `jq -f` are still
|
||||
breaches, as are `python3`, a write and `jq` on any other file. Outside the basis a pipe stage that names no
|
||||
file, as in `jq . <answer> | head`, is counted too (stricter than needed). The task text does not offer `jq`.
|
||||
Against order 13 the A2b rules move 2 of A2's 5 real calls (one `jq` on the own answer after `cd` into its folder,
|
||||
one `sed` script starting with `/`) and leave 3; on B1 they leave 1 of 9 (an escaped backtick in double quotes).
|
||||
- **score** gives per arm and group "x of n", by path count (2-3 and 4 or more, history included), the pairs (both,
|
||||
only filsok, only bundle, neither) with the two-sided exact McNemar p, the difference bundle - filsok with its paired
|
||||
95 % interval (Newcombe 1998, Statist. Med. 17:2635-2650, method 10, tested against its Table III), and the smallest
|
||||
difference the set can show at 80 % power and alpha 0.05 with the measured share of discordant pairs (Connor 1987;
|
||||
tested against the sample sizes in `PLAN.md § Styrke`). It prints no line per question.
|
||||
|
||||
The questions and spans derive from Anthropic documentation, so the set lives in `gates-local/answerset/` and the held
|
||||
out outside the repository (CLAUDE.md rule 3). Only the tools, their tests (synthetic fixtures) and the manifest are
|
||||
tracked. The held-out questions are never opened during development; the manifest's sha256 is the check.
|
||||
|
||||
Tests: `python3 -m unittest discover -s gates/tests`. Types: `mypy --strict gates/`.
|
||||
173
gates/CALIBRATION.md
Normal file
173
gates/CALIBRATION.md
Normal file
|
|
@ -0,0 +1,173 @@
|
|||
# Reader calibration (order 3: 300 of 300, calls of up to ten claims)
|
||||
|
||||
The frozen calibration set for `PLAN.md` criterion 2: before an independent reader may judge the claims on the wiki's
|
||||
pages, it must show that it can tell a claim that follows from its source excerpts from one that does not. From
|
||||
06.10.2026 the reader judges up to ten claims per call and gives one verdict per claim (`CLAUDE.md` § The shape; K3 in
|
||||
the wave-1 contracts), so the calibration measures the reader in calls, not one claim at a time. `calibration.py` holds
|
||||
the tools. None of them calls a model, and the reader pass itself is not here (Step 9 of the v1 plan).
|
||||
|
||||
```
|
||||
python3 gates/calibration.py check --draft <draft.json> --sources <source dir>
|
||||
python3 gates/calibration.py check --set gates-local/calibration/pairs-300.json --key gates-local/calibration/key-300.json \
|
||||
--calls gates-local/calibration/calls-300.json --manifest gates/calibration.manifest.json --sources <source dir>
|
||||
python3 gates/calibration.py build --draft <draft.json> --sources <source dir> --seed <n> --lone 100 --dense 5 \
|
||||
--set-out gates-local/calibration/pairs-300.json --key-out gates-local/calibration/key-300.json \
|
||||
--calls-out gates-local/calibration/calls-300.json --manifest gates/calibration.manifest.json \
|
||||
--set-label <path> --key-label <path> --calls-label <path>
|
||||
python3 gates/calibration.py stub --set <set> (--verdict supported|unsupported | --key <key>) \
|
||||
[--calls <layout> --accept-at <1-10>] --out <verdicts.json>
|
||||
python3 gates/calibration.py gate --set <set> --key <key> --calls <layout> --manifest gates/calibration.manifest.json \
|
||||
--sources <source dir> --verdicts <verdicts.json> [--passed <passed.json>]
|
||||
```
|
||||
|
||||
## The three files and the labels
|
||||
|
||||
- **The set** (`pairs-300.json`) is what the reader sees: per claim an id, the claim and its excerpts (file, line,
|
||||
text). The ids follow the call order and say nothing about the label.
|
||||
- **The key** (`key-300.json`) gives per claim the label, the kind, the draft id and a one-sentence reason.
|
||||
- **The call layout** (`calls-300.json`, schema `okf-wiki/reader-calls/1`): `call_size_max` (10), the calls with their
|
||||
claim ids in position order, and `set_sha256` that binds the layout to the set. Every claim stands in exactly one
|
||||
call; its position is its index in the call plus one (1-10).
|
||||
|
||||
Labels: `unsupported` (planted, must be caught), `supported` (counted controls, must be accepted) and
|
||||
`filler-supported`. Fillers are supported claims that fill the calls so that most calls look like a real page: one
|
||||
wrong claim, if any, among right ones. They are checked like supported claims (verbatim excerpts, not a copy of the
|
||||
excerpt) and audited, are reported apart ("filler accepted x of y"), and never count in the thresholds. A filler may
|
||||
name the unsupported pair it is the twin of (`twin_of`); the build never puts the two in the same call.
|
||||
|
||||
## The tools
|
||||
|
||||
- **check** prints the denominators from the pairs themselves: excerpts that stand verbatim (byte-exact) in their file
|
||||
and, for a built set, at their stated line; pairs per label and per kind. It fails when a kind of the two counted
|
||||
labels has no pair, when a kind does not fit its label, when a `wrong-number` claim has no number that is absent from
|
||||
its excerpts, when a `conflated-sources` pair does not quote two files, when a supported or filler claim is a
|
||||
verbatim copy of its excerpt, or when a `twin_of` does not name an unsupported pair. There is no rule of equal
|
||||
numbers any more: a built set is checked against the manifest instead, which states the counts per label, per kind
|
||||
and per position, the sha256 of set, key, layout and sources, and the layout's own counts. The layout must hold every
|
||||
claim once, no call over ten, the right `set_sha256`, and unsupported claims at every position 1-10. Check prints
|
||||
the layout as a distribution (unsupported per call, unsupported alone in their call, unsupported per position).
|
||||
- **build** checks a labelled draft and lays it out with a fixed seed (the rule is written into the manifest as
|
||||
`shuffle`): `--lone` unsupported pairs stand alone in a call, at positions spread evenly over 1-10, the rest stand
|
||||
`--dense` to a call; the counted supported pairs go one per call, in call order (so into the lone calls first);
|
||||
fillers fill every call to ten, taken in a cycle, never twice in one call and never beside their twin. A filler may
|
||||
therefore stand in several calls under different claim ids; the manifest gives the slots and the distinct fillers.
|
||||
Nothing is written when the layout breaks a rule. A built set is frozen: build refuses to overwrite the set, the key
|
||||
or the layout. An earlier manifest at the same path moves into the new manifest's `history`.
|
||||
- **stub** writes verdicts without a reader: the same verdict for every claim (reject-all, accept-all), or the key's
|
||||
verdicts (the known-positive), and with `--calls` and `--accept-at P` "supported" for every claim at position P (the
|
||||
position-mild stub). A stub names the model `stub`, which no reader transcript names.
|
||||
- **gate** scores one reader's verdicts (`{"reader": "<name>", "models": ["<model>"], "verdicts": [{"id", "verdict"}]}`,
|
||||
a verdict being `supported` or `unsupported`) and prints the model it scored (`model: <name>`), the totals, the
|
||||
caught count per kind, `caught by position: 1 x of y, ..., 10 x of y` and the fillers apart. It never prints or writes a result per call or per pair (K3 point 5; `CLAUDE.md`:
|
||||
the reader is improved only from totals and counts per kind). A missing verdict, an invalid value or an id given
|
||||
twice counts as wrong (fail-closed); a missing verdicts file scores 0. GREEN only when every unsupported claim is
|
||||
caught at every position, at least `accepted_min` counted supported claims are accepted and no verdict is missing,
|
||||
fillers included, and the verdicts name exactly one model; verdicts that name no model, or two or more, are RED
|
||||
(fail-closed). A RED run ends with one `not met:` line that names what failed, positions included. The set, the
|
||||
key, the layout and the sources must match the manifest, or the gate refuses to score.
|
||||
|
||||
## The gate is bound to the reader's model (PLAN § Rammer 23, 31.10.2026)
|
||||
|
||||
The reader's agent file names a model family (`model: opus`), and the family alias moves to a new model when one is
|
||||
released. A pass of the gate holds only for the model that was scored, so the gate and the reading are tied to it:
|
||||
|
||||
1. The gate run's readers are collected with `okfwiki.reader collect ... --calibration`. Collect reads the model of
|
||||
every reader transcript it uses (`message.model` of the assistant records, as Claude Code writes it) and writes the
|
||||
sorted, distinct models into the verdicts file as `models`.
|
||||
2. `gate ... --passed <passed.json>` adds a GREEN run to the record of green runs (schema
|
||||
`okf-wiki/calibration-passed/1`): `{"schema": ..., "passed": [{"model", "set_sha256", "key_sha256",
|
||||
"verdicts_sha256", "reader"}]}`. The record has no time in it, the same entry is not added twice, and a RED run never
|
||||
writes it. A record that cannot be read is a FAIL and is left as it was.
|
||||
3. A reading of the wiki is collected with `okfwiki.reader collect ... --gate-passed <passed.json>`. When the
|
||||
transcripts name no model, or a model that has no entry in the record, collect writes nothing and exits 1. A new
|
||||
model therefore needs a new GREEN gate run before its verdicts count.
|
||||
|
||||
Thresholds (in the manifest): `caught_min` is every unsupported claim (the operator, 06.10: 300 of 300).
|
||||
`accepted_min` is 90 % of the counted supported claims, rounded up (45 of 50), chosen by the PM in E3 because a reader
|
||||
that rejects more than one supported claim in ten makes the writing step unusable.
|
||||
|
||||
Kinds of unsupported claim: `overgeneralisation`, `wrong-number`, `swapped-cause`, `not-mentioned`, `reversed-polarity`,
|
||||
`conflated-sources` (two excerpts from two files, one subject's property given to the other). Kinds of supported and
|
||||
filler claim: `paraphrase`, `inference` (one reasoning step from one excerpt), `synthesis` (needs both excerpts).
|
||||
|
||||
The pairs derive from Anthropic documentation, so the set, the key, the layout and the drafts live in `gates-local/`
|
||||
(CLAUDE.md rule 3). Only the tools, their tests (synthetic fixtures) and the manifest are tracked. The reader is shown
|
||||
the set and the layout only, never the key or the drafts.
|
||||
|
||||
Tests: `python3 -m unittest discover -s gates/tests`. Types: `mypy --strict gates/`.
|
||||
|
||||
## The set (order 3, 08.10.2026)
|
||||
|
||||
Built with seed 20261008, `--lone 100 --dense 5`, against the 599 sources of the copy of 21.09 (the same
|
||||
`tree_sha256` as E3). 1,400 claims in 140 calls of ten:
|
||||
|
||||
- **300 unsupported**, 50 of each kind: E3's 50, and 250 new ones from 150 docs pages E3 did not use (277 of the 300
|
||||
quote a docs page first, 92.3 %; collection 1 is 92.2 % documentation by text). Each new one was written as one small
|
||||
change from a supported reading of its excerpt (its twin).
|
||||
- **50 supported** that count: E3's 50 (paraphrase 15, inference 20, synthesis 15), each once, all in lone calls.
|
||||
- **1,050 filler slots** from 250 distinct fillers: the twins of the 250 new unsupported pairs, each used 3 to 5 times
|
||||
under different claim ids, never in the call of its own unsupported pair. The twins cost almost nothing beyond the
|
||||
error itself, and they never count in the thresholds.
|
||||
- **Layout:** 100 calls with one unsupported claim among nine supported ones (like a real page), 40 calls with five.
|
||||
Unsupported claims per position 1-10: 33, 30, 28, 23, 32, 23, 30, 24, 39, 38.
|
||||
|
||||
Authoring and audit: 25 batches of ten unsupported pairs and their twins, each written by one helper (Opus) from six
|
||||
pages and audited by another with the key visible, judging against the excerpts only. 500 new pairs: 455 OK, 45 FIX,
|
||||
0 WRONG; every FIX took the auditor's wording (most were twins too close to the excerpt's wording, with the unsupported
|
||||
claim reworded to stay one change from the new twin). `check` on the built set: 1,642 of 1,642 excerpts verbatim and
|
||||
at their line.
|
||||
|
||||
Proofs without a reader (verdicts written before 31.10.2026 name no model and are RED under the model rule; a stub
|
||||
written again names the model `stub`): no verdicts 0 of 300, RED; reject-all 300 of 300 and 0 of 50, RED; accept-all 0 of 300, RED;
|
||||
the key accepting everything at one position RED at each of the ten positions, with that position named; the key as
|
||||
judge GREEN.
|
||||
|
||||
## The folder set (motor B16, 08.11.2026)
|
||||
|
||||
The set above is Claude Code documentation and stays out of git. A folder collection (Norwegian and English PDF and
|
||||
Office documents) needs its own measurement, so a second set is tracked in full: `gates/calibration-folder/` holds
|
||||
the documents (`original/`), the text the engine froze from them (`kilde/`, the reader's sources), the draft, the
|
||||
set, the key and the layout; `gates/calibration-folder.manifest.json` is its manifest. The same tools score it:
|
||||
|
||||
```
|
||||
python3 gates/calibration.py gate --set gates/calibration-folder/pairs.json --key gates/calibration-folder/key.json \
|
||||
--calls gates/calibration-folder/calls.json --manifest gates/calibration-folder.manifest.json \
|
||||
--sources gates/calibration-folder/kilde --verdicts <verdicts.json> [--passed gates/calibration.passed.json]
|
||||
```
|
||||
|
||||
- **The documents:** 32 written for the set, four per invented organisation (a housing cooperative, a sports club, a
|
||||
school band, a community farm in Norwegian; a makerspace, an open-source project, an astronomy society, a sailing
|
||||
club in English): per language 4 PDF, 4 Word, 4 Excel, 4 PowerPoint. No text from Anthropic or any other source.
|
||||
Written by eight helpers (Opus) from one brief, rendered with python-docx, openpyxl and python-pptx, PDF through
|
||||
LibreOffice from a Word file, and frozen with `python3 -m okfwiki.sources folder` (32 sources, 0 skipped; `lang`
|
||||
16 no and 16 en, all right).
|
||||
- **The pairs:** 24 batches, three per organisation, each from its own third of every document (spreadsheet parts keep
|
||||
every sheet's title and header row). 300 unsupported (50 per kind), each with its twin as a filler (300), and 50
|
||||
counted supported (paraphrase 15, inference 20, synthesis 15). By the document type of the first excerpt: docx 81,
|
||||
pdf 78, pptx 71, xlsx 70. Audit 1 (one helper per batch, key visible, excerpts only): 487 OK, 163 FIX, 0 WRONG;
|
||||
most fixes took out names or details that stand in the document but not in the excerpt. Audit 2 read the 163 fixed
|
||||
pairs and their twins again: 140 OK, 33 FIX, 0 WRONG. One more pair rested on a word only in the file name, fixed.
|
||||
- **Built** with seed 20261108, `--lone 100 --dense 5`: 1,400 claims in 140 calls, unsupported per position 1-10:
|
||||
32, 33, 25, 27, 30, 32, 28, 31, 33, 29. `check`: 1,666 of 1,666 excerpts verbatim and at their line.
|
||||
- **Proofs without a reader:** no verdicts, reject-all, accept-all RED; the key accepting everything at one position
|
||||
RED at each of the ten positions, with that position named; the key GREEN.
|
||||
- **First reading (run `kalmappe1`, 08.11.2026, claude-opus-5-5): RED.** Unsupported caught 299 of 300 (wrong-number
|
||||
49 of 50, position 5 29 of 30, every other kind and position in full), supported accepted 50 of 50, fillers 1,045
|
||||
of 1,050, 140 of 140 calls answered. The one miss was checked against the key (operator 08.11, the one exception
|
||||
to judging the reader from totals only): the pair was right, the claim gave a difference of two figures that the
|
||||
excerpt did not print. The reader's number check now says to work out a derived figure itself before judging.
|
||||
- **Second reading, both sets, new instructions (runs `kalmappe2`, `kaldoc2`, 08.11.2026, claude-opus-5-5, as one
|
||||
workflow): GREEN on both.** Folder set: caught 300 of 300, accepted 50 of 50, fillers 1,047 of 1,050. Docs set:
|
||||
caught 300 of 300, accepted 49 of 50, fillers 1,047 of 1,050. Both runs are in `gates/calibration.passed.json`.
|
||||
- **Collect needs both:** `okfwiki.reader collect --gate-passed` checks every manifest given with `--manifest`, by
|
||||
default both tracked ones, so a model reads a collection only after a green run on each set.
|
||||
|
||||
## History: E3 (27.09.2026)
|
||||
|
||||
The first set had 100 pairs of (claim, excerpt) judged one claim per call: 50 planted unsupported claims
|
||||
(overgeneralisation 9, wrong-number 9, swapped-cause 8, not-mentioned 8, reversed-polarity 8, conflated-sources 8) and
|
||||
50 supported controls (paraphrase 15, inference 20, synthesis 15), from 90 source files of the copy of 21.09 (42 docs
|
||||
pages, 48 release notes). Five authors wrote the pairs and two auditors read every label with the key visible: 93 OK,
|
||||
7 FIX, 0 WRONG. The thresholds were 50 of 50 and 45 of 50, and `check` required equal numbers of the two labels. Its
|
||||
manifest values stand under `history` in `calibration.manifest.json`; its files (`draft-100.json`, `pairs-100.json`,
|
||||
`key-100.json`) stay in `gates-local/calibration/` untouched. Its 100 pairs are reused in the new set.
|
||||
79
gates/MULTIDOC.md
Normal file
79
gates/MULTIDOC.md
Normal file
|
|
@ -0,0 +1,79 @@
|
|||
# Multi-document set (E2)
|
||||
|
||||
The frozen question set for `PLAN.md` criterion 4 on collection 1: questions that each need at least four
|
||||
different source documents seen together, written before the collection exists. `multidoc.py` holds the
|
||||
three tools. None of them calls a model.
|
||||
|
||||
```
|
||||
python3 gates/multidoc.py check --set gates-local/multidoc/set-60.json --sources <source dir> [--per-question]
|
||||
python3 gates/multidoc.py gate --set gates-local/multidoc/set-60.json --manifest gates/multidoc.manifest.json \
|
||||
--sources <source dir> --arm <name> --answers <answers.json> [--out <results.json>]
|
||||
python3 gates/multidoc.py split --set <all.json> --sources <source dir> --seed <n> --held-out-count 20 \
|
||||
--dev-out gates-local/multidoc/set-60.json --dev-label gates-local/multidoc/set-60.json \
|
||||
--held-out <file OUTSIDE the repository> --manifest gates/multidoc.manifest.json
|
||||
```
|
||||
|
||||
- **check** prints the denominators from the set itself: spans that stand verbatim (byte-exact) in their file,
|
||||
questions whose every span does, questions with at least four distinct files, and questions where every file
|
||||
carries an answer key that stands inside that file's spans (case-insensitive), the answer keys that are
|
||||
wording, and the negated keys that stand negated in their file's spans (below). PASS only when all of them are
|
||||
N of N and N > 0, no key is wording, and every negated key stands negated.
|
||||
- **gate** scores one arm. A question counts when the answer contains every answer key (case-insensitive,
|
||||
backticks and asterisks ignored, whitespace collapsed; a number key matches as a number and a negated key only
|
||||
with a negation before it, below) AND the citations quote every needed file verbatim
|
||||
(whitespace collapsed). A missing answers file scores 0 of N and exits 1 (fail-closed). The set and the source
|
||||
directory must match the manifest's sha256, or the gate refuses to score. `--out` writes per-question results
|
||||
for a paired test (McNemar) between two arms.
|
||||
- **split** draws the held-out questions with `random.Random(seed).sample(sorted(ids), 20)`, renumbers both parts
|
||||
(`d01..`, `h01..`) so the ids tell nothing about the draw, writes the held-out part outside the repository and
|
||||
refuses to overwrite it. The manifest holds only the seed, counts and sha256 values.
|
||||
|
||||
## Answer keys are facts, not wording (E2c)
|
||||
|
||||
An answer key is the smallest fact the answer must hold: a name, a command, a flag, a setting, a number, a version,
|
||||
or one word that carries the meaning. `keyform.py` counts the keys that are wording: two or more words with a
|
||||
function word outside backticks (an article, a pronoun or determiner, a conjunction, a preposition, an auxiliary or
|
||||
modal verb, a negation, or a focusing particle such as "only"). Such a key is split into the runs of its own words
|
||||
without the function words, and each run must still stand in the span: "/vim or /config" becomes `/vim` and
|
||||
`/config`, and "pin your model versions" becomes `pin` and `model version`. A key that is a whole number, a
|
||||
percentage or an English number word matches as a number: its digits or its word, never inside a longer number,
|
||||
version or identifier (`5` does not match `15` or `2.1.5`), and the 3-character minimum does not apply to it.
|
||||
|
||||
```
|
||||
python3 gates/keyform.py --set <set.json> [--set <set.json> ...] [--list]
|
||||
```
|
||||
|
||||
On 27.09.2026 (E2c) the four frozen parts held 104 wording keys of 495; after the rewrite they hold 0 of 523. Only
|
||||
the keys changed. The manifests record the sha256 values from before under `keys_rewritten`.
|
||||
|
||||
## Negated keys (E2d)
|
||||
|
||||
Where a negation carried the fact ("not synced back" became `synced`), a key without it also accepts the opposite
|
||||
answer. Such a key is marked `"negated": true`. It holds only where the answer has a negation at most six words
|
||||
before it in the same sentence (a sentence ends at a line break or at `.`, `!`, `?` or `;` ending a word). The
|
||||
negations are a fixed list: `not`, `no` (so also "no longer"), `never`, `none`, `neither`, `nor`, `without`, `cannot`,
|
||||
and every word ending in `n't` ("isn't", "doesn't", "can't", "don't"; a curly apostrophe counts too). The same
|
||||
function, `key_holds` in `multidoc.py`, scores both gates. A key without the mark holds as before, with or without a
|
||||
negation. `check` requires every negated key to stand negated in one of its file's spans, so the mark is true to the
|
||||
source. The `fasit` answer writes a negated key as "not <key>".
|
||||
|
||||
`polarity.py` counts the polarity keys that accept an answer of the opposite polarity. The polarity keys are listed
|
||||
per set part (`{"schema": "okf-wiki/polarity-keys/1", "keys": [{"id", "atoms"}]}`). For each, the answer from the span is
|
||||
the sentence of the key's spans where it stands negated, else the first sentence that holds it; the opposite answer is
|
||||
that sentence with the negations within the window taken out, or with "not" put in front of the key when there are none.
|
||||
|
||||
```
|
||||
python3 gates/polarity.py --set <set.json> --keys <keys.json> [--set ... --keys ...] [--list]
|
||||
```
|
||||
|
||||
On 27.09.2026 (E2d) the list was E2c's own 22 polarity keys over the four parts. Before the mark, 22 of 22 accepted the
|
||||
opposite answer. The mark went on the 17 that stand negated in their own spans; after it, 5 of 22 accept the opposite
|
||||
answer. Those 5 carry a restriction, not a negation ("only" three times, "still", "keeps ... off"): nothing stands
|
||||
before them to require, and the rule does not require the absence of a negation. The manifests record the sha256
|
||||
values from before under `negated_marked`.
|
||||
|
||||
The questions and spans derive from Anthropic documentation, so they live in `gates-local/` and outside the
|
||||
repository (CLAUDE.md rule 3). Only the tools, their tests (synthetic fixtures) and the manifest are tracked.
|
||||
The held-out questions are never used during development (R54).
|
||||
|
||||
Tests: `python3 -m unittest discover -s gates/tests`. Types: `mypy --strict gates/`.
|
||||
238
gates/README.md
Normal file
238
gates/README.md
Normal file
|
|
@ -0,0 +1,238 @@
|
|||
# gates
|
||||
|
||||
## Mutant gate (E1)
|
||||
|
||||
`mutant_gate.py` runs an auditor command on every case in `mutants.manifest.json`: 37 mutants, each carrying a
|
||||
defect of a named class, plus one known-positive bundle that must come back clean. It calls no model.
|
||||
|
||||
E1b added three link classes: `broken-link`, `orphan-page` and `index-coverage`. They are okf-wiki's own checks,
|
||||
not SPEC § 11 conformance: SPEC § 6.1 says consumers MUST tolerate broken links. Their exact definitions are
|
||||
in the manifest under `provenance.link_rules`, with criterion 6 as it stands (see below).
|
||||
|
||||
```
|
||||
python3 gates/mutant_gate.py --manifest gates/mutants.manifest.json --cases gates-local/mutants \
|
||||
--sources gates-local/mutants-kilder -- <auditor> {bundle} {sources}
|
||||
```
|
||||
|
||||
- `{bundle}` becomes a fresh copy of the case bundle, and `{sources}` becomes the raw-source root.
|
||||
- The auditor prints `{"findings": [{"check": "<class>", ...}]}` and exits 0 when clean or 1 when it raises an alarm.
|
||||
Anything else counts as an error and catches nothing: a missing auditor, a crash, a timeout, output that is not
|
||||
JSON, or an exit status that disagrees with the findings.
|
||||
- The gate is GREEN only when every mutant is caught **with its own class**, the known-positive is clean and
|
||||
nothing errored. Rejecting everything catches every mutant, but it raises a false alarm and matches no class.
|
||||
- Before anything runs, every case file and every raw source listed in the manifest is checked against its
|
||||
sha256. A mismatch stops the gate with exit 2.
|
||||
|
||||
The sources root is `gates-local/mutants-kilder/` (wave 1 order 4): byte-identical copies of the three documentation
|
||||
files the pages cite and one release note, under `kilde/`, with `source-manifest.json` in the form of contract K1
|
||||
(`docs/bolge1/0-felles.local.md` § 4, local). The manifest lists all five files with their sha256. Until 08.10.2026
|
||||
the root was the copy of 21.09 (`~/okf/benk/wiki`); that root has no K1 file, so the gate now stops on it with exit 2.
|
||||
|
||||
The auditor is `okfwiki/audit.py` (L1a). With it the gate was GREEN at 23 of 23 class matched and 0 of 1 false alarms
|
||||
before V1 added the link-rule mutants, and GREEN at 29 of 29 and 0 of 1 on 08.10.2026 with the manifest of that day and
|
||||
`--sources ~/okf/benk/wiki`, before the main-page rule below. The command today (GREEN since wave 2 order 7, see below):
|
||||
|
||||
```
|
||||
python3 gates/mutant_gate.py --manifest gates/mutants.manifest.json --cases gates-local/mutants \
|
||||
--sources gates-local/mutants-kilder -- python3 -m okfwiki.audit '{bundle}' '{sources}'
|
||||
```
|
||||
|
||||
In zsh, quote the placeholders as above when the output is redirected: `{sources} > file` opens a named file descriptor,
|
||||
the auditor gets only `{bundle}`, and every case errors.
|
||||
|
||||
### Criterion 6 under the main-page rule (wave 1 order 4, RED)
|
||||
|
||||
Criterion 6 in `PLAN.md § Ferdig-kriteriet` as the operator set it on 06.10.2026 (choice 2): Google's rule as
|
||||
written, applied to documents. Every documentation page has its main page, and every other concept page has a link from
|
||||
at least one main page. Links are made by the model from meaning; the auditor checks structure only, without a model.
|
||||
The exact definitions are in the manifest under `provenance.link_rules`, and the contract is K2:
|
||||
|
||||
- A concept page lists the sources it is the main page for in its frontmatter, as a YAML list under `x-main-for`. A
|
||||
valid value is `/` + a path with `main_topic: true` in the K1 file at the sources root, and stands in the page's own
|
||||
`sources[].resource`. Without a valid K1 file the rule cannot run (exit 2, never 0).
|
||||
- `main-page-missing`: a main-topic source with no main page. `main-page-duplicate`: two main pages for one source.
|
||||
`main-page-invalid`: not a list, a value not among the page's own sources, or a value that is not a main-topic source.
|
||||
- `no-main-page-link`: a page that is not a main page has no link from a main page's body, in a sentence or in a list
|
||||
of related concepts. `index.md`, `log.md` and the page itself do not count. A main page needs no inbound link.
|
||||
- Unchanged: (a) `broken-link`, (b) `link-misplaced` (a link in code, a heading or a footnote definition),
|
||||
`orphan-page`, `index-coverage`. `link-rule` (an inbound link from any other concept page, the rule of 04.10.2026)
|
||||
is retired in Step 7; no mutant carries it any more.
|
||||
|
||||
The known-positive (normalisation N7): `dontask-modus.md` is the main page of all three documentation sources, and
|
||||
has no inbound link from another concept page (only from `index.md`). Two non-main pages: `bypasspermissions-modus.md`
|
||||
(linked from a sentence on the main page) and `acceptedits-modus.md` (linked from the main page's related-concepts list,
|
||||
and from a sentence on the bypass page; it cites the release note). The main page has exactly one link to each, so a
|
||||
mutant moves or removes "the one link". Every case is rebuilt from this known-positive, so every sha256 changed.
|
||||
|
||||
| Mutant | Class | Defect | Expected extra findings |
|
||||
|---|---|---|---|
|
||||
| m21 | `orphan-page` | acceptEdits no longer linked from any page or `index.md` (its `log.md` entry stays) | `index-coverage`, `no-main-page-link` |
|
||||
| m24 | `no-main-page-link` | bypass linked only from `index.md` | none |
|
||||
| m25 | `link-misplaced` | the main page's one link to bypass inside inline code | `no-main-page-link` |
|
||||
| m26 | `link-misplaced` | the main page's one link to bypass moved into a heading | `no-main-page-link` |
|
||||
| m27 | `link-misplaced` | the main page's one link to bypass moved into a footnote definition | `no-main-page-link` |
|
||||
| m28 | `no-main-page-link` | the only concept-body link to bypass is a self-link (`index.md` and `log.md` still link it) | none |
|
||||
| m29 | `no-main-page-link` | bypass linked only from `log.md` | `index-coverage`, `orphan-page` |
|
||||
| m30 | `no-main-page-link` | acceptEdits linked only from the other non-main page (and `index.md`, `log.md`) | none |
|
||||
| m31 | `main-page-missing` | the main page loses its mark for one documentation source | none |
|
||||
| m32 | `main-page-duplicate` | bypass marks a source the main page marks (and bypass cites) | none |
|
||||
| m33 | `main-page-invalid` | bypass's `x-main-for` is a string, not a list | none |
|
||||
| m34 | `main-page-invalid` | bypass marks a main-topic source it does not cite | none |
|
||||
| m35 | `main-page-invalid` | acceptEdits marks the release note it cites (in K1, not a main topic) | none |
|
||||
| m36 | `main-page-duplicate` | the main page adds an invalid value, and bypass marks a source the main page marks | `main-page-invalid` |
|
||||
| m37 | `no-main-page-link` | bypass marks only a source it does not cite, and the main page's one link to it is removed | `main-page-invalid` |
|
||||
|
||||
m36 and m37 test K2 point 2's last sentence (a page is a main page when at least one value is valid; invalid values do
|
||||
not take that role away). Since the gate counts a mutant matched when its own class is among the findings, each carries a
|
||||
class that only a right reading of the role gives, so each is two edits by design. Measured with two wrong readings of
|
||||
the rule (scratch only): one that makes every page with the key a main page misses m37's class, and one that takes the
|
||||
role from a page with an invalid value misses m36's; on the set without m36 and m37 both missed nothing. Under K2 only
|
||||
links from main pages count, so a self-link on a non-main page can never count: m28 is kept because the order names the
|
||||
case, but it tells K2 auditors apart no better than m24.
|
||||
|
||||
The extras are the classes the builder measures beside the case's own class, as the auditor of Step 7 should report them;
|
||||
they are documentation (`expected_extra` per case), and the gate counts a mutant matched when its own class is among the
|
||||
findings. Two older cases also gain an extra under K2: m11 (`main-page-missing`, `no-main-page-link`: the main page's
|
||||
frontmatter no longer parses) and m12 (`no-main-page-link`: its extra page is linked from `index.md` alone). The
|
||||
manifest maps every K2 case to its mutant under `provenance.k2.coverage`.
|
||||
|
||||
Recorded RED on 08.10.2026 with today's auditor (`okfwiki/audit.py` unchanged), with the command above: **class matched
|
||||
26 of 37**. The old classes are matched 26 of 26 (m01-m23 and m25-m27). The new classes are matched **0 of 11**: m24 and
|
||||
m28-m37 are caught only with the retired `link-rule` (m29 also with `index-coverage` and `orphan-page`). **False alarms
|
||||
1 of 1, by design:** today's auditor reports `link-rule` on the main page of the known-positive, which has no inbound
|
||||
concept link because K2 point 3 does not ask for one; Step 7 retires `link-rule`. Errors 0 of 38, exit 1. Step 7 must
|
||||
turn this into 37 of 37 with 0 false alarms.
|
||||
|
||||
Wave 2 order 7 did (merged into `v1` 09.10.2026, measured in the main copy after the merge): caught 37 of 37 · class
|
||||
matched 37 of 37 · false alarms 0 of 1 · errors 0 of 38, GREEN, exit 0. Since K2 point 15 (a link with `/` does not
|
||||
count), m09 also gives `no-main-page-link` on `acceptedits-modus.md`; the gate counts only each case's own class, and
|
||||
`expected_extra` for m09 in the frozen manifest still says `[]`.
|
||||
|
||||
History: V1 session 2 (05.10.2026) recorded RED at 23 of 29 for the rule of 04.10.2026 (an inbound link from any other
|
||||
concept page, class `link-rule` on m24, m28 and m29), and Step 6 of the v1 plan turned it GREEN at 29 of 29.
|
||||
|
||||
The cases and their sources derive from Anthropic documentation, so they live in `gates-local/` (gitignored, CLAUDE.md
|
||||
rule 3). Only the runner, its tests, the reject-all stub and the manifest (names, classes and sha256) are tracked.
|
||||
|
||||
## File-type gate (E4)
|
||||
|
||||
`filetype_gate.py` asks a reader where each frozen quote occurs, one run per source file. It measures the L1 number
|
||||
"file types read with a locator, of 7". It calls no model.
|
||||
|
||||
```
|
||||
python3 gates/filetype_gate.py --manifest gates/filetypes.manifest.json --pairs gates-local/filetypes/pairs.json \
|
||||
--sources <folder holding blandet/ and wiki/> --cases gates-local/filetypes/files -- <reader> {file} {queries}
|
||||
```
|
||||
|
||||
- `{file}` becomes a fresh copy of one source file, and `{queries}` becomes a JSON file of `{"id", "quote"}` objects.
|
||||
The reader never sees the expected locator. It prints `{"answers": [{"id", "found": [{"locator", "hidden"}]}]}`:
|
||||
every place where the quote occurs, after runs of whitespace are collapsed to one space.
|
||||
- A pair is located when the answer names exactly one place, that place is the frozen locator, and `hidden` is right.
|
||||
Hidden Office content (a hidden sheet, rows or columns, a hidden slide, hidden text, or a tracked deletion) must be
|
||||
marked `hidden: true`, not flattened into ordinary text. Each quote occurs once in its file, so a second place is wrong.
|
||||
- Locators: pdf `{"page": N}`, docx `{"paragraph": N}`, xlsx `{"sheet", "cell"}`, pptx `{"slide": N, "notes": bool}`,
|
||||
and md, txt and html `{"line": N}`. The manifest defines each one and says why it was chosen.
|
||||
- A file type counts as read when every one of its pairs is located. The gate prints this as `file types with every pair
|
||||
located: X of 7`. The hidden classes are reported separately. The gate is GREEN only when every pair is located and no
|
||||
reader run errs.
|
||||
- Before anything runs, the pairs file and every source it names are checked against the manifest's sha256.
|
||||
|
||||
Working set: 112 pairs (16 per type) plus 16 in the hidden cases, from 64 files. A held-out set of 28 pairs (4 per type)
|
||||
lies outside the repository, with only its sha256 here (same form as E2). How the fasit was made is in the manifest under
|
||||
`fasit_method`. The stubs are `stubs/first_locator.py` (always the first page, paragraph, cell, slide or line) and
|
||||
`stubs/fasit_reader.py` (the known-positive, which answers from the pairs file).
|
||||
|
||||
The engine's reader is `python3 -m okfwiki.readers locate {file} {queries}`. On 07.11.2026 it located 128 of 128 in the
|
||||
working set (16 of 16 per type, every hidden class, 4 of 4 controls, 0 reader errors); `first_locator.py` located 7.
|
||||
|
||||
Tests: `python3 -m unittest discover -s gates/tests`. Types: `mypy --strict gates/`.
|
||||
|
||||
## Answer set for collection 1 (wave 2 order 6, K5)
|
||||
|
||||
`answerset.py` and `answers.py` hold the answer set for collection 1 and the harness for the answer measurements
|
||||
(contract K5); see `ANSWERSET.md`. It replaces E2 for the verdict.
|
||||
|
||||
## Whole-document and counting group (E2b)
|
||||
|
||||
`wholedoc.py` holds a small group of questions for the class the E2 set lacks: an answer that needs one long file read
|
||||
as a whole, or a count over many places in one file. It is reported on its own and calls no model. It is a separate
|
||||
gate, not a group inside `multidoc.py`, because that checker requires four or more files per question and E2 is frozen.
|
||||
|
||||
```
|
||||
python3 gates/wholedoc.py cut --e2-set gates-local/multidoc/set-60.json --sources <source dir> --manifest gates/wholedoc.manifest.json
|
||||
python3 gates/wholedoc.py check --set <set> --sources <source dir> --manifest gates/wholedoc.manifest.json --e2-set gates-local/multidoc/set-60.json
|
||||
python3 gates/wholedoc.py split --set <all.json> ... --seed <n> --held-out-per-type 2 --held-out <file OUTSIDE the repository>
|
||||
python3 gates/wholedoc.py gate --set gates-local/wholedoc/set-12.json --manifest gates/wholedoc.manifest.json \
|
||||
--sources <source dir> --arm <name> --answers <answers.json> [--out <results.json>]
|
||||
python3 gates/stubs/wholedoc_arm.py <one-span|fasit> <set> <answers.json>
|
||||
```
|
||||
|
||||
- **cut** takes the sources from the E2 dev set: sorted by id, repeatedly the question needing the fewest new files, while
|
||||
the union stays at 20 files or fewer. It wrote the manifest with the thresholds before any question existed.
|
||||
- **check** measures every property in the source files. Every span stands verbatim and exactly once in its file. A
|
||||
`whole` question sits on a file of 150 lines or more, with at least three keys, spans in at least three parts
|
||||
(a part begins where two span starts lie 15 % of the file or more apart), the first span in the first third and the
|
||||
last one in the last third, no single span holding every key, no key that is wording, and every key marked
|
||||
`negated` standing negated in a span (see `MULTIDOC.md`, "Answer keys are facts, not wording" and "Negated keys").
|
||||
A `count` question has a count of 5 or more that
|
||||
equals its spans and the matches of its `item_pattern` in the file, and the count stands nowhere in the file as a
|
||||
number or an English number word.
|
||||
- **gate**: `whole` counts when the answer holds every key (a number key matches as a number, a negated key only with a
|
||||
negation before it); `count` counts when the answer's `count` field is that exact
|
||||
integer. Both need a citation that quotes the file verbatim. It prints `whole` and `count` separately, and GREEN
|
||||
only when every question is correct. The `one-span` arm must score 0 (RED), and the `fasit` arm is the known-positive
|
||||
(it writes a negated key as "not <key>").
|
||||
|
||||
Frozen 27.09.2026: 16 questions (8 `whole` on the three documentation files of 150 lines or more in the cut, 8 `count`
|
||||
on five files of the cut) were written from the sources after the thresholds were committed. Seed 20260927 held out 2 per
|
||||
type (4 questions, outside the repository, sha256 only here), which leaves 12 for development (6 `whole`, 6 `count`).
|
||||
`check` is PASS on both parts. On the development part the gate is RED at 0 of 12 with the `one-span` arm and GREEN at
|
||||
12 of 12 with the `fasit` arm. Without an answers file it is fail-closed at 0 of 12.
|
||||
|
||||
## Quote rule for K's fragments (V1 Step 4, brief SC3)
|
||||
|
||||
`quote_rule.py` measures two rules that close the path P1 saw (the compiler wrote numbers as words to slip past a rule
|
||||
that asked for a quote only for a digit or a backticked name). It calls no model and prints counts only.
|
||||
|
||||
```
|
||||
python3 -m gates.quote_rule measure ~/okf/benk/wiki/prove-p1/K-fragments ~/okf/benk/wiki/prove-p1/K-work/index.json
|
||||
```
|
||||
|
||||
- **Rule E:** every fragment carries a quote. **Rule L:** a quote is required for a digit, a backticked name or a word
|
||||
from the narrow list (zero to twenty, hundred, thousand, million).
|
||||
- **Answer key**, made apart from rule L: a quoteless fragment whose text holds a word from the broad list of brief SC3
|
||||
(the narrow list plus thirty to ninety, dozen, dozens, once, twice, first to tenth, half, double, pair, couple,
|
||||
several, single), as a whole word. The denominator is the number of hits.
|
||||
- **Choice:** E, unless L lets 0 key fragments through and E rejects at least 33 of 650 (5 %) more than L.
|
||||
|
||||
Measured on 04.10.2026 on P1's 650 fragments (all 650 accepted in P1; 160 have no quote):
|
||||
|
||||
| Count | Value |
|
||||
|---|---|
|
||||
| E rejects | 160 of 650 |
|
||||
| L rejects | 10 of 650 |
|
||||
| Answer-key hits (denominator) | 26 (10 hold a narrow-list word) |
|
||||
| Key hits per file | pack-1.json 19, pack-4.json 7 |
|
||||
| Key fragments E lets through | 0 of 26 |
|
||||
| Key fragments L lets through | 16 of 26 |
|
||||
| E rejects minus L rejects | 150 (threshold 33) |
|
||||
|
||||
**Chose E because L lets 16 of the 26 key fragments through** (the rule needs 0); E rejects 150 more fragments than
|
||||
L, so the cost condition alone would also have allowed L, but the slip condition rules it out. The known-positive holds
|
||||
(26 hits, at least 1). The P1 README says the compiler wrote numbers as words "in one pack" without naming it; the key
|
||||
finds hits in packs 1 and 4, and the four compiler transcripts end in a tool call with no report text, so which pack the
|
||||
README meant is not measured.
|
||||
|
||||
New reject rules, adopted in Step 6 only if each rejects at most 5 % of the 650 (32):
|
||||
|
||||
| Rule | Rejects | Adopt |
|
||||
|---|---|---|
|
||||
| short quote (fewer than 3 words) | 34 of 650 (5.2 %) | no: over 5 %, reported to the PM |
|
||||
| id mismatch (an `existing` id whose entry matches the concept name neither by normalised form nor by Jaccard >= 0.6) | 0 of 650 | yes |
|
||||
|
||||
The 0 id mismatches were checked against a known-positive: with every fragment given its neighbour's id, the check
|
||||
flags 454 of 650, which is every fragment whose neighbour has a different id (196 neighbours share one).
|
||||
|
||||
`NEEDS_QUOTE`, `MERGE_JACCARD` and `norm` are copies of the trial's, kept equal by `TrialConstantsTest`, which reads the
|
||||
trial at run time. Chose copies over an import, because `mypy --strict okfwiki gates tests` does not see the trial
|
||||
package (it is found only when `evals` is listed), so an import would turn the v1 mypy line red once `gates/` is frozen.
|
||||
134
gates/REPAIR.md
Normal file
134
gates/REPAIR.md
Normal file
|
|
@ -0,0 +1,134 @@
|
|||
# Repair gate
|
||||
|
||||
`repair_gate.py` is the judge of the repair command (`python3 -m okfwiki.krepair`, order 34, contracts K18 and K19).
|
||||
It was written and run RED before the command existed (order 34, session A). It calls no model.
|
||||
|
||||
```
|
||||
python3 gates/repair_gate.py --manifest gates/repair.manifest.json -- python3 -m okfwiki.krepair
|
||||
```
|
||||
|
||||
The command after `--` is the program each case's arguments are appended to; it runs from the repository root.
|
||||
Exit 0 GREEN (every case passed), 1 RED, 2 the gate could not run.
|
||||
|
||||
## What it builds
|
||||
|
||||
The gate builds its own collection from the fixtures in `gates/repair/`, with the frozen tools only: `kblocks order`
|
||||
and `blocks`, `kpack skeleton`, `pack` and `accept` (two packs of three sources), `kassemble build --at
|
||||
2026-01-01T00:00:00Z`, `klink accept` and `kassemble build --links`. No generated page is in git.
|
||||
|
||||
- `kilder/`: a K1 source root made by `okfwiki.sources freeze` from text written for the gate about beekeeping: three
|
||||
docs pages (`docs-en-hive.md`, `docs-en-honey.md`, `docs-en-swarm.md`) and one release note
|
||||
(`release-v2.1.0.md`). Nothing in it comes from any documentation (CLAUDE.md rule 3).
|
||||
- `fragments/pack-1.json`, `pack-2.json`: the compiler's output for the two packs, written by hand: 19 fragments,
|
||||
three main pages and five homes.
|
||||
- `links.json`: the five links of the base, by page, sentence text, anchor and target.
|
||||
- `agent.md`, `agent-shared.md`: the agent files the task cases give `--agent`; the second shares a run of eleven words
|
||||
with a line of a source.
|
||||
- `stubs/`: three commands that must each turn the gate RED (below).
|
||||
|
||||
The base has eight pages (`hive`, `frames`, `brood`, `smoker`, `water`, `honey-harvest`, `swarming`,
|
||||
`queen-cells`), 20 claims and five links, and must come out with no auditor finding and criterion 1 at 20 of 20, or
|
||||
the gate cannot run. It holds what the cases need: a fragment of two sentences, a merged fragment, two fragments with
|
||||
context quotes, history from the release note, two fragments in two concepts that share one footnote, two equal
|
||||
sentences on one page (the second linked), a page with one fragment and a link to it, a main page with one fragment
|
||||
and a page whose home it is, and a main page that stands on history alone once its own source's fragment goes.
|
||||
|
||||
## What it measures
|
||||
|
||||
Every case runs in a fresh copy of the base. For each case, by the gate's own code and apart from the command's
|
||||
report:
|
||||
|
||||
1. **Exit code and refusal codes.** Each step's exit code is the case's (0 done, 1 refused). The codes are read from
|
||||
the lines `refused: <code> <claim or record>` on stdout or stderr; the set must be the case's, and a line that names
|
||||
no claim or record fails. The codes are listed with their meaning in the manifest's `refusal_codes`.
|
||||
2. **The pages.** A case that writes is built with `kassemble build --links --replace --at 2026-01-02T00:00:00Z`; the
|
||||
concept pages whose bytes differ from the base (`index.md` and `log.md` left out; a page gone counts) must be
|
||||
exactly the case's `pages_changed`, and the pages gone exactly its `pages_gone`.
|
||||
3. **The claims.** The frozen `claims.extract` runs on both collections. The (id, `excerpt_sha256`) pairs that are
|
||||
gone and those that are new must be, by id, exactly the claims the case names in `claims_before` and
|
||||
`claims_after`; so every other claim keeps its id and its `excerpt_sha256`. Where a case gives `excerpts`, the
|
||||
claim must show exactly those (file, line) excerpts in that order.
|
||||
4. **The auditor and criterion 1.** The frozen auditor's findings on the result, as (check, file), are exactly the
|
||||
case's `findings` (most cases: none), and criterion 1 is N of N with no claim left without a key.
|
||||
5. **The links.** The links file after the build holds exactly the base records, without the case's `links_removed`
|
||||
and with its `links_added` (page, sentence key, anchor, target), and every record's `at` points at its anchor in its
|
||||
sentence on the built page.
|
||||
6. **The command's diff.** After a `fix`, the gate runs `diff --old <base> --new <result> --sources --repairs
|
||||
WORK/K-work/repairs.json --out DIR`: exit 0, `ids.json` (`{"ids": [...]}`, the form `claims calls --ids` reads)
|
||||
must hold exactly the new and changed claim ids the gate measured, and `sider.json` (`{"pages": [...]}`) exactly
|
||||
the changed pages.
|
||||
7. **Nothing written on a refusal.** A refused step changes no file in the work folder, the links file, the collection
|
||||
or the output folder (sha256 of each tree before and after). `task` and `linkset` change none of the first three.
|
||||
No case may change the sources or the copy of the old collection.
|
||||
|
||||
`task` cases: every `<task>.parts.json` in the output folder is `{"parts": [paths], "answer": path}`; every part
|
||||
exists and is at most `kpack.PART_CHARS` characters; no two tasks share an answer path; each listed fragment's
|
||||
footnote stands in exactly one task, and that task holds the page title, every sentence of the fragment, its quote,
|
||||
its context quote and lines of its block that the case names.
|
||||
|
||||
`linkset` cases: the output folder holds `group-<n>.sentences.json` (the `okf-wiki/link-sentences/1` form `klink task
|
||||
--sentences` reads) and `group-<n>.catalogue` (one id per line) per group; the frozen `klink task` runs on each pair;
|
||||
the sentences of all groups are exactly the candidates; and each candidate's catalogue holds its current target.
|
||||
|
||||
## Where the denominator comes from
|
||||
|
||||
Every page, sentence, link and finding a case expects is written by hand in `repair.manifest.json` from the sources in
|
||||
`gates/repair/kilder/`: pages by name, claims as sentence text with its occurrence `n` on the page (counted from 0, as
|
||||
klink counts), links as page, sentence text, anchor and target. The gate turns a sentence into its claim id with the
|
||||
frozen `klink.key_text` (the first 16 hex characters of its sha256, plus `-<n>`), and a fragment into its footnote with
|
||||
the frozen `kassemble.label`, from its file, span, quote and mark. Nothing is taken from the command.
|
||||
|
||||
The findings of the three removal cases follow from K2 as written: a main-topic source no page lists in `x-main-for`
|
||||
is `main-page-missing`, and a page that is not a main page and has no link from one is `no-main-page-link`. When the
|
||||
main page `swarming` goes, its source has no main page and `queen-cells`, whose home it was, has no link from a main
|
||||
page; when `honey-harvest` loses its only fragment from `docs-en-honey.md`, kassemble writes no `x-main-for` (a main
|
||||
page record counts only for a source the page cites), so the source has no main page and the page is no main page and
|
||||
has no home.
|
||||
|
||||
## The formats the gate fixes
|
||||
|
||||
The command did not exist when the gate was written, so the gate fixes what K18 leaves open, and the command follows
|
||||
it:
|
||||
|
||||
- Answers (`okf-wiki/repair-answers/1`): `{"schema", "answers": [{"concept_id", "footnote": "s…", "action": "replace",
|
||||
"fragments": [{"text", "quote", "anchor": {"file", "start", "end"}, "provenance", "context"?}]} | {"concept_id",
|
||||
"footnote", "action": "remove"}]}`.
|
||||
- A repair list in K17's schema, rendered from the base: each claim with its id, `excerpt_sha256`, page, line, mark,
|
||||
kind (`docs` or `notes`), its excerpts as `sources` (`kilde/<file>`, line), origin `o33` and its reasons; each link
|
||||
candidate with the base record's `at`, origin `o33-kandidat`. `claims_sha256` is the sha256 of the base's claim list
|
||||
(`claims list` output, given as `--claims`); `verdicts_sha256` is the sha256 of no bytes.
|
||||
- `<task>.parts.json`, `group-<n>.sentences.json` and `group-<n>.catalogue`, `ids.json` and `sider.json` as above.
|
||||
- Command arguments: `--page` and `--to` are page files (`hive.md`), `--sentence` a sentence key.
|
||||
|
||||
## The cases
|
||||
|
||||
32 cases: 20 that write (replace one with one and with two, a two-sentence fragment, a merged fragment, a release-note
|
||||
fragment, context kept and context replaced, remove one, remove the last on a page and on a main page, remove the last
|
||||
from a source on its main page, a shared footnote, equal sentences, a repair that makes a sentence equal to a later
|
||||
one, a link whose anchor stays and one whose anchor goes, unlink, relink, two repairs in one file, identity), one task,
|
||||
one linkset, and ten refusals (a quote not verbatim beside a good answer, a quote from another source, an unknown key,
|
||||
a collection built before the last repair, two answers for one fragment, a stale list, an agent file that shares eight
|
||||
words, a relink to a page that does not exist, an unlink with the wrong `at` and with the wrong anchor). Each case says
|
||||
in `what` what it tries.
|
||||
|
||||
## The stubs and the gate's own test
|
||||
|
||||
`gates/tests/test_repair_gate.py` shows that each stub turns the gate RED with no case passed: `noop.py` (does
|
||||
nothing, exit 0), `refuse_all.py` (`refused: x`, exit 1, writes nothing) and `touch_generated.py` (gives every page a
|
||||
new `generated.at`; the rebuild keeps it, so the identity case sees every page changed). It also shows that a changed,
|
||||
missing or extra fixture stops the gate (exit 2), that a missing command and a missing module are RED, and it tries
|
||||
the checks a command cannot be run against yet with handmade outputs (task, linkset, a refusal, a link command).
|
||||
|
||||
```
|
||||
python3 -m unittest discover -s gates/tests -p 'test_repair_gate.py'
|
||||
```
|
||||
|
||||
## Not measured
|
||||
|
||||
- The command's `diff` after `unlink` and `relink`: K18 gives them no work folder, so they write no repairs log, and
|
||||
the gate checks their pages, claims and links but not a diff.
|
||||
- `linkset`'s refusal of a stale list: K18 gives it no `--claims`; its cases give a list with the hash of the base's
|
||||
claim list.
|
||||
- A fragment whose block alone is over `kpack.PART_CHARS` (task exit 2), and the instructions' own eight-word check:
|
||||
the command's unit tests try them.
|
||||
- The repairs log's content: only through the command's `diff`.
|
||||
911
gates/answers.py
Normal file
911
gates/answers.py
Normal file
|
|
@ -0,0 +1,911 @@
|
|||
"""The answer harness for PLAN.md criterion 4 (contract K5 points 2-7), adapted from evals/direction_probe.py.
|
||||
|
||||
Each question is answered by one isolated subagent per arm in the user's session, described
|
||||
`S1-<run>-<arm>-<id>`. This module never calls a model and starts nothing:
|
||||
|
||||
setup one arm's basis (its own copy of the root's source files, and for `bundle` the collection beside
|
||||
them) and one task file per question, `<run>/tasks/<arm>/<id>.md`, which carries the question and
|
||||
the basis and never the answer keys, the spans or the number of documents
|
||||
collect the per-question answer files into the arm's answers file (K5 point 4)
|
||||
harvest tokens per question (each message id once, all four usage fields, plus a lower bound for an output
|
||||
the transcript never recorded) and isolation breaches from the subagent transcripts (K5 point 6)
|
||||
score two arms' results from `answerset.py gate --out`: per arm and group, by path count, the pairs with
|
||||
the exact McNemar p, the difference with its paired 95 % interval (Newcombe 1998, method 10), and the
|
||||
smallest difference the set can show at 80 % power and alpha 0.05 (K5 point 7)
|
||||
|
||||
The arms are `filsok` (file search alone over the root's source files) and `bundle` (the collection plus file
|
||||
search over the same files). Exit 0 = done; 1 = could not run, or (collect, harvest) something is missing.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import shlex
|
||||
import shutil
|
||||
import sys
|
||||
from math import comb, sqrt
|
||||
from pathlib import Path
|
||||
from statistics import NormalDist
|
||||
from typing import Any
|
||||
|
||||
from answerset import (
|
||||
ANSWERS_SCHEMA,
|
||||
ARMS,
|
||||
GROUPS,
|
||||
K1_FILE,
|
||||
MANY_PATHS,
|
||||
RESULTS_SCHEMA,
|
||||
Root,
|
||||
RunError,
|
||||
as_list,
|
||||
dump,
|
||||
load_json,
|
||||
load_set,
|
||||
)
|
||||
|
||||
TOKENS_SCHEMA = "okf-wiki/answer-tokens/1"
|
||||
SCORE_SCHEMA = "okf-wiki/answer-score/1"
|
||||
NAME_RE = re.compile(r"^[A-Za-z0-9]+$")
|
||||
ALLOWED_TOOLS = ("Read", "Grep", "Glob", "Bash", "Write", "SubagentHandback", "StructuredOutput", "ToolSearch")
|
||||
READ_ONLY = ("grep", "egrep", "fgrep", "rg", "find", "ls", "wc", "head", "tail", "sed", "cat", "sort", "uniq", "cut",
|
||||
"tr", "nl")
|
||||
LITERAL = ("echo",) # accepted in a chain (order 31 A2: order 13's separators), never offered in the task
|
||||
OWN_ANSWER = ("jq",) # accepted on the helper's own answer file only (order 31 A2b), never offered in the task
|
||||
SEPARATORS = ("|", "&&", ";", "||")
|
||||
FIND_ACTIONS = ("-exec", "-execdir", "-ok", "-okdir", "-delete", "-fprint", "-fprint0", "-fprintf", "-fls")
|
||||
# Order 31 A2b: the search word of grep and rg, the script of sed and the filter of jq are not paths. Per command: the
|
||||
# short options that take a value, those of them whose value is the search word, those whose value is a file of
|
||||
# search words, the long ones of each, and (order 39) every other long option that takes a value, GNU and BSD. Such a
|
||||
# value is consumed and checked as a path, so a file option's value outside the basis (--exclude-from, --ignore-file)
|
||||
# is a breach. An option missing here makes its value the search word, and when that value is a path the path goes
|
||||
# unchecked: the lists must be whole. grep's --context is left out because BSD grep takes its value only after `=`
|
||||
# (`--context 1 /x f` searches for 1 in /x), so a separate value stays the search word and the check stays stricter.
|
||||
GREP_LONG = ("after-context", "before-context", "binary-files", "devices", "directories", "exclude", "exclude-dir",
|
||||
"exclude-from", "group-separator", "include", "include-dir", "label", "max-count")
|
||||
RG_LONG = ("after-context", "before-context", "color", "colors", "context", "context-separator", "dfa-size-limit",
|
||||
"encoding", "engine", "field-context-separator", "field-match-separator", "generate", "glob", "iglob",
|
||||
"ignore-file", "hostname-bin", "hyperlink-format", "max-columns", "max-count", "max-depth", "max-filesize",
|
||||
"path-separator", "pre", "pre-glob", "regex-size-limit", "replace", "sort", "sortr", "threads", "type",
|
||||
"type-add", "type-clear", "type-not")
|
||||
SEARCH_WORDS: dict[str, tuple[str, str, str, tuple[str, ...], tuple[str, ...], tuple[str, ...]]] = {
|
||||
**{grep: ("ABCDdefm", "e", "f", ("regexp",), ("file",), GREP_LONG) for grep in ("grep", "egrep", "fgrep")},
|
||||
"rg": ("ABCEMTdefgjmrt", "e", "f", ("regexp",), ("file",), RG_LONG),
|
||||
"sed": ("efl", "e", "f", ("expression",), ("file",), ("line-length",)),
|
||||
"jq": ("", "", "f", (), ("from-file",), ()), # the filter; -f reads it from the first file instead
|
||||
}
|
||||
RUNS_A_PROGRAM = {"rg": ("pre", "hostname-bin")} # long options whose value is a program the command runs
|
||||
# Order 39: a sed script is exempt from the path check, so its commands are read instead. w, W and e (also as flags
|
||||
# of s) write a file or run a program: a breach. r and R read the file named after them: checked as a path.
|
||||
SED_IN_PLACE = "iI" # GNU -i[SUFFIX], BSD -i SUFFIX and -I SUFFIX: either one in a flag cluster edits in place
|
||||
SED_ADDRESS = re.compile(r"\d+(~\d+)?|\$")
|
||||
SED_PLAIN = set("=dDgGhHlnNpPxzFqQ") # commands without an argument (l, q and Q may take a number)
|
||||
SED_LABEL = set("btT:") # commands whose label runs to a semicolon or a newline
|
||||
SED_TEXT = set("aic") # commands whose text runs to the end of the line
|
||||
SED_S_FLAGS = set("gpiImM0123456789 \t")
|
||||
USAGE_FIELDS = (
|
||||
"input_tokens", # uncached input
|
||||
"cache_creation_input_tokens", # written to the prompt cache
|
||||
"cache_read_input_tokens", # read from it
|
||||
"output_tokens",
|
||||
)
|
||||
INPUT_SIDE = USAGE_FIELDS[:3]
|
||||
CHARS_PER_TOKEN_LOW = 2.5 # injected text counted high, so a derived output is a lower bound (okfwiki/cost.py)
|
||||
ALPHA, POWER = 0.05, 0.80
|
||||
|
||||
|
||||
def write(path: Path, data: Any) -> None:
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_bytes(dump(data))
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- setup
|
||||
|
||||
|
||||
SEARCH = {
|
||||
"filsok": """BASIS holds the source files under `kilde/` and nothing else. Search them with grep and find (Bash,
|
||||
as above), and read what you find with Read.""",
|
||||
"bundle": """BASIS holds `wiki/`, a knowledge collection compiled from the sources, and `kilde/`, the source
|
||||
files. A page names its source as `/kilde/<name>`, which is the file `BASIS/kilde/<name>`. Follow this order:
|
||||
1. Search `BASIS/wiki/index.md` with grep for the question's words: it is never read whole, because it can be over
|
||||
the size one Read allows. Each line is one page: its title and its description.
|
||||
2. Open the main page of the area the question is about: a main page carries `x-main-for` with the source it covers.
|
||||
3. Read it, follow its links to the pages the question needs, and go on from those pages where the question needs it.
|
||||
4. Open the source behind a footnote when the quote has to be checked or when the page does not say enough.
|
||||
You may also search `wiki/` and `kilde/` with grep and find (Bash, as above) at any point.""",
|
||||
}
|
||||
|
||||
|
||||
def render_task(arm: str, basis: Path, qid: str, question: str, answer_path: Path) -> str:
|
||||
return f"""# Answer one question
|
||||
|
||||
BASIS: `{basis}`
|
||||
|
||||
You answer ONE question from the files in BASIS. Read nothing outside BASIS apart from this task file: no other
|
||||
folder, no web, no earlier work. Answer only from what the files say, and say plainly when they do not say something.
|
||||
|
||||
Tools: Read (a file inside BASIS), Bash for read-only search, and Write once, for the answer file. No other tool.
|
||||
Every Bash command starts with `cd {basis} && ` and then uses only {', '.join(READ_ONLY)}, joined by `|` or `&&`,
|
||||
with paths relative to BASIS: never `..`, `~` or a path outside BASIS, no `$` expansion, no output redirection except
|
||||
`2>/dev/null`, no `find -exec` or `-delete`, no `sed -i`. For example: `cd {basis} && grep -rn "term" kilde`
|
||||
|
||||
## How to search
|
||||
|
||||
{SEARCH[arm]}
|
||||
|
||||
## The question
|
||||
|
||||
{question}
|
||||
|
||||
## The answer
|
||||
|
||||
Write one JSON file to `{answer_path}`:
|
||||
|
||||
```json
|
||||
{{"id": "{qid}", "answer": "<your answer in prose>", "citations": [{{"path": "kilde/<name>", "quote": "<text>"}}]}}
|
||||
```
|
||||
|
||||
- `answer`: answer every part of the question.
|
||||
- `citations`: one or more per source file your answer relies on, and cite every such file.
|
||||
- `path`: `kilde/` and the file name, as the file lies in BASIS, for example `kilde/notes.md`: never a page in
|
||||
`wiki/`.
|
||||
- `quote`: text copied character for character from that file (only whitespace may differ), long enough to carry
|
||||
the fact, for example the whole line or sentence.
|
||||
|
||||
When the file is written, reply with the one word `done`.
|
||||
"""
|
||||
|
||||
|
||||
def copy_sources(root: Root, target: Path) -> int:
|
||||
"""Every source K1 lists, copied byte-exact under target/kilde/, each checked against its K1 sha256."""
|
||||
for path in sorted(root.entries):
|
||||
text = root.text(path)
|
||||
if text is None:
|
||||
raise RunError(root.problem(path))
|
||||
dest = target / path
|
||||
dest.parent.mkdir(parents=True, exist_ok=True)
|
||||
shutil.copyfile(root.path / path, dest)
|
||||
return len(root.entries)
|
||||
|
||||
|
||||
def check_part(ref: dict[str, Any], root: Root, part_path: Path) -> None:
|
||||
"""Order 31 A3: the set names the part's K1, the part is a subset_of the root, and every source of the part
|
||||
stands in the root with the same path and sha256."""
|
||||
part = Root(part_path)
|
||||
if ref.get("manifest_sha256") != part.manifest_sha256:
|
||||
raise RunError(f"the set names manifest_sha256 {ref.get('manifest_sha256')!r}, --part-root {part_path} has "
|
||||
f"{part.manifest_sha256}; --part-root is the root the set was made from")
|
||||
raw = json.loads(part.manifest_bytes.decode("utf-8")).get("subset_of")
|
||||
named = raw.get("manifest_sha256") if isinstance(raw, dict) else None
|
||||
if named != root.manifest_sha256:
|
||||
raise RunError(f"--part-root {part_path} names subset_of manifest_sha256 {named!r}, the root has "
|
||||
f"{root.manifest_sha256}")
|
||||
for path, entry in sorted(part.entries.items()):
|
||||
if path not in root.entries:
|
||||
raise RunError(f"{path} of --part-root is not in the root {root.path}")
|
||||
if root.entries[path]["sha256"] != entry["sha256"]:
|
||||
raise RunError(f"{path} has sha256 {entry['sha256']} in --part-root and {root.entries[path]['sha256']} "
|
||||
f"in the root")
|
||||
|
||||
|
||||
def setup_arm(run: Path, name: str, arm: str, set_path: Path, sources: Path,
|
||||
bundle: Path | None, part_root: Path | None = None) -> list[tuple[str, Path]]:
|
||||
if arm not in ARMS:
|
||||
raise RunError(f"unknown arm {arm!r}; the arms are {', '.join(ARMS)}")
|
||||
if not NAME_RE.match(name):
|
||||
raise RunError(f"run name {name!r} must be letters and digits only (it is part of S1-<run>-<arm>-<id>)")
|
||||
if arm == "bundle" and bundle is None:
|
||||
raise RunError("the bundle arm needs --bundle")
|
||||
if arm == "bundle" and bundle is not None and not (bundle / "index.md").is_file():
|
||||
raise RunError(f"{bundle} has no index.md")
|
||||
data = load_set(set_path)
|
||||
root = Root(sources)
|
||||
ref = data["sources"]
|
||||
if part_root is not None:
|
||||
check_part(ref, root, part_root)
|
||||
elif ref.get("manifest_sha256") != root.manifest_sha256:
|
||||
raise RunError(f"the set names manifest_sha256 {ref.get('manifest_sha256')!r}, the root has "
|
||||
f"{root.manifest_sha256}; an arm is set up on the root the set was made from")
|
||||
basis = run / "arms" / arm
|
||||
if basis.exists():
|
||||
raise RunError(f"{basis} exists already; an arm is set up once, from fresh")
|
||||
if arm == "bundle" and bundle is not None:
|
||||
shutil.copytree(bundle, basis / "wiki")
|
||||
copy_sources(root, basis)
|
||||
tasks = []
|
||||
for q in data["questions"]:
|
||||
qid = str(q["id"])
|
||||
task = run / "tasks" / arm / f"{qid}.md"
|
||||
answer = run / "answers" / arm / f"{qid}.json"
|
||||
task.parent.mkdir(parents=True, exist_ok=True)
|
||||
task.write_text(render_task(arm, basis, qid, str(q["question"]), answer), encoding="utf-8")
|
||||
tasks.append((f"S1-{name}-{arm}-{qid}", task))
|
||||
(run / "answers" / arm).mkdir(parents=True, exist_ok=True)
|
||||
return tasks
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- collect
|
||||
|
||||
|
||||
def collect(run: Path, arm: str) -> tuple[Path, list[str]]:
|
||||
"""The per-question answers into one answers file; the id comes from the file name."""
|
||||
if arm not in ARMS:
|
||||
raise RunError(f"unknown arm {arm!r}; the arms are {', '.join(ARMS)}")
|
||||
folder = run / "answers" / arm
|
||||
if not folder.is_dir():
|
||||
raise RunError(f"{folder} does not exist; run setup first")
|
||||
gathered: list[dict[str, Any]] = []
|
||||
problems = []
|
||||
for path in sorted(folder.glob("*.json")):
|
||||
try:
|
||||
data = json.loads(path.read_text(encoding="utf-8"))
|
||||
except (OSError, UnicodeDecodeError, json.JSONDecodeError):
|
||||
data = None
|
||||
if not isinstance(data, dict):
|
||||
problems.append(f"{path.name}: not a JSON object")
|
||||
continue
|
||||
gathered.append({**data, "id": path.stem})
|
||||
out = run / "answers" / f"{arm}.json"
|
||||
write(out, {"schema": ANSWERS_SCHEMA, "arm": arm, "answers": gathered})
|
||||
return out, problems
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- harvest
|
||||
|
||||
|
||||
def read_transcript(path: Path) -> list[dict[str, Any]]:
|
||||
lines = []
|
||||
for raw in path.read_text(encoding="utf-8").splitlines():
|
||||
try:
|
||||
line = json.loads(raw)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
if isinstance(line, dict):
|
||||
lines.append(line)
|
||||
return lines
|
||||
|
||||
|
||||
def first_prompt(lines: list[dict[str, Any]]) -> str:
|
||||
for line in lines:
|
||||
if line.get("type") == "user":
|
||||
content = line.get("message", {}).get("content")
|
||||
if isinstance(content, list):
|
||||
content = " ".join(str(b.get("text", "")) for b in content if isinstance(b, dict))
|
||||
return content if isinstance(content, str) else ""
|
||||
return ""
|
||||
|
||||
|
||||
def inside(path: object, root: Path) -> bool:
|
||||
if not isinstance(path, str) or not os.path.isabs(path):
|
||||
return False
|
||||
norm = Path(os.path.normpath(path))
|
||||
return norm == root or root in norm.parents
|
||||
|
||||
|
||||
def outside_single_quotes(command: str) -> str:
|
||||
"""The command with the text inside single quotes blanked out, where the shell takes every character literally
|
||||
(order 31 A2b). A backslash outside them keeps the next character, so an escaped backtick still shows."""
|
||||
out: list[str] = []
|
||||
quote = ""
|
||||
i = 0
|
||||
while i < len(command):
|
||||
ch = command[i]
|
||||
if quote == "'":
|
||||
quote = "" if ch == "'" else quote
|
||||
out.append(" ")
|
||||
elif ch == "\\":
|
||||
out.append(command[i:i + 2])
|
||||
i += 1
|
||||
elif ch == "'" and not quote:
|
||||
quote = "'"
|
||||
out.append(" ")
|
||||
else:
|
||||
if ch == '"':
|
||||
quote = "" if quote else '"'
|
||||
out.append(ch)
|
||||
i += 1
|
||||
return "".join(out)
|
||||
|
||||
|
||||
def search_words(command: str, args: list[str]) -> set[int]:
|
||||
"""The indexes in args of the search words: the values of -e and --regexp (sed: --expression), or else the first
|
||||
argument that is not an option. With -f or --file the search words come from a file and none is positional,
|
||||
wherever the option stands (GNU getopt reorders the arguments). The value of any other option that takes one is
|
||||
skipped here, so it is checked as a path."""
|
||||
if command not in SEARCH_WORDS:
|
||||
return set()
|
||||
valued, word_opts, file_opts, word_long, file_long, other_long = SEARCH_WORDS[command]
|
||||
words: set[int] = set()
|
||||
first: int | None = None
|
||||
positional, options = True, True
|
||||
i = 0
|
||||
while i < len(args):
|
||||
arg = args[i]
|
||||
if options and arg == "--":
|
||||
options = False
|
||||
elif options and arg.startswith("--"):
|
||||
name, eq, _ = arg[2:].partition("=")
|
||||
if name in word_long + file_long:
|
||||
positional = False
|
||||
if name in word_long:
|
||||
words.add(i if eq else i + 1)
|
||||
if name in word_long + file_long + other_long:
|
||||
i += 0 if eq else 1
|
||||
elif options and arg.startswith("-") and arg != "-":
|
||||
for j, ch in enumerate(arg[1:]):
|
||||
if ch in word_opts + file_opts:
|
||||
positional = False
|
||||
if ch in valued:
|
||||
attached = j + 2 < len(arg)
|
||||
if ch in word_opts:
|
||||
words.add(i if attached else i + 1)
|
||||
i += 0 if attached else 1
|
||||
break
|
||||
elif first is None:
|
||||
first = i
|
||||
i += 1
|
||||
if positional and first is not None:
|
||||
words.add(first)
|
||||
return words
|
||||
|
||||
|
||||
def word_text(command: str, token: str) -> str:
|
||||
"""The search word or script a search-word token carries: the token, or the value attached to -e or --expression."""
|
||||
if token.startswith("--"):
|
||||
return token.partition("=")[2]
|
||||
if token.startswith("-") and token != "-":
|
||||
valued = SEARCH_WORDS[command][0]
|
||||
for j, ch in enumerate(token[1:]):
|
||||
if ch in valued:
|
||||
return token[j + 2:]
|
||||
return token
|
||||
|
||||
|
||||
def in_place(token: str) -> bool:
|
||||
"""A sed flag cluster with -i or -I before the first option that takes a value (`-ni`, `-Ei`, `-I`)."""
|
||||
if not token.startswith("-") or token.startswith("--"):
|
||||
return False
|
||||
for ch in token[1:]:
|
||||
if ch in SED_IN_PLACE:
|
||||
return True
|
||||
if ch in SEARCH_WORDS["sed"][0]:
|
||||
return False
|
||||
return False
|
||||
|
||||
|
||||
class SedParseError(ValueError):
|
||||
pass
|
||||
|
||||
|
||||
def sed_breach(script: str) -> tuple[str | None, list[str]]:
|
||||
"""The breach of one sed script, if any (w, W or e as a command or as a flag of s, or a script that cannot be
|
||||
parsed), and the files it reads with r or R, which the caller checks as paths. POSIX, GNU and BSD commands."""
|
||||
n = len(script)
|
||||
reads: list[str] = []
|
||||
|
||||
def regex_end(k: int, delim: str) -> int:
|
||||
while k < n and script[k] != delim:
|
||||
k += 2 if script[k] == "\\" else 1
|
||||
if k >= n:
|
||||
raise SedParseError(script)
|
||||
return k + 1
|
||||
|
||||
def address(k: int) -> int:
|
||||
m = SED_ADDRESS.match(script, k)
|
||||
if m:
|
||||
return m.end()
|
||||
if k < n and script[k] in "/\\":
|
||||
delim = "/"
|
||||
if script[k] == "\\":
|
||||
if k + 1 >= n:
|
||||
raise SedParseError(script)
|
||||
k += 1
|
||||
delim = script[k]
|
||||
k = regex_end(k + 1, delim)
|
||||
while k < n and script[k] in "IM":
|
||||
k += 1
|
||||
return k
|
||||
|
||||
def line_end(k: int) -> int:
|
||||
end = script.find("\n", k)
|
||||
return n if end < 0 else end
|
||||
|
||||
depth, i = 0, 0
|
||||
try:
|
||||
while i < n:
|
||||
if script[i] in " \t\n;":
|
||||
i += 1
|
||||
continue
|
||||
if script[i] == "}":
|
||||
depth -= 1
|
||||
if depth < 0:
|
||||
raise SedParseError(script)
|
||||
i += 1
|
||||
continue
|
||||
start = i
|
||||
i = address(i)
|
||||
if start < i < n and script[i] == ",":
|
||||
i += 1
|
||||
if i < n and script[i] in "+~":
|
||||
i += 1
|
||||
while i < n and script[i].isdigit():
|
||||
i += 1
|
||||
else:
|
||||
i = address(i)
|
||||
while i < n and script[i] in " \t!":
|
||||
i += 1
|
||||
if i >= n:
|
||||
raise SedParseError(script)
|
||||
cmd = script[i]
|
||||
i += 1
|
||||
if cmd == "{":
|
||||
depth += 1
|
||||
elif cmd in "wWe":
|
||||
return f"Bash sed command {cmd}", reads
|
||||
elif cmd in "rR":
|
||||
end = line_end(i)
|
||||
reads.append(script[i:end].strip())
|
||||
i = end
|
||||
elif cmd in SED_TEXT:
|
||||
i = line_end(i)
|
||||
elif cmd in SED_LABEL:
|
||||
while i < n and script[i] not in ";\n":
|
||||
i += 1
|
||||
elif cmd in SED_PLAIN:
|
||||
while i < n and script[i] in " \t0123456789":
|
||||
i += 1
|
||||
elif cmd in "sy":
|
||||
if i >= n or script[i] in "\\\n":
|
||||
raise SedParseError(script)
|
||||
delim = script[i]
|
||||
i = regex_end(regex_end(i + 1, delim), delim)
|
||||
while cmd == "s" and i < n and script[i] not in ";\n}":
|
||||
if script[i] in "wWe":
|
||||
return f"Bash sed s flag {script[i]}", reads
|
||||
if script[i] not in SED_S_FLAGS:
|
||||
raise SedParseError(script)
|
||||
i += 1
|
||||
else:
|
||||
raise SedParseError(script)
|
||||
except SedParseError:
|
||||
return "Bash sed script that cannot be parsed", reads
|
||||
if depth:
|
||||
return "Bash sed script that cannot be parsed", reads
|
||||
return None, reads
|
||||
|
||||
|
||||
def bash_breach(command: str, basis: Path, answer: Path | None = None) -> str | None:
|
||||
"""None when the command is read-only search that starts with `cd <basis> &&` and stays inside the basis, where
|
||||
the only path outside it may be the answer file; or when it reads the answer file alone, by its absolute path or
|
||||
after `cd` into its folder (order 31 A2b)."""
|
||||
unquoted = outside_single_quotes(command)
|
||||
if "`" in unquoted or re.search(r"\$[A-Za-z_{(]", unquoted):
|
||||
return "Bash with an expansion"
|
||||
try:
|
||||
# Order 39: bash takes '#' as a comment only at the start of a word, so shlex's comment rule is off (a comment
|
||||
# is then checked as arguments, which is stricter); an unquoted newline separates commands as ';' does.
|
||||
lexer = shlex.shlex(re.sub(r"\s2>/dev/null(?=\s|$|[;|&])", "", command), posix=True,
|
||||
punctuation_chars="|&;<>()\n")
|
||||
lexer.whitespace_split = True
|
||||
lexer.commenters = ""
|
||||
lexer.whitespace = " \t\r"
|
||||
tokens = [piece for token in lexer for piece in
|
||||
(re.split(r"(\n)", token) if set(token) <= set("|&;<>()\n") else [token]) if piece]
|
||||
except ValueError:
|
||||
return "Bash that cannot be parsed"
|
||||
segments: list[list[str]] = [[]]
|
||||
newline = False # a newline ends a command; after a separator, on a blank line or at the end it changes nothing
|
||||
for token in tokens:
|
||||
if token == "\n":
|
||||
newline = bool(segments[-1])
|
||||
continue
|
||||
if newline and token not in SEPARATORS:
|
||||
segments.append([])
|
||||
newline = False
|
||||
if token in SEPARATORS:
|
||||
segments.append([])
|
||||
elif token and set(token) <= set("&<>()"):
|
||||
return f"Bash with {token!r}"
|
||||
else:
|
||||
segments[-1].append(token)
|
||||
first = segments[0]
|
||||
own = os.path.normpath(answer) if answer is not None else ""
|
||||
cd = first[1] if len(first) == 2 and first[0] == "cd" and os.path.isabs(first[1]) else ""
|
||||
here: str | None = None # where a relative path is, when it can name the answer file
|
||||
if cd and inside(cd, basis):
|
||||
own_only, chain = False, segments[1:]
|
||||
elif own and cd and os.path.normpath(cd) == os.path.dirname(own):
|
||||
own_only, chain, here = True, segments[1:], os.path.dirname(own)
|
||||
elif own and first and first[0] in READ_ONLY + OWN_ANSWER:
|
||||
own_only, chain = True, segments
|
||||
else:
|
||||
return "Bash that does not start with cd into the basis"
|
||||
for segment in chain:
|
||||
if not segment or segment[0] not in READ_ONLY + LITERAL + OWN_ANSWER:
|
||||
return f"Bash command not allowed: {segment[0] if segment else '(empty)'}"
|
||||
words = search_words(segment[0], segment[1:])
|
||||
names_own = False
|
||||
checked: list[tuple[str, str]] = [] # (token, the value in it that is checked as a path)
|
||||
for index, token in enumerate(segment[1:]):
|
||||
if segment[0] == "sed" and (in_place(token) or token.startswith("--in-place")):
|
||||
return "Bash sed -i"
|
||||
if index in words:
|
||||
if segment[0] == "sed":
|
||||
breach, reads = sed_breach(word_text("sed", token))
|
||||
if breach:
|
||||
return breach
|
||||
checked.extend((name, name) for name in reads)
|
||||
continue
|
||||
if segment[0] == "find" and token in FIND_ACTIONS:
|
||||
return f"Bash find action {token}"
|
||||
if token.startswith("--") and token[2:].partition("=")[0] in RUNS_A_PROGRAM.get(segment[0], ()):
|
||||
return f"Bash {segment[0]} {token.partition('=')[0]} runs a program"
|
||||
checked.append((token, token.split("=", 1)[1] if token.startswith("-") and "=" in token else token))
|
||||
for token, value in checked:
|
||||
if value.startswith("~") or ".." in value.split("/"):
|
||||
return f"Bash path outside the basis: {token}"
|
||||
path = value if os.path.isabs(value) else os.path.join(here, value) if here else ""
|
||||
if own and path and os.path.normpath(path) == own:
|
||||
names_own = True
|
||||
elif own_only and (not token.startswith("-") or "=" in token):
|
||||
return f"Bash path other than the own answer: {token}"
|
||||
elif value.startswith("/") and value != "/dev/null" and not inside(value, basis):
|
||||
return f"Bash path outside the basis: {token}"
|
||||
if own_only and not names_own:
|
||||
return "Bash outside the basis that does not name the own answer"
|
||||
if segment[0] in OWN_ANSWER and not names_own:
|
||||
return f"Bash command not allowed: {segment[0]}"
|
||||
return None
|
||||
|
||||
|
||||
def breaches_of(tools: list[tuple[str, dict[str, Any]]], basis: Path, task: Path, answer: Path) -> list[str]:
|
||||
"""K5 point 3: every tool call that reads or lists a path outside the basis and the task file, every tool that
|
||||
is not allowed, and every write to another file than the answer."""
|
||||
found = []
|
||||
for name, inp in tools:
|
||||
if name not in ALLOWED_TOOLS:
|
||||
found.append(f"{name}: tool not allowed")
|
||||
elif name == "Read":
|
||||
path = inp.get("file_path")
|
||||
if not (inside(path, basis)
|
||||
or (isinstance(path, str) and os.path.normpath(path) in (str(task), str(answer)))):
|
||||
found.append(f"Read outside the basis: {path}")
|
||||
elif name in ("Grep", "Glob"):
|
||||
path = inp.get("path")
|
||||
pattern = inp.get("pattern", "")
|
||||
if path is None:
|
||||
found.append(f"{name} without a path (searches the working directory)")
|
||||
elif not inside(path, basis):
|
||||
found.append(f"{name} outside the basis: {path}")
|
||||
elif name == "Glob" and isinstance(pattern, str) and os.path.isabs(pattern) and not inside(pattern, basis):
|
||||
found.append(f"Glob pattern outside the basis: {pattern}")
|
||||
elif name == "Bash":
|
||||
breach = bash_breach(str(inp.get("command", "")), basis, answer)
|
||||
if breach:
|
||||
found.append(breach)
|
||||
elif name == "Write" and inp.get("file_path") != str(answer):
|
||||
found.append(f"Write to another file: {inp.get('file_path')}")
|
||||
return found
|
||||
|
||||
|
||||
def _injected_chars(line: dict[str, Any]) -> int:
|
||||
"""Text put into the context between two assistant messages: tool results and rendered attachments."""
|
||||
if line.get("type") == "attachment":
|
||||
return len(str(line.get("rendered") or ""))
|
||||
content = line.get("message", {}).get("content") if isinstance(line.get("message"), dict) else None
|
||||
total = 0
|
||||
for item in content if isinstance(content, list) else []:
|
||||
if isinstance(item, dict) and item.get("type") == "tool_result":
|
||||
inner = item.get("content")
|
||||
if isinstance(inner, str):
|
||||
total += len(inner)
|
||||
elif isinstance(inner, list):
|
||||
total += sum(len(str(x.get("text", ""))) for x in inner if isinstance(x, dict))
|
||||
return total
|
||||
|
||||
|
||||
def usage_and_tools(lines: list[dict[str, Any]]) -> tuple[dict[str, int], list[tuple[str, dict[str, Any]]]]:
|
||||
"""The four usage fields summed with each message id once at its largest value, the derived output (a message
|
||||
whose records all lack a stop reason never got its final usage written: its output is the growth of the next
|
||||
message's input side less the text injected between them, counted high), and every tool call."""
|
||||
per: dict[str, dict[str, int]] = {}
|
||||
order: list[str] = []
|
||||
complete: set[str] = set()
|
||||
between: dict[str, int] = {}
|
||||
tools: list[tuple[str, dict[str, Any]]] = []
|
||||
for line in lines:
|
||||
message = line.get("message")
|
||||
if line.get("type") != "assistant" or not isinstance(message, dict):
|
||||
if order:
|
||||
between[order[-1]] = between.get(order[-1], 0) + _injected_chars(line)
|
||||
continue
|
||||
mid = str(message.get("id", ""))
|
||||
if mid not in per:
|
||||
order.append(mid)
|
||||
seen = per.setdefault(mid, {})
|
||||
fields = message.get("usage") or {}
|
||||
for f in USAGE_FIELDS:
|
||||
value = fields.get(f)
|
||||
if isinstance(value, int):
|
||||
seen[f] = max(seen.get(f, 0), value)
|
||||
if isinstance(message.get("stop_reason"), str):
|
||||
complete.add(mid)
|
||||
for block in message.get("content") or []:
|
||||
if isinstance(block, dict) and block.get("type") == "tool_use":
|
||||
inp = block.get("input")
|
||||
tools.append((str(block.get("name")), inp if isinstance(inp, dict) else {}))
|
||||
out = {f: sum(v.get(f, 0) for v in per.values()) for f in USAGE_FIELDS}
|
||||
derived = 0
|
||||
for i, mid in enumerate(order[:-1]):
|
||||
if mid in complete:
|
||||
continue
|
||||
growth = (sum(per[order[i + 1]].get(f, 0) for f in INPUT_SIDE) - sum(per[mid].get(f, 0) for f in INPUT_SIDE)
|
||||
- int(between.get(mid, 0) / CHARS_PER_TOKEN_LOW))
|
||||
derived += max(0, growth - per[mid].get("output_tokens", 0))
|
||||
out["output_tokens_derived"] = derived
|
||||
return out, tools
|
||||
|
||||
|
||||
def harvest(run: Path, name: str, arm: str, transcripts: Path) -> tuple[dict[str, Any], list[str]]:
|
||||
"""Per task of the arm: tokens over all its runs and every isolation breach. Matched by the subagent's
|
||||
description S1-<run>-<arm>-<id> (the .meta.json beside the transcript); a transcript with no description is
|
||||
matched by the task path in its first prompt, one with another description is not this arm's. Returns the K5 tokens document and the tasks no transcript was found for."""
|
||||
if arm not in ARMS:
|
||||
raise RunError(f"unknown arm {arm!r}; the arms are {', '.join(ARMS)}")
|
||||
if not transcripts.is_dir():
|
||||
raise RunError(f"{transcripts} is not a directory")
|
||||
tasks_dir = run / "tasks" / arm
|
||||
ids = sorted(p.stem for p in tasks_dir.glob("*.md"))
|
||||
if not ids:
|
||||
raise RunError(f"no tasks in {tasks_dir}; run setup first")
|
||||
described = re.compile(rf"^S1-{re.escape(name)}-{re.escape(arm)}-(\S+)$")
|
||||
prompted = re.compile(re.escape(str(tasks_dir)) + r"/([A-Za-z0-9]+)\.md")
|
||||
basis = run / "arms" / arm
|
||||
rows: dict[str, dict[str, Any]] = {}
|
||||
for path in sorted(transcripts.rglob("*.jsonl")):
|
||||
meta_path = path.with_name(path.stem + ".meta.json")
|
||||
description = ""
|
||||
if meta_path.is_file():
|
||||
meta = load_json(meta_path, "agent meta")
|
||||
description = str(meta.get("description", "")) if isinstance(meta, dict) else ""
|
||||
lines = read_transcript(path)
|
||||
m = described.match(description) if description else prompted.search(first_prompt(lines))
|
||||
if m is None or m.group(1) not in ids:
|
||||
continue
|
||||
qid = m.group(1)
|
||||
counts, tools = usage_and_tools(lines)
|
||||
breaches = breaches_of(tools, basis, tasks_dir / f"{qid}.md", run / "answers" / arm / f"{qid}.json")
|
||||
row = rows.setdefault(qid, {"id": qid, **{f: 0 for f in counts}, "isolation_breaches": 0,
|
||||
"breach_detail": [], "runs": 0})
|
||||
for f, value in counts.items():
|
||||
row[f] += value
|
||||
row["breach_detail"] += breaches
|
||||
row["isolation_breaches"] = len(row["breach_detail"])
|
||||
row["runs"] += 1
|
||||
doc = {"schema": TOKENS_SCHEMA, "arm": arm, "questions": [rows[i] for i in sorted(rows)]}
|
||||
return doc, [i for i in ids if i not in rows]
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- score
|
||||
|
||||
|
||||
def mcnemar_exact(only_a: int, only_b: int) -> float:
|
||||
"""Two-sided exact McNemar: the binomial test over the discordant pairs, p = 0.5."""
|
||||
n, k = only_a + only_b, min(only_a, only_b)
|
||||
if n == 0:
|
||||
return 1.0
|
||||
return min(1.0, 2 * sum(comb(n, i) for i in range(k + 1)) / float(2 ** n))
|
||||
|
||||
|
||||
def wilson(x: int, n: int, z: float) -> tuple[float, float]:
|
||||
centre = (x + z * z / 2) / (n + z * z)
|
||||
half = z * sqrt(x * (n - x) / n + z * z / 4) / (n + z * z)
|
||||
return centre - half, centre + half
|
||||
|
||||
|
||||
def paired_difference_ci(e: int, f: int, g: int, h: int, level: float = 0.95) -> tuple[float, float]:
|
||||
"""The interval for (e + f)/n - (e + g)/n from paired data: e both, f first only, g second only, h neither.
|
||||
Newcombe (1998), Statist. Med. 17:2635-2650, method 10: Wilson score limits for each margin combined with the
|
||||
phi coefficient, its numerator max(eh - fg - n/2, 0) when eh > fg, and phi 0 when a margin is 0."""
|
||||
n = e + f + g + h
|
||||
if n == 0:
|
||||
raise RunError("no pairs")
|
||||
z = NormalDist().inv_cdf(1 - (1 - level) / 2)
|
||||
p1, p2 = (e + f) / n, (e + g) / n
|
||||
l1, u1 = wilson(e + f, n, z)
|
||||
l2, u2 = wilson(e + g, n, z)
|
||||
denominator = (e + f) * (g + h) * (e + g) * (f + h)
|
||||
numerator = float(e * h - f * g)
|
||||
if numerator > 0:
|
||||
numerator = max(numerator - n / 2, 0.0)
|
||||
phi = numerator / sqrt(denominator) if denominator else 0.0
|
||||
dl1, du1, dl2, du2 = p1 - l1, u1 - p1, p2 - l2, u2 - p2
|
||||
low = sqrt(max(0.0, dl1 * dl1 - 2 * phi * dl1 * du2 + du2 * du2))
|
||||
high = sqrt(max(0.0, du1 * du1 - 2 * phi * du1 * dl2 + dl2 * dl2))
|
||||
return p1 - p2 - low, p1 - p2 + high
|
||||
|
||||
|
||||
def min_detectable(n: int, discordant_share: float, alpha: float = ALPHA, power: float = POWER) -> float | None:
|
||||
"""The smallest difference between paired proportions that n pairs show at the given power with a two-sided
|
||||
McNemar test, for a share of discordant pairs psi (Connor 1987: n = (z_a sqrt(psi) + z_b sqrt(psi - d^2))^2 /
|
||||
d^2, solved for d by bisection). None when psi is 0 or no d up to psi is visible."""
|
||||
if n <= 0 or discordant_share <= 0:
|
||||
return None
|
||||
za, zb = NormalDist().inv_cdf(1 - alpha / 2), NormalDist().inv_cdf(power)
|
||||
psi = min(discordant_share, 1.0)
|
||||
|
||||
def needed(d: float) -> float:
|
||||
return (za * sqrt(psi) + zb * sqrt(max(0.0, psi - d * d))) ** 2 / (d * d)
|
||||
|
||||
if needed(psi) > n:
|
||||
return None
|
||||
low, high = 1e-9, psi
|
||||
for _ in range(200):
|
||||
mid = (low + high) / 2
|
||||
if needed(mid) > n:
|
||||
low = mid
|
||||
else:
|
||||
high = mid
|
||||
return high
|
||||
|
||||
|
||||
def load_results(path: Path) -> dict[str, Any]:
|
||||
data = load_json(path, "results")
|
||||
if not isinstance(data, dict) or data.get("schema") != RESULTS_SCHEMA or not isinstance(data.get("results"), list):
|
||||
raise RunError(f"{path} is not a results file with schema {RESULTS_SCHEMA}")
|
||||
if data.get("arm") not in ARMS:
|
||||
raise RunError(f"{path} is for arm {data.get('arm')!r}; the arms are {', '.join(ARMS)}")
|
||||
return data
|
||||
|
||||
|
||||
def path_band(paths: int) -> str:
|
||||
return "4+" if paths >= MANY_PATHS else "2-3"
|
||||
|
||||
|
||||
def score(results: list[dict[str, Any]], tokens: list[dict[str, Any]]) -> dict[str, Any]:
|
||||
by_arm = {r["arm"]: r for r in results}
|
||||
if sorted(by_arm) != sorted(ARMS) or len(results) != len(ARMS):
|
||||
raise RunError(f"score needs one results file for each arm: {', '.join(ARMS)}")
|
||||
a, b = by_arm["filsok"], by_arm["bundle"]
|
||||
if a.get("set_sha256") != b.get("set_sha256"):
|
||||
raise RunError("the two results files are for different sets")
|
||||
rows = {arm: {str(r["id"]): r for r in by_arm[arm]["results"]} for arm in ARMS}
|
||||
if sorted(rows["filsok"]) != sorted(rows["bundle"]):
|
||||
raise RunError("the two results files hold different questions")
|
||||
ids = sorted(rows["filsok"])
|
||||
doc: dict[str, Any] = {"schema": SCORE_SCHEMA, "set_sha256": a.get("set_sha256"), "questions": len(ids),
|
||||
"arms": {}, "pairs": {}}
|
||||
for arm in ARMS:
|
||||
groups = {g: [sum(1 for i in ids if rows[arm][i]["group"] == g and rows[arm][i]["correct"] is True),
|
||||
sum(1 for i in ids if rows[arm][i]["group"] == g)] for g in GROUPS}
|
||||
bands = {band: [sum(1 for i in ids if path_band(int(rows[arm][i]["paths_needed"])) == band
|
||||
and rows[arm][i]["correct"] is True),
|
||||
sum(1 for i in ids if path_band(int(rows[arm][i]["paths_needed"])) == band)]
|
||||
for band in ("2-3", "4+")}
|
||||
doc["arms"][arm] = {"groups": groups, "paths": bands,
|
||||
"all": [sum(1 for i in ids if rows[arm][i]["correct"] is True), len(ids)]}
|
||||
for label, members in [("all", ids)] + [(g, [i for i in ids if rows["filsok"][i]["group"] == g]) for g in GROUPS]:
|
||||
counts = {"both": 0, "only filsok": 0, "only bundle": 0, "neither": 0}
|
||||
for i in members:
|
||||
fa, bu = rows["filsok"][i]["correct"] is True, rows["bundle"][i]["correct"] is True
|
||||
counts["both" if fa and bu else "only filsok" if fa else "only bundle" if bu else "neither"] += 1
|
||||
doc["pairs"][label] = counts
|
||||
pairs = doc["pairs"]["all"]
|
||||
doc["mcnemar_p"] = mcnemar_exact(pairs["only filsok"], pairs["only bundle"])
|
||||
n = len(ids)
|
||||
if n:
|
||||
low, high = paired_difference_ci(pairs["both"], pairs["only bundle"], pairs["only filsok"], pairs["neither"])
|
||||
doc["difference"] = {"of": "bundle - filsok", "point": (pairs["only bundle"] - pairs["only filsok"]) / n,
|
||||
"low": low, "high": high, "level": 0.95, "method": "Newcombe 1998 method 10"}
|
||||
share = (pairs["only filsok"] + pairs["only bundle"]) / n
|
||||
doc["smallest_visible"] = {"discordant_share": share, "difference": min_detectable(n, share),
|
||||
"power": POWER, "alpha": ALPHA}
|
||||
doc["tokens"] = {}
|
||||
for t in tokens:
|
||||
if not isinstance(t, dict) or t.get("schema") != TOKENS_SCHEMA or t.get("arm") not in ARMS:
|
||||
raise RunError(f"a tokens file is not {TOKENS_SCHEMA} for one of the arms")
|
||||
qs = as_list(t.get("questions"))
|
||||
total = sum(int(q.get(f, 0)) for q in qs for f in (*USAGE_FIELDS, "output_tokens_derived"))
|
||||
doc["tokens"][t["arm"]] = {"questions": len(qs), "per_question": total / len(qs) if qs else None,
|
||||
"isolation_breaches": sum(int(q.get("isolation_breaches", 0)) for q in qs)}
|
||||
return doc
|
||||
|
||||
|
||||
def render(doc: dict[str, Any]) -> list[str]:
|
||||
lines = []
|
||||
for arm in ARMS:
|
||||
row = doc["arms"][arm]
|
||||
lines.append(f"{arm}: " + ", ".join(f"{g} {c} of {n}" for g, (c, n) in row["groups"].items())
|
||||
+ f", all {row['all'][0]} of {row['all'][1]}")
|
||||
for arm in ARMS:
|
||||
lines.append(f"{arm} by paths: " + ", ".join(f"{b} {c} of {n}" for b, (c, n) in doc["arms"][arm]["paths"].items()))
|
||||
for label, counts in doc["pairs"].items():
|
||||
lines.append(("pairs" if label == "all" else f"pairs {label}") + ": "
|
||||
+ ", ".join(f"{k} {v}" for k, v in counts.items()))
|
||||
lines.append(f"exact McNemar p = {doc['mcnemar_p']:.3g}")
|
||||
if "difference" in doc:
|
||||
d = doc["difference"]
|
||||
lines.append(f"bundle - filsok = {d['point']:.3f}, 95 % CI {d['low']:.3f} to {d['high']:.3f} "
|
||||
f"(paired, Newcombe 1998 method 10)")
|
||||
s = doc["smallest_visible"]
|
||||
visible = "none" if s["difference"] is None else f"{s['difference']:.3f}"
|
||||
lines.append(f"smallest difference visible at 80 % power, alpha 0.05: {visible} "
|
||||
f"(discordant share measured {s['discordant_share']:.3f}, N = {doc['questions']})")
|
||||
for arm, t in doc["tokens"].items():
|
||||
per = "none" if t["per_question"] is None else f"{t['per_question']:.0f}"
|
||||
lines.append(f"{arm} tokens: {t['questions']} questions harvested, {per} per question, "
|
||||
f"isolation breaches {t['isolation_breaches']}")
|
||||
return lines
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- main
|
||||
|
||||
|
||||
def build_parser() -> argparse.ArgumentParser:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
sub = p.add_subparsers(dest="command", required=True)
|
||||
s = sub.add_parser("setup", help="one arm's basis and task files")
|
||||
s.add_argument("--run", required=True)
|
||||
s.add_argument("--name", required=True, help="the run's name, letters and digits (S1-<name>-<arm>-<id>)")
|
||||
s.add_argument("--arm", required=True)
|
||||
s.add_argument("--set", required=True)
|
||||
s.add_argument("--sources", required=True, help=f"the K1 root measured (its kilde/ files and {K1_FILE})")
|
||||
s.add_argument("--bundle", help="the collection, for the bundle arm")
|
||||
s.add_argument("--part-root", help="the K1 root the set was made from, when it is a subset_of --sources: the arm "
|
||||
"searches the whole of --sources")
|
||||
c = sub.add_parser("collect", help="the answer files into the arm's answers file")
|
||||
c.add_argument("--run", required=True)
|
||||
c.add_argument("--arm", required=True)
|
||||
h = sub.add_parser("harvest", help="tokens and isolation breaches per question from the transcripts")
|
||||
h.add_argument("--run", required=True)
|
||||
h.add_argument("--name", required=True)
|
||||
h.add_argument("--arm", required=True)
|
||||
h.add_argument("--transcripts", required=True, help="a directory searched for *.jsonl at any depth")
|
||||
r = sub.add_parser("score", help="two arms' results: groups, paths, pairs, McNemar, interval, power")
|
||||
r.add_argument("--results", action="append", required=True, help="answerset.py gate --out, once per arm")
|
||||
r.add_argument("--tokens", action="append", default=[], help="harvest output, once per arm (optional)")
|
||||
r.add_argument("--out", default="")
|
||||
return p
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args = build_parser().parse_args(argv)
|
||||
try:
|
||||
if args.command == "setup":
|
||||
run = Path(args.run).resolve()
|
||||
tasks = setup_arm(run, args.name, args.arm, Path(args.set), Path(args.sources),
|
||||
Path(args.bundle).resolve() if args.bundle else None,
|
||||
Path(args.part_root) if args.part_root else None)
|
||||
for description, task in tasks:
|
||||
print(f"{description}\t{task}")
|
||||
return 0
|
||||
if args.command == "collect":
|
||||
out, problems = collect(Path(args.run).resolve(), args.arm)
|
||||
print(out)
|
||||
for problem in problems:
|
||||
print(f"problem: {problem}")
|
||||
return 1 if problems else 0
|
||||
if args.command == "harvest":
|
||||
run = Path(args.run).resolve()
|
||||
doc, missing = harvest(run, args.name, args.arm, Path(args.transcripts))
|
||||
write(run / "tokens" / f"{args.arm}.json", doc)
|
||||
breaches = sum(q["isolation_breaches"] for q in doc["questions"])
|
||||
print(f"tasks harvested: {len(doc['questions'])}; breaches: {breaches}")
|
||||
if missing:
|
||||
print(f"tasks without a transcript: {len(missing)}: {', '.join(missing)}")
|
||||
return 1 if missing else 0
|
||||
doc = score([load_results(Path(p)) for p in args.results],
|
||||
[load_json(Path(p), "tokens") for p in args.tokens])
|
||||
print("\n".join(render(doc)))
|
||||
if args.out:
|
||||
write(Path(args.out), doc)
|
||||
return 0
|
||||
except RunError as exc:
|
||||
print(f"FAIL: {exc}", file=sys.stderr)
|
||||
return 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
32
gates/answerset-del1-hard.manifest.json
Normal file
32
gates/answerset-del1-hard.manifest.json
Normal file
|
|
@ -0,0 +1,32 @@
|
|||
{
|
||||
"draw": "not split: random.Random(seed).shuffle(sorted ids), then renumbered x01.. in that order; the set is the part of its root",
|
||||
"groups": {
|
||||
"part": {
|
||||
"docs-2-3": 4,
|
||||
"docs-4": 26,
|
||||
"history": 0
|
||||
}
|
||||
},
|
||||
"part": {
|
||||
"file": "gates-local/answerset/part-del1-hard.json",
|
||||
"questions": 30,
|
||||
"sha256": "032635992a24a27b309c8c334bf56565e458ccf50ab7d0f26e8da6513fc4b414"
|
||||
},
|
||||
"rules": {
|
||||
"many_paths": 4,
|
||||
"min_many": 20,
|
||||
"min_paths": 3
|
||||
},
|
||||
"schema": "okf-wiki/answer-set-manifest/1",
|
||||
"seed": 20261011,
|
||||
"sources": {
|
||||
"copy": {
|
||||
"frozen_at": "2026-10-08T12:47:05Z",
|
||||
"manifest_sha256": "8c9f87fdc9e776f31cd8dca8c6a8593ba9f90889b52a43683187709af454a34f"
|
||||
},
|
||||
"part": {
|
||||
"frozen_at": "2026-10-08T19:13:44Z",
|
||||
"manifest_sha256": "0483d05250aeea186ec8f85e0728659461e97502618af5561abb6737dc5389fd"
|
||||
}
|
||||
}
|
||||
}
|
||||
51
gates/answerset.manifest.json
Normal file
51
gates/answerset.manifest.json
Normal file
|
|
@ -0,0 +1,51 @@
|
|||
{
|
||||
"draw": "per group in the order docs-2-3, docs-4, history: random.Random(seed).sample(sorted(ids of the group), quota) with one generator; quotas proportional to the group sizes (largest remainder); both parts renumbered d../h.. in the order of the original ids",
|
||||
"groups": {
|
||||
"all": {
|
||||
"docs-2-3": 35,
|
||||
"docs-4": 34,
|
||||
"history": 11
|
||||
},
|
||||
"held_out": {
|
||||
"docs-2-3": 9,
|
||||
"docs-4": 8,
|
||||
"history": 3
|
||||
},
|
||||
"part": {
|
||||
"docs-2-3": 9,
|
||||
"docs-4": 9,
|
||||
"history": 0
|
||||
},
|
||||
"set": {
|
||||
"docs-2-3": 26,
|
||||
"docs-4": 26,
|
||||
"history": 8
|
||||
}
|
||||
},
|
||||
"held_out": {
|
||||
"questions": 20,
|
||||
"sha256": "0971053aec5373e88be48fd7d732f6f075ac4efeb5b97462d51abfe5adab3470"
|
||||
},
|
||||
"part": {
|
||||
"file": "gates-local/answerset/part-del1.json",
|
||||
"questions": 18,
|
||||
"sha256": "5f8ec090a0d19af4e8ba4d72d9ca5774050581b8bd58f01b108f71860ef124e7"
|
||||
},
|
||||
"schema": "okf-wiki/answer-set-manifest/1",
|
||||
"seed": 20261009,
|
||||
"set": {
|
||||
"file": "gates-local/answerset/set-60.json",
|
||||
"questions": 60,
|
||||
"sha256": "100571b9e73d41d60e1ea2424256b375123c1a83333fd49a87fc8533aba0581e"
|
||||
},
|
||||
"sources": {
|
||||
"copy": {
|
||||
"frozen_at": "2026-10-08T12:47:05Z",
|
||||
"manifest_sha256": "8c9f87fdc9e776f31cd8dca8c6a8593ba9f90889b52a43683187709af454a34f"
|
||||
},
|
||||
"part": {
|
||||
"frozen_at": "2026-10-08T19:13:44Z",
|
||||
"manifest_sha256": "0483d05250aeea186ec8f85e0728659461e97502618af5561abb6737dc5389fd"
|
||||
}
|
||||
}
|
||||
}
|
||||
694
gates/answerset.py
Normal file
694
gates/answerset.py
Normal file
|
|
@ -0,0 +1,694 @@
|
|||
"""The question set for PLAN.md criterion 4 on collection 1 (contract K5 point 1): checker, part count, held-out
|
||||
split, answer gate and the stub arms. No model call.
|
||||
|
||||
check every span stands byte for byte in its file in a K1 root, every question has at least two distinct
|
||||
paths, its group follows K1 and the path count, every path carries an answer key that stands in that
|
||||
path's spans, no key is wording, every negated key stands negated. PASS only N of N with N > 0.
|
||||
For a harder set, --min-paths raises the paths every question needs and --min-many sets how many
|
||||
questions need 4 or more.
|
||||
inside the questions whose every span stands in a part root (a K1 subset root, here the first part).
|
||||
split draws the held-out questions with a fixed seed, stratified by group, writes them OUTSIDE the
|
||||
repository, renumbers both parts and writes the manifest (seed, counts, sha256 values only).
|
||||
freeze freezes a set not made by split: it must pass check against its own root with the thresholds and
|
||||
hold exactly --count questions. It is renumbered in a seeded order and gets a manifest of its own that
|
||||
names it the part of that root (seed, rules, counts, sha256 values, the roots' K1 sha256; no text),
|
||||
so gate judges it as it judges the part. A frozen set or manifest is never overwritten.
|
||||
gate scores one arm: a question is correct when the answer holds every key (multidoc.key_holds) AND the
|
||||
citations quote, verbatim, every path the question needs. A missing answers file scores 0
|
||||
(fail-closed). The set and the root must match the manifest.
|
||||
stub writes a known arm: `fasit` answers with the keys and quotes the spans; `no-citations` answers with
|
||||
the keys and cites nothing.
|
||||
|
||||
The group is set by the sources, never by the writer: `history` when a path has `main_topic: false` in K1,
|
||||
otherwise `docs-2-3` (2-3 distinct paths) or `docs-4` (4 or more).
|
||||
|
||||
Exit 0 = PASS / scored completely; 1 = FAIL or could not run. A check that cannot run is a FAIL.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import math
|
||||
import random
|
||||
import re
|
||||
import sys
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from keyform import count_keys
|
||||
from multidoc import KEY_MAX, KEY_MIN, SPAN_MAX, SPAN_MIN, key_holds, key_in, negated_in, normalise_quote, number_key
|
||||
|
||||
SET_SCHEMA = "okf-wiki/answer-set/1"
|
||||
MANIFEST_SCHEMA = "okf-wiki/answer-set-manifest/1"
|
||||
ANSWERS_SCHEMA = "okf-wiki/answers/1"
|
||||
RESULTS_SCHEMA = "okf-wiki/answer-results/1"
|
||||
K1_SCHEMA = "okf-wiki/source-manifest/1"
|
||||
K1_FILE = "source-manifest.json"
|
||||
GROUPS = ("docs-2-3", "docs-4", "history")
|
||||
ARMS = ("filsok", "bundle")
|
||||
MIN_PATHS = 2
|
||||
MANY_PATHS = 4
|
||||
PATH_RE = re.compile(r"^kilde/[a-z0-9._-]+$")
|
||||
REPO_ROOT = Path(__file__).resolve().parents[1]
|
||||
|
||||
|
||||
class RunError(Exception):
|
||||
"""The command cannot run; reported as FAIL with exit 1."""
|
||||
|
||||
|
||||
def dump(data: Any) -> bytes:
|
||||
"""The K5 file form: two writings of the same content are byte-identical."""
|
||||
return (json.dumps(data, indent=1, sort_keys=True, ensure_ascii=False) + "\n").encode("utf-8")
|
||||
|
||||
|
||||
def sha256_bytes(data: bytes) -> str:
|
||||
return hashlib.sha256(data).hexdigest()
|
||||
|
||||
|
||||
def load_json(path: Path, what: str) -> Any:
|
||||
try:
|
||||
return json.loads(path.read_text(encoding="utf-8"))
|
||||
except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
|
||||
raise RunError(f"cannot read {what} {path}: {exc}") from exc
|
||||
|
||||
|
||||
def as_list(value: object) -> list[Any]:
|
||||
return value if isinstance(value, list) else []
|
||||
|
||||
|
||||
def group_of(paths: list[str], main_topic: dict[str, bool]) -> str:
|
||||
"""The group K1 gives a question with these distinct paths (see the module text)."""
|
||||
distinct = set(paths)
|
||||
if any(main_topic.get(p) is False for p in distinct):
|
||||
return "history"
|
||||
return "docs-4" if len(distinct) >= MANY_PATHS else "docs-2-3"
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- K1 root
|
||||
|
||||
|
||||
class Root:
|
||||
"""A K1 root read with this module's own code (contract K1; okfwiki.sources is not imported)."""
|
||||
|
||||
def __init__(self, path: Path) -> None:
|
||||
self.path = path
|
||||
manifest = path / K1_FILE
|
||||
if not path.is_dir() or not manifest.is_file():
|
||||
raise RunError(f"{path} is not a K1 root: no {K1_FILE}")
|
||||
self.manifest_bytes = manifest.read_bytes()
|
||||
self.manifest_sha256 = sha256_bytes(self.manifest_bytes)
|
||||
try:
|
||||
data = json.loads(self.manifest_bytes.decode("utf-8"))
|
||||
except (UnicodeDecodeError, json.JSONDecodeError) as exc:
|
||||
raise RunError(f"{manifest} is not JSON: {exc}") from exc
|
||||
if not isinstance(data, dict) or data.get("schema") != K1_SCHEMA or not isinstance(data.get("sources"), list):
|
||||
raise RunError(f"{manifest} is not a K1 manifest with schema {K1_SCHEMA}")
|
||||
self.frozen_at = str(data.get("frozen_at", ""))
|
||||
self.entries: dict[str, dict[str, Any]] = {}
|
||||
for entry in data["sources"]:
|
||||
if not (isinstance(entry, dict) and isinstance(entry.get("path"), str)
|
||||
and isinstance(entry.get("main_topic"), bool) and isinstance(entry.get("sha256"), str)):
|
||||
raise RunError(f"{manifest} has an entry without path, main_topic or sha256: {entry!r}")
|
||||
self.entries[entry["path"]] = entry
|
||||
self.main_topic = {p: bool(e["main_topic"]) for p, e in self.entries.items()}
|
||||
self._text: dict[str, str | None] = {}
|
||||
self._bad: dict[str, str] = {}
|
||||
|
||||
def ref(self) -> dict[str, str]:
|
||||
return {"frozen_at": self.frozen_at, "manifest_sha256": self.manifest_sha256}
|
||||
|
||||
def text(self, path: str, verify: bool = True) -> str | None:
|
||||
"""The text of a listed source, or None (with the reason in problem()) when it is not listed, missing, not
|
||||
UTF-8, or (with verify) its sha256 differs from K1."""
|
||||
key = f"{verify}:{path}"
|
||||
if key not in self._text:
|
||||
self._text[key] = self._read(path, verify, key)
|
||||
return self._text[key]
|
||||
|
||||
def _read(self, path: str, verify: bool, key: str) -> str | None:
|
||||
if not PATH_RE.match(path) or path not in self.entries:
|
||||
self._bad[key] = f"{path} is not a path K1 lists in {self.path}"
|
||||
return None
|
||||
target = self.path / path
|
||||
try:
|
||||
data = target.read_bytes()
|
||||
except OSError:
|
||||
self._bad[key] = f"{path} is missing in {self.path}"
|
||||
return None
|
||||
if verify and sha256_bytes(data) != self.entries[path]["sha256"]:
|
||||
self._bad[key] = f"{path} sha256 differs from K1 in {self.path}"
|
||||
return None
|
||||
try:
|
||||
return data.decode("utf-8")
|
||||
except UnicodeDecodeError:
|
||||
self._bad[key] = f"{path} is not UTF-8"
|
||||
return None
|
||||
|
||||
def problem(self, path: str, verify: bool = True) -> str:
|
||||
return self._bad.get(f"{verify}:{path}", "")
|
||||
|
||||
|
||||
def load_set(path: Path) -> dict[str, Any]:
|
||||
data = load_json(path, "set")
|
||||
if (not isinstance(data, dict) or data.get("schema") != SET_SCHEMA or not isinstance(data.get("questions"), list)
|
||||
or not isinstance(data.get("sources"), dict)):
|
||||
raise RunError(f"{path} is not a set with schema {SET_SCHEMA}")
|
||||
return data
|
||||
|
||||
|
||||
def evidence_of(q: dict[str, Any]) -> list[tuple[str, str]]:
|
||||
"""(path, span) for every well-formed evidence entry."""
|
||||
out = []
|
||||
for e in as_list(q.get("evidence")):
|
||||
if isinstance(e, dict) and isinstance(e.get("path"), str) and isinstance(e.get("span"), str):
|
||||
out.append((e["path"], e["span"]))
|
||||
return out
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- check
|
||||
|
||||
|
||||
@dataclass
|
||||
class CheckReport:
|
||||
questions: int = 0
|
||||
spans: int = 0
|
||||
spans_verbatim: int = 0
|
||||
all_spans_verbatim: int = 0
|
||||
min_paths: int = 0
|
||||
group_ok: int = 0
|
||||
key_per_path: int = 0
|
||||
keys: int = 0
|
||||
wording_keys: int = 0
|
||||
negated_keys: int = 0
|
||||
negated_ok: int = 0
|
||||
groups: dict[str, int] = field(default_factory=lambda: {g: 0 for g in GROUPS})
|
||||
errors: list[str] = field(default_factory=list)
|
||||
paths_per_question: dict[str, int] = field(default_factory=dict)
|
||||
need_paths: int = MIN_PATHS
|
||||
need_many: int = 0
|
||||
many: int = 0
|
||||
|
||||
@property
|
||||
def passed(self) -> bool:
|
||||
n = self.questions
|
||||
return (n > 0 and not self.errors and self.spans_verbatim == self.spans and self.all_spans_verbatim == n
|
||||
and self.min_paths == n and self.group_ok == n and self.key_per_path == n
|
||||
and self.wording_keys == 0 and self.negated_ok == self.negated_keys and self.many >= self.need_many)
|
||||
|
||||
def lines(self) -> list[str]:
|
||||
n = self.questions
|
||||
out = [f"error: {e}" for e in self.errors]
|
||||
out += [
|
||||
f"questions: {n}",
|
||||
"groups: " + ", ".join(f"{g} {self.groups[g]}" for g in GROUPS),
|
||||
f"spans verbatim: {self.spans_verbatim} of {self.spans}",
|
||||
f"questions with every span verbatim: {self.all_spans_verbatim} of {n}",
|
||||
f"questions with >= {self.need_paths} distinct paths: {self.min_paths} of {n}",
|
||||
]
|
||||
if self.need_many:
|
||||
out.append(f"questions with >= {MANY_PATHS} distinct paths: {self.many} of {n} (required >= {self.need_many})")
|
||||
out += [
|
||||
f"questions whose group matches K1 and the path count: {self.group_ok} of {n}",
|
||||
f"questions with a key in every path's spans: {self.key_per_path} of {n}",
|
||||
f"answer keys that are wording: {self.wording_keys} of {self.keys}",
|
||||
f"negated keys negated in their span: {self.negated_ok} of {self.negated_keys}",
|
||||
"PASS" if self.passed else "FAIL",
|
||||
]
|
||||
return out
|
||||
|
||||
|
||||
def check_set(data: dict[str, Any], root: Root, need_paths: int = MIN_PATHS, need_many: int = 0) -> CheckReport:
|
||||
questions = data["questions"]
|
||||
rep = CheckReport(questions=len(questions), need_paths=need_paths, need_many=need_many)
|
||||
ref = data["sources"]
|
||||
if ref.get("manifest_sha256") != root.manifest_sha256 or ref.get("frozen_at") != root.frozen_at:
|
||||
rep.errors.append(f"the set names sources {ref.get('frozen_at')!r} manifest_sha256 "
|
||||
f"{ref.get('manifest_sha256')!r}, the root has {root.frozen_at!r} {root.manifest_sha256}")
|
||||
form = count_keys(questions)
|
||||
rep.keys, rep.wording_keys = form.keys, form.wording
|
||||
seen: set[str] = set()
|
||||
for q in questions:
|
||||
qid = q.get("id") if isinstance(q, dict) else None
|
||||
if not isinstance(qid, str) or not qid:
|
||||
rep.errors.append("question without an id")
|
||||
continue
|
||||
if qid in seen:
|
||||
rep.errors.append(f"duplicate id {qid}")
|
||||
seen.add(qid)
|
||||
if not isinstance(q.get("question"), str) or not q["question"].strip():
|
||||
rep.errors.append(f"{qid}: empty question")
|
||||
raw_evidence = as_list(q.get("evidence"))
|
||||
evidence = evidence_of(q)
|
||||
rep.spans += len(raw_evidence)
|
||||
if len(evidence) != len(raw_evidence):
|
||||
rep.errors.append(f"{qid}: an evidence entry without a path or a span")
|
||||
spans_by_path: dict[str, list[str]] = {}
|
||||
verbatim = 0
|
||||
for path, span in evidence:
|
||||
if not SPAN_MIN <= len(span) <= SPAN_MAX:
|
||||
rep.errors.append(f"{qid}: span length {len(span)} in {path} outside {SPAN_MIN}-{SPAN_MAX}")
|
||||
text = root.text(path)
|
||||
if text is None:
|
||||
rep.errors.append(f"{qid}: {root.problem(path)}")
|
||||
elif span in text:
|
||||
verbatim += 1
|
||||
spans_by_path.setdefault(path, []).append(span)
|
||||
rep.spans_verbatim += verbatim
|
||||
if raw_evidence and verbatim == len(raw_evidence):
|
||||
rep.all_spans_verbatim += 1
|
||||
paths = sorted(spans_by_path)
|
||||
rep.paths_per_question[qid] = len(paths)
|
||||
if len(paths) >= need_paths:
|
||||
rep.min_paths += 1
|
||||
if len(paths) >= MANY_PATHS:
|
||||
rep.many += 1
|
||||
group = q.get("group")
|
||||
if group in GROUPS:
|
||||
rep.groups[group] += 1
|
||||
else:
|
||||
rep.errors.append(f"{qid}: group {group!r} is not one of {', '.join(GROUPS)}")
|
||||
if paths and all(p in root.entries for p in paths) and group == group_of(paths, root.main_topic):
|
||||
rep.group_ok += 1
|
||||
keyed: set[str] = set()
|
||||
keys_ok = bool(as_list(q.get("answer_keys")))
|
||||
for k in as_list(q.get("answer_keys")):
|
||||
kpath, key = (k.get("path"), k.get("text")) if isinstance(k, dict) else (None, None)
|
||||
if not isinstance(key, str) or not isinstance(kpath, str):
|
||||
keys_ok = False
|
||||
rep.errors.append(f"{qid}: malformed answer key")
|
||||
continue
|
||||
if kpath not in spans_by_path:
|
||||
rep.errors.append(f"{qid}: answer key path {kpath} is not an evidence path")
|
||||
if number_key(key) is None and not KEY_MIN <= len(key.strip()) <= KEY_MAX:
|
||||
rep.errors.append(f"{qid}: key length {len(key.strip())} outside {KEY_MIN}-{KEY_MAX}")
|
||||
if any(key_in(key, s) for s in spans_by_path.get(kpath, [])):
|
||||
keyed.add(kpath)
|
||||
else:
|
||||
keys_ok = False
|
||||
if "negated" in k:
|
||||
if k["negated"] is not True:
|
||||
rep.errors.append(f"{qid}: negated must be true or absent")
|
||||
continue
|
||||
rep.negated_keys += 1
|
||||
rep.negated_ok += any(negated_in(key, s) for s in spans_by_path.get(kpath, []))
|
||||
if keys_ok and paths and keyed == set(paths):
|
||||
rep.key_per_path += 1
|
||||
return rep
|
||||
|
||||
|
||||
def thresholds(args: argparse.Namespace) -> tuple[int, int]:
|
||||
if args.min_paths < MIN_PATHS or args.min_many < 0:
|
||||
raise RunError(f"--min-paths must be at least {MIN_PATHS} and --min-many at least 0")
|
||||
return args.min_paths, args.min_many
|
||||
|
||||
|
||||
def cmd_check(args: argparse.Namespace) -> int:
|
||||
rep = check_set(load_set(Path(args.set)), Root(Path(args.sources)), *thresholds(args))
|
||||
if args.per_question:
|
||||
for qid, n in rep.paths_per_question.items():
|
||||
print(f"{qid}\t{n} paths")
|
||||
print("\n".join(rep.lines()))
|
||||
return 0 if rep.passed else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- inside
|
||||
|
||||
|
||||
def inside_ids(questions: list[dict[str, Any]], part: Root) -> list[str]:
|
||||
"""The ids of the questions whose every span stands in its file in the part root (path listed there)."""
|
||||
out = []
|
||||
for q in questions:
|
||||
evidence = evidence_of(q) if isinstance(q, dict) else []
|
||||
if evidence and len(evidence) == len(as_list(q.get("evidence"))):
|
||||
texts = [(part.text(p, verify=False), s) for p, s in evidence]
|
||||
if all(t is not None and s in t for t, s in texts):
|
||||
out.append(str(q.get("id")))
|
||||
return out
|
||||
|
||||
|
||||
def cmd_inside(args: argparse.Namespace) -> int:
|
||||
questions = load_set(Path(args.set))["questions"]
|
||||
part = Root(Path(args.part_root))
|
||||
ids = set(inside_ids(questions, part))
|
||||
n = len(questions)
|
||||
print(f"questions with every span in the part root: {len(ids)} of {n}")
|
||||
for g in GROUPS:
|
||||
members = [q for q in questions if isinstance(q, dict) and q.get("group") == g]
|
||||
print(f" {g}: {sum(1 for q in members if q.get('id') in ids)} of {len(members)}")
|
||||
if not args.require_all:
|
||||
return 0
|
||||
passed = n > 0 and len(ids) == n
|
||||
print("PASS" if passed else "FAIL")
|
||||
return 0 if passed else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- split
|
||||
|
||||
|
||||
def inside_repo(path: Path) -> bool:
|
||||
try:
|
||||
path.resolve().relative_to(REPO_ROOT)
|
||||
return True
|
||||
except ValueError:
|
||||
return False
|
||||
|
||||
|
||||
def quotas(counts: dict[str, int], n_held: int) -> dict[str, int]:
|
||||
"""Held-out questions per group, proportional to the group's size (largest remainder, ties in GROUPS order)."""
|
||||
total = sum(counts.values())
|
||||
exact = {g: n_held * counts[g] / total for g in GROUPS}
|
||||
out = {g: math.floor(exact[g]) for g in GROUPS}
|
||||
for g in sorted(GROUPS, key=lambda g: (-(exact[g] - out[g]), GROUPS.index(g)))[: n_held - sum(out.values())]:
|
||||
out[g] += 1
|
||||
return out
|
||||
|
||||
|
||||
def group_counts(questions: list[dict[str, Any]]) -> dict[str, int]:
|
||||
return {g: sum(1 for q in questions if q.get("group") == g) for g in GROUPS}
|
||||
|
||||
|
||||
def cmd_split(args: argparse.Namespace) -> int:
|
||||
held_path = Path(args.held_out)
|
||||
if inside_repo(held_path):
|
||||
raise RunError(f"held-out path {held_path} is inside the repository; it must live outside it")
|
||||
if held_path.exists():
|
||||
raise RunError(f"held-out file {held_path} already exists; a frozen draw is never redrawn")
|
||||
root, part_root = Root(Path(args.sources)), Root(Path(args.part_sources))
|
||||
data, part_data = load_set(Path(args.set)), load_set(Path(args.part))
|
||||
for label, d, r in (("set", data, root), ("part", part_data, part_root)):
|
||||
rep = check_set(d, r)
|
||||
if not rep.passed:
|
||||
print("\n".join(rep.lines()))
|
||||
raise RunError(f"the {label} fails the check; nothing was split")
|
||||
part_q = part_data["questions"]
|
||||
if len(inside_ids(part_q, part_root)) != len(part_q):
|
||||
raise RunError("a question of the part has a span outside the part root; nothing was split")
|
||||
questions = data["questions"]
|
||||
n_held = args.held_out_count
|
||||
if not 0 < n_held < len(questions):
|
||||
raise RunError(f"cannot hold out {n_held} of {len(questions)}")
|
||||
counts = group_counts(questions)
|
||||
per_group = quotas(counts, n_held)
|
||||
rng = random.Random(args.seed)
|
||||
held_ids: set[str] = set()
|
||||
for g in GROUPS:
|
||||
ids = sorted(q["id"] for q in questions if q["group"] == g)
|
||||
held_ids.update(rng.sample(ids, per_group[g]))
|
||||
by_id = {q["id"]: q for q in questions}
|
||||
ordered = sorted(by_id)
|
||||
dev = [dict(by_id[i], id=f"d{n:02d}") for n, i in enumerate((i for i in ordered if i not in held_ids), 1)]
|
||||
held = [dict(by_id[i], id=f"h{n:02d}") for n, i in enumerate((i for i in ordered if i in held_ids), 1)]
|
||||
dev_bytes = dump({"schema": SET_SCHEMA, "sources": data["sources"], "questions": dev})
|
||||
held_bytes = dump({"schema": SET_SCHEMA, "sources": data["sources"], "questions": held})
|
||||
part_bytes = Path(args.part).read_bytes()
|
||||
manifest = {
|
||||
"schema": MANIFEST_SCHEMA,
|
||||
"seed": args.seed,
|
||||
"draw": "per group in the order " + ", ".join(GROUPS) + ": random.Random(seed).sample(sorted(ids of the "
|
||||
"group), quota) with one generator; quotas proportional to the group sizes (largest remainder); "
|
||||
"both parts renumbered d../h.. in the order of the original ids",
|
||||
"groups": {"all": counts, "set": group_counts(dev), "held_out": group_counts(held),
|
||||
"part": group_counts(part_q)},
|
||||
"set": {"file": args.set_label or Path(args.set_out).name, "sha256": sha256_bytes(dev_bytes),
|
||||
"questions": len(dev)},
|
||||
"held_out": {"sha256": sha256_bytes(held_bytes), "questions": len(held)},
|
||||
"part": {"file": args.part_label or Path(args.part).name, "sha256": sha256_bytes(part_bytes),
|
||||
"questions": len(part_q)},
|
||||
"sources": {"copy": root.ref(), "part": part_root.ref()},
|
||||
}
|
||||
held_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
held_path.write_bytes(held_bytes)
|
||||
set_out = Path(args.set_out)
|
||||
set_out.parent.mkdir(parents=True, exist_ok=True)
|
||||
set_out.write_bytes(dev_bytes)
|
||||
manifest_path = Path(args.manifest)
|
||||
manifest_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
manifest_path.write_bytes(dump(manifest))
|
||||
print(f"seed: {args.seed}")
|
||||
print("held out per group: " + ", ".join(f"{g} {per_group[g]}" for g in GROUPS))
|
||||
print(f"set: {len(dev)} questions, sha256 {manifest['set']['sha256']}")
|
||||
print(f"held out: {len(held)} questions, sha256 {manifest['held_out']['sha256']}")
|
||||
print(f"part: {len(part_q)} questions, sha256 {manifest['part']['sha256']}")
|
||||
print(f"manifest: {manifest_path}")
|
||||
return 0
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- freeze
|
||||
|
||||
|
||||
def cmd_freeze(args: argparse.Namespace) -> int:
|
||||
set_out, manifest_path = Path(args.set_out), Path(args.manifest)
|
||||
for target in (set_out, manifest_path):
|
||||
if target.exists():
|
||||
raise RunError(f"{target} already exists; a frozen set is never frozen again")
|
||||
need_paths, need_many = thresholds(args)
|
||||
root = Root(Path(args.sources))
|
||||
data = load_set(Path(args.set))
|
||||
rep = check_set(data, root, need_paths, need_many)
|
||||
print("\n".join(rep.lines()))
|
||||
if not rep.passed:
|
||||
raise RunError("the set fails the check; nothing was frozen")
|
||||
questions = data["questions"]
|
||||
if len(questions) != args.count:
|
||||
raise RunError(f"the set holds {len(questions)} questions, not {args.count}; nothing was frozen")
|
||||
order = sorted(q["id"] for q in questions)
|
||||
random.Random(args.seed).shuffle(order)
|
||||
by_id = {q["id"]: q for q in questions}
|
||||
frozen = [dict(by_id[i], id=f"{args.prefix}{n:02d}") for n, i in enumerate(order, 1)]
|
||||
set_bytes = dump({"schema": SET_SCHEMA, "sources": data["sources"], "questions": frozen})
|
||||
raw = json.loads(root.manifest_bytes.decode("utf-8")).get("subset_of")
|
||||
sources = {"part": root.ref()}
|
||||
if isinstance(raw, dict) and isinstance(raw.get("manifest_sha256"), str):
|
||||
sources["copy"] = {"frozen_at": str(raw.get("frozen_at", "")), "manifest_sha256": raw["manifest_sha256"]}
|
||||
manifest = {
|
||||
"schema": MANIFEST_SCHEMA,
|
||||
"seed": args.seed,
|
||||
"draw": "not split: random.Random(seed).shuffle(sorted ids), then renumbered "
|
||||
f"{args.prefix}01.. in that order; the set is the part of its root",
|
||||
"rules": {"min_paths": need_paths, "many_paths": MANY_PATHS, "min_many": need_many},
|
||||
"groups": {"part": group_counts(frozen)},
|
||||
"part": {"file": args.set_label or set_out.name, "sha256": sha256_bytes(set_bytes), "questions": len(frozen)},
|
||||
"sources": sources,
|
||||
}
|
||||
set_out.parent.mkdir(parents=True, exist_ok=True)
|
||||
set_out.write_bytes(set_bytes)
|
||||
manifest_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
manifest_path.write_bytes(dump(manifest))
|
||||
print(f"seed: {args.seed}")
|
||||
print(f"frozen: {len(frozen)} questions, sha256 {manifest['part']['sha256']}")
|
||||
print(f"manifest: {manifest_path}")
|
||||
return 0
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- gate
|
||||
|
||||
|
||||
@dataclass
|
||||
class Scored:
|
||||
qid: str
|
||||
group: str
|
||||
answered: bool
|
||||
keys_held: int
|
||||
keys: int
|
||||
paths_cited: int
|
||||
paths_needed: int
|
||||
|
||||
@property
|
||||
def answer_ok(self) -> bool:
|
||||
return self.answered and self.keys_held == self.keys
|
||||
|
||||
@property
|
||||
def source_ok(self) -> bool:
|
||||
return self.answered and self.paths_cited == self.paths_needed
|
||||
|
||||
@property
|
||||
def correct(self) -> bool:
|
||||
return self.answer_ok and self.source_ok
|
||||
|
||||
def row(self) -> dict[str, Any]:
|
||||
return {"id": self.qid, "group": self.group, "correct": self.correct, "keys_held": self.keys_held,
|
||||
"keys": self.keys, "paths_cited": self.paths_cited, "paths_needed": self.paths_needed}
|
||||
|
||||
|
||||
def cited_path(value: object) -> str | None:
|
||||
"""A citation's path as K1 writes it; a leading "/" (a page's resource form) is accepted."""
|
||||
if not isinstance(value, str):
|
||||
return None
|
||||
return value[1:] if value.startswith("/") else value
|
||||
|
||||
|
||||
def score_one(q: dict[str, Any], a: dict[str, Any] | None, root: Root) -> Scored:
|
||||
keys = as_list(q.get("answer_keys"))
|
||||
needed = sorted({p for p, _ in evidence_of(q)})
|
||||
if a is None:
|
||||
return Scored(q["id"], str(q.get("group")), False, 0, len(keys), 0, len(needed))
|
||||
answer = a.get("answer") if isinstance(a.get("answer"), str) else ""
|
||||
held = sum(1 for k in keys if isinstance(k, dict) and key_holds(k, str(answer)))
|
||||
supported: set[str] = set()
|
||||
for c in as_list(a.get("citations")):
|
||||
path = cited_path(c.get("path")) if isinstance(c, dict) else None
|
||||
quote = c.get("quote") if isinstance(c, dict) else None
|
||||
if path not in needed or not isinstance(quote, str) or not quote.strip():
|
||||
continue
|
||||
text = root.text(path, verify=False)
|
||||
if text is not None and normalise_quote(quote) in normalise_quote(text):
|
||||
supported.add(path)
|
||||
return Scored(q["id"], str(q.get("group")), True, held, len(keys), len(supported), len(needed))
|
||||
|
||||
|
||||
def manifest_part(manifest: dict[str, Any], set_hash: str) -> tuple[str, str]:
|
||||
"""(part name, the K1 manifest sha256 its root must have)."""
|
||||
sources = manifest.get("sources", {})
|
||||
for name, root_name in (("set", "copy"), ("held_out", "copy"), ("part", "part")):
|
||||
if manifest.get(name, {}).get("sha256") == set_hash:
|
||||
return name, str(sources.get(root_name, {}).get("manifest_sha256"))
|
||||
raise RunError(f"set sha256 {set_hash or '(missing)'} matches no part of the manifest")
|
||||
|
||||
|
||||
def cmd_gate(args: argparse.Namespace) -> int:
|
||||
if args.arm not in ARMS:
|
||||
raise RunError(f"unknown arm {args.arm!r}; the arms are {', '.join(ARMS)}")
|
||||
set_path = Path(args.set)
|
||||
manifest = load_json(Path(args.manifest), "manifest")
|
||||
if not isinstance(manifest, dict) or manifest.get("schema") != MANIFEST_SCHEMA:
|
||||
raise RunError(f"{args.manifest} is not a manifest with schema {MANIFEST_SCHEMA}")
|
||||
set_hash = sha256_bytes(set_path.read_bytes()) if set_path.is_file() else ""
|
||||
part, root_sha = manifest_part(manifest, set_hash)
|
||||
root = Root(Path(args.sources))
|
||||
if root.manifest_sha256 != root_sha:
|
||||
raise RunError(f"the root's {K1_FILE} sha256 {root.manifest_sha256} is not the manifest's {root_sha} "
|
||||
f"for the {part}")
|
||||
questions = load_set(set_path)["questions"]
|
||||
n = len(questions)
|
||||
answers: dict[str, dict[str, Any]] = {}
|
||||
fail_closed = ""
|
||||
answers_path = Path(args.answers)
|
||||
if not answers_path.is_file():
|
||||
fail_closed = f"no answers for arm {args.arm} at {answers_path}"
|
||||
else:
|
||||
data = load_json(answers_path, "answers")
|
||||
if not isinstance(data, dict) or data.get("schema") != ANSWERS_SCHEMA or not isinstance(data.get("answers"),
|
||||
list):
|
||||
raise RunError(f"{answers_path} is not an answers file with schema {ANSWERS_SCHEMA}")
|
||||
if data.get("arm") != args.arm:
|
||||
raise RunError(f"answers file is for arm {data.get('arm')!r}, not {args.arm!r}")
|
||||
for a in data["answers"]:
|
||||
if isinstance(a, dict) and isinstance(a.get("id"), str):
|
||||
answers[a["id"]] = a
|
||||
results = sorted((score_one(q, answers.get(q["id"]), root) for q in questions), key=lambda r: r.qid)
|
||||
missing = sum(1 for r in results if not r.answered)
|
||||
correct = sum(r.correct for r in results)
|
||||
print(f"set: {part}, {n} questions; arm: {args.arm}")
|
||||
if fail_closed:
|
||||
print(f"fail-closed: {fail_closed}")
|
||||
print(f"missing answers: {missing}")
|
||||
print(f"correct answer: {sum(r.answer_ok for r in results)} of {n}")
|
||||
print(f"correct source: {sum(r.source_ok for r in results)} of {n}")
|
||||
print(f"correct answer and source: {correct} of {n}")
|
||||
for g in GROUPS:
|
||||
members = [r for r in results if r.group == g]
|
||||
print(f" {g}: {sum(r.correct for r in members)} of {len(members)}")
|
||||
if args.out:
|
||||
out = {"schema": RESULTS_SCHEMA, "arm": args.arm, "set_sha256": set_hash, "results": [r.row() for r in results]}
|
||||
Path(args.out).parent.mkdir(parents=True, exist_ok=True)
|
||||
Path(args.out).write_bytes(dump(out))
|
||||
scored = not fail_closed and missing == 0
|
||||
print("GREEN" if scored and n > 0 and correct == n else "RED")
|
||||
print("SCORED" if scored else "UNSCORED (fail-closed)" if fail_closed else "INCOMPLETE")
|
||||
return 0 if scored else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- stub
|
||||
|
||||
|
||||
def stub_answer(q: dict[str, Any], kind: str) -> dict[str, Any]:
|
||||
"""fasit: every key as its own sentence ("not <key>" when negated) and every span quoted; no-citations: the
|
||||
same answer with no citations."""
|
||||
keys = [k for k in as_list(q.get("answer_keys")) if isinstance(k, dict) and isinstance(k.get("text"), str)]
|
||||
text = " ".join(("not " if k.get("negated") is True else "") + k["text"] + "." for k in keys)
|
||||
citations = [{"path": p, "quote": s} for p, s in evidence_of(q)] if kind == "fasit" else []
|
||||
return {"id": q["id"], "answer": text, "citations": citations}
|
||||
|
||||
|
||||
def cmd_stub(args: argparse.Namespace) -> int:
|
||||
if args.arm not in ARMS:
|
||||
raise RunError(f"unknown arm {args.arm!r}; the arms are {', '.join(ARMS)}")
|
||||
questions = load_set(Path(args.set))["questions"]
|
||||
doc = {"schema": ANSWERS_SCHEMA, "arm": args.arm, "answers": [stub_answer(q, args.kind) for q in questions]}
|
||||
out = Path(args.out)
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_bytes(dump(doc))
|
||||
print(f"{args.kind}: {len(questions)} answers for arm {args.arm} at {out}")
|
||||
return 0
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- main
|
||||
|
||||
|
||||
def build_parser() -> argparse.ArgumentParser:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
sub = p.add_subparsers(dest="command", required=True)
|
||||
c = sub.add_parser("check", help="check spans, paths, groups and answer keys against a K1 root")
|
||||
c.add_argument("--set", required=True)
|
||||
c.add_argument("--sources", required=True, help="the K1 root the set was made from")
|
||||
c.add_argument("--per-question", action="store_true", help="also list distinct paths per question")
|
||||
c.add_argument("--min-paths", type=int, default=MIN_PATHS, help="distinct paths every question needs")
|
||||
c.add_argument("--min-many", type=int, default=0, help=f"questions that need {MANY_PATHS} or more paths")
|
||||
i = sub.add_parser("inside", help="count the questions whose every span stands in a part root")
|
||||
i.add_argument("--set", required=True)
|
||||
i.add_argument("--part-root", required=True)
|
||||
i.add_argument("--require-all", action="store_true", help="FAIL unless every question is inside")
|
||||
s = sub.add_parser("split", help="draw the held-out questions with a fixed seed, stratified by group")
|
||||
s.add_argument("--set", required=True)
|
||||
s.add_argument("--sources", required=True)
|
||||
s.add_argument("--seed", type=int, required=True)
|
||||
s.add_argument("--held-out-count", type=int, default=20)
|
||||
s.add_argument("--set-out", required=True)
|
||||
s.add_argument("--set-label", default="", help="path of the set as written in the manifest")
|
||||
s.add_argument("--held-out", required=True, help="file path OUTSIDE the repository; never overwritten")
|
||||
s.add_argument("--part", required=True, help="the separate set for the first part (not split)")
|
||||
s.add_argument("--part-sources", required=True, help="the part's K1 root")
|
||||
s.add_argument("--part-label", default="", help="path of the part as written in the manifest")
|
||||
s.add_argument("--manifest", required=True)
|
||||
f = sub.add_parser("freeze", help="freeze a set not made by split, with a manifest of its own")
|
||||
f.add_argument("--set", required=True)
|
||||
f.add_argument("--sources", required=True, help="the K1 root the set names (a subset root such as the part)")
|
||||
f.add_argument("--seed", type=int, required=True)
|
||||
f.add_argument("--count", type=int, required=True, help="the number of questions the set must hold")
|
||||
f.add_argument("--min-paths", type=int, required=True)
|
||||
f.add_argument("--min-many", type=int, required=True)
|
||||
f.add_argument("--prefix", required=True, help="id prefix of the frozen questions")
|
||||
f.add_argument("--set-out", required=True)
|
||||
f.add_argument("--set-label", default="", help="path of the set as written in the manifest")
|
||||
f.add_argument("--manifest", required=True)
|
||||
g = sub.add_parser("gate", help="score one arm's answers")
|
||||
g.add_argument("--set", required=True)
|
||||
g.add_argument("--manifest", required=True)
|
||||
g.add_argument("--sources", required=True)
|
||||
g.add_argument("--arm", required=True)
|
||||
g.add_argument("--answers", required=True)
|
||||
g.add_argument("--out", default="", help="write the per-question results (K5 point 5)")
|
||||
t = sub.add_parser("stub", help="write a known arm's answers (fasit or no-citations)")
|
||||
t.add_argument("--set", required=True)
|
||||
t.add_argument("--arm", required=True)
|
||||
t.add_argument("--kind", required=True, choices=("fasit", "no-citations"))
|
||||
t.add_argument("--out", required=True)
|
||||
return p
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args = build_parser().parse_args(argv)
|
||||
commands = {"check": cmd_check, "inside": cmd_inside, "split": cmd_split, "freeze": cmd_freeze, "gate": cmd_gate,
|
||||
"stub": cmd_stub}
|
||||
try:
|
||||
return commands[args.command](args)
|
||||
except RunError as exc:
|
||||
print(f"FAIL: {exc}", file=sys.stderr)
|
||||
return 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
107
gates/calibration-folder.manifest.json
Normal file
107
gates/calibration-folder.manifest.json
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
{
|
||||
"schema": "okf-wiki/calibration-manifest/2",
|
||||
"seed": 20261108,
|
||||
"shuffle": "random.Random(seed): each label's pairs sorted by draft id and shuffled; the first `lone` unsupported pairs alone in a call at positions spread evenly over 1-10, the rest `dense` to a call; counted supported pairs one per call in call order; filler pairs fill every call to ten in a cycle, never twice in a call and never beside the unsupported pair they are the twin of; the other positions shuffled; ids renumbered c001.. in call order, calls k001..",
|
||||
"set": {
|
||||
"file": "gates/calibration-folder/pairs.json",
|
||||
"sha256": "852b50d8fcd7ef248b40d43ec7b8dc28379163e6ddd64c37e40f3b170b97b61e",
|
||||
"pairs": 1400
|
||||
},
|
||||
"key": {
|
||||
"file": "gates/calibration-folder/key.json",
|
||||
"sha256": "af728bba35f4fe2947163adf3053f0b7bc3390ca5179df19f3b8abdd8fe0f641"
|
||||
},
|
||||
"calls": {
|
||||
"file": "gates/calibration-folder/calls.json",
|
||||
"sha256": "92125fd9772358fe02a4fd62fbfd09d4b64b24954b4be8fdb4464733d9125728",
|
||||
"calls": 140,
|
||||
"call_size_max": 10,
|
||||
"unsupported_alone": 100,
|
||||
"unsupported_per_call": {
|
||||
"1": 100,
|
||||
"5": 40
|
||||
},
|
||||
"built_with": {
|
||||
"lone": 100,
|
||||
"dense": 5
|
||||
}
|
||||
},
|
||||
"counts": {
|
||||
"unsupported": 300,
|
||||
"supported": 50,
|
||||
"filler-supported": 1050,
|
||||
"filler_distinct": 300,
|
||||
"by_kind": {
|
||||
"overgeneralisation": 50,
|
||||
"wrong-number": 50,
|
||||
"swapped-cause": 50,
|
||||
"not-mentioned": 50,
|
||||
"reversed-polarity": 50,
|
||||
"conflated-sources": 50,
|
||||
"paraphrase": 15,
|
||||
"inference": 20,
|
||||
"synthesis": 15
|
||||
},
|
||||
"by_position": {
|
||||
"1": {
|
||||
"unsupported": 32,
|
||||
"supported": 9,
|
||||
"filler-supported": 99
|
||||
},
|
||||
"2": {
|
||||
"unsupported": 33,
|
||||
"supported": 6,
|
||||
"filler-supported": 101
|
||||
},
|
||||
"3": {
|
||||
"unsupported": 25,
|
||||
"supported": 5,
|
||||
"filler-supported": 110
|
||||
},
|
||||
"4": {
|
||||
"unsupported": 27,
|
||||
"supported": 4,
|
||||
"filler-supported": 109
|
||||
},
|
||||
"5": {
|
||||
"unsupported": 30,
|
||||
"supported": 3,
|
||||
"filler-supported": 107
|
||||
},
|
||||
"6": {
|
||||
"unsupported": 32,
|
||||
"supported": 4,
|
||||
"filler-supported": 104
|
||||
},
|
||||
"7": {
|
||||
"unsupported": 28,
|
||||
"supported": 6,
|
||||
"filler-supported": 106
|
||||
},
|
||||
"8": {
|
||||
"unsupported": 31,
|
||||
"supported": 2,
|
||||
"filler-supported": 107
|
||||
},
|
||||
"9": {
|
||||
"unsupported": 33,
|
||||
"supported": 3,
|
||||
"filler-supported": 104
|
||||
},
|
||||
"10": {
|
||||
"unsupported": 29,
|
||||
"supported": 8,
|
||||
"filler-supported": 103
|
||||
}
|
||||
}
|
||||
},
|
||||
"thresholds": {
|
||||
"caught_min": 300,
|
||||
"accepted_min": 45
|
||||
},
|
||||
"sources": {
|
||||
"files": 32,
|
||||
"tree_sha256": "79c4c8e59ae932284ba009d7177593c5bbb9692c3bbc9d6e492b092536ab31f7"
|
||||
},
|
||||
"history": []
|
||||
}
|
||||
2107
gates/calibration-folder/calls.json
Normal file
2107
gates/calibration-folder/calls.json
Normal file
File diff suppressed because it is too large
Load diff
9421
gates/calibration-folder/draft.json
Normal file
9421
gates/calibration-folder/draft.json
Normal file
File diff suppressed because it is too large
Load diff
9805
gates/calibration-folder/key.json
Normal file
9805
gates/calibration-folder/key.json
Normal file
File diff suppressed because it is too large
Load diff
9
gates/calibration-folder/kilde/aarsberetning-2025.pdf.md
Normal file
9
gates/calibration-folder/kilde/aarsberetning-2025.pdf.md
Normal file
|
|
@ -0,0 +1,9 @@
|
|||
Årsberetning og regnskapsomtale 2025 Årsberetning og regnskapsomtale 2025 for Kvernhaugen borettslag Denne årsberetningen er skrevet av styret og ble godkjent i styremøtet 10. mars 2026. Den legges fram for den ordinære generalforsamlingen 21. april 2026 sammen med årsregnskapet og revisors beretning. Beretningen omtaler regnskapsåret fra 1. januar til 31. desember 2025. 1. Om borettslaget Kvernhaugen borettslag ble stiftet i 1978, og de første beboerne flyttet inn i 1979. Borettslaget har 72 leiligheter fordelt på fire blokker, A, B, C og D, med adressene Kvernhaugen 2, 4, 6 og 8. Blokk A og blokk B har 18 leiligheter hver, blokk C har 20 leiligheter, og blokk D har 16 leiligheter. Alle blokkene har fire etasjer og kjeller, men ingen av dem har heis. Leilighetstype Antall Bruksareal (m²) 2-roms 24 52 3-roms 36 71 4-roms 12 88 Sum 72 Til eiendommen hører en felles hage på 1 450 m² mellom blokkene, en lekeplass, en felles grillplass, to sykkelboder, et fellesvaskeri i kjelleren i blokk B, et fellesrom i første etasje i blokk D og et hobbyverksted i kjelleren i blokk C. Hver leilighet har i tillegg en kjellerbod på mellom 4 og 6 m². I 2025 ble 9 leiligheter solgt, mot 6 i 2024. Borettslaget har en vaktmester, Kåre Myrvold, i 80 % stilling. Forretningsførsel og revisjon kjøpes inn som tjenester. Avtalen om forretningsførsel ble reforhandlet høsten 2025, og den nye avtalen gjelder fra 1. januar 2026 med en årlig pris som er om lag 3,6 % høyere enn før. 2. Styrets arbeid i 2025 Den ordinære generalforsamlingen ble holdt 23. april 2025. Der var 41 stemmeberettigede andelseiere til stede, og 9 andelseiere var representert ved fullmakt, til sammen 50 stemmer. Generalforsamlingen fastsatte styrehonoraret for 2025 til 180 000 kr, mot 165 000 kr for 2024. Etter generalforsamlingen 2025 har styret bestått av: Ingrid Solbakk, styreleder Torbjørn Leirvik, nestleder og ansvarlig for bygninger Amira Hassani, styremedlem og ansvarlig for økonomi Henrik Vågsæther, styremedlem og ansvarlig for uteområder Marit Kolstø-Berge, styremedlem og sekretær
|
||||
|
||||
Liv Brennhaug og Simen Ørbekk, varamedlemmer Styrehonoraret på 180 000 kr ble fordelt slik: styrelederen fikk 60 000 kr, nestlederen 35 000 kr, de tre andre styremedlemmene 25 000 kr hver og de to varamedlemmene 5 000 kr hver. Styret holdt 11 styremøter i 2025 og behandlet 96 saker. De største sakene var forprosjektet for rehabilitering av balkonger og fasader, utskiftingen av belysningen i trapperommene og et økende antall klager på støy. Styret mottok 14 skriftlige klager på støy i 2025, mot 9 i 2024. De fleste klagene gjaldt støy mellom kl. 22:00 og kl. 23:00 på kvelder fra søndag til torsdag. Som følge av dette foreslår styret for generalforsamlingen 2026 at ro-tiden på kvelder fra søndag til torsdag skal begynne kl. 22:00 i stedet for kl. 23:00. En ekstraordinær generalforsamling ble holdt 18. november 2025. Den vedtok å sette i gang et forprosjekt for balkonger og fasader med en kostnadsramme på 1,2 mill. kr. Vedtaket ble gjort med 47 stemmer for og 3 stemmer mot. Styret gjennomførte en beboerundersøkelse i september 2025. 51 av 72 husstander svarte, som gir en svarprosent på 71. Det mest etterspurte tiltaket var et bedre kjøkken i fellesrommet, og deretter flere plasser i sykkelbodene. 3. Fellesarealer og dugnader Fellesvaskeriet har fire vaskemaskiner, to tørketromler og ett tørkerom. I september 2025 ble de to eldste vaskemaskinene byttet ut for til sammen 64 800 kr. De to andre maskinene er fra 2014 og skal etter planen byttes i 2027. Prisene i vaskeriet har vært uendret siden 2023. Inntektene fra vaskeriet var 61 750 kr i 2025, mot 58 900 kr i 2024. Fellesrommet i blokk D er 62 m² og har plass til 40 personer. Fra 1. januar 2025 ble leien økt fra 400 kr til 450 kr per døgn, fordi rengjøring etter utleie ble satt bort. Rommet ble leid ut i 64 døgn i 2025, mot 51 døgn i 2024. Leieinntektene var 28 800 kr i 2025 og 20 400 kr i 2024. Rommet leies bare ut til beboere. Hobbyverkstedet i kjelleren i blokk C åpnet 1. mars 2025 i et tidligere lagerrom på 24 m². Ombyggingen kostet 52 300 kr. Verkstedet er gratis for beboerne, og det har vært brukt jevnt gjennom året, særlig av beboere som reparerer sykler og møbler. Borettslaget har to sykkelboder. Sykkelbod nord ved blokk A har 48 plasser, og sykkelbod sør ved blokk D har 36 plasser. I mai 2025 fikk sykkelbod nord en ny port med nøkkelbrikkeleser, og i sykkelbod sør ble det satt opp et ladeskap med 8 låsbare rom for elsykkelbatterier. Ved den årlige ryddeaksjonen i oktober ble 23 umerkede sykler fjernet. 5 av dem ble hentet av eierne innen fristen, og de 18 andre ble gitt bort til gjenbruk. Det ble holdt to dugnader i 2025. På vårdugnaden 26. april deltok 41 av 72 husstander, og på høstdugnaden 18. oktober deltok 35 husstander. Hagekomiteen, som har fem medlemmer og ledes av Vibeke Strand, organiserer dugnadene. De 12 hageparsellene på 8
|
||||
|
||||
m² hver ble fordelt ved loddtrekning i mars 2025 for sesongene 2025 og 2026. Ved årsskiftet sto 9 husstander på venteliste for parsell. I januar og februar 2025 ble det registrert 4 innbrudd i kjellerboder i blokk C. I alle tilfellene var kjellerdøra blitt stående åpen. Styret sendte derfor ut et skriv om at kjellerdørene alltid skal være lukket og låst, og etter mars 2025 er det ikke registrert flere innbrudd. Ingen av innbruddene førte til kostnader for borettslaget, fordi de stjålne gjenstandene var dekket av beboernes egne forsikringer. Vaktmesteren registrerte 312 feilmeldinger fra beboerne i 2025, mot 278 i 2024. 87 % av feilmeldingene ble utbedret innen 14 dager. De vanligste feilene gjaldt dørpumper, belysning i kjellerboder og vaskemaskiner som stoppet midt i et program. Det ble solgt 24 ekstra nøkkelbrikker i 2025 til 400 kr per stykk, til sammen 9 600 kr. I 2024 ble det solgt 30 ekstra brikker til samme pris. 4. Vedlikehold og energi Kostnadene til vedlikehold og reparasjoner var 1 236 800 kr i 2025, mot 874 500 kr i 2024. Økningen skyldes i hovedsak tre store poster: ny LED-belysning i trapperommene, reparasjonen av en taklekkasje i blokk C og tilstandsrapporten for balkonger og fasader. Tilstandsrapporten kostet 148 500 kr og er første del av forprosjektet. LED-belysning med bevegelsessensor i alle trapperom ble montert i mars 2025 for 186 000 kr. Taklekkasjen over oppgang 2 i blokk C ble oppdaget etter kraftig regn 2. juni og reparert for 214 600 kr. Trapperommene i blokk D ble malt i august 2025 for 167 300 kr. Alle stigeledninger ble inspisert med kamera i oktober 2025 for 72 500 kr, som grunnlag for rørfornying. Det oppsto tre vannskader i leiligheter i blokk A og C. Borettslaget betalte egenandel på 10 000 kr for hver skade. Strømforbruket i fellesarealene gikk ned fra 241 300 kWh i 2024 til 214 600 kWh i 2025, en nedgang på 11 %. Nedgangen skyldes først og fremst at den nye belysningen i trapperommene bare er på når noen beveger seg der. Energikostnaden i fellesarealene falt fra 352 100 kr til 318 700 kr. Forsikringspremien økte med 10 %, fra 441 800 kr til 486 000 kr. Forsikringsselskapet begrunnet økningen med antall vannskader i borettslaget de siste tre årene. Styret minner om at beboeren selv betaler egenandelen når en vannskade skyldes utstyr i leiligheten som beboeren har ansvar for, for eksempel en vaskemaskin eller oppvaskmaskin. Ingen av de tre skadene i 2025 hadde en slik årsak, og derfor ble egenandelene belastet borettslaget.
|
||||
|
||||
\5. Regnskap 2025 Årsregnskapet viser et årsresultat på 745 950 kr, mot 992 000 kr i 2024. Driftsresultatet ble 1 195 850 kr. Nedgangen fra året før skyldes i hovedsak høyere vedlikeholdskostnader. Tabellen viser hovedtallene. Post 2025 (kr) 2024 (kr) Felleskostnader 5 184 000 4 984 600 Andre driftsinntekter 100 150 91 300 Sum driftsinntekter 5 284 150 5 075 900 Gebyrer for vann, avløp og renovasjon 812 400 768 900 Forsikring 486 000 441 800 Energi i fellesarealer 318 700 352 100 Vaktmester (lønn og sosiale kostnader) 642 500 618 200 Forretningsførsel og revisjon 214 300 206 900 Vedlikehold og reparasjoner 1 236 800 874 500 Styrehonorar 180 000 165 000 Andre driftskostnader 197 600 183 400 Sum driftskostnader 4 088 300 3 610 800 Driftsresultat 1 195 850 1 465 100 Renteinntekter 48 300 39 600 Rentekostnader 498 200 512 700 Årsresultat 745 950 992 000 Felleskostnadene ble økt med 4 % fra 1. januar 2025. Gjennomsnittlig felleskostnad var 6 000 kr per leilighet per måned i 2025. Andre driftsinntekter består av inntekter fra vaskeriet, utleie av fellesrommet og salg av ekstra nøkkelbrikker. Borettslaget har ett nedbetalingslån med flytende rente. Lånet ble nedbetalt med 620 000 kr i 2025, og fellesgjelden gikk ned fra 12 260 000 kr ved inngangen av året til 11 640 000 kr per 31. desember 2025. Det tilsvarer i gjennomsnitt om lag 161 700 kr per leilighet. Renten var 4,40 % ved utgangen av 2024 og 4,15 % ved utgangen av 2025. Lånet skal etter dagens plan være nedbetalt i 2044. Disposisjonsfondet var 3 412 500 kr per 31. desember 2025. Styret mener at likviditeten er god nok til å betale resten av forprosjektet i 2026 uten nye lån. Revisor har ikke hatt merknader til regnskapet. 6. Rehabilitering av balkonger og fasader Tilstandsrapporten fra desember 2025 viser at 14 balkonger har betongskader i den alvorligste tilstandsgraden. Alle disse balkongene ligger i blokk A og blokk B. Rapporten viser også at fasadene på alle fire blokkene har sprekker og dårlig isolasjon, fordi de aldri har vært etterisolert siden byggeåret. Styret har derfor sperret 3 balkonger i blokk A for bruk inntil videre, slik at ingen skal skades av løse betongbiter.
|
||||
|
||||
Styret legger fram et forslag for generalforsamlingen 21. april 2026 om å rehabilitere balkonger og fasader på alle fire blokkene. Hovedpunktene i forslaget er: Kostnadsoverslaget er 41,0 mill. kr inkludert merverdiavgift. Arbeidet skal finansieres med et nytt lån på inntil 39,5 mill. kr med 30 års nedbetalingstid og 1,5 mill. kr fra disposisjonsfondet. Blokk A og blokk B rehabiliteres fra april til oktober 2027, og blokk C og blokk D fra april til oktober 2028. Felleskostnadene må etter styrets beregning økes med om lag 12 % fra 1. januar 2027. Hver balkong vil være utilgjengelig i om lag 10 uker mens arbeidet pågår. Forprosjektet ledes av en byggekomité med fire medlemmer: Ingrid Solbakk, Torbjørn Leirvik og to beboere som ikke sitter i styret. Byggekomiteen hadde sitt første møte 2. desember 2025. I forprosjektets andre del, som gjennomføres i 2026, skal det lages et anbudsgrunnlag, og minst tre entreprenører skal inviteres til å gi tilbud. Styret vil ikke inngå kontrakt før anbudene er vurdert av byggekomiteen og en uavhengig rådgiver. Beboerne har hatt mulighet til å stille spørsmål om prosjektet på to informasjonsmøter, 14. januar og 25. februar 2026. Til sammen 57 personer deltok på de to møtene. De vanligste spørsmålene gjaldt støy i byggeperioden, om balkongene blir større, og hvor mye felleskostnadene vil øke. Balkongene blir ikke større enn i dag, fordi en utvidelse ville kreve nye fundamenter. Forslaget krever to tredels flertall av de avgitte stemmene, fordi tiltaket går ut over vanlig forvaltning og vedlikehold. Hvis forslaget ikke vedtas, må styret likevel utbedre de 14 skadde balkongene. Det er anslått å koste om lag 6,8 mill. kr og vil ikke gi bedre isolasjon av fasadene. 7. Utsikter for 2026 Styret har vedtatt å øke felleskostnadene med 6 % fra 1. januar 2026. Økningen skyldes resten av forprosjektet, som er budsjettert til 1 051 500 kr, en ventet økning i forsikringspremien på 8 % og høyere gebyrer for vann, avløp og renovasjon. Gjennomsnittlig felleskostnad blir dermed om lag 6 360 kr per leilighet per måned i 2026. Budsjettet for 2026 viser et årsresultat på 173 500 kr. Vårdugnaden i 2026 holdes 25. april og høstdugnaden 17. oktober. Styret oppfordrer alle husstander til å delta, fordi dugnadene holder kostnadene til hagestell nede. Styret takker beboerne, vaktmesteren og hagekomiteen for innsatsen i 2025.
|
||||
|
|
@ -0,0 +1,235 @@
|
|||
# Brantwater Sailing Club - Club Rules, Safety and Duty Handbook, 2026 edition
|
||||
|
||||
# Club Rules, Safety and Duty Handbook - 2026 edition
|
||||
|
||||
This edition was approved by the committee on Monday 9 March 2026 and replaces the 2024 edition. It applies to every member, every guest and every visiting sailor while they are on club grounds or afloat on club business. Where this handbook and a notice posted by the Officer of the Day disagree on the day, the Officer of the Day's notice applies.
|
||||
|
||||
## 1. Membership and fees
|
||||
|
||||
The club year for subscriptions runs from 1 March to the last day of February. Subscriptions fall due on 1 March. A member whose subscription has not been paid by 31 March may not sail club boats, use the safety boat cover or sign in guests until it is paid.
|
||||
|
||||
Category
|
||||
|
||||
Who it covers
|
||||
|
||||
Annual subscription 2026 (£)
|
||||
|
||||
Duties per season
|
||||
|
||||
Full adult
|
||||
|
||||
One person aged 26 or over
|
||||
|
||||
285
|
||||
|
||||
3
|
||||
|
||||
Family
|
||||
|
||||
Two adults and their children under 18 at one address
|
||||
|
||||
430
|
||||
|
||||
4, shared by the adults
|
||||
|
||||
Junior
|
||||
|
||||
One person under 18
|
||||
|
||||
95
|
||||
|
||||
0
|
||||
|
||||
Young adult
|
||||
|
||||
One person aged 18 to 25
|
||||
|
||||
140
|
||||
|
||||
2
|
||||
|
||||
Social
|
||||
|
||||
One person who does not sail from the club
|
||||
|
||||
75
|
||||
|
||||
0
|
||||
|
||||
A joining fee of £60 is charged once per membership in the year of joining, except for junior and young adult memberships.
|
||||
|
||||
A member who joins on or after 1 September pays half the annual subscription for the rest of that club year.
|
||||
|
||||
Members aged 75 or over on 1 March are exempt from duties but may still volunteer for them.
|
||||
|
||||
A junior member must have a parent or guardian who is a full, family or social member, unless the junior is 16 or 17 and the committee has agreed otherwise in writing.
|
||||
|
||||
## 2. Clubhouse and bar
|
||||
|
||||
The clubhouse has a main room with the bar, the Lookout Room upstairs, a galley and two changing rooms. The main room holds up to 80 people and the Lookout Room up to 24. The Lookout Room is used as the race office and for training; food and drink other than water are not allowed in it, because the training charts and radios are kept there.
|
||||
|
||||
Bar opening hours
|
||||
|
||||
Summer (1 April to 31 October)
|
||||
|
||||
Winter (1 November to 31 March)
|
||||
|
||||
Wednesday
|
||||
|
||||
18:00 to 22:30
|
||||
|
||||
Closed
|
||||
|
||||
Saturday
|
||||
|
||||
12:00 to 18:00
|
||||
|
||||
Closed
|
||||
|
||||
Sunday
|
||||
|
||||
12:00 to 17:00
|
||||
|
||||
12:00 to 16:00
|
||||
|
||||
Other days
|
||||
|
||||
Closed, except for booked events
|
||||
|
||||
Closed, except for booked events
|
||||
|
||||
The bar serves only members aged 18 or over and their signed-in guests.
|
||||
|
||||
A member may sign in at most 3 guests on any one day, and the same guest may be signed in at most 6 times in a season.
|
||||
|
||||
Members under 18 may be in the main room until 20:00; after 20:00 they may stay only when a parent or guardian is with them.
|
||||
|
||||
Dogs are not allowed inside the clubhouse, except assistance dogs. Elsewhere on club grounds dogs must be kept on a lead.
|
||||
|
||||
Smoking and vaping are not allowed anywhere on club grounds except the marked area behind the north fence.
|
||||
|
||||
The changing rooms are open from 07:30 to 21:30. The door code changes on 1 April and 1 October and is sent to paid-up members only.
|
||||
|
||||
## 3. Boat shed, dinghy park and moorings
|
||||
|
||||
### Boat shed
|
||||
|
||||
The boat shed holds the club dinghies, the safety boats and the workboat. Keys are held only by instructors, the boatman, committee members and the safety boat crew on duty. Fuel must not be kept in the boat shed or the clubhouse; it is stored only in the fuel locker outside the boat shed, which holds no more than 100 litres of petrol.
|
||||
|
||||
### Dinghy park and racks
|
||||
|
||||
The dinghy park has 64 spaces and costs £95 for the storage year from 1 April to 31 March. The racks have 24 spaces for kayaks, paddleboards and tenders and cost £45 for the same period. Storage fees are due by 1 April.
|
||||
|
||||
Every boat and launching trolley must carry a label with the owner's name and membership number.
|
||||
|
||||
If a storage fee is still unpaid on 30 June, the committee may remove the boat after giving 28 days' written notice.
|
||||
|
||||
A boat that has been removed and not claimed within 6 months may be sold, and the proceeds after costs are kept for the owner for one further year.
|
||||
|
||||
Launching trolleys must not be taken beyond the yellow marker on the slipway, because the lower section is damaged.
|
||||
|
||||
### Swinging moorings
|
||||
|
||||
The club has 38 swinging moorings in two trots, north and south. The mooring season runs from 1 April to 31 October, and boats must be off the moorings by 15 November. The fee is £210 for a boat up to 6.5 metres long and £265 for a boat over 6.5 metres. No boat over 8 metres long may use a club mooring.
|
||||
|
||||
Applications for a mooring must reach the Moorings Officer by 31 January. The committee allocates moorings by 28 February.
|
||||
|
||||
Existing holders keep their mooring only if they pay the fee by 1 April.
|
||||
|
||||
Every moored boat must have third-party liability insurance of at least £3 million.
|
||||
|
||||
The boatman inspects the chain and sinker of every mooring every two years. A holder must not change or add to the mooring tackle.
|
||||
|
||||
Six winter hardstanding spaces are available from 1 November to 31 March for £150.
|
||||
|
||||
## 4. Using club boats
|
||||
|
||||
Club dinghies may be used free of charge by members who hold club Level 2 or above. Members with Level 1 may use them only on a supervised session run by an instructor. Bookings are made on the board in the boat shed.
|
||||
|
||||
One booking may last up to 3 hours. On Saturdays and Sundays between 10:00 and 16:00, a booking may last only 2 hours.
|
||||
|
||||
Club boats may not be booked for racing on Wednesday evenings unless the Rear Commodore (Sailing) agrees.
|
||||
|
||||
Juniors may sail club boats without an instructor present only from the age of 14 and only if they hold Level 3.
|
||||
|
||||
The Osprey 16 trainers are kept for courses and may be booked by members only outside course days.
|
||||
|
||||
Any damage must be written in the defects book in the boat shed before the user leaves the club.
|
||||
|
||||
## 5. Safety on the water
|
||||
|
||||
Everyone afloat must wear a buoyancy aid, whether in a club boat, a private boat or a safety boat. Safety boat crews wear lifejackets and the helm must attach the kill cord whenever the engine is running. Juniors must also wear a buoyancy aid on the pontoons at all times.
|
||||
|
||||
Situation
|
||||
|
||||
Limit
|
||||
|
||||
Junior training
|
||||
|
||||
No sailing when the mean wind is over 18 knots or gusts are over 25 knots
|
||||
|
||||
Club boats, adult use
|
||||
|
||||
No sailing when the mean wind is over 22 knots
|
||||
|
||||
Safety cover for racing
|
||||
|
||||
1 safety boat for up to 12 boats racing; 2 safety boats for 13 or more
|
||||
|
||||
Safety cover for junior training
|
||||
|
||||
1 safety boat for every 6 junior boats on the water; no more than 12 junior boats on the water at once
|
||||
|
||||
Water temperature below 10 °C
|
||||
|
||||
Wetsuit or drysuit required for juniors; strongly advised for adults
|
||||
|
||||
Members may sail their own boats outside organised sessions only if another adult is ashore at the club and they have written their name on the afloat board.
|
||||
|
||||
Nobody may sail after sunset.
|
||||
|
||||
From 1 November to 31 March, single-handed sailing is allowed only when a safety boat is staffed.
|
||||
|
||||
The swimming zone off Brantholme beach is out of bounds to all boats.
|
||||
|
||||
Junior boats must stay south of the yellow buoy line during training.
|
||||
|
||||
When the Officer of the Day flies the red flag, no boat may launch.
|
||||
|
||||
## 6. The duty system
|
||||
|
||||
The club has no paid race or safety staff, so every activity on the water depends on members doing their duties. There are four kinds of duty: Officer of the Day, safety boat helm, safety boat crew and shore duty (sign-in, galley and teas). The rota for spring and summer is published by 1 March, and the rota for autumn by 1 August.
|
||||
|
||||
A safety boat helm must be 18 or over and hold the Club Powerboat Certificate.
|
||||
|
||||
Safety boat crew must be 16 or over.
|
||||
|
||||
An Officer of the Day must have been a member for at least one full season.
|
||||
|
||||
If you cannot do a duty, arrange a swap with another member and tell the Duty Coordinator at least 48 hours before the duty.
|
||||
|
||||
A duty that is missed without a swap is charged at £35, which is added to the member's next invoice.
|
||||
|
||||
A duty counts only if the member signs the duty sheet in the race office on the day.
|
||||
|
||||
The Duty Coordinator is Neil Harrowby. The Club Powerboat Certificate course is run twice a year by the Safety Officer and is free to members who agree to do at least two safety boat duties in the following season.
|
||||
|
||||
## 7. Junior members
|
||||
|
||||
Junior training runs on Saturday mornings from 09:30 to 12:30 between May and September. The minimum age is 8, and every junior must be able to swim 50 metres in clothes while wearing a buoyancy aid. The season fee for junior training is £120 for junior members; children who are not members cannot join the programme.
|
||||
|
||||
A parent or guardian must sign the junior in and out at the race office for every session.
|
||||
|
||||
Juniors under 14 must be collected from the clubhouse by a parent or guardian, and may not leave on their own.
|
||||
|
||||
Instructors may cancel a session at short notice because of wind, water temperature or lack of safety cover; cancellations are posted by 08:00.
|
||||
|
||||
Dogs are not allowed in the boat park or on the pontoons on junior training days.
|
||||
|
||||
## 8. Incidents, complaints and changes
|
||||
|
||||
Every incident, near miss or injury on the water or ashore must be reported to the Safety Officer, Fiona Brackley, within 24 hours, using the incident book in the race office. In an emergency, call the emergency services first and then the Officer of the Day.
|
||||
|
||||
Complaints about another member's behaviour go to the Secretary in writing. The committee will reply within 28 days. A member who breaks these rules more than once in a season may have their right to use club boats suspended for up to 3 months, and serious breaches may lead to the end of membership after a hearing.
|
||||
|
||||
The committee may change this handbook during the season. Changes take effect 14 days after they are posted on the clubhouse noticeboard and sent to members, except safety changes, which take effect at once.
|
||||
|
|
@ -0,0 +1,11 @@
|
|||
Brantwater Sailing Club - Commodore's Annual Report 2025-26 Brantwater Sailing Club: Commodore's Annual Report 2025-26 This report covers the club year from 1 October 2025 to 30 September 2026. It will be presented to the Annual General Meeting on Saturday 21 November 2026 at 19:00 in the main room of the clubhouse. The accounts summary in section 6 has been prepared by the Treasurer, Graham Oakeshott, and checked by the club's two honorary examiners, who signed them off on 27 October 2026. 1. Introduction from the Commodore This was my second year as Commodore, and it was a busy one. We grew to 312 members, the highest figure since the club was founded in 1971. We completed 22 races across our three main series, hosted the Brantwater Open Regatta in July and trained more juniors than in any previous season. We also met our first serious problem with the slipway, which cracked after the high water of February 2026. Most of this report is good news, but the slipway repair will shape the coming winter, and I ask every member to read section 5 before the AGM. None of this happens without volunteers. In 2026 members completed 533 duties on the water and ashore. I want to thank in particular our Rear Commodore (Sailing), Priya Castellan, our Principal Instructor, Tomasz Wilde, and our Safety Officer, Fiona Brackley, who between them were at the club on more than 140 days of the season. Helen Marchbank, Commodore 2. Membership On 30 September 2026 the club had 312 members, compared with 296 a year earlier. During the year 47 new members joined and 31 left, most of them because they had moved away from the area. Members are counted as individuals, so a family membership counts every person it covers. The 96 people in the family category belong to 31 family memberships, up from 30 last year. Category Members at 30 Sept 2025 Members at 30 Sept 2026 Change Full adult 141 148 +7 Family (individuals) 92 96 +4 Junior (under 18) 25 27 +2 Young adult (18 to 25) 17 19 +2 Social (non-sailing) 21 22 +1 Total 296 312 +16
|
||||
|
||||
The full adult subscription rose from £270 in 2025 to £285 in 2026. Despite the increase, only four full adult members gave cost as their reason for leaving. The joining fee stayed at £60 per membership and was again waived for junior and young adult memberships, because the committee wants to make it as easy as possible for younger sailors to join. We now have a waiting list for moorings for the first time since 2019. Nine members are on it, and the committee expects that most of them will not be offered a mooring before the 2028 season. 3. Sailing, racing and training Racing Racing at Brantwater uses the club's own yardstick handicap, so that all classes can race together in one fleet. Each series allows one discard when 4 to 7 races have been completed and two discards when 8 or more have been completed. The season results were as follows. Series Dates (2026) Races planned Races completed Average boats per race Spring Series Sundays 12 April to 31 May 8 7 11 Summer Evening Series Wednesdays 3 June to 5 August 10 9 14 Autumn Series Sundays 6 September to 11 October 6 6 9 Brantwater Open Regatta 18 to 19 July 4 4 37 The Spring Series race on 26 April was cancelled because no qualified safety boat helm was available, which is the first time in six years that a race has been lost for that reason. The Summer Evening Series race on 15 July was abandoned after the first lap because of a thunderstorm. The Summer Evening Series was won by Owen Fairclough and Lily Fairclough in their Heron 14, with 9 points net after two discards. The Spring Series was won by Duncan Ferrier and the Autumn Series by Rosa Lindqvist, both sailing single- handers. The Open Regatta attracted 37 boats, 15 of them from six visiting clubs, which is 8 more boats than in 2025. Visiting sailors used the dinghy park free of charge for the weekend, and the bar took £4,120 over the two days. Junior training The junior programme ran on Saturday mornings from 09:30 to 12:30 between 2 May and 19 September. Of the 21 Saturdays in that period, 20 were used for training; there was no session on 18 July because of the regatta. 41 juniors aged 8 to 17 were enrolled, against 34 in 2025. The season fee for junior members stayed at £120.
|
||||
|
||||
Of the 41 juniors, 23 passed at least one club level during the season: 12 passed Level 1, 8 passed Level 2 and 3 passed Level 3. Training was led by 6 senior instructors and 9 assistant instructors, up from 5 and 7 in 2025. Two members qualified as senior instructors at the club's expense, which is the main reason training costs rose from £3,720 to £4,860. Adult courses Eighteen adults completed the two-weekend Start Sailing course and thirteen completed the four-evening Improver course. Start Sailing courses use the two Osprey 16 trainers, because their size allows an instructor and two learners to sail together. Eleven of the eighteen Start Sailing graduates have since joined the club as full adult members. 4. Safety and duties Our safety boats, Gannet and Puffin, were launched on 74 days and ran for a combined 212 engine hours. The safety boat crews attended 19 capsizes that needed help, towed 3 boats back to the club and brought one sailor ashore with early signs of cold. There were no injuries that needed hospital treatment. Two incidents were reported to the Safety Officer within the required 24 hours; one further incident on 9 August was reported after four days, and the committee has reminded all officers of the day of the deadline. The duty system asked for 564 duties during the season. Members completed 533 of them and 31 were missed without a swap, each of which was charged at £35 under the current rules. Beyond the cancelled race on 26 April, on four other days a start was delayed by more than 30 minutes while a replacement safety boat helm was found. Because a missed safety boat duty can stop a race or a training session for everyone, the committee has decided to raise the missed duty charge from £35 to £50 from 1 April 2027. The number of duties each member owes does not change. Members who cannot do a duty will still be able to avoid the charge by arranging a swap and telling the Duty Coordinator, Neil Harrowby, at least 48 hours before the duty. 5. Facilities, fleet and the slipway repair Club fleet The club owns 15 sailing dinghies (eight Tern 12 single-handers, five Heron 14 double- handers and two Osprey 16 trainers) and three powerboats (the safety boats Gannet and Puffin, and the workboat Mallard, which is used for mooring work and is not a primary safety boat). Two new Tern 12s were bought in May 2026 for £3,900 each, paid from the Boat Replacement Fund. Heron 14 number H2 has been out of use since 22 August because of a leaking centreboard case; it will be repaired over the winter. Moorings and storage 34 of the club's 38 swinging moorings were let in 2026, against 36 in 2025. Two south trot moorings are always kept for visiting boats, and two north trot moorings, N09 and N17, have been out of use since their sinkers failed inspection in March 2026; they will be replaced over the winter, and their former holders will be offered them first. 57 of the 64
|
||||
|
||||
dinghy park spaces and 20 of the 24 rack spaces were taken. All six winter hardstanding spaces have been booked for the period from 1 November 2026 to 31 March 2027. Clubhouse The clubhouse roof over the changing rooms was patched in June at a cost of £3,260, which is included in clubhouse repairs and cleaning. Utility costs fell from £8,050 to £7,420 because the new heating controls in the main room and the Lookout Room now switch off automatically at 22:45. The slipway repair The slipway was built in 1984 and is 34 metres long and 4.5 metres wide. After the high water in February 2026, the lower 9 metres cracked and settled by up to 70 millimetres, so that the wheels of launching trolleys now catch on the step. Since March 2026 members have been asked not to take trolleys beyond the yellow marker at the end of the sound concrete. The committee obtained three quotes and accepted the lowest, £34,900, for breaking out and recasting the lower 12 metres. With a 10% contingency of £3,490, the total budget is £38,390. The work will take place from Monday 4 January to Friday 26 February 2027, because the lake level is normally at its lowest in those weeks, so that the contractor can work on the lower section in the dry. Slipway funding source Amount (£) Status Development Fund 22,000 Held at 30 September 2026 Grant from the Thistleby Charitable Trust 9,000 Confirmed on 2 October 2026 General reserve 7,390 Subject to AGM approval Total 38,390 The slipway will be closed to everyone from 2 January to 28 February 2027, which is two days longer at each end than the contractor's works, so that the site can be set up and cleared safely. During that time the safety boats will launch from the temporary pontoon at the north beach, and members will not be able to launch their own boats at all. As a result, the Frostbite Series will run only from 8 November to 20 December 2026, with seven races instead of the usual sixteen. 6. Accounts summary The club made a surplus of £31,830 in the year to 30 September 2026, compared with £24,015 in the year before. The main reasons were higher subscription income, from both the larger membership and the new fees, and a better bar result. Bar sales of £38,900 less bar stock of £16,800 gave a gross bar margin of £22,100 before wages. Income 2024-25 (£) 2025-26 (£) Subscriptions 56,385 62,385 Joining fees 1,980 2,640 Moorings and boat storage 13,920 14,355 Bar sales 35,400 38,900
|
||||
|
||||
Training courses 8,120 9,810 Race entry fees 2,300 2,150 Social events 3,950 4,600 Donations 750 1,200 Total income 122,805 136,040 Expenditure 2024-25 (£) 2025-26 (£) Bar stock 15,300 16,800 Wages (bar steward and boatman) 20,600 21,400 Ground rent to the Brantwater Estate 9,500 9,500 Insurance 6,300 6,850 Utilities 8,050 7,420 Club boat and safety boat maintenance 7,640 8,930 Safety boat fuel 2,140 2,310 Training costs 3,720 4,860 Clubhouse repairs and cleaning 11,880 12,340 Affiliation and licences 1,890 1,950 Social events 2,960 3,180 Administration 2,610 2,470 Depreciation 6,200 6,200 Total expenditure 98,790 104,210 Surplus 24,015 31,830 Insurance rose by 8.7%, from £6,300 to £6,850, after the insurer added cover for the new Tern 12s and raised its base rate. Race entry fees fell slightly, by £150, because one Spring Series race was cancelled and one Summer Evening race was abandoned. The ground rent is fixed at £9,500 a year until the next rent review in April 2029 under the lease, which runs until 2041. The committee allocated the surplus as follows: £15,000 to the Development Fund for the slipway, £6,000 to the Boat Replacement Fund and the remaining £10,830 to the general reserve. Reserve At 30 Sept 2025 (£) Added (£) Spent (£) At 30 Sept 2026 (£) General reserve 37,770 10,830 0 48,600 Development Fund 7,000 15,000 0 22,000 Boat Replacement Fund 13,300 6,000 7,800 11,500 Total reserves 58,070 31,830 7,800 82,100 The committee's policy is that the general reserve must not fall below £35,000, which is roughly four months of running costs. If the AGM approves the use of £7,390 for the slipway, the general reserve will fall to £41,210 and will stay above that floor.
|
||||
|
||||
\7. Looking ahead and proposals to the AGM The committee will ask the AGM to approve two resolutions. The first is the use of £7,390 from the general reserve for the slipway repair. The second is a set of new subscription rates, which would apply from 1 March 2027, the date on which subscriptions fall due. The proposed increases are needed because insurance and wages are rising faster than membership numbers. Membership category 2026 rate (£) Proposed 2027 rate (£) Full adult 285 298 Family 430 450 Junior (under 18) 95 95 Young adult (18 to 25) 140 145 Social (non-sailing) 75 75 Junior and social rates would not change, because the committee wants to keep junior sailing affordable and because social members do not use the boats or the safety cover. The joining fee would stay at £60. Mooring and storage fees are set by the committee rather than the AGM and will stay at their 2026 levels for 2027. Three committee posts are up for election this year: Vice Commodore, Secretary and Moorings Officer. Duncan Ferrier has agreed to stand again as Vice Commodore. Ruth Penhallow is stepping down as Secretary after six years, and Alan Treadgold is stepping down as Moorings Officer after four. Nominations must reach the Secretary by 19:00 on Saturday 14 November 2026, one week before the meeting. Finally, the 2027 season will begin with the fitting-out work party on Saturday 27 March 2027. The boatman will lay the mooring buoys from the workboat on Saturday 3 April, so moored boats can go on their moorings from that afternoon, and the season opening day is Saturday 10 April. I look forward to seeing you on the water.
|
||||
|
|
@ -0,0 +1,183 @@
|
|||
## Copperfold Makerspace Annual General Meeting 2026
|
||||
|
||||
## Annual general meeting, 28 November 2026
|
||||
|
||||
The meeting starts at 14:00 in Unit 4, Copperfold Yard.
|
||||
|
||||
The Saturday open session ends at 13:00 today so that the workshop can be cleared for seating.
|
||||
|
||||
Only Full and Concession members who joined at least three months ago may vote.
|
||||
|
||||
Proxy votes had to reach the membership secretary by 14:00 on Thursday 26 November.
|
||||
|
||||
## Agenda
|
||||
|
||||
Annual report and accounts for 2025–26.
|
||||
|
||||
Budget for 2026–27.
|
||||
|
||||
Vote on the move to Unit 11, Larchmere Works.
|
||||
|
||||
Vote on membership fees from 1 August 2027.
|
||||
|
||||
Election of the chair and two ordinary committee members.
|
||||
|
||||
## Membership in 2025–26
|
||||
|
||||
We had 214 members on 31 March 2026, up from 187 a year earlier.
|
||||
|
||||
74 new members joined and 47 left during the year.
|
||||
|
||||
The split was 121 Full, 49 Concession, 35 Associate and 9 Youth members.
|
||||
|
||||
The new Wednesday session from 13:00 to 16:00 started on 7 January 2026 and averaged 11 attendees.
|
||||
|
||||
## Fees since 1 September 2026
|
||||
|
||||
Full membership rose from £26 to £30 a month.
|
||||
|
||||
Concession membership rose from £16 to £18 a month.
|
||||
|
||||
Associate (£12) and Youth (£8) fees and the £15 joining fee did not change.
|
||||
|
||||
The rise was approved on 14 June 2026 because electricity costs rose 19 per cent and insurance 15 per cent.
|
||||
|
||||
## Inductions
|
||||
|
||||
Members completed 360 inductions in 2025–26, up from 297 in 2024–25.
|
||||
|
||||
Woodshop basic was the most common induction, with 96 completed.
|
||||
|
||||
On 1 October 2026, 46 members were waiting for woodshop basic, with an average wait of 9 weeks.
|
||||
|
||||
Only 6 of our 21 inductors are approved for woodshop machines.
|
||||
|
||||
## Safety record
|
||||
|
||||
Members recorded 17 incidents in 2025–26, down from 23 the year before.
|
||||
|
||||
Cuts and grazes made up 12 of the 17 incidents.
|
||||
|
||||
One member needed three stitches at a walk-in clinic after a chisel cut in March 2026.
|
||||
|
||||
Nine members hold a current first-aid certificate.
|
||||
|
||||
## The laser fire and what changed
|
||||
|
||||
On 12 August 2025, 3 mm plywood caught fire in the laser while the operator was out of the room.
|
||||
|
||||
Another member put the fire out with the CO2 extinguisher, and nobody was hurt.
|
||||
|
||||
The laser was out of use for 4 days and the operator's induction was suspended for 30 days.
|
||||
|
||||
Since handbook version 4 took effect on 1 September 2026, the laser must be paused before anyone leaves the room.
|
||||
|
||||
## Income in 2025–26
|
||||
|
||||
Total income was £69,010, up from £60,240 in 2024–25.
|
||||
|
||||
Membership and joining fees brought in £52,480, or 76 per cent of income.
|
||||
|
||||
Laser time and materials brought in £6,420, and courses and Junior Saturdays £4,950.
|
||||
|
||||
A £3,000 grant from the Wexcombe Community Trust could only be spent on dust extraction.
|
||||
|
||||
## Spending and reserves
|
||||
|
||||
Total expenditure was £64,615, of which rent for Unit 4 was £27,600.
|
||||
|
||||
We spent £10,630 on new equipment and £3,585 on repairs and maintenance.
|
||||
|
||||
The year ended with a surplus of £4,395.
|
||||
|
||||
Reserves stood at £22,940 on 31 March 2026, including a £15,000 move fund ring-fenced on 9 January 2026.
|
||||
|
||||
## Budget for 2026–27
|
||||
|
||||
Budgeted income is £87,100, including a £12,000 move appeal that is only collected if the move is approved.
|
||||
|
||||
Budgeted expenditure is £73,950.
|
||||
|
||||
The budget includes a £7,800 deposit for Unit 11 and £4,500 for the electrical survey and design.
|
||||
|
||||
Rent at Unit 4 stays at £2,300 a month until the lease ends.
|
||||
|
||||
## Why we need to move
|
||||
|
||||
Unit 4 has only 186 square metres and no room for more benches.
|
||||
|
||||
The single-phase supply means the table saw and planer-thicknesser cannot run at the same time.
|
||||
|
||||
Because of the flats above the yard, fob access ends at 23:00 and noisy machines stop at 22:00.
|
||||
|
||||
The laser was booked for 87 per cent of bookable hours in 2025–26.
|
||||
|
||||
## Unit 11, Larchmere Works
|
||||
|
||||
Unit 11 is a ground-floor unit of 410 square metres, about 900 metres from Copperfold Yard.
|
||||
|
||||
It has a three-phase supply, an 18 square metre dust-extraction room and a 4.2-metre ceiling.
|
||||
|
||||
There are no homes within 150 metres of the unit.
|
||||
|
||||
The rent is £3,900 a month on a ten-year lease with a break clause after five years.
|
||||
|
||||
## Fit-out cost and funding
|
||||
|
||||
The fit-out is estimated at £38,500 in total.
|
||||
|
||||
The largest item is the three-phase electrical work at £14,200, followed by dust-extraction ducting at £9,800.
|
||||
|
||||
It will be funded by the £15,000 move fund, a £12,000 members' appeal and £11,500 in grant applications.
|
||||
|
||||
The move appeal opens on 1 December 2026, but only if the move is approved today.
|
||||
|
||||
## Timetable if the move is approved
|
||||
|
||||
The lease for Unit 11 is to be signed by 15 January 2027 and the deposit paid in February 2027.
|
||||
|
||||
We get the keys on 1 May 2027 and fit out the unit in May and June.
|
||||
|
||||
Unit 4 closes to members after 2 July 2027, and machines move over two weekends in July.
|
||||
|
||||
Unit 11 opens to members on Saturday 24 July 2027.
|
||||
|
||||
## Handing back Unit 4
|
||||
|
||||
The landlord has agreed that we may hand back Unit 4 on 31 July 2027 instead of 30 September 2027.
|
||||
|
||||
This requires eight weeks' written notice, so notice must be given by 5 June 2027.
|
||||
|
||||
Handing back early saves two months' rent, or £4,600.
|
||||
|
||||
We will still pay rent on both units in May, June and July 2027.
|
||||
|
||||
## Proposed fees from 1 August 2027
|
||||
|
||||
Full membership would rise from £30 to £34 a month.
|
||||
|
||||
Concession would rise from £18 to £20, and Associate from £12 to £14.
|
||||
|
||||
The Youth fee would stay at £8 a month.
|
||||
|
||||
These fees only take effect if the move is approved.
|
||||
|
||||
## How today's votes work
|
||||
|
||||
The move needs a two-thirds majority of the votes cast, including proxy votes.
|
||||
|
||||
The fee proposal needs a simple majority and is voted on separately after the move.
|
||||
|
||||
Associate and Youth members may speak but may not vote.
|
||||
|
||||
If the move is rejected, the committee will ask the landlord to extend the Unit 4 lease beyond 30 September 2027.
|
||||
|
||||
## Committee elections
|
||||
|
||||
Imogen Halvorsen-Pike is stepping down as chair after four years.
|
||||
|
||||
Sunniva Okafor, who led the search for new premises, is the only candidate for chair.
|
||||
|
||||
Callum Ferrand, Hettie Galbraith and Dov Pennick-Ash are standing for the two ordinary seats.
|
||||
|
||||
Tobias Wrenfield continues as treasurer until 2027.
|
||||
|
|
@ -0,0 +1,11 @@
|
|||
Copperfold Makerspace Annual Report 2025–26 Copperfold Makerspace Annual Report 2025–26 This report covers the financial year from 1 April 2025 to 31 March 2026. It was prepared by the committee for the annual general meeting on Saturday 28 November 2026. 1. From the chair This is my fourth and final annual report as chair of Copperfold Makerspace. When I was first elected in November 2022, we had 132 members, one cramped woodshop and a laser cutter that was booked for barely a third of its available hours. On 31 March 2026 we had 214 members, 21 approved inductors and a laser that members queue for. None of this would have been possible without our 38 active volunteers, who logged 2,940 hours between them during the year. Two themes ran through 2025–26. The first was growth. Membership rose by 27 people and the number of completed inductions rose from 297 to 360, yet the waiting list for the woodshop grew faster than our inductors could clear it. The second was a growing recognition that Unit 4 can no longer hold us. Our 186 square metres were laid out for a small joinery business, the electricity supply is single-phase, and the residents of the flats above Copperfold Yard have every right to a quiet night. For these reasons the committee spent much of the second half of the year looking for a new home, and this report ends with the proposal that members will vote on at the general meeting: a move to Unit 11 at Larchmere Works in July 2027. I will not stand for re-election in November. Thank you for trusting me with the role for four years. Imogen Halvorsen-Pike, Chair. 2. Membership and access On 31 March 2026 the makerspace had 214 paying members, compared with 187 on 31 March 2025. We welcomed 74 new members during the year and 47 members left, most of them because they moved away from the area or because their studies ended. The table below shows the split by tier and the monthly fee that applied throughout 2025–26. Tier Members on 31 March 2025 Members on 31 March 2026 Monthly fee in 2025 –26 Full 108 121 £26 Concession 44 49 £16 Associate 29 35 £12 Youth (ages 14–17) 6 9 £8 Total 187 214 –
|
||||
|
||||
Concession membership is open to students in full-time education, to members aged 65 or over and to members on a low income, who self-declare. Concession members have exactly the same access as Full members. Associate members may use the space only during staffed open sessions, and they may not use the laser cutter or the resin printer. Youth members may only be in the building when their parent or guardian, who must be a Full or Concession member, is also present. Until 31 May 2025, Full and Concession members could let themselves in with their key fob at any hour of the day or night. From 1 June 2025 fob access has been limited to 07:00– 23:00 every day, because three noise complaints were received from residents of the flats above Copperfold Yard between February and April 2025. No further complaints have been received since the change took effect. Staffed open sessions ran on Tuesdays and Thursdays from 18:00 to 21:30 and on Saturdays from 10:00 to 16:00. From Wednesday 7 January 2026 we added a Wednesday afternoon session from 13:00 to 16:00, mainly for members who work shifts or are retired. The Wednesday session averaged 11 attendees in its first twelve weeks, against an average of 26 on Tuesday and Thursday evenings and 34 on Saturdays. Fees did not change during 2025–26. At the general meeting on 14 June 2026, members approved a rise in the Full fee from £26 to £30 a month and in the Concession fee from £16 to £18 a month, with effect from 1 September 2026. The rise was needed because our electricity costs rose by 19 per cent between 2024–25 and 2025–26 and our insurance premium rose by 15 per cent at the September 2025 renewal. The Associate fee of £12 and the Youth fee of £8 were not changed, and the one-off joining fee of £15 also stayed the same. 3. Workshops, equipment and inductions Members completed 360 inductions in 2025–26, against 297 the year before. Every induction is run by one of our volunteer inductors. Inductions are free for Full, Concession and Youth members, while Associate members pay £10 per induction. The table shows the length of each induction and how many were completed in each of the last two years. Induction Length Completed 2024–25 Completed 2025–26 Woodshop basic 3 hours 81 96 Table saw 1.5 hours 36 41 Planer-thicknesser and router table 1 hour 17 22 Laser cutter 2 hours plus one supervised job 52 58 FDM 3D printing 1 hour 61 73 Resin printing 1.5 hours 9 14 Electronics bench 1 hour 30 37 Overlocker and embroidery machine 45 minutes 11 19
|
||||
|
||||
Total 297 360 Demand for the woodshop continues to outstrip what we can offer. On 1 October 2026, 46 members were on the waiting list for the woodshop basic induction, and the average wait was 9 weeks. Only 6 of our 21 inductors are approved to induct on woodshop machines, because a woodshop inductor must have been inducted on those machines for at least two years and must be signed off by the woodshop lead, Ines Calloway. We spent £10,630 on new equipment during the year: A 310 mm planer-thicknesser, bought in October 2025 for £2,950. It replaced the 260 mm planer that was built in 2011 and donated to us in 2017, which was retired after its motor failed in September 2025. Two FDM 3D printers with 300 × 300 mm beds, bought in November 2025 for £1,240 in total. Our four older printers have 250 × 250 mm beds. A wash-and-cure station for the resin printer, bought in June 2025 for £310. A two-bag dust extractor with new ducting, installed in August 2025 for £4,100. Two bench power supplies for the electronics bench, bought in January 2026 for £380. An embroidery machine with a maximum hoop size of 200 × 300 mm, bought in March 2026 for £1,650. Repairs and maintenance cost £3,585. The largest single item was a new laser tube, fitted in February 2026 at a cost of £1,180 after the old tube had logged about 2,300 firing hours. Blades, belts, filters and small tools made up the remaining £2,405. The laser cutter was booked for 87 per cent of the hours in which it could be booked, compared with 71 per cent in 2024–25. Firing time is charged at £4 an hour, rounded up to the nearest 15 minutes. Laser firing time brought in £3,860 of the £6,420 we received for laser time and materials; the remaining £2,560 came from filament, resin and timber offcuts sold to members at cost. Junior Saturdays for children aged 8 to 13 ran on the first and third Saturday of each month from 10:00 to 12:00. We held 22 sessions with 196 child attendances, an average of 8.9 children per session against a cap of 12. We also ran 31 paid courses, ranging from building a simple stool to an introduction to microcontrollers. Courses are open to non-members, and 248 people took part. Courses and Junior Saturdays together brought in £4,950. 4. Safety Members recorded 17 incidents in the incident book in 2025–26, down from 23 in 2024–25. Nobody was admitted to hospital. All but one incident were dealt with by first aid on site; the exception was a member who cut a finger on a chisel in March 2026 and needed three stitches at a walk-in clinic. Type of incident 2024–25 2025–26 Cuts and grazes 15 12 Burns from soldering irons 4 2
|
||||
|
||||
and glue guns Laser flare-ups 0 1 Other (strains, dust in the eye) 4 2 Total 23 17 On 12 August 2025 a sheet of 3 mm plywood caught fire inside the laser cutter while the operator was in the kitchen fetching a drink. Another member saw flames through the lid and put the fire out with the CO2 extinguisher kept beside the machine. Nobody was hurt, but the honeycomb bed and one lens had to be replaced, and the laser was out of use for 4 days. Because the operator had left the room while the laser was firing, the operator's laser induction was suspended for 30 days, as the handbook requires for a first breach of a safety rule. As a result of the incident, the committee made two changes. The laser room door now carries a sign stating that the laser must be paused before anyone leaves the room, and this rule was written into version 4 of the member handbook, which took effect on 1 September 2026. The laser induction also now includes a practical exercise with the extinguisher. We kept the two-person rule for the table saw and the planer-thicknesser: these machines may only be used when at least one other adult member is in the building. Nine members hold a current first-aid certificate, and at least one of them was present at 81 per cent of staffed open sessions during the year. 5. Finances The figures below summarise the accounts for 2025–26, which were checked by our independent examiner, Laurence Pemberton-Day, on 2 October 2026. Item Amount 2025–26 (£) Membership fees 51,370 Joining fees 1,110 Laser time and materials 6,420 Courses and Junior Saturdays 4,950 Donations 2,160 Grant from the Wexcombe Community Trust 3,000 Total income 69,010 Rent of Unit 4 27,600 Electricity 8,140 Insurance 3,480 Equipment purchases 10,630 Repairs and maintenance 3,585 Consumables 6,890 Cleaning, bank charges and sundries 4,290 Total expenditure 64,615 Surplus for the year 4,395 Total income rose from £60,240 in 2024–25 to £69,010 in 2025–26, mainly because of the larger membership. Total expenditure rose from £58,910 to £64,615. Electricity rose from
|
||||
|
||||
£6,840 to £8,140 and insurance from £3,020 to £3,480. Membership and joining fees together came to £52,480, which is 76 per cent of total income. The grant of £3,000 from the Wexcombe Community Trust was given for the dust- extraction upgrade only, so it could not be spent on anything else. It covered most of the £4,100 that the new extractor and ducting cost. Our reserves rose from £18,545 on 31 March 2025 to £22,940 on 31 March 2026. On 9 January 2026 the committee ring-fenced £15,000 of the reserves as a move fund, which may only be spent on moving to and fitting out new premises. The remaining general reserve of £7,940 is above the committee's minimum, which is two months' rent at Unit 4, or £4,600. The rent of Unit 4 is £2,300 a month and is fixed until the lease ends on 30 September 2027. 6. Planning the move to Larchmere Works The committee concluded in December 2025 that Unit 4 cannot support the makerspace beyond the end of its lease. The main reasons are: Space: at 186 square metres, Unit 4 has no room for more benches, and the woodshop waiting list stood at 46 members on 1 October 2026. Power: the supply is single-phase, so the table saw and the planer-thicknesser cannot run at the same time without tripping the main breaker. Neighbours: the flats above the yard are the reason fob access ends at 23:00 and noisy machines must stop at 22:00. Demand: the laser was booked for 87 per cent of bookable hours, and there is no room for a second enclosed machine room. After viewing five units between January and May 2026, the committee recommends Unit 11 at Larchmere Works, about 900 metres from Copperfold Yard. Unit 11 is a ground-floor unit of 410 square metres with a three-phase supply, a separate 18 square metre room for dust extraction, a ceiling height of 4.2 metres and a 3-metre-wide roller door that will let us bring in full sheets of plywood. There are no homes within 150 metres. The rent is £3,900 a month on a ten-year lease, with a break clause after five years. The fit-out has been estimated at £38,500: Fit-out item Estimate (£) Three-phase distribution board, cabling and sockets 14,200 Dust-extraction ducting for the whole woodshop 9,800 Enclosed laser and resin room with its own extraction 6,700 Flooring and fire doors 4,300 Removal crew and moving heavy machines 3,500 Total 38,500
|
||||
|
||||
The plan is to pay for the fit-out from three sources: the £15,000 move fund, a members' move appeal with a target of £12,000, and grant applications to two charitable trusts totalling £11,500. The move appeal will open on 1 December 2026, but only if members approve the move. The two months' rent deposit of £7,800 is not part of the fit-out budget and will be paid from the 2026–27 operating budget. If the move is approved, the timetable is: Lease for Unit 11 signed by Friday 15 January 2027, and the deposit paid in February 2027. Keys handed over on Saturday 1 May 2027, followed by fit-out in May and June 2027. Unit 4 closes to members after Friday 2 July 2027, and machines are moved over the weekends of 3–4 July and 10–11 July 2027. Unit 11 opens to members on Saturday 24 July 2027. Unit 4 is handed back on 31 July 2027. The landlord of Unit 4 has agreed that we may hand the unit back on 31 July 2027 instead of 30 September 2027, provided we give eight weeks' written notice, which means notice by 5 June 2027. This saves two months' rent, or £4,600, but we will still pay rent on both units for May, June and July 2027, which costs £6,900 at Unit 4. Because the rent at Unit 11 is £1,600 a month higher than at Unit 4, the committee also proposes that from 1 August 2027 the Full fee rises from £30 to £34, the Concession fee from £18 to £20 and the Associate fee from £12 to £14, while the Youth fee stays at £8. The fee change will only take effect if the move is approved. The move needs a two-thirds majority of the votes cast, while the fee change needs a simple majority, so the two are voted on separately. 7. Volunteers and the year ahead The makerspace has no paid staff. In 2025–26, 38 active volunteers logged 2,940 hours running open sessions, inducting members, repairing machines, cleaning and keeping the accounts. Twenty-one of them are approved inductors, and nine hold a first-aid certificate. The committee has seven members, and three seats are up for election in November 2026: the chair and two ordinary seats. Nominations closed on 14 November 2026 with four candidates for the three seats. Sunniva Okafor, who has led the search for new premises, is the only candidate for chair. Tobias Wrenfield continues as treasurer until 2027. Special thanks go to Priya Okonkwo-Bell, our safety officer, and to the area leads: Ines Calloway in the woodshop, Joss Tannery on the laser, Ruth Abernethy-Quint in 3D printing, Kwame Lindqvist at the electronics bench and Hettie Galbraith in the textile corner.
|
||||
|
|
@ -0,0 +1,94 @@
|
|||
#### Equipment register
|
||||
|
||||
A: ID | B: Item | C: Area | D: Acquired (year) | E: Purchase cost (£) | F: Rated power (W) | G: Induction needed | H: Status | I: Responsible | J: Notes
|
||||
A: WS-01 | B: Table saw, 254 mm blade | C: Woodshop | D: 2018 | E: 1890 | F: 2000 | G: Table saw | H: In service | I: Ines Calloway | J: Two-person rule; never run at the same time as WS-03
|
||||
A: WS-02 | B: Bandsaw, 350 mm throat | C: Woodshop | D: 2018 | E: 920 | F: 750 | G: Woodshop basic | H: In service | I: Ines Calloway | J: Youth members only under direct supervision of a woodshop inductor
|
||||
A: WS-03 | B: Planer-thicknesser, 310 mm | C: Woodshop | D: 2025 | E: 2950 | F: 2200 | G: Planer-thicknesser and router table | H: In service | I: Ines Calloway | J: Bought October 2025 to replace WS-09; two-person rule
|
||||
A: WS-04 | B: Router table | C: Woodshop | D: 2020 | E: 640 | F: 2100 | G: Planer-thicknesser and router table | H: In service | I: Ines Calloway | J: Fence rebuilt June 2026
|
||||
A: WS-05 | B: Pillar drill | C: Woodshop | D: 2017 | E: 380 | F: 550 | G: Woodshop basic | H: In service | I: Ines Calloway | J: Youth members only under direct supervision of a woodshop inductor
|
||||
A: WS-06 | B: Mitre saw, 216 mm | C: Woodshop | D: 2021 | E: 410 | F: 1500 | G: Woodshop basic | H: In service | I: Ines Calloway | J: Not for Youth members
|
||||
A: WS-07 | B: Belt and disc sander | C: Woodshop | D: 2019 | E: 290 | F: 600 | G: Woodshop basic | H: Out of service | I: Ines Calloway | J: Belt tracking fault reported 3 October 2026
|
||||
A: WS-08 | B: Dust extractor, two-bag, with ducting | C: Woodshop | D: 2025 | E: 4100 | F: 2200 | G: Woodshop basic | H: In service | I: Ines Calloway | J: Installed August 2025; £3,000 of the cost covered by the Wexcombe Community Trust grant
|
||||
A: WS-09 | B: Planer-thicknesser, 260 mm | C: Woodshop | D: 2017 | E: 0 | F: 1500 | G: None | H: Retired | I: Ines Calloway | J: Built 2011 and donated 2017; motor failed September 2025; retired October 2025
|
||||
A: LC-01 | B: Laser cutter, 60 W CO2, bed 900 x 600 mm | C: Laser room | D: 2019 | E: 6800 | F: 1200 | G: Laser cutter | H: In service | I: Joss Tannery | J: Tube replaced February 2026 for £1,180; bed and one lens replaced after the fire on 12 August 2025
|
||||
A: LC-02 | B: Laser fume extractor | C: Laser room | D: 2019 | E: 1150 | F: 900 | G: Laser cutter | H: In service | I: Joss Tannery | J: Filters changed every 300 firing hours
|
||||
A: LC-03 | B: CO2 fire extinguisher, 2 kg | C: Laser room | D: 2019 | E: 65 | F: 0 | G: None | H: In service | I: Priya Okonkwo-Bell | J: Inspected 12 September 2026; next inspection due September 2027
|
||||
A: 3D-01 | B: FDM printer, bed 250 x 250 mm | C: 3D printing | D: 2020 | E: 450 | F: 350 | G: FDM 3D printing | H: In service | I: Ruth Abernethy-Quint | J: On the steel fire shelf; overnight prints allowed
|
||||
A: 3D-02 | B: FDM printer, bed 250 x 250 mm | C: 3D printing | D: 2020 | E: 450 | F: 350 | G: FDM 3D printing | H: In service | I: Ruth Abernethy-Quint | J: On the steel fire shelf; overnight prints allowed
|
||||
A: 3D-03 | B: FDM printer, bed 250 x 250 mm | C: 3D printing | D: 2021 | E: 470 | F: 350 | G: FDM 3D printing | H: Out of service | I: Ruth Abernethy-Quint | J: Extruder fault; replacement part ordered 21 October 2026
|
||||
A: 3D-04 | B: FDM printer, bed 250 x 250 mm | C: 3D printing | D: 2022 | E: 490 | F: 350 | G: FDM 3D printing | H: In service | I: Ruth Abernethy-Quint | J: Nozzle replaced July 2026
|
||||
A: 3D-05 | B: FDM printer, bed 300 x 300 mm | C: 3D printing | D: 2025 | E: 620 | F: 400 | G: FDM 3D printing | H: In service | I: Ruth Abernethy-Quint | J: Bought November 2025
|
||||
A: 3D-06 | B: FDM printer, bed 300 x 300 mm | C: 3D printing | D: 2025 | E: 620 | F: 400 | G: FDM 3D printing | H: In service | I: Ruth Abernethy-Quint | J: Bought November 2025
|
||||
A: 3D-07 | B: Resin printer | C: Resin room | D: 2023 | E: 540 | F: 120 | G: Resin printing | H: In service | I: Ruth Abernethy-Quint | J: Must never run when nobody is in the building
|
||||
A: 3D-08 | B: Wash-and-cure station | C: Resin room | D: 2025 | E: 310 | F: 60 | G: Resin printing | H: In service | I: Ruth Abernethy-Quint | J: Bought June 2025
|
||||
|
||||
A: ID | B: Item | C: Area | D: Acquired (year) | E: Purchase cost (£) | F: Rated power (W) | G: Induction needed | H: Status | I: Responsible | J: Notes
|
||||
A: EL-01 | B: Soldering stations (4) | C: Electronics bench | D: 2018 | E: 480 | F: 320 | G: Electronics bench | H: In service | I: Kwame Lindqvist | J: Four stations of 80 W each; lead-free solder only
|
||||
A: EL-02 | B: Bench power supplies, 0-30 V (2) | C: Electronics bench | D: 2026 | E: 380 | F: 300 | G: Electronics bench | H: In service | I: Kwame Lindqvist | J: Bought January 2026
|
||||
A: EL-03 | B: Oscilloscope, 100 MHz, 2-channel | C: Electronics bench | D: 2020 | E: 360 | F: 40 | G: Electronics bench | H: In service | I: Kwame Lindqvist | J: Probes replaced May 2026
|
||||
A: EL-04 | B: Solder fume extractor | C: Electronics bench | D: 2018 | E: 140 | F: 45 | G: Electronics bench | H: In service | I: Kwame Lindqvist | J: Must run while soldering
|
||||
A: TX-01 | B: Sewing machines (3) | C: Textile corner | D: 2017 | E: 600 | F: 210 | G: None for members aged 18 or over | H: In service | I: Hettie Galbraith | J: Three machines of 70 W each; speed limiters fitted for Junior Saturdays
|
||||
A: TX-02 | B: Overlocker, 4-thread | C: Textile corner | D: 2021 | E: 430 | F: 120 | G: Overlocker and embroidery machine | H: In service | I: Hettie Galbraith | J: Youth members may use it after induction
|
||||
A: TX-03 | B: Embroidery machine, max hoop 200 x 300 mm | C: Textile corner | D: 2026 | E: 1650 | F: 90 | G: Overlocker and embroidery machine | H: In service | I: Hettie Galbraith | J: Bought March 2026; bookable for up to 3 hours at a time
|
||||
A: TX-04 | B: Steam iron and board | C: Textile corner | D: 2022 | E: 85 | F: 2400 | G: None | H: In service | I: Hettie Galbraith | J: Switch off and unplug when leaving the corner
|
||||
|
||||
#### Induction log Sep-Oct 2026
|
||||
|
||||
A: Date | B: Member no. | C: Tier | D: Induction | E: Duration (min) | F: Inductor | G: Fee charged (£) | H: Result | I: Lapses if unused (months) | J: Notes
|
||||
A: 2026-09-01 | B: M-0412 | C: Full | D: Woodshop basic | E: 180 | F: Ines Calloway | G: 0 | H: Passed | I: 12
|
||||
A: 2026-09-02 | B: M-0398 | C: Concession | D: FDM 3D printing | E: 60 | F: Ruth Abernethy-Quint | G: 0 | H: Passed | I: 12 | J: Wednesday afternoon session
|
||||
A: 2026-09-03 | B: M-0377 | C: Full | D: Laser cutter | E: 120 | F: Joss Tannery | G: 0 | H: Passed - supervised job pending | I: 6 | J: Supervised job booked for 10 September 2026
|
||||
A: 2026-09-05 | B: M-0421 | C: Associate | D: Electronics bench | E: 60 | F: Kwame Lindqvist | G: 10 | H: Passed | I: 12 | J: Associate induction fee paid
|
||||
A: 2026-09-05 | B: M-0405 | C: Youth | D: Woodshop basic (youth) | E: 120 | F: Theo Vantongeren | G: 0 | H: Passed | I: 12 | J: Bandsaw, pillar drill and hand tools only; parent M-0287 present
|
||||
A: 2026-09-08 | B: M-0356 | C: Full | D: Table saw | E: 90 | F: Ines Calloway | G: 0 | H: Passed | I: 6 | J: Woodshop basic completed March 2026
|
||||
A: 2026-09-10 | B: M-0377 | C: Full | D: Laser supervised job | E: 40 | F: Amara Leclercq-Hood | G: 0 | H: Passed | I: 6 | J: Laser induction now complete
|
||||
A: 2026-09-10 | B: M-0390 | C: Concession | D: Overlocker and embroidery machine | E: 45 | F: Hettie Galbraith | G: 0 | H: Passed | I: 12
|
||||
A: 2026-09-12 | B: M-0412 | C: Full | D: Table saw | E: 90 | F: Theo Vantongeren | G: 0 | H: Not passed - repeat booked | I: 6 | J: Hands too close to the blade on rip cuts; repeat on 26 September 2026
|
||||
A: 2026-09-15 | B: M-0233 | C: Full | D: Laser refresher | E: 30 | F: Joss Tannery | G: 0 | H: Passed | I: 6 | J: Last laser use March 2026, so the induction had lapsed after 6 months
|
||||
A: 2026-09-16 | B: M-0419 | C: Associate | D: Woodshop basic | E: 180 | F: Ines Calloway | G: 10 | H: Passed | I: 12 | J: Associate induction fee paid
|
||||
A: 2026-09-17 | B: M-0388 | C: Full | D: Resin printing | E: 90 | F: Ruth Abernethy-Quint | G: 0 | H: Passed | I: 12 | J: Nitrile gloves issued
|
||||
A: 2026-09-19 | B: M-0401 | C: Concession | D: FDM 3D printing | E: 60 | F: Ruth Abernethy-Quint | G: 0 | H: Passed | I: 12
|
||||
A: 2026-09-22 | B: M-0356 | C: Full | D: Planer-thicknesser and router table | E: 60 | F: Ines Calloway | G: 0 | H: Passed | I: 12 | J: Trained on WS-03
|
||||
A: 2026-09-24 | B: M-0415 | C: Full | D: Electronics bench | E: 60 | F: Mirela Banks | G: 0 | H: Passed | I: 12
|
||||
A: 2026-09-26 | B: M-0412 | C: Full | D: Table saw | E: 90 | F: Ines Calloway | G: 0 | H: Passed | I: 6 | J: Repeat of 12 September 2026
|
||||
A: 2026-09-29 | B: M-0402 | C: Concession | D: Laser cutter | E: 120 | F: Amara Leclercq-Hood | G: 0 | H: Passed - supervised job pending | I: 6 | J: Included the extinguisher exercise
|
||||
A: 2026-09-30 | B: M-0422 | C: Associate | D: FDM 3D printing | E: 60 | F: Ruth Abernethy-Quint | G: 10 | H: Passed | I: 12 | J: Associate induction fee paid
|
||||
A: 2026-10-01 | B: M-0198 | C: Full | D: Woodshop refresher | E: 30 | F: Theo Vantongeren | G: 0 | H: Passed | I: 12 | J: Woodshop basic had lapsed in August 2026
|
||||
A: 2026-10-03 | B: M-0406 | C: Youth | D: Electronics bench | E: 60 | F: Kwame Lindqvist | G: 0 | H: Passed | I: 12 | J: No mains work; parent M-0310 present
|
||||
A: 2026-10-06 | B: M-0418 | C: Full | D: Woodshop basic | E: 180 | F: Ines Calloway | G: 0 | H: Passed | I: 12 | J: On the waiting list since 14 July 2026
|
||||
A: 2026-10-08 | B: M-0402 | C: Concession | D: Laser supervised job | E: 35 | F: Joss Tannery | G: 0 | H: Passed | I: 6 | J: Laser induction now complete
|
||||
A: 2026-10-10 | B: M-0393 | C: Full | D: Resin printing | E: 90 | F: Ruth Abernethy-Quint | G: 0 | H: Not passed - repeat booked | I: 12 | J: Removed a print without gloves; repeat on 24 October 2026
|
||||
A: 2026-10-14 | B: M-0424 | C: Concession | D: Overlocker and embroidery machine | E: 45 | F: Hettie Galbraith | G: 0 | H: Passed | I: 12 | J: Wednesday afternoon session
|
||||
A: 2026-10-15 | B: M-0420 | C: Full | D: FDM 3D printing | E: 60 | F: Ruth Abernethy-Quint | G: 0 | H: Passed | I: 12 | J: Trained on 3D-05
|
||||
|
||||
A: Date | B: Member no. | C: Tier | D: Induction | E: Duration (min) | F: Inductor | G: Fee charged (£) | H: Result | I: Lapses if unused (months) | J: Notes
|
||||
A: 2026-10-17 | B: M-0399 | C: Full | D: Laser cutter | E: 120 | F: Joss Tannery | G: 0 | H: Passed - supervised job pending | I: 6 | J: Supervised job booked for 29 October 2026
|
||||
A: 2026-10-20 | B: M-0418 | C: Full | D: Table saw | E: 90 | F: Theo Vantongeren | G: 0 | H: Passed | I: 6
|
||||
A: 2026-10-24 | B: M-0393 | C: Full | D: Resin printing | E: 90 | F: Ruth Abernethy-Quint | G: 0 | H: Passed | I: 12 | J: Repeat of 10 October 2026
|
||||
A: 2026-10-27 | B: M-0425 | C: Associate | D: Overlocker and embroidery machine | E: 45 | F: Hettie Galbraith | G: 10 | H: Passed | I: 12 | J: Associate induction fee paid
|
||||
A: 2026-10-31 | B: M-0409 | C: Youth | D: FDM 3D printing | E: 60 | F: Ruth Abernethy-Quint | G: 0 | H: Passed | I: 12 | J: Parent M-0344 present
|
||||
|
||||
#### Budget 2026-27
|
||||
|
||||
A: Line | B: Type | C: Budget 2026-27 (£) | D: Actual 2025-26 (£) | E: Responsible | F: Notes
|
||||
A: Membership fees | B: Income | C: 59600 | D: 51370 | E: Callum Ferrand | F: Full £30 and Concession £18 from 1 September 2026
|
||||
A: Joining fees | B: Income | C: 1200 | D: 1110 | E: Callum Ferrand | F: £15 per new member; 74 members joined in 2025-26
|
||||
A: Laser firing time | B: Income | C: 4300 | D: 3860 | E: Joss Tannery | F: £4 per hour, rounded up to the nearest 15 minutes
|
||||
A: Materials sold at cost | B: Income | C: 2600 | D: 2560 | E: Tobias Wrenfield | F: Filament 5p per gram, resin 12p per ml, timber offcuts
|
||||
A: Courses | B: Income | C: 4500 | D: 4166 | E: Hettie Galbraith | F: Courses are open to non-members
|
||||
A: Junior Saturdays | B: Income | C: 900 | D: 784 | E: Hettie Galbraith | F: £4 per child; at most 12 children per session
|
||||
A: Donations | B: Income | C: 2000 | D: 2160 | E: Tobias Wrenfield
|
||||
A: Grants | B: Income | C: 0 | D: 3000 | E: Tobias Wrenfield | F: 2025-26 grant from the Wexcombe Community Trust was for dust extraction only
|
||||
A: Move appeal (ring-fenced) | B: Income | C: 12000 | D: 0 | E: Sunniva Okafor | F: Opens 1 December 2026, only if members approve the move
|
||||
A: Total income | B: Income | C: 87100 | D: 69010 | E: Tobias Wrenfield
|
||||
A: Rent, Unit 4 | B: Expenditure | C: 27600 | D: 27600 | E: Tobias Wrenfield | F: £2,300 a month, fixed until the lease ends on 30 September 2027
|
||||
A: Electricity | B: Expenditure | C: 8900 | D: 8140 | E: Tobias Wrenfield | F: Single-phase supply at Unit 4
|
||||
A: Insurance | B: Expenditure | C: 3650 | D: 3480 | E: Tobias Wrenfield | F: Renewed each September
|
||||
A: Equipment purchases | B: Expenditure | C: 6500 | D: 10630 | E: Ines Calloway | F: No large purchases planned before the move
|
||||
A: Repairs and maintenance | B: Expenditure | C: 3200 | D: 3585 | E: Ines Calloway | F: 2025-26 included the new laser tube at £1,180
|
||||
A: Consumables | B: Expenditure | C: 7200 | D: 6890 | E: Ruth Abernethy-Quint | F: Filament, resin, solder, sandpaper, blades
|
||||
A: Cleaning, bank charges and sundries | B: Expenditure | C: 4600 | D: 4290 | E: Tobias Wrenfield
|
||||
A: Deposit for Unit 11 | B: Expenditure | C: 7800 | D: 0 | E: Sunniva Okafor | F: Two months' rent at £3,900; due February 2027; paid from the operating budget, not the move fund
|
||||
A: Electrical survey and design, Unit 11 | B: Expenditure | C: 4500 | D: 0 | E: Sunniva Okafor | F: First part of the £14,200 three-phase work; paid from the move fund
|
||||
A: Total expenditure | B: Expenditure | C: 73950 | D: 64615 | E: Tobias Wrenfield
|
||||
A: Surplus | B: Result | C: 13150 | D: 4395 | E: Tobias Wrenfield | F: The 2026-27 surplus includes the £12,000 ring-fenced move appeal
|
||||
|
|
@ -0,0 +1,259 @@
|
|||
# Copperfold Makerspace Member Handbook and Safety Rules
|
||||
|
||||
# Copperfold Makerspace Member Handbook and Safety Rules
|
||||
|
||||
Version 4, effective from 1 September 2026. Approved by the committee on 18 August 2026.
|
||||
|
||||
## 1. About this handbook
|
||||
|
||||
This handbook applies to every member, guest and volunteer at Unit 4, Copperfold Yard. Version 4 replaces version 3 of March 2025. The main changes are the new fees for Full and Concession members, the rule that the laser must be paused before you leave the laser room, and the conditions for running FDM printers overnight. If members approve the move to Larchmere Works, a version 5 will be issued before the new unit opens.
|
||||
|
||||
Every member must read this handbook before their first induction and sign the declaration at the back. Questions about the rules go to the safety officer, Priya Okonkwo-Bell, or to the area lead concerned.
|
||||
|
||||
## 2. Membership tiers
|
||||
|
||||
Tier
|
||||
|
||||
Monthly fee from 1 September 2026
|
||||
|
||||
Who can join
|
||||
|
||||
Access
|
||||
|
||||
Full
|
||||
|
||||
£30
|
||||
|
||||
Anyone aged 18 or over
|
||||
|
||||
Key fob 07:00–23:00 every day; all areas after induction
|
||||
|
||||
Concession
|
||||
|
||||
£18
|
||||
|
||||
Students in full-time education, people aged 65 or over, people on a low income (self-declared)
|
||||
|
||||
Same as Full
|
||||
|
||||
Associate
|
||||
|
||||
£12
|
||||
|
||||
Anyone aged 18 or over
|
||||
|
||||
Staffed open sessions only; no laser cutter, no resin printer, no storage box
|
||||
|
||||
Youth
|
||||
|
||||
£8
|
||||
|
||||
Ages 14–17 with a parent or guardian who is a Full or Concession member
|
||||
|
||||
Only while the parent or guardian is in the building (see section 7)
|
||||
|
||||
A one-off joining fee of £15 applies to every tier except Youth. Fees are paid monthly in advance on the 1st of each month. A member who is more than 30 days behind with fees has their fob switched off until the arrears are paid.
|
||||
|
||||
You may pause your membership for up to 3 months in any 12-month period, for example while travelling or during illness. No fee is due during a pause, but you must empty your storage box unless you pay £3 a month to keep it.
|
||||
|
||||
## 3. Opening hours and access
|
||||
|
||||
Staffed open sessions: Tuesday and Thursday 18:00–21:30, Wednesday 13:00–16:00 and Saturday 10:00–16:00.
|
||||
|
||||
Fob access for Full and Concession members: 07:00–23:00 every day. Nobody may be in the building between 23:00 and 07:00.
|
||||
|
||||
Junior Saturdays: the first and third Saturday of each month, 10:00–12:00, for children aged 8–13.
|
||||
|
||||
Fob access is limited to 07:00–23:00 because of our agreement with the residents of the flats above the yard. For the same reason, the noisy machines (the table saw, planer-thicknesser, router table, mitre saw and dust extractor) must be switched off by 22:00, even though members may stay until 23:00.
|
||||
|
||||
Members may bring up to two guests at a time, but only during staffed open sessions. Guests must sign the visitors' book and must be aged 16 or over, unless they are attending a Junior Saturday. Guests must not use any machine that needs an induction.
|
||||
|
||||
A member who is alone in the building may use the 3D printers, the textile corner and the electronics bench, but must not use any woodshop machine or the laser cutter.
|
||||
|
||||
## 4. Inductions
|
||||
|
||||
You must complete an induction before using any machine marked with a yellow 'induction required' tag. Inductions are free for Full, Concession and Youth members. Associate members pay £10 per induction and cannot book the laser cutter or resin printing inductions.
|
||||
|
||||
Induction
|
||||
|
||||
Length
|
||||
|
||||
Covers
|
||||
|
||||
Lapses if unused for
|
||||
|
||||
Woodshop basic
|
||||
|
||||
3 hours
|
||||
|
||||
Bandsaw, pillar drill, mitre saw, belt and disc sander, dust extractor, hand tools
|
||||
|
||||
12 months
|
||||
|
||||
Table saw
|
||||
|
||||
1.5 hours
|
||||
|
||||
Table saw
|
||||
|
||||
6 months
|
||||
|
||||
Planer-thicknesser and router table
|
||||
|
||||
1 hour
|
||||
|
||||
Planer-thicknesser, router table
|
||||
|
||||
12 months
|
||||
|
||||
Laser cutter
|
||||
|
||||
2 hours plus one supervised job
|
||||
|
||||
Laser cutter and its fume extractor
|
||||
|
||||
6 months
|
||||
|
||||
FDM 3D printing
|
||||
|
||||
1 hour
|
||||
|
||||
Printers 3D-01 to 3D-06
|
||||
|
||||
12 months
|
||||
|
||||
Resin printing
|
||||
|
||||
1.5 hours
|
||||
|
||||
Resin printer and wash-and-cure station
|
||||
|
||||
12 months
|
||||
|
||||
Electronics bench
|
||||
|
||||
1 hour
|
||||
|
||||
Soldering stations, bench power supplies, oscilloscope
|
||||
|
||||
12 months
|
||||
|
||||
Overlocker and embroidery machine
|
||||
|
||||
45 minutes
|
||||
|
||||
Overlocker, embroidery machine
|
||||
|
||||
12 months
|
||||
|
||||
Woodshop basic must be completed before you can book the table saw induction or the planer-thicknesser and router table induction. A lapsed induction can be renewed with a 30-minute refresher, but only if it lapsed less than 12 months ago; after that, the full induction must be repeated. The sewing machines need no induction for members aged 18 or over.
|
||||
|
||||
Inductors are volunteers approved by the area lead. To induct on woodshop machines, an inductor must have been inducted on those machines for at least two years and must be signed off by the woodshop lead.
|
||||
|
||||
## 5. General safety rules
|
||||
|
||||
Wear safety glasses in the woodshop and the laser room at all times, even when you are not operating a machine.
|
||||
|
||||
Tie back long hair and remove loose jewellery and scarves before using any rotating machine.
|
||||
|
||||
Do not wear gloves when using the table saw, the bandsaw or the pillar drill.
|
||||
|
||||
Wear closed-toe shoes everywhere in the building.
|
||||
|
||||
Do not use any machine after drinking alcohol. Alcohol is not allowed in the workshop areas; the kitchen during an approved social event is the only exception.
|
||||
|
||||
Food and drink may be consumed only in the kitchen.
|
||||
|
||||
Leave every machine clean and switched off at its isolator.
|
||||
|
||||
Report any damage or fault at once by fixing a red 'do not use' tag to the machine and writing in the fault book.
|
||||
|
||||
## 6. Area rules: woodshop
|
||||
|
||||
The dust extractor must run whenever a woodshop machine is cutting.
|
||||
|
||||
The table saw and the planer-thicknesser may only be used when at least one other adult member is in the building (the two-person rule).
|
||||
|
||||
Because the supply in Unit 4 is single-phase, the table saw and the planer-thicknesser must never run at the same time; doing so trips the main breaker.
|
||||
|
||||
Wear hearing protection when using the table saw, planer-thicknesser, router table or mitre saw.
|
||||
|
||||
Timber in the rack must be no longer than 2.4 metres and must be labelled with your name and the date. Unlabelled timber becomes shared offcuts after 14 days.
|
||||
|
||||
## 7. Area rules: laser cutter
|
||||
|
||||
Never leave the laser room while the laser is firing. If you need to leave, even briefly, pause the job first. This rule was added in version 4 after a fire in the machine on 12 August 2025.
|
||||
|
||||
The fume extractor must be switched on before the laser starts firing.
|
||||
|
||||
Permitted materials: plywood and MDF up to 6 mm, acrylic up to 8 mm, card, paper, vegetable-tanned leather and cotton fabric.
|
||||
|
||||
Prohibited materials: PVC, vinyl, polycarbonate, ABS, fibreglass, anything containing chlorine, and any material you cannot identify.
|
||||
|
||||
Bookings are limited to 2 hours per booking and 4 hours per member per week, and may be made up to 7 days ahead.
|
||||
|
||||
Firing time is charged at £4 an hour, rounded up to the nearest 15 minutes.
|
||||
|
||||
## 8. Area rules: 3D printing and the resin room
|
||||
|
||||
The FDM area and the resin room have different rules, because uncured resin is a skin irritant and gives off fumes.
|
||||
|
||||
Prints expected to last longer than 8 hours must be booked in the print calendar.
|
||||
|
||||
An FDM printer may keep running after 23:00 only if the print was started before 21:00, the person who started it checked the first layer, and the printer stands on the steel fire shelf under the smoke detector.
|
||||
|
||||
Filament from the shared shelf costs 5p per gram, paid by the weight of the finished print including supports.
|
||||
|
||||
The resin room is for Full and Concession members aged 18 or over who have completed the resin printing induction.
|
||||
|
||||
Nitrile gloves and safety glasses must be worn in the resin room at all times, and the extraction fan must be on.
|
||||
|
||||
The resin printer must never run when nobody is in the building.
|
||||
|
||||
Resin from the shelf costs 12p per ml. Used cleaning alcohol goes in the labelled waste container and never down the sink.
|
||||
|
||||
## 9. Area rules: electronics bench and textile corner
|
||||
|
||||
Use lead-free solder only, and run the solder fume extractor while soldering.
|
||||
|
||||
Work on mains voltage is not allowed unless the electronics lead, Kwame Lindqvist, or his deputy, Mirela Banks, is present.
|
||||
|
||||
Lithium polymer batteries may only be charged inside a fire-proof bag on the steel tray and must never be left charging unattended.
|
||||
|
||||
The overlocker and the embroidery machine need an induction. The embroidery machine may be booked for up to 3 hours at a time.
|
||||
|
||||
Switch off and unplug the iron whenever you leave the textile corner.
|
||||
|
||||
## 10. Young people
|
||||
|
||||
Members must be 18 or over, except Youth members aged 14–17. A Youth member's parent or guardian must be in the building throughout the visit. Youth members take a shortened woodshop basic induction of 2 hours that covers only the bandsaw, the pillar drill and hand tools.
|
||||
|
||||
Youth members may use the FDM printers, the electronics bench (but never mains voltage), the sewing machines and, after induction, the overlocker. They may use the bandsaw and the pillar drill only under the direct supervision of a woodshop inductor. Youth members must not use the table saw, planer-thicknesser, router table, mitre saw, laser cutter or resin printer.
|
||||
|
||||
Junior Saturdays are for children aged 8–13 and cost £4 per child per session. Each session takes at most 12 children, with at least one adult volunteer for every 4 children, and a parent or carer must stay in the building. The textile corner is the only area where children at a Junior Saturday may use powered machines, and then only the sewing machines fitted with speed limiters.
|
||||
|
||||
## 11. Accidents and incidents
|
||||
|
||||
Enter every incident, including near-misses, in the incident book within 24 hours.
|
||||
|
||||
If an injury needs more than a plaster, tell the safety officer the same day.
|
||||
|
||||
First-aid kits are kept in the woodshop, the laser room and the kitchen. The eye-wash station is beside the woodshop door.
|
||||
|
||||
CO2 extinguishers are kept beside the laser cutter and the electronics bench, and a fire blanket hangs in the kitchen.
|
||||
|
||||
In a fire, leave by the nearest exit and gather at the far end of the yard by the green gate.
|
||||
|
||||
The names of the members who hold a current first-aid certificate are on the noticeboard by the entrance.
|
||||
|
||||
## 12. Storage and leaving the space
|
||||
|
||||
Each Full and Concession member gets one 54-litre storage box. A second box costs £3 a month. Associate members get no box, and Youth members share their parent's box.
|
||||
|
||||
Projects left on a bench for more than 72 hours get an orange tag. If they are not collected within a further 7 days, they go to the clear-out shelf, and after 14 days on the clear-out shelf they may be given away.
|
||||
|
||||
The last member to leave must check that every machine is off at its isolator, the laser and resin rooms are locked, and the roller door is shut.
|
||||
|
||||
## 13. Breaches and appeals
|
||||
|
||||
A first breach of a safety rule leads to a 30-day suspension of the induction for the machine concerned. After a second serious breach within 12 months, the committee may end the membership, and fees for the current month are not refunded. A member may appeal in writing to the chair within 14 days of the decision. The appeal is decided by three committee members who were not involved in the original decision.
|
||||
|
|
@ -0,0 +1,86 @@
|
|||
#### Dyrkingsplan 2026
|
||||
|
||||
A: Skifte | B: Areal skifte (m²) | C: Kultur | D: Areal kultur (m²) | E: Såing/utplanting (uke) | F: Første høst (uke) | G: Forventet avling (kg) | H: Ansvarlig | I: Merknad
|
||||
A: A | B: 2400 | C: Gulrøtter | D: 1200 | E: 17 | F: 27 | G: 2000 | H: Ingrid Solhaug-Berntsen | I: Tre sådatoer
|
||||
A: A | B: 2400 | C: Pastinakk | D: 400 | E: 17 | F: 36 | G: 450 | H: Ingrid Solhaug-Berntsen
|
||||
A: A | B: 2400 | C: Rødbeter | D: 600 | E: 19 | F: 28 | G: 800 | H: Ingrid Solhaug-Berntsen
|
||||
A: B | B: 2200 | C: Salat | D: 700 | E: 16 | F: 23 | G: 1100 | H: Mira Holten | I: Snegleskjerm: 1 m bar stripe rundt bedene
|
||||
A: B | B: 2200 | C: Spinat | D: 300 | E: 16 | F: 23 | G: 300 | H: Mira Holten
|
||||
A: B | B: 2200 | C: Reddik | D: 200 | E: 17 | F: 23 | G: 150 | H: Mira Holten | I: Sås hver andre uke
|
||||
A: B | B: 2200 | C: Erter | D: 600 | E: 18 | F: 28 | G: 350 | H: Mira Holten
|
||||
A: C | B: 2300 | C: Hodekål | D: 800 | E: 20 | F: 30 | G: 1000 | H: Ingrid Solhaug-Berntsen | I: Fiberduk mot kålflue til uke 26
|
||||
A: C | B: 2300 | C: Rødkål | D: 400 | E: 20 | F: 34 | G: 450 | H: Ingrid Solhaug-Berntsen
|
||||
A: C | B: 2300 | C: Grønnkål | D: 500 | E: 21 | F: 33 | G: 550 | H: Ingrid Solhaug-Berntsen
|
||||
A: C | B: 2300 | C: Brokkoli | D: 400 | E: 21 | F: 30 | G: 300 | H: Ingrid Solhaug-Berntsen | I: Ny kultur 2026
|
||||
A: D | B: 2200 | C: Løk | D: 900 | E: 18 | F: 32 | G: 950 | H: Mira Holten | I: Poteter flyttet ut av skiftet
|
||||
A: D | B: 2200 | C: Purre | D: 700 | E: 21 | F: 35 | G: 650 | H: Mira Holten
|
||||
A: D | B: 2200 | C: Hvitløk | D: 300 | E: 42 | F: 29 | G: 120 | H: Mira Holten | I: Satt høsten 2025
|
||||
A: E | B: 2300 | C: Squash | D: 900 | E: 22 | F: 28 | G: 1100 | H: Ingrid Solhaug-Berntsen
|
||||
A: E | B: 2300 | C: Gresskar | D: 800 | E: 22 | F: 37 | G: 900 | H: Ingrid Solhaug-Berntsen
|
||||
A: E | B: 2300 | C: Bønner | D: 500 | E: 23 | F: 30 | G: 520 | H: Ingrid Solhaug-Berntsen
|
||||
A: F | B: 2200 | C: Jordbær | D: 1000 | F: 25 | G: 450 | H: Mira Holten | I: Flerårig; fiberduk ved frostvarsel
|
||||
A: F | B: 2200 | C: Blomster | D: 600 | E: 20 | F: 27 | G: 0 | H: Mira Holten | I: Ca. 2 100 buketter
|
||||
A: F | B: 2200 | C: Urter | D: 400 | E: 19 | F: 24 | G: 250 | H: Mira Holten | I: Bigården ligger bak skiftet
|
||||
A: G | B: 1400 | C: Grønngjødsling (kløver) | D: 1400 | E: 18 | G: 0 | H: Ingrid Solhaug-Berntsen | I: Hvileår, ingen høsting
|
||||
A: H | B: 3000 | C: Poteter | D: 2800 | E: 19 | F: 30 | G: 2900 | H: Ingrid Solhaug-Berntsen | I: Nytt skifte 2026, leid av Arne Dyrstad
|
||||
A: Drivhus | B: 180 | C: Tomat | D: 110 | E: 15 | F: 26 | G: 1500 | H: Mira Holten
|
||||
A: Drivhus | B: 180 | C: Agurk | D: 50 | E: 17 | F: 26 | G: 750 | H: Mira Holten
|
||||
A: Drivhus | B: 180 | C: Paprika | D: 20 | E: 15 | F: 30 | G: 260 | H: Mira Holten
|
||||
A: Tunnelhus | B: 120 | C: Tomat | D: 60 | E: 21 | F: 29 | G: 650 | H: Mira Holten | I: Nytt tunnelhus, i bruk fra uke 21
|
||||
A: Tunnelhus | B: 120 | C: Chili | D: 20 | E: 21 | F: 31 | G: 60 | H: Mira Holten
|
||||
A: Tunnelhus | B: 120 | C: Sukkermelon | D: 40 | E: 21 | F: 33 | G: 90 | H: Mira Holten | I: Forsøk første år
|
||||
|
||||
#### Høstelogg 2025
|
||||
|
||||
A: Uke | B: Andeler hentet (av 131) | C: Høstet friland (kg) | D: Høstet drivhus (kg) | E: Egg til eggandeler (stk) | F: Nedbør (mm) | G: Merknad
|
||||
A: 24 | B: 104 | C: 180 | D: 0 | E: 300 | F: 12 | G: Første henting tirsdag 10. juni; salat, reddik og spinat
|
||||
A: 25 | B: 110 | C: 260 | D: 0 | E: 300 | F: 8
|
||||
A: 26 | B: 115 | C: 340 | D: 20 | E: 300 | F: 15 | G: Første tomater fra drivhuset
|
||||
A: 27 | B: 118 | C: 420 | D: 45 | E: 300 | F: 22
|
||||
A: 28 | B: 112 | C: 520 | D: 80 | E: 300 | F: 6
|
||||
A: 29 | B: 98 | C: 600 | D: 120 | E: 300 | F: 3 | G: Fellesferie, lavt oppmøte
|
||||
A: 30 | B: 96 | C: 660 | D: 160 | E: 300 | F: 18 | G: Laveste oppmøte i sesongen
|
||||
A: 31 | B: 101 | C: 720 | D: 190 | E: 300 | F: 25
|
||||
A: 32 | B: 114 | C: 760 | D: 210 | E: 300 | F: 31 | G: Tørråte oppdaget i skifte D onsdag 6. august
|
||||
A: 33 | B: 117 | C: 780 | D: 230 | E: 300 | F: 44 | G: Potetriset slått ned i skifte D 12. august
|
||||
A: 34 | B: 119 | C: 740 | D: 240 | E: 300 | F: 38
|
||||
A: 35 | B: 118 | C: 800 | D: 235 | E: 300 | F: 29 | G: Brunsnegl i salaten i skifte B
|
||||
A: 36 | B: 121 | C: 860 | D: 220 | E: 300 | F: 20 | G: Høyeste oppmøte i sesongen
|
||||
A: 37 | B: 120 | C: 900 | D: 200 | E: 300 | F: 14 | G: Honning delt ut, ett glass per andel
|
||||
A: 38 | B: 116 | C: 880 | D: 170 | E: 300 | F: 9
|
||||
A: 39 | B: 114 | C: 820 | D: 130 | E: 300 | F: 26 | G: Høstfest lørdag 27. september
|
||||
A: 40 | B: 112 | C: 760 | D: 80 | E: 300 | F: 33
|
||||
A: 41 | B: 109 | C: 640 | D: 15 | E: 300 | F: 19 | G: Rev tok 3 høner natt til 9. oktober
|
||||
A: 42 | B: 107 | C: 210 | D: 5 | E: 300 | F: 22 | G: Siste henting fredag 17. oktober
|
||||
|
||||
#### Andelsøkonomi
|
||||
|
||||
A: Post | B: Type | C: Regnskap 2025 (kr) | D: Budsjett 2026 (kr) | E: Merknad
|
||||
A: Helandeler | B: Inntekt | C: 401800 | D: 492800 | E: 98 × 4 100 kr i 2025, 112 × 4 400 kr i 2026
|
||||
A: Halvandeler | B: Inntekt | C: 79200 | D: 93600 | E: 33 × 2 400 kr i 2025, 36 × 2 600 kr i 2026
|
||||
A: Eggandeler | B: Inntekt | C: 18000 | D: 26000 | E: 30 × 600 kr i 2025, 40 × 650 kr i 2026
|
||||
A: Honning solgt på høstfesten | B: Inntekt | C: 8320 | D: 9000 | E: 52 kg à 160 kr i 2025
|
||||
A: Frikjøp av arbeidstimer | B: Inntekt | C: 33000 | D: 18000 | E: 132 timer à 250 kr i 2025; maks 4 timer per andel fra 2026
|
||||
A: Høstfest | B: Inntekt | C: 6400 | D: 7000
|
||||
A: Overført fra reservefondet | B: Inntekt | C: 0 | D: 26000 | E: Til tunnelhuset
|
||||
A: Sum inntekter | B: Sum | C: 546720 | D: 672400
|
||||
A: Lønn gartner og gartnerassistent | B: Utgift | C: 362000 | D: 418000 | E: 1,4 årsverk i 2025, 1,6 årsverk i 2026
|
||||
A: Frø og planter | B: Utgift | C: 41500 | D: 44000
|
||||
A: Gjødsel og kompost | B: Utgift | C: 18200 | D: 19500
|
||||
A: Fôr til høner | B: Utgift | C: 14600 | D: 17800 | E: 28 høner i 2025, 34 høner i 2026
|
||||
A: Birøkt, utstyr | B: Utgift | C: 9800 | D: 12600 | E: Fra 4 til 6 bikuber
|
||||
A: Vanning og strøm | B: Utgift | C: 22300 | D: 24000
|
||||
A: Vedlikehold drivhus | B: Utgift | C: 16900 | D: 9000 | E: Ny plastduk etter storm i februar 2025
|
||||
A: Nytt tunnelhus (investering) | B: Utgift | C: 0 | D: 64000 | E: 120 m², 26 000 kr fra reservefondet
|
||||
A: Forsikring | B: Utgift | C: 12400 | D: 13100
|
||||
A: Leie av jord | B: Utgift | C: 24000 | D: 30000 | E: Skifte H koster 6 000 kr per år
|
||||
A: Diverse | B: Utgift | C: 8700 | D: 9000
|
||||
A: Sum utgifter | B: Sum | C: 530400 | D: 661000
|
||||
A: Resultat | B: Sum | C: 16320 | D: 11400 | E: Overskudd 2025 satt av til reservefondet
|
||||
|
||||
#### Andeler og priser
|
||||
|
||||
A: Andelstype | B: Pris 2025 (kr) | C: Pris 2026 (kr) | D: Antall 2025 | E: Antall 2026 | F: Arbeidsplikt 2025 (timer) | G: Arbeidsplikt 2026 (timer) | H: Merknad
|
||||
A: Helandel | B: 4100 | C: 4400 | D: 98 | E: 112 | F: 10 | G: 12 | H: Kan betales i to rater: 15. april og 15. juli
|
||||
A: Halvandel | B: 2400 | C: 2600 | D: 33 | E: 36 | F: 6 | G: 8 | H: Betales i én rate innen 15. april
|
||||
A: Eggandel | B: 600 | C: 650 | D: 30 | E: 40 | F: 0 | G: 0 | H: Krever hel- eller halvandel; frist 1. mai
|
||||
|
|
@ -0,0 +1,175 @@
|
|||
## First Light: Introduction to the Beginners' Course, Spring 2026
|
||||
|
||||
## Welcome to First Light, spring 2026
|
||||
|
||||
This course runs on six Tuesday evenings from 3 February to 10 March 2026.
|
||||
|
||||
Each session runs from 19:30 to 21:30 in the warm room at Tallow Hill Observatory.
|
||||
|
||||
Your course leader is Callum Steadman.
|
||||
|
||||
This spring run has 16 places, and all 16 are taken.
|
||||
|
||||
## About the society
|
||||
|
||||
The Tallow Hill Astronomical Society was founded in 1987 and opened its observatory in 1994.
|
||||
|
||||
At the end of 2025 the society had 186 members, up from 171 a year earlier.
|
||||
|
||||
Members run public stargazing evenings, a telescope loan scheme and a schools programme.
|
||||
|
||||
In 2025, 14 of our 32 new members joined after taking this course.
|
||||
|
||||
## The observatory
|
||||
|
||||
The observatory stands 214 m above sea level, 3 km north of Wrexley Green.
|
||||
|
||||
The 4.2 m dome houses the 406 mm Pennick Telescope.
|
||||
|
||||
The Annexe has a roll-off roof and three telescope piers.
|
||||
|
||||
The warm room seats 20 and is the only building with white light, food and hot drinks.
|
||||
|
||||
## The six sessions
|
||||
|
||||
Session 1, 3 February: welcome, the society and how the night sky works.
|
||||
|
||||
Session 2, 10 February: finding your way with a planisphere and star charts.
|
||||
|
||||
Session 3, 17 February: binoculars and telescopes, with hands-on practice indoors.
|
||||
|
||||
Session 4, 24 February: the Moon and the planets.
|
||||
|
||||
Session 5, 3 March: star clusters, nebulae and galaxies.
|
||||
|
||||
Session 6, 10 March: a practical night at the telescopes and the certificates.
|
||||
|
||||
## Completing the course
|
||||
|
||||
You complete the course by attending at least five of the six sessions.
|
||||
|
||||
Everyone who completes receives a First Light certificate at the end of session 6.
|
||||
|
||||
In 2025, 24 of the 29 people who enrolled completed the course.
|
||||
|
||||
Completing the course also shortens the wait before you can borrow a telescope.
|
||||
|
||||
## If the weather is bad
|
||||
|
||||
Sessions 1 to 5 go ahead in any weather, because they are held indoors.
|
||||
|
||||
Session 6 needs a clear sky, because it is held at the telescopes.
|
||||
|
||||
If session 6 is clouded out on 10 March, it moves to Tuesday 17 March 2026.
|
||||
|
||||
The duty director decides at 16:00 on the day and tells you by 17:00.
|
||||
|
||||
## Age and accompanying adults
|
||||
|
||||
The minimum age for the course is 14.
|
||||
|
||||
Participants aged 14 or 15 must be accompanied by an adult at every session.
|
||||
|
||||
The accompanying adult attends free of charge but does not receive a certificate.
|
||||
|
||||
Participants aged 16 or over may attend on their own.
|
||||
|
||||
## Course fees
|
||||
|
||||
The course fee is £20 for members and £35 for non-members.
|
||||
|
||||
The non-member fee includes three months of free membership from the first session.
|
||||
|
||||
Course fees are the same as in 2025.
|
||||
|
||||
Red torches and planispheres are lent to you for the length of the course.
|
||||
|
||||
## Dark adaptation and red light
|
||||
|
||||
Your eyes need about 20 to 30 minutes in darkness to adapt fully.
|
||||
|
||||
Outside the warm room, only red light is allowed anywhere on the site.
|
||||
|
||||
A single glance at a white phone screen spoils the dark adaptation of everyone nearby.
|
||||
|
||||
Please return your red torch at the end of session 6.
|
||||
|
||||
## Safety on the hill
|
||||
|
||||
Nobody may be on the site alone after sunset: the two-person rule applies to everyone.
|
||||
|
||||
The footpath from the lower gate is about 400 m long and takes about ten minutes.
|
||||
|
||||
In frost or snow the duty director may close the footpath to guests.
|
||||
|
||||
Report any fall or injury to the session leader on the same evening.
|
||||
|
||||
## Inside the dome
|
||||
|
||||
The dome holds at most 8 people, including the operator.
|
||||
|
||||
Only qualified operators may move the Pennick Telescope; you may look but not touch the controls.
|
||||
|
||||
The dome is closed if the sustained wind exceeds 35 km/h or humidity rises above 90%.
|
||||
|
||||
The Annexe roof can stay open in stronger wind, up to a sustained 45 km/h.
|
||||
|
||||
## Choosing a first instrument
|
||||
|
||||
Binoculars of 7x50 or 10x50 are a good first instrument for most beginners.
|
||||
|
||||
In session 3 you can try a refractor, a reflector and a Maksutov side by side.
|
||||
|
||||
Larger Dobsonians of 150 mm and 200 mm show more but weigh 17.5 to 21 kg.
|
||||
|
||||
Small refractors of 70 mm to 80 mm weigh around 5 kg and set up in minutes.
|
||||
|
||||
## Borrowing a society telescope
|
||||
|
||||
Members can normally borrow a telescope after three months of membership.
|
||||
|
||||
If you complete First Light, you may borrow after just one month of membership.
|
||||
|
||||
Loans last 14 days if they start between 1 November and 28 February, and 28 days at other times.
|
||||
|
||||
The refundable deposit has been £40 since 1 January 2026; it was £30 before that.
|
||||
|
||||
## Heavier telescopes and younger borrowers
|
||||
|
||||
Telescopes weighing 15 kg or more need a 45-minute loan induction first.
|
||||
|
||||
Members aged 14 to 17 may borrow only telescopes under 10 kg, with an adult member co-signing.
|
||||
|
||||
Late returns cost £2 per day, up to a maximum of £30 per loan.
|
||||
|
||||
Only one society telescope may be on loan to a household at a time.
|
||||
|
||||
## Observing after the course
|
||||
|
||||
Public evenings are held on the first Friday of each month from September to April.
|
||||
|
||||
The next public evening is on Friday 6 February 2026, and bookings are capped at 40.
|
||||
|
||||
Members may book Annexe piers 2 and 3 for their own telescopes, at most twice a month.
|
||||
|
||||
In 2025 members logged 61 observing nights with 214 clear hours.
|
||||
|
||||
## Keep Tallow Dark
|
||||
|
||||
Our dark-sky campaign had collected 312 household pledges by the end of 2025.
|
||||
|
||||
The target is 400 pledges by 31 December 2026.
|
||||
|
||||
Sky readings at our six monitoring sites averaged 20.8 magnitudes per square arcsecond in 2025.
|
||||
|
||||
A higher reading means a darker sky; the 2024 average was 20.6.
|
||||
|
||||
## Before next Tuesday
|
||||
|
||||
Bring warm clothes, your red torch and a notebook to every session.
|
||||
|
||||
Session 2 on 10 February covers finding your way around the sky with a planisphere.
|
||||
|
||||
If you will miss a session, tell Callum Steadman beforehand so that the materials can be kept for you.
|
||||
|
||||
Remember that you may miss only one session and still complete the course.
|
||||
|
|
@ -0,0 +1,85 @@
|
|||
#### Club fleet
|
||||
|
||||
A: Boat ID | B: Class | C: Sail no. | D: Year built | E: Length (m) | F: Hull weight (kg) | G: Engine (hp) | H: Condition | I: Last serviced | J: Hours afloat 2026 (h) | K: Notes
|
||||
A: T1 | B: Tern 12 | C: 2101 | D: 2014 | E: 3.6 | F: 52 | G: 0 | H: Fair | I: 2026-03-14 | J: 96 | K: Mast step replaced in March 2026
|
||||
A: T2 | B: Tern 12 | C: 2102 | D: 2014 | E: 3.6 | F: 52 | G: 0 | H: Fair | I: 2026-03-14 | J: 88 | K: Small gelcoat crack on transom; check monthly
|
||||
A: T3 | B: Tern 12 | C: 2103 | D: 2016 | E: 3.6 | F: 52 | G: 0 | H: Good | I: 2026-03-15 | J: 104
|
||||
A: T4 | B: Tern 12 | C: 2104 | D: 2016 | E: 3.6 | F: 52 | G: 0 | H: Good | I: 2026-03-15 | J: 99
|
||||
A: T5 | B: Tern 12 | C: 2105 | D: 2018 | E: 3.6 | F: 52 | G: 0 | H: Good | I: 2026-03-15 | J: 112 | K: Most used junior boat in 2026
|
||||
A: T6 | B: Tern 12 | C: 2106 | D: 2018 | E: 3.6 | F: 52 | G: 0 | H: Good | I: 2026-03-21 | J: 91
|
||||
A: T7 | B: Tern 12 | C: 2107 | D: 2026 | E: 3.6 | F: 52 | G: 0 | H: New | I: 2026-05-09 | J: 64 | K: Bought May 2026 for £3,900 from Boat Replacement Fund
|
||||
A: T8 | B: Tern 12 | C: 2108 | D: 2026 | E: 3.6 | F: 52 | G: 0 | H: New | I: 2026-05-09 | J: 58 | K: Bought May 2026 for £3,900 from Boat Replacement Fund
|
||||
A: H1 | B: Heron 14 | C: 1401 | D: 2011 | E: 4.3 | F: 98 | G: 0 | H: Fair | I: 2026-03-21 | J: 71 | K: New mainsail needed for 2027
|
||||
A: H2 | B: Heron 14 | C: 1402 | D: 2011 | E: 4.3 | F: 98 | G: 0 | H: Poor | I: 2026-03-21 | J: 38 | K: Out of use since 22 August 2026: centreboard case leak; winter repair
|
||||
A: H3 | B: Heron 14 | C: 1403 | D: 2015 | E: 4.3 | F: 98 | G: 0 | H: Good | I: 2026-03-22 | J: 83
|
||||
A: H4 | B: Heron 14 | C: 1404 | D: 2019 | E: 4.3 | F: 98 | G: 0 | H: Good | I: 2026-03-22 | J: 90
|
||||
A: H5 | B: Heron 14 | C: 1405 | D: 2019 | E: 4.3 | F: 98 | G: 0 | H: Good | I: 2026-03-22 | J: 86
|
||||
A: O1 | B: Osprey 16 | C: 1601 | D: 2009 | E: 4.9 | F: 145 | G: 0 | H: Fair | I: 2026-03-28 | J: 62 | K: Kept for adult Start Sailing courses
|
||||
A: O2 | B: Osprey 16 | C: 1602 | D: 2013 | E: 4.9 | F: 145 | G: 0 | H: Good | I: 2026-03-28 | J: 67 | K: Kept for adult Start Sailing courses
|
||||
A: Gannet | B: Safety boat (RIB) | C: n/a | D: 2020 | E: 4.8 | F: 420 | G: 40 | H: Good | I: 2026-03-07 | J: 128 | K: Primary safety boat; helm needs Club Powerboat Certificate
|
||||
A: Puffin | B: Safety boat (RIB) | C: n/a | D: 2015 | E: 4.2 | F: 310 | G: 25 | H: Fair | I: 2026-03-07 | J: 84 | K: Second safety boat; new kill cord fitted April 2026
|
||||
A: Mallard | B: Workboat (aluminium) | C: n/a | D: 2008 | E: 5.5 | F: 520 | G: 20 | H: Fair | I: 2026-02-28 | J: 46 | K: Mooring work only; not a primary safety boat
|
||||
|
||||
#### Moorings north trot
|
||||
|
||||
A: Mooring no. | B: Holder | C: Boat name | D: Boat type | E: Boat length (m) | F: Season fee (£) | G: Paid by 1 April | H: Last chain inspection | I: Status | J: Notes
|
||||
A: N01 | B: A. Treadgold | C: Whimbrel | D: Day boat | E: 5.8 | F: 210 | G: Yes | H: 2025-03 | I: Let | J: Moorings Officer
|
||||
A: N02 | B: J. Whitcombe | C: Second Wind | D: Small cruiser | E: 7.4 | F: 265 | G: Yes | H: 2025-03 | I: Let
|
||||
A: N03 | B: M. Okonjo | C: Bramble | D: Keelboat | E: 6.2 | F: 210 | G: Yes | H: 2025-03 | I: Let
|
||||
A: N04 | B: S. Pellow | C: Halcyon | D: Small cruiser | E: 7.9 | F: 265 | G: Yes | H: 2025-03 | I: Let | J: Largest boat on the north trot
|
||||
A: N05 | B: D. Ferrier | C: Curlew | D: Keelboat | E: 6 | F: 210 | G: Yes | H: 2026-03 | I: Let
|
||||
A: N06 | B: E. Rasmussen | C: Fair Isle | D: Day boat | E: 5.5 | F: 210 | G: Yes | H: 2026-03 | I: Let
|
||||
A: N07 | B: C. Dewhurst | C: Merrybell | D: Small cruiser | E: 6.8 | F: 265 | G: No | H: 2026-03 | I: Let | J: Paid 14 April 2026; committee allowed late payment once
|
||||
A: N08 | B: R. Penhallow | C: Linnet | D: Keelboat | E: 6.1 | F: 210 | G: Yes | H: 2026-03 | I: Let
|
||||
A: N09 | B: - | C: - | D: - | E: 0 | F: 0 | G: - | H: 2026-03 | I: Out of use | J: Sinker failed inspection March 2026; replace winter 2026-27; former holder offered it first
|
||||
A: N10 | B: P. Kaminski | C: Saltire | D: Small cruiser | E: 7.2 | F: 265 | G: Yes | H: 2026-03 | I: Let
|
||||
A: N11 | B: H. Oduya | C: Teal | D: Day boat | E: 5.2 | F: 210 | G: Yes | H: 2025-03 | I: Let | J: Smallest boat on the north trot
|
||||
A: N12 | B: G. Oakeshott | C: Ledger | D: Keelboat | E: 6.4 | F: 210 | G: Yes | H: 2025-03 | I: Let
|
||||
A: N13 | B: L. Brannigan | C: Morwenna | D: Small cruiser | E: 7.6 | F: 265 | G: Yes | H: 2025-03 | I: Let
|
||||
A: N14 | B: T. Ashworth | C: Kittiwake | D: Day boat | E: 5.9 | F: 210 | G: Yes | H: 2025-03 | I: Let
|
||||
A: N15 | B: F. Brackley | C: Sea Pink | D: Keelboat | E: 6.3 | F: 210 | G: Yes | H: 2026-03 | I: Let
|
||||
A: N16 | B: W. Harker | C: Orion | D: Small cruiser | E: 7 | F: 265 | G: Yes | H: 2026-03 | I: Let | J: Chain renewed March 2026
|
||||
A: N17 | B: - | C: - | D: - | E: 0 | F: 0 | G: - | H: 2026-03 | I: Out of use | J: Sinker failed inspection March 2026; replace winter 2026-27; former holder offered it first
|
||||
A: N18 | B: B. Castellan | C: Puzzle | D: Day boat | E: 5.6 | F: 210 | G: Yes | H: 2025-03 | I: Let
|
||||
A: N19 | B: K. Ashdown | C: Driftwood | D: Keelboat | E: 6.5 | F: 210 | G: Yes | H: 2025-03 | I: Let | J: At the 6.5 m limit for the lower fee
|
||||
A: N20 | B: I. Marchetti | C: Solstice | D: Small cruiser | E: 7.1 | F: 265 | G: Yes | H: 2025-03 | I: Let
|
||||
|
||||
#### Summer Evening Series 2026
|
||||
|
||||
A: Position | B: Helm | C: Crew | D: Class | E: Sail no. | F: Club yardstick | G: Races sailed | H: Net points | I: Notes
|
||||
A: 1 | B: Owen Fairclough | C: Lily Fairclough | D: Heron 14 | E: 3317 | F: 1085 | G: 9 | H: 9 | I: Series winners; two discards
|
||||
A: 2 | B: Rosa Lindqvist | C: - | D: Swift 15 | E: 772 | F: 1020 | G: 9 | H: 14
|
||||
A: 3 | B: Duncan Ferrier | C: - | D: Lapwing 13 | E: 4410 | F: 1110 | G: 8 | H: 19
|
||||
A: 4 | B: Priya Castellan | C: Ben Castellan | D: Heron 14 | E: 3290 | F: 1085 | G: 9 | H: 22
|
||||
A: 5 | B: Mark Okonjo | C: - | D: Tern 12 | E: 2088 | F: 1140 | G: 9 | H: 27
|
||||
A: 6 | B: Hannah Oduya | C: Joel Oduya | D: Osprey 16 | E: 1588 | F: 1060 | G: 7 | H: 31
|
||||
A: 7 | B: Stuart Pellow | C: - | D: Lapwing 13 | E: 4377 | F: 1110 | G: 8 | H: 36
|
||||
A: 8 | B: Eva Rasmussen | C: Kit Rasmussen | D: Heron 14 | E: 3104 | F: 1085 | G: 8 | H: 41
|
||||
A: 9 | B: Tomasz Wilde | C: - | D: Swift 15 | E: 801 | F: 1020 | G: 5 | H: 47 | I: Missed four races on instructor duty
|
||||
A: 10 | B: Laura Brannigan | C: - | D: Tern 12 | E: 2105 | F: 1140 | G: 7 | H: 52 | I: Sailed club boat T5 with the Rear Commodore's agreement
|
||||
A: 11 | B: William Harker | C: Amy Harker | D: Heron 14 | E: 1403 | F: 1085 | G: 6 | H: 58 | I: Sailed club boat H3 with the Rear Commodore's agreement
|
||||
A: 12 | B: Ivo Marchetti | C: - | D: Lapwing 13 | E: 4392 | F: 1110 | G: 6 | H: 63
|
||||
A: 13 | B: Peter Kaminski | C: - | D: Tern 12 | E: 2059 | F: 1140 | G: 4 | H: 71 | I: One discard only
|
||||
A: 14 | B: Kerry Ashdown | C: Maisie Ashdown | D: Heron 14 | E: 3256 | F: 1085 | G: 3 | H: 80 | I: Too few races for any discard
|
||||
|
||||
#### Duty rota Spring 2027
|
||||
|
||||
A: Date | B: Day | C: Activity | D: First warning or start | E: Officer of the day | F: Safety boat helm | G: Safety boat crew | H: Shore duty | I: Safety boats planned | J: Notes
|
||||
A: 2027-03-27 | B: Saturday | C: Fitting-out work party | D: 09:00 | E: Duncan Ferrier | F: - | G: - | H: Colin Dewhurst | I: 0 | J: No boats afloat; ashore work only
|
||||
A: 2027-04-03 | B: Saturday | C: Mooring laying day | D: 09:30 | E: Alan Treadgold | F: Sam Ridley | G: Will Harker | H: - | I: 0 | J: Workboat Mallard only; not a sailing day
|
||||
A: 2027-04-10 | B: Saturday | C: Season opening and safety boat refresher | D: 10:00 | E: Fiona Brackley | F: Fiona Brackley | G: Joel Oduya | H: Ruth Penhallow | I: 1 | J: Refresher required for all 2027 safety boat helms
|
||||
A: 2027-04-11 | B: Sunday | C: Spring Series race 1 | D: 10:30 | E: Priya Castellan | F: Mark Okonjo | G: Kit Rasmussen | H: Hannah Oduya | I: 2
|
||||
A: 2027-04-17 | B: Saturday | C: Start Sailing course weekend 1 | D: 09:30 | E: Tomasz Wilde | F: Instructor team | G: Instructor team | H: - | I: 1 | J: Course continues Sunday 18 April
|
||||
A: 2027-04-18 | B: Sunday | C: Spring Series race 2 | D: 10:30 | E: Stuart Pellow | F: Graham Oakeshott | G: Amy Harker | H: Laura Brannigan | I: 2
|
||||
A: 2027-04-24 | B: Saturday | C: Start Sailing course weekend 2 | D: 09:30 | E: Tomasz Wilde | F: Instructor team | G: Instructor team | H: - | I: 1 | J: Course continues Sunday 25 April
|
||||
A: 2027-04-25 | B: Sunday | C: Spring Series race 3 | D: 10:30 | E: Eva Rasmussen | F: Ivo Marchetti | G: Ben Castellan | H: Peter Kaminski | I: 2
|
||||
A: 2027-05-01 | B: Saturday | C: Junior training session 1 | D: 09:30 | E: Kerry Ashdown | F: Owen Fairclough | G: Joel Oduya | H: Ruth Penhallow | I: 2 | J: First junior session of 2027
|
||||
A: 2027-05-02 | B: Sunday | C: Spring Series race 4 | D: 10:30 | E: Rosa Lindqvist | F: William Harker | G: Lily Fairclough | H: Colin Dewhurst | I: 2
|
||||
A: 2027-05-08 | B: Saturday | C: Junior training session 2 | D: 09:30 | E: Kerry Ashdown | F: Mark Okonjo | G: Kit Rasmussen | H: Laura Brannigan | I: 2
|
||||
A: 2027-05-09 | B: Sunday | C: Spring Series race 5 | D: 10:30 | E: Duncan Ferrier | F: Peter Kaminski | G: Amy Harker | H: Hannah Oduya | I: 2
|
||||
A: 2027-05-15 | B: Saturday | C: Junior training session 3 | D: 09:30 | E: Kerry Ashdown | F: Graham Oakeshott | G: Ben Castellan | H: Eva Rasmussen | I: 2
|
||||
A: 2027-05-16 | B: Sunday | C: Spring Series race 6 | D: 10:30 | E: Priya Castellan | F: Owen Fairclough | G: Joel Oduya | H: Ivo Marchetti | I: 2 | J: Swap agreed: Owen Fairclough replaces Ivo Marchetti
|
||||
A: 2027-05-22 | B: Saturday | C: Junior training session 4 | D: 09:30 | E: Kerry Ashdown | F: Ivo Marchetti | G: Lily Fairclough | H: Stuart Pellow | I: 2
|
||||
A: 2027-05-23 | B: Sunday | C: Spring Series race 7 | D: 10:30 | E: Stuart Pellow | F: Fiona Brackley | G: Kit Rasmussen | H: Rosa Lindqvist | I: 2
|
||||
A: 2027-05-29 | B: Saturday | C: Junior training session 5 | D: 09:30 | E: Kerry Ashdown | F: William Harker | G: Amy Harker | H: Peter Kaminski | I: 2
|
||||
A: 2027-05-30 | B: Sunday | C: Spring Series race 8 | D: 10:30 | E: Eva Rasmussen | F: Mark Okonjo | G: Ben Castellan | H: Graham Oakeshott | I: 2 | J: Last race of the Spring Series; prizegiving at 15:00
|
||||
|
|
@ -0,0 +1,231 @@
|
|||
# Foreldrehåndbok og regler for instrumentlån 2026/2027
|
||||
|
||||
# Foreldrehåndbok og regler for instrumentlån 2026/2027
|
||||
|
||||
Denne håndboka gjelder fra 1. august 2026 og erstatter håndboka for 2025/2026. Den er vedtatt av foreldreutvalget i Tindvika skolekorps 18. juni 2026. Håndboka forklarer hvordan korpset er organisert, hva som forventes av foresatte, og hvilke regler som gjelder når et barn låner instrument av korpset. Alle foresatte skal ha lest den før første øvelse.
|
||||
|
||||
## 1. Avdelinger og opptak
|
||||
|
||||
Korpset har tre avdelinger. Hver avdeling har sin egen dirigent eller instruktør og sin egen kontaktperson i foreldreutvalget.
|
||||
|
||||
Aspirantkorpset tar imot barn fra og med 3. trinn. Aspirantene får gruppetimer på sitt instrument og spiller samspill hver mandag.
|
||||
|
||||
Hovedkorpset er for musikanter fra og med 5. trinn som har bestått overgangsprøven. Overgangsprøven holdes i mai, og normalt har aspiranten da spilt i minst to år.
|
||||
|
||||
Drill- og slaggruppa tar imot barn fra og med 4. trinn. Det kreves ingen prøve, men nye medlemmer starter bare i august og januar.
|
||||
|
||||
En musikant kan være med i både Hovedkorpset og Drill- og slaggruppa. Det er ikke mulig å være med i både Aspirantkorpset og Hovedkorpset samtidig. Innmelding skjer på skjema fra kontaktpersonen, og fristen for innmelding til høstsemesteret 2026 er 20. september 2026.
|
||||
|
||||
Nye aspiranter har seks ukers prøvetid fra første øvelse. Melder et barn seg ut i løpet av prøvetiden, betaler familien ikke kontingent for semesteret, heller ikke når utmeldingen kommer etter 1. oktober. Instrumentet skal da leveres tilbake innen 14 dager, som ved annen utmelding. Prøvetiden gjelder bare Aspirantkorpset og ikke nye medlemmer i Drill- og slaggruppa, fordi de ikke låner instrument med hjem.
|
||||
|
||||
## 2. Øvelser og rom
|
||||
|
||||
Alle øvelser holdes på Tindvika skole. Korpset låner rommene gratis mot at vi følger reglene for hvert rom. Brudd på romreglene kan føre til at korpset mister retten til rommet, og derfor må foresatte som har foreldrevakt, passe på at reglene blir fulgt.
|
||||
|
||||
Rom
|
||||
|
||||
Brukes av
|
||||
|
||||
Tid
|
||||
|
||||
Særlige regler
|
||||
|
||||
Storsalen
|
||||
|
||||
Hovedkorpset
|
||||
|
||||
Tirsdager 18:00–20:00
|
||||
|
||||
Ingen mat i salen. Stolene settes tilbake i stabler langs veggen etter øvelse.
|
||||
|
||||
Musikkrommet (rom 214)
|
||||
|
||||
Aspirantkorpset
|
||||
|
||||
Mandager 16:30–18:30
|
||||
|
||||
Maks 25 personer i rommet. Kun vann er tillatt. Foresatte venter i kantina.
|
||||
|
||||
Gymsalen
|
||||
|
||||
Drill- og slaggruppa
|
||||
|
||||
Onsdager 17:00–18:30
|
||||
|
||||
Innesko er påbudt. Slagverk på stativ skal ikke stå uten tilsyn.
|
||||
|
||||
Instrumentlageret (rom K12)
|
||||
|
||||
Instrumentansvarlig
|
||||
|
||||
Tirsdager 17:30–18:00
|
||||
|
||||
Musikanter under 13 år skal ikke være alene i lageret.
|
||||
|
||||
Aspirantenes gruppetimer går fra 16:30 til 17:15, og samspillet går fra 17:30 til 18:30. Det er maks fire aspiranter i hver gruppetime. Hovedkorpset har i tillegg registerøvelse kl. 17:00–17:50 den første tirsdagen i hver måned.
|
||||
|
||||
## 3. Kontingent og betaling
|
||||
|
||||
Kontingenten ble økt fra 1. august 2026 etter vedtak i årsmøtet 4. juni 2026. Økningen skyldes at den nye avtalen med dirigentene gir 8 prosent høyere honorar fra samme dato. Kontingenten dekker dirigenter, instruktører, noter og den ordinære driften, men ikke seminarer eller turer.
|
||||
|
||||
Avdeling
|
||||
|
||||
Kontingent 2025/2026 (kr per semester)
|
||||
|
||||
Kontingent 2026/2027 (kr per semester)
|
||||
|
||||
Aspirantkorpset
|
||||
|
||||
1 100
|
||||
|
||||
1 200
|
||||
|
||||
Hovedkorpset
|
||||
|
||||
1 650
|
||||
|
||||
1 800
|
||||
|
||||
Drill- og slaggruppa
|
||||
|
||||
1 650
|
||||
|
||||
1 800
|
||||
|
||||
En musikant som er med i både Hovedkorpset og Drill- og slaggruppa, betaler kontingent bare én gang.
|
||||
|
||||
Barn nummer to i samme familie får 25 prosent søskenmoderasjon. Barn nummer tre og flere betaler ikke kontingent.
|
||||
|
||||
Faktura sendes 15. september for høstsemesteret og 15. februar for vårsemesteret, med 30 dagers forfall.
|
||||
|
||||
Familier som har vanskelig for å betale, kan kontakte kassereren. Ingen barn skal måtte slutte på grunn av økonomi, og saken behandles fortrolig.
|
||||
|
||||
### Utmelding
|
||||
|
||||
Utmelding skal skje skriftlig til sekretæren. Kontingenten for høstsemesteret betales fullt ut hvis utmeldingen kommer etter 1. oktober, og kontingenten for vårsemesteret betales fullt ut hvis utmeldingen kommer etter 1. mars. Kommer utmeldingen før disse datoene, faktureres ikke semesteret. En musikant som er påmeldt turen til København, er likevel bundet av reglene for avbestilling av turen, som står i reiseplanen fra årsmøtet.
|
||||
|
||||
## 4. Dugnad og foreldrevakt
|
||||
|
||||
Dugnad er korpsets nest største inntektskilde etter kontingenten. Hver familie skal ta tre ordinære dugnadsvakter per sesong, uansett hvor mange barn familien har i korpset. For hver ordinær vakt som ikke blir gjennomført, faktureres familien 300 kroner. Familier med barn på turen til København i juni 2027 skal i tillegg ta tre ekstra turvakter, og en turvakt som ikke blir gjennomført, faktureres med 400 kroner. Dugnadsansvarlig Amina Rasoul setter opp vaktlistene.
|
||||
|
||||
Foreldrevakt er noe annet enn dugnad. Foreldrevakten låser opp, passer på romreglene og er til stede i pausen. Familier i Hovedkorpset og Drill- og slaggruppa har foreldrevakt to ganger per semester, mens familier i Aspirantkorpset har foreldrevakt én gang per semester. Den som ikke kan ta vakten sin, må selv bytte med en annen familie og gi beskjed til kontaktpersonen.
|
||||
|
||||
## 5. Fravær og beskjeder
|
||||
|
||||
Fravær meldes til avdelingens kontaktperson innen kl. 14:00 samme dag som øvelsen.
|
||||
|
||||
Mer enn tre fravær uten gyldig grunn i ett semester fører til en samtale mellom dirigent, musikant og foresatte.
|
||||
|
||||
Før konserter er generalprøven obligatorisk. En musikant som uteblir fra generalprøven uten gyldig grunn, kan ikke spille på konserten.
|
||||
|
||||
Beskjeder fra korpset sendes på e-post senest fem dager før en endring, med unntak av avlysning på grunn av sykdom hos dirigenten.
|
||||
|
||||
## 6. Seminarer
|
||||
|
||||
Korpset holder to helgeseminarer i 2026/2027 på Granlia leirskole: høstseminaret 16.–18. oktober 2026 og vinterseminaret 5.–7. februar 2027. Deltakeravgiften er 700 kroner per seminar for musikanter i Hovedkorpset og Drill- og slaggruppa. Påmeldingsfristen er 25. september 2026 for høstseminaret og 10. januar 2027 for vinterseminaret.
|
||||
|
||||
Seminarene er obligatoriske for musikanter i Hovedkorpset, fordi konsertprogrammet øves inn der. Påmeldte som melder avbud senere enn sju dager før seminaret, betaler full deltakeravgift, med mindre avbudet skyldes sykdom. To foresatte er på vakt hele natten, og musikantene leverer mobiltelefonen til vaktene kl. 22:00.
|
||||
|
||||
Aspirantene deltar bare lørdag fra 10:00 til 16:00. Fra høsten 2026 betaler aspiranter 250 kroner for seminardagen. Tidligere dekket korpset dette, men foreldreutvalget har endret ordningen fordi antallet aspiranter har økt fra 15 til 21 på ett år. Foresatte henter aspirantene på Granlia leirskole kl. 16:00.
|
||||
|
||||
## 7. Konserter, uniform og bilder
|
||||
|
||||
Musikanter i Hovedkorpset låner korpsuniform. Uniformen deles ut på første øvelse i september og leveres inn sammen med instrumentet innen 15. juni. Aspiranter har ikke uniform; de spiller i korpsets t-skjorte og svarte bukser. Musikanter i Drill- og slaggruppa bruker drilldrakt, som også lånes av korpset.
|
||||
|
||||
Uniformen skal ikke vaskes i maskin. Korpset sender alle uniformer til rens én gang per år, i juli.
|
||||
|
||||
Små reparasjoner, som en løs knapp, kan familien gjøre selv.
|
||||
|
||||
Hvis en uniformsjakke blir borte, betaler familien 1 500 kroner. Bukse og lue erstattes med 400 kroner hver.
|
||||
|
||||
Oppmøte før konserter er 60 minutter før start for Hovedkorpset og 45 minutter før start for Aspirantkorpset.
|
||||
|
||||
Korpset tar bilder og korte filmopptak på konserter og seminarer. Bilder der et barn kan gjenkjennes, publiseres bare hvis en foresatt har gitt skriftlig samtykke i innmeldingsskjemaet. Samtykket kan trekkes tilbake når som helst ved å gi beskjed til sekretæren, og bilder som allerede er publisert, fjernes da innen sju dager. Foresatte som filmer på konserter, skal ikke dele opptak der andre barn kan gjenkjennes.
|
||||
|
||||
## 8. Regler for instrumentlån
|
||||
|
||||
Korpset låner ut instrumenter til musikanter som ikke har eget instrument. En foresatt må signere en låneavtale før instrumentet blir utlevert. Låneavtalen gjelder ett skoleår om gangen, fra 1. august til 15. juni. Instrumentet tilhører korpset hele tiden, og korpset kan kreve det tilbake hvis reglene ikke blir fulgt.
|
||||
|
||||
### Leie
|
||||
|
||||
Leien er 350 kroner per semester for hvert instrument og faktureres sammen med kontingenten.
|
||||
|
||||
Aspiranter betaler ikke leie i sitt første år i korpset. Fra det andre året betaler de ordinær leie.
|
||||
|
||||
Søskenmoderasjonen gjelder bare kontingenten, ikke instrumentleien.
|
||||
|
||||
Stikker, øvingspad, ventilolje og notestativ til hjemmebruk kjøper familien selv. Klarinett- og saksofonspillere får tre rør per semester fra korpset.
|
||||
|
||||
### Hva musikanten skal og ikke skal gjøre
|
||||
|
||||
Instrumentet skal alltid oppbevares i kassen når det ikke er i bruk.
|
||||
|
||||
Instrumentet skal ikke oppbevares i rom som er kaldere enn 5 grader eller står i direkte sol.
|
||||
|
||||
Musikanten har ikke lov til å låne instrumentet videre til andre, heller ikke til søsken.
|
||||
|
||||
Familien skal ikke reparere instrumentet selv eller levere det til reparasjon. All service bestilles av instrumentansvarlig.
|
||||
|
||||
Skader skal meldes til instrumentansvarlig innen sju dager.
|
||||
|
||||
### Slagverk
|
||||
|
||||
Store slagverksinstrumenter som stortromme, pauker og xylofon skal ikke tas med hjem. Marsjtrommer kan tas med hjem bare i de to ukene før en konkurranse eller mønstring, og bare når instruktøren har gitt tillatelse. Slagverkere i Drill- og slaggruppa øver hjemme på øvingspad.
|
||||
|
||||
### Skade og tap
|
||||
|
||||
Korpsets forsikring dekker skader på instrumenter som skjer ved vanlig bruk. Ved skade som skyldes uaktsomhet, betaler familien en egenandel på inntil 1 000 kroner. Hvis et instrument blir borte fordi det ble etterlatt uten tilsyn, betaler familien en egenandel på 2 500 kroner. Egenandelen faktureres etter at instrumentansvarlig har vurdert saken.
|
||||
|
||||
### Instrumentsjekk og innlevering
|
||||
|
||||
Alle utlånte instrumenter skal til instrumentsjekk i uke 34, før første øvelse i august. Alle instrumenter skal leveres inn innen 15. juni slik at korpset kan sende dem til årlig service om sommeren. Musikanter som skal på turen til København, beholder instrumentet til etter turen og leverer det innen 25. juni 2027. Instrumentlageret er åpent for innlevering tirsdager 17:30–18:00.
|
||||
|
||||
Ved utmelding skal instrumentet leveres tilbake innen 14 dager. Hvis instrumentet ikke er levert innen fristen, faktureres leie for et helt semester, og korpset kan i tillegg kreve erstatning hvis instrumentet ikke blir levert.
|
||||
|
||||
## 9. Foreldreutvalget og kontaktpersoner
|
||||
|
||||
Foreldreutvalget er korpsets styre og velges på årsmøtet i juni. Utvalget har seks medlemmer i 2026/2027.
|
||||
|
||||
Navn
|
||||
|
||||
Rolle
|
||||
|
||||
Kontaktperson for
|
||||
|
||||
Henrik Bråtveit
|
||||
|
||||
Leder
|
||||
|
||||
Hovedkorpset
|
||||
|
||||
Mona Fjellstad
|
||||
|
||||
Kasserer
|
||||
|
||||
Fakturaer og betaling
|
||||
|
||||
Siri Nyhagen
|
||||
|
||||
Instrumentansvarlig
|
||||
|
||||
Instrumentlån og instrumentlageret
|
||||
|
||||
Kari Enerhaug
|
||||
|
||||
Leder av turkomiteen
|
||||
|
||||
Turen til København 2027
|
||||
|
||||
Øystein Lunde
|
||||
|
||||
Sekretær
|
||||
|
||||
Aspirantkorpset
|
||||
|
||||
Amina Rasoul
|
||||
|
||||
Dugnadsansvarlig
|
||||
|
||||
Drill- og slaggruppa
|
||||
|
||||
Dirigentene og instruktøren er ansatt av korpset og er ikke medlemmer av foreldreutvalget. Spørsmål om musikk og plassering i korpset går til dirigenten, mens spørsmål om penger, dugnad og regler går til foreldreutvalget.
|
||||
191
gates/calibration-folder/kilde/generalforsamling-2026.pptx.md
Normal file
191
gates/calibration-folder/kilde/generalforsamling-2026.pptx.md
Normal file
|
|
@ -0,0 +1,191 @@
|
|||
## Ordinær generalforsamling 2026
|
||||
|
||||
## Ordinær generalforsamling i Kvernhaugen borettslag
|
||||
|
||||
Generalforsamlingen holdes tirsdag 21. april 2026 kl. 18:30.
|
||||
|
||||
Møtet holdes i et leid lokale i nabolaget, fordi fellesrommet bare har plass til 40 personer.
|
||||
|
||||
Registrering av stemmer og fullmakter starter kl. 18:00.
|
||||
|
||||
Hver leilighet har én stemme, også når den eies av flere.
|
||||
|
||||
## Dagsorden
|
||||
|
||||
Konstituering og godkjenning av innkallingen.
|
||||
|
||||
Årsberetning og årsregnskap for 2025.
|
||||
|
||||
Styrehonorar for 2026.
|
||||
|
||||
Forslag om rehabilitering av balkonger og fasader.
|
||||
|
||||
Forslag om endret ro-tid i husordensreglene.
|
||||
|
||||
Valg av styremedlemmer og valgkomité.
|
||||
|
||||
## Borettslaget i tall
|
||||
|
||||
Borettslaget har 72 leiligheter i blokkene A, B, C og D.
|
||||
|
||||
Det er 24 toroms, 36 treroms og 12 fireroms leiligheter.
|
||||
|
||||
9 leiligheter skiftet eier i 2025, mot 6 i 2024.
|
||||
|
||||
Styret holdt 11 styremøter og behandlet 96 saker i 2025.
|
||||
|
||||
## Regnskap 2025: inntekter
|
||||
|
||||
Felleskostnadene utgjorde 5 184 000 kr etter en økning på 4 % fra 1. januar 2025.
|
||||
|
||||
Vaskeriet ga 61 750 kr i inntekter, mot 58 900 kr året før.
|
||||
|
||||
Fellesrommet var utleid i 64 døgn og ga 28 800 kr.
|
||||
|
||||
Salg av 24 ekstra nøkkelbrikker ga 9 600 kr.
|
||||
|
||||
Sum driftsinntekter ble 5 284 150 kr.
|
||||
|
||||
## Regnskap 2025: kostnader og resultat
|
||||
|
||||
Sum driftskostnader ble 4 088 300 kr, mot 3 610 800 kr i 2024.
|
||||
|
||||
Vedlikehold og reparasjoner kostet 1 236 800 kr og var den største posten.
|
||||
|
||||
Forsikringen økte med 10 % til 486 000 kr etter flere vannskader.
|
||||
|
||||
Årsresultatet ble 745 950 kr, mot 992 000 kr i 2024.
|
||||
|
||||
Revisor har ikke hatt merknader til regnskapet.
|
||||
|
||||
## Lån og disposisjonsfond
|
||||
|
||||
Fellesgjelden var 11 640 000 kr per 31. desember 2025.
|
||||
|
||||
Lånet ble nedbetalt med 620 000 kr i løpet av 2025.
|
||||
|
||||
Renten falt fra 4,40 % til 4,15 % gjennom året.
|
||||
|
||||
Disposisjonsfondet var 3 412 500 kr ved årsskiftet.
|
||||
|
||||
## Budsjett 2026
|
||||
|
||||
Felleskostnadene ble økt med 6 % fra 1. januar 2026 til om lag 6 360 kr i snitt per måned.
|
||||
|
||||
Resten av forprosjektet er budsjettert til 1 051 500 kr.
|
||||
|
||||
Ordinært vedlikehold er budsjettert til 980 000 kr.
|
||||
|
||||
Budsjettet viser et årsresultat på 173 500 kr.
|
||||
|
||||
Styret foreslår et styrehonorar på 186 000 kr for 2026, mot 180 000 kr for 2025.
|
||||
|
||||
## Vedlikehold utført i 2025
|
||||
|
||||
LED-belysning med bevegelsessensor ble montert i alle trapperom i mars for 186 000 kr.
|
||||
|
||||
Taklekkasjen i blokk C ble reparert i juni for 214 600 kr.
|
||||
|
||||
Trapperommene i blokk D ble malt i august for 167 300 kr.
|
||||
|
||||
To av fire vaskemaskiner ble byttet i september for 64 800 kr.
|
||||
|
||||
Tilstandsrapporten for balkonger og fasader kostet 148 500 kr.
|
||||
|
||||
## Energi i fellesarealene
|
||||
|
||||
Strømforbruket gikk ned fra 241 300 kWh i 2024 til 214 600 kWh i 2025.
|
||||
|
||||
Nedgangen på 11 % skyldes først og fremst den nye belysningen i trapperommene.
|
||||
|
||||
Energikostnaden falt fra 352 100 kr til 318 700 kr.
|
||||
|
||||
Budsjettet for 2026 regner med 305 000 kr til energi.
|
||||
|
||||
## Fellesrom, verksted og sykkelboder
|
||||
|
||||
Leien for fellesrommet har vært 450 kr per døgn siden 1. januar 2025.
|
||||
|
||||
Hobbyverkstedet i blokk C åpnet 1. mars 2025 og er gratis for beboerne.
|
||||
|
||||
Sykkelbod sør fikk ladeskap med 8 låsbare rom for elsykkelbatterier i mai 2025.
|
||||
|
||||
23 umerkede sykler ble fjernet ved ryddeaksjonen i oktober 2025.
|
||||
|
||||
## Forslag: ro-tid fra kl. 22:00 søndag til torsdag
|
||||
|
||||
Styret mottok 14 skriftlige klager på støy i 2025, mot 9 i 2024.
|
||||
|
||||
De fleste klagene gjaldt støy mellom kl. 22:00 og kl. 23:00 på kvelder fra søndag til torsdag.
|
||||
|
||||
Styret foreslår at ro-tiden søndag til torsdag begynner kl. 22:00 i stedet for kl. 23:00.
|
||||
|
||||
Ro-tiden på fredager og lørdager foreslås uendret.
|
||||
|
||||
Hvis forslaget vedtas, gjelder de nye reglene fra 1. mai 2026.
|
||||
|
||||
## Rehabilitering: hvorfor nå
|
||||
|
||||
Tilstandsrapporten fra desember 2025 fant alvorlige betongskader på 14 balkonger.
|
||||
|
||||
Alle de 14 skadde balkongene ligger i blokk A og blokk B.
|
||||
|
||||
3 balkonger i blokk A er sperret for bruk inntil videre.
|
||||
|
||||
Fasadene har aldri vært etterisolert siden byggeåret.
|
||||
|
||||
Bare å utbedre de 14 balkongene vil koste om lag 6,8 mill. kr uten bedre isolasjon.
|
||||
|
||||
## Rehabilitering: kostnad og finansiering
|
||||
|
||||
Kostnadsoverslaget er 41,0 mill. kr inkludert merverdiavgift.
|
||||
|
||||
Et nytt lån på inntil 39,5 mill. kr skal nedbetales over 30 år.
|
||||
|
||||
1,5 mill. kr hentes fra disposisjonsfondet.
|
||||
|
||||
Felleskostnadene må økes med om lag 12 % fra 1. januar 2027.
|
||||
|
||||
Forslaget krever to tredels flertall av de avgitte stemmene.
|
||||
|
||||
## Rehabilitering: tidsplan
|
||||
|
||||
Anbudsgrunnlaget skal være ferdig og anbud innhentes høsten 2026.
|
||||
|
||||
Blokk A og blokk B rehabiliteres fra april til oktober 2027.
|
||||
|
||||
Blokk C og blokk D rehabiliteres fra april til oktober 2028.
|
||||
|
||||
Hver balkong blir utilgjengelig i om lag 10 uker.
|
||||
|
||||
Beboerne varsles minst 4 uker før stillaset settes opp.
|
||||
|
||||
## Dugnader og hage i 2026
|
||||
|
||||
Vårdugnaden holdes lørdag 25. april og høstdugnaden lørdag 17. oktober.
|
||||
|
||||
I 2025 deltok 41 husstander på vårdugnaden og 35 på høstdugnaden.
|
||||
|
||||
Hagekomiteen har fem medlemmer og ledes av Vibeke Strand.
|
||||
|
||||
9 husstander står på venteliste for en av de 12 hageparsellene.
|
||||
|
||||
## Valg
|
||||
|
||||
Ingrid Solbakk er på valg og stiller til gjenvalg som styreleder for 2 år.
|
||||
|
||||
Henrik Vågsæther ønsker ikke gjenvalg etter 4 år i styret.
|
||||
|
||||
Valgkomiteen foreslår Petter Nygård-Ulvik som nytt styremedlem for 2 år.
|
||||
|
||||
Valgkomiteen består av Oddbjørg Rise og Jonas Fjellstad.
|
||||
|
||||
## Spørsmål og avstemning
|
||||
|
||||
Spørsmål til regnskapet besvares av Amira Hassani, som har ansvar for økonomi.
|
||||
|
||||
Spørsmål om rehabiliteringen besvares av Ingrid Solbakk og Torbjørn Leirvik.
|
||||
|
||||
Avstemningen om rehabiliteringen skjer skriftlig.
|
||||
|
||||
Protokollen fra møtet legges i alle postkasser innen 2 uker.
|
||||
|
|
@ -0,0 +1,205 @@
|
|||
# Husordensregler og praktisk informasjon
|
||||
|
||||
# Husordensregler og praktisk informasjon for Kvernhaugen borettslag
|
||||
|
||||
Husordensreglene gjelder fra 1. mai 2026. De ble vedtatt av generalforsamlingen 21. april 2026 og erstatter reglene fra 2019. Reglene gjelder for alle som bor i borettslaget, også leietakere og gjester. Andelseieren er ansvarlig for at husstanden, leietakere og gjester følger reglene.
|
||||
|
||||
## 1. Ro og støy
|
||||
|
||||
Fra 1. mai 2026 begynner ro-tiden kl. 22:00 på søndager til torsdager, mot kl. 23:00 tidligere. Endringen ble vedtatt på generalforsamlingen 21. april 2026 med 49 stemmer for og 17 stemmer mot, fordi styret mottok 14 skriftlige klager på støy i 2025. Ro-tiden på fredager og lørdager er uendret.
|
||||
|
||||
Kveld
|
||||
|
||||
Ro-tid fra 1. mai 2026
|
||||
|
||||
Ro-tid før 1. mai 2026
|
||||
|
||||
Søndag til torsdag
|
||||
|
||||
Kl. 22:00–07:00
|
||||
|
||||
Kl. 23:00–07:00
|
||||
|
||||
Fredag og lørdag
|
||||
|
||||
Kl. 23:00–09:00
|
||||
|
||||
Kl. 23:00–09:00
|
||||
|
||||
Støyende arbeid er bare tillatt i disse tidsrommene:
|
||||
|
||||
Mandag til fredag kl. 08:00–20:00.
|
||||
|
||||
Lørdag kl. 10:00–16:00.
|
||||
|
||||
Aldri på søndager og helligdager.
|
||||
|
||||
Med støyende arbeid menes boring, banking, sliping og annet arbeid som høres i andre leiligheter. Den som skal pusse opp i mer enn én dag, skal varsle naboene i samme oppgang med oppslag minst 3 dager før arbeidet starter. Oppslaget skal vise hvor lenge arbeidet varer.
|
||||
|
||||
Den som skal ha selskap i leiligheten etter kl. 22:00, bør varsle naboene i oppgangen på forhånd. Et varsel gir likevel ikke rett til å bryte ro-tiden.
|
||||
|
||||
## 2. Fellesrommet og hobbyverkstedet
|
||||
|
||||
Fellesrommet ligger i første etasje i blokk D, og hobbyverkstedet ligger i kjelleren i blokk C. De to rommene har ulike regler, som vist i tabellen.
|
||||
|
||||
Regel
|
||||
|
||||
Fellesrommet (blokk D)
|
||||
|
||||
Hobbyverkstedet (blokk C)
|
||||
|
||||
Størrelse
|
||||
|
||||
62 m², plass til 40 personer
|
||||
|
||||
24 m²
|
||||
|
||||
Pris
|
||||
|
||||
450 kr per døgn og 1 000 kr i depositum
|
||||
|
||||
Gratis for beboere
|
||||
|
||||
Booking
|
||||
|
||||
Hos vaktmester, minst 7 dager og høyst 3 måneder før
|
||||
|
||||
Bare nødvendig ved bruk i mer enn 3 timer
|
||||
|
||||
Aldersgrense
|
||||
|
||||
Leietaker må være minst 20 år
|
||||
|
||||
Barn under 15 år bare sammen med en voksen
|
||||
|
||||
Tider
|
||||
|
||||
Forlatt innen kl. 22:00 søndag til torsdag og innen kl. 01:00 natt til lørdag og søndag
|
||||
|
||||
Åpent kl. 08:00–21:00 alle dager
|
||||
|
||||
Fellesrommet leies bare ut til beboere, og leietakeren må være til stede under hele arrangementet. Fredag og lørdag skal musikken dempes fra kl. 23:00. Leietakeren skal rydde og tørke av bord og benker før rommet forlates. Gulvvask etter utleie er inkludert i leien. Depositumet betales tilbake når vaktmesteren har kontrollert rommet, og bare dersom ingenting er skadet eller mangler.
|
||||
|
||||
Avbestilling av fellesrommet senest 48 timer før leieperioden starter gir full refusjon av leien. Ved senere avbestilling betales leien ikke tilbake, med unntak av sykdom som kan dokumenteres. Det er ikke tillatt å leie rommet videre til andre.
|
||||
|
||||
I hobbyverkstedet er elektrisk verktøy bare tillatt på hverdager kl. 10:00–20:00 og på lørdager kl. 11:00–16:00. Elektrisk verktøy skal aldri brukes på søndager og helligdager. Verkstedet skal ikke brukes til lakkering eller sprøytemaling, fordi rommet ikke har avtrekk for løsemidler. Den siste som går, skal slå av hovedbryteren ved døra.
|
||||
|
||||
## 3. Fellesvaskeriet
|
||||
|
||||
Vaskeriet ligger i kjelleren i blokk B og har fire vaskemaskiner, to tørketromler og ett tørkerom. Vaskeriet er åpent kl. 07:00–22:00 mandag til lørdag og kl. 10:00–18:00 på søndager.
|
||||
|
||||
En vask koster 22 kr, og en runde i tørketrommelen koster 15 kr. Beløpet trekkes via nøkkelbrikken.
|
||||
|
||||
Hver booking kan vare i høyst 3 timer.
|
||||
|
||||
Hver leilighet kan ha høyst 2 bookinger per uke.
|
||||
|
||||
Tøy skal tas ut av tørkerommet senest 24 timer etter at bookingen er slutt.
|
||||
|
||||
Barn under 12 år får ikke bruke vaskeriet uten en voksen.
|
||||
|
||||
Tepper, hundesenger og tøy med sand eller jord skal ikke vaskes i maskinene.
|
||||
|
||||
Dersom en maskin ikke virker, skal feilen meldes til vaktmesteren. Tiden som går tapt, kan bookes på nytt samme uke selv om leiligheten da får mer enn 2 bookinger.
|
||||
|
||||
## 4. Sykkelboder og kjellerboder
|
||||
|
||||
Sykkelbod nord ved blokk A har 48 plasser, og sykkelbod sør ved blokk D har 36 plasser. Begge åpnes med nøkkelbrikke. Sykler skal merkes med leilighetsnummer. Merkelapper deles ut av vaktmesteren i september hvert år.
|
||||
|
||||
Umerkede sykler fjernes ved ryddeaksjonen i oktober, etter fire ukers varsel på oppslagstavlene.
|
||||
|
||||
Fjernede sykler lagres i 3 måneder før de gis bort til gjenbruk.
|
||||
|
||||
Batterier til elsykler skal bare lades i ladeskapet i sykkelbod sør. Lading er ikke tillatt i sykkelbod nord eller i kjellerbodene.
|
||||
|
||||
Ladeskapet har 8 låsbare rom, og hvert rom kan brukes i høyst 12 timer om gangen.
|
||||
|
||||
Hver leilighet har en kjellerbod på mellom 4 og 6 m². Det skal ikke lagres brannfarlige væsker, gassflasker eller maling i kjellerbodene. Gangene i kjelleren skal holdes fri, fordi de er rømningsvei.
|
||||
|
||||
## 5. Trapperom, balkonger og røyking
|
||||
|
||||
Trapperommene er rømningsvei og skal holdes fri for gjenstander. Det eneste som kan stå utenfor leilighetsdøra, er en dørmatte. Sko, esker, møbler og avfallsposer skal ikke settes i trapperommet, heller ikke for en kort periode. Vaktmesteren fjerner gjenstander som blir stående, etter et varsel på 7 dager.
|
||||
|
||||
Røyking er ikke tillatt i trapperom, kjellere, vaskeri, fellesrom, hobbyverksted eller sykkelboder.
|
||||
|
||||
Røyking er heller ikke tillatt nærmere enn 5 meter fra inngangsdørene.
|
||||
|
||||
Røyking på egen balkong er tillatt, men aske og sneiper skal ikke kastes over rekkverket.
|
||||
|
||||
Blomsterkasser skal henge på innsiden av rekkverket.
|
||||
|
||||
Tøy kan tørkes på balkongen bare når det henger lavere enn toppen av rekkverket.
|
||||
|
||||
Det er ikke tillatt å mate fugler fra balkongene, fordi foring trekker til seg rotter og duer.
|
||||
|
||||
Den som flytter inn eller ut, skal melde dette til styret senest 1 uke før flyttedagen. Fordi ingen av blokkene har heis, skal flyttelass bæres i trappen, og vegger og rekkverk som blir skadet under flyttingen, skal utbedres for flytterens regning.
|
||||
|
||||
## 6. Hage, grill og lek
|
||||
|
||||
Hagen på 1 450 m² er felles for alle beboere. Hagekomiteen har ansvar for planting og stell, og vaktmesteren har ansvar for plenklipping. Vårdugnaden holdes 25. april 2026 og høstdugnaden 17. oktober 2026. Det ilegges ikke gebyr for å utebli fra dugnad.
|
||||
|
||||
På balkongene er bare elektrisk grill og gassgrill tillatt. Kullgrill er ikke tillatt på balkongene.
|
||||
|
||||
Kullgrill kan brukes på den felles grillplassen i hagen fram til kl. 22:00.
|
||||
|
||||
Lekeplassen er for barn under 12 år.
|
||||
|
||||
Ballspill på plenen er tillatt fram til kl. 20:00.
|
||||
|
||||
De 12 hageparsellene på 8 m² fordeles ved loddtrekning for to sesonger om gangen. Neste loddtrekning holdes i mars 2027.
|
||||
|
||||
## 7. Dyrehold
|
||||
|
||||
Det er tillatt å holde hund, katt og andre smådyr, men dyreholdet skal meldes skriftlig til styret innen 2 uker etter at dyret er kommet til leiligheten. Hunder skal holdes i bånd på hele eiendommen og skal ikke inn på lekeplassen. Eieren skal plukke opp etter dyret. Styret kan kreve at et dyr fjernes dersom det er til vesentlig sjenanse for naboene og eieren ikke retter forholdet etter skriftlig advarsel.
|
||||
|
||||
## 8. Avfall
|
||||
|
||||
Avfall skal sorteres i de nedgravde containerne for restavfall, papir, plast og glass og metall. Matavfall skal legges i grønne poser i containeren for restavfall. Det er ikke tillatt å sette avfall ved siden av containerne.
|
||||
|
||||
Borettslaget bestiller container for grovavfall to ganger i 2026: 9. til 11. mai og 12. til 14. september. Elektriske produkter, maling, løsemidler og andre kjemikalier skal ikke legges i grovavfallscontaineren.
|
||||
|
||||
## 9. Fremleie og endringer i leiligheten
|
||||
|
||||
Fremleie av hele leiligheten krever skriftlig godkjenning fra styret. Søknaden skal sendes senest 4 uker før leieforholdet starter, og den skal oppgi navn på leietakeren og leieperioden. Korttidsutleie er tillatt i høyst 30 døgn per kalenderår, og hver utleie skal meldes til styret før den starter.
|
||||
|
||||
Den som vil installere varmepumpe, flytte eller fjerne vegger eller gjøre endringer i våtrom, skal søke styret minst 6 uker før arbeidet starter. Innglassing av balkong er ikke tillatt før rehabiliteringen av balkongene er ferdig i oktober 2028, fordi de nye balkongene får en felles løsning for innglassing.
|
||||
|
||||
## 10. Skader, vaktmester og nøkler
|
||||
|
||||
Vaktmester Kåre Myrvold arbeider mandag til torsdag kl. 07:30–15:00. Han har kontortid på vaktmesterkontoret i kjelleren i blokk B tirsdager kl. 08:00–09:00 og torsdager kl. 14:00–15:00. Saker som ikke haster, legges i feilmeldingsboksen ved vaktmesterkontoret.
|
||||
|
||||
Akutte skader utenom arbeidstiden, som vannlekkasje, strømbrudd i fellesarealer og innbrudd i fellesrom, meldes til vakttelefonen. Nummeret står på oppslagstavla i hver oppgang.
|
||||
|
||||
Egenandelen ved forsikringsskader er 10 000 kr. Beboeren betaler egenandelen når skaden skyldes utstyr eller forhold i leiligheten som beboeren har ansvar for.
|
||||
|
||||
Røykvarsleren i leiligheten skal testes av beboeren hvert år innen 1. desember.
|
||||
|
||||
Hver leilighet har 3 nøkkelbrikker. Ekstra brikker koster 400 kr per stykk og bestilles hos vaktmesteren.
|
||||
|
||||
En tapt nøkkelbrikke skal meldes straks, slik at den kan sperres.
|
||||
|
||||
## 11. Byggeperioden 2027–2028
|
||||
|
||||
Generalforsamlingen 21. april 2026 vedtok å rehabilitere balkonger og fasader på alle fire blokkene. Vedtaket ble gjort med 58 stemmer for og 8 stemmer mot, av i alt 66 avgitte stemmer. Kostnadsrammen er 41,0 mill. kr. Blokk A og blokk B rehabiliteres fra april til oktober 2027, og blokk C og blokk D fra april til oktober 2028.
|
||||
|
||||
Beboerne i hver blokk varsles minst 4 uker før stillaset settes opp.
|
||||
|
||||
Balkongen skal være tømt for møbler, planter og annet utstyr innen datoen i varselet.
|
||||
|
||||
Hver balkong er utilgjengelig i om lag 10 uker.
|
||||
|
||||
Vinduer på fasaden som rehabiliteres, kan i perioder på inntil 2 uker ikke åpnes.
|
||||
|
||||
Byggearbeid foregår bare på hverdager kl. 07:00–17:00.
|
||||
|
||||
Felleskostnadene blir ikke redusert i byggeperioden.
|
||||
|
||||
De 3 balkongene i blokk A som styret har sperret, skal fortsatt ikke brukes før de er rehabilitert.
|
||||
|
||||
## 12. Vedlikeholdsansvar
|
||||
|
||||
Andelseieren har ansvar for vedlikehold innvendig i leiligheten, blant annet maling, gulv, kjøkkeninnredning, hvitevarer og sluk i bad. Borettslaget har ansvar for felles rør fram til stoppekranen i leiligheten, for vinduer og ytterdører, for balkonger og for fasader. Det betyr at beboeren ikke skal bytte vindu eller balkongdør på egen hånd. Den som oppdager fukt, råte eller sprekker i bærende konstruksjoner, skal melde fra til vaktmesteren så snart som mulig, slik at skaden ikke blir større.
|
||||
|
||||
## 13. Brudd på reglene
|
||||
|
||||
Klager på brudd på husordensreglene skal være skriftlige og leveres i styrets postkasse i blokk B, oppgang 2. Styret behandler klagen i første styremøte etter at den er mottatt. Ved første brudd sender styret en skriftlig påminnelse. Ved gjentatte eller alvorlige brudd kan styret gi skriftlig advarsel, og i ytterste fall kan vesentlige brudd føre til krav om salg av andelen.
|
||||
|
|
@ -0,0 +1,183 @@
|
|||
## Informasjonsmøte for nye foreldre – Tindvika skolekorps, 2. september 2026
|
||||
|
||||
## Velkommen til Tindvika skolekorps
|
||||
|
||||
Informasjonsmøtet holdes onsdag 2. september 2026 kl. 18:30 i storsalen på Tindvika skole.
|
||||
|
||||
Møtet er for foresatte til nye aspiranter og nye medlemmer i Drill- og slaggruppa.
|
||||
|
||||
Møtet ledes av Henrik Bråtveit, som er leder av foreldreutvalget.
|
||||
|
||||
Etter møtet kan alle prøve instrumenter i musikkrommet fram til kl. 20:00.
|
||||
|
||||
## Om korpset
|
||||
|
||||
Korpset ble stiftet i 1971 og øver på Tindvika skole.
|
||||
|
||||
Korpset hadde 69 medlemmer per 1. september 2025, mot 58 året før.
|
||||
|
||||
Vi har tre avdelinger: Aspirantkorpset, Hovedkorpset og Drill- og slaggruppa.
|
||||
|
||||
Foreldreutvalget er korpsets styre og har seks medlemmer.
|
||||
|
||||
## Aspirantkorpset
|
||||
|
||||
Aspirantkorpset tar imot barn fra og med 3. trinn.
|
||||
|
||||
Gruppetimene går mandager 16:30–17:15 med maks fire aspiranter per gruppe.
|
||||
|
||||
Samspillet går mandager 17:30–18:30 i musikkrommet, rom 214.
|
||||
|
||||
Dirigent for Aspirantkorpset er Tobias Rønneberg.
|
||||
|
||||
## Veien videre i korpset
|
||||
|
||||
Overgangsprøven til Hovedkorpset holdes i mai hvert år.
|
||||
|
||||
Hovedkorpset er for musikanter fra og med 5. trinn som har bestått prøven.
|
||||
|
||||
Normalt spiller en aspirant i minst to år før overgangsprøven.
|
||||
|
||||
Drill- og slaggruppa tar imot barn fra og med 4. trinn, uten prøve.
|
||||
|
||||
Nye medlemmer i Drill- og slaggruppa starter bare i august og januar.
|
||||
|
||||
## Øvingstider og rom
|
||||
|
||||
Hovedkorpset øver tirsdager 18:00–20:00 i storsalen.
|
||||
|
||||
Drill- og slaggruppa øver onsdager 17:00–18:30 i gymsalen, og innesko er påbudt.
|
||||
|
||||
I musikkrommet er kun vann tillatt, og det kan være maks 25 personer i rommet.
|
||||
|
||||
Foresatte som venter på aspiranter, sitter i kantina.
|
||||
|
||||
## Dirigenter og instruktør
|
||||
|
||||
Ingeborg Haugstøl er dirigent for Hovedkorpset og begynner nå sin femte sesong.
|
||||
|
||||
Tobias Rønneberg leder Aspirantkorpset.
|
||||
|
||||
Live Aasmyr er instruktør for Drill- og slaggruppa.
|
||||
|
||||
Dirigentene er ansatt av korpset og sitter ikke i foreldreutvalget.
|
||||
|
||||
## Kontingent 2026/2027
|
||||
|
||||
Aspirantkorpset koster 1 200 kroner per semester, mot 1 100 kroner i fjor.
|
||||
|
||||
Hovedkorpset og Drill- og slaggruppa koster 1 800 kroner per semester, mot 1 650 kroner i fjor.
|
||||
|
||||
Økningen gjelder fra 1. august 2026 fordi dirigentene har fått 8 prosent høyere honorar.
|
||||
|
||||
En musikant som er med i to avdelinger, betaler kontingent bare én gang.
|
||||
|
||||
## Søskenmoderasjon og betaling
|
||||
|
||||
Barn nummer to i samme familie får 25 prosent moderasjon på kontingenten.
|
||||
|
||||
Barn nummer tre og flere betaler ikke kontingent.
|
||||
|
||||
Faktura sendes 15. september og 15. februar med 30 dagers forfall.
|
||||
|
||||
Familier som har vanskelig for å betale, kan ta kontakt med kasserer Mona Fjellstad.
|
||||
|
||||
## Instrumentlån for nye aspiranter
|
||||
|
||||
Korpset låner ut instrument til alle aspiranter som ikke har eget instrument.
|
||||
|
||||
Aspiranter betaler ikke leie det første året i korpset.
|
||||
|
||||
Fra det andre året er leien 350 kroner per semester.
|
||||
|
||||
En foresatt må signere låneavtalen før instrumentet blir utlevert.
|
||||
|
||||
Låneavtalen gjelder fra 1. august til 15. juni.
|
||||
|
||||
## Slik tar dere vare på instrumentet
|
||||
|
||||
Instrumentet skal ligge i kassen når det ikke er i bruk.
|
||||
|
||||
Instrumentet skal ikke stå i rom under 5 grader eller i direkte sol.
|
||||
|
||||
Musikanten har ikke lov til å låne instrumentet bort, heller ikke til søsken.
|
||||
|
||||
Skader meldes til instrumentansvarlig Siri Nyhagen innen sju dager.
|
||||
|
||||
Ved skade som skyldes uaktsomhet, betaler familien en egenandel på inntil 1 000 kroner.
|
||||
|
||||
## Seminarer 2026/2027
|
||||
|
||||
Høstseminaret er 16.–18. oktober 2026, og vinterseminaret er 5.–7. februar 2027, begge på Granlia leirskole.
|
||||
|
||||
Aspirantene deltar bare lørdag fra 10:00 til 16:00.
|
||||
|
||||
Seminardagen koster 250 kroner for aspiranter fra denne høsten.
|
||||
|
||||
Musikanter i de andre avdelingene betaler 700 kroner per seminar.
|
||||
|
||||
Påmeldingsfristen for høstseminaret er 25. september 2026.
|
||||
|
||||
## Konserter denne sesongen
|
||||
|
||||
Høstkonserten er søndag 22. november 2026.
|
||||
|
||||
Julekonserten er søndag 13. desember 2026, og der spiller alle tre avdelingene.
|
||||
|
||||
Støttekonserten for turen til København er søndag 7. mars 2027.
|
||||
|
||||
Vårkonserten er søndag 25. april 2027.
|
||||
|
||||
Generalprøven før hver konsert er obligatorisk.
|
||||
|
||||
## Dugnad
|
||||
|
||||
Hver familie tar tre ordinære dugnadsvakter per sesong, uansett hvor mange barn familien har i korpset.
|
||||
|
||||
En ordinær vakt som ikke blir gjennomført, faktureres med 300 kroner.
|
||||
|
||||
Første dugnad er loppemarkedet i gymsalen lørdag 26. september 2026.
|
||||
|
||||
Dugnadsansvarlig Amina Rasoul setter opp vaktlistene.
|
||||
|
||||
## Foreldrevakt
|
||||
|
||||
Foreldrevakt er noe annet enn dugnad.
|
||||
|
||||
Familier i Aspirantkorpset har foreldrevakt én gang per semester.
|
||||
|
||||
Familier i Hovedkorpset og Drill- og slaggruppa har foreldrevakt to ganger per semester.
|
||||
|
||||
Den som ikke kan ta vakten sin, bytter selv med en annen familie.
|
||||
|
||||
## Fravær og beskjeder
|
||||
|
||||
Fravær meldes til avdelingens kontaktperson innen kl. 14:00 samme dag.
|
||||
|
||||
Kontaktperson for Aspirantkorpset er sekretær Øystein Lunde.
|
||||
|
||||
Kontaktperson for Drill- og slaggruppa er Amina Rasoul.
|
||||
|
||||
Mer enn tre fravær uten gyldig grunn i ett semester fører til en samtale med dirigenten.
|
||||
|
||||
## Turen til København i juni 2027
|
||||
|
||||
Hovedkorpset og Drill- og slaggruppa reiser til Nordisk ungdomskorpstreff 17.–21. juni 2027.
|
||||
|
||||
Aspirantkorpset deltar ikke på turen.
|
||||
|
||||
Alle musikanter som reiser, må ha fylt 11 år innen 1. juni 2027.
|
||||
|
||||
Påmeldingen er bindende, og fristen er 1. desember 2026.
|
||||
|
||||
Egenandelen er 3 900 kroner, og familier med barn på turen tar tre ekstra dugnadsvakter.
|
||||
|
||||
## Neste steg for nye familier
|
||||
|
||||
Innmeldingsskjemaet må være levert innen 20. september 2026.
|
||||
|
||||
Låneavtalen for instrument signeres før første øvelse.
|
||||
|
||||
Første øvelse for nye aspiranter er mandag 7. september 2026 kl. 16:30.
|
||||
|
||||
Spørsmål om musikk går til dirigenten, og spørsmål om penger og regler går til foreldreutvalget.
|
||||
|
|
@ -0,0 +1,81 @@
|
|||
#### Instrumenter per 01.10.2026
|
||||
|
||||
A: Instrument-ID | B: Instrument | C: Avdeling | D: Innkjøpsår | E: Verdi (kr) | F: Status | G: Utlånt til | H: Siste service (dato) | I: Merknad
|
||||
A: TS-01 | B: Kornett | C: Hovedkorpset | D: 2009 | E: 6500 | F: Utlånt | G: Nora B. | H: 12.07.2026
|
||||
A: TS-02 | B: Kornett | C: Hovedkorpset | D: 2012 | E: 7200 | F: Utlånt | G: Elias K. | H: 12.07.2026 | I: Bulk i munnstykkerøret reparert våren 2026
|
||||
A: TS-03 | B: Kornett | C: Aspirantkorpset | D: 2015 | E: 7800 | F: Utlånt | G: Ida M. | H: 12.07.2026 | I: Aspirant, første år uten leie
|
||||
A: TS-04 | B: Kornett | C: Aspirantkorpset | D: 2018 | E: 8400 | F: Utlånt | G: Filip H. | H: 12.07.2026 | I: Aspirant, første år uten leie
|
||||
A: TS-05 | B: Kornett | C: Aspirantkorpset | D: 2021 | E: 9100 | F: På lager | H: 12.07.2026 | I: Reserve for nye aspiranter
|
||||
A: TS-06 | B: Trompet | C: Hovedkorpset | D: 2016 | E: 9800 | F: Utlånt | G: Sander L. | H: 12.07.2026
|
||||
A: TS-07 | B: Trompet | C: Hovedkorpset | D: 2003 | E: 5200 | F: Til service | H: 02.10.2026 | I: Ventil nr. 2 henger, ventet tilbake uke 43
|
||||
A: AH-01 | B: Althorn | C: Hovedkorpset | D: 2006 | E: 11500 | F: Utlånt | G: Maja S. | H: 12.07.2026
|
||||
A: AH-02 | B: Althorn | C: Hovedkorpset | D: 2011 | E: 12800 | F: Utlånt | G: Tiril Ø. | H: 12.07.2026
|
||||
A: AH-03 | B: Althorn | C: Hovedkorpset | D: 2019 | E: 14200 | F: Utlånt | G: Johannes R. | H: 12.07.2026 | I: Gikk over fra Aspirantkorpset august 2026
|
||||
A: AH-04 | B: Althorn | C: Hovedkorpset | D: 2026 | E: 16900 | F: Utlånt | G: Selma A. | I: Kjøpt august 2026 etter vedtak på årsmøtet
|
||||
A: BA-01 | B: Baryton | C: Hovedkorpset | D: 2008 | E: 15400 | F: Utlånt | G: Oskar N. | H: 12.07.2026
|
||||
A: BA-02 | B: Baryton | C: Hovedkorpset | D: 2014 | E: 17100 | F: På lager | H: 12.07.2026 | I: Ledig fra høsten 2026
|
||||
A: EU-01 | B: Eufonium | C: Hovedkorpset | D: 2010 | E: 21500 | F: Utlånt | G: Linnea F. | H: 12.07.2026
|
||||
A: EU-02 | B: Eufonium | C: Hovedkorpset | D: 2017 | E: 24300 | F: Utlånt | G: Mathias J. | H: 12.07.2026
|
||||
A: TR-01 | B: Trombone | C: Hovedkorpset | D: 2005 | E: 7600 | F: Utlånt | G: Ella V. | H: 12.07.2026
|
||||
A: TR-02 | B: Trombone | C: Hovedkorpset | D: 2013 | E: 8900 | F: Utlånt | G: Henrik G. | H: 12.07.2026
|
||||
A: TR-03 | B: Trombone | C: Aspirantkorpset | D: 2020 | E: 9500 | F: Utlånt | G: Vilde T. | H: 12.07.2026 | I: Aspirant, andre år, betaler leie
|
||||
A: TU-01 | B: Tuba | C: Hovedkorpset | D: 1998 | E: 19000 | F: Utlånt | G: Aksel D. | H: 12.07.2026 | I: Brukes bare på øvelser, står i storsalen
|
||||
A: TU-02 | B: Tuba | C: Hovedkorpset | D: 2026 | E: 27900 | F: Utlånt | G: Emil W. | H: 12.07.2026 | I: Kjøpt mars 2026, erstatter tuba fra 1984
|
||||
A: VH-01 | B: Valthorn | C: Hovedkorpset | D: 2007 | E: 18800 | F: Utlånt | G: Sofie E. | H: 12.07.2026
|
||||
A: VH-02 | B: Valthorn | C: Hovedkorpset | D: 2015 | E: 21000 | F: På lager | H: 12.07.2026
|
||||
A: KL-01 | B: Klarinett | C: Hovedkorpset | D: 2004 | E: 4900 | F: Utlånt | G: Thea P. | H: 12.07.2026 | I: Var til service fem uker våren 2026
|
||||
A: KL-02 | B: Klarinett | C: Hovedkorpset | D: 2012 | E: 6100 | F: Utlånt | G: Leah C. | H: 12.07.2026 | I: Var til service seks uker våren 2026
|
||||
A: KL-03 | B: Klarinett | C: Aspirantkorpset | D: 2019 | E: 6800 | F: Utlånt | G: Mia K. | H: 12.07.2026 | I: Aspirant, første år uten leie
|
||||
A: KL-04 | B: Klarinett | C: Aspirantkorpset | D: 2022 | E: 7300 | F: Utlånt | G: Noah B. | H: 12.07.2026 | I: Aspirant, første år uten leie
|
||||
A: FL-01 | B: Fløyte | C: Hovedkorpset | D: 2010 | E: 5600 | F: Utlånt | G: Ingrid H. | H: 12.07.2026
|
||||
A: FL-02 | B: Fløyte | C: Aspirantkorpset | D: 2023 | E: 6900 | F: Utlånt | G: Astrid L. | H: 12.07.2026 | I: Aspirant, første år uten leie
|
||||
A: FL-03 | B: Fløyte | C: Hovedkorpset | D: 2014 | E: 6200 | F: På lager | H: 12.07.2026 | I: Ledig, lånes ut til ny musikant ved behov
|
||||
A: AS-01 | B: Altsaksofon | C: Hovedkorpset | D: 2011 | E: 13400 | F: Utlånt | G: Jonas M. | H: 12.07.2026
|
||||
A: AS-02 | B: Altsaksofon | C: Hovedkorpset | D: 2018 | E: 15800 | F: Til service | H: 28.09.2026 | I: Ny pute på G-klaff, ventet tilbake uke 42
|
||||
A: TSX-01 | B: Tenorsaksofon | C: Hovedkorpset | D: 2016 | E: 19600 | F: Utlånt | G: Kasper O. | H: 12.07.2026
|
||||
|
||||
A: Instrument-ID | B: Instrument | C: Avdeling | D: Innkjøpsår | E: Verdi (kr) | F: Status | G: Utlånt til | H: Siste service (dato) | I: Merknad
|
||||
A: SL-01 | B: Marsjtromme | C: Drill- og slaggruppa | D: 2014 | E: 4800 | F: På lager | H: 12.07.2026 | I: Hjemlån bare før konkurranse, med tillatelse
|
||||
A: SL-02 | B: Marsjtromme | C: Drill- og slaggruppa | D: 2014 | E: 4800 | F: På lager | H: 12.07.2026 | I: Hjemlån bare før konkurranse, med tillatelse
|
||||
A: SL-03 | B: Marsjtromme | C: Drill- og slaggruppa | D: 2020 | E: 5900 | F: På lager | H: 12.07.2026 | I: Nytt skinn juli 2026
|
||||
A: SL-04 | B: Stortromme | C: Drill- og slaggruppa | D: 2012 | E: 8700 | F: På lager | H: 12.07.2026 | I: Skal ikke tas med hjem
|
||||
A: SL-05 | B: Cymbaler (par) | C: Drill- og slaggruppa | D: 2017 | E: 3600 | F: På lager | H: 12.07.2026
|
||||
A: SL-06 | B: Pauker (sett med 2) | C: Hovedkorpset | D: 2002 | E: 42000 | F: På lager | H: 12.07.2026 | I: Står fast i storsalen, skal ikke flyttes uten instrumentansvarlig
|
||||
A: SL-07 | B: Xylofon | C: Hovedkorpset | D: 2009 | E: 23500 | F: På lager | H: 12.07.2026 | I: Skal ikke tas med hjem
|
||||
A: SL-08 | B: Glockenspiel | C: Aspirantkorpset | D: 2021 | E: 4200 | F: På lager | H: 12.07.2026 | I: Brukes på samspill mandager
|
||||
|
||||
#### Reisebudsjett v2
|
||||
|
||||
A: Post | B: Type | C: Antall | D: Pris per enhet (kr) | E: Beløp (kr) | F: Ansvarlig | G: Merknad
|
||||
A: Reisekostnad tur/retur | B: Utgift | C: 45 | D: 1450 | E: 65250 | F: Kari Enerhaug | G: 38 musikanter og 7 voksne
|
||||
A: Overnatting Herberget Kastanjegården | B: Utgift | C: 180 | D: 310 | E: 55800 | F: Kari Enerhaug | G: 45 personer x 4 netter, frokost inkludert
|
||||
A: Lunsj | B: Utgift | C: 180 | D: 120 | E: 21600 | F: Kari Enerhaug | G: 4 dager
|
||||
A: Middag | B: Utgift | C: 180 | D: 190 | E: 34200 | F: Kari Enerhaug | G: 4 dager
|
||||
A: Festivalavgift | B: Utgift | C: 38 | D: 650 | E: 24700 | F: Ingeborg Haugstøl | G: Økt fra 550 kr i oktober 2026; voksne ledere betaler ikke
|
||||
A: Forsendelse av slagverk i kasser | B: Utgift | C: 1 | D: 18500 | E: 18500 | F: Siri Nyhagen | G: Sendes to uker før avreise
|
||||
A: Fellesaktivitet badeland | B: Utgift | C: 45 | D: 220 | E: 9900 | F: Kari Enerhaug | G: Søndag 20.06.2027
|
||||
A: Felles t-skjorter | B: Utgift | C: 45 | D: 160 | E: 7200 | F: Amina Rasoul | G: Bestilles innen 1. april 2027
|
||||
A: Tilleggsforsikring instrumenter | B: Utgift | C: 1 | D: 4800 | E: 4800 | F: Mona Fjellstad | G: Gjelder bare korpsets instrumenter
|
||||
A: Diett dirigent og instruktør | B: Utgift | C: 12 | D: 400 | E: 4800 | F: Mona Fjellstad | G: 3 personer x 4 dager
|
||||
A: Uforutsett | B: Utgift | C: 1 | D: 10000 | E: 10000 | F: Henrik Bråtveit | G: Kan bare brukes etter vedtak i foreldreutvalget
|
||||
A: Sum utgifter | B: Utgift | E: 256750 | F: Mona Fjellstad | G: Versjon 2
|
||||
A: Egenandel musikanter | B: Inntekt | C: 38 | D: 3900 | E: 148200 | F: Mona Fjellstad | G: Tre rater à 1 300 kr; ingen søskenmoderasjon
|
||||
A: Dugnader til reisen | B: Inntekt | C: 1 | D: 70000 | E: 70000 | F: Amina Rasoul | G: Se arket Dugnader 2026-2027
|
||||
A: Fra reisefondet | B: Inntekt | C: 1 | D: 25000 | E: 25000 | F: Mona Fjellstad | G: Saldo 41 000 kr per 31.05.2026
|
||||
A: Støttekonsert 07.03.2027 | B: Inntekt | C: 1 | D: 14000 | E: 14000 | F: Ingeborg Haugstøl | G: Billetter; kafé er ført under dugnader
|
||||
A: Sum inntekter | B: Inntekt | E: 257200 | F: Mona Fjellstad | G: Uendret fra versjon 1
|
||||
A: Resultat | B: Resultat | E: 450 | F: Mona Fjellstad | G: Versjon 2 revidert 12.10.2026; var 4 250 kr i versjon 1 fordi festivalavgiften økte med 100 kr per musikant
|
||||
|
||||
#### Dugnader 2026-2027
|
||||
|
||||
A: Dato | B: Dugnad | C: Antall vakter | D: Forventet inntekt (kr) | E: Andel til reisen (%) | F: Beløp til reisen (kr) | G: Ansvarlig | H: Merknad
|
||||
A: 26.09.2026 | B: Loppemarked i gymsalen | C: 24 | D: 32000 | E: 50 | F: 16000 | G: Amina Rasoul | H: Resten går til driften
|
||||
A: 22.11.2026 | B: Kafé ved høstkonserten | C: 8 | D: 6500 | E: 100 | F: 6500 | G: Øystein Lunde | H: Turvakt
|
||||
A: 05.12.2026 | B: Juletresalg på skoleplassen | C: 18 | D: 21000 | E: 100 | F: 21000 | G: Kari Enerhaug | H: Turvakt; bare voksne kan stå i salget
|
||||
A: 13.12.2026 | B: Kafé ved julekonserten | C: 8 | D: 5500 | E: 0 | F: 0 | G: Øystein Lunde | H: Ordinær vakt, alt går til driften
|
||||
A: 09.01.2027 | B: Varetelling hos lokal dagligvarebutikk | C: 12 | D: 9000 | E: 100 | F: 9000 | G: Henrik Bråtveit | H: Turvakt; deltakere må være minst 16 år
|
||||
A: 07.03.2027 | B: Kafé ved støttekonserten | C: 6 | D: 4000 | E: 100 | F: 4000 | G: Kari Enerhaug | H: Turvakt
|
||||
A: 17.04.2027 | B: Vårrydding for velforeningen | C: 15 | D: 4500 | E: 100 | F: 4500 | G: Amina Rasoul | H: Turvakt; fast honorar
|
||||
A: 25.04.2027 | B: Kafé ved vårkonserten | C: 8 | D: 6000 | E: 0 | F: 0 | G: Øystein Lunde | H: Ordinær vakt
|
||||
A: 17.05.2027 | B: Kiosk på skoleplassen 17. mai | C: 20 | D: 18000 | E: 50 | F: 9000 | G: Mona Fjellstad | H: Ordinær vakt og turvakt
|
||||
A: 22.05.2027 | B: Plantesalg | C: 10 | D: 7000 | E: 0 | F: 0 | G: Siri Nyhagen | H: Ordinær vakt, alt går til driften
|
||||
|
|
@ -0,0 +1,9 @@
|
|||
The Kettlewick Project: Annual Report 2025 The Kettlewick Project: Annual Report 2025 Kettlewick is a command-line tool for organising and searching personal notes stored as plain text files. It has been developed by volunteers since its first public release in April 2019. This report covers the calendar year 2025, from 1 January to 31 December, and was prepared by the maintainers for the annual maintainers' meeting on 14 March 2026. Figures are as of 31 December 2025 unless another date is given. The report was finalised on 2 February 2026, so it does not include any decision taken at the meeting itself. 1. The year in summary 2025 was the busiest year in the project's history. We shipped 11 releases, compared with 9 in 2024, and merged 1,086 pull requests, up from 812 the year before. The number of people with at least one merged pull request rose from 64 to 87. Two serious incidents, a search index corruption bug in June and an expired release signing key in September, led directly to rule changes that take effect in 2026. 11 releases: 4 minor releases and 7 patch releases (2024: 1 major, 2 minor and 6 patch releases). 1,086 merged pull requests from 87 distinct contributors, 31 of whom had their first pull request merged in 2025. 1,138 issues opened and 1,021 closed; the median time to a first response fell from 52 hours in 2024 to 31 hours. 2,416 automated tests at year end, up from 1,874, with line coverage of 84.2% (end of 2024: 79.5%). Median search latency on the 100,000-note benchmark corpus fell from 131 ms in release 2.3.0 to 72 ms in release 2.6.0. 2 security advisories and 5 code of conduct reports, all closed by the end of the year. 2. Releases Minor releases follow a fixed 13-week cadence and are always published on a Tuesday. The planned dates are set a year in advance and do not move when a release is late; a delayed release simply shortens the following cycle. The project skipped the December 2024 slot so that no release would fall in the holiday period, and the cadence restarted on 18 February 2025. Patch releases are normally published on a Tuesday two weeks after the minor release they fix, unless a fix is urgent. In 2025 the four minor releases were planned for 18 February, 20 May, 19 August and 18 November. Releases 2.3.0 and 2.4.0 shipped on their planned dates. Releases 2.5.0 and 2.6.0
|
||||
|
||||
each shipped 7 days late, on 26 August and 25 November, because testing of the first release candidate found blocking bugs and a second release candidate was needed. Each minor release had its own release manager, except that Tobias Wren managed both 2.3.0 and 2.5.0. Version Release date Type Release candidates Merged pull requests Release manager 2.3.0 18 February 2025 Minor 1 241 Tobias Wren 2.4.0 20 May 2025 Minor 1 276 Priya Halloran 2.5.0 26 August 2025 Minor 2 259 Tobias Wren 2.6.0 25 November 2025 Minor 2 228 Dagny Oyelaran The seven patch releases were 2.3.1, 2.4.1, 2.4.2, 2.5.1, 2.5.2, 2.6.1 and 2.6.2. Two of them were unscheduled: 2.4.2 was published on Saturday 7 June to fix the index corruption bug described in section 5, and 2.5.2 was the security release for advisory KW-SA-2025-02. Release 2.5.1 was planned for 9 September but shipped on 12 September, 3 days late, because the release signing key had expired the day before the planned date. Because two of the four minor releases needed a second release candidate, the maintainers voted in December 2025 to lengthen the feature freeze from 7 days to 14 days before each minor release. In 2025 the translation string freeze fell on the same day as the feature freeze; from release 2.7.0 it falls 10 days before release, so that text changed by fixes in the first four days of the freeze can still be translated. Both changes apply from release 2.7.0, planned for 17 February 2026, whose release manager is Saoirse Lindqvist. 3. People: contributors, maintainers and conduct 87 people had at least one pull request merged in 2025, against 64 in 2024. Of these, 31 were first-time contributors. Based on the optional location given by 71 of the 87 contributors, they live in 23 countries. Documentation and translation changes made up 27% of merged pull requests, compared with 22% in 2024. The project started the year with 7 maintainers and ended it with 9. Saoirse Lindqvist became a maintainer on 8 April 2025 with responsibility for testing and continuous integration, and Ngozi Barraclough became a maintainer on 14 October 2025 with responsibility for Windows support. No maintainer stepped down. Commit access is offered to contributors who have had at least 10 pull requests merged over a period of at least 6 months and who are nominated by 2 existing maintainers; a nomination is accepted unless a maintainer objects within 7 days. The Seedling mentoring programme ran two cohorts of 12 weeks each. Applicants must be at least 16 years old, and each cohort takes at most 6 mentees, each paired with one maintainer. A maintainer mentors only one person at a time.
|
||||
|
||||
Cohort Dates Applications Mentees accepted Mentees completed Spring 2025 3 March to 25 May 2025 19 6 5 Autumn 2025 1 September to 23 November 2025 14 5 5 One spring mentee withdrew in week 4 because of school exams. Of the 10 mentees who completed the programme, 7 had at least one pull request merged by 31 December 2025. Code of conduct reports are handled by a committee of three: Ruth Abernethy (chair), Dagny Oyelaran and Hollis Fenwick, who is a contributor but not a maintainer. The committee received 5 reports in 2025, compared with 2 in 2024. Two reports led to no action, two led to a written warning and one led to a temporary ban of 30 days from all project spaces. All 5 reports were acknowledged within the 48-hour target. Four were decided within the 14-day target; the fifth took 19 days because Dagny Oyelaran had taken part in the discussion concerned and had to step aside, and a substitute had to be found. As a result, Felix Ondrasek was named standing substitute in January 2026. 4. Quality: tests, benchmarks and triage The test suite grew from 1,874 tests at the end of 2024 to 2,416 at the end of 2025, an increase of 542. Line coverage rose from 79.5% to 84.2%. A full run takes 6 minutes 40 seconds on the Linux runner and 9 minutes 12 seconds on the macOS runner. Every pull request is tested on Linux and macOS. Windows is tested only in the nightly run at 03:00 UTC, because the project has a single Windows runner. The coverage gate fails a pull request if total coverage drops by more than 0.5 percentage points or falls below 82%. During 2025, 14 flaky tests were moved to quarantine. A quarantined test must be fixed or deleted within 21 days; 12 were fixed and 2 were deleted, and none stayed in quarantine longer than the limit. Benchmarks run on a dedicated machine with 8 cores and 32 GB of memory, every Sunday at 02:00 UTC and on every release candidate. There are two corpora: a small corpus of 10,000 notes (48 MB) and a large corpus of 100,000 notes (471 MB). The table shows the large corpus for each minor release since 2.2.0. Release Index build time (s) Median search latency (ms) 95th percentile latency (ms) Peak memory (MB) 2.2.0 171.2 148 412 1,240 2.3.0 160.4 131 377 1,205 2.4.0 139.7 112 330 1,318 2.5.0 142.9 104 298 1,104 2.6.0 108.5 72 214 1,061 The largest single gain came in 2.6.0, where a new trigram pre-filter cut the median search latency on the large corpus by about 31% compared with 2.5.0. Peak memory rose by 9.4%
|
||||
|
||||
in 2.4.0 because of the new tag cache, just below the 10% regression threshold; the cache was made smaller in 2.5.0, which brought peak memory down to 1,104 MB. A pull request that makes median search latency or index build time more than 5% slower, or raises peak memory by more than 10%, is labelled perf-regression and may be merged only with the approval of the performance lead, Mateus Kowalczyk. In 2025, 6 pull requests received the label: 4 were merged after approval and 2 were closed. 1,138 issues were opened in 2025: 602 bug reports, 341 feature requests and 195 questions and other issues. 1,021 issues were closed, so the number of open issues rose from 412 at the start of the year to 529 at the end. The project aims to give every new issue a first response within 72 hours and a label within 7 days. Until July, triage was done by whoever had time. A weekly triage rota began on Monday 4 August 2025, with one maintainer on duty from Monday to Sunday; it covered 22 weeks by the end of the year. Before the rota, the monthly median first response ranged from 27 to 44 hours. From August to November it stayed between 24 and 29 hours; it rose to 39 hours in December, when fewer people were active over the holidays. In total, 64 issues were still unlabelled after 7 days, 50 of them in January to July. 5. Incidents and security Index corruption in 2.4.1. Release 2.4.1, published on 3 June 2025, contained a bug in the store module. Users with more than 50,000 notes who ran an incremental reindex could end up with a corrupted search index. Note files themselves were never damaged, because the index is stored separately from the notes. The first report arrived on 4 June, and 37 issue reports were filed in total. The fix shipped in 2.4.2 on 7 June, 4 days after 2.4.1; until then users were told to run a full reindex. The change that caused the bug had been approved by one maintainer, and no test used a corpus larger than 20,000 notes. As a result, changes to the index and store modules need approval from two maintainers from 1 January 2026, and a test with a 60,000-note corpus was added to the nightly run. Expired signing key. The release signing key, created on 8 September 2023 with a lifetime of 2 years, expired on 8 September 2025. This was discovered while signing 2.5.1 and delayed that release by 3 days. The new key, created on 11 September 2025, has a lifetime of 3 years and expires on 11 September 2028. A reminder issue is now opened automatically 90 days before the key expires. Continuous integration outage. The self-hosted Linux runners were offline from 14 to 22 October 2025, 9 days in all, after a disk failure. During the outage only documentation-only pull requests were merged, and 41 other pull requests waited until the runners were back. Two security advisories were published in 2025. Both were fixed within the 30-day embargo set by the security policy. Advisory Problem Reported Fixed in Days to fix KW-SA-2025- 01 Crash when a note contains 19 February 2025 2.3.1 (4 March 2025) 13
|
||||
|
||||
tags nested more than 500 levels deep KW-SA-2025- 02 Import of a crafted archive could write files outside the notes folder 2 September 2025 2.5.2 (30 September 2025) 28 The reporter of KW-SA-2025-01 is credited by name in the advisory. The reporter of KW- SA-2025-02 asked to remain anonymous. 6. Looking ahead Since 2019 the founder, Imogen Varga, has served as lead maintainer with the final say when maintainers cannot agree. She used this power 3 times in 2025. She has herself proposed that the role be replaced by an elected steering council. The proposal will be put to a vote at the annual meeting on 14 March 2026 and needs the support of two thirds of the maintainers, that is at least 6 of 9. The steering council would have 5 seats with terms of 2 years; in the first election 3 seats would be for 2 years and 2 seats for 1 year, so that later elections are staggered. Only maintainers who have held the role for at least 12 months could stand as candidates. Anyone with at least 5 merged pull requests in the 12 months before voting opens could vote. Decisions would need a quorum of 4 of the 5 members and a simple majority, except changes to the licence or the code of conduct, which would need 4 votes. If the proposal is approved, nominations would run from 1 to 21 April 2026, voting from 1 to 14 May 2026, and the council would take office on 1 July 2026. The lead maintainer role would end on 30 June 2026. The release plan for 2026 keeps the 13-week cadence: 2.7.0 on 17 February, 2.8.0 on 19 May, 2.9.0 on 18 August and 3.0.0 on 17 November 2026. Release 3.0.0 will remove support for configuration format version 1, which has been deprecated since 2.2.0, and will raise the minimum Python version from 3.9 to 3.10. The maintainers propose four goals for 2026: line coverage of at least 86% by the end of the year, a median first response to issues below 24 hours, 100 distinct contributors with a merged pull request, and Windows tests on every pull request by release 2.9.0.
|
||||
|
|
@ -0,0 +1,237 @@
|
|||
# Kettlewick Contributor Guide and Release Process
|
||||
|
||||
# Kettlewick Contributor Guide and Release Process
|
||||
|
||||
Version 4.2, effective 1 April 2026. Approved by the maintainers on 24 March 2026. This version replaces version 4.1 of 6 January 2026.
|
||||
|
||||
## 1. About this guide
|
||||
|
||||
This guide applies to everyone who contributes code, documentation, translations or reviews to the Kettlewick project. Maintainers must follow it in full. Contributors without commit access need to read sections 2 to 5 before opening their first pull request. Where this guide and an older wiki page disagree, this guide applies.
|
||||
|
||||
Changes in recent versions:
|
||||
|
||||
Version 4.1 (6 January 2026): changes to the index and store modules need two approvals; the feature freeze is 14 days instead of 7; the string freeze is 10 days before release.
|
||||
|
||||
Version 4.2 (1 April 2026): new section 8 on the steering council, after the governance proposal was approved on 14 March 2026; Felix Ondrasek added as standing substitute on the conduct committee.
|
||||
|
||||
## 2. Before you start
|
||||
|
||||
Every commit must carry a Signed-off-by line confirming that you have the right to submit the work under the project licence. Pull requests with unsigned commits are not merged.
|
||||
|
||||
Open an issue before you start on a large change. A change that touches more than 400 lines outside the tests, or that adds a new command, must have an issue with the label accepted before the pull request is opened.
|
||||
|
||||
Issues labelled good-first-issue are reserved for first-time contributors for 14 days after they are labelled.
|
||||
|
||||
Do not report security problems in public issues. Use the private security address described in section 6.
|
||||
|
||||
The Seedling mentoring programme pairs a new contributor with one maintainer for 12 weeks. Applicants must be at least 16 years old. Each cohort takes at most 6 mentees, and a maintainer may mentor only one mentee at a time. Applications close 4 weeks before a cohort starts.
|
||||
|
||||
Cohort
|
||||
|
||||
Application deadline
|
||||
|
||||
Start
|
||||
|
||||
End
|
||||
|
||||
Spring 2026
|
||||
|
||||
2 February 2026
|
||||
|
||||
2 March 2026
|
||||
|
||||
24 May 2026
|
||||
|
||||
Autumn 2026
|
||||
|
||||
3 August 2026
|
||||
|
||||
31 August 2026
|
||||
|
||||
22 November 2026
|
||||
|
||||
Commit access is offered to contributors who have had at least 10 pull requests merged over a period of at least 6 months and who are nominated by 2 maintainers. The nomination is accepted unless a maintainer objects within 7 days.
|
||||
|
||||
## 3. Pull requests and review
|
||||
|
||||
The number of approvals a pull request needs depends on what it changes. If a pull request touches more than one area, the strictest rule applies.
|
||||
|
||||
Area
|
||||
|
||||
Approvals needed
|
||||
|
||||
Who may approve
|
||||
|
||||
Notes
|
||||
|
||||
Core modules (index and store)
|
||||
|
||||
2
|
||||
|
||||
Maintainers
|
||||
|
||||
Since 1 January 2026; before that 1 approval
|
||||
|
||||
Command-line interface and other code
|
||||
|
||||
1
|
||||
|
||||
Maintainers
|
||||
|
||||
Plugins
|
||||
|
||||
1
|
||||
|
||||
Maintainers or plugin reviewers
|
||||
|
||||
Documentation only
|
||||
|
||||
1
|
||||
|
||||
Any maintainer
|
||||
|
||||
Must stay open for at least 48 hours before merging
|
||||
|
||||
Translations
|
||||
|
||||
1
|
||||
|
||||
The translation coordinator for that language, or any maintainer
|
||||
|
||||
Version bump before a release
|
||||
|
||||
1
|
||||
|
||||
Any maintainer other than the release manager
|
||||
|
||||
Exception to the self-merge rule below
|
||||
|
||||
A maintainer must not merge their own pull request. The only exception is the release manager, who may merge the version bump pull request after one approval from another maintainer.
|
||||
|
||||
Every pull request must pass the tests on Linux and macOS before it is merged.
|
||||
|
||||
Pull requests are merged by squash merge only.
|
||||
|
||||
Reviewers aim to give a first review within 5 days. If you have had no review after 5 days, you may ask in the review channel.
|
||||
|
||||
A pull request with no activity from its author for 60 days is closed with the label stale. The author may reopen it at any time.
|
||||
|
||||
## 4. Tests and benchmarks
|
||||
|
||||
Every bug fix must include a test that fails without the fix.
|
||||
|
||||
The coverage gate fails a pull request if total line coverage drops by more than 0.5 percentage points or falls below 82%.
|
||||
|
||||
Windows is tested only in the nightly run at 03:00 UTC. If the nightly Windows run fails because of your merged pull request, you are expected to help fix it within 7 days; otherwise the pull request may be reverted.
|
||||
|
||||
A flaky test may be moved to quarantine for at most 21 days, after which it must be fixed or deleted. No more than 10 tests may be in quarantine at the same time.
|
||||
|
||||
Do not add tests that need network access; they are skipped in the nightly run and do not count towards coverage.
|
||||
|
||||
Benchmarks run every Sunday at 02:00 UTC and on every release candidate, using a small corpus of 10,000 notes and a large corpus of 100,000 notes. You can ask for a benchmark run on your pull request by commenting /bench. Each pull request may have at most 3 requested runs, because the benchmark machine is shared by the whole project.
|
||||
|
||||
A pull request that makes median search latency or index build time more than 5% slower, or raises peak memory by more than 10%, on either corpus, is labelled perf-regression. It may be merged only with the approval of the performance lead, Mateus Kowalczyk, even if it already has the approvals required in section 3.
|
||||
|
||||
## 5. Issue triage
|
||||
|
||||
Triage is done by a weekly rota. The maintainer on duty covers Monday 00:00 UTC to Sunday 23:59 UTC. Every maintainer takes at least one week per quarter, and swaps must be recorded in the rota file before the week starts.
|
||||
|
||||
Every new issue should get a first response within 72 hours and a label within 7 days.
|
||||
|
||||
An issue labelled needs-info is closed if the reporter has not replied within 30 days. It can be reopened when the information arrives.
|
||||
|
||||
Duplicates are closed with a link to the older issue, even if the newer issue has more detail; the detail is copied across.
|
||||
|
||||
Feature requests are labelled feature and are not given a milestone until a maintainer agrees to review the work.
|
||||
|
||||
The person on duty must not close an issue they reported themselves.
|
||||
|
||||
## 6. Release process
|
||||
|
||||
Minor releases follow a fixed 13-week cadence and are published on a Tuesday. The planned dates for 2026 are 17 February (2.7.0), 19 May (2.8.0), 18 August (2.9.0) and 17 November (3.0.0). A late release does not move the next planned date. Each minor release has a release manager, who must have been a maintainer for at least 6 months. No one may manage two minor releases in a row. The release manager for 2.8.0 is Felix Ondrasek.
|
||||
|
||||
Step
|
||||
|
||||
When
|
||||
|
||||
Who
|
||||
|
||||
What it means
|
||||
|
||||
Feature freeze
|
||||
|
||||
14 days before release (a Tuesday)
|
||||
|
||||
Release manager
|
||||
|
||||
Only bug fixes and documentation may be merged into the release branch
|
||||
|
||||
String freeze
|
||||
|
||||
10 days before release (a Saturday)
|
||||
|
||||
Release manager and translation coordinators
|
||||
|
||||
No changes to user-visible text
|
||||
|
||||
First release candidate
|
||||
|
||||
7 days before release (a Tuesday)
|
||||
|
||||
Release manager
|
||||
|
||||
Published for testing; benchmarks run automatically
|
||||
|
||||
Final release
|
||||
|
||||
The planned Tuesday
|
||||
|
||||
Release manager
|
||||
|
||||
Only if no bug labelled blocker is open
|
||||
|
||||
Patch release
|
||||
|
||||
Normally 2 weeks after the minor release
|
||||
|
||||
Release manager of that minor release
|
||||
|
||||
Bug fixes only; no release candidate
|
||||
|
||||
A second release candidate is required if any change other than a documentation change is merged after the first release candidate.
|
||||
|
||||
The feature freeze was 7 days until the end of 2025. It was extended to 14 days from release 2.7.0 because two of the four minor releases in 2025 needed a second release candidate.
|
||||
|
||||
Releases are signed with the project key. The current key was created on 11 September 2025, has a lifetime of 3 years and expires on 11 September 2028. A reminder issue is opened automatically 90 days before expiry.
|
||||
|
||||
Security problems are reported to the private security address. The security team acknowledges a report within 3 working days. A fix is released and the advisory published within 30 days of the report, unless the reporter agrees to a longer embargo, which may never exceed 90 days. Security releases may be published on any day of the week and do not need a release candidate. Reporters are credited by name only if they agree.
|
||||
|
||||
## 7. Code of conduct reports
|
||||
|
||||
The code of conduct applies in all project spaces: the repository, the issue tracker, the chat channels and the annual meeting. Reports go to the conduct committee: Ruth Abernethy (chair), Dagny Oyelaran and Hollis Fenwick. Felix Ondrasek is the standing substitute.
|
||||
|
||||
The committee acknowledges a report within 48 hours and decides within 14 days.
|
||||
|
||||
A committee member who is involved in a report must step aside, and the standing substitute takes their place.
|
||||
|
||||
Possible outcomes are no action, a written warning, a temporary ban of up to 90 days from all project spaces, or a permanent ban.
|
||||
|
||||
A permanent ban needs a unanimous decision of the committee. From 1 July 2026 the person banned may ask the steering council to review it.
|
||||
|
||||
The reporter is told the outcome but not necessarily the details. Reports must not be discussed in public channels.
|
||||
|
||||
## 8. Governance from 1 July 2026
|
||||
|
||||
On 14 March 2026 the maintainers approved the proposal to replace the lead maintainer with an elected steering council. Eight maintainers voted for, none voted against and one abstained; at least 6 of 9 votes were needed. Until 30 June 2026 the lead maintainer, Imogen Varga, keeps the final say when maintainers cannot agree.
|
||||
|
||||
The steering council has 5 seats. In the first election 3 seats are for 2 years and 2 seats for 1 year; after that all terms are 2 years.
|
||||
|
||||
Candidates must have been maintainers for at least 12 months when nominations close.
|
||||
|
||||
Anyone with at least 5 merged pull requests in the 12 months before voting opens may vote.
|
||||
|
||||
No more than 2 council members may work for the same organisation.
|
||||
|
||||
The council decides by simple majority with a quorum of 4 members. Changes to the licence, the code of conduct or this section need 4 votes.
|
||||
|
||||
Nominations run from 1 to 21 April 2026 and voting from 1 to 14 May 2026. The council takes office on 1 July 2026.
|
||||
|
|
@ -0,0 +1,175 @@
|
|||
## Kettlewick Annual Maintainers' Meeting 2026
|
||||
|
||||
## Kettlewick annual maintainers' meeting 2026
|
||||
|
||||
The meeting is held online on Saturday 14 March 2026 from 14:00 to 17:00 UTC.
|
||||
|
||||
All 9 maintainers are invited, and Ruth Abernethy chairs the meeting.
|
||||
|
||||
The vote on the steering council proposal is the last item on the agenda.
|
||||
|
||||
## Agenda
|
||||
|
||||
Review of 2025: releases, incidents, quality and people.
|
||||
|
||||
First experience with the longer feature freeze in release 2.7.0.
|
||||
|
||||
The steering council proposal, followed by questions and the vote.
|
||||
|
||||
Plans and goals for 2026.
|
||||
|
||||
## 2025 in numbers
|
||||
|
||||
We shipped 11 releases in 2025, two more than the 9 in 2024.
|
||||
|
||||
We merged 1,086 pull requests, up from 812 in 2024.
|
||||
|
||||
87 people had a pull request merged, and 31 of them were new contributors.
|
||||
|
||||
1,138 issues were opened and 1,021 were closed, so 529 issues were open at year end.
|
||||
|
||||
## Releases in 2025
|
||||
|
||||
Four minor releases followed the 13-week cadence, always on a Tuesday.
|
||||
|
||||
Releases 2.5.0 and 2.6.0 each shipped 7 days late because they needed a second release candidate.
|
||||
|
||||
Seven patch releases shipped, two of them unscheduled: 2.4.2 and the security release 2.5.2.
|
||||
|
||||
Tobias Wren managed 2.3.0 and 2.5.0, Priya Halloran managed 2.4.0 and Dagny Oyelaran managed 2.6.0.
|
||||
|
||||
## Release 2.7.0: the first with a 14-day freeze
|
||||
|
||||
Release 2.7.0 shipped on its planned date, Tuesday 17 February 2026.
|
||||
|
||||
It needed only one release candidate.
|
||||
|
||||
The feature freeze started on 3 February and the string freeze on 7 February.
|
||||
|
||||
Saoirse Lindqvist was release manager, and Felix Ondrasek manages 2.8.0 on 19 May.
|
||||
|
||||
## Incident: index corruption in 2.4.1
|
||||
|
||||
Users with more than 50,000 notes could get a corrupted index after an incremental reindex.
|
||||
|
||||
Note files were never damaged, because the index is stored separately.
|
||||
|
||||
37 issue reports came in, and the fix shipped in 2.4.2 four days later, on 7 June.
|
||||
|
||||
Since 1 January 2026, changes to the index and store modules need two approvals instead of one.
|
||||
|
||||
## Incidents: signing key and CI outage
|
||||
|
||||
The signing key expired on 8 September 2025 and delayed 2.5.1 by 3 days.
|
||||
|
||||
The new key lasts 3 years, and a reminder issue opens 90 days before it expires.
|
||||
|
||||
The Linux runners were offline for 9 days, from 14 to 22 October, after a disk failure.
|
||||
|
||||
During the outage only documentation-only pull requests were merged, and 41 others had to wait.
|
||||
|
||||
## Security advisories
|
||||
|
||||
KW-SA-2025-01, a crash on deeply nested tags, was reported on 19 February and fixed in 2.3.1 after 13 days.
|
||||
|
||||
KW-SA-2025-02, a path problem in import, was reported on 2 September and fixed in 2.5.2 after 28 days.
|
||||
|
||||
Both fixes stayed within the 30-day embargo.
|
||||
|
||||
## Test suite
|
||||
|
||||
The suite grew from 1,874 to 2,416 tests during 2025.
|
||||
|
||||
Line coverage rose from 79.5% to 84.2%.
|
||||
|
||||
A full run takes 6 minutes 40 seconds on Linux and 9 minutes 12 seconds on macOS.
|
||||
|
||||
14 flaky tests went into quarantine; 12 were fixed and 2 were deleted within the 21-day limit.
|
||||
|
||||
## Benchmarks
|
||||
|
||||
On the large corpus, median search latency fell from 104 ms in 2.5.0 to 72 ms in 2.6.0.
|
||||
|
||||
Since 2.2.0, median latency on the large corpus has fallen from 148 ms to 72 ms.
|
||||
|
||||
Peak memory rose 9.4% in 2.4.0, just under the 10% threshold, and fell again in 2.5.0.
|
||||
|
||||
Six pull requests were labelled perf-regression in 2025; four were merged after approval.
|
||||
|
||||
## Issue triage and the rota
|
||||
|
||||
The median first response fell from 52 hours in 2024 to 31 hours in 2025.
|
||||
|
||||
The weekly rota began on 4 August 2025 and covered 22 weeks by year end.
|
||||
|
||||
From August to November the monthly median stayed between 24 and 29 hours.
|
||||
|
||||
64 issues were unlabelled after 7 days, 50 of them before the rota began.
|
||||
|
||||
## Contributors and the Seedling programme
|
||||
|
||||
Saoirse Lindqvist joined as maintainer in April and Ngozi Barraclough in October 2025.
|
||||
|
||||
The two Seedling cohorts accepted 11 mentees, and 10 completed the 12 weeks.
|
||||
|
||||
7 of the 10 who completed had a pull request merged by the end of 2025.
|
||||
|
||||
The spring 2026 cohort started on 2 March with applications closed on 2 February.
|
||||
|
||||
## Code of conduct
|
||||
|
||||
The committee received 5 reports in 2025, compared with 2 in 2024.
|
||||
|
||||
The outcomes were 2 with no action, 2 written warnings and 1 temporary ban of 30 days.
|
||||
|
||||
One decision took 19 days instead of 14 because a committee member had to step aside.
|
||||
|
||||
Felix Ondrasek has been the standing substitute since January 2026.
|
||||
|
||||
## Proposal: a steering council
|
||||
|
||||
The lead maintainer role would be replaced by an elected council of 5 members.
|
||||
|
||||
Imogen Varga used the lead maintainer's final say 3 times in 2025.
|
||||
|
||||
Terms would be 2 years, with 2 of the first 5 seats lasting only 1 year.
|
||||
|
||||
No more than 2 council members could work for the same organisation.
|
||||
|
||||
## Proposal: who can stand and who can vote
|
||||
|
||||
Candidates must have been maintainers for at least 12 months when nominations close.
|
||||
|
||||
Eight of the nine current maintainers would be eligible to stand in the first election.
|
||||
|
||||
Anyone with at least 5 merged pull requests in the 12 months before voting opens could vote.
|
||||
|
||||
Changes to the licence or the code of conduct would need 4 of the 5 council votes.
|
||||
|
||||
## Proposal: timeline if approved
|
||||
|
||||
Nominations would run from 1 to 21 April 2026.
|
||||
|
||||
Voting would run from 1 to 14 May 2026.
|
||||
|
||||
The council would take office on 1 July 2026, and the lead maintainer role would end on 30 June.
|
||||
|
||||
Today the proposal needs at least 6 of the 9 maintainers to vote for it.
|
||||
|
||||
## Plans for 2026
|
||||
|
||||
Minor releases are planned for 19 May (2.8.0), 18 August (2.9.0) and 17 November (3.0.0).
|
||||
|
||||
Release 3.0.0 removes configuration format version 1 and requires Python 3.10 or newer.
|
||||
|
||||
Seedling autumn 2026 runs from 31 August to 22 November, with applications closing on 3 August.
|
||||
|
||||
## Goals for 2026
|
||||
|
||||
Reach line coverage of at least 86% by the end of 2026.
|
||||
|
||||
Bring the median first response to issues below 24 hours.
|
||||
|
||||
Reach 100 distinct contributors with a merged pull request.
|
||||
|
||||
Run Windows tests on every pull request by release 2.9.0.
|
||||
|
|
@ -0,0 +1,66 @@
|
|||
#### Release history
|
||||
|
||||
A: Version | B: Release date | C: Type | D: Planned date | E: Delay (days) | F: Release candidates (count) | G: Merged pull requests (count) | H: Release manager | I: Notes
|
||||
A: 2.0.0 | B: 2024-03-12 | C: Major | D: 2024-03-12 | E: 0 | F: 3 | G: 302 | H: Imogen Varga | I: New plugin interface; configuration format version 2 introduced
|
||||
A: 2.0.1 | B: 2024-03-26 | C: Patch | D: 2024-03-26 | E: 0 | F: 0 | G: 21 | H: Imogen Varga | I: Fixes for plugin loading
|
||||
A: 2.1.0 | B: 2024-06-11 | C: Minor | D: 2024-06-11 | E: 0 | F: 1 | G: 198 | H: Tobias Wren | I: Saved searches
|
||||
A: 2.1.1 | B: 2024-06-25 | C: Patch | D: 2024-06-25 | E: 0 | F: 0 | G: 15 | H: Tobias Wren
|
||||
A: 2.1.2 | B: 2024-07-16 | C: Patch | D: 2024-07-16 | E: 0 | F: 0 | G: 9 | H: Tobias Wren | I: Encoding fix for note titles
|
||||
A: 2.2.0 | B: 2024-09-10 | C: Minor | D: 2024-09-10 | E: 0 | F: 1 | G: 214 | H: Priya Halloran | I: Configuration format version 1 deprecated
|
||||
A: 2.2.1 | B: 2024-09-24 | C: Patch | D: 2024-09-24 | E: 0 | F: 0 | G: 17 | H: Priya Halloran
|
||||
A: 2.2.2 | B: 2024-10-29 | C: Patch | D: 2024-10-29 | E: 0 | F: 0 | G: 20 | H: Priya Halloran
|
||||
A: 2.2.3 | B: 2024-12-03 | C: Patch | D: 2024-12-03 | E: 0 | F: 0 | G: 16 | H: Priya Halloran | I: Last release of 2024; December minor slot skipped
|
||||
A: 2.3.0 | B: 2025-02-18 | C: Minor | D: 2025-02-18 | E: 0 | F: 1 | G: 241 | H: Tobias Wren | I: Cadence restarted
|
||||
A: 2.3.1 | B: 2025-03-04 | C: Patch | D: 2025-03-04 | E: 0 | F: 0 | G: 18 | H: Tobias Wren | I: Includes fix for KW-SA-2025-01
|
||||
A: 2.4.0 | B: 2025-05-20 | C: Minor | D: 2025-05-20 | E: 0 | F: 1 | G: 276 | H: Priya Halloran | I: Tag cache added
|
||||
A: 2.4.1 | B: 2025-06-03 | C: Patch | D: 2025-06-03 | E: 0 | F: 0 | G: 22 | H: Priya Halloran | I: Introduced index corruption bug for corpora over 50,000 notes
|
||||
A: 2.4.2 | B: 2025-06-07 | C: Patch | D: unscheduled | E: n/a | F: 0 | G: 3 | H: Priya Halloran | I: Emergency fix for index corruption, published on a Saturday
|
||||
A: 2.5.0 | B: 2025-08-26 | C: Minor | D: 2025-08-19 | E: 7 | F: 2 | G: 259 | H: Tobias Wren | I: Second release candidate needed; tag cache made smaller
|
||||
A: 2.5.1 | B: 2025-09-12 | C: Patch | D: 2025-09-09 | E: 3 | F: 0 | G: 14 | H: Tobias Wren | I: Delayed by expired signing key
|
||||
A: 2.5.2 | B: 2025-09-30 | C: Patch | D: unscheduled | E: n/a | F: 0 | G: 2 | H: Tobias Wren | I: Security release for KW-SA-2025-02
|
||||
A: 2.6.0 | B: 2025-11-25 | C: Minor | D: 2025-11-18 | E: 7 | F: 2 | G: 228 | H: Dagny Oyelaran | I: Second release candidate needed; trigram pre-filter
|
||||
A: 2.6.1 | B: 2025-12-09 | C: Patch | D: 2025-12-09 | E: 0 | F: 0 | G: 19 | H: Dagny Oyelaran
|
||||
A: 2.6.2 | B: 2025-12-23 | C: Patch | D: 2025-12-23 | E: 0 | F: 0 | G: 4 | H: Dagny Oyelaran | I: Last release of 2025
|
||||
|
||||
#### Benchmarks
|
||||
|
||||
A: Release | B: Corpus | C: Notes in corpus (count) | D: Corpus size (MB) | E: Index build time (s) | F: Median search latency (ms) | G: 95th percentile latency (ms) | H: Peak memory (MB) | I: Change in median latency vs previous release (%) | J: Notes
|
||||
A: 2.2.0 | B: Small | C: 10000 | D: 48 | E: 14.8 | F: 21 | G: 63 | H: 182 | I: n/a | J: Baseline for this sheet
|
||||
A: 2.2.0 | B: Large | C: 100000 | D: 471 | E: 171.2 | F: 148 | G: 412 | H: 1240 | I: n/a | J: Baseline for this sheet
|
||||
A: 2.3.0 | B: Small | C: 10000 | D: 48 | E: 13.9 | F: 19 | G: 58 | H: 176 | I: -9.5
|
||||
A: 2.3.0 | B: Large | C: 100000 | D: 471 | E: 160.4 | F: 131 | G: 377 | H: 1205 | I: -11.5
|
||||
A: 2.4.0 | B: Small | C: 10000 | D: 48 | E: 12.1 | F: 17 | G: 51 | H: 188 | I: -10.5 | J: Peak memory +6.8% from tag cache
|
||||
A: 2.4.0 | B: Large | C: 100000 | D: 471 | E: 139.7 | F: 112 | G: 330 | H: 1318 | I: -14.5 | J: Peak memory +9.4%, just under the 10% threshold
|
||||
A: 2.5.0 | B: Small | C: 10000 | D: 48 | E: 12.4 | F: 16 | G: 47 | H: 171 | I: -5.9 | J: Index build +2.5%, under the 5% threshold
|
||||
A: 2.5.0 | B: Large | C: 100000 | D: 471 | E: 142.9 | F: 104 | G: 298 | H: 1104 | I: -7.1 | J: Smaller tag cache lowered peak memory
|
||||
A: 2.6.0 | B: Small | C: 10000 | D: 48 | E: 9.6 | F: 11 | G: 34 | H: 165 | I: -31.3 | J: Trigram pre-filter
|
||||
A: 2.6.0 | B: Large | C: 100000 | D: 471 | E: 108.5 | F: 72 | G: 214 | H: 1061 | I: -30.8 | J: Trigram pre-filter
|
||||
|
||||
#### Issue triage 2025
|
||||
|
||||
A: Month (2025) | B: Issues opened (count) | C: Issues closed (count) | D: Bug reports (count) | E: Feature requests (count) | F: Questions and other (count) | G: Median first response (hours) | H: Unlabelled after 7 days (count) | I: Triage arrangement | J: Notes
|
||||
A: January | B: 84 | C: 71 | D: 44 | E: 25 | F: 15 | G: 44 | H: 9 | I: No rota
|
||||
A: February | B: 79 | C: 75 | D: 41 | E: 24 | F: 14 | G: 41 | H: 7 | I: No rota
|
||||
A: March | B: 102 | C: 88 | D: 55 | E: 30 | F: 17 | G: 38 | H: 6 | I: No rota | J: Spike after 2.3.0 release
|
||||
A: April | B: 88 | C: 80 | D: 46 | E: 27 | F: 15 | G: 35 | H: 5 | I: No rota
|
||||
A: May | B: 97 | C: 83 | D: 50 | E: 30 | F: 17 | G: 33 | H: 4 | I: No rota
|
||||
A: June | B: 131 | C: 122 | D: 82 | E: 28 | F: 21 | G: 27 | H: 11 | I: No rota | J: 37 reports of index corruption after 2.4.1
|
||||
A: July | B: 86 | C: 79 | D: 45 | E: 26 | F: 15 | G: 36 | H: 8 | I: No rota
|
||||
A: August | B: 104 | C: 90 | D: 56 | E: 31 | F: 17 | G: 29 | H: 3 | I: Weekly rota | J: Weekly rota began Monday 4 August
|
||||
A: September | B: 95 | C: 87 | D: 52 | E: 27 | F: 16 | G: 28 | H: 2 | I: Weekly rota
|
||||
A: October | B: 92 | C: 84 | D: 47 | E: 29 | F: 16 | G: 26 | H: 2 | I: Weekly rota | J: CI outage 14 to 22 October
|
||||
A: November | B: 99 | C: 86 | D: 45 | E: 34 | F: 20 | G: 24 | H: 1 | I: Weekly rota
|
||||
A: December | B: 81 | C: 76 | D: 39 | E: 30 | F: 12 | G: 39 | H: 6 | I: Weekly rota | J: Holiday period; fewer people active
|
||||
|
||||
#### Maintainers
|
||||
|
||||
A: Name | B: Role | C: Maintainer since | D: Main area | E: Triage weeks in 2025 (count) | F: Pull request reviews in 2025 (count) | G: Release manager for | H: Notes
|
||||
A: Imogen Varga | B: Lead maintainer until 30 June 2026 | C: 2019-04 | D: Index module | E: 2 | F: 188 | G: 2.0.0 | H: Founder of the project
|
||||
A: Tobias Wren | B: Maintainer | C: 2020-11 | D: Release tooling | E: 3 | F: 214 | G: 2.1.0, 2.3.0, 2.5.0
|
||||
A: Priya Halloran | B: Maintainer | C: 2021-03 | D: Store module and sync | E: 3 | F: 201 | G: 2.2.0, 2.4.0
|
||||
A: Dagny Oyelaran | B: Maintainer | C: 2021-09 | D: Documentation and translations | E: 2 | F: 167 | G: 2.6.0 | H: Conduct committee member
|
||||
A: Mateus Kowalczyk | B: Performance lead | C: 2022-06 | D: Benchmarks and search speed | E: 3 | F: 139 | H: Approves perf-regression pull requests
|
||||
A: Ruth Abernethy | B: Maintainer | C: 2023-02 | D: Community | E: 2 | F: 96 | H: Chair of the conduct committee
|
||||
A: Felix Ondrasek | B: Maintainer | C: 2024-01 | D: Plugins | E: 3 | F: 158 | H: Standing conduct substitute from January 2026; release manager for 2.8.0
|
||||
A: Saoirse Lindqvist | B: Maintainer | C: 2025-04-08 | D: Testing and continuous integration | E: 3 | F: 172 | H: Release manager for 2.7.0
|
||||
A: Ngozi Barraclough | B: Maintainer | C: 2025-10-14 | D: Windows support | E: 1 | F: 41 | H: Joined in October 2025
|
||||
|
|
@ -0,0 +1,9 @@
|
|||
Lyngmo Idrettslag – Årsmelding 2025 Lyngmo Idrettslag – Årsmelding 2025 Årsmeldingen gjelder perioden 1. januar til 31. desember 2025 og legges fram for årsmøtet torsdag 19. mars 2026 kl. 19:00 i klubbhuset på Lyngmoen. Den er skrevet av styret med bidrag fra lederne i langrennsgruppa, orienteringsgruppa, idrettsskolen, hyttekomiteen og løypekomiteen. Regnskapet er revidert av lagets to valgte revisorer, Per Ottar Lunde og Gunvor Aaslie, som leverte sin beretning 2. mars 2026. 1. Styret og organisasjon Styret har i 2025 bestått av leder Torunn Bakkevold, nestleder Eivind Rustad, kasserer Sigrun Høiby, sekretær Halvard Moen og styremedlem Ingrid Fjelstad. Varamedlemmer har vært Rolf Engebakken og Synnøve Bratlie. Styret har holdt 11 styremøter og behandlet 74 saker, mot 9 møter og 61 saker året før. Alle styremøter holdes første tirsdag i måneden, med unntak av juli, da det ikke holdes møte. Under styret ligger fem faste utvalg. Langrennsgruppa ledes av Marte Kvamsdal, orienteringsgruppa av Bjørnar Teigen og idrettsskolen av Hanne Sæther. Hyttekomiteen ledes av hyttesjef Kåre Lien, og løypekomiteen ledes av løypesjef Arnstein Vold. Rennleder for Lyngmorennet var Øystein Brekke, som har hatt rollen siden 2021. Laget hadde 684 medlemmer per 31. desember 2025, mot 641 ett år tidligere. Det er en økning på 43 medlemmer, eller 6,7 prosent. Veksten kom i hovedsak blant barn under 13 år, fordi idrettsskolen tok inn flere barn enn før. Tabellen viser medlemmene fordelt på kategori. Kategori Alder Medlemmer 31.12.2025 Medlemmer 31.12.2024 Barn 0–12 år 231 204 Ungdom 13–19 år 118 112 Voksen 20–66 år 268 262 Honnør 67 år og eldre 67 63 Sum 684 641 I tillegg hadde laget 29 støttemedlemmer. Støttemedlemmer betaler 200 kr i året, men de har ikke stemmerett på årsmøtet og kan ikke velges til tillitsverv. De er derfor ikke tatt med i tallene over. Av de ordinære medlemmene var 142 husstander registrert med familiemedlemskap. 2. Aktivitet i gruppene Langrennsgruppa hadde 96 aktive utøvere mellom 8 og 16 år i 2025, mot 88 året før. Gruppa er delt i tre treningsgrupper: 8–10 år, 11–12 år og konkurransegruppa for 13–16 år. Konkurransegruppa hadde 27 utøvere og deltok til sammen i 34 renn utenfor Lyngmoen. Gruppa hadde 21 trenere, og 9 av dem fullførte lagets grunnkurs for trenere i løpet av året.
|
||||
|
||||
Fellestreningene for 8–12 år ble holdt i lysløypa på tirsdager og torsdager fra 1. desember 2025, så snart det var nok snø. På grunn av lite snø i begynnelsen av desember ble de fire første treningene holdt som løpetrening på stiene ved Tjernåsen i stedet. Barmarkstreningen fra august til november samlet i snitt 41 barn per kveld. Sommerens samling for konkurransegruppa ble holdt på Granlihytta 15.–17. august 2025 med 22 deltakere. Egenandelen var 650 kr per deltaker, og laget dekket resten av kostnaden på 17 900 kr. Utlånsordningen for ski hadde 46 par klassiske ski og 30 par skøyteski ved utgangen av 2025. Ordningen er bare for medlemmer under 13 år. Familiene betaler et depositum på 300 kr per par, og depositumet betales tilbake når skiene leveres innen 30. april. Skiene ble lånt ut 61 ganger i løpet av sesongen, og 4 par ble kassert fordi de var for slitt til å repareres. Orienteringsgruppa hadde 41 aktive medlemmer i 2025. Gruppa arrangerte 13 karusellløp på tirsdager fra 6. mai til 2. september, med sommerpause i hele juli. Karusellen hadde til sammen 1 126 starter, eller i snitt 87 starter per løp. Det beste enkeltløpet var 10. juni med 121 starter. Karusellen er åpen for alle, men bare medlemmer kan delta i sammenlagtkonkurransen. Det nye kartet over Tjernåsen nord ble ferdig i august 2025. Kartet dekker 3,8 kvadratkilometer og er tegnet i målestokk 1:7 500 for barn og 1:10 000 for voksne. Kartarbeidet kostet 45 000 kr og ble i sin helhet betalt med gaver fra lokale givere som ønsket å være anonyme. Derfor belastet kartet ikke driftsbudsjettet. Idrettsskolen er for barn fra 6 til og med 11 år og holdes på onsdager kl. 17:30–18:45 ved klubbhuset. Vårsemesteret varte fra 8. januar til 7. mai 2025 og hadde 52 barn. Høstsemesteret varte fra 3. september til 26. november 2025 og hadde 58 barn, som er nær taket på 60 barn. Idrettsskolen hadde sju faste instruktører. Om vinteren går idrettsskolen på ski i Barnerunden, og resten av året veksler barna mellom løping, ballaktiviteter og enkel orientering. Tre onsdager i 2025 ble avlyst: to fordi temperaturen var lavere enn minus 12 grader, og én fordi klubbhuset hadde vannlekkasje. 3. Anlegg: lysløypa og løypenettet Lagets løypenett ligger ved Tjernåsen og består av fire løyper. Storrunden er 4,2 km lang og er opplyst. Barnerunden er 1,1 km lang og er også opplyst. Granlirunden er 9,5 km og går fra lysløypa til Granlihytta og tilbake, men den har ikke lys. Hundesporet er 3,0 km og er heller ikke opplyst. Løype Lengde Lys Stilart Hund Storrunden 4,2 km Ja, 06:00–22:00 Klassisk og skøyting Ikke tillatt 1. des.–15. april Barnerunden 1,1 km Ja, 16:00–21:00 Bare klassisk Ikke tillatt 1. des.–15. april
|
||||
|
||||
Granlirunden 9,5 km Nei Klassisk og skøyting Tillatt i bånd Hundesporet 3,0 km Nei Klassisk Tillatt Fra 1. januar 2025 slås lyset i Storrunden av kl. 22:00 i stedet for kl. 23:00, som var tidligere praksis. Endringen ble vedtatt for å spare strøm. Som følge av endringen falt strømkostnaden for lysløypa fra 71 200 kr i 2024 til 61 200 kr i 2025, en nedgang på 14 prosent. Lyset i Barnerunden står på mellom kl. 16:00 og 21:00, fordi runden først og fremst brukes av barnefamilier tidlig på kvelden. Løypekomiteen preparerte lysløypa 61 ganger i sesongen 2024/2025, mot 54 ganger sesongen før. Frivillige brukte til sammen 486 timer på preparering. Lysløypa prepareres bare når snødybden er minst 20 cm, og Granlirunden prepareres bare når snødybden er minst 30 cm, fordi terrenget der har mer stein og røtter. Skøytetraseen i Storrunden kjøres opp bare på tirsdager, torsdager og lørdager, slik at klassisksporet holder seg fint de andre dagene. Hunder har ikke lov til å være i Storrunden eller Barnerunden fra 1. desember til 15. april. Hundesporet ble lagt i 2023 nettopp for at hundeeiere skal ha et eget tilbud. I Granlirunden er hund tillatt i bånd, med unntak av dagen for Lyngmorennet, da løypa er stengt for alle andre enn løpere og funksjonærer. 4. Granlihytta Granlihytta ligger ved Granlitjernet, 4,6 km fra startområdet ved lysløypa. Hytta har 22 sengeplasser fordelt på 6 soverom, en stor stue med vedovn og et kjøkken med plass til 40 personer. Hytta har strøm, men ikke innlagt vann om vinteren. Vann hentes da fra kilden 150 meter nedenfor hytta. Hytta hadde 312 gjestedøgn i 2025, mot 287 i 2024. Overnatting er forbeholdt medlemmer og gjester som er med et medlem. Leietakeren må være minst 25 år og være medlem av laget. Prisen i 2025 var 150 kr per natt for medlemmer og 250 kr per natt for gjester. Hele hytta kan leies for en helg, fra fredag kl. 16:00 til søndag kl. 15:00, for 3 500 kr. Helgeleie av hele hytta er kun for medlemmer. Booking skal gjøres hos hyttesjefen senest 14 dager før ankomst. Avbestilling er gratis inntil 7 dager før ankomst; senere avbestilling koster 50 prosent av leien. Leietakeren skal vaske hytta før avreise. Blir hytta ikke vasket, belastes et vaskegebyr på 600 kr. Røyking er forbudt både inne i hytta og på verandaen. Hund kan være i stua og gangen, men ikke på soverommene. Kafeen i Granlihytta var åpen 13 søndager fra 5. januar til 30. mars 2025, kl. 11:00–15:00. Kafeen hadde 2 870 besøkende og omsatte for 96 300 kr. Vaffelsalget alene utgjorde omtrent en tredjedel av omsetningen. Kafeen drives helt av frivillige, og hver søndag krever fire vakter.
|
||||
|
||||
Sommeren 2025 ble det oppdaget lekkasje i taket over to av soverommene. Hyttekomiteen gjorde en midlertidig reparasjon for 38 000 kr, men taktekkeren som besiktiget hytta i september anbefalte å skifte hele taket innen to år. Kostnaden er anslått til 420 000 kr. Styret foreslår derfor for årsmøtet at taket skiftes sommeren 2026, og at arbeidet finansieres med overskuddet fra 2025 og en del av egenkapitalen. Vedlikeholdskostnaden for hytta ble 188 400 kr i 2025, nesten tre ganger så høy som i 2024, fordi også pipa ble murt om og vedboden fikk ny grunnmur. 5. Lyngmorennet 2025 Det 31. Lyngmorennet ble arrangert søndag 9. februar 2025 med start og mål ved lysløypa. Rennet hadde 368 startende, mot 341 i 2024. Klassene for barn opp til 12 år gikk 2 km eller 3 km i Storrunden og Barnerunden, mens ungdom og voksne gikk 10 km eller 20 km med Granlirunden som hoveddel. Første start gikk kl. 10:00. Startkontingenten var 100 kr for barn opp til 12 år, 150 kr for ungdom 13–16 år og 300 kr for alle fra 17 år. Etteranmelding kostet 100 kr ekstra uansett klasse. Påmeldingsfristen var onsdag 5. februar kl. 23:59, fire dager før rennet. Rennet hadde 71 funksjonærer, og alle var frivillige. Rennet ga inntekter på 164 200 kr og kostnader på 76 800 kr, slik at overskuddet ble 87 400 kr. Overskuddet går uavkortet til drift av lysløypa. Neste renn, det 32. Lyngmorennet, ble fastsatt til søndag 8. februar 2026. 6. Dugnad Medlemmene la ned 3 140 dugnadstimer i 2025, mot 2 870 timer i 2024. Den største enkeltdugnaden var Lyngmorennet, med omtrent 640 timer. Hyttedugnaden 7. juni samlet 34 personer, som hogg og stablet om lag 18 favner ved. Laget kunne ikke drevet Granlihytta, lysløypa og rennet uten denne innsatsen. Dugnadsordningen er uendret fra 2024: hver familie med barn i langrenn eller orientering skal ta to dugnadsvakter per sesong, og familier som bare har barn i idrettsskolen skal ta én vakt. En vakt varer minst fire timer. Familier som ikke tar vaktene sine, betaler en dugnadsavgift på 800 kr per vakt. I 2025 ble det fakturert dugnadsavgift for 23 vakter, til sammen 18 400 kr, som er inntektsført under treningsavgifter. 7. Økonomi og veien videre Regnskapet for 2025 viser et overskudd på 285 800 kr, mot 120 700 kr i 2024. Overskuddet i 2024 var lavere fordi laget det året kjøpte nytt tidtakingsutstyr for 168 000 kr. Det store overskuddet i 2025 skyldes særlig økte inntekter fra kontingent, treningsavgifter og hytteutleie, og at gavene til kartet over Tjernåsen nord kom inn som ekstraordinær inntekt. Tabellen viser hovedpostene. Post 2025 (kr) 2024 (kr) Inntekt: Medlemskontingent 312 400 281 600
|
||||
|
||||
Inntekt: Treningsavgifter 198 600 172 400 Inntekt: Lyngmorennet 164 200 149 500 Inntekt: Hytteutleie Granlihytta 141 800 128 300 Inntekt: Kafé i Granlihytta 96 300 88 700 Inntekt: Gaver og innsamling 45 000 12 000 Sum inntekter 958 300 832 500 Kostnad: Vedlikehold Granlihytta 188 400 64 200 Kostnad: Utstyr 97 500 283 100 Kostnad: Lyngmorennet 76 800 71 900 Kostnad: Løypepreparering 74 800 69 500 Kostnad: Strøm lysløypa 61 200 71 200 Kostnad: Idrettsskolen 51 300 47 600 Kostnad: Trenerutdanning 42 000 28 000 Kostnad: Kafévarer 41 900 39 400 Kostnad: Administrasjon 38 600 36 900 Sum kostnader 672 500 711 800 Årsresultat 285 800 120 700 Egenkapitalen var 1 146 000 kr per 31. desember 2025. Laget har ingen lån. Styret foreslår at hele overskuddet for 2025 settes av til nytt tak på Granlihytta. Medlemskontingenten for 2025 var 600 kr for voksne og 1 300 kr for familier. Styret vil foreslå for årsmøtet at voksenkontingenten økes til 650 kr og familiekontingenten til 1 400 kr fra 2026, mens kontingenten for barn, ungdom og honnør holdes uendret. Kostnaden til trenerutdanning økte fra 28 000 kr til 42 000 kr, fordi ni trenere tok grunnkurset. Laget betaler kurset bare når treneren forplikter seg til å være trener i minst to sesonger etter kurset. Slutter treneren før det, må halvparten av kursavgiften betales tilbake. Styret takker alle trenere, instruktører, funksjonærer, hyttevakter og løypekjørere for innsatsen i 2025. Det viktigste målet for 2026 er å skifte taket på Granlihytta uten å svekke tilbudet til barn og unge. Styret ønsker også å utvide idrettsskolen til 70 plasser fra høsten 2027, men bare dersom laget finner minst tre nye instruktører. Lyngmoen, 2. mars 2026. Torunn Bakkevold, Eivind Rustad, Sigrun Høiby, Halvard Moen, Ingrid Fjelstad.
|
||||
173
gates/calibration-folder/kilde/lyngmo-il-aarsmote-2026.pptx.md
Normal file
173
gates/calibration-folder/kilde/lyngmo-il-aarsmote-2026.pptx.md
Normal file
|
|
@ -0,0 +1,173 @@
|
|||
## Årsmøte i Lyngmo Idrettslag 19. mars 2026
|
||||
|
||||
## Årsmøte i Lyngmo Idrettslag 2026
|
||||
|
||||
Årsmøtet holdes torsdag 19. mars 2026 kl. 19:00 i klubbhuset på Lyngmoen.
|
||||
|
||||
Bare medlemmer som har fylt 15 år og har betalt kontingenten for 2025, har stemmerett.
|
||||
|
||||
Støttemedlemmer kan være til stede, men de kan ikke stemme.
|
||||
|
||||
Møteleder er foreslått å være Rolf Engebakken.
|
||||
|
||||
## Saksliste
|
||||
|
||||
Sak 1–3: Godkjenning av innkalling, saksliste og møteleder.
|
||||
|
||||
Sak 4: Årsmelding for 2025.
|
||||
|
||||
Sak 5: Regnskap for 2025 med revisorenes beretning.
|
||||
|
||||
Sak 6: Forslag om ny kontingent fra 2026.
|
||||
|
||||
Sak 7: Forslag om å skifte taket på Granlihytta sommeren 2026.
|
||||
|
||||
Sak 8 og 9: Budsjett for 2026 og valg.
|
||||
|
||||
## Medlemmer
|
||||
|
||||
Laget hadde 684 medlemmer per 31. desember 2025, mot 641 året før.
|
||||
|
||||
Økningen på 43 medlemmer kom mest blant barn under 13 år.
|
||||
|
||||
Barnegruppen 0–12 år vokste fra 204 til 231 medlemmer.
|
||||
|
||||
I tillegg hadde laget 29 støttemedlemmer, som ikke telles med i medlemstallet.
|
||||
|
||||
## Langrenn
|
||||
|
||||
Langrennsgruppa hadde 96 aktive utøvere i 2025, og 101 per mars 2026.
|
||||
|
||||
Konkurransegruppa for 13–16 år hadde 27 utøvere og gikk 34 renn utenfor Lyngmoen i 2025.
|
||||
|
||||
Ni trenere tok grunnkurset for trenere i 2025.
|
||||
|
||||
De fire første skitreningene i desember 2025 ble løpetrening fordi det var for lite snø.
|
||||
|
||||
## Orientering
|
||||
|
||||
Tirsdagskarusellen hadde 13 løp fra 6. mai til 2. september 2025.
|
||||
|
||||
Karusellen fikk 1 126 starter, i snitt 87 per løp.
|
||||
|
||||
Det nye kartet over Tjernåsen nord dekker 3,8 kvadratkilometer.
|
||||
|
||||
Kartet kostet 45 000 kr og ble betalt fullt ut med gaver.
|
||||
|
||||
## Idrettsskolen
|
||||
|
||||
Idrettsskolen hadde 52 barn våren 2025 og 58 barn høsten 2025.
|
||||
|
||||
Taket er 60 barn per semester, og vårsemesteret 2026 er fullt.
|
||||
|
||||
Tre onsdager ble avlyst i 2025, to av dem på grunn av kulde under minus 12 grader.
|
||||
|
||||
Styret vil utvide til 70 plasser fra høsten 2027 hvis tre nye instruktører melder seg.
|
||||
|
||||
## Lysløypa og løypenettet
|
||||
|
||||
Lyset i Storrunden slås av kl. 22:00 fra 1. januar 2025, mot kl. 23:00 tidligere.
|
||||
|
||||
Strømkostnaden falt fra 71 200 kr i 2024 til 61 200 kr i 2025.
|
||||
|
||||
Lysløypa ble preparert 61 ganger i sesongen 2024/2025, mot 54 ganger sesongen før.
|
||||
|
||||
Vinteren 2026 ble Granlirunden først preparert 17. januar, fordi den krever minst 30 cm snø.
|
||||
|
||||
## Granlihytta
|
||||
|
||||
Hytta hadde 312 gjestedøgn i 2025, mot 287 i 2024.
|
||||
|
||||
Kafeen hadde 2 870 besøkende på 13 søndager i 2025.
|
||||
|
||||
Per 15. mars 2026 har kafeen hatt 2 460 besøkende på 11 søndager.
|
||||
|
||||
Vedlikeholdet kostet 188 400 kr i 2025 på grunn av taklekkasje, ny pipe og ny grunnmur under vedboden.
|
||||
|
||||
## Lyngmorennet 2025
|
||||
|
||||
Det 31. Lyngmorennet gikk søndag 9. februar 2025 med 368 startende.
|
||||
|
||||
Rennet hadde 71 frivillige funksjonærer.
|
||||
|
||||
Inntektene var 164 200 kr og kostnadene 76 800 kr.
|
||||
|
||||
Overskuddet på 87 400 kr gikk til drift av lysløypa.
|
||||
|
||||
## Lyngmorennet 2026
|
||||
|
||||
Det 32. Lyngmorennet gikk søndag 8. februar 2026 med 412 startende.
|
||||
|
||||
Rennet hadde 83 funksjonærer, som er 12 flere enn i 2025.
|
||||
|
||||
Foreløpig regnskap viser et overskudd på 94 100 kr.
|
||||
|
||||
Det 33. Lyngmorennet er satt til søndag 7. februar 2027.
|
||||
|
||||
## Dugnad
|
||||
|
||||
Medlemmene la ned 3 140 dugnadstimer i 2025, mot 2 870 timer i 2024.
|
||||
|
||||
Lyngmorennet alene krevde omtrent 640 dugnadstimer.
|
||||
|
||||
Det ble fakturert dugnadsavgift for 23 vakter i 2025, til sammen 18 400 kr.
|
||||
|
||||
Fra sesongen 2026–2027 teller en seks timers vakt under Lyngmorennet som to vakter.
|
||||
|
||||
## Regnskap 2025
|
||||
|
||||
Inntektene var 958 300 kr og kostnadene 672 500 kr.
|
||||
|
||||
Årsresultatet ble et overskudd på 285 800 kr, mot 120 700 kr i 2024.
|
||||
|
||||
Egenkapitalen var 1 146 000 kr per 31. desember 2025, og laget har ingen lån.
|
||||
|
||||
Revisorene Per Ottar Lunde og Gunvor Aaslie anbefaler at regnskapet godkjennes.
|
||||
|
||||
## Sak 6: Ny kontingent fra 2026
|
||||
|
||||
Styret foreslår å øke voksenkontingenten fra 600 kr til 650 kr.
|
||||
|
||||
Styret foreslår å øke familiekontingenten fra 1 300 kr til 1 400 kr.
|
||||
|
||||
Kontingenten for barn (300 kr), ungdom (400 kr) og honnør (300 kr) foreslås uendret.
|
||||
|
||||
Økningen skal bygge opp egenkapitalen igjen etter takarbeidet på Granlihytta.
|
||||
|
||||
## Sak 7: Nytt tak på Granlihytta
|
||||
|
||||
Taktekkeren anbefalte i september 2025 å skifte hele taket innen to år.
|
||||
|
||||
Kostnaden er anslått til 420 000 kr.
|
||||
|
||||
Arbeidet foreslås gjort sommeren 2026, og hytta stenges for utleie fra 15. juni til 20. august 2026.
|
||||
|
||||
Hele overskuddet for 2025 settes av til taket, og resten tas fra egenkapitalen.
|
||||
|
||||
## Sak 8: Budsjett 2026
|
||||
|
||||
Budsjetterte inntekter er 1 012 000 kr.
|
||||
|
||||
Budsjetterte kostnader er 1 154 000 kr, medregnet 420 000 kr til nytt tak.
|
||||
|
||||
Budsjettet gir et underskudd på 142 000 kr, som dekkes av egenkapitalen.
|
||||
|
||||
Uten takarbeidet ville budsjettet vist et overskudd på 278 000 kr.
|
||||
|
||||
## Sak 9: Valg
|
||||
|
||||
Torunn Bakkevold foreslås gjenvalgt som leder for to år.
|
||||
|
||||
Kasserer Sigrun Høiby går av etter seks år, og Leif Opsahl foreslås som ny kasserer for to år.
|
||||
|
||||
Halvard Moen foreslås gjenvalgt som sekretær for ett år.
|
||||
|
||||
Eivind Rustad og Ingrid Fjelstad er ikke på valg i år.
|
||||
|
||||
## Takk for innsatsen
|
||||
|
||||
Laget hadde 21 langrennstrenere og sju instruktører i idrettsskolen i 2025.
|
||||
|
||||
Uten 3 140 dugnadstimer kunne vi ikke drevet hytta, lysløypa og rennet.
|
||||
|
||||
Neste store dugnad er hyttedugnaden lørdag 6. juni 2026.
|
||||
|
|
@ -0,0 +1,295 @@
|
|||
# Håndbok for trenere og foreldre – sesongen 2026–2027
|
||||
|
||||
# Håndbok for trenere og foreldre – sesongen 2026–2027
|
||||
|
||||
Denne håndboka gjelder for langrennsgruppa, orienteringsgruppa og idrettsskolen i Lyngmo Idrettslag fra 1. august 2026 til 31. juli 2027. Den erstatter håndboka for 2025–2026. Håndboka ble vedtatt av styret 1. september 2026 og bygger på vedtakene fra årsmøtet 19. mars 2026. Alle trenere skal ha lest håndboka før første trening, og foreldre får den tilsendt ved påmelding.
|
||||
|
||||
## 1. Hvem vi er og hva vi vil
|
||||
|
||||
Lyngmo Idrettslag skal gi barn og unge på Lyngmoen et trygt og morsomt idrettstilbud nær der de bor. Vi ønsker at flest mulig skal fortsette med idrett lengst mulig. Derfor rangerer vi ikke barn under 13 år i resultatlister fra våre egne arrangementer, og alle barn i gruppene 8–12 år får like mye oppfølging uansett nivå.
|
||||
|
||||
Alle som møter på trening, skal bli sett og hilst på av en trener.
|
||||
|
||||
Ingen barn skal sendes hjem alene før foreldre har bekreftet at det er i orden.
|
||||
|
||||
Trenere og foreldre skal snakke positivt om andre utøvere, trenere og dommere.
|
||||
|
||||
Konkurranse er frivillig for alle under 13 år.
|
||||
|
||||
## 2. Grupper og treningstider
|
||||
|
||||
Tabellen viser de faste treningstidene for sesongen. Skitreningene starter i lysløypa så snart det er preparert, og senest fra 1. desember 2026 dersom snøen tillater det. Før det trener langrennsgruppene på barmark fra samme sted. Orienteringsgruppene trener fra 4. mai til 31. august 2027, med pause i juli.
|
||||
|
||||
Gruppe
|
||||
|
||||
Alder
|
||||
|
||||
Dager
|
||||
|
||||
Tid
|
||||
|
||||
Sted
|
||||
|
||||
Idrettsskolen
|
||||
|
||||
6–11 år
|
||||
|
||||
Onsdag
|
||||
|
||||
17:30–18:45
|
||||
|
||||
Klubbhuset
|
||||
|
||||
Langrenn 8–10
|
||||
|
||||
8–10 år
|
||||
|
||||
Tirsdag og torsdag
|
||||
|
||||
17:30–18:30
|
||||
|
||||
Lysløypa, Barnerunden
|
||||
|
||||
Langrenn 11–12
|
||||
|
||||
11–12 år
|
||||
|
||||
Tirsdag og torsdag
|
||||
|
||||
18:00–19:15
|
||||
|
||||
Lysløypa, Storrunden
|
||||
|
||||
Konkurransegruppa
|
||||
|
||||
13–16 år
|
||||
|
||||
Mandag og onsdag, lørdag
|
||||
|
||||
18:30–20:00, lørdag 10:00–12:00
|
||||
|
||||
Lysløypa og Granlirunden
|
||||
|
||||
Orientering 9–12
|
||||
|
||||
9–12 år
|
||||
|
||||
Tirsdag
|
||||
|
||||
18:00–19:00
|
||||
|
||||
Tjernåsen nord
|
||||
|
||||
Orientering 13+
|
||||
|
||||
13 år og eldre
|
||||
|
||||
Torsdag
|
||||
|
||||
18:00–19:30
|
||||
|
||||
Tjernåsen nord
|
||||
|
||||
Idrettsskolen har høstsemester fra 2. september til 25. november 2026 og vårsemester fra 6. januar til 5. mai 2027. Det er plass til 60 barn per semester. Er det flere påmeldte, går barn som var med forrige semester foran nye barn, og deretter gjelder rekkefølgen for påmelding. Barn kan prøve to treninger gratis før de må være medlem og ha betalt treningsavgift.
|
||||
|
||||
Konkurransegruppa kan også bruke Granlirunden på lørdagstreningene, fordi den er lagets lengste løype. Barn under 12 år har ikke lov til å gå Granlirunden uten en voksen, og etter mørkets frembrudd skal alle i Granlirunden ha hodelykt, fordi løypa ikke er opplyst.
|
||||
|
||||
## 3. Trenerens ansvar
|
||||
|
||||
Hver treningsgruppe skal ha en hovedtrener som er minst 18 år. Hjelpetrenere må være minst 15 år og kan bare lede en trening når en hovedtrener er til stede. Det skal alltid være minst to voksne på hver trening, slik at én kan følge et barn som trenger hjelp mens den andre blir med gruppa.
|
||||
|
||||
I gruppene 8–12 år skal det være minst én voksen per 8 barn.
|
||||
|
||||
I idrettsskolen skal det være minst én voksen per 6 barn under 9 år og én voksen per 8 barn fra 9 år.
|
||||
|
||||
Treneren skal føre oppmøte for hver trening innen kl. 22:00 samme kveld.
|
||||
|
||||
Treneren skal ha med førstehjelpsveske, mobiltelefon og ekstra votter på alle treninger.
|
||||
|
||||
Alle trenere skal ha gjennomført førstehjelpskurs i løpet av de siste tre årene.
|
||||
|
||||
Laget holder førstehjelpskurs for trenere lørdag 10. oktober 2026 kl. 10:00–15:00 i klubbhuset. Kurset er gratis for trenere. Laget betaler også grunnkurset for trenere, men bare når treneren forplikter seg til å være trener i minst to sesonger etter kurset. Slutter treneren før det, må halvparten av kursavgiften betales tilbake.
|
||||
|
||||
Trenere har ikke lov til å legge ut bilder av barn på lagets sider eller i lagets kanaler uten skriftlig samtykke fra foreldrene. Samtykkeskjemaet skal leveres innen 1. oktober 2026. Bilder der barn kan gjenkjennes, skal aldri publiseres sammen med fullt navn, heller ikke når samtykke er gitt.
|
||||
|
||||
## 4. Foreldrenes ansvar
|
||||
|
||||
Foreldre har ansvar for at barnet har riktig utstyr og klær etter været, og for at barnet blir hentet presis. Foreldre til barn under 10 år skal bli på området under treningen eller avtale med treneren hvem som henter. Barn som ikke er hentet 15 minutter etter at treningen er slutt, blir sammen med en trener inne i klubbhuset til en voksen kommer.
|
||||
|
||||
Meld fra til treneren før kl. 15:00 hvis barnet ikke kommer, når gruppa har færre enn 10 påmeldte.
|
||||
|
||||
Gi beskjed til treneren om allergier, astma eller andre forhold som er viktige ved skade eller kulde.
|
||||
|
||||
Foreldre skal ikke stå i løypa og heie under trening, men gjerne ved start og mål.
|
||||
|
||||
Spørsmål om uttak og konkurranser tas med hovedtreneren, ikke på trening mens barna hører på.
|
||||
|
||||
## 5. Sikkerhet, kulde og vær
|
||||
|
||||
Temperaturen måles ved klubbhuset kl. 16:00 på treningsdager. Er det for kaldt, avlyses treningen, og beskjed sendes i gruppas meldingskanal senest kl. 16:30. Grensene er satt fordi barn mister varme raskere enn voksne, og fordi frostskader i ansiktet kan oppstå uten at barnet merker det.
|
||||
|
||||
Gruppe
|
||||
|
||||
Utetrening avlyses når temperaturen er lavere enn
|
||||
|
||||
Idrettsskolen (6–11 år)
|
||||
|
||||
minus 12 °C
|
||||
|
||||
Langrenn og orientering 8–12 år
|
||||
|
||||
minus 15 °C
|
||||
|
||||
Konkurransegruppa 13–16 år
|
||||
|
||||
minus 20 °C
|
||||
|
||||
Når treningen avlyses på grunn av kulde, kan gruppene 8–12 år i stedet ha innetrening i klubbhuset dersom minst to trenere er tilgjengelige. Idrettsskolen har ingen innetrening ved kulde, fordi klubbhuset ikke har plass til 60 barn samtidig. Ved torden avbrytes orienteringstreningen straks, og alle samles ved startpunktet.
|
||||
|
||||
## 6. Løypene og regler i løypa
|
||||
|
||||
Lyset i Storrunden står på fra kl. 06:00 til 22:00 hver dag i skisesongen, og lyset i Barnerunden står på fra kl. 16:00 til 21:00. Skøyting er bare tillatt i Storrunden og Granlirunden. Barnerunden er bare for klassisk stil, slik at de minste kan gå i sporet uten å bli forstyrret.
|
||||
|
||||
Hunder har ikke lov til å være i Storrunden eller Barnerunden fra 1. desember til 15. april. Det gjelder også trenere og foreldre som venter på barn. Hundesporet på 3,0 km kan brukes med hund hele vinteren. Løypekomiteen preparerer lysløypa bare når snødybden er minst 20 cm. Er det mindre snø, flyttes skitreningen til barmark.
|
||||
|
||||
## 7. Utstyr og utlån
|
||||
|
||||
Laget låner ut ski til medlemmer under 13 år. Depositumet er 300 kr per par og betales tilbake når skiene leveres innen 30. april 2027. Skiene hentes i utstyrsboden ved klubbhuset på tirsdager kl. 17:00–18:00 i november. Ski som leveres etter 30. april, gir ikke depositumet tilbake, med unntak av når barnet har vært sykt og lederen av langrennsgruppa har godkjent det.
|
||||
|
||||
Smøreboden ved lysløypa er åpen for alle medlemmer, men smørejern og fluorfrie pulver skal bare brukes når en trener er til stede. Det er forbudt å bruke smøremidler med fluor på lagets ski.
|
||||
|
||||
## 8. Dugnad
|
||||
|
||||
Laget drives av frivillige. Hver familie med barn i langrenn eller orientering skal ta to dugnadsvakter per sesong, og familier som bare har barn i idrettsskolen skal ta én vakt. En vanlig vakt varer minst fire timer. Vakter under Lyngmorennet varer seks timer og teller derfor som to vakter. Familier som ikke tar vaktene sine, betaler en dugnadsavgift på 800 kr per vakt. Dugnadsavgiften faktureres 15. mai 2027.
|
||||
|
||||
Vakter settes opp av dugnadsansvarlig Wenche Tollefsrud, og bytte av vakt skal avtales direkte med en annen familie.
|
||||
|
||||
Ungdom fra 14 år kan ta vakt på vegne av familien, men bare sammen med en voksen på samme vakt.
|
||||
|
||||
Kafévakter i Granlihytta går fra kl. 10:30 til 15:30, fordi kafeen skal være klar ved åpning kl. 11:00.
|
||||
|
||||
Trenere og styremedlemmer er fritatt for dugnad det året de har vervet.
|
||||
|
||||
Det 33. Lyngmorennet arrangeres søndag 7. februar 2027. Påmeldingsfristen er onsdag 3. februar 2027 kl. 23:59. Startkontingenten er uendret: 100 kr for barn opp til 12 år, 150 kr for ungdom 13–16 år og 300 kr fra 17 år, og etteranmelding koster 100 kr ekstra. Barn i langrennsgruppa som tar en dugnadsvakt sammen med en forelder, starter gratis.
|
||||
|
||||
Kafeen i Granlihytta er åpen søndager kl. 11:00–15:00 fra 3. januar til 21. mars 2027. I påsken er kafeen åpen hver dag fra skjærtorsdag 25. mars til andre påskedag 29. mars 2027, kl. 11:00–16:00.
|
||||
|
||||
## 9. Granlihytta for treningsgrupper
|
||||
|
||||
Treningsgruppene kan låne Granlihytta gratis i inntil to helger per sesong. Det gjelder bare når minst to voksne er med, og når den som står ansvarlig, er over 25 år. Treningsgrupper må booke hytta hos hyttesjefen senest 30 dager før, fordi hyttehelgene ellers raskt blir leid ut til medlemmer. Vanlige regler for hytta gjelder også for treningsgrupper: hytta skal vaskes før avreise, og røyking er forbudt inne og på verandaen.
|
||||
|
||||
Taket på Granlihytta ble skiftet sommeren 2026. Hytta var derfor stengt for all utleie fra 15. juni til 20. august 2026, og samlingen for konkurransegruppa ble flyttet til helgen 28.–30. august 2026.
|
||||
|
||||
## 10. Avgifter og frister
|
||||
|
||||
Medlemskontingenten for 2026 ble vedtatt på årsmøtet 19. mars 2026. Kontingenten for voksne økte fra 600 til 650 kr og familiekontingenten fra 1 300 til 1 400 kr, fordi laget trenger å bygge opp egenkapitalen igjen etter takarbeidet på Granlihytta. Treningsavgiftene er uendret fra forrige sesong.
|
||||
|
||||
Avgift
|
||||
|
||||
Beløp
|
||||
|
||||
Frist
|
||||
|
||||
Medlemskontingent barn 0–12 år
|
||||
|
||||
300 kr per år
|
||||
|
||||
\31. mars
|
||||
|
||||
Medlemskontingent ungdom 13–19 år
|
||||
|
||||
400 kr per år
|
||||
|
||||
\31. mars
|
||||
|
||||
Medlemskontingent voksen 20–66 år
|
||||
|
||||
650 kr per år
|
||||
|
||||
\31. mars
|
||||
|
||||
Medlemskontingent honnør 67 år og eldre
|
||||
|
||||
300 kr per år
|
||||
|
||||
\31. mars
|
||||
|
||||
Familiekontingent
|
||||
|
||||
1 400 kr per år
|
||||
|
||||
\31. mars
|
||||
|
||||
Treningsavgift langrenn 8–12 år
|
||||
|
||||
900 kr per sesong
|
||||
|
||||
\15. oktober 2026
|
||||
|
||||
Treningsavgift konkurransegruppa 13–16 år
|
||||
|
||||
1 200 kr per sesong
|
||||
|
||||
\15. oktober 2026
|
||||
|
||||
Treningsavgift orientering
|
||||
|
||||
600 kr per sesong
|
||||
|
||||
\1. mai 2027
|
||||
|
||||
Idrettsskolen
|
||||
|
||||
450 kr per semester
|
||||
|
||||
\15. september 2026 og 20. januar 2027
|
||||
|
||||
Søskenmoderasjon gis på treningsavgiften: barn nummer to får 25 prosent rabatt, og barn nummer tre og videre trener gratis. Moderasjonen gjelder ikke idrettsskolen og ikke medlemskontingenten. Lagets ulykkesforsikring dekker medlemmer under trening, men bare når medlemskontingenten for året er betalt.
|
||||
|
||||
## 11. Kontakt
|
||||
|
||||
Rolle
|
||||
|
||||
Navn
|
||||
|
||||
Leder
|
||||
|
||||
Torunn Bakkevold
|
||||
|
||||
Nestleder
|
||||
|
||||
Eivind Rustad
|
||||
|
||||
Kasserer (fra 19. mars 2026)
|
||||
|
||||
Leif Opsahl
|
||||
|
||||
Leder langrennsgruppa
|
||||
|
||||
Marte Kvamsdal
|
||||
|
||||
Leder orienteringsgruppa
|
||||
|
||||
Bjørnar Teigen
|
||||
|
||||
Leder idrettsskolen
|
||||
|
||||
Hanne Sæther
|
||||
|
||||
Hyttesjef
|
||||
|
||||
Kåre Lien
|
||||
|
||||
Løypesjef
|
||||
|
||||
Arnstein Vold
|
||||
|
||||
Rennleder Lyngmorennet
|
||||
|
||||
Øystein Brekke
|
||||
|
||||
Dugnadsansvarlig
|
||||
|
||||
Wenche Tollefsrud
|
||||
|
||||
Spørsmål om faktura og betaling rettes til kassereren. Spørsmål om treningstider og grupper rettes til lederen av gruppa. Endringer i håndboka i løpet av sesongen vedtas av styret og varsles i alle gruppenes meldingskanaler.
|
||||
|
|
@ -0,0 +1,86 @@
|
|||
#### Kontingent og avgifter
|
||||
|
||||
A: Avgift | B: Gruppe | C: Beløp 2025 (kr) | D: Beløp 2026 (kr) | E: Antall 2025 | F: Enhet for antall | G: Merknad
|
||||
A: Medlemskontingent | B: Barn 0–12 år | C: 300 | D: 300 | E: 231 | F: medlemmer per 31.12.2025 | G: Uendret
|
||||
A: Medlemskontingent | B: Ungdom 13–19 år | C: 400 | D: 400 | E: 118 | F: medlemmer per 31.12.2025 | G: Uendret
|
||||
A: Medlemskontingent | B: Voksen 20–66 år | C: 600 | D: 650 | E: 268 | F: medlemmer per 31.12.2025 | G: Økt etter vedtak på årsmøtet 19.03.2026
|
||||
A: Medlemskontingent | B: Honnør 67 år og eldre | C: 300 | D: 300 | E: 67 | F: medlemmer per 31.12.2025 | G: Uendret
|
||||
A: Familiekontingent | B: Hele husstanden | C: 1300 | D: 1400 | E: 142 | F: husstander per 31.12.2025 | G: Økt etter vedtak på årsmøtet 19.03.2026
|
||||
A: Støttemedlem | B: Alle aldre | C: 200 | D: 200 | E: 29 | F: støttemedlemmer per 31.12.2025 | G: Ingen stemmerett, kan ikke velges til verv
|
||||
A: Treningsavgift | B: Langrenn 8–12 år | C: 900 | D: 900 | E: 69 | F: aktive utøvere 2025 | G: Frist 15. oktober
|
||||
A: Treningsavgift | B: Konkurransegruppa 13–16 år | C: 1200 | D: 1200 | E: 27 | F: aktive utøvere 2025 | G: Frist 15. oktober
|
||||
A: Treningsavgift | B: Orientering | C: 600 | D: 600 | E: 41 | F: aktive medlemmer 2025 | G: Frist 1. mai
|
||||
A: Treningsavgift per semester | B: Idrettsskolen 6–11 år | C: 450 | D: 450 | E: 58 | F: barn høsten 2025 | G: Søskenmoderasjon gjelder ikke
|
||||
A: Dugnadsavgift per vakt | B: Familier med aktive barn | C: 800 | D: 800 | E: 23 | F: vakter fakturert 2025 | G: Faktureres 15. mai
|
||||
A: Depositum utlånsski per par | B: Medlemmer under 13 år | C: 300 | D: 300 | E: 61 | F: utlån sesongen 2025 | G: Tilbake ved levering innen 30. april
|
||||
A: Startkontingent Lyngmorennet | B: Barn opp til 12 år | C: 100 | D: 100 | E: 142 | F: startende 2025 | G: Uendret for 2026 og 2027
|
||||
A: Startkontingent Lyngmorennet | B: Ungdom 13–16 år | C: 150 | D: 150 | E: 61 | F: startende 2025 | G: Uendret for 2026 og 2027
|
||||
A: Startkontingent Lyngmorennet | B: Fra 17 år | C: 300 | D: 300 | E: 165 | F: startende 2025 | G: Uendret for 2026 og 2027
|
||||
A: Tillegg for etteranmelding | B: Alle klasser | C: 100 | D: 100 | E: 19 | F: etteranmeldte 2025 | G: Kommer i tillegg til startkontingenten
|
||||
A: Overnatting Granlihytta per natt | B: Medlem | C: 150 | D: 150 | E: 248 | F: gjestedøgn 2025 | G: Leietaker må være medlem og minst 25 år
|
||||
A: Overnatting Granlihytta per natt | B: Gjest av medlem | C: 250 | D: 250 | E: 64 | F: gjestedøgn 2025 | G: Bare sammen med et medlem
|
||||
A: Helgeleie hele Granlihytta | B: Kun medlemmer | C: 3500 | D: 3500 | E: 9 | F: helger 2025 | G: Fredag 16:00 til søndag 15:00
|
||||
A: Vaskegebyr Granlihytta | B: Leietaker | C: 600 | D: 600 | E: 2 | F: ganger 2025 | G: Når hytta ikke er vasket før avreise
|
||||
|
||||
#### Dugnadsliste 2025-26
|
||||
|
||||
A: Dato | B: Ukedag | C: Oppgave | D: Sted | E: Ansvarlig | F: Antall personer | G: Varighet per vakt (t) | H: Teller som vakter | I: Status | J: Merknad
|
||||
A: 13.09.2025 | B: lørdag | C: Rydding av kratt langs Storrunden | D: Lysløypa | E: Arnstein Vold | F: 14 | G: 4 | H: 1 | I: Utført | J: Ryddet 1,8 km
|
||||
A: 27.09.2025 | B: lørdag | C: Rydding og merking av Granlirunden | D: Granlirunden | E: Arnstein Vold | F: 11 | G: 5 | H: 1 | I: Utført | J: Ny merking ved Granlitjernet
|
||||
A: 11.10.2025 | B: lørdag | C: Kontroll av lysarmaturer | D: Storrunden og Barnerunden | E: Eivind Rustad | F: 4 | G: 4 | H: 1 | I: Utført | J: 6 armaturer skiftet
|
||||
A: 25.10.2025 | B: lørdag | C: Vedbæring og vask | D: Granlihytta | E: Kåre Lien | F: 9 | G: 4 | H: 1 | I: Utført
|
||||
A: 08.11.2025 | B: lørdag | C: Utlevering av utlånsski | D: Utstyrsboden | E: Marte Kvamsdal | F: 3 | G: 4 | H: 1 | I: Utført | J: 38 par levert ut
|
||||
A: 22.11.2025 | B: lørdag | C: Klargjøring av smøreboden | D: Lysløypa | E: Marte Kvamsdal | F: 5 | G: 4 | H: 1 | I: Utført
|
||||
A: 06.12.2025 | B: lørdag | C: Julegrøt for idrettsskolen | D: Klubbhuset | E: Hanne Sæther | F: 6 | G: 4 | H: 1 | I: Avlyst | J: Vannlekkasje i klubbhuset
|
||||
A: 04.01.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført | J: Sesongåpning
|
||||
A: 11.01.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført
|
||||
A: 18.01.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført
|
||||
A: 25.01.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 3 | G: 5 | H: 1 | I: Utført – manglet 1 person | J: Familien fakturert dugnadsavgift
|
||||
A: 01.02.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført
|
||||
A: 08.02.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført | J: Rennsdag, ekstra vaffelsalg
|
||||
A: 15.02.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført | J: Vinterferie
|
||||
A: 22.02.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført
|
||||
A: 01.03.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført
|
||||
A: 08.03.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 3 | G: 5 | H: 1 | I: Utført – manglet 1 person | J: Byttet uten å melde fra
|
||||
A: 15.03.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Utført
|
||||
A: 22.03.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Planlagt
|
||||
A: 29.03.2026 | B: søndag | C: Kafévakt 10:30–15:30 | D: Granlihytta | E: Wenche Tollefsrud | F: 4 | G: 5 | H: 1 | I: Planlagt | J: Siste kafésøndag
|
||||
A: 07.02.2026 | B: lørdag | C: Rigging av start og mål, Lyngmorennet | D: Lysløypa | E: Øystein Brekke | F: 18 | G: 6 | H: 1 | I: Utført | J: Fra 2026–27 teller en slik vakt som to
|
||||
A: 08.02.2026 | B: søndag | C: Funksjonær under Lyngmorennet | D: Lysløypa og Granlirunden | E: Øystein Brekke | F: 83 | G: 6 | H: 1 | I: Utført | J: 412 startende; teller som to vakter fra 2026–27
|
||||
A: 08.02.2026 | B: søndag | C: Nedrigging etter Lyngmorennet | D: Lysløypa | E: Øystein Brekke | F: 12 | G: 3 | H: 1 | I: Utført | J: Kortere vakt
|
||||
A: 18.04.2026 | B: lørdag | C: Innsamling av merkebånd | D: Granlirunden | E: Arnstein Vold | F: 6 | G: 4 | H: 1 | I: Planlagt
|
||||
A: 28.04.2026 | B: tirsdag | C: Innlevering og sortering av utlånsski | D: Utstyrsboden | E: Marte Kvamsdal | F: 4 | G: 4 | H: 1 | I: Planlagt | J: Frist for levering er 30. april
|
||||
|
||||
#### Løypepreparering vinter 2026
|
||||
|
||||
A: Dato | B: Ukedag | C: Løype | D: Preparert lengde (km) | E: Skøytetrasé | F: Snødybde (cm) | G: Temperatur (°C) | H: Timer brukt (t) | I: Utført av | J: Merknad
|
||||
A: 02.01.2026 | B: fredag | C: Storrunden | D: 0 | E: Nei | F: 17 | G: -4.5 | H: 0 | I: Arnstein Vold | J: Ikke preparert – for lite snø (krav 20 cm)
|
||||
A: 06.01.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 24 | G: -8 | H: 2 | I: Arnstein Vold | J: Første preparering i 2026
|
||||
A: 06.01.2026 | B: tirsdag | C: Barnerunden | D: 1.1 | E: Nei | F: 24 | G: -8 | H: 0.5 | I: Arnstein Vold
|
||||
A: 08.01.2026 | B: torsdag | C: Storrunden | D: 4.2 | E: Ja | F: 26 | G: -11.5 | H: 2 | I: Torgeir Haugland
|
||||
A: 10.01.2026 | B: lørdag | C: Granlirunden | D: 0 | E: Nei | F: 27 | G: -6 | H: 0 | I: Torgeir Haugland | J: Ikke preparert – for lite snø (krav 30 cm)
|
||||
A: 10.01.2026 | B: lørdag | C: Storrunden | D: 4.2 | E: Ja | F: 27 | G: -6 | H: 2 | I: Torgeir Haugland
|
||||
A: 13.01.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 31 | G: -3 | H: 2 | I: Arnstein Vold
|
||||
A: 13.01.2026 | B: tirsdag | C: Barnerunden | D: 1.1 | E: Nei | F: 31 | G: -3 | H: 0.5 | I: Arnstein Vold
|
||||
A: 15.01.2026 | B: torsdag | C: Hundesporet | D: 3 | E: Nei | F: 33 | G: -2.5 | H: 1 | I: Liv Marit Sundby
|
||||
A: 17.01.2026 | B: lørdag | C: Granlirunden | D: 9.5 | E: Ja | F: 34 | G: -9 | H: 4.5 | I: Torgeir Haugland | J: Første preparering av Granlirunden
|
||||
A: 17.01.2026 | B: lørdag | C: Storrunden | D: 4.2 | E: Ja | F: 34 | G: -9 | H: 2 | I: Torgeir Haugland
|
||||
A: 20.01.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 36 | G: -16.5 | H: 2.5 | I: Arnstein Vold | J: Tung snø etter vindfall
|
||||
A: 22.01.2026 | B: torsdag | C: Barnerunden | D: 1.1 | E: Nei | F: 36 | G: -14 | H: 0.5 | I: Liv Marit Sundby
|
||||
A: 24.01.2026 | B: lørdag | C: Granlirunden | D: 9.5 | E: Ja | F: 38 | G: -7 | H: 4.5 | I: Torgeir Haugland
|
||||
A: 27.01.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 41 | G: -5.5 | H: 2 | I: Arnstein Vold
|
||||
A: 31.01.2026 | B: lørdag | C: Storrunden | D: 4.2 | E: Ja | F: 43 | G: -10 | H: 2 | I: Liv Marit Sundby
|
||||
A: 03.02.2026 | B: tirsdag | C: Hundesporet | D: 3 | E: Nei | F: 44 | G: -12.5 | H: 1 | I: Liv Marit Sundby
|
||||
A: 05.02.2026 | B: torsdag | C: Barnerunden | D: 1.1 | E: Nei | F: 45 | G: -8.5 | H: 0.5 | I: Arnstein Vold
|
||||
A: 07.02.2026 | B: lørdag | C: Storrunden | D: 4.2 | E: Ja | F: 46 | G: -6 | H: 2.5 | I: Arnstein Vold | J: Klargjort for Lyngmorennet
|
||||
A: 07.02.2026 | B: lørdag | C: Granlirunden | D: 9.5 | E: Ja | F: 46 | G: -6 | H: 5 | I: Torgeir Haugland | J: Klargjort for Lyngmorennet
|
||||
A: 12.02.2026 | B: torsdag | C: Storrunden | D: 4.2 | E: Ja | F: 47 | G: -2 | H: 2 | I: Arnstein Vold
|
||||
A: 14.02.2026 | B: lørdag | C: Granlirunden | D: 9.5 | E: Ja | F: 49 | G: -4 | H: 4.5 | I: Torgeir Haugland | J: Vinterferie
|
||||
A: 19.02.2026 | B: torsdag | C: Barnerunden | D: 1.1 | E: Nei | F: 50 | G: -1 | H: 0.5 | I: Liv Marit Sundby
|
||||
A: 24.02.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 48 | G: 1.5 | H: 2 | I: Arnstein Vold | J: Mildvær, sporet frøs til om kvelden
|
||||
A: 28.02.2026 | B: lørdag | C: Granlirunden | D: 9.5 | E: Ja | F: 44 | G: -3.5 | H: 4.5 | I: Torgeir Haugland
|
||||
A: 03.03.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 42 | G: -5 | H: 2 | I: Arnstein Vold
|
||||
A: 07.03.2026 | B: lørdag | C: Hundesporet | D: 3 | E: Nei | F: 40 | G: -7.5 | H: 1 | I: Liv Marit Sundby
|
||||
A: 10.03.2026 | B: tirsdag | C: Storrunden | D: 4.2 | E: Ja | F: 37 | G: 0.5 | H: 2 | I: Arnstein Vold
|
||||
A: 14.03.2026 | B: lørdag | C: Granlirunden | D: 0 | E: Nei | F: 29 | G: 2 | H: 0 | I: Torgeir Haugland | J: Ikke preparert – under 30 cm på de høyeste partiene
|
||||
A: 14.03.2026 | B: lørdag | C: Storrunden | D: 4.2 | E: Ja | F: 29 | G: 2 | H: 2 | I: Torgeir Haugland | J: Siste skøytetrasé denne sesongen
|
||||
257
gates/calibration-folder/kilde/medlemshaandbok-2026.docx.md
Normal file
257
gates/calibration-folder/kilde/medlemshaandbok-2026.docx.md
Normal file
|
|
@ -0,0 +1,257 @@
|
|||
# Medlemshåndbok for sesongen 2026
|
||||
|
||||
# Medlemshåndbok for sesongen 2026
|
||||
|
||||
Denne håndboka gjelder for alle andelshavere i Kvernlia Andelslandbruk SA fra 1. april til 31. desember 2026 og erstatter håndboka for 2025. Den ble vedtatt av styret 9. februar 2026. Regler som er endret fra 2025, er merket med «Nytt i 2026». Har du spørsmål om reglene, kan du spørre gartneren på hentedagene.
|
||||
|
||||
## 1. Om andelslaget
|
||||
|
||||
Kvernlia Andelslandbruk ble startet i 2019 og driver på leid jord ved Kvernlia gård. I 2026 dyrker vi 1,8 hektar friland fordelt på åtte skifter, A til H, og har et drivhus på 180 m² og et nytt tunnelhus på 120 m². Andelslaget har 34 høner og seks bikuber.
|
||||
|
||||
Gartner Ingrid Solhaug-Berntsen (100 prosent stilling) leder dyrkingen og oppdaterer høstetavla.
|
||||
|
||||
Gartnerassistent Mira Holten (60 prosent stilling, opp fra 40 prosent i 2025) har ansvar for drivhuset og tunnelhuset.
|
||||
|
||||
Styreleder Torbjørn Aasvang svarer for styret og medlemsmøtene.
|
||||
|
||||
Kasserer Halvard Moe-Fjeld har ansvar for betaling, refusjon og frikjøp.
|
||||
|
||||
Vaktansvarlig Sunniva Rakkestad fører vaktlista og har ansvar for hønsegården.
|
||||
|
||||
Birøkter Eivind Lauvstad har ansvar for bigården og arbeider frivillig.
|
||||
|
||||
## 2. Andeler, betaling og oppsigelse
|
||||
|
||||
Andelslaget har 148 andeler i 2026: 112 helandeler og 36 halvandeler. I tillegg er det 40 eggandeler. Andelen betales på forskudd, fordi frø, planter og lønn må betales før første henting.
|
||||
|
||||
Andelstype
|
||||
|
||||
Pris 2026 (kr)
|
||||
|
||||
Mengde
|
||||
|
||||
Arbeidsplikt (timer)
|
||||
|
||||
Betalingsfrist
|
||||
|
||||
Helandel
|
||||
|
||||
4 400
|
||||
|
||||
Full mengde på høstetavla
|
||||
|
||||
12
|
||||
|
||||
\15. april, eller i to like rater 15. april og 15. juli
|
||||
|
||||
Halvandel
|
||||
|
||||
2 600
|
||||
|
||||
Halv mengde på høstetavla
|
||||
|
||||
8
|
||||
|
||||
\15. april, kun én rate
|
||||
|
||||
Eggandel (tillegg)
|
||||
|
||||
650
|
||||
|
||||
10 egg per uke
|
||||
|
||||
0
|
||||
|
||||
\1. mai
|
||||
|
||||
Andelshaveren må være minst 18 år.
|
||||
|
||||
Eggandel kan bare tegnes av den som har hel- eller halvandel. Påmeldingsfristen er 1. mai, og de 40 plassene fordeles etter ventelisteplass.
|
||||
|
||||
Er andelen ikke betalt innen 14 dager etter fristen, går den til neste person på ventelista.
|
||||
|
||||
Andelen refunderes ikke etter 15. april. Unntaket er flytting mer enn 50 km fra gården før 1. juli; da refunderes 50 prosent.
|
||||
|
||||
Andelen kan overdras til en annen person for resten av sesongen, men bare etter skriftlig melding til kassereren.
|
||||
|
||||
Vil du fornye andelen for 2027, må du melde fra innen 1. desember 2026. Ellers går plassen til ventelista.
|
||||
|
||||
## 3. Henting og innhøsting
|
||||
|
||||
Sesongen 2026 har 21 hentuker, fra uke 23 til uke 43. Første henting er tirsdag 2. juni og siste henting lørdag 24. oktober. Hentetidene er tirsdag kl. 16.00–20.00 og lørdag kl. 09.00–13.00. Nytt i 2026: lørdag erstatter fredag som hentedag. Hver andel henter én gang per uke, enten tirsdag eller lørdag.
|
||||
|
||||
Høstetavla henger ved redskapsboden og oppdateres av gartneren senest én time før hentingen starter. Tavla viser hva som kan høstes, i hvilket skifte, og hvor mye en helandel skal ta. Halvandeler tar halvparten av mengden på tavla.
|
||||
|
||||
Du skal bare høste det som står på tavla, og bare den mengden som står der.
|
||||
|
||||
Grønt flagg betyr at du kan høste fritt innenfor mengden på tavla.
|
||||
|
||||
Gult flagg betyr at du bare skal skjære med kniv eller saks, minst 3 cm over bakken, slik at planten vokser videre. Det gjelder blant annet salat og urter.
|
||||
|
||||
Rødt flagg betyr at bedet er nysådd eller dekket med fiberduk, og at det ikke skal høstes der.
|
||||
|
||||
Jordbær: høyst 1 liter per helandel og 0,5 liter per halvandel per henting.
|
||||
|
||||
Blomster: én bukett per helandel hver uke. Halvandeler kan plukke en bukett i uker med partall.
|
||||
|
||||
Urtehagen i skifte F kan høstes fritt i små mengder, unntatt i bed med rødt flagg.
|
||||
|
||||
Medlemmer høster ikke selv i drivhuset eller tunnelhuset. Derfra settes grønnsakene fram ferdig plukket i kasser ved hentebordet.
|
||||
|
||||
Poteter og lagringsgrønnsaker graves av gartnerne fra uke 38 og deles ut i sekker.
|
||||
|
||||
Kniver, sakser og kurver lånes i redskapsboden og skal leveres vasket før du går.
|
||||
|
||||
Barn under 12 år kan være med og høste, men bare sammen med en voksen. Hunder er ikke tillatt på åkeren. De kan bindes ved benken innenfor porten, men aldri i nærheten av hønsegården.
|
||||
|
||||
## 4. Andelsvenn, glemt henting og ferie
|
||||
|
||||
Kan du ikke hente selv, kan en andelsvenn hente for deg. Du må melde navnet til gartneren innen kl. 12.00 samme dag. Andelsvennen må følge de samme reglene som deg, og en andelsvenn under 18 år må ha med seg en voksen.
|
||||
|
||||
Hentes ikke andelen innen hentetiden er slutt, faller den bort for den uka. Det som står igjen, høstes av gartnerne og gis til Nabolagskjøkkenet ved Dorvikstjern. Andelen kan ikke tas igjen en senere uke.
|
||||
|
||||
Pakket andel: Melder du fra minst 48 timer før hentingen starter, pakker gartnerne andelen din. Den står i kjølerommet og må hentes innen kl. 21.00 samme dag.
|
||||
|
||||
Du kan få pakket andel høyst tre ganger per sesong.
|
||||
|
||||
Det gis ikke pause eller refusjon for ferie. I ferien kan du i stedet la en andelsvenn hente.
|
||||
|
||||
## 5. Arbeidsplikt og vaktordning
|
||||
|
||||
Nytt i 2026: Arbeidsplikten er 12 timer for helandel og 8 timer for halvandel, mot 10 og 6 timer i 2025. Plikten er økt fordi arealet er større, og fordi mange åkervakter sto tomme i juli i fjor. Timene skal gjøres mellom 1. mai og 31. oktober. Minst halvparten av timene skal være booket innen 1. juni.
|
||||
|
||||
Vaktlista henger på tavla i redskapsboden. Når vakta er gjort, kvitterer du i vaktboka, og vaktansvarlig fører timene inn hver mandag.
|
||||
|
||||
Vakttype
|
||||
|
||||
Tidspunkt
|
||||
|
||||
Teller (timer)
|
||||
|
||||
Krav
|
||||
|
||||
Aldersgrense
|
||||
|
||||
Åkervakt
|
||||
|
||||
Tirsdag og torsdag 08.00–12.00, lørdag 13.00–17.00
|
||||
|
||||
4
|
||||
|
||||
Ingen
|
||||
|
||||
16 år alene, 12–15 år sammen med en voksen
|
||||
|
||||
Hentevakt
|
||||
|
||||
Tirsdag 15.30–20.00, lørdag 08.30–13.00
|
||||
|
||||
4,5
|
||||
|
||||
Gjennomgang på 30 minutter første gang
|
||||
|
||||
18 år
|
||||
|
||||
Drivhusvakt
|
||||
|
||||
Onsdag og søndag 17.00–19.00
|
||||
|
||||
2
|
||||
|
||||
Ingen
|
||||
|
||||
12 år
|
||||
|
||||
Hønsevakt
|
||||
|
||||
Kveld alle dager 19.30–21.00, morgen lørdag og søndag 07.00–08.30
|
||||
|
||||
1,5
|
||||
|
||||
Hønsekurs
|
||||
|
||||
15 år alene, 10–14 år sammen med en voksen
|
||||
|
||||
Dugnad
|
||||
|
||||
Lørdag 9. mai og lørdag 31. oktober, 10.00–13.00
|
||||
|
||||
3
|
||||
|
||||
Ingen
|
||||
|
||||
Ingen
|
||||
|
||||
Timer som barn mellom 12 og 17 år gjør, teller på husstandens andel. Timer gjort av barn under 12 år teller ikke.
|
||||
|
||||
En vakt må avbestilles minst 48 timer før den starter. Avbestiller du senere, teller vakta som ikke utført, med mindre du selv skaffer en erstatter som er medlem.
|
||||
|
||||
Nytt i 2026: Frikjøp koster 300 kr per time, og hver andel kan kjøpe seg fri fra høyst 4 timer.
|
||||
|
||||
Mangler du mer enn 4 timer per 31. oktober, faktureres de timene som går ut over 4, med 600 kr per time, og du mister fortrinnsretten til å fornye andelen.
|
||||
|
||||
Faktura for frikjøp og manglende timer sendes ut 15. november.
|
||||
|
||||
## 6. Hønsegården
|
||||
|
||||
Bare medlemmer som har tatt hønsekurset, kan ta hønsevakt. Kurset holdes lørdag 18. april og lørdag 25. april kl. 10.00–12.00, med plass til høyst 12 deltakere per kurs. Nytt i 2026: medlemmene tar også morgenvakta lørdag og søndag. Morgenvakta på hverdager tas fortsatt av gartnerne.
|
||||
|
||||
Kveldsvakta skal lukke alle tre lukene, telle hønene, samle egg, fylle vann og kontrollere at lampa på strømgjerdet lyser.
|
||||
|
||||
Morgenvakta åpner lukene tidligst kl. 07.00, gir fôr og samler egg.
|
||||
|
||||
Hønene skal bare ha fôr fra sekkene i boden og grønnsaksavfall fra gården, aldri matrester hjemmefra.
|
||||
|
||||
Egg som samles inn på vakt, legges i kjølerommet. Vaktene har ikke lov til å ta med egg hjem.
|
||||
|
||||
Eggandelene hentes i kjølerommet på hentedagene.
|
||||
|
||||
Rømte høner, skader eller hull i gjerdet meldes til vaktansvarlig samme kveld.
|
||||
|
||||
## 7. Bigården
|
||||
|
||||
Bigården ligger bak skifte F og er merket med gule skilt. Fordi to medlemmer ble stukket i 2025, er reglene for bigården skjerpet og bedre skiltet.
|
||||
|
||||
Hold minst 10 meter avstand til kubene, med mindre du er i følge med birøkteren eller en av de to birøkthjelperne.
|
||||
|
||||
Barn under 16 år skal ikke inn i bigården, heller ikke i følge med voksne.
|
||||
|
||||
Kubene åpnes bare av birøkteren og birøkthjelperne, og aldri mens hentingen pågår.
|
||||
|
||||
Medlemmer med kjent allergi mot bistikk skal melde fra til gartneren før første henting.
|
||||
|
||||
Førstehjelpsskrinet henger på innsiden av døra til redskapsboden.
|
||||
|
||||
Hver andel, både hel- og halvandel, får ett glass honning på 250 gram i september.
|
||||
|
||||
## 8. Drivhus og tunnelhus
|
||||
|
||||
Medlemmer går bare inn i drivhuset og tunnelhuset når de har drivhusvakt, eller sammen med en av gartnerne. Barn under 8 år skal ikke inn i drivhuset eller tunnelhuset. Tunnelhuset settes opp på vårdugnaden 9. mai og skal være i bruk fra uke 21.
|
||||
|
||||
Dørene skal lukkes om kvelden når det er meldt under 12 °C om natta.
|
||||
|
||||
Takluker og dører åpnes når temperaturen inne er over 26 °C.
|
||||
|
||||
Vask støvlene i fotbadet ved døra før du går inn, fordi sykdommer på tomat og potet smitter med jord.
|
||||
|
||||
Vanningen går automatisk kl. 07.00. Drivhusvakta vanner bare for hånd når gartneren har skrevet det på tavla inne i drivhuset.
|
||||
|
||||
## 9. Fellesarrangementer
|
||||
|
||||
Andelslaget har to dugnader, én høstfest og ett medlemsmøte i året. Dugnadene teller på arbeidsplikten, men høstfesten og medlemsmøtet gjør det ikke.
|
||||
|
||||
Vårdugnaden er lørdag 9. mai kl. 10.00–13.00. Da settes tunnelhuset opp, og poteter settes i skifte H.
|
||||
|
||||
Høstfesten er lørdag 26. september fra kl. 15.00 i låven. Hver husstand kan ta med høyst fire gjester.
|
||||
|
||||
Høstdugnaden er lørdag 31. oktober kl. 10.00–13.00. Da ryddes åkeren, og drivhuset vaskes ned for vinteren.
|
||||
|
||||
Medlemsmøtet før sesongen holdes i mars. Der legger styret fram sesongevalueringen, budsjettet og endringer i reglene.
|
||||
|
||||
Hver andel har én stemme på medlemsmøtet, uansett om det er en hel- eller halvandel.
|
||||
|
||||
Det er ikke lov å grille eller gjøre opp ild på gården, med unntak av bålpanna ved låven på høstfesten. Røyking er ikke tillatt noe sted på gården.
|
||||
|
||||
## 10. Avvik og kontakt
|
||||
|
||||
Skader, stikk, rømte dyr og ødelagt utstyr skal meldes til gartneren samme dag og føres i avviksboka i redskapsboden. Spørsmål om betaling og refusjon går til kassereren, og spørsmål om vakter går til vaktansvarlig. Endringer i reglene kan bare vedtas av styret, og de skal legges fram på et medlemsmøte.
|
||||
189
gates/calibration-folder/kilde/medlemsmote-mars-2026.pptx.md
Normal file
189
gates/calibration-folder/kilde/medlemsmote-mars-2026.pptx.md
Normal file
|
|
@ -0,0 +1,189 @@
|
|||
## Medlemsmøte før sesongen 2026
|
||||
|
||||
## Velkommen til medlemsmøtet 2026
|
||||
|
||||
Møtet holdes i låven torsdag 12. mars 2026 kl. 18.30.
|
||||
|
||||
Vi ser tilbake på sesongen 2025 og går gjennom planen for 2026.
|
||||
|
||||
Styret legger fram endringene som gjelder fra den nye sesongen.
|
||||
|
||||
Det blir tid til spørsmål etter hver del.
|
||||
|
||||
## Sesongen 2025 i tall
|
||||
|
||||
131 andeler hentet grønnsaker i 19 uker, fra 10. juni til 17. oktober.
|
||||
|
||||
Vi høstet 14 200 kg, som er 96 prosent av planen på 14 800 kg.
|
||||
|
||||
En helandel fikk i snitt 124 kg grønnsaker, bær og urter.
|
||||
|
||||
I snitt hentet 85 prosent av andelene hver uke.
|
||||
|
||||
610 kg som ikke ble hentet, gikk til Nabolagskjøkkenet.
|
||||
|
||||
## Det som gikk dårlig i 2025
|
||||
|
||||
Tørråte i skifte D ga bare 1 900 kg poteter, 700 kg under planen.
|
||||
|
||||
Brunsnegl i skifte B kostet oss 240 kg salat.
|
||||
|
||||
Nattefrost 14. mai skadet en tredel av jordbærplantene.
|
||||
|
||||
En rev tok tre høner fordi en luke sto åpen etter kveldsvakta.
|
||||
|
||||
## Det som gikk bra i 2025
|
||||
|
||||
Drivhuset ga 2 350 kg, 150 kg mer enn planlagt.
|
||||
|
||||
Squash og gulrøtter ga de største avlingene vi har hatt.
|
||||
|
||||
Fire bikuber ga 85 kg honning.
|
||||
|
||||
Regnskapet viste et overskudd på 16 320 kr.
|
||||
|
||||
## Hva medlemmene svarte
|
||||
|
||||
104 av 131 andeler svarte på undersøkelsen i november.
|
||||
|
||||
81 prosent var fornøyd eller svært fornøyd, mot 76 prosent året før.
|
||||
|
||||
64 prosent ønsket henting på lørdag i stedet for fredag.
|
||||
|
||||
Det vanligste ønsket var jevnere tilgang på salat.
|
||||
|
||||
## Andeler i 2026
|
||||
|
||||
Vi får 148 andeler: 112 helandeler og 36 halvandeler.
|
||||
|
||||
119 andeler fornyet, og 29 kommer inn fra ventelista.
|
||||
|
||||
Eggandelene økes fra 30 til 40.
|
||||
|
||||
Andelshaveren må være minst 18 år.
|
||||
|
||||
## Priser og betaling
|
||||
|
||||
Helandel koster 4 400 kr, mot 4 100 kr i fjor.
|
||||
|
||||
Halvandel koster 2 600 kr, mot 2 400 kr i fjor.
|
||||
|
||||
Eggandel koster 650 kr og må tegnes innen 1. mai.
|
||||
|
||||
Helandel kan betales i to rater, 15. april og 15. juli.
|
||||
|
||||
Halvandel betales i én rate innen 15. april.
|
||||
|
||||
## Ny hentedag: lørdag
|
||||
|
||||
Fredag utgår som hentedag fordi flertallet ønsket lørdag.
|
||||
|
||||
Lørdag hentes det kl. 09.00–13.00.
|
||||
|
||||
Tirsdag beholdes med henting kl. 16.00–20.00.
|
||||
|
||||
Hver andel henter fortsatt én gang per uke.
|
||||
|
||||
## Sesongen 2026
|
||||
|
||||
Sesongen får 21 hentuker, fra uke 23 til uke 43.
|
||||
|
||||
Første henting er tirsdag 2. juni.
|
||||
|
||||
Siste henting er lørdag 24. oktober.
|
||||
|
||||
Målet er en samlet avling på 18 600 kg.
|
||||
|
||||
## Arbeidsplikten øker
|
||||
|
||||
Helandeler skal gjøre 12 timer, mot 10 timer i 2025.
|
||||
|
||||
Halvandeler skal gjøre 8 timer, mot 6 timer i 2025.
|
||||
|
||||
I 2025 sto 11 av 32 åkervakter tomme i ukene 28 til 31.
|
||||
|
||||
Minst halvparten av timene skal være booket innen 1. juni.
|
||||
|
||||
## Tak på frikjøp
|
||||
|
||||
Frikjøp koster 300 kr per time, mot 250 kr i fjor.
|
||||
|
||||
Hver andel kan kjøpe seg fri fra høyst 4 timer.
|
||||
|
||||
Timer som mangler ut over 4, faktureres med 600 kr per time.
|
||||
|
||||
Ni helandeler kjøpte seg helt fri i 2025, og det ga tomme vakter i juli.
|
||||
|
||||
## Vaktene i 2026
|
||||
|
||||
En åkervakt varer 4 timer og en hentevakt 4,5 timer.
|
||||
|
||||
En drivhusvakt varer 2 timer og en hønsevakt 1,5 timer.
|
||||
|
||||
Medlemmene tar også morgenvakta i hønsegården i helgene.
|
||||
|
||||
Vårdugnaden er lørdag 9. mai og teller 3 timer.
|
||||
|
||||
En vakt må avbestilles minst 48 timer før den starter.
|
||||
|
||||
## Mer jord: skifte H
|
||||
|
||||
Vi leier 0,3 hektar av Arne Dyrstad for 6 000 kr per år.
|
||||
|
||||
Dyrket areal øker fra 1,5 til 1,8 hektar.
|
||||
|
||||
Potetene flyttes til skifte H for å redusere risikoen for tørråte.
|
||||
|
||||
Skifte G ligger i grønngjødsling med kløver også i 2026.
|
||||
|
||||
## Nytt tunnelhus
|
||||
|
||||
Tunnelhuset blir 120 m² og står ved siden av drivhuset på 180 m².
|
||||
|
||||
Det koster 64 000 kr, og 26 000 kr tas fra reservefondet.
|
||||
|
||||
Tunnelhuset settes opp på vårdugnaden 9. mai.
|
||||
|
||||
Der skal vi dyrke tomat, chili og et forsøk med sukkermelon.
|
||||
|
||||
## Hønene
|
||||
|
||||
Vi kjøper inn 9 unghøner i april, slik at flokken blir på 34.
|
||||
|
||||
Et strømgjerde på 1,2 meter står nå rundt hønsegården.
|
||||
|
||||
Hønsekurs holdes 18. april og 25. april kl. 10.00–12.00.
|
||||
|
||||
Bare medlemmer med hønsekurs kan ta hønsevakt.
|
||||
|
||||
## Biene
|
||||
|
||||
To kuber deles i vår, så vi får seks bikuber.
|
||||
|
||||
Målet for honning i 2026 er 110 kg.
|
||||
|
||||
Hold minst 10 meter avstand til kubene uten følge av birøkteren.
|
||||
|
||||
Barn under 16 år skal ikke inn i bigården.
|
||||
|
||||
## Budsjett 2026
|
||||
|
||||
Budsjettet har inntekter på 672 400 kr og utgifter på 661 000 kr.
|
||||
|
||||
Det gir et planlagt overskudd på 11 400 kr.
|
||||
|
||||
Lønn er budsjettert til 418 000 kr for 1,6 årsverk.
|
||||
|
||||
Frikjøp er budsjettert til 18 000 kr, mot 33 000 kr i 2025.
|
||||
|
||||
## Viktige datoer
|
||||
|
||||
Betalingsfristen for andelene er 15. april.
|
||||
|
||||
Påmeldingsfristen for eggandel er 1. mai.
|
||||
|
||||
Høstfesten er lørdag 26. september.
|
||||
|
||||
Høstdugnaden er lørdag 31. oktober.
|
||||
|
||||
Fornyelse for 2027 må meldes innen 1. desember.
|
||||
|
|
@ -0,0 +1,199 @@
|
|||
## Brantwater Sailing Club - New Members' Welcome Evening
|
||||
|
||||
## Welcome to Brantwater Sailing Club
|
||||
|
||||
This welcome evening is on Friday 13 November 2026 at 19:30 in the main room of the clubhouse.
|
||||
|
||||
The club was founded in 1971 and sails on Brantwater, next to Brantholme village.
|
||||
|
||||
Tonight we cover fees, facilities, training, racing, safety and your duties.
|
||||
|
||||
Questions are welcome at any point, and there is a drink on the club at the bar afterwards.
|
||||
|
||||
## The club in numbers
|
||||
|
||||
We had 312 members on 30 September 2026, up from 296 a year earlier.
|
||||
|
||||
47 new members joined during the 2025-26 club year.
|
||||
|
||||
The club owns 15 sailing dinghies and three powerboats.
|
||||
|
||||
We have 38 swinging moorings, 64 dinghy park spaces and 24 rack spaces.
|
||||
|
||||
Members completed 533 duties in the 2026 season.
|
||||
|
||||
## Who's who on the committee
|
||||
|
||||
Helen Marchbank is Commodore and chairs the committee.
|
||||
|
||||
Priya Castellan is Rear Commodore (Sailing) and runs the racing programme.
|
||||
|
||||
Tomasz Wilde is Principal Instructor, and Kerry Ashdown coordinates the juniors.
|
||||
|
||||
Fiona Brackley is Safety Officer, and Neil Harrowby is Duty Coordinator.
|
||||
|
||||
Graham Oakeshott is Treasurer, and Ruth Penhallow is Secretary until the AGM.
|
||||
|
||||
## Your membership in 2026
|
||||
|
||||
A full adult membership costs £285 and a family membership costs £430.
|
||||
|
||||
Junior membership costs £95, young adult (18 to 25) £140 and social membership £75.
|
||||
|
||||
The £60 joining fee is not charged for junior and young adult memberships.
|
||||
|
||||
Subscriptions fall due on 1 March each year.
|
||||
|
||||
If you joined on or after 1 September, you paid half the annual subscription.
|
||||
|
||||
## What the AGM will decide on 21 November
|
||||
|
||||
The AGM starts at 19:00 on Saturday 21 November 2026 in the main room.
|
||||
|
||||
The committee proposes a full adult rate of £298 and a family rate of £450 from 1 March 2027.
|
||||
|
||||
The young adult rate would rise to £145, while junior and social rates would stay the same.
|
||||
|
||||
The AGM is also asked to approve £7,390 from the general reserve for the slipway repair.
|
||||
|
||||
## The clubhouse and bar
|
||||
|
||||
In summer the bar opens on Wednesdays 18:00 to 22:30, Saturdays 12:00 to 18:00 and Sundays 12:00 to 17:00.
|
||||
|
||||
From 1 November to 31 March the bar opens only on Sundays, from 12:00 to 16:00.
|
||||
|
||||
You may sign in up to 3 guests a day, and the same guest up to 6 times a season.
|
||||
|
||||
Members under 18 may stay in the main room after 20:00 only with a parent or guardian.
|
||||
|
||||
The Lookout Room upstairs is the race office and training room, and only water is allowed there.
|
||||
|
||||
## Storing your boat
|
||||
|
||||
A dinghy park space costs £95 and a rack space £45 for the year from 1 April to 31 March.
|
||||
|
||||
Every boat and trolley must be labelled with your name and membership number.
|
||||
|
||||
A swinging mooring costs £210 for a boat up to 6.5 metres and £265 for a longer boat.
|
||||
|
||||
Mooring applications must reach the Moorings Officer by 31 January.
|
||||
|
||||
There is a waiting list of nine members for moorings, so a mooring is unlikely before 2028.
|
||||
|
||||
## Using the club boats
|
||||
|
||||
Club dinghies are free to use for members who hold club Level 2 or above.
|
||||
|
||||
A booking may last up to 3 hours, or 2 hours on weekends between 10:00 and 16:00.
|
||||
|
||||
The two Osprey 16 trainers are kept for courses on course days.
|
||||
|
||||
Write any damage in the defects book in the boat shed before you leave the club.
|
||||
|
||||
## Learning to sail as an adult
|
||||
|
||||
Start Sailing is a two-weekend course sailed in the Osprey 16 trainers.
|
||||
|
||||
The next Start Sailing course runs on 17-18 and 24-25 April 2027.
|
||||
|
||||
The Improver course runs over four evenings.
|
||||
|
||||
In 2026, 18 adults completed Start Sailing and 13 completed the Improver course.
|
||||
|
||||
## Juniors at Brantwater
|
||||
|
||||
Junior training runs on Saturdays from 09:30 to 12:30, from May to September.
|
||||
|
||||
Juniors must be at least 8 years old and able to swim 50 metres in clothes with a buoyancy aid.
|
||||
|
||||
The season fee is £120, and only junior members can take part.
|
||||
|
||||
41 juniors trained in 2026, and the first session of 2027 is on Saturday 1 May.
|
||||
|
||||
Juniors under 14 must be collected from the clubhouse by a parent or guardian.
|
||||
|
||||
## Racing for everyone
|
||||
|
||||
All classes race together in one fleet on the club's own yardstick handicap.
|
||||
|
||||
The Spring Series runs on Sundays, with the first warning signal at 10:30.
|
||||
|
||||
The Summer Evening Series runs on Wednesdays, with the first warning signal at 18:45.
|
||||
|
||||
The 2027 Spring Series starts on Sunday 11 April and ends on Sunday 30 May.
|
||||
|
||||
In 2026, the Open Regatta drew 37 boats, 15 of them from six visiting clubs.
|
||||
|
||||
## Safety on the water
|
||||
|
||||
Everyone afloat must wear a buoyancy aid, and juniors must wear one on the pontoons too.
|
||||
|
||||
Juniors do not sail when the mean wind is over 18 knots; adults may not take club boats out above 22 knots.
|
||||
|
||||
When the Officer of the Day flies the red flag, no boat may launch.
|
||||
|
||||
Nobody sails after sunset, and from November to March single-handers sail only when a safety boat is staffed.
|
||||
|
||||
Report any incident to the Safety Officer within 24 hours, using the book in the race office.
|
||||
|
||||
## Your duties
|
||||
|
||||
A full adult member does 3 duties a season, a family membership 4 and a young adult 2.
|
||||
|
||||
Junior and social members, and members aged 75 or over, have no duties.
|
||||
|
||||
A safety boat helm must be 18 or over and hold the Club Powerboat Certificate; crew must be 16 or over.
|
||||
|
||||
If you cannot do a duty, arrange a swap and tell Neil Harrowby at least 48 hours before.
|
||||
|
||||
## The missed duty charge is going up
|
||||
|
||||
In 2026, 31 of 564 duties were missed without a swap.
|
||||
|
||||
One Spring Series race was cancelled because no safety boat helm was available.
|
||||
|
||||
For that reason the charge for a missed duty rises from £35 to £50 from 1 April 2027.
|
||||
|
||||
The number of duties you owe does not change, and a notified swap still avoids the charge.
|
||||
|
||||
## The slipway repair this winter
|
||||
|
||||
The lower 9 metres of the slipway cracked and settled by up to 70 millimetres after the February 2026 high water.
|
||||
|
||||
The contractor will recast the lower 12 metres from 4 January to 26 February 2027.
|
||||
|
||||
The slipway is closed to everyone from 2 January to 28 February 2027.
|
||||
|
||||
Members cannot launch their own boats during the closure; the safety boats will use the north beach pontoon.
|
||||
|
||||
The total budget is £38,390, including a 10% contingency.
|
||||
|
||||
## Winter at the club
|
||||
|
||||
The Frostbite Series runs on Sundays from 8 November to 20 December 2026, with first warning at 11:00.
|
||||
|
||||
This winter there are seven Frostbite races instead of the usual sixteen, because of the slipway works.
|
||||
|
||||
Boats must be off the swinging moorings by 15 November.
|
||||
|
||||
All six winter hardstanding spaces are booked until 31 March 2027.
|
||||
|
||||
## Key dates for your diary
|
||||
|
||||
Nominations for the committee close at 19:00 on Saturday 14 November 2026.
|
||||
|
||||
The AGM is on Saturday 21 November 2026 at 19:00.
|
||||
|
||||
Subscriptions are due on 1 March 2027, and storage fees on 1 April 2027.
|
||||
|
||||
The fitting-out work party is on Saturday 27 March 2027, and the season opening day is Saturday 10 April.
|
||||
|
||||
## Questions and contacts
|
||||
|
||||
Ask any committee member tonight, or leave a note in the Secretary's tray in the race office.
|
||||
|
||||
Duty swaps go to Neil Harrowby, and mooring questions go to the Moorings Officer.
|
||||
|
||||
Junior questions go to Kerry Ashdown, and course questions to Tomasz Wilde.
|
||||
|
||||
The rota for spring and summer 2027 will be published by 1 March 2027.
|
||||
|
|
@ -0,0 +1,9 @@
|
|||
Sesongevaluering 2025 Sesongevaluering 2025 Kvernlia Andelslandbruk SA. Vedtatt av styret 20. januar 2026. Utarbeidet av gartner Ingrid Solhaug-Berntsen og styreleder Torbjørn Aasvang. 1. Sammendrag Sesongen 2025 var den sjuende siden Kvernlia Andelslandbruk ble startet våren 2019, og den største så langt. 131 andeler, fordelt på 98 helandeler og 33 halvandeler, hentet grønnsaker i 19 uker, fra tirsdag 10. juni til fredag 17. oktober. Til sammen ble det høstet 14 200 kg grønnsaker, bær og urter, mot en plan på 14 800 kg. Det tilsvarer 96 prosent av planen og i snitt 124 kg per helandel. Sesongen ble preget av en våt august. I ukene 32 til 35 kom det til sammen 142 mm nedbør, mot normalt rundt 90 mm, og som følge av fuktigheten fikk potetene i skifte D tørråte. Potetavlingen endte derfor på 1 900 kg, 700 kg under planen. Drivhuset leverte derimot mer enn ventet, og både squash og gulrøtter ga de største avlingene vi har hatt. Økonomisk gikk andelslaget med et overskudd på 16 320 kr, som i sin helhet er satt av til reservefondet. Fondet var ved årsskiftet på 64 300 kr. På bakgrunn av erfaringene og medlemsundersøkelsen har styret vedtatt fem hovedendringer for 2026: ny hentedag lørdag i stedet for fredag, økt arbeidsplikt, et tak på frikjøp, et nytt tunnelhus og et nytt skifte H for potetene. Endringene er beskrevet i kapittel 7 og legges fram på medlemsmøtet torsdag 12. mars 2026. 2. Medlemmer, andeler og henting Ved sesongstart hadde andelslaget 131 betalende andeler, 12 flere enn i 2024. En helandel kostet 4 100 kr og en halvandel 2 400 kr. En helandel er beregnet for en husstand på tre til fire personer, mens en halvandel gir halvparten av mengden på høstetavla og passer for én til to personer. I tillegg kunne medlemmer med hel- eller halvandel tegne en eggandel. Andelstype Antall 2025 Pris 2025 (kr) Arbeidsplikt 2025 (timer) Mengde Helandel 98 4 100 10 Full mengde på høstetavla Halvandel 33 2 400 6 Halv mengde på høstetavla Eggandel (tillegg) 30 600 0 10 egg per uke Hentingen foregikk to dager i uka: tirsdag kl. 16.00–20.00 og fredag kl. 15.00–19.00. Hver andel kunne hente én gang per uke, enten tirsdag eller fredag. I snitt hentet 112 av 131
|
||||
|
||||
andeler hver uke, altså 85 prosent. Det laveste oppmøtet var i uke 30, da bare 96 andeler hentet fordi mange var på ferie. Det høyeste var i uke 36 med 121 andeler. Grønnsaker som ikke ble hentet, ble høstet av gartnerne på lørdagen og levert til Nabolagskjøkkenet ved Dorvikstjern, som drives av frivillige. I løpet av sesongen gikk 610 kg til nabolagskjøkkenet. Ventelista for nye andeler hadde 41 navn per 1. desember 2025, som var fristen for å fornye. Av de 131 andelene ble 119 fornyet for 2026, som gir en fornyelsesgrad på 91 prosent. Dermed får 29 personer fra ventelista andel i 2026, og antallet andeler øker til 148. De 12 som ikke fornyet, oppga oftest flytting eller for lite tid som grunn. 3. Dyrking, avling og drivhus Andelslaget dyrket 1,5 hektar friland fordelt på sju skifter, A til G, og hadde i tillegg et drivhus på 180 m². Vekstskiftet går over fire år, slik at ingen kulturfamilie står på samme skifte oftere enn hvert fjerde år. Skifte G lå i grønngjødsling med kløver hele sesongen og ga derfor ingen avling. Tabellen viser planlagt og faktisk avling per kultur. Kultur Plan (kg) Høstet (kg) Avvik (kg) Kommentar Poteter 2 600 1 900 −700 Tørråte i skifte D Gulrøtter 1 800 1 950 +150 Største avling hittil Løk 900 840 −60 Kål (hode-, rød- og spisskål) 1 400 1 310 −90 Kålflue i juni Salat 1 100 860 −240 Brunsnegl i skifte B Squash 900 1 180 +280 Største avling hittil Bønner 500 460 −40 Rødbeter 700 760 +60 Grønnkål 450 520 +70 Purre 600 570 −30 Jordbær 400 310 −90 Nattefrost 14. mai Urter 200 240 +40 Annet friland (spinat, reddik, erter m.m.) 1 050 950 −100 Drivhus (tomat, agurk, paprika) 2 200 2 350 +150 Første tomater i uke 26 Sum 14 800 14 200 −600 96 prosent av planen Tørråten ble oppdaget i skifte D onsdag 6. august. Fordi sykdommen sprer seg raskt i fuktig vær, ble potetriset slått ned 12. august. Det stoppet smitten, men det stoppet også veksten,
|
||||
|
||||
og knollene ble mindre enn normalt. Lagringspotetene som skulle deles ut i uke 41 og 42, ble derfor erstattet av ekstra gulrøtter og kål. Salaten i skifte B ble angrepet av brunsnegl fra midten av august. Gartnerne plukket snegler for hånd hver morgen i tre uker, men tapet ble likevel 240 kg. For 2026 legges det inn en stripe på én meter uten vegetasjon rundt salatbedene, fordi sneglene ikke liker å krysse tørr, bar jord. Nattefrosten 14. mai skadet blomstene på omtrent en tredel av jordbærplantene. Jordbærene ble derfor begrenset til én liter per helandel per henting, og bare i fire av ukene. Drivhuset på 180 m² ga 2 350 kg, fordelt på 1 420 kg tomat, 690 kg agurk og 240 kg paprika. Drivhuset høstes bare av gartnerne og drivhusvaktene, og grønnsakene ble satt fram ferdig plukket i kasser på hentedagene. Sommeren viste at drivhuset er fullt utnyttet, og at medlemmene ønsker flere tomater enn vi kan dyrke der. Det er bakgrunnen for tunnelhuset på 120 m² som bygges i 2026. 4. Høner og bier Hønseflokken talte 28 høner ved sesongstart. Hønene ga til sammen 6 380 egg i 2025. Av disse gikk 5 700 egg til de 30 eggandelene, som fikk 10 egg per uke i 19 uker. Resten ble brukt på dugnadene og høstfesten 27. september. Medlemmene hadde kveldsvakta i hønsegården kl. 19.30–21.00, mens gartnerne tok morgenstellet. Av 133 kveldsvakter i sesongen ble 114 tatt av medlemmer, og gartnerne dekket de 19 andre. Hønsevakta ble gjort av 22 medlemmer som hadde tatt hønsekurset. Ingen vakter falt ut, men 14 vakter ble byttet med under ett døgns varsel. Natt til 9. oktober tok en rev tre høner fordi en luke i hønsegården ikke var lukket etter kveldsvakta. Etter hendelsen er det satt opp et strømgjerde på 1,2 meter rundt hønsegården, og sjekklista for kveldsvakta har fått et eget punkt om lukene. Flokken telte 25 høner ved årsskiftet. Styret har vedtatt å kjøpe inn 9 unghøner i april 2026, slik at flokken blir på 34 høner. Birøkter Eivind Lauvstad hadde fire bikuber i bigården bak skifte F. Det ble høstet 85 kg honning, i snitt drøyt 21 kg per kube. Hver andel, både hel- og halvandeler, fikk ett glass på 250 gram i uke 37, til sammen 131 glass eller omtrent 33 kg. De resterende 52 kg ble solgt til medlemmene på høstfesten for 160 kr per kilo, som ga 8 320 kr. To medlemmer ble stukket av bier i 2025. Begge var innenfor ti meter fra kubene uten følge av birøkter, og regelen om avstand til bigården blir derfor tydeligere skiltet i 2026. Alle fire kubene var sterke ved siste kontroll 3. desember. Til våren deles to av kubene, slik at andelslaget får seks bikuber i 2026.
|
||||
|
||||
\5. Arbeidsplikt og vakter Arbeidsplikten i 2025 var 10 timer for en helandel og 6 timer for en halvandel. Til sammen skulle medlemmene dermed gjøre 1 178 timer. Det ble utført 1 046 timer, eller 89 prosent. De resterende 132 timene ble kjøpt fri for 250 kr per time, til sammen 33 000 kr. Frikjøpet var ujevnt fordelt. Ni helandeler kjøpte seg fri fra alle sine timer, og 14 andeler kjøpte seg fri fra en del av timene. Som følge av dette var det for få folk på åkervaktene i juli, da ugraset vokser raskest. I ukene 28 til 31 sto 11 av 32 åkervakter ubesatt, og gartnerne måtte bruke 60 timer overtid på luking. Tabellen viser hvordan de utførte timene fordelte seg på vakttypene. Vakttype Lengde per vakt (timer) Antall vakter 2025 Utførte timer 2025 Åkervakt 4 137 548 Hentevakt 4,5 44 198 Hønsevakt (kveld) 1,5 114 171 Drivhusvakt 2 33 66 Dugnad (10. mai og 25. oktober) 3 21 63 Sum 349 1 046 Åkervaktene sto for over halvparten av de utførte timene. Hentevaktene var de mest populære og var fullbooket allerede i april, mens drivhusvaktene på onsdager var vanskeligst å fylle. 6. Økonomi Regnskapet for 2025 viser inntekter på 546 720 kr og kostnader på 530 400 kr. Overskuddet på 16 320 kr er overført til reservefondet, som skal brukes til større investeringer i drivhus og utstyr. Post Regnskap 2025 (kr) Helandeler (98 × 4 100 kr) 401 800 Halvandeler (33 × 2 400 kr) 79 200 Eggandeler (30 × 600 kr) 18 000 Honning solgt på høstfesten 8 320 Frikjøp av arbeidstimer 33 000 Høstfest 6 400 Sum inntekter 546 720 Lønn gartner og gartnerassistent 362 000 Frø og planter 41 500 Gjødsel og kompost 18 200 Fôr til høner 14 600 Birøkt, utstyr 9 800 Vanning og strøm 22 300 Vedlikehold drivhus 16 900 Forsikring 12 400 Leie av jord 24 000
|
||||
|
||||
Diverse 8 700 Sum kostnader 530 400 Resultat 16 320 Lønn er den klart største posten og utgjorde 68 prosent av kostnadene. Gartneren er ansatt i 100 prosent stilling og gartnerassistenten i 40 prosent, til sammen 1,4 årsverk. Vedlikehold av drivhuset ble høyere enn budsjettert fordi plastduken på taket måtte skiftes etter stormen i februar. Jordleien på 24 000 kr tilsvarer 16 000 kr per hektar for de 1,5 hektarene vi leier av Kvernlia gård. Frikjøpet ga 33 000 kr, men styret mener at frikjøp ikke skal være en fast inntektskilde. Pengene dekker ikke de 60 timene overtid gartnerne måtte ta, og gjør det vanskeligere å planlegge arbeidet. 7. Medlemsundersøkelsen og endringer for 2026 Medlemsundersøkelsen ble sendt ut 10. november og var åpen til 24. november 2025. 104 av 131 andeler svarte, en svarprosent på 79. Av dem var 81 prosent fornøyd eller svært fornøyd med sesongen, mot 76 prosent i 2024. 64 prosent ønsket å kunne hente på lørdag i stedet for fredag, og 38 prosent ville ha eggandel hvis det var plass. Det vanligste ønsket i fritekstsvarene var jevnere tilgang på salat. Styret har vedtatt følgende endringer, som gjelder fra sesongen 2026: Hentetiden på fredag utgår og erstattes av lørdag kl. 09.00–13.00, fordi flertallet i undersøkelsen ønsket dette. Tirsdag kl. 16.00–20.00 beholdes. Arbeidsplikten økes til 12 timer for helandel (fra 10) og 8 timer for halvandel (fra 6), fordi arealet øker og åkervaktene i juli må fylles. Frikjøp koster 300 kr per time (fra 250 kr), og hver andel kan kjøpe seg fri fra høyst 4 timer. Prisene blir 4 400 kr for helandel, 2 600 kr for halvandel og 650 kr for eggandel. Økningen skyldes hovedsakelig at gartnerassistenten går opp fra 40 til 60 prosent stilling, og høyere jordleie. Skifte H på 0,3 hektar leies av nabogården til Arne Dyrstad for 6 000 kr per år. Potetene flyttes dit, fordi jorda ikke har hatt poteter på minst ti år og risikoen for tørråte da er lavere. Et tunnelhus på 120 m² bygges ved siden av drivhuset. Det koster 64 000 kr, hvorav 26 000 kr tas fra reservefondet. Hønseflokken økes til 34 høner, og antallet eggandeler økes fra 30 til 40. Sesongen 2026 får 21 hentuker, fra uke 23 til uke 43. Første henting blir tirsdag 2. juni og siste henting lørdag 24. oktober. Med 148 andeler og 1,8 hektar friland har gartnerne satt et mål om en samlet avling på 18 600 kg, inkludert drivhus og tunnelhus. Styret takker alle medlemmer for innsatsen i 2025, og særlig de 22 som tok hønsevakter gjennom hele sesongen.
|
||||
|
|
@ -0,0 +1,11 @@
|
|||
Tindvika skolekorps – Sesongrapport 2025/2026 og reiseplan for København 2027 Tindvika skolekorps – Sesongrapport 2025/2026 og reiseplan for København 2027 Rapporten er skrevet av foreldreutvalget og ble lagt fram for årsmøtet torsdag 4. juni 2026 i storsalen på Tindvika skole. Den dekker perioden 1. august 2025 til 31. mai 2026. Tallene for medlemmer gjelder per 1. september hvert år, mens regnskapstallene gjelder hele perioden. Siste del av rapporten er den foreløpige reiseplanen for turen til Nordisk ungdomskorpstreff i København i juni 2027, slik den ble vedtatt på årsmøtet. 1. Sesongen i korte trekk Tindvika skolekorps ble stiftet i 1971 og har tre avdelinger: Aspirantkorpset, Hovedkorpset og Drill- og slaggruppa. Sesongen 2025/2026 ble den største på over ti år. Korpset hadde 69 medlemmer per 1. september 2025, mot 58 medlemmer ett år tidligere. Veksten kom først og fremst i Aspirantkorpset, som gikk fra 15 til 21 aspiranter fordi vi for første gang holdt en åpen instrumentdag på skolen i juni 2025. Avdeling Medlemmer 2024/2025 Medlemmer 2025/2026 Endring Aspirantkorpset 15 21 +6 Hovedkorpset 31 34 +3 Drill- og slaggruppa 12 14 +2 Totalt 58 69 +11 I løpet av sesongen meldte tre musikanter i Hovedkorpset seg ut, alle fordi familiene flyttet fra Tindvika. Ingen aspiranter sluttet. Fire aspiranter avla overgangsprøven i mai 2026, og alle fire består og går over i Hovedkorpset fra 1. august 2026. To av dem har i tillegg meldt seg på Drill- og slaggruppa. Ingeborg Haugstøl var dirigent for Hovedkorpset for fjerde sesong på rad. Tobias Rønneberg ledet Aspirantkorpset, og Live Aasmyr var instruktør for Drill- og slaggruppa. Foreldreutvalget hadde seks medlemmer og holdt ti møter i perioden. 2. Konserter og opptredener Korpset holdt tre konserter med billettsalg. Voksne betalte 100 kroner i inngang, mens barn under 16 år hadde gratis adgang. Publikumstallet nedenfor omfatter derfor også barn som ikke betalte. Billettinntektene gikk til korpsets drift og ikke til reisefondet. Konsert Dato Sted Publikum Billettinntekt
|
||||
|
||||
(kr) Høstkonsert 16.11.2025 Tindvika kirke 312 18 400 Julekonsert 14.12.2025 Storsalen, Tindvika skole 245 12 900 Vårkonsert 26.04.2026 Tindvika kirke 288 16 350 Sum 845 47 650 Høstkonserten hadde flest tilhørere, men julekonserten fikk mest ros i evalueringen fordi alle tre avdelingene spilte sammen i det siste nummeret. Julekonserten ble holdt i storsalen og ikke i kirken, fordi kirken var opptatt den helgen. Som følge av den mindre salen ble det solgt færre billetter enn året før. I tillegg til konsertene spilte Hovedkorpset og Drill- og slaggruppa i barnetoget 17. mai 2026. Drill- og slaggruppa deltok også på en regional drillmønstring i Bergen 21. mars 2026, der gruppa fikk 81 av 100 poeng og endte på tredjeplass av ni grupper i juniorklassen. 3. Seminarer Korpset holdt to helgeseminarer på Granlia leirskole. Deltakeravgiften var 650 kroner per musikant for hvert seminar. Avgiften dekket overnatting, alle måltider og instruktører. Aspirantene deltok bare på lørdagen og betalte derfor ikke for seminarene denne sesongen; korpset dekket deres lunsj. Høstseminaret 10.–12. oktober 2025 hadde 41 deltakere fra Hovedkorpset og Drill- og slaggruppa og fire innleide instruktører. Vinterseminaret 6.–8. februar 2026 hadde 38 deltakere og tre innleide instruktører, fordi slagverksinstruktøren måtte melde avbud. Begge seminarene endte med en kort konsert for foresatte søndag kl. 13:00. Seminarene er den viktigste arbeidsformen for å lære nye stykker. På høstseminaret øvde Hovedkorpset inn hele programmet til høstkonserten, mens Drill- og slaggruppa brukte tolv timer på et nytt drillprogram. Foreldreutvalget hadde to foresatte på vakt hele natten på begge seminarene, og ingen hendelser ble meldt. Musikantene hadde ikke lov til å ha mobiltelefon etter kl. 22:00, og telefonene ble samlet inn av vaktene. Evalueringen etter vinterseminaret viste at 34 av 38 deltakere ønsket å beholde ordningen med to seminarer per sesong. Foreldreutvalget har derfor satt opp to seminarer også i 2026/2027. Deltakeravgiften øker til 700 kroner per seminar fra høsten 2026 fordi Granlia leirskole har varslet høyere pris for overnatting. 4. Instrumenter og utlån Per 31. mai 2026 eide korpset 39 instrumenter, medregnet slagverk. Av disse var 32 utlånt til musikanter gjennom hele sesongen. Leien var 350 kroner per semester for hvert instrument, og aspiranter i sitt første år betalte ikke leie. Leieinntektene ble 22 400 kroner. Den største enkeltutgiften var en ny tuba til 27 900 kroner, kjøpt i mars 2026 fordi den eldste tubaen fra 1984 hadde en sprekk i klokka og ikke lenger kunne repareres. To klarinetter var til service i til sammen elleve uker. Ingen instrumenter ble meldt tapt, men
|
||||
|
||||
en kornett fikk en bulk i munnstykkerøret, og familien betalte egenandelen på 1 000 kroner etter reglene for instrumentlån. Instrumentsjekken i uke 34 i august 2025 viste at fem instrumenter manglet kasse eller hadde en ødelagt kasse. Korpset kjøpte fem nye kasser for til sammen 4 750 kroner. Fordi tre av kassene var ødelagt ved uaktsomhet, har foreldreutvalget bestemt at familien fra 2026/2027 skal betale for ny kasse i slike tilfeller, innenfor egenandelen på inntil 1 000 kroner. Kasser som er slitt ved vanlig bruk, erstattes fortsatt av korpset. Instrumentansvarlig Siri Nyhagen har gått gjennom alle instrumentene. Hun anbefaler at korpset kjøper ett nytt althorn før høsten 2026, fordi to nye musikanter fra Aspirantkorpset skal spille althorn i Hovedkorpset. Årsmøtet sluttet seg til anbefalingen. 5. Økonomi Regnskapet for 2025/2026 viser et underskudd på 3 200 kroner. Underskuddet skyldes kjøpet av den nye tubaen, som ikke var budsjettert. Uten tubaen ville korpset ha hatt et overskudd på 24 700 kroner. Korpset satte av 15 000 kroner til reisefondet, som dermed hadde en saldo på 41 000 kroner per 31. mai 2026. Post Beløp (kr) Medlemskontingent 196 350 Instrumentleie 22 400 Billettinntekter konserter 47 650 Dugnader 61 200 Seminaravgift fra deltakere 51 350 Tilskudd fra Tindvika Musikklegat 40 000 Sum inntekter 418 950 Honorar dirigent Hovedkorpset 172 000 Honorar dirigent Aspirantkorpset 58 000 Honorar instruktør Drill- og slaggruppa 46 000 Seminarer (opphold, mat, instruktører) 63 800 Instrumentservice og innkjøp 41 500 Noter og arrangementer 12 300 Konsertutgifter 9 400 Avsetning til reisefondet 15 000 Diverse 4 150 Sum utgifter 422 150 Resultat -3 200 Kontingenten i 2025/2026 var 1 650 kroner per semester for Hovedkorpset og Drill- og slaggruppa og 1 100 kroner per semester for Aspirantkorpset. Årsmøtet vedtok å øke kontingenten fra 1. august 2026 til 1 800 kroner for Hovedkorpset og Drill- og slaggruppa og 1 200 kroner for Aspirantkorpset. Økningen skyldes at den nye avtalen med dirigentene gir 8 prosent høyere honorar fra samme dato. Søskenmoderasjonen på 25 prosent for barn nummer to videreføres, og barn nummer tre og flere betaler fortsatt ikke kontingent. Foreldreutvalget har budsjettert sesongen 2026/2027 med 446 000 kroner i inntekter og 442 000 kroner i utgifter. Det settes ikke av penger til reisefondet i 2026/2027, fordi 25
|
||||
|
||||
000 kroner av fondet skal brukes på turen til København. De resterende 16 000 kronene i fondet skal stå urørt som reserve for neste tur. Årsmøtet ga foreldreutvalget fullmakt til å kjøpe instrumenter for inntil 20 000 kroner uten nytt årsmøtevedtak; større kjøp må godkjennes av et ekstraordinært årsmøte. 6. Dugnad og foreldrevakt Dugnadene ga 61 200 kroner i 2025/2026. Alle familier skulle ta tre dugnadsvakter i løpet av sesongen. Sju familier gjennomførte ikke alle vaktene sine, til sammen elleve vakter, og ble fakturert 300 kroner per vakt. Disse 3 300 kronene er tatt med i dugnadsinntekten nedenfor. Dugnad Dato Inntekt (kr) Loppemarked i gymsalen 27.09.2025 29 800 Juletresalg på skoleplassen 06.12.2025 18 600 Kafé ved tre konserter november 2025 – april 2026 9 500 Fakturerte vakter som ikke ble gjennomført 3 300 Sum 61 200 Loppemarkedet ga mest, men krevde også flest vakter: 26 vakter fordelt på fredag kveld og lørdag. Juletresalget ga nest mest, selv om salget ble stengt to timer tidligere enn planlagt fordi alle trærne var solgt. Foreldreutvalget foreslår at juletresalget i 2026 bestiller flere trær, og at hele inntekten går til turen til København. Ordningen med foreldrevakt på øvelsene ble innført i januar 2026, etter at storsalen to ganger sto ulåst etter øvelse i høstsemesteret. Foreldrevakten låser opp og igjen, passer på romreglene og er til stede i pausen. Fra høsten 2026 har familier i Hovedkorpset og Drill- og slaggruppa foreldrevakt to ganger per semester, mens familier i Aspirantkorpset har foreldrevakt én gang per semester. Foreldrevakt regnes ikke som dugnadsvakt. 7. Reiseplan: Nordisk ungdomskorpstreff i København 2027 Årsmøtet vedtok at Hovedkorpset og Drill- og slaggruppa skal reise til Nordisk ungdomskorpstreff i København fra torsdag 17. juni til mandag 21. juni 2027. Turen varer i fem dager med fire overnattinger. Turkomiteen ledes av Kari Enerhaug, og planen nedenfor er foreløpig. Endelig program blir sendt ut senest 1. april 2027. Hvem kan reise Turen gjelder bare musikanter i Hovedkorpset og Drill- og slaggruppa. Aspirantkorpset deltar ikke. Musikanten må ha fylt 11 år innen 1. juni 2027. Musikanten må ha vært medlem av Hovedkorpset eller Drill- og slaggruppa senest fra 1. januar 2027. Musikanten må ha møtt på minst 80 prosent av øvelsene i vårsemesteret 2027, med unntak av dokumentert sykdom.
|
||||
|
||||
Foreldreutvalget regner med 38 musikanter. Korpset skal ha minst én voksen per seks musikanter, og derfor reiser sju voksne: dirigent Ingeborg Haugstøl, instruktør Live Aasmyr, turleder Kari Enerhaug og fire foresatte som trekkes blant dem som melder seg. Voksne ledere betaler ikke egenandel. Søsken som ikke er medlemmer, kan ikke bli med på turen. Program Dag Dato Innhold Torsdag 17.06.2027 Oppmøte ved Tindvika skole kl. 06:45, ankomst og innkvartering på Herberget Kastanjegården om kvelden Fredag 18.06.2027 Festivalparade kl. 11:00 og marsjkonkurranse for Drill- og slaggruppa kl. 15:30 Lørdag 19.06.2027 Konsertkonkurranse for Hovedkorpset, spilletid maks 20 minutter Søndag 20.06.2027 Fellesaktivitet i badeland før lunsj, avslutningskonsert kl. 19:00 Mandag 21.06.2027 Hjemreise, forventet ankomst Tindvika skole ca. kl. 21:00 Kostnader og betaling Turen er budsjettert med 252 950 kroner i utgifter og 257 200 kroner i inntekter, altså et overskudd på 4 250 kroner som buffer. Hver musikant betaler en egenandel på 3 900 kroner. Resten finansieres med 70 000 kroner fra dugnader, 25 000 kroner fra reisefondet og 14 000 kroner fra en støttekonsert i mars 2027. Festivalavgiften er budsjettert med 550 kroner per musikant, slik den var oppgitt av arrangøren i mai 2026. Påmeldingen er bindende og må være levert innen 1. desember 2026. Egenandelen betales i tre like rater på 1 300 kroner: innen 15. januar, 15. mars og 15. mai 2027. Ved avbestilling før 1. mars 2027 betales alt innbetalt beløp tilbake, med fradrag av et gebyr på 300 kroner. Ved avbestilling fra og med 1. mars 2027 betales egenandelen bare tilbake mot legeattest. Søskenmoderasjonen på kontingenten gjelder ikke egenandelen for turen. Familier med barn på turen skal ta tre ekstra dugnadsvakter i sesongen 2026/2027, i tillegg til de tre ordinære vaktene alle familier har. For hver ekstra turvakt som ikke blir gjennomført, faktureres familien 400 kroner, fordi dugnadsinntekten er en forutsetning for at egenandelen kan holdes på 3 900 kroner.
|
||||
|
||||
Regler under turen Musikanter under 13 år skal være på rommet kl. 22:30, og musikanter som er 13 år eller eldre kl. 23:00. Mobiltelefoner leveres til voksen leder ved leggetid og hentes ved frokost. Musikantene skal alltid gå i grupper på minst tre, og de har ikke lov til å forlate festivalområdet uten en voksen. Alkohol, energidrikk og snus er forbudt under hele turen. Brudd kan føre til at musikanten sendes hjem for familiens regning. Anbefalte lommepenger er inntil 800 kroner. Forberedelser Turkomiteen består av Kari Enerhaug og to andre foresatte og rapporterer til foreldreutvalget hver måned. Komiteen holder et eget foreldremøte om turen i januar 2027, etter at påmeldingsfristen har gått ut. Hovedkorpset skal øve inn konkurranseprogrammet fra vinterseminaret i februar 2027, og Drill- og slaggruppa skal ha fire ekstra øvelser på lørdager i mai 2027. Alle musikanter som reiser, må ha gyldig pass eller nasjonalt ID-kort. Foresatte skal fylle ut et helseskjema med opplysninger om allergier og medisiner innen 1. mai 2027. Instrumentene som korpset eier, er forsikret gjennom korpsets ordinære forsikring også på turen. Private instrumenter er ikke dekket, med mindre familien selv har forsikret dem. Store slagverksinstrumenter sendes i kasser fra skolen to uker før avreise og skal pakkes av instrumentansvarlig, ikke av musikantene.
|
||||
|
|
@ -0,0 +1,9 @@
|
|||
Tallow Hill Astronomical Society: Annual Report 2025 Tallow Hill Astronomical Society: Annual Report 2025 This report covers the work of the Tallow Hill Astronomical Society and its observatory from 1 January to 31 December 2025. It was approved by the committee on 11 February 2026 and will be presented to the Annual General Meeting on Saturday 14 March 2026 in Wrexley Green village hall. All money figures are in pounds sterling and all visitor figures are head counts taken at the gate. 1. Chair's foreword 2025 was the busiest year the society has had since it was founded in 1987. Membership rose from 171 to 186, members borrowed a society telescope 97 times, and 894 people came up the hill to a public evening, the solar open day or a school group visit. Our 34 active volunteers gave an estimated 2,640 hours of their time, compared with 2,210 hours in 2024. The year also brought the most serious equipment failure since the observatory opened in 1994. On 19 February the dome rotation motor burnt out while the dome was being closed in a sustained wind of 42 km/h. The dome stayed out of use until 26 March, and the repair cost £2,350. As a result, the committee lowered the wind limit for operating the dome from 40 km/h to 35 km/h in edition 4 of the operating handbook, which took effect on 1 February 2026. Two decisions taken in 2025 affect members in 2026. First, the refundable deposit for a loan telescope rose from £30 to £40 on 1 January 2026. Second, the committee is asking the Annual General Meeting to approve higher membership fees from 1 April 2026; the reasons are set out in section 7. I would like to thank every volunteer, and the Corran family, who own Tallow Hill and have leased the observatory site to us since 1994. Rosalind Fairweather, Chair. 2. Membership and volunteers On 31 December 2025 the society had 186 members, an increase of 15 on the 171 members a year earlier. Of the 171 members at the end of 2024, 154 renewed, a renewal rate of 90%. A further 32 people joined during the year, and 14 of those 32 new members came to us through the First Light beginners' course. Family memberships cover up to two adults and their children under 16 living at the same address, and each person in a family membership is counted as a member. Category Fee in 2025 (£ per year) Members on 31 Dec 2024 Members on 31 Dec 2025 Adult 42 95 102
|
||||
|
||||
Concession (students and people aged 65 or over) 24 38 41 Junior (under 16) 10 18 19 Family (up to 2 adults and their children under 16) 65 20 people in 10 memberships 24 people in 12 memberships Total 171 186 The society depends on volunteers for everything it does. At the end of 2025 there were 19 keyholders, two more than a year earlier, who are trained to open and close the observatory on their own responsibility. Eight of the keyholders are qualified to operate the Pennick Telescope in the main dome. Fourteen volunteers hold the society's safeguarding certificate, which is needed for any work with school groups and must be renewed every three years. Active volunteers: 34 (2024: 31). Volunteer hours: about 2,640 (2024: about 2,210). Keyholders: 19, of whom 2 qualified in 2025. Pennick Telescope operators: 8, unchanged from 2024. Volunteers with a current safeguarding certificate: 14. 3. The observatory and its telescopes Tallow Hill Observatory stands 214 m above sea level, 3 km north of Wrexley Green. The main dome is 4.2 m in diameter and houses the 406 mm Pennick Telescope, named after the society's founder Margaret Pennick. The roll-off roof building known as the Annexe has three piers: pier 1 carries the society's 254 mm Schmidt-Cassegrain telescope, and piers 2 and 3 can be booked by members for their own instruments. The warm room seats 20 people and is used for talks, the beginners' course and hot drinks. Because the dome was out of use from 19 February to 26 March, the public evening on 7 March and the school visit on 12 March used only the Annexe telescopes. Both went ahead, and the 7 March evening was fully booked with 40 visitors. Members logged 61 observing nights in 2025 with a total of 214 clear hours, against 49 nights and 176 clear hours in 2024. Members borrowed a society telescope 97 times in 2025, compared with 81 times in 2024, and 58 different members used the scheme. The most borrowed instrument was the 150 mm Dobsonian T06, with 14 loans. Of the 97 loans, 19 were returned late, and late fees brought in £146. Late fees are £2 per day and are capped at £30 for a single loan. A grant of £4,500 from the Elsworth Family Trust paid for two new loan telescopes, which joined the fleet in September 2025: T13, a second 150 mm Dobsonian, and T14, an 80 mm refractor. The 200 mm Dobsonian T07 was withdrawn on 18 November 2025 after a borrower reported a dull primary mirror; it has gone away for recoating and is expected back in April 2026. The society now owns 14 telescopes, of which 11 are part of the loan
|
||||
|
||||
fleet. The Pennick Telescope, the Annexe pier telescope and the solar telescope T12 are never lent out. On 10 September 2025 the committee decided to raise the refundable loan deposit from £30 to £40 for all loans starting on or after 1 January 2026. The reason is that the replacement cost of the eyepiece and finder set lent with each telescope has risen to about £45, so a £30 deposit no longer covered a lost set. The concession deposit is £25. 4. Public stargazing and the solar open day The 2025 programme had 14 public evenings: the regular evening on the first Friday of each month from January to April and from September to December, and six special evenings for events such as the Perseid and Geminid meteor showers. Nine evenings were held and five were cancelled, four because of cloud and one, on 12 April, because of high wind. Public evenings are free, but places must be booked, and bookings are capped at 40 per evening. The nine evenings that went ahead drew 318 visitors, just over 35 per evening, and two of them, on 7 March and 5 September, were fully booked. The summer solar open day was held on Saturday 21 June 2025 from 10:00 to 16:00. Visitors looked at the Sun safely through the hydrogen-alpha solar telescope T12 and through projection screens, and 412 people attended, against 365 in 2024. The table below gives visitor numbers for all events held at the observatory. Pupils reached in their own classrooms are not included here; they are reported in section 5. Activity Events held 2025 Visitors 2025 Events held 2024 Visitors 2024 Public stargazing evenings 9 (14 planned) 318 8 (13 planned) 279 Summer solar open day 1 412 1 365 School group evening visits 6 164 4 109 Total 16 894 13 753 5. First Light course and schools outreach First Light is the society's beginners' course: six sessions on Tuesday evenings from 19:30 to 21:30 in the warm room, with a practical night at the telescopes as the last session. A participant completes the course by attending at least five of the six sessions. The fee in 2025 was £20 for members and £35 for non-members, and the non-member fee includes three months of free membership. Spring run, 4 February to 11 March 2025: 16 enrolled (the maximum), 13 completed. Autumn run, 7 October to 11 November 2025: 13 enrolled, 11 completed. In total 29 people enrolled and 24 completed; 11 were already members and 18 were not. 14 of the 18 non-member participants went on to take out full membership.
|
||||
|
||||
Our schools programme, Sky in the Classroom, grew strongly. Volunteers made 17 classroom visits to 9 schools and reached 1,040 pupils, compared with 12 visits and 760 pupils in 2024. The inflatable planetarium, bought in 2023, was used at 11 of those visits for a total of 31 shows; it holds at most 25 children per show. Schools contribute £40 towards materials for each classroom visit, which brought in £680. Six school groups, with 164 pupils in all, also came to the observatory for an evening visit. Evening visits are free for schools, but they are open only to schools within 25 km of the observatory and only to pupils aged 9 or over. A group may bring at most 30 pupils, and there must be at least one adult for every 8 pupils. Every volunteer who helps with a school group must hold a current safeguarding certificate. 6. Keep Tallow Dark Keep Tallow Dark is the society's dark-sky campaign, started in 2022. Households are asked to sign a pledge to shield their outdoor lights so that no light shines above the horizontal, to use warm-white lamps of 3,000 K or lower, and to fit motion sensors or switch outdoor lights off by 23:00. By 31 December 2025 the campaign had 312 pledges, up from 205 a year earlier. The target is 400 pledges by 31 December 2026. Eleven volunteers took 102 sky brightness readings at six monitoring sites during 2025. A reading is counted only if the Moon is below the horizon and the reading is taken at least 90 minutes after sunset, because moonlight and twilight would otherwise swamp the artificial light we are trying to measure. Readings are given in magnitudes per square arcsecond, and a higher number means a darker sky. The mean of the six site averages rose from 20.6 in 2024 to 20.8 in 2025. Every site recorded a darker average sky than in 2024, by between 0.1 and 0.3. The darkest site was the observatory itself, at 21.3, and the brightest was Wrexley Green village hall, at 20.4. The committee is cautious about these results: a single year's change can partly reflect weather and the timing of readings, and three more years of data are needed before the campaign can claim a lasting improvement. 7. Finances and plans for 2026 The society ended 2025 with a surplus of £1,420 on income of £13,110 and expenditure of £11,690. In 2024 income was £11,920 and expenditure £11,310, a surplus of £610. Reserves stood at £9,830 on 31 December 2025, against £8,410 a year earlier. Of the reserves, £1,370 is a restricted fund: it is the unspent part of the Elsworth Family Trust grant and may be used only for schools outreach in 2026. Item Income 2025 (£) Expenditure 2025 (£) Membership fees 6,238 First Light course fees 850 Grant from the Elsworth Family Trust 4,500 School contributions to 680
|
||||
|
||||
classroom visits Tea stall at the solar open day 612 Loan late fees 146 Bank interest 84 Insurance 1,840 Site lease from the Corran family 1,200 Electricity 960 Dome motor repair 2,350 Two new loan telescopes (T13 and T14) 1,980 Telescope maintenance and eyepieces 640 Schools outreach kits and printing 1,150 First Light course materials 310 Keep Tallow Dark leaflets and sky meter 540 Hall hire for winter talks 720 Total 13,110 11,690 Our insurer has quoted £2,171 for 2026, an increase of 18% on the £1,840 paid in 2025. The committee also wants to build a dome reserve of £3,000 by the end of 2027, so that a future motor failure can be repaired without delay. For these two reasons the committee proposes the following membership fees from 1 April 2026, subject to approval at the Annual General Meeting on 14 March 2026: Adult: from £42 to £45 per year. Concession: from £24 to £26 per year. Junior: unchanged at £10 per year. Family: from £65 to £70 per year. The First Light course fees and the late fee of £2 per day are not affected by the proposal. Our main plans for 2026 are: Return T07 to the loan fleet after mirror recoating, expected in April 2026. Reach 400 Keep Tallow Dark pledges by 31 December 2026. Make at least 20 classroom visits, funded partly from the restricted fund of £1,370. Train at least three new keyholders in a course starting in May 2026. Replace the warm room heater, which failed twice during the winter of 2025.
|
||||
|
|
@ -0,0 +1,251 @@
|
|||
# Tallow Hill Observatory Operating Handbook and Telescope Loan Rules, Edition 4
|
||||
|
||||
# Tallow Hill Observatory Operating Handbook and Telescope Loan Rules
|
||||
|
||||
Edition 4. Approved by the committee of the Tallow Hill Astronomical Society on 21 January 2026 and in force from 1 February 2026. This edition replaces edition 3 of March 2023. Changes from edition 3 are marked 'Changed in edition 4' with the reason.
|
||||
|
||||
## 1. Scope of this handbook
|
||||
|
||||
This handbook applies to every member, volunteer and guest at Tallow Hill Observatory, and to every member who borrows a society telescope. Part A (sections 2 to 6) covers the observatory site. Part B (sections 7 to 10) covers the telescope loan scheme. Where this handbook and a verbal instruction from the duty director disagree on the night, the duty director's instruction applies until the end of that night, and the duty director must report the disagreement to the Observatory Director within 7 days.
|
||||
|
||||
The duty director is the keyholder named on the rota as responsible for a given night. On public evenings the duty director must be one of the eight qualified Pennick Telescope operators.
|
||||
|
||||
## 2. Access and keyholders
|
||||
|
||||
Only keyholders may unlock the lower gate, the dome, the Annexe and the warm room. To become a keyholder a member must have been a member for at least 6 months, complete the three-session keyholder training and be approved by the committee. There are currently 19 keyholders.
|
||||
|
||||
Two-person rule: after sunset, nobody may be on the site alone. At least two adults must be present, and at least one of them must be a keyholder.
|
||||
|
||||
The first keyholder to arrive must sign in the observatory log in the warm room, and the last to leave must record the closing time and the state of the dome and the Annexe roof.
|
||||
|
||||
The site must be vacated by 03:00, unless the duty director has approved an all-night session at least 24 hours in advance.
|
||||
|
||||
The code on the lower gate is changed every 6 months, on 1 January and 1 July, and is given only to keyholders.
|
||||
|
||||
A keyholder who has not opened the observatory for 12 months must repeat the final training session before opening it again.
|
||||
|
||||
## 3. Buildings, capacity and age limits
|
||||
|
||||
The observatory has three buildings, each with its own limits. The limits apply on every night, including public evenings, member nights and school visits.
|
||||
|
||||
Building
|
||||
|
||||
Maximum people inside
|
||||
|
||||
Lighting allowed
|
||||
|
||||
Age rule
|
||||
|
||||
Food and drink
|
||||
|
||||
Main dome (Pennick Telescope)
|
||||
|
||||
8, including the operator
|
||||
|
||||
Red light only
|
||||
|
||||
Children under 10 not admitted; under 16 only with an adult
|
||||
|
||||
Not allowed
|
||||
|
||||
Annexe (roll-off roof, 3 piers)
|
||||
|
||||
15
|
||||
|
||||
Red light only
|
||||
|
||||
Under 16 only with an adult
|
||||
|
||||
Not allowed
|
||||
|
||||
Warm room
|
||||
|
||||
20 seated
|
||||
|
||||
White light allowed
|
||||
|
||||
No limit
|
||||
|
||||
Allowed
|
||||
|
||||
The warm room is the only building where white light, food and hot drinks are allowed. Phone screens used outside the warm room must be set to a red night mode or kept in a pocket, because a single glance at a white screen spoils the dark adaptation of everyone nearby.
|
||||
|
||||
Piers 2 and 3 in the Annexe may be booked by members for their own telescopes. A booking is for one night, a member may hold at most two bookings in any calendar month, and bookings close at 12:00 on the day. Pier 1 carries the society's 254 mm Schmidt-Cassegrain telescope and may not be booked.
|
||||
|
||||
## 4. Weather and closing rules
|
||||
|
||||
The dome and the Annexe roof are closed according to the limits below. Any keyholder present may order the dome or the Annexe closed, and every person on site must follow that order at once.
|
||||
|
||||
Condition
|
||||
|
||||
Main dome
|
||||
|
||||
Annexe roof
|
||||
|
||||
Sustained wind
|
||||
|
||||
Close above 35 km/h
|
||||
|
||||
Close above 45 km/h
|
||||
|
||||
Gusts
|
||||
|
||||
Close above 50 km/h
|
||||
|
||||
Close above 60 km/h
|
||||
|
||||
Relative humidity
|
||||
|
||||
Close above 90%
|
||||
|
||||
Close above 90%
|
||||
|
||||
Rain, sleet or snow
|
||||
|
||||
Close at once
|
||||
|
||||
Close at once
|
||||
|
||||
Changed in edition 4: the sustained wind limit for the dome was 40 km/h in edition 3. It was lowered to 35 km/h because the dome rotation motor failed on 19 February 2025 while the dome was being closed in a sustained wind of 42 km/h. The Annexe limits are unchanged, because the Annexe roof slides on rails and is not driven by a motor.
|
||||
|
||||
The cover must be put on the Pennick Telescope whenever relative humidity rises above 85%, even if the dome itself stays open, so that dew does not form on the mirror.
|
||||
|
||||
The duty director decides at 16:00 on the day whether a public evening goes ahead. The evening is cancelled if the forecast cloud cover for 19:00 to 22:00 is above 70%, or if the forecast sustained wind is above 35 km/h. Booked visitors must be told of a cancellation by 17:00. An evening that goes ahead under cloud is not cancelled later, and the programme then moves to a talk in the warm room.
|
||||
|
||||
## 5. Operating the Pennick Telescope
|
||||
|
||||
Only the eight qualified Pennick Telescope operators may move the telescope, rotate the dome or change the instruments on the telescope. Every session on the Pennick Telescope must be entered in the telescope log with the start time, the end time and the objects observed.
|
||||
|
||||
The telescope must not be pointed below 15 degrees above the horizon, because the tube would then strike the lower wall of the dome.
|
||||
|
||||
The telescope must never be pointed at the Sun or within 20 degrees of it, and it may not be used in daylight.
|
||||
|
||||
Guests may look through the eyepiece but must not touch the hand controller or the mount.
|
||||
|
||||
Members aged under 14 must not use the dome rotation controls, even under supervision.
|
||||
|
||||
Any fault must be written in the fault book before the operator leaves, and a fault that stops the telescope from moving must be reported to the Observatory Director by 12:00 the next day.
|
||||
|
||||
## 6. Safety and first aid
|
||||
|
||||
The first-aid kit, the defibrillator and the fire extinguishers are kept in the warm room, next to the door. Every keyholder is shown where they are during training. The footpath from the lower gate to the buildings is about 400 m long and takes about ten minutes to walk; in frost or snow the duty director may close the path to guests.
|
||||
|
||||
Any accident or near miss must be written in the accident book on the night and reported to the Safety Officer within 24 hours.
|
||||
|
||||
Alcohol is not allowed anywhere on the site, except at the summer solar open day, when it is allowed in the warm room only.
|
||||
|
||||
Children under 16 must be accompanied by an adult at all times outside the warm room.
|
||||
|
||||
School groups must bring at least one adult for every 8 pupils, and no more than 30 pupils may visit at one time.
|
||||
|
||||
Every volunteer working with a school group must hold a current safeguarding certificate, renewed every three years.
|
||||
|
||||
## 7. Telescope loans: who may borrow
|
||||
|
||||
The society keeps a loan fleet of 11 telescopes. The Pennick Telescope, the Annexe pier telescope and the solar telescope T12 are never lent, and solar filters are not lent with any loan telescope. A member may borrow a loan telescope only if all the following apply:
|
||||
|
||||
The member has been a member for at least 3 months. Exception: a member who has completed the First Light course in the last 12 months may borrow after 1 month of membership.
|
||||
|
||||
The member is aged 18 or over. Members aged 14 to 17 may borrow only a telescope weighing less than 10 kg, and only with an adult member co-signing the loan form. Members under 14 may not borrow.
|
||||
|
||||
No one else in the same household has a society telescope on loan. Only one telescope may be on loan to a household at a time.
|
||||
|
||||
The member has no unpaid late fees from an earlier loan.
|
||||
|
||||
Telescopes weighing 15 kg or more, which at present are the 150 mm and 200 mm Dobsonians, may be lent only to a member who has completed the 45-minute loan induction. The induction is run by the Loans Officer on the first Saturday of each month and is valid for as long as the member keeps their membership.
|
||||
|
||||
## 8. Loan periods, deposits and late fees
|
||||
|
||||
Item
|
||||
|
||||
Rule
|
||||
|
||||
Standard loan period
|
||||
|
||||
28 days for loans starting 1 March to 31 October
|
||||
|
||||
Peak season loan period
|
||||
|
||||
14 days for loans starting 1 November to 28 February
|
||||
|
||||
Renewal
|
||||
|
||||
One renewal of 14 days, only if no other member is on the waiting list for that telescope
|
||||
|
||||
Deposit
|
||||
|
||||
£40, or £25 for concession members; refunded when the telescope is returned complete
|
||||
|
||||
Late fee
|
||||
|
||||
£2 per day, capped at £30 for one loan
|
||||
|
||||
Overdue more than 21 days
|
||||
|
||||
Deposit kept and the case referred to the committee
|
||||
|
||||
Changed in edition 4: the deposit was £30 for loans starting before 1 January 2026. It was raised to £40 because the replacement cost of the eyepiece and finder set lent with each telescope has risen to about £45. The concession deposit of £25 is unchanged.
|
||||
|
||||
Renewals must be asked for before the due date, by message to the Loans Officer. A renewal starts on the original due date, so it does not extend a loan that is already overdue. The Loans Officer may waive a late fee only if the observatory was closed on the due date and on the following collection session.
|
||||
|
||||
## 9. Collection, care and return
|
||||
|
||||
Telescopes are collected and returned in the warm room on Saturdays from 10:00 to 12:00 and on Wednesdays from 19:00 to 20:00. No other times are possible unless the Loans Officer agrees in writing beforehand. On collection the Loans Officer checks the telescope against its contents list with the borrower, and both sign the loan form.
|
||||
|
||||
Never point a loan telescope at the Sun, with or without a filter.
|
||||
|
||||
Do not leave a telescope outdoors overnight or store it in an unheated outbuilding, because condensation damages the optics.
|
||||
|
||||
Do not try to clean or adjust a mirror or lens; report any problem instead.
|
||||
|
||||
Report any damage or loss to the Loans Officer within 48 hours.
|
||||
|
||||
The telescope may not be lent on to anyone else, including family members who are not covered by the loan.
|
||||
|
||||
A borrower who reports damage within 48 hours is charged only the actual repair cost, up to the amount of the deposit. Damage reported later, or found only at return, may be charged at the full repair cost.
|
||||
|
||||
## 10. Roles and contacts
|
||||
|
||||
Role
|
||||
|
||||
Name
|
||||
|
||||
Responsible for
|
||||
|
||||
Chair
|
||||
|
||||
Rosalind Fairweather
|
||||
|
||||
Committee meetings and the annual report
|
||||
|
||||
Observatory Director
|
||||
|
||||
Priya Lindgren
|
||||
|
||||
The buildings, the duty rota and the Pennick Telescope
|
||||
|
||||
Loans Officer
|
||||
|
||||
Tomasz Wren
|
||||
|
||||
The loan fleet, inductions, deposits and late fees
|
||||
|
||||
Safety Officer
|
||||
|
||||
Owen Pryce
|
||||
|
||||
The accident book, first aid and safety checks
|
||||
|
||||
Outreach Coordinator
|
||||
|
||||
Gwen Haddow
|
||||
|
||||
School visits and safeguarding certificates
|
||||
|
||||
Course Leader
|
||||
|
||||
Callum Steadman
|
||||
|
||||
The First Light beginners' course
|
||||
|
||||
Questions about this handbook go to the Observatory Director. Proposals for changes may be sent by any member, and the committee reviews the handbook at least once every three years.
|
||||
|
|
@ -0,0 +1,79 @@
|
|||
#### Inventory 31 Jan 2026
|
||||
|
||||
A: Telescope ID | B: Type | C: Aperture (mm) | D: Focal length (mm) | E: Weight (kg) | F: Year acquired | G: Condition | H: Loanable | I: Status on 31 Jan 2026 | J: Notes
|
||||
A: T01 | B: Reflector (Pennick Telescope) | C: 406 | D: 1830 | E: 68 | F: 2006 | G: Good | H: No | I: Observatory use | J: Fixed in the main dome; qualified operators only
|
||||
A: T02 | B: Schmidt-Cassegrain | C: 254 | D: 2540 | E: 27.5 | F: 2011 | G: Good | H: No | I: Observatory use | J: Fixed on Annexe pier 1
|
||||
A: T03 | B: Refractor | C: 70 | D: 700 | E: 4.5 | F: 2009 | G: Fair | H: Yes | I: On loan | J: Beginner kit; Moon and planets
|
||||
A: T04 | B: Refractor | C: 80 | D: 600 | E: 6 | F: 2012 | G: Good | H: Yes | I: Available
|
||||
A: T05 | B: Tabletop Dobsonian | C: 114 | D: 450 | E: 5.8 | F: 2015 | G: Good | H: Yes | I: On loan | J: Needs a sturdy table or stool
|
||||
A: T06 | B: Dobsonian | C: 150 | D: 1200 | E: 17.5 | F: 2016 | G: Good | H: Yes | I: Available | J: Induction required; most borrowed in 2025 (14 loans)
|
||||
A: T07 | B: Dobsonian | C: 200 | D: 1200 | E: 21 | F: 2017 | G: Under repair | H: Yes | I: Withdrawn | J: Withdrawn 18 Nov 2025 for primary mirror recoating; expected back April 2026
|
||||
A: T08 | B: Dobsonian | C: 200 | D: 1200 | E: 21 | F: 2019 | G: Good | H: Yes | I: On loan | J: Induction required
|
||||
A: T09 | B: Maksutov | C: 90 | D: 1250 | E: 4.2 | F: 2018 | G: Good | H: Yes | I: Available
|
||||
A: T10 | B: Refractor | C: 102 | D: 660 | E: 8.300000000000001 | F: 2020 | G: Excellent | H: Yes | I: Available
|
||||
A: T11 | B: Reflector on equatorial mount | C: 130 | D: 900 | E: 11.4 | F: 2014 | G: Fair | H: Yes | I: Available | J: Mount slips in declination; lent with a spare clamp
|
||||
A: T12 | B: Hydrogen-alpha solar telescope | C: 40 | D: 400 | E: 3.1 | F: 2021 | G: Good | H: No | I: Observatory use | J: Solar open day and observatory use only; never lent
|
||||
A: T13 | B: Dobsonian | C: 150 | D: 1200 | E: 17.5 | F: 2025 | G: Excellent | H: Yes | I: Available | J: Added Sept 2025; Elsworth Family Trust grant; induction required
|
||||
A: T14 | B: Refractor | C: 80 | D: 480 | E: 5.2 | F: 2025 | G: Excellent | H: Yes | I: Available | J: Added Sept 2025; Elsworth Family Trust grant
|
||||
|
||||
#### Loan log Oct 2025-Jan 2026
|
||||
|
||||
A: Loan no. | B: Telescope ID | C: Member no. | D: Borrower age group | E: Start date | F: Due date | G: Returned date | H: Days late (days) | I: Late fee (£) | J: Renewed | K: Notes
|
||||
A: L25-081 | B: T06 | C: M0412 | D: Adult | E: 2025-10-04 | F: 2025-11-01 | G: 2025-11-01 | H: 0 | I: 0 | J: No | K: Induction completed 6 Sep 2025
|
||||
A: L25-082 | B: T03 | C: M0388 | D: 14-17 | E: 2025-10-04 | F: 2025-11-01 | G: 2025-10-29 | H: 0 | I: 0 | J: No | K: Co-signed by parent, member M0387
|
||||
A: L25-083 | B: T10 | C: M0451 | D: Adult | E: 2025-10-08 | F: 2025-11-05 | G: 2025-11-08 | H: 3 | I: 6 | J: No
|
||||
A: L25-084 | B: T07 | C: M0299 | D: Adult | E: 2025-10-11 | F: 2025-11-08 | G: 2025-11-15 | H: 7 | I: 14 | J: No | K: Borrower reported dull primary mirror; T07 withdrawn 18 Nov
|
||||
A: L25-085 | B: T05 | C: M0467 | D: Adult | E: 2025-10-11 | F: 2025-11-08 | G: 2025-11-05 | H: 0 | I: 0 | J: No | K: First Light completer; 1-month rule applied
|
||||
A: L25-086 | B: T09 | C: M0320 | D: Adult | E: 2025-10-15 | F: 2025-11-26 | G: 2025-11-26 | H: 0 | I: 0 | J: Yes | K: Renewed once for 14 days; no waiting list
|
||||
A: L25-087 | B: T13 | C: M0402 | D: Adult | E: 2025-10-18 | F: 2025-11-15 | G: 2025-11-15 | H: 0 | I: 0 | J: No | K: First loan of T13
|
||||
A: L25-088 | B: T04 | C: M0475 | D: Adult | E: 2025-10-25 | F: 2025-11-22 | G: 2025-11-19 | H: 0 | I: 0 | J: No
|
||||
A: L25-089 | B: T08 | C: M0356 | D: Adult | E: 2025-10-29 | F: 2025-11-26 | G: 2025-12-10 | H: 14 | I: 28 | J: No
|
||||
A: L25-090 | B: T11 | C: M0441 | D: Adult | E: 2025-11-01 | F: 2025-11-15 | G: 2025-11-15 | H: 0 | I: 0 | J: No | K: Peak season: 14-day loan; spare clamp issued
|
||||
A: L25-091 | B: T06 | C: M0360 | D: Adult | E: 2025-11-08 | F: 2025-11-22 | G: 2025-11-22 | H: 0 | I: 0 | J: No
|
||||
A: L25-092 | B: T03 | C: M0479 | D: Adult | E: 2025-11-12 | F: 2025-11-26 | G: 2025-11-26 | H: 0 | I: 0 | J: No
|
||||
A: L25-093 | B: T10 | C: M0403 | D: Adult | E: 2025-11-15 | F: 2025-11-29 | G: 2025-12-03 | H: 4 | I: 8 | J: No
|
||||
A: L25-094 | B: T05 | C: M0333 | D: Adult | E: 2025-11-22 | F: 2025-12-06 | G: 2025-12-06 | H: 0 | I: 0 | J: No
|
||||
A: L25-095 | B: T13 | C: M0412 | D: Adult | E: 2025-11-29 | F: 2025-12-27 | G: 2025-12-27 | H: 0 | I: 0 | J: Yes | K: Renewed once for 14 days
|
||||
A: L25-096 | B: T14 | C: M0488 | D: 14-17 | E: 2025-12-06 | F: 2025-12-20 | G: 2025-12-20 | H: 0 | I: 0 | J: No | K: Co-signed by parent, member M0489
|
||||
A: L25-097 | B: T09 | C: M0451 | D: Adult | E: 2025-12-13 | F: 2025-12-27 | G: 2026-01-14 | H: 18 | I: 30 | J: No | K: Last loan of 2025; late fee capped at £30
|
||||
A: L26-001 | B: T04 | C: M0502 | D: Adult | E: 2026-01-03 | F: 2026-01-17 | G: 2026-01-17 | H: 0 | I: 0 | J: No | K: First Light completer (autumn 2025); 1-month rule applied
|
||||
A: L26-002 | B: T06 | C: M0377 | D: Adult | E: 2026-01-03 | F: 2026-01-17 | G: 2026-01-21 | H: 4 | I: 8 | J: No
|
||||
A: L26-003 | B: T08 | C: M0299 | D: Adult | E: 2026-01-10 | F: 2026-02-07 | J: Yes | K: Renewal asked for on 21 Jan 2026; still on loan
|
||||
A: L26-004 | B: T10 | C: M0415 | D: Adult | E: 2026-01-14 | F: 2026-01-28 | G: 2026-01-28 | H: 0 | I: 0 | J: No
|
||||
A: L26-005 | B: T13 | C: M0467 | D: Adult | E: 2026-01-17 | F: 2026-01-31 | G: 2026-01-31 | H: 0 | I: 0 | J: No | K: Induction completed 3 Jan 2026
|
||||
A: L26-006 | B: T03 | C: M0510 | D: Adult | E: 2026-01-24 | F: 2026-02-07 | J: No | K: Still on loan
|
||||
A: L26-007 | B: T05 | C: M0520 | D: Adult | E: 2026-01-31 | F: 2026-02-14 | J: No | K: Still on loan
|
||||
|
||||
#### Sessions 2025
|
||||
|
||||
A: Date | B: Day | C: Event type | D: Event name | E: Status | F: Reason if cancelled | G: Clear time (min) | H: Visitors (count) | I: Sky reading (mag/arcsec²) | J: Lead | K: Notes
|
||||
A: 2025-01-03 | B: Fri | C: Public evening | D: New Year sky | E: Cancelled | F: Cloud | G: 0 | H: 0 | J: Priya Lindgren | K: Forecast cloud 90% at 16:00
|
||||
A: 2025-01-21 | B: Tue | C: School visit | D: Ashcombe Primary, Year 6 | E: Held | G: 75 | H: 28 | I: 20.9 | J: Gwen Haddow | K: Planetarium used for the first 40 minutes
|
||||
A: 2025-02-07 | B: Fri | C: Public evening | D: Winter clusters | E: Held | G: 110 | H: 38 | I: 21 | J: Priya Lindgren
|
||||
A: 2025-02-13 | B: Thu | C: School visit | D: Millbeck Primary, Years 5-6 | E: Held | G: 60 | H: 30 | I: 20.7 | J: Gwen Haddow | K: Maximum group size; 4 adults
|
||||
A: 2025-02-22 | B: Sat | C: Special evening | D: Planet line-up evening | E: Cancelled | F: Cloud | G: 0 | H: 0 | J: Hamish Ottley | K: Dome also out of use (motor failure 19 Feb)
|
||||
A: 2025-03-07 | B: Fri | C: Public evening | D: Spring Moon | E: Held | G: 95 | H: 40 | J: Priya Lindgren | K: Fully booked; Annexe telescopes only; no sky reading as Moon above horizon
|
||||
A: 2025-03-12 | B: Wed | C: School visit | D: Wrexley Green Primary, Year 5 | E: Held | G: 50 | H: 24 | J: Gwen Haddow | K: Annexe telescopes only; Moon above horizon
|
||||
A: 2025-04-04 | B: Fri | C: Public evening | D: Spring sky tour | E: Held | G: 85 | H: 31 | I: 20.8 | J: Callum Steadman
|
||||
A: 2025-04-12 | B: Sat | C: Special evening | D: Spring galaxies | E: Cancelled | F: High wind | G: 0 | H: 0 | J: Callum Steadman | K: Gusts of 60 km/h forecast
|
||||
A: 2025-08-12 | B: Tue | C: Special evening | D: Perseid evening | E: Held | G: 140 | H: 36 | J: Hamish Ottley | K: Bright Moon; 23 meteors counted by visitors
|
||||
A: 2025-09-05 | B: Fri | C: Public evening | D: Autumn sky tour | E: Held | G: 120 | H: 40 | I: 20.9 | J: Priya Lindgren | K: Fully booked
|
||||
A: 2025-09-20 | B: Sat | C: Special evening | D: Saturn evening | E: Held | G: 70 | H: 27 | I: 21.1 | J: Callum Steadman | K: Cloud from 22:10
|
||||
A: 2025-10-03 | B: Fri | C: Public evening | D: Harvest sky | E: Held | G: 100 | H: 33 | J: Hamish Ottley | K: Moon above horizon; no sky reading
|
||||
A: 2025-10-16 | B: Thu | C: School visit | D: Kestrel Down School, Year 7 | E: Held | G: 80 | H: 29 | I: 21 | J: Gwen Haddow
|
||||
A: 2025-10-25 | B: Sat | C: Special evening | D: Andromeda evening | E: Cancelled | F: Cloud | G: 0 | H: 0 | J: Priya Lindgren
|
||||
A: 2025-11-07 | B: Fri | C: Public evening | D: Autumn deep sky | E: Held | G: 90 | H: 35 | J: Callum Steadman | K: Moon above horizon; no sky reading
|
||||
A: 2025-11-13 | B: Thu | C: School visit | D: Hartsmoor Primary, Year 6 | E: Held | G: 65 | H: 27 | I: 20.8 | J: Gwen Haddow
|
||||
A: 2025-12-03 | B: Wed | C: School visit | D: Ashcombe Primary, Year 5 | E: Held | G: 55 | H: 26 | J: Gwen Haddow | K: Moon above horizon
|
||||
A: 2025-12-05 | B: Fri | C: Public evening | D: Winter returns | E: Cancelled | F: Cloud | G: 0 | H: 0 | J: Priya Lindgren
|
||||
A: 2025-12-13 | B: Sat | C: Special evening | D: Geminid evening | E: Held | G: 130 | H: 38 | I: 21.2 | J: Hamish Ottley | K: 41 meteors counted by visitors
|
||||
|
||||
#### Dark-sky sites 2024-2025
|
||||
|
||||
A: Site code | B: Site name | C: Distance from observatory (km) | D: Mean reading 2024 (mag/arcsec²) | E: Mean reading 2025 (mag/arcsec²) | F: Readings in 2025 (count) | G: Change (mag/arcsec²) | H: Volunteer responsible | I: Notes
|
||||
A: S1 | B: Tallow Hill Observatory | C: 0 | D: 21.1 | E: 21.3 | F: 22 | G: 0.2 | H: Marta Ilves | I: Darkest site
|
||||
A: S2 | B: Kestrel Down | C: 3.4 | D: 20.9 | E: 21 | F: 18 | G: 0.1 | H: Hamish Ottley
|
||||
A: S3 | B: Millbeck Common | C: 5.1 | D: 20.5 | E: 20.7 | F: 15 | G: 0.2 | H: Marta Ilves
|
||||
A: S4 | B: Wrexley Green village hall | C: 3.1 | D: 20.2 | E: 20.4 | F: 19 | G: 0.2 | H: Owen Pryce | I: Brightest site; 41 pledges in the surrounding streets
|
||||
A: S5 | B: Hartsmoor Farm | C: 6.8 | D: 20.6 | E: 20.8 | F: 12 | G: 0.2 | H: Gwen Haddow | I: Lit farmyard 300 m to the south
|
||||
A: S6 | B: Ashcombe churchyard | C: 4 | D: 20.3 | E: 20.6 | F: 16 | G: 0.3 | H: Tomasz Wren | I: Largest improvement
|
||||
|
|
@ -0,0 +1,67 @@
|
|||
#### Vedlikeholdsplan 2026-2030
|
||||
|
||||
A: År | B: Tiltak | C: Bygning | D: Estimert kostnad (kr) | E: Prioritet | F: Ansvarlig | G: Status
|
||||
A: 2026 | B: Forprosjekt balkonger og fasader, del 2 (anbudsgrunnlag og anbudsrunde) | C: A, B, C, D | D: 1051500 | E: Høy | F: Ingrid Solbakk | G: Pågår; anbud innhentes høsten 2026
|
||||
A: 2026 | B: Nytt tak på sykkelbod sør | C: Sykkelbod sør | D: 118000 | E: Høy | F: Kåre Myrvold | G: Bestilt; utføres i juni
|
||||
A: 2026 | B: Maling av trapperom | C: A | D: 172000 | E: Middels | F: Torbjørn Leirvik | G: Planlagt august
|
||||
A: 2026 | B: Utskifting av ytterdører i oppganger | C: B | D: 236000 | E: Middels | F: Torbjørn Leirvik | G: Tilbud mottatt
|
||||
A: 2026 | B: Service på ventilasjonsanlegg | C: A, B, C, D | D: 48000 | E: Høy | F: Kåre Myrvold | G: Årlig service i oktober
|
||||
A: 2026 | B: Nye benker og bord ved grillplassen | C: Hage | D: 64000 | E: Lav | F: Petter Nygård-Ulvik | G: Hagekomiteen velger modell
|
||||
A: 2026 | B: Drenering langs nordveggen | C: D | D: 198000 | E: Høy | F: Torbjørn Leirvik | G: Fuktmålinger viste behov
|
||||
A: 2026 | B: Diverse småreparasjoner | C: A, B, C, D | D: 144000 | E: Middels | F: Kåre Myrvold | G: Løpende
|
||||
A: 2027 | B: Rehabilitering av balkonger og fasader | C: A | D: 10400000 | E: Høy | F: Ingrid Solbakk | G: Byggestart april 2027 (vedtatt 21.04.2026)
|
||||
A: 2027 | B: Rehabilitering av balkonger og fasader | C: B | D: 10400000 | E: Høy | F: Ingrid Solbakk | G: Byggestart april 2027 (vedtatt 21.04.2026)
|
||||
A: 2027 | B: Utskifting av de to siste vaskemaskinene | C: B | D: 68000 | E: Middels | F: Kåre Myrvold | G: Maskinene er fra 2014
|
||||
A: 2027 | B: Maling av trapperom | C: B | D: 176000 | E: Middels | F: Torbjørn Leirvik | G: Ikke startet
|
||||
A: 2027 | B: Service på ventilasjonsanlegg | C: A, B, C, D | D: 49000 | E: Høy | F: Kåre Myrvold | G: Årlig service
|
||||
A: 2028 | B: Rehabilitering av balkonger og fasader | C: C | D: 11300000 | E: Høy | F: Ingrid Solbakk | G: Byggestart april 2028 (vedtatt 21.04.2026)
|
||||
A: 2028 | B: Rehabilitering av balkonger og fasader | C: D | D: 8900000 | E: Høy | F: Ingrid Solbakk | G: Byggestart april 2028 (vedtatt 21.04.2026)
|
||||
A: 2028 | B: Nytt dekke på gangstiene i hagen | C: Uteområde | D: 142000 | E: Lav | F: Petter Nygård-Ulvik | G: Venter til stillaser er fjernet
|
||||
A: 2028 | B: Utskifting av tørketromler | C: B | D: 54000 | E: Lav | F: Kåre Myrvold | G: Vurderes etter tilstand
|
||||
A: 2029 | B: Nytt taktekke | C: A | D: 1640000 | E: Middels | F: Torbjørn Leirvik | G: Tekket er fra 1996
|
||||
A: 2029 | B: Maling av trapperom | C: C | D: 181000 | E: Lav | F: Torbjørn Leirvik | G: Ikke startet
|
||||
A: 2029 | B: Nye lekeapparater | C: Uteområde | D: 210000 | E: Middels | F: Petter Nygård-Ulvik | G: Dagens apparater er fra 2009
|
||||
A: 2030 | B: Rørfornying av stigeledninger (strømpeforing) | C: A, B, C, D | D: 4850000 | E: Høy | F: Ingrid Solbakk | G: Basert på rørinspeksjonen i oktober 2025
|
||||
A: 2030 | B: Oppgradering av fellesrommet (kjøkken og gulv) | C: D | D: 285000 | E: Lav | F: Marit Kolstø-Berge | G: Ønske fra beboerundersøkelsen 2025
|
||||
|
||||
#### Budsjett 2026
|
||||
|
||||
A: Post | B: Regnskap 2025 (kr) | C: Budsjett 2026 (kr) | D: Endring (%) | E: Merknad
|
||||
A: Felleskostnader | B: 5184000 | C: 5495000 | D: 6 | E: Økning på 6 % fra 1. januar 2026
|
||||
A: Andre driftsinntekter (vaskeri, fellesrom, nøkkelbrikker) | B: 100150 | C: 104000 | D: 3.8 | E: Uendrede priser i vaskeri og fellesrom
|
||||
A: Gebyrer for vann, avløp og renovasjon | B: 812400 | C: 853000 | D: 5 | E: Varslet økning på om lag 5 %
|
||||
A: Forsikring | B: 486000 | C: 525000 | D: 8 | E: Økning på 8 % etter vannskadene i 2025
|
||||
A: Energi i fellesarealer | B: 318700 | C: 305000 | D: -4.3 | E: Lavere forbruk etter LED-belysning
|
||||
A: Vaktmester (lønn og sosiale kostnader) | B: 642500 | C: 668000 | D: 4 | E: Lønnsoppgjør
|
||||
A: Forretningsførsel og revisjon | B: 214300 | C: 222000 | D: 3.6 | E: Ny avtale fra 1. januar 2026
|
||||
A: Vedlikehold og reparasjoner | B: 1236800 | C: 980000 | D: -20.8 | E: Se arket Vedlikeholdsplan
|
||||
A: Forprosjekt rehabilitering | B: 0 | C: 1051500 | D: Ny post | E: Restbeløp av rammen på 1,2 mill. kr; 148 500 kr ble ført som vedlikehold i 2025
|
||||
A: Styrehonorar | B: 180000 | C: 186000 | D: 3.3 | E: Vedtatt av generalforsamlingen 21. april 2026
|
||||
A: Andre driftskostnader | B: 197600 | C: 205000 | D: 3.7
|
||||
A: Renteinntekter | B: 48300 | C: 40000 | D: -17.2 | E: Lavere innskudd når forprosjektet betales
|
||||
A: Rentekostnader | B: 498200 | C: 470000 | D: -5.7 | E: Lavere lånesaldo og rente
|
||||
A: Sum driftsinntekter | B: 5284150 | C: 5599000 | D: 6
|
||||
A: Sum driftskostnader | B: 4088300 | C: 4995500 | D: 22.2 | E: Inkluderer forprosjektet
|
||||
A: Årsresultat | B: 745950 | C: 173500 | D: -76.7 | E: Budsjettert overskudd
|
||||
|
||||
#### Utført vedlikehold 2025
|
||||
|
||||
A: Dato | B: Tiltak | C: Bygning | D: Kostnad (kr) | E: Utført av | F: Merknad
|
||||
A: 2025-02-14 | B: Ombygging av tidligere lagerrom til hobbyverksted | C: C | D: 52300 | E: Vaktmester og innleid snekker | F: Åpnet for beboere 1. mars 2025
|
||||
A: 2025-03-21 | B: LED-belysning med bevegelsessensor i alle trapperom | C: A, B, C, D | D: 186000 | E: Innleid elektriker | F: Ga lavere strømforbruk i fellesarealer
|
||||
A: 2025-04-09 | B: Nytt fallunderlag under huskestativet på lekeplassen | C: Uteområde | D: 38200 | E: Innleid entreprenør | F: Gammelt underlag var slitt ned
|
||||
A: 2025-05-16 | B: Ladeskap med 8 låsbare rom for elsykkelbatterier | C: Sykkelbod sør | D: 39800 | E: Innleid elektriker | F: Lading er bare tillatt i disse skapene
|
||||
A: 2025-05-28 | B: Ny port med nøkkelbrikkeleser i sykkelbod nord | C: Sykkelbod nord | D: 31500 | E: Innleid låsesmed | F: Erstattet port med hengelås
|
||||
A: 2025-06-11 | B: Reparasjon av taklekkasje over oppgang 2 | C: C | D: 214600 | E: Innleid taktekker | F: Lekkasje oppdaget etter kraftig regn 2. juni
|
||||
A: 2025-06-30 | B: Nye busker langs gangstien og reparasjon av plen | C: Hage | D: 24300 | E: Hagekomiteen og vaktmester | F: Plantet på vårdugnaden og etterpå
|
||||
A: 2025-08-22 | B: Maling av vegger og tak i trapperom | C: D | D: 167300 | E: Innleid malerfirma | F: Arbeidet tok tre uker
|
||||
A: 2025-09-05 | B: To nye vaskemaskiner i fellesvaskeriet | C: B | D: 64800 | E: Leverandør monterte | F: De to eldste av fire maskiner ble byttet
|
||||
A: 2025-09-19 | B: Utskifting av to sluk i vaskeriet | C: B | D: 28400 | E: Innleid rørlegger | F: Gjort samtidig med maskinbyttet
|
||||
A: 2025-10-03 | B: Rørinspeksjon med kamera i alle stigeledninger | C: A, B, C, D | D: 72500 | E: Innleid rørlegger | F: Grunnlag for rørfornying i 2030
|
||||
A: 2025-10-24 | B: Service på ventilasjonsanlegg | C: A, B, C, D | D: 46800 | E: Innleid servicefirma | F: Årlig service
|
||||
A: 2025-11-07 | B: Nye nøkkelbrikkelesere ved fellesrom og vaskeri | C: B, D | D: 42600 | E: Innleid låsesmed | F: Gamle lesere sviktet i kulde
|
||||
A: 2025-11-27 | B: Kontroll og service av brannslukningsapparater | C: A, B, C, D | D: 18900 | E: Innleid servicefirma | F: Kontroll hvert femte år
|
||||
A: 2025-12-12 | B: Egenandeler for tre vannskader | C: A, C | D: 30000 | E: Forsikringsoppgjør | F: 10 000 kr per skade
|
||||
A: 2025-12-17 | B: Tilstandsrapport for balkonger og fasader (forprosjekt, del 1) | C: A, B, C, D | D: 148500 | E: Innleid rådgivende ingeniør | F: Del av forprosjektets ramme på 1,2 mill. kr
|
||||
A: 2025-12-31 | B: Diverse småreparasjoner gjennom året | C: A, B, C, D | D: 30300 | E: Vaktmester | F: Dørpumper, vinduhengsler, lysrør i boder m.m.
|
||||
B: Sum vedlikehold og reparasjoner 2025 | D: 1236800 | F: Stemmer med årsregnskapet
|
||||
BIN
gates/calibration-folder/original/aarsberetning-2025.pdf
Normal file
BIN
gates/calibration-folder/original/aarsberetning-2025.pdf
Normal file
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
BIN
gates/calibration-folder/original/generalforsamling-2026.pptx
Normal file
BIN
gates/calibration-folder/original/generalforsamling-2026.pptx
Normal file
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
BIN
gates/calibration-folder/original/lyngmo-il-aarsmelding-2025.pdf
Normal file
BIN
gates/calibration-folder/original/lyngmo-il-aarsmelding-2025.pdf
Normal file
Binary file not shown.
BIN
gates/calibration-folder/original/lyngmo-il-aarsmote-2026.pptx
Normal file
BIN
gates/calibration-folder/original/lyngmo-il-aarsmote-2026.pptx
Normal file
Binary file not shown.
Binary file not shown.
Binary file not shown.
BIN
gates/calibration-folder/original/medlemshaandbok-2026.docx
Normal file
BIN
gates/calibration-folder/original/medlemshaandbok-2026.docx
Normal file
Binary file not shown.
BIN
gates/calibration-folder/original/medlemsmote-mars-2026.pptx
Normal file
BIN
gates/calibration-folder/original/medlemsmote-mars-2026.pptx
Normal file
Binary file not shown.
Binary file not shown.
BIN
gates/calibration-folder/original/sesongevaluering-2025.pdf
Normal file
BIN
gates/calibration-folder/original/sesongevaluering-2025.pdf
Normal file
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
16735
gates/calibration-folder/pairs.json
Normal file
16735
gates/calibration-folder/pairs.json
Normal file
File diff suppressed because it is too large
Load diff
145
gates/calibration.manifest.json
Normal file
145
gates/calibration.manifest.json
Normal file
|
|
@ -0,0 +1,145 @@
|
|||
{
|
||||
"schema": "okf-wiki/calibration-manifest/2",
|
||||
"seed": 20261008,
|
||||
"shuffle": "random.Random(seed): each label's pairs sorted by draft id and shuffled; the first `lone` unsupported pairs alone in a call at positions spread evenly over 1-10, the rest `dense` to a call; counted supported pairs one per call in call order; filler pairs fill every call to ten in a cycle, never twice in a call and never beside the unsupported pair they are the twin of; the other positions shuffled; ids renumbered c001.. in call order, calls k001..",
|
||||
"set": {
|
||||
"file": "gates-local/calibration/pairs-300.json",
|
||||
"sha256": "fd0c9c975bafd8770bab749e1faf9b38b0711491a4f9f433b40ef2004b4c27f7",
|
||||
"pairs": 1400
|
||||
},
|
||||
"key": {
|
||||
"file": "gates-local/calibration/key-300.json",
|
||||
"sha256": "97c6fada731d5718b94a543867d1adac81547e96072e27fc3c0c7fae324fb4d2"
|
||||
},
|
||||
"calls": {
|
||||
"file": "gates-local/calibration/calls-300.json",
|
||||
"sha256": "a356abdab4e0eb868fbcdb77e0a634c1bfa28648c5a5d6af1ac5a09748e52709",
|
||||
"calls": 140,
|
||||
"call_size_max": 10,
|
||||
"unsupported_alone": 100,
|
||||
"unsupported_per_call": {
|
||||
"1": 100,
|
||||
"5": 40
|
||||
},
|
||||
"built_with": {
|
||||
"lone": 100,
|
||||
"dense": 5
|
||||
}
|
||||
},
|
||||
"counts": {
|
||||
"unsupported": 300,
|
||||
"supported": 50,
|
||||
"filler-supported": 1050,
|
||||
"filler_distinct": 250,
|
||||
"by_kind": {
|
||||
"overgeneralisation": 50,
|
||||
"wrong-number": 50,
|
||||
"swapped-cause": 50,
|
||||
"not-mentioned": 50,
|
||||
"reversed-polarity": 50,
|
||||
"conflated-sources": 50,
|
||||
"paraphrase": 15,
|
||||
"inference": 20,
|
||||
"synthesis": 15
|
||||
},
|
||||
"by_position": {
|
||||
"1": {
|
||||
"unsupported": 33,
|
||||
"supported": 8,
|
||||
"filler-supported": 99
|
||||
},
|
||||
"2": {
|
||||
"unsupported": 30,
|
||||
"supported": 5,
|
||||
"filler-supported": 105
|
||||
},
|
||||
"3": {
|
||||
"unsupported": 28,
|
||||
"supported": 6,
|
||||
"filler-supported": 106
|
||||
},
|
||||
"4": {
|
||||
"unsupported": 23,
|
||||
"supported": 5,
|
||||
"filler-supported": 112
|
||||
},
|
||||
"5": {
|
||||
"unsupported": 32,
|
||||
"supported": 6,
|
||||
"filler-supported": 102
|
||||
},
|
||||
"6": {
|
||||
"unsupported": 23,
|
||||
"supported": 4,
|
||||
"filler-supported": 113
|
||||
},
|
||||
"7": {
|
||||
"unsupported": 30,
|
||||
"supported": 6,
|
||||
"filler-supported": 104
|
||||
},
|
||||
"8": {
|
||||
"unsupported": 24,
|
||||
"supported": 3,
|
||||
"filler-supported": 113
|
||||
},
|
||||
"9": {
|
||||
"unsupported": 39,
|
||||
"supported": 1,
|
||||
"filler-supported": 100
|
||||
},
|
||||
"10": {
|
||||
"unsupported": 38,
|
||||
"supported": 6,
|
||||
"filler-supported": 96
|
||||
}
|
||||
}
|
||||
},
|
||||
"thresholds": {
|
||||
"caught_min": 300,
|
||||
"accepted_min": 45
|
||||
},
|
||||
"sources": {
|
||||
"files": 599,
|
||||
"tree_sha256": "6daf3ed73335f3e6f332c7b188adb5aebbc659b6fdd1d7591280e129648a06d5"
|
||||
},
|
||||
"history": [
|
||||
{
|
||||
"schema": "okf-wiki/calibration-manifest/1",
|
||||
"seed": 20260927,
|
||||
"shuffle": "random.Random(seed).shuffle(pairs sorted by draft id); renumbered c001..",
|
||||
"set": {
|
||||
"file": "gates-local/calibration/pairs-100.json",
|
||||
"sha256": "bb9d303840dd43eff8c55304c9ce9d935922678a84fe97022c6c721e9fa61d0d",
|
||||
"pairs": 100
|
||||
},
|
||||
"key": {
|
||||
"file": "gates-local/calibration/key-100.json",
|
||||
"sha256": "b30a37019dfe224474a0cf1b1aa4a82ccda2c332a035ecd4fb1b4500f78343ca"
|
||||
},
|
||||
"counts": {
|
||||
"unsupported": 50,
|
||||
"supported": 50,
|
||||
"by_kind": {
|
||||
"overgeneralisation": 9,
|
||||
"wrong-number": 9,
|
||||
"swapped-cause": 8,
|
||||
"not-mentioned": 8,
|
||||
"reversed-polarity": 8,
|
||||
"conflated-sources": 8,
|
||||
"paraphrase": 15,
|
||||
"inference": 20,
|
||||
"synthesis": 15
|
||||
}
|
||||
},
|
||||
"thresholds": {
|
||||
"caught_min": 50,
|
||||
"accepted_min": 45
|
||||
},
|
||||
"sources": {
|
||||
"files": 599,
|
||||
"tree_sha256": "6daf3ed73335f3e6f332c7b188adb5aebbc659b6fdd1d7591280e129648a06d5"
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
26
gates/calibration.passed.json
Normal file
26
gates/calibration.passed.json
Normal file
|
|
@ -0,0 +1,26 @@
|
|||
{
|
||||
"schema": "okf-wiki/calibration-passed/1",
|
||||
"passed": [
|
||||
{
|
||||
"model": "claude-opus-5-5",
|
||||
"set_sha256": "fd0c9c975bafd8770bab749e1faf9b38b0711491a4f9f433b40ef2004b4c27f7",
|
||||
"key_sha256": "97c6fada731d5718b94a543867d1adac81547e96072e27fc3c0c7fae324fb4d2",
|
||||
"verdicts_sha256": "189cf3a41e41b60c376844e3412aa57fffb47df13aa6b51a1871ffd00fa51aeb",
|
||||
"reader": "okf-wiki-reader"
|
||||
},
|
||||
{
|
||||
"model": "claude-opus-5-5",
|
||||
"set_sha256": "852b50d8fcd7ef248b40d43ec7b8dc28379163e6ddd64c37e40f3b170b97b61e",
|
||||
"key_sha256": "af728bba35f4fe2947163adf3053f0b7bc3390ca5179df19f3b8abdd8fe0f641",
|
||||
"verdicts_sha256": "2d5f04a9b39b5c7ac58f92201144b1999cfbdd8be4a844b5439c057679392383",
|
||||
"reader": "okf-wiki-reader"
|
||||
},
|
||||
{
|
||||
"model": "claude-opus-5-5",
|
||||
"set_sha256": "fd0c9c975bafd8770bab749e1faf9b38b0711491a4f9f433b40ef2004b4c27f7",
|
||||
"key_sha256": "97c6fada731d5718b94a543867d1adac81547e96072e27fc3c0c7fae324fb4d2",
|
||||
"verdicts_sha256": "e959492441c9f41b8ce1e697e33c96be2bc11990a4a46eba73cf8eff4cf96d25",
|
||||
"reader": "okf-wiki-reader"
|
||||
}
|
||||
]
|
||||
}
|
||||
860
gates/calibration.py
Normal file
860
gates/calibration.py
Normal file
|
|
@ -0,0 +1,860 @@
|
|||
"""The reader calibration for PLAN.md criterion 2: set checker, seeded build with a call layout, stub reader and verdict
|
||||
gate. No model call.
|
||||
|
||||
The reader judges up to ten claims per call and gives one verdict per claim (K3 in docs/bolge1/0-felles.local.md). The
|
||||
calibration therefore comes as three frozen files beside the manifest: the set (claims and excerpts, what the reader
|
||||
sees), the key (label and kind per claim) and the call layout (which claims stand together in a call, and where).
|
||||
|
||||
Labels: "unsupported" (planted, must be caught), "supported" (counted controls, must be accepted) and
|
||||
"filler-supported" (supported claims that fill the calls so they resemble a real page; reported apart and never
|
||||
counted in the thresholds).
|
||||
|
||||
check every excerpt stands verbatim in its source file (and, in a built set, at its stated line), every kind of
|
||||
the two counted labels is present and fits its label, a wrong-number claim carries a number the excerpts
|
||||
do not, a supported or filler claim is not a verbatim copy of its excerpt, and a twin_of names an
|
||||
unsupported pair. A built set is checked with its layout against the manifest: sha256 of set, key, layout
|
||||
and sources, the layout rules, and the counts per label, kind and position.
|
||||
build checks a labelled draft, lays it out in calls with a fixed seed, renumbers the claims (c001.., in call
|
||||
order) and the calls (k001..), and writes the set, the key, the layout and a manifest with sha256, counts
|
||||
and thresholds only. A built set is frozen: build never overwrites it. An earlier manifest at the same
|
||||
path moves into the new one's history.
|
||||
stub writes a verdict file without a reader: the same verdict for every claim, or the key's verdicts, and
|
||||
optionally "supported" for every claim at one position (the position-mild stub). Its model is "stub".
|
||||
gate scores one reader's verdicts against the key: totals, caught per kind and caught per position, and the
|
||||
fillers apart; never a result per call or per pair. A missing, invalid or duplicated verdict counts as
|
||||
wrong (fail-closed). GREEN only when every unsupported claim is caught at every position, the counted
|
||||
supported claims reach the manifest's threshold, no verdict is missing, and the verdicts name exactly
|
||||
one model (their "models", written by okfwiki.reader collect from the readers' transcripts; PLAN
|
||||
§ Rammer 23). It prints the model it scored. With --passed FILE a GREEN run is added to FILE, the record
|
||||
of green runs (okf-wiki/calibration-passed/1: model, set_sha256, key_sha256, verdicts_sha256, reader; no
|
||||
time, and an entry already there is not added twice); a RED run never writes it. okfwiki.reader collect
|
||||
--gate-passed FILE turns a reading into verdicts only for a model in FILE with a run on the set and key
|
||||
of the tracked manifest, so a model change, or a new set, needs a new gate run. Both read FILE with
|
||||
okfwiki.reader.read_passed.
|
||||
|
||||
Exit 0 = PASS / GREEN / written; 1 = FAIL / RED or could not run. A check that cannot run is a FAIL.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import math
|
||||
import random
|
||||
import re
|
||||
import sys
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
|
||||
|
||||
from okfwiki import reader # noqa: E402
|
||||
|
||||
DRAFT_SCHEMA = "okf-wiki/calibration-draft/1"
|
||||
SET_SCHEMA = "okf-wiki/calibration-set/1"
|
||||
KEY_SCHEMA = "okf-wiki/calibration-key/1"
|
||||
CALLS_SCHEMA = "okf-wiki/reader-calls/1"
|
||||
MANIFEST_SCHEMA = "okf-wiki/calibration-manifest/2"
|
||||
PASSED_SCHEMA = reader.PASSED_SCHEMA
|
||||
STUB_MODEL = "stub"
|
||||
VERDICTS_SCHEMA_NOTE = (
|
||||
'{"reader": "<name>", "models": ["<model>"], "verdicts": [{"id": "c001", "verdict": "supported" | "unsupported"}]}'
|
||||
)
|
||||
|
||||
UNSUPPORTED = "unsupported"
|
||||
SUPPORTED = "supported"
|
||||
FILLER = "filler-supported"
|
||||
LABELS = (UNSUPPORTED, SUPPORTED, FILLER)
|
||||
COUNTED = (UNSUPPORTED, SUPPORTED)
|
||||
VERDICTS = (SUPPORTED, UNSUPPORTED)
|
||||
UNSUPPORTED_KINDS = (
|
||||
"overgeneralisation",
|
||||
"wrong-number",
|
||||
"swapped-cause",
|
||||
"not-mentioned",
|
||||
"reversed-polarity",
|
||||
"conflated-sources",
|
||||
)
|
||||
SUPPORTED_KINDS = ("paraphrase", "inference", "synthesis")
|
||||
KINDS = {UNSUPPORTED: UNSUPPORTED_KINDS, SUPPORTED: SUPPORTED_KINDS, FILLER: SUPPORTED_KINDS}
|
||||
TWO_EXCERPT_KINDS = ("conflated-sources", "synthesis")
|
||||
MAX_EXCERPTS = 2
|
||||
CLAIM_MIN, CLAIM_MAX = 10, 400
|
||||
EXCERPT_MIN, EXCERPT_MAX = 20, 1500
|
||||
CALL_SIZE_MAX = 10 # K3 point 1: up to ten claims per call, positions 1-10
|
||||
ACCEPTED_SHARE = 0.9 # PM's choice (order E3): a reader that rejects more than 10 % of supported claims is unusable.
|
||||
NUMBER = re.compile(r"\d+(?:[.,]\d+)*")
|
||||
LAYOUT_STATS = ("calls", "call_size_max", "unsupported_alone", "unsupported_per_call")
|
||||
LAYOUT_RULE = (
|
||||
"random.Random(seed): each label's pairs sorted by draft id and shuffled; the first `lone` unsupported pairs alone"
|
||||
" in a call at positions spread evenly over 1-10, the rest `dense` to a call; counted supported pairs one per call"
|
||||
" in call order; filler pairs fill every call to ten in a cycle, never twice in a call and never beside the"
|
||||
" unsupported pair they are the twin of; the other positions shuffled; ids renumbered c001.. in call order,"
|
||||
" calls k001.."
|
||||
)
|
||||
|
||||
|
||||
class RunError(Exception):
|
||||
"""The command cannot run; reported as FAIL with exit 1."""
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- helpers
|
||||
|
||||
|
||||
def sha256_bytes(data: bytes) -> str:
|
||||
return hashlib.sha256(data).hexdigest()
|
||||
|
||||
|
||||
def tree_sha256(sources: Path) -> str:
|
||||
"""One hash over the flat sources directory: sorted "name NUL sha256 LF" lines (same as gates/multidoc.py)."""
|
||||
lines = []
|
||||
for p in sorted(sources.iterdir(), key=lambda p: p.name):
|
||||
if p.is_file():
|
||||
lines.append(f"{p.name}\0{sha256_bytes(p.read_bytes())}\n")
|
||||
return sha256_bytes("".join(lines).encode("utf-8"))
|
||||
|
||||
|
||||
def count_files(sources: Path) -> int:
|
||||
return sum(1 for p in sources.iterdir() if p.is_file())
|
||||
|
||||
|
||||
def normalise(text: str) -> str:
|
||||
"""Case-folded, whitespace collapsed, outer punctuation stripped. Used for the verbatim-copy test only."""
|
||||
return re.sub(r"\s+", " ", text).strip().strip(".,;:!?").strip().casefold()
|
||||
|
||||
|
||||
def numbers(text: str) -> set[str]:
|
||||
return {n.replace(",", "") for n in NUMBER.findall(text)}
|
||||
|
||||
|
||||
def as_list(value: object) -> list[Any]:
|
||||
return value if isinstance(value, list) else []
|
||||
|
||||
|
||||
def as_dict(value: object) -> dict[str, Any]:
|
||||
return value if isinstance(value, dict) else {}
|
||||
|
||||
|
||||
def is_plain_name(name: object) -> bool:
|
||||
return isinstance(name, str) and name != "" and "/" not in name and "\\" not in name and name not in (".", "..")
|
||||
|
||||
|
||||
def load_json(path: Path, what: str) -> Any:
|
||||
try:
|
||||
return json.loads(path.read_text(encoding="utf-8"))
|
||||
except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
|
||||
raise RunError(f"cannot read {what} {path}: {exc}") from exc
|
||||
|
||||
|
||||
def load_list(path: Path, schema: str, key: str, what: str) -> list[dict[str, Any]]:
|
||||
data = load_json(path, what)
|
||||
if not isinstance(data, dict) or data.get("schema") != schema or not isinstance(data.get(key), list):
|
||||
raise RunError(f"{path} is not a {what} with schema {schema}")
|
||||
items: list[dict[str, Any]] = data[key]
|
||||
return items
|
||||
|
||||
|
||||
def require_dir(path: Path, what: str) -> None:
|
||||
if not path.is_dir():
|
||||
raise RunError(f"{what} {path} is not a directory")
|
||||
|
||||
|
||||
def file_sha256(path: Path) -> str:
|
||||
return sha256_bytes(path.read_bytes()) if path.is_file() else "(missing)"
|
||||
|
||||
|
||||
def line_of(text: str, excerpt: str, line: int) -> bool:
|
||||
"""True when an occurrence of the excerpt starts on the given 1-based line."""
|
||||
start = text.find(excerpt)
|
||||
while start != -1:
|
||||
if text.count("\n", 0, start) + 1 == line:
|
||||
return True
|
||||
start = text.find(excerpt, start + 1)
|
||||
return False
|
||||
|
||||
|
||||
def first_line(text: str, excerpt: str) -> int:
|
||||
return text.count("\n", 0, text.find(excerpt)) + 1
|
||||
|
||||
|
||||
def verdict_for(label: object) -> str:
|
||||
"""The right verdict for a label: only an unsupported claim should be rejected."""
|
||||
return UNSUPPORTED if label == UNSUPPORTED else SUPPORTED
|
||||
|
||||
|
||||
class SourceCache:
|
||||
def __init__(self, root: Path) -> None:
|
||||
self.root = root
|
||||
self._text: dict[str, str | None] = {}
|
||||
|
||||
def text(self, name: str) -> str | None:
|
||||
"""File text, or None when the name is not a plain file directly in the sources directory."""
|
||||
if not is_plain_name(name):
|
||||
return None
|
||||
if name not in self._text:
|
||||
p = self.root / name
|
||||
try:
|
||||
self._text[name] = p.read_text(encoding="utf-8") if p.is_file() else None
|
||||
except (OSError, UnicodeDecodeError):
|
||||
self._text[name] = None
|
||||
return self._text[name]
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- check
|
||||
|
||||
|
||||
@dataclass
|
||||
class CheckReport:
|
||||
pairs: int = 0
|
||||
unsupported: int = 0
|
||||
supported: int = 0
|
||||
filler: int = 0
|
||||
excerpts: int = 0
|
||||
excerpts_verbatim: int = 0
|
||||
lines_checked: int = 0
|
||||
lines_correct: int = 0
|
||||
by_kind: dict[str, int] = field(default_factory=lambda: {k: 0 for lab in COUNTED for k in KINDS[lab]})
|
||||
filler_by_kind: dict[str, int] = field(default_factory=lambda: {k: 0 for k in KINDS[FILLER]})
|
||||
errors: list[str] = field(default_factory=list)
|
||||
notes: list[str] = field(default_factory=list)
|
||||
|
||||
@property
|
||||
def passed(self) -> bool:
|
||||
return (
|
||||
self.pairs > 0
|
||||
and not self.errors
|
||||
and self.excerpts_verbatim == self.excerpts
|
||||
and self.lines_correct == self.lines_checked
|
||||
)
|
||||
|
||||
def lines(self) -> list[str]:
|
||||
out = [f"error: {e}" for e in self.errors]
|
||||
out += [
|
||||
f"pairs: {self.pairs} ({self.unsupported} {UNSUPPORTED}, {self.supported} {SUPPORTED}, {self.filler} {FILLER})",
|
||||
f"excerpts verbatim: {self.excerpts_verbatim} of {self.excerpts}",
|
||||
]
|
||||
if self.lines_checked:
|
||||
out.append(f"excerpt lines correct: {self.lines_correct} of {self.lines_checked}")
|
||||
for label in COUNTED:
|
||||
out.append(f"{label} by kind: " + ", ".join(f"{k} {self.by_kind[k]}" for k in KINDS[label]))
|
||||
out.append(f"{FILLER} by kind: " + ", ".join(f"{k} {self.filler_by_kind[k]}" for k in KINDS[FILLER]))
|
||||
out += self.notes
|
||||
out.append("PASS" if self.passed else "FAIL")
|
||||
return out
|
||||
|
||||
|
||||
def check_pairs(pairs: list[dict[str, Any]], sources: Path, with_lines: bool = False) -> CheckReport:
|
||||
"""Check labelled pairs ({id, label, kind, claim, excerpts: [{file, text[, line]}][, twin_of]}) against the sources."""
|
||||
require_dir(sources, "sources")
|
||||
cache = SourceCache(sources)
|
||||
rep = CheckReport(pairs=len(pairs))
|
||||
unsupported_ids = {p.get("id") for p in pairs if isinstance(p, dict) and p.get("label") == UNSUPPORTED}
|
||||
seen: set[str] = set()
|
||||
for p in pairs:
|
||||
pid = p.get("id") if isinstance(p, dict) else None
|
||||
if not isinstance(pid, str) or not pid:
|
||||
rep.errors.append("pair without an id")
|
||||
continue
|
||||
if pid in seen:
|
||||
rep.errors.append(f"duplicate id {pid}")
|
||||
seen.add(pid)
|
||||
label, kind, claim = p.get("label"), p.get("kind"), p.get("claim")
|
||||
if label not in KINDS:
|
||||
rep.errors.append(f"{pid}: label {label!r} is not one of {', '.join(LABELS)}")
|
||||
elif kind not in KINDS[label]:
|
||||
rep.errors.append(f"{pid}: kind {kind!r} does not belong to label {label}")
|
||||
elif label == UNSUPPORTED:
|
||||
rep.unsupported += 1
|
||||
rep.by_kind[kind] += 1
|
||||
elif label == SUPPORTED:
|
||||
rep.supported += 1
|
||||
rep.by_kind[kind] += 1
|
||||
else:
|
||||
rep.filler += 1
|
||||
rep.filler_by_kind[kind] += 1
|
||||
twin = p.get("twin_of")
|
||||
if twin is not None and (label != FILLER or twin not in unsupported_ids):
|
||||
rep.errors.append(f"{pid}: twin_of {twin!r} does not name an unsupported pair of a {FILLER} pair")
|
||||
if not isinstance(claim, str) or not CLAIM_MIN <= len(claim.strip()) <= CLAIM_MAX:
|
||||
rep.errors.append(f"{pid}: claim missing or length outside {CLAIM_MIN}-{CLAIM_MAX}")
|
||||
claim = ""
|
||||
excerpts = as_list(p.get("excerpts"))
|
||||
if not 1 <= len(excerpts) <= MAX_EXCERPTS:
|
||||
rep.errors.append(f"{pid}: {len(excerpts)} excerpts, expected 1-{MAX_EXCERPTS}")
|
||||
texts: list[str] = []
|
||||
files: set[str] = set()
|
||||
for e in excerpts:
|
||||
rep.excerpts += 1
|
||||
name, text = (e.get("file"), e.get("text")) if isinstance(e, dict) else (None, None)
|
||||
if not is_plain_name(name) or not isinstance(name, str):
|
||||
rep.errors.append(f"{pid}: excerpt file {name!r} is not a plain file name")
|
||||
continue
|
||||
if not isinstance(text, str):
|
||||
rep.errors.append(f"{pid}: excerpt in {name} is not text")
|
||||
continue
|
||||
if not EXCERPT_MIN <= len(text) <= EXCERPT_MAX:
|
||||
rep.errors.append(f"{pid}: excerpt length {len(text)} in {name} outside {EXCERPT_MIN}-{EXCERPT_MAX}")
|
||||
files.add(name)
|
||||
texts.append(text)
|
||||
source = cache.text(name)
|
||||
if source is not None and text in source:
|
||||
rep.excerpts_verbatim += 1
|
||||
if with_lines:
|
||||
rep.lines_checked += 1
|
||||
line = e.get("line")
|
||||
if source is not None and isinstance(line, int) and line_of(source, text, line):
|
||||
rep.lines_correct += 1
|
||||
if kind in TWO_EXCERPT_KINDS and len(excerpts) != 2:
|
||||
rep.errors.append(f"{pid}: kind {kind} needs two excerpts")
|
||||
if kind == "conflated-sources" and len(files) != 2:
|
||||
rep.errors.append(f"{pid}: kind conflated-sources needs excerpts from two different files")
|
||||
if kind == "wrong-number" and not numbers(claim) - set().union(*(numbers(t) for t in texts)):
|
||||
rep.errors.append(f"{pid}: wrong-number claim has no number that is absent from the excerpts")
|
||||
if label in (SUPPORTED, FILLER) and claim and any(normalise(claim) in normalise(t) for t in texts):
|
||||
rep.errors.append(f"{pid}: {label} claim is a verbatim copy of its excerpt")
|
||||
for label in COUNTED:
|
||||
for k in KINDS[label]:
|
||||
if rep.by_kind[k] == 0:
|
||||
rep.errors.append(f"kind {k} has no pair")
|
||||
return rep
|
||||
|
||||
|
||||
def join_set_and_key(set_path: Path, key_path: Path) -> list[dict[str, Any]]:
|
||||
pairs = load_list(set_path, SET_SCHEMA, "pairs", "set")
|
||||
labels = load_list(key_path, KEY_SCHEMA, "labels", "key")
|
||||
by_id = {k.get("id"): k for k in labels if isinstance(k, dict)}
|
||||
ids = [p.get("id") for p in pairs if isinstance(p, dict)]
|
||||
if set(ids) != set(by_id) or len(ids) != len(labels) or len(ids) != len(pairs):
|
||||
raise RunError("the key does not label exactly the pairs of the set")
|
||||
return [
|
||||
dict(p, label=by_id[p["id"]].get("label"), kind=by_id[p["id"]].get("kind"), draft_id=by_id[p["id"]].get("draft_id"))
|
||||
for p in pairs
|
||||
]
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- layout
|
||||
|
||||
|
||||
@dataclass
|
||||
class Layout:
|
||||
"""What a call layout says, measured against the labels: positions, counts, and every rule it breaks."""
|
||||
|
||||
position: dict[str, int] = field(default_factory=dict)
|
||||
stats: dict[str, Any] = field(default_factory=dict)
|
||||
by_position: dict[str, dict[str, int]] = field(default_factory=dict)
|
||||
errors: list[str] = field(default_factory=list)
|
||||
|
||||
|
||||
def shown(ids: list[str]) -> str:
|
||||
return ", ".join(ids[:5]) + (", ..." if len(ids) > 5 else "")
|
||||
|
||||
|
||||
def check_layout(layout: object, labels: dict[str, str], set_sha256: str) -> Layout:
|
||||
"""K3 point 1: every claim of the set in exactly one call, at most call_size_max (<= 10) claims per call, the
|
||||
layout bound to the set by its sha256, and unsupported claims at every position 1-10."""
|
||||
out = Layout()
|
||||
if not isinstance(layout, dict) or layout.get("schema") != CALLS_SCHEMA:
|
||||
out.errors.append(f"the layout is not a layout with schema {CALLS_SCHEMA}")
|
||||
return out
|
||||
size_max = layout.get("call_size_max")
|
||||
if not isinstance(size_max, int) or isinstance(size_max, bool) or not 1 <= size_max <= CALL_SIZE_MAX:
|
||||
out.errors.append(f"call_size_max {size_max!r} is not a whole number from 1 to {CALL_SIZE_MAX}")
|
||||
size_max = CALL_SIZE_MAX
|
||||
if layout.get("set_sha256") != set_sha256:
|
||||
out.errors.append(f"the layout's set_sha256 {layout.get('set_sha256')!r} is not the set's sha256 {set_sha256}")
|
||||
calls = layout.get("calls")
|
||||
if not isinstance(calls, list) or not calls:
|
||||
out.errors.append("the layout has no calls")
|
||||
return out
|
||||
seen: dict[str, int] = {}
|
||||
names: set[str] = set()
|
||||
per_call: dict[int, int] = {}
|
||||
for c in calls:
|
||||
name, ids = (c.get("call"), c.get("ids")) if isinstance(c, dict) else (None, None)
|
||||
if not isinstance(name, str) or not isinstance(ids, list) or not ids or not all(isinstance(i, str) for i in ids):
|
||||
out.errors.append(f"call {name!r} is not a named call with a list of claim ids")
|
||||
continue
|
||||
if name in names:
|
||||
out.errors.append(f"call name {name} is given twice")
|
||||
names.add(name)
|
||||
if len(ids) > size_max:
|
||||
out.errors.append(f"call {name} has {len(ids)} claims, more than {size_max}")
|
||||
for n, i in enumerate(ids, 1):
|
||||
seen[i] = seen.get(i, 0) + 1
|
||||
out.position[i] = n
|
||||
u = sum(1 for i in ids if labels.get(i) == UNSUPPORTED)
|
||||
per_call[u] = per_call.get(u, 0) + 1
|
||||
twice = sorted(i for i, n in seen.items() if n > 1)
|
||||
if twice:
|
||||
out.errors.append(f"{len(twice)} claim id(s) stand in more than one place: {shown(twice)}")
|
||||
unknown = sorted(i for i in seen if i not in labels)
|
||||
if unknown:
|
||||
out.errors.append(f"{len(unknown)} claim id(s) of the layout are not in the set: {shown(unknown)}")
|
||||
missing = sorted(i for i in labels if i not in seen)
|
||||
if missing:
|
||||
out.errors.append(f"{len(missing)} claim id(s) of the set stand in no call: {shown(missing)}")
|
||||
out.by_position = {str(n): {lab: 0 for lab in LABELS} for n in range(1, CALL_SIZE_MAX + 1)}
|
||||
for i, n in out.position.items():
|
||||
if i in labels and str(n) in out.by_position and labels[i] in LABELS:
|
||||
out.by_position[str(n)][labels[i]] += 1
|
||||
for n in range(1, CALL_SIZE_MAX + 1):
|
||||
if out.by_position[str(n)][UNSUPPORTED] == 0:
|
||||
out.errors.append(f"position {n} has no unsupported claim")
|
||||
out.stats = {
|
||||
"calls": len(calls),
|
||||
"call_size_max": size_max,
|
||||
"unsupported_alone": per_call.get(1, 0),
|
||||
"unsupported_per_call": {str(k): per_call[k] for k in sorted(per_call)},
|
||||
}
|
||||
return out
|
||||
|
||||
|
||||
def layout_lines(lay: Layout) -> list[str]:
|
||||
per_call = lay.stats.get("unsupported_per_call", {})
|
||||
return [
|
||||
f"calls: {lay.stats.get('calls', 0)} (at most {lay.stats.get('call_size_max', CALL_SIZE_MAX)} claims each)",
|
||||
"unsupported per call: " + ", ".join(f"{k} in {v} calls" for k, v in per_call.items()),
|
||||
f"unsupported alone in their call: {lay.stats.get('unsupported_alone', 0)}",
|
||||
"unsupported by position: "
|
||||
+ ", ".join(f"{n} {lay.by_position.get(str(n), {}).get(UNSUPPORTED, 0)}" for n in range(1, CALL_SIZE_MAX + 1)),
|
||||
]
|
||||
|
||||
|
||||
def positions_of(layout: object) -> dict[str, int]:
|
||||
"""Claim id -> position 1.. in its call, read leniently (the stub's use; the gate checks the layout itself)."""
|
||||
out: dict[str, int] = {}
|
||||
for c in as_list(as_dict(layout).get("calls")):
|
||||
for n, i in enumerate(as_list(as_dict(c).get("ids")), 1):
|
||||
if isinstance(i, str):
|
||||
out[i] = n
|
||||
return out
|
||||
|
||||
|
||||
def tally(pairs: list[dict[str, Any]]) -> dict[str, Any]:
|
||||
"""Counts per label and per kind of the two counted labels, and the distinct filler pairs behind the slots."""
|
||||
out: dict[str, Any] = {lab: sum(1 for p in pairs if p.get("label") == lab) for lab in LABELS}
|
||||
out["filler_distinct"] = len({str(p.get("draft_id", p.get("id"))) for p in pairs if p.get("label") == FILLER})
|
||||
out["by_kind"] = {
|
||||
k: sum(1 for p in pairs if p.get("label") == lab and p.get("kind") == k) for lab in COUNTED for k in KINDS[lab]
|
||||
}
|
||||
return out
|
||||
|
||||
|
||||
@dataclass
|
||||
class Frozen:
|
||||
pairs: list[dict[str, Any]]
|
||||
layout: Layout
|
||||
manifest: dict[str, Any]
|
||||
errors: list[str]
|
||||
|
||||
|
||||
def load_frozen(set_path: Path, key_path: Path, calls_path: Path, manifest_path: Path, sources: Path) -> Frozen:
|
||||
"""The set, key and layout with every disagreement with the manifest listed (sha256, sources, layout, counts)."""
|
||||
require_dir(sources, "sources")
|
||||
manifest = load_json(manifest_path, "manifest")
|
||||
if not isinstance(manifest, dict) or manifest.get("schema") != MANIFEST_SCHEMA:
|
||||
raise RunError(f"{manifest_path} is not a manifest with schema {MANIFEST_SCHEMA}")
|
||||
errors: list[str] = []
|
||||
for path, part in ((set_path, "set"), (key_path, "key"), (calls_path, "calls")):
|
||||
digest = file_sha256(path)
|
||||
if digest != as_dict(manifest.get(part)).get("sha256"):
|
||||
errors.append(f"{part} {path} sha256 {digest} does not match the manifest")
|
||||
if tree_sha256(sources) != as_dict(manifest.get("sources")).get("tree_sha256"):
|
||||
errors.append(f"sources {sources} differ from the manifest's tree_sha256; the set is frozen against them")
|
||||
pairs = join_set_and_key(set_path, key_path)
|
||||
labels = {str(p["id"]): str(p.get("label")) for p in pairs}
|
||||
layout = check_layout(load_json(calls_path, "layout"), labels, file_sha256(set_path))
|
||||
errors += layout.errors
|
||||
counts = as_dict(manifest.get("counts"))
|
||||
for k, v in {**tally(pairs), "by_position": layout.by_position}.items():
|
||||
if counts.get(k) != v:
|
||||
errors.append(f"the manifest's counts.{k} differs from the set and layout" + ("" if isinstance(v, dict) else f" ({counts.get(k)!r}, measured {v!r})"))
|
||||
calls = as_dict(manifest.get("calls"))
|
||||
for k in LAYOUT_STATS:
|
||||
if k in layout.stats and calls.get(k) != layout.stats[k]:
|
||||
errors.append(f"the manifest's calls.{k} differs from the layout ({calls.get(k)!r}, measured {layout.stats[k]!r})")
|
||||
return Frozen(pairs, layout, manifest, errors)
|
||||
|
||||
|
||||
def cmd_check(args: argparse.Namespace) -> int:
|
||||
sources = Path(args.sources)
|
||||
if args.draft:
|
||||
rep = check_pairs(load_list(Path(args.draft), DRAFT_SCHEMA, "pairs", "draft"), sources)
|
||||
elif args.set and args.key and args.calls and args.manifest:
|
||||
frozen = load_frozen(Path(args.set), Path(args.key), Path(args.calls), Path(args.manifest), sources)
|
||||
rep = check_pairs(frozen.pairs, sources, with_lines=True)
|
||||
rep.errors = frozen.errors + rep.errors
|
||||
rep.notes = layout_lines(frozen.layout)
|
||||
else:
|
||||
raise RunError("check needs --draft, or --set, --key, --calls and --manifest")
|
||||
print("\n".join(rep.lines()))
|
||||
return 0 if rep.passed else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- build
|
||||
|
||||
|
||||
def dump(schema: str, key: str, items: list[dict[str, Any]]) -> bytes:
|
||||
return (json.dumps({"schema": schema, key: items}, ensure_ascii=False, indent=1) + "\n").encode("utf-8")
|
||||
|
||||
|
||||
def lay_out(draft: list[dict[str, Any]], seed: int, lone: int, dense: int) -> list[list[dict[str, Any]]]:
|
||||
"""Calls of draft pairs, each in position order. With random.Random(seed): each label's pairs sorted by draft id and
|
||||
shuffled; the first `lone` unsupported pairs stand alone in a call, at positions spread evenly over 1-10; the rest
|
||||
stand `dense` to a call; counted supported pairs go one per call in call order; filler pairs fill every call to
|
||||
ten, taken in a cycle, never twice in one call and never in the call of the unsupported pair they are the twin of;
|
||||
the other positions are shuffled."""
|
||||
rng = random.Random(seed)
|
||||
groups = {lab: sorted((p for p in draft if p.get("label") == lab), key=lambda p: str(p["id"])) for lab in LABELS}
|
||||
for lab in LABELS:
|
||||
rng.shuffle(groups[lab])
|
||||
unsupported, supported, fillers = groups[UNSUPPORTED], groups[SUPPORTED], groups[FILLER]
|
||||
if not 0 <= lone <= len(unsupported):
|
||||
raise RunError(f"--lone {lone} is outside 0-{len(unsupported)} (the unsupported pairs in the draft)")
|
||||
if not 1 <= dense <= CALL_SIZE_MAX:
|
||||
raise RunError(f"--dense {dense} is outside 1-{CALL_SIZE_MAX}")
|
||||
rest = unsupported[lone:]
|
||||
calls = [[u] for u in unsupported[:lone]] + [rest[i : i + dense] for i in range(0, len(rest), dense)]
|
||||
pending = list(supported)
|
||||
while pending:
|
||||
free = [c for c in calls if len(c) < CALL_SIZE_MAX]
|
||||
if not free:
|
||||
raise RunError(f"{len(pending)} supported pairs find no free place in the calls")
|
||||
for c in free[: len(pending)]:
|
||||
c.append(pending.pop(0))
|
||||
cursor = 0
|
||||
for c in calls:
|
||||
need = CALL_SIZE_MAX - len(c)
|
||||
if need and not fillers:
|
||||
raise RunError(f"a call has {need} free places and the draft has no {FILLER} pairs to fill them")
|
||||
taken = {str(p["id"]) for p in c}
|
||||
twins_of = {str(p["id"]) for p in c if p.get("label") == UNSUPPORTED}
|
||||
misses = 0
|
||||
while need > 0:
|
||||
f = fillers[cursor % len(fillers)]
|
||||
cursor += 1
|
||||
if str(f["id"]) in taken or f.get("twin_of") in twins_of:
|
||||
misses += 1
|
||||
if misses >= len(fillers):
|
||||
raise RunError(f"too few distinct {FILLER} pairs to fill a call without repeating one or meeting its twin")
|
||||
continue
|
||||
c.append(f)
|
||||
taken.add(str(f["id"]))
|
||||
need -= 1
|
||||
misses = 0
|
||||
at = [n % CALL_SIZE_MAX for n in range(lone)]
|
||||
rng.shuffle(at)
|
||||
ordered: list[list[dict[str, Any]]] = []
|
||||
for n, c in enumerate(calls):
|
||||
if n < lone:
|
||||
head, others = c[0], c[1:]
|
||||
rng.shuffle(others)
|
||||
ordered.append(others[: at[n]] + [head] + others[at[n] :])
|
||||
else:
|
||||
rng.shuffle(c)
|
||||
ordered.append(c)
|
||||
return ordered
|
||||
|
||||
|
||||
def history_of(manifest_path: Path) -> list[Any]:
|
||||
"""The history for a new manifest: the old manifest's history, then the old manifest itself without it."""
|
||||
if not manifest_path.exists():
|
||||
return []
|
||||
old = load_json(manifest_path, "manifest")
|
||||
if not isinstance(old, dict):
|
||||
raise RunError(f"{manifest_path} is not a manifest; it would be lost from the history")
|
||||
return as_list(old.get("history")) + [{k: v for k, v in old.items() if k != "history"}]
|
||||
|
||||
|
||||
def cmd_build(args: argparse.Namespace) -> int:
|
||||
sources = Path(args.sources)
|
||||
set_path, key_path, calls_path = Path(args.set_out), Path(args.key_out), Path(args.calls_out)
|
||||
manifest_path = Path(args.manifest)
|
||||
for out_path in (set_path, key_path, calls_path):
|
||||
if out_path.exists():
|
||||
raise RunError(f"{out_path} already exists; a built set is frozen and never rebuilt")
|
||||
draft = load_list(Path(args.draft), DRAFT_SCHEMA, "pairs", "draft")
|
||||
rep = check_pairs(draft, sources)
|
||||
if not rep.passed:
|
||||
print("\n".join(rep.lines()))
|
||||
raise RunError("the draft fails the check; nothing was built")
|
||||
calls = lay_out(draft, args.seed, args.lone, args.dense)
|
||||
cache = SourceCache(sources)
|
||||
total = sum(len(c) for c in calls)
|
||||
width, call_width = max(3, len(str(total))), max(3, len(str(len(calls))))
|
||||
pairs: list[dict[str, Any]] = []
|
||||
labels: list[dict[str, Any]] = []
|
||||
layout_calls: list[dict[str, Any]] = []
|
||||
for k, members in enumerate(calls, 1):
|
||||
ids = []
|
||||
for p in members:
|
||||
cid = f"c{len(pairs) + 1:0{width}d}"
|
||||
excerpts = []
|
||||
for e in p["excerpts"]:
|
||||
source = cache.text(e["file"]) or ""
|
||||
excerpts.append({"file": e["file"], "line": first_line(source, e["text"]), "text": e["text"]})
|
||||
pairs.append({"id": cid, "claim": p["claim"], "excerpts": excerpts})
|
||||
labels.append({"id": cid, "label": p["label"], "kind": p["kind"], "draft_id": p["id"], "why": p.get("why", "")})
|
||||
ids.append(cid)
|
||||
layout_calls.append({"call": f"k{k:0{call_width}d}", "ids": ids})
|
||||
set_bytes, key_bytes = dump(SET_SCHEMA, "pairs", pairs), dump(KEY_SCHEMA, "labels", labels)
|
||||
layout = {"call_size_max": CALL_SIZE_MAX, "calls": layout_calls, "schema": CALLS_SCHEMA, "set_sha256": sha256_bytes(set_bytes)}
|
||||
calls_bytes = (json.dumps(layout, indent=1, sort_keys=True, ensure_ascii=False) + "\n").encode("utf-8")
|
||||
lay = check_layout(layout, {lab["id"]: lab["label"] for lab in labels}, sha256_bytes(set_bytes))
|
||||
if lay.errors:
|
||||
raise RunError("the layout breaks K3 point 1, nothing was built: " + "; ".join(lay.errors))
|
||||
joined = [dict(p, label=lab["label"], kind=lab["kind"], draft_id=lab["draft_id"]) for p, lab in zip(pairs, labels)]
|
||||
counts = tally(joined)
|
||||
manifest = {
|
||||
"schema": MANIFEST_SCHEMA,
|
||||
"seed": args.seed,
|
||||
"shuffle": LAYOUT_RULE,
|
||||
"set": {"file": args.set_label or set_path.name, "sha256": sha256_bytes(set_bytes), "pairs": len(pairs)},
|
||||
"key": {"file": args.key_label or key_path.name, "sha256": sha256_bytes(key_bytes)},
|
||||
"calls": {
|
||||
"file": args.calls_label or calls_path.name,
|
||||
"sha256": sha256_bytes(calls_bytes),
|
||||
**lay.stats,
|
||||
"built_with": {"lone": args.lone, "dense": args.dense},
|
||||
},
|
||||
"counts": {**counts, "by_position": lay.by_position},
|
||||
"thresholds": {
|
||||
"caught_min": counts[UNSUPPORTED],
|
||||
"accepted_min": math.ceil(ACCEPTED_SHARE * counts[SUPPORTED]),
|
||||
},
|
||||
"sources": {"files": count_files(sources), "tree_sha256": tree_sha256(sources)},
|
||||
"history": history_of(manifest_path),
|
||||
}
|
||||
for path, data in ((set_path, set_bytes), (key_path, key_bytes), (calls_path, calls_bytes)):
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_bytes(data)
|
||||
manifest_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
manifest_path.write_text(json.dumps(manifest, indent=1) + "\n", encoding="utf-8")
|
||||
rep.notes = layout_lines(lay)
|
||||
print("\n".join(rep.lines()))
|
||||
print(f"seed: {args.seed}")
|
||||
print(f"set: {len(pairs)} claims, sha256 {manifest['set']['sha256']}")
|
||||
print(f"key: sha256 {manifest['key']['sha256']}")
|
||||
print(f"layout: {len(layout_calls)} calls, sha256 {manifest['calls']['sha256']}")
|
||||
print(f"manifest: {manifest_path}")
|
||||
return 0
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- stub
|
||||
|
||||
|
||||
def cmd_stub(args: argparse.Namespace) -> int:
|
||||
pairs = load_list(Path(args.set), SET_SCHEMA, "pairs", "set")
|
||||
if args.key:
|
||||
verdict = {str(p["id"]): verdict_for(p["label"]) for p in join_set_and_key(Path(args.set), Path(args.key))}
|
||||
reader = "stub-key"
|
||||
else:
|
||||
verdict = {str(p["id"]): args.verdict for p in pairs}
|
||||
reader = f"stub-all-{args.verdict}"
|
||||
if args.accept_at is not None:
|
||||
if not args.calls:
|
||||
raise RunError("--accept-at needs --calls, the layout that says where each claim stands")
|
||||
if not 1 <= args.accept_at <= CALL_SIZE_MAX:
|
||||
raise RunError(f"--accept-at {args.accept_at} is outside 1-{CALL_SIZE_MAX}")
|
||||
for i, n in positions_of(load_json(Path(args.calls), "layout")).items():
|
||||
if n == args.accept_at and i in verdict:
|
||||
verdict[i] = SUPPORTED
|
||||
reader += f"-accept-at-{args.accept_at}"
|
||||
reader = args.reader or reader
|
||||
out = {"reader": reader, "models": [STUB_MODEL],
|
||||
"verdicts": [{"id": p["id"], "verdict": verdict[str(p["id"])]} for p in pairs]}
|
||||
Path(args.out).write_text(json.dumps(out, indent=1) + "\n", encoding="utf-8")
|
||||
print(f"stub {reader}: {len(pairs)} verdicts written to {args.out}")
|
||||
return 0
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- gate
|
||||
|
||||
|
||||
def read_verdicts(path: Path) -> tuple[str, dict[str, str], str]:
|
||||
"""(reader, verdict per id, fail-closed reason). An id given twice, or with an invalid verdict, is left out."""
|
||||
if not path.is_file():
|
||||
return "", {}, f"no verdicts at {path}"
|
||||
data = load_json(path, "verdicts")
|
||||
if not isinstance(data, dict) or not isinstance(data.get("verdicts"), list):
|
||||
raise RunError(f"{path} is not a verdicts file {VERDICTS_SCHEMA_NOTE}")
|
||||
counts: dict[str, int] = {}
|
||||
given: dict[str, str] = {}
|
||||
for v in data["verdicts"]:
|
||||
if isinstance(v, dict) and isinstance(v.get("id"), str):
|
||||
counts[v["id"]] = counts.get(v["id"], 0) + 1
|
||||
if v.get("verdict") in VERDICTS:
|
||||
given[v["id"]] = v["verdict"]
|
||||
verdicts = {i: val for i, val in given.items() if counts[i] == 1}
|
||||
reader = data.get("reader")
|
||||
return reader if isinstance(reader, str) else "", verdicts, ""
|
||||
|
||||
|
||||
def read_models(path: Path) -> list[str]:
|
||||
"""The sorted, distinct models a verdicts file names in its "models" list; none when the file or the list is
|
||||
missing or holds no model name."""
|
||||
if not path.is_file():
|
||||
return []
|
||||
data = load_json(path, "verdicts")
|
||||
raw = data.get("models") if isinstance(data, dict) else None
|
||||
return sorted({m for m in as_list(raw) if isinstance(m, str) and m})
|
||||
|
||||
|
||||
def record_passed(path: Path, entry: dict[str, str]) -> bool:
|
||||
"""Add one green run to the record at path (created when missing); False when the same entry is there already.
|
||||
A record that cannot be read, or does not have the schema, is a RunError and is left as it was."""
|
||||
entries: list[Any] = []
|
||||
if path.exists():
|
||||
try:
|
||||
entries = reader.read_passed(path) # the parser collect --gate-passed reads the record with
|
||||
except reader.RunError as exc:
|
||||
raise RunError(str(exc)) from exc
|
||||
if entry in entries:
|
||||
return False
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_text(json.dumps({"schema": PASSED_SCHEMA, "passed": [*entries, entry]}, indent=1) + "\n", encoding="utf-8")
|
||||
return True
|
||||
|
||||
|
||||
def cmd_gate(args: argparse.Namespace) -> int:
|
||||
frozen = load_frozen(Path(args.set), Path(args.key), Path(args.calls), Path(args.manifest), Path(args.sources))
|
||||
if frozen.errors:
|
||||
raise RunError("the frozen set does not hold: " + "; ".join(frozen.errors))
|
||||
thresholds = as_dict(frozen.manifest.get("thresholds"))
|
||||
caught_min, accepted_min = thresholds.get("caught_min"), thresholds.get("accepted_min")
|
||||
if not isinstance(caught_min, int) or not isinstance(accepted_min, int):
|
||||
raise RunError("the manifest has no integer thresholds caught_min and accepted_min")
|
||||
pairs, position = frozen.pairs, frozen.layout.position
|
||||
reader, verdicts, fail_closed = read_verdicts(Path(args.verdicts))
|
||||
models = read_models(Path(args.verdicts))
|
||||
model_reason = "" if len(models) == 1 else (
|
||||
"the verdicts name no model" if not models else f"the verdicts name {len(models)} models")
|
||||
n = {lab: sum(1 for p in pairs if p["label"] == lab) for lab in LABELS}
|
||||
caught = sum(1 for p in pairs if p["label"] == UNSUPPORTED and verdicts.get(p["id"]) == UNSUPPORTED)
|
||||
accepted = sum(1 for p in pairs if p["label"] == SUPPORTED and verdicts.get(p["id"]) == SUPPORTED)
|
||||
filler_accepted = sum(1 for p in pairs if p["label"] == FILLER and verdicts.get(p["id"]) == SUPPORTED)
|
||||
missing = sum(1 for p in pairs if p["id"] not in verdicts)
|
||||
unknown = sum(1 for i in verdicts if i not in position)
|
||||
per_kind = []
|
||||
for k in UNSUPPORTED_KINDS:
|
||||
of_kind = [p for p in pairs if p["label"] == UNSUPPORTED and p["kind"] == k]
|
||||
hit = sum(1 for p in of_kind if verdicts.get(p["id"]) == UNSUPPORTED)
|
||||
per_kind.append(f"{k} {hit} of {len(of_kind)}")
|
||||
per_position, short = [], []
|
||||
for at in range(1, CALL_SIZE_MAX + 1):
|
||||
there = [p for p in pairs if p["label"] == UNSUPPORTED and position.get(p["id"]) == at]
|
||||
hit = sum(1 for p in there if verdicts.get(p["id"]) == UNSUPPORTED)
|
||||
per_position.append(f"{at} {hit} of {len(there)}")
|
||||
if hit < len(there) or not there:
|
||||
short.append(f"position {at}: caught {hit} of {len(there)}")
|
||||
green = (
|
||||
not fail_closed
|
||||
and not model_reason
|
||||
and missing == 0
|
||||
and caught >= caught_min
|
||||
and caught == n[UNSUPPORTED]
|
||||
and not short
|
||||
and accepted >= accepted_min
|
||||
)
|
||||
print(
|
||||
f"set: {len(pairs)} claims ({n[UNSUPPORTED]} {UNSUPPORTED}, {n[SUPPORTED]} {SUPPORTED}, {n[FILLER]} {FILLER})"
|
||||
f" in {frozen.layout.stats['calls']} calls; reader: {reader or '(none)'}"
|
||||
)
|
||||
print(f"model: {', '.join(models) if models else '(none)'}")
|
||||
if fail_closed:
|
||||
print(f"fail-closed: {fail_closed}")
|
||||
if model_reason:
|
||||
print(f"fail-closed: {model_reason}")
|
||||
print(f"missing verdicts: {missing}" + (f"; verdicts for unknown ids: {unknown}" if unknown else ""))
|
||||
print(f"unsupported caught {caught} of {n[UNSUPPORTED]} · supported accepted {accepted} of {n[SUPPORTED]}")
|
||||
print("caught by kind: " + ", ".join(per_kind))
|
||||
print("caught by position: " + ", ".join(per_position))
|
||||
print(f"filler accepted {filler_accepted} of {n[FILLER]} (reported apart, never counted in the thresholds)")
|
||||
print(
|
||||
f"required: caught >= {caught_min} of {n[UNSUPPORTED]} with every position 1-{CALL_SIZE_MAX} caught in full,"
|
||||
f" accepted >= {accepted_min} of {n[SUPPORTED]}, no verdict missing"
|
||||
)
|
||||
if not green:
|
||||
reasons = []
|
||||
if fail_closed:
|
||||
reasons.append("fail-closed")
|
||||
if model_reason:
|
||||
reasons.append(model_reason)
|
||||
if missing:
|
||||
reasons.append(f"{missing} verdicts missing")
|
||||
if caught < caught_min or caught < n[UNSUPPORTED]:
|
||||
reasons.append(f"caught {caught} of {n[UNSUPPORTED]}")
|
||||
reasons += short
|
||||
if accepted < accepted_min:
|
||||
reasons.append(f"accepted {accepted} of {n[SUPPORTED]}, needs {accepted_min}")
|
||||
print("not met: " + "; ".join(reasons))
|
||||
if green and args.passed:
|
||||
entry = {"model": models[0], "set_sha256": file_sha256(Path(args.set)), "key_sha256": file_sha256(Path(args.key)),
|
||||
"verdicts_sha256": file_sha256(Path(args.verdicts)), "reader": reader}
|
||||
added = record_passed(Path(args.passed), entry)
|
||||
print(f"passed: {models[0]} {'recorded in' if added else 'already recorded in'} {args.passed}")
|
||||
print("GREEN" if green else "RED")
|
||||
return 0 if green else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- main
|
||||
|
||||
|
||||
def build_parser() -> argparse.ArgumentParser:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
sub = p.add_subparsers(dest="command", required=True)
|
||||
c = sub.add_parser("check", help="check a labelled draft, or a built set with its key, layout and manifest")
|
||||
c.add_argument("--draft", default="")
|
||||
c.add_argument("--set", default="")
|
||||
c.add_argument("--key", default="")
|
||||
c.add_argument("--calls", default="", help="the call layout (K3 point 1)")
|
||||
c.add_argument("--manifest", default="")
|
||||
c.add_argument("--sources", required=True)
|
||||
b = sub.add_parser("build", help="lay out a checked draft in calls with a fixed seed: set, key, layout and manifest")
|
||||
b.add_argument("--draft", required=True)
|
||||
b.add_argument("--sources", required=True)
|
||||
b.add_argument("--seed", type=int, required=True)
|
||||
b.add_argument("--lone", type=int, required=True, help="unsupported pairs that stand alone in their call")
|
||||
b.add_argument("--dense", type=int, required=True, help="unsupported pairs per call for the rest")
|
||||
b.add_argument("--set-out", required=True)
|
||||
b.add_argument("--key-out", required=True)
|
||||
b.add_argument("--calls-out", required=True)
|
||||
b.add_argument("--set-label", default="", help="path of the set as written in the manifest")
|
||||
b.add_argument("--key-label", default="", help="path of the key as written in the manifest")
|
||||
b.add_argument("--calls-label", default="", help="path of the layout as written in the manifest")
|
||||
b.add_argument("--manifest", required=True)
|
||||
s = sub.add_parser("stub", help="write verdicts without a reader")
|
||||
s.add_argument("--set", required=True)
|
||||
which = s.add_mutually_exclusive_group(required=True)
|
||||
which.add_argument("--verdict", choices=VERDICTS, help="the same verdict for every claim")
|
||||
which.add_argument("--key", help="the key's verdict for every claim (the known-positive)")
|
||||
s.add_argument("--calls", default="", help="the call layout, for --accept-at")
|
||||
s.add_argument("--accept-at", type=int, default=None, help="answer supported for every claim at this position")
|
||||
s.add_argument("--reader", default="")
|
||||
s.add_argument("--out", required=True)
|
||||
g = sub.add_parser("gate", help="score one reader's verdicts against the key")
|
||||
g.add_argument("--set", required=True)
|
||||
g.add_argument("--key", required=True)
|
||||
g.add_argument("--calls", required=True, help="the call layout (K3 point 1)")
|
||||
g.add_argument("--manifest", required=True)
|
||||
g.add_argument("--sources", required=True)
|
||||
g.add_argument("--verdicts", required=True, help=VERDICTS_SCHEMA_NOTE)
|
||||
g.add_argument("--passed", default="", metavar="FILE",
|
||||
help=f"on GREEN, add the run to this record of green runs ({PASSED_SCHEMA}); RED never writes it")
|
||||
return p
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args = build_parser().parse_args(argv)
|
||||
commands = {"check": cmd_check, "build": cmd_build, "stub": cmd_stub, "gate": cmd_gate}
|
||||
try:
|
||||
return commands[args.command](args)
|
||||
except RunError as exc:
|
||||
print(f"FAIL: {exc}", file=sys.stderr)
|
||||
return 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
302
gates/filetype_gate.py
Normal file
302
gates/filetype_gate.py
Normal file
|
|
@ -0,0 +1,302 @@
|
|||
"""File-type gate for the reader (E4): does a reader find each frozen quote at its one true place, per file type?
|
||||
|
||||
No model call happens here. The gate is fail-closed: a reader that is missing, crashes, times out or answers in
|
||||
the wrong shape has located nothing.
|
||||
|
||||
Reader contract (the L1b reader, or any stub standing in for it):
|
||||
- argv is the command after `--`, with `{file}` replaced by a fresh copy of one source file (its name and
|
||||
extension kept), `{queries}` by a JSON file {"file": "<name>", "queries": [{"id": "...", "quote": "..."}]},
|
||||
and `{sources}` by the --sources root. The reader never sees the expected locator.
|
||||
- stdout is one JSON object {"answers": [{"id": "...", "found": [{"locator": {...}, "hidden": false}]}]}:
|
||||
every place in the file where the quote occurs, compared with runs of whitespace collapsed to one space.
|
||||
- exit 0. Anything else is an error for every pair on that file.
|
||||
|
||||
A pair is *located* when its answer names exactly one place, that place is the expected locator, and the
|
||||
`hidden` flag equals the expected one (hidden Office content must be marked, not flattened into ordinary text).
|
||||
Every quote was chosen to occur exactly once in its file, so a second place is a wrong answer.
|
||||
|
||||
Locators (see the manifest for why each was chosen): pdf {"page": N} · docx {"paragraph": N} ·
|
||||
xlsx {"sheet": name, "cell": "B7"} · pptx {"slide": N, "notes": bool} · md, txt, html {"line": N}.
|
||||
Numbers are 1-based integers (a boolean is not a number). Cell references ignore case; `notes` defaults to false.
|
||||
|
||||
Before anything runs, the pairs file and every source file a pair names are checked against the manifest's
|
||||
sha256; a mismatch, a missing file or a file the manifest does not list stops the gate (exit 2).
|
||||
|
||||
Usage:
|
||||
python3 gates/filetype_gate.py --manifest gates/filetypes.manifest.json --pairs gates-local/filetypes/pairs.json \\
|
||||
--sources <root of blandet/ and wiki/> --cases gates-local/filetypes/files [--timeout 120] \\
|
||||
-- <reader> {file} {queries}
|
||||
Exit: 0 GREEN, 1 RED, 2 the gate could not run.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import shutil
|
||||
import subprocess
|
||||
import sys
|
||||
import tempfile
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
PAIRS_SCHEMA = "okf-wiki/filetype-pairs/1"
|
||||
MANIFEST_SCHEMA = "okf-wiki/filetype-manifest/1"
|
||||
TYPES = ("pdf", "docx", "xlsx", "pptx", "md", "txt", "html")
|
||||
ROOTS = ("sources", "cases")
|
||||
GROUPS = ("filetype", "hidden", "hidden-control")
|
||||
LOCATOR_KEYS: dict[str, frozenset[str]] = {
|
||||
"pdf": frozenset({"page"}),
|
||||
"docx": frozenset({"paragraph"}),
|
||||
"xlsx": frozenset({"sheet", "cell"}),
|
||||
"pptx": frozenset({"slide", "notes"}),
|
||||
"md": frozenset({"line"}),
|
||||
"txt": frozenset({"line"}),
|
||||
"html": frozenset({"line"}),
|
||||
}
|
||||
|
||||
|
||||
class GateError(Exception):
|
||||
"""The gate cannot produce a measurement (bad manifest, drifted fixtures, bad arguments)."""
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Pair:
|
||||
id: str
|
||||
type: str
|
||||
root: str
|
||||
path: str
|
||||
group: str
|
||||
hidden_class: str | None
|
||||
quote: str
|
||||
locator: dict[str, Any]
|
||||
hidden: bool
|
||||
|
||||
@property
|
||||
def source(self) -> tuple[str, str]:
|
||||
return (self.root, self.path)
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Verdict:
|
||||
located: bool
|
||||
reason: str = ""
|
||||
|
||||
|
||||
def sha256_file(path: Path) -> str:
|
||||
return hashlib.sha256(path.read_bytes()).hexdigest()
|
||||
|
||||
|
||||
def _is_int(v: object) -> bool:
|
||||
return isinstance(v, int) and not isinstance(v, bool)
|
||||
|
||||
|
||||
def canonical_locator(ftype: str, loc: object) -> dict[str, Any] | None:
|
||||
"""The locator in comparable form, or None when it does not have the type's shape."""
|
||||
if not isinstance(loc, dict):
|
||||
return None
|
||||
out = dict(loc)
|
||||
if ftype == "pptx":
|
||||
out.setdefault("notes", False)
|
||||
if set(out) != LOCATOR_KEYS[ftype]:
|
||||
return None
|
||||
if ftype == "xlsx":
|
||||
if not isinstance(out["sheet"], str) or not isinstance(out["cell"], str):
|
||||
return None
|
||||
out["cell"] = out["cell"].upper()
|
||||
return out
|
||||
for key in LOCATOR_KEYS[ftype] - {"notes"}:
|
||||
if not _is_int(out[key]):
|
||||
return None
|
||||
if ftype == "pptx" and not isinstance(out["notes"], bool):
|
||||
return None
|
||||
return out
|
||||
|
||||
|
||||
def same_locator(ftype: str, expected: object, got: object) -> bool:
|
||||
a, b = canonical_locator(ftype, expected), canonical_locator(ftype, got)
|
||||
return a is not None and a == b
|
||||
|
||||
|
||||
def load_json(path: Path, what: str) -> Any:
|
||||
try:
|
||||
return json.loads(path.read_text(encoding="utf-8"))
|
||||
except (OSError, json.JSONDecodeError) as exc:
|
||||
raise GateError(f"cannot read {what} {path}: {exc}") from exc
|
||||
|
||||
|
||||
def load_pairs(path: Path) -> tuple[str, list[Pair]]:
|
||||
data = load_json(path, "pairs file")
|
||||
if not isinstance(data, dict) or data.get("schema") != PAIRS_SCHEMA:
|
||||
raise GateError(f"pairs file schema must be {PAIRS_SCHEMA!r}")
|
||||
pairs: list[Pair] = []
|
||||
for raw in data.get("pairs", []):
|
||||
try:
|
||||
p = Pair(id=str(raw["id"]), type=str(raw["type"]), root=str(raw["root"]), path=str(raw["path"]),
|
||||
group=str(raw["group"]), hidden_class=raw.get("hidden_class"), quote=str(raw["quote"]),
|
||||
locator=raw["locator"], hidden=raw["hidden"])
|
||||
except (KeyError, TypeError) as exc:
|
||||
raise GateError(f"pair {raw!r} is malformed: {exc}") from exc
|
||||
if p.type not in TYPES or p.root not in ROOTS or p.group not in GROUPS or not isinstance(p.hidden, bool):
|
||||
raise GateError(f"pair {p.id}: bad type, root, group or hidden flag")
|
||||
if canonical_locator(p.type, p.locator) is None:
|
||||
raise GateError(f"pair {p.id}: locator {p.locator!r} does not have the {p.type} shape")
|
||||
if p.group != "filetype" and not p.hidden_class:
|
||||
raise GateError(f"pair {p.id}: a hidden case needs a hidden_class")
|
||||
if not p.quote.strip():
|
||||
raise GateError(f"pair {p.id}: empty quote")
|
||||
pairs.append(p)
|
||||
ids = [p.id for p in pairs]
|
||||
if not pairs or len(ids) != len(set(ids)):
|
||||
raise GateError("pairs file has no pairs or repeats an id")
|
||||
return str(data.get("set", "")), pairs
|
||||
|
||||
|
||||
def verify_fixtures(manifest_path: Path, pairs_path: Path, pairs: list[Pair], roots: dict[str, Path]) -> None:
|
||||
manifest = load_json(manifest_path, "manifest")
|
||||
if not isinstance(manifest, dict) or manifest.get("schema") != MANIFEST_SCHEMA:
|
||||
raise GateError(f"manifest schema must be {MANIFEST_SCHEMA!r}")
|
||||
problems: list[str] = []
|
||||
pairs_sha = sha256_file(pairs_path)
|
||||
known_sets = {str(s.get("sha256")) for s in manifest.get("sets", {}).values()}
|
||||
if pairs_sha not in known_sets:
|
||||
problems.append(f"pairs file {pairs_path} matches no set in the manifest (sha256 {pairs_sha[:12]}...)")
|
||||
listed = {(str(s["root"]), str(s["path"])): str(s["sha256"]) for s in manifest.get("sources", [])}
|
||||
for root, rel in sorted({p.source for p in pairs}):
|
||||
if (root, rel) not in listed:
|
||||
problems.append(f"{root}:{rel} is not in the manifest")
|
||||
continue
|
||||
f = roots[root] / rel
|
||||
if not f.is_file():
|
||||
problems.append(f"source missing: {f}")
|
||||
elif sha256_file(f) != listed[(root, rel)]:
|
||||
problems.append(f"source sha256 mismatch: {f}")
|
||||
if problems:
|
||||
raise GateError("fixtures do not match the manifest:\n " + "\n ".join(problems))
|
||||
|
||||
|
||||
def run_reader(template: list[str], pairs: list[Pair], source: Path, sources_root: Path,
|
||||
timeout: float) -> tuple[dict[str, list[Any]] | None, str]:
|
||||
"""Answers per pair id for one file, or (None, reason) when the run is an error."""
|
||||
with tempfile.TemporaryDirectory(prefix="filetype-gate-") as tmp:
|
||||
copy = Path(tmp) / "file" / source.name
|
||||
copy.parent.mkdir()
|
||||
shutil.copyfile(source, copy)
|
||||
queries = Path(tmp) / "queries.json"
|
||||
queries.write_text(json.dumps({"file": source.name,
|
||||
"queries": [{"id": p.id, "quote": p.quote} for p in pairs]}), encoding="utf-8")
|
||||
argv = [a.replace("{file}", str(copy)).replace("{queries}", str(queries)).replace("{sources}", str(sources_root))
|
||||
for a in template]
|
||||
try:
|
||||
proc = subprocess.run(argv, capture_output=True, text=True, timeout=timeout, check=False)
|
||||
except FileNotFoundError:
|
||||
return None, f"reader not found: {argv[0]}"
|
||||
except PermissionError:
|
||||
return None, f"reader not executable: {argv[0]}"
|
||||
except subprocess.TimeoutExpired:
|
||||
return None, f"reader timed out after {timeout:g}s"
|
||||
if proc.returncode != 0:
|
||||
tail = proc.stderr.strip().splitlines()[-1:] or [""]
|
||||
return None, f"reader exited {proc.returncode} {tail[0]}".rstrip()
|
||||
try:
|
||||
payload = json.loads(proc.stdout)
|
||||
except json.JSONDecodeError:
|
||||
return None, "reader output is not valid JSON"
|
||||
answers = payload.get("answers") if isinstance(payload, dict) else None
|
||||
if not isinstance(answers, list) or not all(isinstance(a, dict) and isinstance(a.get("id"), str)
|
||||
and isinstance(a.get("found"), list) for a in answers):
|
||||
return None, 'reader output lacks an "answers" list of {"id", "found": [...]} objects'
|
||||
return {str(a["id"]): list(a["found"]) for a in answers}, ""
|
||||
|
||||
|
||||
def judge(pair: Pair, found: list[Any] | None) -> Verdict:
|
||||
if found is None:
|
||||
return Verdict(False, "no answer")
|
||||
if not found:
|
||||
return Verdict(False, "not found")
|
||||
if len(found) > 1:
|
||||
return Verdict(False, f"{len(found)} places claimed")
|
||||
place = found[0]
|
||||
if not isinstance(place, dict) or not same_locator(pair.type, pair.locator, place.get("locator")):
|
||||
got = place.get("locator") if isinstance(place, dict) else place
|
||||
return Verdict(False, f"wrong place {json.dumps(got)}")
|
||||
if place.get("hidden") is not pair.hidden:
|
||||
return Verdict(False, f"hidden flag wrong (expected {str(pair.hidden).lower()})")
|
||||
return Verdict(True)
|
||||
|
||||
|
||||
def report(set_name: str, pairs: list[Pair], verdicts: dict[str, Verdict], errors: dict[tuple[str, str], str],
|
||||
files: int, command: list[str]) -> bool:
|
||||
print(f"filetype gate: {len(pairs)} pairs in {files} files (set {set_name}); reader: {' '.join(command)}")
|
||||
for (root, rel), reason in sorted(errors.items()):
|
||||
print(f" ERROR {root}:{rel}: {reason}")
|
||||
for p in pairs:
|
||||
v = verdicts[p.id]
|
||||
if not v.located:
|
||||
print(f" {p.id:<8} {p.type:<5} {json.dumps(p.locator):<40} NOT LOCATED ({v.reason})")
|
||||
|
||||
def line(label: str, group: list[Pair]) -> str:
|
||||
return f"{label}: {sum(verdicts[p.id].located for p in group)} of {len(group)}"
|
||||
|
||||
ordinary = [p for p in pairs if p.group == "filetype"]
|
||||
present = [t for t in TYPES if any(p.type == t for p in ordinary)]
|
||||
print("per file type: " + " · ".join(line(t, [p for p in ordinary if p.type == t]) for t in present))
|
||||
full = [t for t in present if all(verdicts[p.id].located for p in ordinary if p.type == t)]
|
||||
print(f"file types with every pair located: {len(full)} of {len(present)}")
|
||||
hidden = [p for p in pairs if p.group == "hidden"]
|
||||
if hidden:
|
||||
classes = sorted({str(p.hidden_class) for p in hidden})
|
||||
print("hidden, per class: " + " · ".join(line(c, [p for p in hidden if p.hidden_class == c]) for c in classes))
|
||||
controls = [p for p in pairs if p.group == "hidden-control"]
|
||||
if controls:
|
||||
print(line("visible controls in the hidden cases", controls))
|
||||
located = sum(v.located for v in verdicts.values())
|
||||
print(f"located {located} of {len(pairs)} · reader errors {len(errors)} of {files} files")
|
||||
green = located == len(pairs) and not errors
|
||||
print(f"verdict: {'GREEN' if green else 'RED'}")
|
||||
return green
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args_in = sys.argv[1:] if argv is None else argv
|
||||
if "--" in args_in:
|
||||
split = args_in.index("--")
|
||||
own, command = args_in[:split], args_in[split + 1:]
|
||||
else:
|
||||
own, command = args_in, []
|
||||
parser = argparse.ArgumentParser(description="Run the reader's file-type gate (fail-closed, no model calls).")
|
||||
parser.add_argument("--manifest", type=Path, required=True)
|
||||
parser.add_argument("--pairs", type=Path, required=True, help="the working set or the held-out set")
|
||||
parser.add_argument("--sources", type=Path, required=True, help="root for pairs with root 'sources'")
|
||||
parser.add_argument("--cases", type=Path, required=True, help="root for pairs with root 'cases' (gates-local)")
|
||||
parser.add_argument("--timeout", type=float, default=120.0, help="seconds per reader run (one run per file)")
|
||||
args = parser.parse_args(own)
|
||||
if not command:
|
||||
print("usage error: give the reader command after --, e.g. -- python3 reader.py {file} {queries}")
|
||||
return 2
|
||||
roots = {"sources": args.sources, "cases": args.cases}
|
||||
try:
|
||||
set_name, pairs = load_pairs(args.pairs)
|
||||
verify_fixtures(args.manifest, args.pairs, pairs, roots)
|
||||
except GateError as exc:
|
||||
print(f"GATE CANNOT RUN: {exc}")
|
||||
return 2
|
||||
by_file: dict[tuple[str, str], list[Pair]] = {}
|
||||
for p in pairs:
|
||||
by_file.setdefault(p.source, []).append(p)
|
||||
verdicts: dict[str, Verdict] = {}
|
||||
errors: dict[tuple[str, str], str] = {}
|
||||
for (root, rel), group in sorted(by_file.items()):
|
||||
answers, reason = run_reader(command, group, roots[root] / rel, args.sources, args.timeout)
|
||||
if answers is None:
|
||||
errors[(root, rel)] = reason
|
||||
verdicts.update({p.id: Verdict(False, "reader error") for p in group})
|
||||
continue
|
||||
verdicts.update({p.id: judge(p, answers.get(p.id)) for p in group})
|
||||
return 0 if report(set_name, pairs, verdicts, errors, len(by_file), command) else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
683
gates/filetypes.manifest.json
Normal file
683
gates/filetypes.manifest.json
Normal file
|
|
@ -0,0 +1,683 @@
|
|||
{
|
||||
"schema": "okf-wiki/filetype-manifest/1",
|
||||
"description": "E4 file-type gate for L1b: frozen (quote, locator) pairs per file type plus hidden Office content. Pairs live in gates-local/ and outside the repository; only this manifest is tracked.",
|
||||
"seed": 20260927,
|
||||
"draw": "random.Random(seed): pdf/docx/xlsx/pptx 10 files x 2 pairs at distinct locators, files in seeded order from the 22 per type, a file giving fewer than 2 is skipped; md 10 files x 2 from the 599; txt and html 6 files, 4+4+3+3+3+3. Per type 20 pairs, shuffled, the first 4 (20 %) held out. Hidden cases only in the working set. Both sets shuffled and renumbered p../h..",
|
||||
"sets": {
|
||||
"working": {
|
||||
"file": "gates-local/filetypes/pairs.json",
|
||||
"sha256": "08815bfeda03b2eebbcaf1bb2c00e58710c5c45a16da6ae8c547c0762c46dc5e",
|
||||
"pairs": 128
|
||||
},
|
||||
"held_out": {
|
||||
"sha256": "0c6ea4149941f25c443d7db9bab57d5c99ad5ec0b8cbfaa410b6ff04e4caf30b",
|
||||
"pairs": 28,
|
||||
"note": "outside the repository, same form as E2; no build session reads it (R54)"
|
||||
}
|
||||
},
|
||||
"counts": {
|
||||
"per_type": {
|
||||
"pdf": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 10,
|
||||
"files_total": 10
|
||||
},
|
||||
"docx": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 9,
|
||||
"files_total": 10
|
||||
},
|
||||
"xlsx": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 9,
|
||||
"files_total": 10
|
||||
},
|
||||
"pptx": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 10,
|
||||
"files_total": 10,
|
||||
"notes_working": 3
|
||||
},
|
||||
"md": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 10,
|
||||
"files_total": 10
|
||||
},
|
||||
"txt": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 6,
|
||||
"files_total": 6
|
||||
},
|
||||
"html": {
|
||||
"working": 16,
|
||||
"held_out": 4,
|
||||
"files_working": 6,
|
||||
"files_total": 6
|
||||
}
|
||||
},
|
||||
"hidden_per_class": {
|
||||
"docx-hidden-text": {
|
||||
"pairs": 2,
|
||||
"files": [
|
||||
"hidden/hidden-text-and-deletions.docx"
|
||||
],
|
||||
"origin": "constructed"
|
||||
},
|
||||
"docx-tracked-deletion": {
|
||||
"pairs": 2,
|
||||
"files": [
|
||||
"hidden/hidden-text-and-deletions.docx"
|
||||
],
|
||||
"origin": "constructed"
|
||||
},
|
||||
"pptx-hidden-slide": {
|
||||
"pairs": 2,
|
||||
"files": [
|
||||
"hidden/hidden-slide.pptx"
|
||||
],
|
||||
"origin": "constructed"
|
||||
},
|
||||
"xlsx-hidden-columns": {
|
||||
"pairs": 2,
|
||||
"files": [
|
||||
"blandet/kilde/xlsx/nist-sp800-201_fra_data_sample.xlsx"
|
||||
],
|
||||
"origin": "natural"
|
||||
},
|
||||
"xlsx-hidden-rows": {
|
||||
"pairs": 2,
|
||||
"files": [
|
||||
"hidden/hidden-sheet-and-rows.xlsx"
|
||||
],
|
||||
"origin": "constructed"
|
||||
},
|
||||
"xlsx-hidden-sheet": {
|
||||
"pairs": 2,
|
||||
"files": [
|
||||
"hidden/hidden-sheet-and-rows.xlsx"
|
||||
],
|
||||
"origin": "constructed"
|
||||
}
|
||||
},
|
||||
"hidden_controls": 4
|
||||
},
|
||||
"gate": {
|
||||
"green_when": "every pair located (exactly one place, the frozen locator, the right hidden flag) and 0 reader errors",
|
||||
"contract": "reader {file} {queries} -> {\"answers\": [{\"id\", \"found\": [{\"locator\", \"hidden\"}]}]}; whitespace runs collapse to one space when matching; see gates/filetype_gate.py"
|
||||
},
|
||||
"locators": {
|
||||
"pdf": {
|
||||
"shape": {
|
||||
"page": "N"
|
||||
},
|
||||
"why": "physical page, 1-based, as a viewer opens it (#page=N); printed page labels differ from it and are missing in many files"
|
||||
},
|
||||
"docx": {
|
||||
"shape": {
|
||||
"paragraph": "N"
|
||||
},
|
||||
"definition": "1-based position among w:p elements of word/document.xml's w:body in document order, counting paragraphs in table cells and content controls, not counting paragraphs inside a text box (w:txbxContent) or a compatibility fallback (mc:Fallback); empty paragraphs count",
|
||||
"why": "chosen over a heading path because only 6 of the 22 docx have any paragraph styled as a heading (measured: a w:pStyle value beginning 'Heading' or 'heading'), so a heading path is undefined for 16; a paragraph index exists in every file and is unique. Table paragraphs are counted so table text has a locator, but no quote is drawn from a table (tables are not in v1)"
|
||||
},
|
||||
"xlsx": {
|
||||
"shape": {
|
||||
"sheet": "name",
|
||||
"cell": "B7"
|
||||
},
|
||||
"why": "sheet name + A1 reference is what Excel shows and what a reader opens; string cells only (shared, inline, formula strings), since number formatting makes the displayed text of a number cell ambiguous"
|
||||
},
|
||||
"pptx": {
|
||||
"shape": {
|
||||
"slide": "N",
|
||||
"notes": "bool"
|
||||
},
|
||||
"why": "1-based position in the presentation's slide list (hidden slides count); notes true when the text is in the slide's notes page"
|
||||
},
|
||||
"md": {
|
||||
"shape": {
|
||||
"line": "N"
|
||||
},
|
||||
"why": "the line of the raw file, as L1a's auditor already checks"
|
||||
},
|
||||
"txt": {
|
||||
"shape": {
|
||||
"line": "N"
|
||||
},
|
||||
"why": "the line of the raw file"
|
||||
},
|
||||
"html": {
|
||||
"shape": {
|
||||
"line": "N"
|
||||
},
|
||||
"why": "the line of the raw source (view-source opens it), where the quote lies inside one text node; chosen over a DOM path because parsers repair markup differently, so a path is not parser-independent, and because it is the same shape as md and txt. Text in head, title, script, style, noscript, template and svg is not visible and gives no quotes"
|
||||
}
|
||||
},
|
||||
"fasit_method": {
|
||||
"principle": "not the method L1b will use (a library reader): package XML read directly, or two independent PDF extractors, then a second method must agree",
|
||||
"pdf": "poppler pdftotext (whole file, pages split at form feed) and pypdf extract_text per page; the quote is verbatim on one pdftotext line of page N, in pypdf's page N, and on no other page in either (so the page count of both must agree)",
|
||||
"docx": "zipfile + ElementTree on word/*.xml; second method python-docx Paragraph.text (font.hidden for hidden text, w:delText for deletions)",
|
||||
"xlsx": "zipfile + ElementTree on workbook, rels, sharedStrings and worksheet XML (sheet state, row and col hidden); second method openpyxl cell value, sheet_state, row and column dimensions",
|
||||
"pptx": "zipfile + ElementTree on presentation, slide and notes XML (sldIdLst order, show=0); second method python-pptx shape and notes paragraphs",
|
||||
"md": "raw line; second method a fresh read of the line; frontmatter, fenced code, headings, tables and HTML lines give no quotes",
|
||||
"txt": "raw line; second method a fresh read of the line",
|
||||
"html": "stdlib HTMLParser line of each text node; second method tag-stripped, unescaped raw source line, and lxml text_content for uniqueness",
|
||||
"quote_rule": "6-10 consecutive words, or a whole cell or paragraph of 3-12 words and at least 15 characters; only [A-Za-z0-9 ,.;:()/%'-]; verbatim with single spaces in the raw unit text; occurs exactly once in the file, counted over every text span in the package (all parts, all pages of both PDF extractors)",
|
||||
"builder": {
|
||||
"file": "gates-local/filetypes/build_e4.py",
|
||||
"sha256": "ac8c515a39b19e53254fb35f20370b26a06a27343e3140e0c73ff9168311a759"
|
||||
},
|
||||
"verifier": {
|
||||
"file": "gates-local/filetypes/verify_e4.py",
|
||||
"sha256": "54dc8c0ba76de835413a56d93a84960f2f9205d8899bec1d337106b6d56f9d81",
|
||||
"known_negative": "14 of 14 pairs with a moved locator or a flipped hidden flag fail it"
|
||||
}
|
||||
},
|
||||
"measured": {
|
||||
"working": {
|
||||
"verbatim": 128,
|
||||
"unique": 128,
|
||||
"second": 128,
|
||||
"hidden": 128,
|
||||
"of": 128
|
||||
},
|
||||
"held_out": {
|
||||
"verbatim": 28,
|
||||
"unique": 28,
|
||||
"second": 28,
|
||||
"hidden": 28,
|
||||
"of": 28
|
||||
},
|
||||
"source_sha256_agrees_with_collection_manifests": "51 of 51"
|
||||
},
|
||||
"sources": [
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "hidden/hidden-sheet-and-rows.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "2d0975f5ecdb6426c64532194000f503f4ebecd526ec7159ef60c1b99e4b6d66",
|
||||
"origin": "constructed",
|
||||
"base": "blandet/kilde/xlsx/usgs-myb1-2022-cobal-ert.xlsx",
|
||||
"base_sha256": "c2df472e47050d85ef03bba211ed1a75633ba5ba997e1e9ce48230dca749f9df",
|
||||
"edit": "workbook.xml: sheet 3 ('T2') state=hidden; xl/worksheets/sheet2.xml: rows [18, 30] on sheet 2 ('T1') hidden=1"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "hidden/hidden-slide.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "bcb7c5ef6b883e968cd0e3397849ca12f49142874dd2aa13d247e802c033ef3e",
|
||||
"origin": "constructed",
|
||||
"base": "blandet/kilde/pptx/ntrs-20240015731.pptx",
|
||||
"base_sha256": "60b9f510bdaa71349d946d0918f33c4bee92f2e5109c74c973333c35c0defa9e",
|
||||
"edit": "ppt/slides/slide4.xml, ppt/slides/slide5.xml (slides [4, 5]): show=0"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "hidden/hidden-text-and-deletions.docx",
|
||||
"type": "docx",
|
||||
"sha256": "f25b067d32e85581cf204c100a1cc0fb7b7353a9e5ced237c12e3322ac5871eb",
|
||||
"origin": "constructed",
|
||||
"base": "blandet/kilde/docx/ntrs-20230002032.docx",
|
||||
"base_sha256": "cae4a29fadbe0065cb80251c837a0e562ff49f52b5894366576637fcbd2db97a",
|
||||
"edit": "paragraphs [1, 7]: w:vanish on every run; paragraphs [5, 9]: every run wrapped in w:del with w:t renamed w:delText"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "html/rfc9111.html",
|
||||
"type": "html",
|
||||
"sha256": "ce91ee9848d2b9ac46386b0f0cbd4bfd9c0cd1948ebc363834cadc7e0997f3d7",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9111.html"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "html/rfc9112.html",
|
||||
"type": "html",
|
||||
"sha256": "587ec383523df5a9713f2f4192199876af1d6d4a260ce996e0a1f9e4f19dde42",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9112.html"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "html/rfc9204.html",
|
||||
"type": "html",
|
||||
"sha256": "06f0e084ebc9b1a54983a133985f1c3341c39faa89e00c6e49285ca47ae2b544",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9204.html"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "html/rfc9293.html",
|
||||
"type": "html",
|
||||
"sha256": "0be501b21b790ae4fa1db0cfe3d3ee935a94f69dc8e0ef7e32a04250eaa30f13",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9293.html"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "html/rfc9562.html",
|
||||
"type": "html",
|
||||
"sha256": "e2d10634480471b56fe6e032091c5986eeb95b456475c714fbfe520efa97b656",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9562.html"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "html/rfc9651.html",
|
||||
"type": "html",
|
||||
"sha256": "04d169deea262559f23c5261f63f81b79169320ecd7b16a3b24363a0f95c03c7",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9651.html"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "txt/rfc3986.txt",
|
||||
"type": "txt",
|
||||
"sha256": "3102dae4b68cebe40337730312fcb612297b8928547267e8b3d1ee6002b2d683",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc3986.txt"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "txt/rfc5322.txt",
|
||||
"type": "txt",
|
||||
"sha256": "cb459acca15a2aa895863eaeff6f1d5b9c4ca15578ae3e3706f757787a385969",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc5322.txt"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "txt/rfc6749.txt",
|
||||
"type": "txt",
|
||||
"sha256": "f204fc8661d6c92d2ec6e0b54808f961a9ad26e792f57f312d9528335519bd71",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc6749.txt"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "txt/rfc7519.txt",
|
||||
"type": "txt",
|
||||
"sha256": "fecd930e9ccf2276b95c0017c6c4ff5d09352e4bc3c7629946447894e0f97248",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc7519.txt"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "txt/rfc8259.txt",
|
||||
"type": "txt",
|
||||
"sha256": "61a5378f4255c720beb2a4b4a63b29540147c140f36988bf086291989b4cd2d7",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc8259.txt"
|
||||
},
|
||||
{
|
||||
"root": "cases",
|
||||
"path": "txt/rfc9110.txt",
|
||||
"type": "txt",
|
||||
"sha256": "21c1cdce6ab0e5509b04d84a28000836c7a087cf786efe6f04877ebfff47232a",
|
||||
"origin": "public, IETF RFC fetched 2026-09-27",
|
||||
"url": "https://www.rfc-editor.org/rfc/rfc9110.txt"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20205008786.docx",
|
||||
"type": "docx",
|
||||
"sha256": "d1c2644b2da7eedae7c61fe974e855ba0e4302ea5407ef1e9178887f7e06364f",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20205009574.docx",
|
||||
"type": "docx",
|
||||
"sha256": "530d7320ba0fd1d9f9b40a55f37fc5fad030ed9e7436a4e6db0a1d652d704bdf",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20210010866.docx",
|
||||
"type": "docx",
|
||||
"sha256": "be21ae9200acf15d9923107b63f95d4c4ec567439f7b9812ae59b8e44b712b2c",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20210024410.docx",
|
||||
"type": "docx",
|
||||
"sha256": "539fe8f7541cd26e841faadd56c6d6b8189a7eb8505bbe6aef6cf4ef79c1a3d1",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20220000312.docx",
|
||||
"type": "docx",
|
||||
"sha256": "ca89a94c76b41b4fd357b2c0febf9ceb246b241923d8c4762489f086e9c3bddb",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20220015605.docx",
|
||||
"type": "docx",
|
||||
"sha256": "12d3c45379996e7c66c9f3c0b350059cde772af85ea061f5261f10d90b919872",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20220016968.docx",
|
||||
"type": "docx",
|
||||
"sha256": "a7e8b4a637bcd95d0734d3b3519f16a6d7db46504e8922e560bd41f2fa5e11fb",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20230003166.docx",
|
||||
"type": "docx",
|
||||
"sha256": "e4ba28940e708ae3b0d7a16cb152034e326b0a008fb37664a02afa4b1fb155b6",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20230013967.docx",
|
||||
"type": "docx",
|
||||
"sha256": "f5ec5b5a07d1a3061d67c41dc8eb72a52c0cc2eb2c0391459b86b2c9ca5cf626",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/docx/ntrs-20240002302.docx",
|
||||
"type": "docx",
|
||||
"sha256": "baf6fa67a3b9bf73e126a0dfdfd0b3b84062259fbb4f4448d52c4f5387151202",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-19650012761.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "abc437645f2d42e16f006370fd124862f7afdf271a8e0a4c88278c26125147c6",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-19900003140.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "3f8d6ea00c59cd6e07910d54415f8c60b8de443cc99d0b3ee7884d8158895816",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-19970003047.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "09a8f1f8d4ba8ac746d808533ef12981c788d791a83ff7518e2db634bd14d8a2",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20010073812.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "d456715b474113aec36b8336b45ad86b55bf4d4d20e24250a23e2a4267769fbf",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20050207381.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "d31dff490ed4593b217faa93c2569c2a913a7971c229a31f715cef6fea22e677",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20110007792.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "f23f6a718326076e6c72801b3564374b49e8cb5d0bfc99548311d9ab6f24d3c4",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20180004759.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "ac9bb2e7a6d7130d2ebde0e1bf0f96835fca331c1c2c0a471f39c3e3b2c8eace",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20205003902.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "4bf7920d3d62b9bfc548a768575d47fd54d58730f3a9c0b8cd94f874a8c400c8",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20220008919.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "5d834740c6f136f04e73f27c7c704b806edafbf1ac2bf0ae165ad90ba1d59302",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pdf/ntrs-20240007574.pdf",
|
||||
"type": "pdf",
|
||||
"sha256": "50e1654e76746f35575f4d800d4930c47aeecad4789df7c765523a067d2f1dd9",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20205007521.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "1322b9af7b2fe5f08c37940a38d9b7d5f7164a0293b0105749ff05e12babbcad",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20205007851.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "538a3fccda9314c1e05828cf5eea75016327764718dd8cfb4e13255b0270076c",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20210012920.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "50e24adc845d5df65c53ba37cd6414c29e7e68d6ee0e5f8109e20259d51786b3",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20210014153.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "a9c6827393591e732c9c62d0f2b5ac21f14125cc5b31d563830cdfa264335f53",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20210023316.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "fb658cd952883ec6291b26bc61ce73c555a73e66e4ab0c8cb15761211d23c799",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20230007913.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "a1a34b88d68a1f09dbaaa7f6f559de54a22875f519b2adba330264f04b8af077",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20230013937.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "9f22c3b8785469115876e45bfa3108a3fe6e2fca1b6a5d8d4be3d486ef67424f",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20230017811.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "d6c1a39fba65882b4c750146a39e3eff3a8702937a2397af77704ff2973ef9f6",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20240015731.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "60b9f510bdaa71349d946d0918f33c4bee92f2e5109c74c973333c35c0defa9e",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/pptx/ntrs-20260007869.pptx",
|
||||
"type": "pptx",
|
||||
"sha256": "f41f3e613854ff7d858b950e8ceb622c2ec70726763bdf01a439feac86c9f9ea",
|
||||
"origin": "public, NASA NTRS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/nist-sp800-201_fra_data_sample.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "e87cb5f3120f20819a65da3f73db4976184b33262329c5676a1beb19da48dd17",
|
||||
"origin": "public, NIST (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/nist-sp800-53r4-to-r5-comparison-workbook.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "94465ffc5584e0c968342cfe0492f1c31b2854bed8b61914bf517651a5664570",
|
||||
"origin": "public, NIST (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/nist-sp800-53r5-control-catalog.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "1777ed6f839bccd55516e6e703bfe75e3cf63efb2ec80bb9309ed960af2c9ea4",
|
||||
"origin": "public, NIST (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-cobal-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "c2df472e47050d85ef03bba211ed1a75633ba5ba997e1e9ce48230dca749f9df",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-germa-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "5659778e60e7769f8ff66da8812199d39766fdc41d8f24b7a1d5fa2118c6fe31",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-gold-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "97cd9d361ae7b7e4a408109a3ba4d0e3a51480f008ca318d0a3c7ad5402d50e8",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-graph-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "20904e85a0019537f63e6cba4494216ce813b4a2339ae2b1687cf4b46befbb6e",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-lead-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "f3a7bfd684fd892efd71eb788d17b6322611b97b306104207e1be05020137baf",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-mica-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "5901884a879c5ed9854b9bc1ab8cd78b0cf17800287d7a6f29e6dc85b31994bf",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-sulfu-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "59c66830cb77bd04b2a2f2c6e45b91234651b85ef6c514d3581df8be901a04b3",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "blandet/kilde/xlsx/usgs-myb1-2022-zinc-ert.xlsx",
|
||||
"type": "xlsx",
|
||||
"sha256": "1857b943aa7617b0c21148dde2b366d289f7b7e9ab8ac1e941962f0a1096b0de",
|
||||
"origin": "public, USGS (the mixed collection's source folder)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/docs-en-permissions.md",
|
||||
"type": "md",
|
||||
"sha256": "d3e0f513776bc0a0351489596a341575e25b175d33be4ef6ef834216f792bfcf",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/docs-en-sandbox-environments.md",
|
||||
"type": "md",
|
||||
"sha256": "03730f093b2f1f78d8af65775f3516244af56ce3782b709f8e79bc34fb257072",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/docs-en-terminal-config.md",
|
||||
"type": "md",
|
||||
"sha256": "caf020a9955995c88fbf5d208cac9e78c054ad6c742975d98bd10424bd2a8fdb",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/docs-en-whats-new-index.md",
|
||||
"type": "md",
|
||||
"sha256": "854e3ed48564385023b3ed25bbdb703e9a84c41a6abc312be38bc4efbce3a59b",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/release-v1.0.86.md",
|
||||
"type": "md",
|
||||
"sha256": "76bc2feec3fdce6a3b302a894304c48bde202481ec9d86a852aa85f3613a2c79",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/release-v2.1.122.md",
|
||||
"type": "md",
|
||||
"sha256": "40d0686ce5fb73feb04e525d67ef57565c56eb9665c111a54bbe6e09a1fe869e",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/release-v2.1.147.md",
|
||||
"type": "md",
|
||||
"sha256": "cb3199d49c9e92dcbadf1e201606f0e0f629ec6d9246d17e3477d33f19ece83f",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/release-v2.1.218.md",
|
||||
"type": "md",
|
||||
"sha256": "16a45ae72c677bd4f91b497e3d96edf849dd476170db15c19f5039667e397e75",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/release-v2.1.257.md",
|
||||
"type": "md",
|
||||
"sha256": "76e629f50f6eb2e77e9c39f03cadcd6c765070e32e92d92ae8b5aa68bb687dc9",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
},
|
||||
{
|
||||
"root": "sources",
|
||||
"path": "wiki/kilde/release-v2.1.92.md",
|
||||
"type": "md",
|
||||
"sha256": "0fd18d8bca192d5b1a17a7bded3b65fc8fcd5a72dd630f8eb43e2b52342a8434",
|
||||
"origin": "collection 1 source folder (Anthropic documentation, never committed: rule 3)"
|
||||
}
|
||||
]
|
||||
}
|
||||
93
gates/keyform.py
Normal file
93
gates/keyform.py
Normal file
|
|
@ -0,0 +1,93 @@
|
|||
"""Answer-key form (E2c): counts answer keys that are wording rather than a fact. No model call.
|
||||
|
||||
An answer key is the smallest fact an answer must hold: a name, a command, a flag, a setting, a number, a version or
|
||||
one word that carries the meaning. A key that joins several words with a function word (an article, a pronoun or
|
||||
determiner, a conjunction, a preposition, an auxiliary or modal verb, a negation, or a focusing particle) is wording:
|
||||
a correct answer can say the same thing in other words and miss the key. Text inside backticks is code, not wording.
|
||||
|
||||
python3 gates/keyform.py --set <set.json> [--set <set.json> ...] [--list]
|
||||
|
||||
Prints "answer keys that are wording: X of N" over every key in the sets (either schema). Exit 0 = PASS (X = 0 and
|
||||
N > 0); 1 = FAIL or could not run.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import re
|
||||
import sys
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
FUNCTION_WORDS = frozenset("""
|
||||
a an the
|
||||
i me my you your he him his she her it its we us our they them their this that these those who whom whose which what
|
||||
each every all any some
|
||||
and or but nor so if when while because than as whether then
|
||||
to of in on at by for from with without within into onto over under about after before between through beyond via
|
||||
per up down out off against during until
|
||||
is are was were be been being am can could may might will would shall should must do does did has have had
|
||||
isn't aren't wasn't weren't don't doesn't didn't can't cannot won't wouldn't shouldn't hasn't haven't
|
||||
not no never only also still just
|
||||
""".split())
|
||||
|
||||
|
||||
def is_wording(text: str) -> bool:
|
||||
"""True when the key has two or more words and a word outside backticks is a function word."""
|
||||
if len(text.split()) < 2:
|
||||
return False
|
||||
outside = re.sub(r"`[^`]*`", " ", text)
|
||||
words = (w.strip(".,;:!?()[]{}\"'").casefold() for w in outside.split())
|
||||
return any(w in FUNCTION_WORDS for w in words)
|
||||
|
||||
|
||||
@dataclass
|
||||
class KeyReport:
|
||||
keys: int = 0
|
||||
wording: int = 0
|
||||
listed: list[str] = field(default_factory=list)
|
||||
|
||||
|
||||
def count_keys(questions: list[dict[str, Any]]) -> KeyReport:
|
||||
rep = KeyReport()
|
||||
for q in questions:
|
||||
for k in q.get("answer_keys", []) if isinstance(q, dict) else []:
|
||||
text = k.get("text") if isinstance(k, dict) else None
|
||||
rep.keys += 1
|
||||
if isinstance(text, str) and is_wording(text):
|
||||
rep.wording += 1
|
||||
rep.listed.append(f"{q.get('id')}: {text}")
|
||||
return rep
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
p.add_argument("--set", action="append", required=True)
|
||||
p.add_argument("--list", action="store_true", help="list every wording key")
|
||||
args = p.parse_args(argv)
|
||||
total = KeyReport()
|
||||
for path in args.set:
|
||||
try:
|
||||
data = json.loads(Path(path).read_text(encoding="utf-8"))
|
||||
except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
|
||||
print(f"FAIL: cannot read set {path}: {exc}", file=sys.stderr)
|
||||
return 1
|
||||
if not isinstance(data, dict) or not isinstance(data.get("questions"), list):
|
||||
print(f"FAIL: {path} is not a set", file=sys.stderr)
|
||||
return 1
|
||||
rep = count_keys(data["questions"])
|
||||
total.keys += rep.keys
|
||||
total.wording += rep.wording
|
||||
total.listed += [f"{Path(path).name} {line}" for line in rep.listed]
|
||||
if args.list:
|
||||
print("\n".join(total.listed))
|
||||
print(f"answer keys that are wording: {total.wording} of {total.keys}")
|
||||
passed = total.keys > 0 and total.wording == 0
|
||||
print("PASS" if passed else "FAIL")
|
||||
return 0 if passed else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
34
gates/multidoc.manifest.json
Normal file
34
gates/multidoc.manifest.json
Normal file
|
|
@ -0,0 +1,34 @@
|
|||
{
|
||||
"schema": "okf-wiki/multidoc-manifest/1",
|
||||
"seed": 20260926,
|
||||
"draw": "random.Random(seed).sample(sorted(ids), held_out.questions); both parts renumbered d../h..",
|
||||
"dev": {
|
||||
"file": "gates-local/multidoc/set-60.json",
|
||||
"sha256": "bc98c9a16d18f7a851c2ce319a32325244076f8be30ea3d8266bdea7a21aad68",
|
||||
"questions": 60
|
||||
},
|
||||
"held_out": {
|
||||
"sha256": "ca7c77fbe02a674728c0fb2023b4d64a8b2003bf7ad303e7cc1177040129bf72",
|
||||
"questions": 20
|
||||
},
|
||||
"sources": {
|
||||
"files": 599,
|
||||
"tree_sha256": "6daf3ed73335f3e6f332c7b188adb5aebbc659b6fdd1d7591280e129648a06d5"
|
||||
},
|
||||
"keys_rewritten": {
|
||||
"order": "E2c",
|
||||
"rule": "answer keys rewritten as facts (gates/keyform.py): every key of two or more words with a function word outside backticks became the runs of its own words without the function words, each still inside its span; questions, files, spans, counts and patterns unchanged",
|
||||
"before": {
|
||||
"dev_sha256": "342686e100f2016e376bc4ecb28c14657646d1c35ba8e3bcad3593f4204a001a",
|
||||
"held_out_sha256": "dd061056f8dadb62831623348a339fb1849ca8714bd83abb2b059428342438b8"
|
||||
}
|
||||
},
|
||||
"negated_marked": {
|
||||
"order": "E2d",
|
||||
"rule": "\"negated\": true added to the polarity keys of E2c's list that stand negated in their own spans (gates/multidoc.py negated_in): 17 of 22; nothing else changed",
|
||||
"before": {
|
||||
"dev_sha256": "b4832f93d10090d439554d1a159a5140b03eecf6eb453faafa6638aec50f23b0",
|
||||
"held_out_sha256": "b062fd3a8b97f7599c29f9c47e0313ee96e424f03cd24f6e28d650fd831639c7"
|
||||
}
|
||||
}
|
||||
}
|
||||
492
gates/multidoc.py
Normal file
492
gates/multidoc.py
Normal file
|
|
@ -0,0 +1,492 @@
|
|||
"""The multi-document set for PLAN.md criterion 4: checker, held-out split and answer gate. No model call.
|
||||
|
||||
check every answer span stands verbatim in its source file, every question needs >= 4 distinct files,
|
||||
and every file carries an answer key that stands inside that file's spans. The counts printed
|
||||
here are the denominators; they come from the set, not from whoever wrote it.
|
||||
split draws the held-out questions with a fixed seed, writes them OUTSIDE the repository, renumbers
|
||||
both parts, and writes a manifest with sha256 only (nothing about the held-out questions).
|
||||
gate scores one arm's answers: a question counts when the answer holds every answer key AND the
|
||||
citations quote, verbatim, every file the answer key needs. A missing arm scores 0 (fail-closed).
|
||||
|
||||
Exit 0 = PASS / scored completely; 1 = FAIL or could not run. A check that cannot run is a FAIL.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import random
|
||||
import re
|
||||
import sys
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from keyform import count_keys
|
||||
|
||||
SET_SCHEMA = "okf-wiki/multidoc-set/1"
|
||||
MANIFEST_SCHEMA = "okf-wiki/multidoc-manifest/1"
|
||||
ANSWERS_SCHEMA_NOTE = '{"arm": "<name>", "answers": [{"id", "answer", "citations": [{"file", "quote"}]}]}'
|
||||
MIN_FILES = 4
|
||||
SPAN_MIN, SPAN_MAX = 20, 400
|
||||
KEY_MIN, KEY_MAX = 3, 60
|
||||
NUMBER_WORDS = ("zero one two three four five six seven eight nine ten eleven twelve thirteen fourteen fifteen "
|
||||
"sixteen seventeen eighteen nineteen twenty").split()
|
||||
NEGATION_WORDS = frozenset("not no never none neither nor without cannot".split())
|
||||
NEGATION_WINDOW = 6
|
||||
REPO_ROOT = Path(__file__).resolve().parents[1]
|
||||
|
||||
|
||||
class RunError(Exception):
|
||||
"""The command cannot run; reported as FAIL with exit 1."""
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- helpers
|
||||
|
||||
|
||||
def sha256_bytes(data: bytes) -> str:
|
||||
return hashlib.sha256(data).hexdigest()
|
||||
|
||||
|
||||
def tree_sha256(sources: Path) -> str:
|
||||
"""One hash over the flat sources directory: sorted "name NUL sha256 LF" lines."""
|
||||
lines = []
|
||||
for p in sorted(sources.iterdir(), key=lambda p: p.name):
|
||||
if p.is_file():
|
||||
lines.append(f"{p.name}\0{sha256_bytes(p.read_bytes())}\n")
|
||||
return sha256_bytes("".join(lines).encode("utf-8"))
|
||||
|
||||
|
||||
def count_files(sources: Path) -> int:
|
||||
return sum(1 for p in sources.iterdir() if p.is_file())
|
||||
|
||||
|
||||
def normalise_answer(text: str) -> str:
|
||||
"""Case-folded, without backticks and asterisks, whitespace collapsed. Used for answer keys only."""
|
||||
return re.sub(r"\s+", " ", text.replace("`", "").replace("*", "")).strip().casefold()
|
||||
|
||||
|
||||
def number_key(key: str) -> tuple[int, bool] | None:
|
||||
"""(n, is_percentage) when the key is a whole number, a percentage or an English number word; otherwise None."""
|
||||
text = normalise_answer(key)
|
||||
m = re.fullmatch(r"(\d+) ?(%?)", text)
|
||||
if m:
|
||||
return int(m.group(1)), m.group(2) == "%"
|
||||
return (NUMBER_WORDS.index(text), False) if text in NUMBER_WORDS else None
|
||||
|
||||
|
||||
def key_positions(key: str, norm: str) -> list[int]:
|
||||
"""Start offsets of the answer key in normalised text. A number key matches as a number: its digits or its
|
||||
English word, never inside a longer number, version or identifier. Any other key matches as normalised text."""
|
||||
num = number_key(key)
|
||||
if num is None:
|
||||
needle = normalise_answer(key)
|
||||
return [m.start() for m in re.finditer(re.escape(needle), norm)] if needle else []
|
||||
n, percent = num
|
||||
if percent:
|
||||
return [m.start() for m in re.finditer(rf"(?<![\w.]){n} ?(?:%|percent\b)", norm)]
|
||||
forms = [str(n)] + ([NUMBER_WORDS[n]] if n < len(NUMBER_WORDS) else [])
|
||||
return [m.start() for m in re.finditer(rf"(?<![\w.])(?:{'|'.join(forms)})(?![\w%]|\.\d)", norm)]
|
||||
|
||||
|
||||
def key_in(key: str, text: str) -> bool:
|
||||
"""True when the answer key stands in the text (see key_positions)."""
|
||||
return bool(key_positions(key, normalise_answer(text)))
|
||||
|
||||
|
||||
def is_negation(word: str) -> bool:
|
||||
"""True when the word is a negation: one of NEGATION_WORDS or a word ending in n't."""
|
||||
w = word.strip(".,;:!?()[]{}\"'").replace("’", "'").casefold()
|
||||
return w in NEGATION_WORDS or w.endswith("n't")
|
||||
|
||||
|
||||
def sentences(text: str) -> list[str]:
|
||||
"""The text split at line breaks and at a full stop, question mark, exclamation mark or semicolon ending a word."""
|
||||
return [s for s in re.split(r"\n|[.!?;](?=\s|$)", text) if s.strip()]
|
||||
|
||||
|
||||
def negated_at(norm: str, pos: int) -> bool:
|
||||
"""True when one of the NEGATION_WINDOW words before offset pos in a normalised sentence is a negation."""
|
||||
return any(is_negation(w) for w in norm[:pos].split()[-NEGATION_WINDOW:])
|
||||
|
||||
|
||||
def negated_in(key: str, text: str) -> bool:
|
||||
"""True when, within one sentence of the text, the key stands with a negation at most NEGATION_WINDOW words
|
||||
before it."""
|
||||
for s in sentences(text):
|
||||
norm = normalise_answer(s)
|
||||
if any(negated_at(norm, pos) for pos in key_positions(key, norm)):
|
||||
return True
|
||||
return False
|
||||
|
||||
|
||||
def key_holds(key: dict[str, Any], text: str) -> bool:
|
||||
"""The gates' one test of an answer key against an answer. A key marked "negated": true holds only where it
|
||||
stands negated (negated_in); any other key holds where it stands (key_in), with or without a negation."""
|
||||
if key.get("negated") is True:
|
||||
return negated_in(key["text"], text)
|
||||
return key_in(key["text"], text)
|
||||
|
||||
|
||||
def normalise_quote(text: str) -> str:
|
||||
"""Whitespace collapsed, otherwise exact. Used for an arm's citations."""
|
||||
return re.sub(r"\s+", " ", text).strip()
|
||||
|
||||
|
||||
def as_list(value: object) -> list[Any]:
|
||||
return value if isinstance(value, list) else []
|
||||
|
||||
|
||||
def is_plain_name(name: object) -> bool:
|
||||
return isinstance(name, str) and name != "" and "/" not in name and "\\" not in name and name not in (".", "..")
|
||||
|
||||
|
||||
def load_json(path: Path, what: str) -> Any:
|
||||
try:
|
||||
return json.loads(path.read_text(encoding="utf-8"))
|
||||
except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
|
||||
raise RunError(f"cannot read {what} {path}: {exc}") from exc
|
||||
|
||||
|
||||
def load_questions(path: Path) -> list[dict[str, Any]]:
|
||||
data = load_json(path, "set")
|
||||
if not isinstance(data, dict) or data.get("schema") != SET_SCHEMA or not isinstance(data.get("questions"), list):
|
||||
raise RunError(f"{path} is not a set with schema {SET_SCHEMA}")
|
||||
questions: list[dict[str, Any]] = data["questions"]
|
||||
return questions
|
||||
|
||||
|
||||
def require_dir(path: Path, what: str) -> None:
|
||||
if not path.is_dir():
|
||||
raise RunError(f"{what} {path} is not a directory")
|
||||
|
||||
|
||||
class SourceCache:
|
||||
def __init__(self, root: Path) -> None:
|
||||
self.root = root
|
||||
self._text: dict[str, str | None] = {}
|
||||
|
||||
def text(self, name: str) -> str | None:
|
||||
"""File text, or None when the name is not a plain file directly in the sources directory."""
|
||||
if not is_plain_name(name):
|
||||
return None
|
||||
if name not in self._text:
|
||||
p = self.root / name
|
||||
try:
|
||||
self._text[name] = p.read_text(encoding="utf-8") if p.is_file() else None
|
||||
except (OSError, UnicodeDecodeError):
|
||||
self._text[name] = None
|
||||
return self._text[name]
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- check
|
||||
|
||||
|
||||
@dataclass
|
||||
class CheckReport:
|
||||
questions: int = 0
|
||||
spans: int = 0
|
||||
spans_verbatim: int = 0
|
||||
all_spans_verbatim: int = 0
|
||||
min_files: int = 0
|
||||
key_per_file: int = 0
|
||||
keys: int = 0
|
||||
wording_keys: int = 0
|
||||
negated_keys: int = 0
|
||||
negated_ok: int = 0
|
||||
errors: list[str] = field(default_factory=list)
|
||||
files_per_question: dict[str, int] = field(default_factory=dict)
|
||||
|
||||
@property
|
||||
def passed(self) -> bool:
|
||||
n = self.questions
|
||||
return (
|
||||
n > 0
|
||||
and not self.errors
|
||||
and self.spans_verbatim == self.spans
|
||||
and self.all_spans_verbatim == n
|
||||
and self.min_files == n
|
||||
and self.key_per_file == n
|
||||
and self.wording_keys == 0
|
||||
and self.negated_ok == self.negated_keys
|
||||
)
|
||||
|
||||
def lines(self) -> list[str]:
|
||||
n = self.questions
|
||||
out = [f"error: {e}" for e in self.errors]
|
||||
out += [
|
||||
f"questions: {n}",
|
||||
f"spans verbatim: {self.spans_verbatim} of {self.spans}",
|
||||
f"questions with every span verbatim: {self.all_spans_verbatim} of {n}",
|
||||
f"questions with >= {MIN_FILES} distinct files: {self.min_files} of {n}",
|
||||
f"questions with a key per file: {self.key_per_file} of {n}",
|
||||
f"answer keys that are wording: {self.wording_keys} of {self.keys}",
|
||||
f"negated keys negated in their span: {self.negated_ok} of {self.negated_keys}",
|
||||
"PASS" if self.passed else "FAIL",
|
||||
]
|
||||
return out
|
||||
|
||||
|
||||
def check_questions(questions: list[dict[str, Any]], sources: Path) -> CheckReport:
|
||||
require_dir(sources, "sources")
|
||||
cache = SourceCache(sources)
|
||||
rep = CheckReport(questions=len(questions))
|
||||
form = count_keys(questions)
|
||||
rep.keys, rep.wording_keys = form.keys, form.wording
|
||||
seen: set[str] = set()
|
||||
for q in questions:
|
||||
qid = q.get("id") if isinstance(q, dict) else None
|
||||
if not isinstance(qid, str) or not qid:
|
||||
rep.errors.append("question without an id")
|
||||
continue
|
||||
if qid in seen:
|
||||
rep.errors.append(f"duplicate id {qid}")
|
||||
seen.add(qid)
|
||||
if not isinstance(q.get("question"), str) or not q["question"].strip():
|
||||
rep.errors.append(f"{qid}: empty question")
|
||||
evidence = as_list(q.get("evidence"))
|
||||
keys = as_list(q.get("answer_keys"))
|
||||
spans_by_file: dict[str, list[str]] = {}
|
||||
verbatim = 0
|
||||
for e in evidence:
|
||||
name, span = (e.get("file"), e.get("span")) if isinstance(e, dict) else (None, None)
|
||||
rep.spans += 1
|
||||
if not isinstance(name, str) or not is_plain_name(name):
|
||||
rep.errors.append(f"{qid}: evidence file {name!r} is not a plain file name")
|
||||
continue
|
||||
if not isinstance(span, str):
|
||||
rep.errors.append(f"{qid}: evidence span in {name} is not text")
|
||||
continue
|
||||
if not SPAN_MIN <= len(span) <= SPAN_MAX:
|
||||
rep.errors.append(f"{qid}: span length {len(span)} in {name} outside {SPAN_MIN}-{SPAN_MAX}")
|
||||
text = cache.text(name)
|
||||
if text is not None and span in text:
|
||||
verbatim += 1
|
||||
spans_by_file.setdefault(name, []).append(span)
|
||||
rep.spans_verbatim += verbatim
|
||||
if evidence and verbatim == len(evidence):
|
||||
rep.all_spans_verbatim += 1
|
||||
files = set(spans_by_file)
|
||||
rep.files_per_question[qid] = len(files)
|
||||
if len(files) >= MIN_FILES:
|
||||
rep.min_files += 1
|
||||
keyed: set[str] = set()
|
||||
keys_ok = bool(keys)
|
||||
for k in keys:
|
||||
name, key = (k.get("file"), k.get("text")) if isinstance(k, dict) else (None, None)
|
||||
if not isinstance(key, str) or not isinstance(name, str):
|
||||
keys_ok = False
|
||||
rep.errors.append(f"{qid}: malformed answer key")
|
||||
continue
|
||||
if number_key(key) is None and not KEY_MIN <= len(key.strip()) <= KEY_MAX:
|
||||
rep.errors.append(f"{qid}: key length {len(key.strip())} outside {KEY_MIN}-{KEY_MAX}")
|
||||
if any(key_in(key, s) for s in spans_by_file.get(name, [])):
|
||||
keyed.add(name)
|
||||
else:
|
||||
keys_ok = False
|
||||
if "negated" in k:
|
||||
if k["negated"] is not True:
|
||||
rep.errors.append(f"{qid}: negated must be true or absent")
|
||||
continue
|
||||
rep.negated_keys += 1
|
||||
rep.negated_ok += any(negated_in(key, s) for s in spans_by_file.get(name, []))
|
||||
if keys_ok and files and keyed == files:
|
||||
rep.key_per_file += 1
|
||||
return rep
|
||||
|
||||
|
||||
def cmd_check(args: argparse.Namespace) -> int:
|
||||
rep = check_questions(load_questions(Path(args.set)), Path(args.sources))
|
||||
if args.per_question:
|
||||
for qid, n in rep.files_per_question.items():
|
||||
print(f"{qid}\t{n} files")
|
||||
print("\n".join(rep.lines()))
|
||||
return 0 if rep.passed else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- split
|
||||
|
||||
|
||||
def inside_repo(path: Path) -> bool:
|
||||
try:
|
||||
path.resolve().relative_to(REPO_ROOT)
|
||||
return True
|
||||
except ValueError:
|
||||
return False
|
||||
|
||||
|
||||
def dump_set(questions: list[dict[str, Any]]) -> bytes:
|
||||
return (json.dumps({"schema": SET_SCHEMA, "questions": questions}, ensure_ascii=False, indent=1) + "\n").encode("utf-8")
|
||||
|
||||
|
||||
def cmd_split(args: argparse.Namespace) -> int:
|
||||
sources, held_path = Path(args.sources), Path(args.held_out)
|
||||
dev_path, manifest_path = Path(args.dev_out), Path(args.manifest)
|
||||
if inside_repo(held_path):
|
||||
raise RunError(f"held-out path {held_path} is inside the repository; it must live outside it")
|
||||
if held_path.exists():
|
||||
raise RunError(f"held-out file {held_path} already exists; a frozen draw is never redrawn")
|
||||
questions = load_questions(Path(args.set))
|
||||
rep = check_questions(questions, sources)
|
||||
if not rep.passed:
|
||||
print("\n".join(rep.lines()))
|
||||
raise RunError("the set fails the check; nothing was split")
|
||||
n_held = args.held_out_count
|
||||
if not 0 < n_held < len(questions):
|
||||
raise RunError(f"cannot hold out {n_held} of {len(questions)}")
|
||||
ids = sorted(q["id"] for q in questions)
|
||||
held_ids = set(random.Random(args.seed).sample(ids, n_held))
|
||||
by_id = {q["id"]: q for q in questions}
|
||||
dev = [dict(by_id[i], id=f"d{n:02d}") for n, i in enumerate((i for i in ids if i not in held_ids), 1)]
|
||||
held = [dict(by_id[i], id=f"h{n:02d}") for n, i in enumerate(sorted(held_ids), 1)]
|
||||
dev_bytes, held_bytes = dump_set(dev), dump_set(held)
|
||||
held_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
held_path.write_bytes(held_bytes)
|
||||
dev_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
dev_path.write_bytes(dev_bytes)
|
||||
manifest = {
|
||||
"schema": MANIFEST_SCHEMA,
|
||||
"seed": args.seed,
|
||||
"draw": "random.Random(seed).sample(sorted(ids), held_out.questions); both parts renumbered d../h..",
|
||||
"dev": {"file": args.dev_label or dev_path.name, "sha256": sha256_bytes(dev_bytes), "questions": len(dev)},
|
||||
"held_out": {"sha256": sha256_bytes(held_bytes), "questions": len(held)},
|
||||
"sources": {"files": count_files(sources), "tree_sha256": tree_sha256(sources)},
|
||||
}
|
||||
manifest_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
manifest_path.write_text(json.dumps(manifest, indent=1) + "\n", encoding="utf-8")
|
||||
print(f"seed: {args.seed}")
|
||||
print(f"dev: {len(dev)} questions, sha256 {manifest['dev']['sha256']}")
|
||||
print(f"held out: {len(held)} questions, sha256 {manifest['held_out']['sha256']}")
|
||||
print(f"manifest: {manifest_path}")
|
||||
return 0
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- gate
|
||||
|
||||
|
||||
@dataclass
|
||||
class GateResult:
|
||||
qid: str
|
||||
answered: bool
|
||||
answer_ok: bool
|
||||
source_ok: bool
|
||||
|
||||
@property
|
||||
def correct(self) -> bool:
|
||||
return self.answer_ok and self.source_ok
|
||||
|
||||
|
||||
def score_one(q: dict[str, Any], a: dict[str, Any] | None, cache: SourceCache) -> GateResult:
|
||||
if a is None:
|
||||
return GateResult(q["id"], False, False, False)
|
||||
answer = a.get("answer", "") if isinstance(a.get("answer"), str) else ""
|
||||
answer_ok = all(key_holds(k, answer) for k in q["answer_keys"])
|
||||
supported: set[str] = set()
|
||||
citations = as_list(a.get("citations"))
|
||||
for c in citations:
|
||||
name, quote = (c.get("file"), c.get("quote")) if isinstance(c, dict) else (None, None)
|
||||
if not isinstance(name, str) or not isinstance(quote, str) or not quote.strip():
|
||||
continue
|
||||
text = cache.text(name)
|
||||
if text is not None and normalise_quote(quote) in normalise_quote(text):
|
||||
supported.add(name)
|
||||
needed = {e["file"] for e in q["evidence"]}
|
||||
return GateResult(q["id"], True, answer_ok, needed <= supported)
|
||||
|
||||
|
||||
def cmd_gate(args: argparse.Namespace) -> int:
|
||||
set_path, sources, manifest_path = Path(args.set), Path(args.sources), Path(args.manifest)
|
||||
require_dir(sources, "sources")
|
||||
manifest = load_json(manifest_path, "manifest")
|
||||
if not isinstance(manifest, dict) or manifest.get("schema") != MANIFEST_SCHEMA:
|
||||
raise RunError(f"{manifest_path} is not a manifest with schema {MANIFEST_SCHEMA}")
|
||||
set_hash = sha256_bytes(set_path.read_bytes()) if set_path.is_file() else ""
|
||||
known = {manifest.get("dev", {}).get("sha256"): "dev", manifest.get("held_out", {}).get("sha256"): "held_out"}
|
||||
part = known.get(set_hash)
|
||||
if part is None:
|
||||
raise RunError(f"set {set_path} sha256 {set_hash or '(missing)'} matches neither part of the manifest")
|
||||
if tree_sha256(sources) != manifest.get("sources", {}).get("tree_sha256"):
|
||||
raise RunError(f"sources {sources} differ from the manifest's tree_sha256; the set is frozen against them")
|
||||
questions = load_questions(set_path)
|
||||
n = len(questions)
|
||||
cache = SourceCache(sources)
|
||||
answers: dict[str, dict[str, Any]] = {}
|
||||
fail_closed = ""
|
||||
answers_path = Path(args.answers)
|
||||
if not answers_path.is_file():
|
||||
fail_closed = f"no answers for arm {args.arm} at {answers_path}"
|
||||
else:
|
||||
data = load_json(answers_path, "answers")
|
||||
if not isinstance(data, dict) or not isinstance(data.get("answers"), list):
|
||||
raise RunError(f"{answers_path} is not an answers file {ANSWERS_SCHEMA_NOTE}")
|
||||
if data.get("arm") != args.arm:
|
||||
raise RunError(f"answers file is for arm {data.get('arm')!r}, not {args.arm!r}")
|
||||
for a in data["answers"]:
|
||||
if isinstance(a, dict) and isinstance(a.get("id"), str):
|
||||
answers[a["id"]] = a
|
||||
results = [score_one(q, answers.get(q["id"]), cache) for q in questions]
|
||||
missing = sum(1 for r in results if not r.answered)
|
||||
print(f"set: {part}, {n} questions; arm: {args.arm}")
|
||||
if fail_closed:
|
||||
print(f"fail-closed: {fail_closed}")
|
||||
print(f"missing answers: {missing}")
|
||||
print(f"correct answer: {sum(r.answer_ok for r in results)} of {n}")
|
||||
print(f"correct source: {sum(r.source_ok for r in results)} of {n}")
|
||||
print(f"correct answer and source: {sum(r.correct for r in results)} of {n}")
|
||||
if args.out:
|
||||
out = {
|
||||
"arm": args.arm,
|
||||
"set_sha256": set_hash,
|
||||
"results": [
|
||||
{"id": r.qid, "answered": r.answered, "answer_ok": r.answer_ok, "source_ok": r.source_ok, "correct": r.correct}
|
||||
for r in results
|
||||
],
|
||||
}
|
||||
Path(args.out).write_text(json.dumps(out, indent=1) + "\n", encoding="utf-8")
|
||||
scored = not fail_closed and missing == 0
|
||||
print("SCORED" if scored else "UNSCORED (fail-closed)" if fail_closed else "INCOMPLETE")
|
||||
return 0 if scored else 1
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------------------------- main
|
||||
|
||||
|
||||
def build_parser() -> argparse.ArgumentParser:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
sub = p.add_subparsers(dest="command", required=True)
|
||||
c = sub.add_parser("check", help="check spans, file counts and answer keys")
|
||||
c.add_argument("--set", required=True)
|
||||
c.add_argument("--sources", required=True)
|
||||
c.add_argument("--per-question", action="store_true", help="also list distinct files per question")
|
||||
s = sub.add_parser("split", help="draw the held-out questions with a fixed seed")
|
||||
s.add_argument("--set", required=True)
|
||||
s.add_argument("--sources", required=True)
|
||||
s.add_argument("--seed", type=int, required=True)
|
||||
s.add_argument("--held-out-count", type=int, default=20)
|
||||
s.add_argument("--dev-out", required=True)
|
||||
s.add_argument("--dev-label", default="", help="path of the dev set as written in the manifest")
|
||||
s.add_argument("--held-out", required=True, help="file path OUTSIDE the repository")
|
||||
s.add_argument("--manifest", required=True)
|
||||
g = sub.add_parser("gate", help="score one arm's answers against the answer key")
|
||||
g.add_argument("--set", required=True)
|
||||
g.add_argument("--manifest", required=True)
|
||||
g.add_argument("--sources", required=True)
|
||||
g.add_argument("--arm", required=True)
|
||||
g.add_argument("--answers", required=True, help=ANSWERS_SCHEMA_NOTE)
|
||||
g.add_argument("--out", default="", help="write per-question results (for paired tests)")
|
||||
return p
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args = build_parser().parse_args(argv)
|
||||
commands = {"check": cmd_check, "split": cmd_split, "gate": cmd_gate}
|
||||
try:
|
||||
return commands[args.command](args)
|
||||
except RunError as exc:
|
||||
print(f"FAIL: {exc}", file=sys.stderr)
|
||||
return 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
242
gates/mutant_gate.py
Normal file
242
gates/mutant_gate.py
Normal file
|
|
@ -0,0 +1,242 @@
|
|||
"""Mutant gate for the auditor: runs an auditor command on every mutant and on one known-positive bundle.
|
||||
|
||||
No model call happens here. The gate is fail-closed: an auditor that is missing, crashes, times out or answers
|
||||
in the wrong shape has caught nothing.
|
||||
|
||||
Auditor contract (provisional until L1 builds the auditor; the command is a template, so L1 may ignore a field):
|
||||
- argv is the command after `--`, with `{bundle}` replaced by a fresh copy of the case bundle and `{sources}`
|
||||
by the raw-source root that `sources[].resource` paths resolve against.
|
||||
- stdout is one JSON object: {"findings": [{"check": "<class>", ...}, ...]}.
|
||||
- exit 0 with no findings means clean; exit 1 with at least one finding means alarm. Anything else is an error.
|
||||
|
||||
A mutant is *caught* when the auditor raises an alarm on it, and *class matched* when one finding's `check`
|
||||
equals the mutant's class in the manifest. The gate is GREEN only when every mutant is caught and class
|
||||
matched, the known-positive is clean, and nothing errored.
|
||||
|
||||
Before anything runs, every case file and every listed source file is checked against the manifest's sha256;
|
||||
a mismatch, a missing file or an unexpected file stops the gate (exit 2), because a measurement on other
|
||||
fixtures than the manifest names is no measurement of them.
|
||||
|
||||
Usage:
|
||||
python3 gates/mutant_gate.py --manifest gates/mutants.manifest.json --cases gates-local/mutants \\
|
||||
--sources <raw-source root> [--timeout 60] -- <auditor> {bundle} {sources}
|
||||
Exit: 0 GREEN, 1 RED, 2 the gate could not run.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import shutil
|
||||
import subprocess
|
||||
import sys
|
||||
import tempfile
|
||||
from collections import Counter
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
|
||||
KNOWN_POSITIVE = "known-positive"
|
||||
MUTANT = "mutant"
|
||||
|
||||
|
||||
class GateError(Exception):
|
||||
"""The gate cannot produce a measurement (bad manifest, drifted fixtures, bad arguments)."""
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Case:
|
||||
name: str
|
||||
kind: str
|
||||
cls: str | None
|
||||
files: dict[str, str]
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Manifest:
|
||||
sources: dict[str, str]
|
||||
cases: list[Case]
|
||||
|
||||
@property
|
||||
def known_positive(self) -> Case:
|
||||
return next(c for c in self.cases if c.kind == KNOWN_POSITIVE)
|
||||
|
||||
@property
|
||||
def mutants(self) -> list[Case]:
|
||||
return [c for c in self.cases if c.kind == MUTANT]
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Outcome:
|
||||
"""What one auditor run said: 'clean', 'alarm' or 'error'."""
|
||||
|
||||
status: str
|
||||
checks: frozenset[str] = field(default_factory=frozenset)
|
||||
reason: str = ""
|
||||
|
||||
|
||||
def sha256_file(path: Path) -> str:
|
||||
return hashlib.sha256(path.read_bytes()).hexdigest()
|
||||
|
||||
|
||||
def hash_tree(root: Path) -> dict[str, str]:
|
||||
"""sha256 of every regular file under root, keyed by its POSIX path relative to root."""
|
||||
out: dict[str, str] = {}
|
||||
for p in sorted(root.rglob("*")):
|
||||
if p.is_symlink():
|
||||
raise GateError(f"symlink not allowed in a case: {p}")
|
||||
if p.is_file():
|
||||
out[p.relative_to(root).as_posix()] = sha256_file(p)
|
||||
return out
|
||||
|
||||
|
||||
def load_manifest(path: Path) -> Manifest:
|
||||
try:
|
||||
data = json.loads(path.read_text(encoding="utf-8"))
|
||||
except (OSError, json.JSONDecodeError) as exc:
|
||||
raise GateError(f"cannot read manifest {path}: {exc}") from exc
|
||||
if data.get("schema") != 1:
|
||||
raise GateError("manifest schema must be 1")
|
||||
cases: list[Case] = []
|
||||
for raw in data.get("cases", []):
|
||||
kind = raw.get("kind")
|
||||
if kind not in (KNOWN_POSITIVE, MUTANT):
|
||||
raise GateError(f"case {raw.get('name')!r}: kind must be {KNOWN_POSITIVE!r} or {MUTANT!r}")
|
||||
cls = raw.get("class")
|
||||
if kind == MUTANT and not cls:
|
||||
raise GateError(f"mutant {raw.get('name')!r} has no class")
|
||||
files = raw.get("files")
|
||||
if not isinstance(files, dict) or not files:
|
||||
raise GateError(f"case {raw.get('name')!r} lists no files")
|
||||
cases.append(Case(name=str(raw["name"]), kind=kind, cls=cls, files={str(k): str(v) for k, v in files.items()}))
|
||||
names = [c.name for c in cases]
|
||||
if len(names) != len(set(names)):
|
||||
raise GateError("case names are not unique")
|
||||
if sum(c.kind == KNOWN_POSITIVE for c in cases) != 1:
|
||||
raise GateError("manifest must have exactly one known-positive case")
|
||||
if not any(c.kind == MUTANT for c in cases):
|
||||
raise GateError("manifest has no mutants")
|
||||
sources = {str(s["path"]): str(s["sha256"]) for s in data.get("sources", [])}
|
||||
return Manifest(sources=sources, cases=cases)
|
||||
|
||||
|
||||
def verify_fixtures(manifest: Manifest, cases_root: Path, sources_root: Path) -> None:
|
||||
problems: list[str] = []
|
||||
for case in manifest.cases:
|
||||
d = cases_root / case.name
|
||||
if not d.is_dir():
|
||||
problems.append(f"{case.name}: case directory missing: {d}")
|
||||
continue
|
||||
actual = hash_tree(d)
|
||||
for rel in sorted(set(case.files) - set(actual)):
|
||||
problems.append(f"{case.name}: missing file {rel}")
|
||||
for rel in sorted(set(actual) - set(case.files)):
|
||||
problems.append(f"{case.name}: unexpected file {rel}")
|
||||
for rel in sorted(set(actual) & set(case.files)):
|
||||
if actual[rel] != case.files[rel]:
|
||||
problems.append(f"{case.name}: sha256 mismatch in {rel}")
|
||||
for rel, digest in sorted(manifest.sources.items()):
|
||||
p = sources_root / rel
|
||||
if not p.is_file():
|
||||
problems.append(f"source missing: {p}")
|
||||
elif sha256_file(p) != digest:
|
||||
problems.append(f"source sha256 mismatch: {p}")
|
||||
if problems:
|
||||
raise GateError("fixtures do not match the manifest:\n " + "\n ".join(problems))
|
||||
|
||||
|
||||
def run_auditor(template: list[str], bundle: Path, sources_root: Path, timeout: float) -> Outcome:
|
||||
argv = [a.replace("{bundle}", str(bundle)).replace("{sources}", str(sources_root)) for a in template]
|
||||
try:
|
||||
proc = subprocess.run(argv, capture_output=True, text=True, timeout=timeout, check=False)
|
||||
except FileNotFoundError:
|
||||
return Outcome("error", reason=f"auditor not found: {argv[0]}")
|
||||
except PermissionError:
|
||||
return Outcome("error", reason=f"auditor not executable: {argv[0]}")
|
||||
except subprocess.TimeoutExpired:
|
||||
return Outcome("error", reason=f"auditor timed out after {timeout:g}s")
|
||||
if proc.returncode not in (0, 1):
|
||||
tail = proc.stderr.strip().splitlines()[-1:] or [""]
|
||||
return Outcome("error", reason=f"auditor exited {proc.returncode} {tail[0]}".rstrip())
|
||||
try:
|
||||
payload = json.loads(proc.stdout)
|
||||
except json.JSONDecodeError:
|
||||
return Outcome("error", reason=f"auditor output is not valid JSON (exit {proc.returncode})")
|
||||
findings = payload.get("findings") if isinstance(payload, dict) else None
|
||||
if not isinstance(findings, list) or not all(isinstance(f, dict) and isinstance(f.get("check"), str) for f in findings):
|
||||
return Outcome("error", reason='auditor output lacks a "findings" list of objects with a string "check"')
|
||||
checks = frozenset(str(f["check"]) for f in findings)
|
||||
if proc.returncode == 0 and not findings:
|
||||
return Outcome("clean")
|
||||
if proc.returncode == 1 and findings:
|
||||
return Outcome("alarm", checks=checks)
|
||||
return Outcome("error", reason=f"exit {proc.returncode} disagrees with {len(findings)} finding(s)")
|
||||
|
||||
|
||||
def run_case(case: Case, template: list[str], cases_root: Path, sources_root: Path, timeout: float) -> Outcome:
|
||||
with tempfile.TemporaryDirectory(prefix="mutant-gate-") as tmp:
|
||||
bundle = Path(tmp) / case.name
|
||||
shutil.copytree(cases_root / case.name, bundle)
|
||||
return run_auditor(template, bundle, sources_root, timeout)
|
||||
|
||||
|
||||
def report(manifest: Manifest, outcomes: dict[str, Outcome], command: list[str]) -> bool:
|
||||
mutants = manifest.mutants
|
||||
kp = manifest.known_positive
|
||||
n = len(mutants)
|
||||
print(f"mutant gate: {n} mutants, 1 known-positive; auditor: {' '.join(command)}")
|
||||
for case in mutants:
|
||||
o = outcomes[case.name]
|
||||
if o.status == "alarm":
|
||||
tag = "CAUGHT" if case.cls in o.checks else "CAUGHT (wrong class: " + ", ".join(sorted(o.checks)) + ")"
|
||||
elif o.status == "error":
|
||||
tag = f"MISSED (error: {o.reason})"
|
||||
else:
|
||||
tag = "MISSED (auditor said clean)"
|
||||
print(f" {case.name:<40} [{case.cls}] {tag}")
|
||||
o = outcomes[kp.name]
|
||||
kp_tag = {"clean": "CLEAN", "alarm": "FALSE ALARM (" + ", ".join(sorted(o.checks)) + ")"}.get(
|
||||
o.status, f"ERROR ({o.reason})")
|
||||
print(f" {kp.name:<40} [{KNOWN_POSITIVE}] {kp_tag}")
|
||||
print("mutants per class: " + ", ".join(f"{c}: {k}" for c, k in sorted(Counter(m.cls for m in mutants).items())))
|
||||
caught = [m for m in mutants if outcomes[m.name].status == "alarm"]
|
||||
matched = [m for m in caught if m.cls in outcomes[m.name].checks]
|
||||
false_alarms = int(o.status == "alarm")
|
||||
errors = sum(outcomes[c.name].status == "error" for c in manifest.cases)
|
||||
print(f"caught {len(caught)} of {n} · class matched {len(matched)} of {n} · "
|
||||
f"false alarms {false_alarms} of 1 · errors {errors} of {n + 1}")
|
||||
slipped = [m.name for m in mutants if m not in matched]
|
||||
print("slipped through: " + (", ".join(slipped) if slipped else "none"))
|
||||
green = len(matched) == n and o.status == "clean" and errors == 0
|
||||
print(f"verdict: {'GREEN' if green else 'RED'}")
|
||||
return green
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args_in = sys.argv[1:] if argv is None else argv
|
||||
if "--" in args_in:
|
||||
split = args_in.index("--")
|
||||
own, command = args_in[:split], args_in[split + 1:]
|
||||
else:
|
||||
own, command = args_in, []
|
||||
parser = argparse.ArgumentParser(description="Run the auditor's mutant gate (fail-closed, no model calls).")
|
||||
parser.add_argument("--manifest", type=Path, required=True)
|
||||
parser.add_argument("--cases", type=Path, required=True, help="directory holding one subdirectory per case")
|
||||
parser.add_argument("--sources", type=Path, required=True, help="raw-source root, substituted for {sources}")
|
||||
parser.add_argument("--timeout", type=float, default=60.0, help="seconds per auditor run")
|
||||
args = parser.parse_args(own)
|
||||
if not command:
|
||||
print("usage error: give the auditor command after --, e.g. -- python3 audit.py {bundle} {sources}")
|
||||
return 2
|
||||
try:
|
||||
manifest = load_manifest(args.manifest)
|
||||
verify_fixtures(manifest, args.cases, args.sources)
|
||||
except GateError as exc:
|
||||
print(f"GATE CANNOT RUN: {exc}")
|
||||
return 2
|
||||
outcomes = {c.name: run_case(c, command, args.cases, args.sources, args.timeout) for c in manifest.cases}
|
||||
return 0 if report(manifest, outcomes, command) else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
670
gates/mutants.manifest.json
Normal file
670
gates/mutants.manifest.json
Normal file
|
|
@ -0,0 +1,670 @@
|
|||
{
|
||||
"schema": 1,
|
||||
"description": "E1/E1b/V1/K2 mutant gate. Cases live in gates-local/mutants and their sources in gates-local/mutants-kilder (gitignored, Anthropic-derived); only names, classes and sha256 are tracked.",
|
||||
"provenance": {
|
||||
"sample": {
|
||||
"name": "dontask-modus.md",
|
||||
"sha256": "56ef9ce2fe6ce1df7ef3e47e0abd020c8f339608b47297b384c13600023a7f93"
|
||||
},
|
||||
"normalisation": [
|
||||
"N1: every sources[].resource given a leading / (PLAN.md format choice a).",
|
||||
"N2: three body sentences that shared a footnote with the next sentence given their own footnote (criterion 1 counts sentences).",
|
||||
"N3: the writer's two-sentence meta-statement on the contradictions section removed: it cannot carry a quote.",
|
||||
"N4: the related-concepts line of plain names turned into a list of relative links (format choice b; pages not written yet, SPEC section 6.1).",
|
||||
"index.md (SPEC section 8) and log.md (SPEC section 9) added so section 11.3 can be mutated and the page is not an orphan.",
|
||||
"N5 (E1b): the known-positive is link-clean. Five related-concepts links to pages not written are removed, and the sixth target is written as a second cited page (bypasspermissions-modus.md, 6 of 6 quotes verbatim at their line) that links back and is in index.md and log.md. Chose this over removing all six, because a page that is missing from the index but still linked (so not an orphan) needs a second page; chose it over writing all six, because more pages add Anthropic-derived fixture text and no class. m12's extra file is now a heading only and listed in the index, so it carries no link defect.",
|
||||
"N6 (V1): the known-positive meets the link rule. Each page gets an x-aliases entry that its own sources name, and one cited sentence that links to the other page with a footnote whose x-quote names that page's alias. Chose one link each way over a single link, because the plan's N6 asks it of each page; the m24-m28 mutants therefore break both links. m09 now moves the related-concepts list link, the link it moved before.",
|
||||
"N6 is history (V1 session 2): the source-backed rule it served was replaced on 04.10.2026 by a structure rule (an inbound link from any other concept page), which the operator's rule of 06.10.2026 replaced in turn (N7). The aliases stay. N7 unlinks the N6 link from bypasspermissions-modus.md to dontask-modus.md and keeps the one back.",
|
||||
"N7 (wave 1 order 4, contract K2 of docs/bolge1/0-felles.local.md, the operator's rule of 06.10.2026): the known-positive meets K2. dontask-modus.md is the main page of all three documentation sources (x-main-for lists them; it is the only page that cites all three, and the only one that cites docs-en-agent-sdk-permissions.md, so K2 points 1-2 require it there). Two non-main pages: bypasspermissions-modus.md and a new acceptedits-modus.md (5 of 5 quotes verbatim at their line: four from two of the same documentation sources, one from the release note kilde/release-v1.0.117.md, so a non-main page cites a release note). The main page has exactly one link to each non-main page, so a mutant can move or remove the one link: to bypass from the N6 sentence, to acceptedits from the related-concepts list (which no longer lists bypass). bypass links to acceptedits from a new cited sentence (source pm-554), so a page linked only from the other non-main page takes one edit (m30). The main page has no inbound link from a concept page: the N6 link from bypass is unlinked to plain text, and its log.md entry is plain text (SPEC section 9 allows an entry without a link; log.md never counts, so this only keeps to the order's 'only from index.md'), so it is reached from index.md only and K2 point 3 (a main page needs no inbound link) is tried. Today's auditor reports link-rule on it, the one false alarm, by design, until Step 7 retires link-rule. Every case is rebuilt from this known-positive, so every case's sha256 has changed. The sources moved to their own root in K1 form (provenance.k2.sources_root)."
|
||||
],
|
||||
"link_rules": [
|
||||
"These are okf-wiki's own checks (PLAN.md L1), not SPEC section 11: SPEC section 6.1 says consumers MUST tolerate broken links, and section 11 lists broken cross-links and missing index.md files among what a consumer MUST NOT reject.",
|
||||
"broken-link: a markdown link in a concept body whose target (relative to the file, or bundle-absolute with a leading /) is not a file in the bundle. External URLs and pure #fragments are not bundle links. A link inside a frontmatter value is not a body link (SPEC section 2: the body is everything after the frontmatter); the known-positive has one, inside a verbatim x-quote, and it must not raise an alarm.",
|
||||
"orphan-page: a concept that no other concept body and no index.md links to. log.md does not count: it is history, not navigation. An orphan is therefore always also absent from the index.",
|
||||
"index-coverage: a concept in the directory that its index.md does not list (SPEC section 8: an index enumerates the directory's contents), or an index.md entry whose target is not a file.",
|
||||
"Criterion 6 (the operator's choice 2 of 06.10.2026, verbatim: 'Lenkeregelen, Googles slik den står: hver dokumentasjonsside har sin hovedside, og hver tilleggsside har lenke fra minst én hovedside. Modellen avgjør hvor et begrep hører hjemme, koden sjekker, og en vakt spør på nytt for dem som mangler. Brudd stopper fortsatt.'). Google's rule as written ('verify no reference you minted is orphaned'), applied to documents. The auditor checks structure only, without a model.",
|
||||
"(a) broken-link: a body link to a concept page that does not exist (m19, m20). (b) link-misplaced: a markdown link to a bundle file inside code (an inline span or a fenced block), a heading or a footnote definition of a concept body (m25 code, m26 heading, m27 footnote: each moves the main page's one link to a non-main page).",
|
||||
"The main-page rule (contract K2). A concept page lists the sources it is the main page for in its frontmatter as x-main-for, a YAML list of strings; a valid value is '/' + a path with main_topic true in the K1 source manifest at the sources root, and stands in the page's own sources[].resource. Every main-topic source has exactly one page that lists it validly: main-page-missing (m31), main-page-duplicate (m32). main-page-invalid: not a list (m33), a value not among the page's own sources[].resource (m34), a value in K1 that is not main_topic, here a release note (m35). A page is a main page when at least one value is valid; invalid values do not take that role away (m36: the main page keeps its role beside an invalid value, so a second mark is a duplicate; m37: a page whose every value is invalid is not a main page, so it still needs a link). Without K1 at the sources root, or with an invalid K1, the rule cannot run: exit 2, never 0.",
|
||||
"no-main-page-link: a concept page that is not a main page (a non-main page) has no link from a main page's body outside code, headings and footnote definitions, in a sentence or in a list of related concepts. index.md, log.md and the page itself do not count; a link with #anchor counts for its file; a main page needs no inbound link (the known-positive). m30 (only from the other non-main page), m24 (only from index.md), m29 (only from log.md), m28 (in a concept body, only from itself; index.md and log.md still link it), m37. Under K2 only links from main pages count, so a self-link on a non-main page can never count, and m28 tells K2 auditors apart no better than m24; it stays because the order names the case.",
|
||||
"link-rule (an inbound link from any other concept page, the rule of 04.10.2026) is retired in Step 7, and no mutant carries it any more. Today's auditor still emits it: on the main page of the known-positive (the one false alarm, by design) and of every mutant. The gate counts a mutant matched when its own class is among the findings, so the extra finding hides no match and fakes none.",
|
||||
"Separate classes for (b) and the main-page link: with one class, m25-m27 would also be matched by no-main-page-link, and the gate could not show that check (b) exists. expected_extra on each case lists the link and K2 classes the builder measures beside the case's own class, as the auditor of Step 7 should report them (link-rule left out). It is documentation; the gate checks only the own class."
|
||||
],
|
||||
"k2": {
|
||||
"sources_root": "gates-local/mutants-kilder",
|
||||
"k1_file": "source-manifest.json",
|
||||
"k1_sha256": "b4cb0e24a762b41ca82fc3f9e138ebaf59a5aad250713ca93dad291c578bf0c8",
|
||||
"main_topic": [
|
||||
"kilde/docs-en-agent-sdk-permissions.md",
|
||||
"kilde/docs-en-permission-modes.md",
|
||||
"kilde/docs-en-permissions.md"
|
||||
],
|
||||
"not_main_topic": [
|
||||
"kilde/release-v1.0.117.md"
|
||||
],
|
||||
"known_positive": {
|
||||
"x-main-for": {
|
||||
"dontask-modus.md": [
|
||||
"/kilde/docs-en-agent-sdk-permissions.md",
|
||||
"/kilde/docs-en-permission-modes.md",
|
||||
"/kilde/docs-en-permissions.md"
|
||||
]
|
||||
},
|
||||
"non_main_pages": [
|
||||
"acceptedits-modus.md",
|
||||
"bypasspermissions-modus.md"
|
||||
],
|
||||
"links_from_main_pages": {
|
||||
"acceptedits-modus.md": [
|
||||
"dontask-modus.md"
|
||||
],
|
||||
"bypasspermissions-modus.md": [
|
||||
"dontask-modus.md"
|
||||
]
|
||||
},
|
||||
"concept_links_into_main_pages": {
|
||||
"dontask-modus.md": []
|
||||
},
|
||||
"cites_release_notes": [
|
||||
"acceptedits-modus.md"
|
||||
]
|
||||
},
|
||||
"coverage": {
|
||||
"main-page-missing: the main page loses its mark for one source": "m31-main-page-missing",
|
||||
"main-page-duplicate: a non-main page marks the same source": "m32-main-page-duplicate",
|
||||
"main-page-invalid: the value is not a list": "m33-main-page-invalid-not-a-list",
|
||||
"main-page-invalid: a value that is not among the page's own sources[].resource": "m34-main-page-invalid-not-own-source",
|
||||
"main-page-invalid: a release-note path, in K1 but not main_topic": "m35-main-page-invalid-release-note",
|
||||
"no-main-page-link: linked only from the other non-main page": "m30-no-main-link-only-from-non-main",
|
||||
"no-main-page-link: linked only from index.md": "m24-no-main-link-only-from-index",
|
||||
"no-main-page-link: linked only from log.md": "m29-no-main-link-only-from-log",
|
||||
"no-main-page-link: linked only from itself": "m28-no-main-link-only-self",
|
||||
"link-misplaced: the main page's only link to a non-main page sits in code": "m25-link-misplaced-in-code",
|
||||
"link-misplaced: the main page's only link to a non-main page sits in a heading": "m26-link-misplaced-in-heading",
|
||||
"link-misplaced: the main page's only link to a non-main page sits in a footnote definition": "m27-link-misplaced-in-footnote",
|
||||
"main-page-duplicate: an invalid value does not take the main role from a page's valid values": "m36-main-role-kept-despite-invalid-value",
|
||||
"no-main-page-link: a page whose every value is invalid is not a main page": "m37-only-invalid-marks-not-a-main-page",
|
||||
"broken-link: a link to a concept page that does not exist": "m19-broken-link-list, m20-broken-link-inline",
|
||||
"orphan-page: no link from a concept page or index.md (log.md does not count)": "m21-orphan-page",
|
||||
"a main page needs no inbound link (K2 point 3)": "baseline"
|
||||
}
|
||||
}
|
||||
},
|
||||
"sources": [
|
||||
{
|
||||
"path": "kilde/docs-en-agent-sdk-permissions.md",
|
||||
"sha256": "66c661342bb2326b5e5c4a8ee057a0c74144e9279baeb55b313d234c78ec7647"
|
||||
},
|
||||
{
|
||||
"path": "kilde/docs-en-permission-modes.md",
|
||||
"sha256": "71e1498d768380e3e23eae8d7ce387ef6960b0d1db4d80a47574c50da6015cbd"
|
||||
},
|
||||
{
|
||||
"path": "kilde/docs-en-permissions.md",
|
||||
"sha256": "d3e0f513776bc0a0351489596a341575e25b175d33be4ef6ef834216f792bfcf"
|
||||
},
|
||||
{
|
||||
"path": "kilde/release-v1.0.117.md",
|
||||
"sha256": "324a56574c0974a17658ea578c0aed4e6f16132ae64b70197506d36588fedb2d"
|
||||
},
|
||||
{
|
||||
"path": "source-manifest.json",
|
||||
"sha256": "b4cb0e24a762b41ca82fc3f9e138ebaf59a5aad250713ca93dad291c578bf0c8"
|
||||
}
|
||||
],
|
||||
"cases": [
|
||||
{
|
||||
"name": "baseline",
|
||||
"kind": "known-positive",
|
||||
"class": null,
|
||||
"mutation": "the sample page normalised (N1-N6) and brought under K2 (N7), a second and a third cited page, and a conforming index.md and log.md",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m01-uncited-sentence-paragraph",
|
||||
"kind": "mutant",
|
||||
"class": "uncited-sentence",
|
||||
"mutation": "a sentence with no footnote appended to a cited paragraph",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "e26d885472892d5cae049ca3cbfecd4b630731e72f3b1f27c4973dc64f6a5d10",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m02-uncited-sentence-list-item",
|
||||
"kind": "mutant",
|
||||
"class": "uncited-sentence",
|
||||
"mutation": "a list item with one sentence and no footnote",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "863fd9cf51cb82b056cf4f7618ff484916e86689769f8426209a291767db3972",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m03-quote-not-verbatim",
|
||||
"kind": "mutant",
|
||||
"class": "quote-not-verbatim",
|
||||
"mutation": "x-quote of source pm-493a changed by one word (not on its line)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "e55800e4e7b32be50ab5d79c6cfdba377ce8c843f1f63506ab04bb10e56322d4",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m04-quote-wrong-locator",
|
||||
"kind": "mutant",
|
||||
"class": "quote-wrong-locator",
|
||||
"mutation": "source pm-131: right text, locator moved to the line above",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "e9006b18166f20a2a0def746d29606f3a988eaff9bc28cf3e452a7c7a7c6b89d",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m05-source-missing-id",
|
||||
"kind": "mutant",
|
||||
"class": "source-missing-id",
|
||||
"mutation": "the id line of source sdk-90 removed",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "b2447a18b289c7a6ef27543a5d1fa897850d7c376be76a2aa5eadd5beb6f1c79",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m06-source-duplicate-id",
|
||||
"kind": "mutant",
|
||||
"class": "source-duplicate-id",
|
||||
"mutation": "a second sources entry reusing id pm-493a (its own quote verifies)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "a36af8e46bba4e637759afb77634632d8022ae84084af92c017612eb807a5dfc",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m07-footnote-unknown-id",
|
||||
"kind": "mutant",
|
||||
"class": "footnote-unknown-id",
|
||||
"mutation": "a footnote [^pm-433] with a definition line but no sources entry",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "de307f310b6253b70e0edb3d0afc7359931a4d0a1549faaa6c1e79a639cf999c",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m08-resource-not-bundle-absolute",
|
||||
"kind": "mutant",
|
||||
"class": "resource-not-bundle-absolute",
|
||||
"mutation": "the first sources[].resource naming docs-en-agent-sdk-permissions.md written without the leading / (format choice a)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "4105ea8058ab2f68bddeb93ce0f557bb0ffc3ec0659c069944edc9949bf6bcb0",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m09-body-link-bundle-absolute",
|
||||
"kind": "mutant",
|
||||
"class": "body-link-bundle-absolute",
|
||||
"mutation": "one body link written with a leading / (format choice b): the related-concepts link",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "d805e2bb8f41c36de87363aedd3a03a79e835cc9c38cc424643cc91034e95315",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m10-verified-not-human",
|
||||
"kind": "mutant",
|
||||
"class": "verified-not-human",
|
||||
"mutation": "verified written by a process actor, as an auditor would (format choice e)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "9a34677d1ea78d92c143ef8af3c31927b7d70cc0c63adf56e0a8f2bf4495463e",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m11-spec11-frontmatter-unparseable",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-1-frontmatter",
|
||||
"mutation": "frontmatter YAML no longer parses (unclosed flow list)",
|
||||
"expected_extra": [
|
||||
"main-page-missing",
|
||||
"no-main-page-link"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "d6d48cd2ee1e076faf5470611d6793d1563545c4ba6a45876a2d4c7a87c717a6",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m12-spec11-frontmatter-missing",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-1-frontmatter",
|
||||
"mutation": "a fourth non-reserved .md file with no frontmatter, a heading only and listed in the index (E1b: so it is no orphan and not missing from the index)",
|
||||
"expected_extra": [
|
||||
"no-main-page-link"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "a4b195f67457a9290e352cab1433c83d19b9d3ad4ad5be1873f02e388928c7af",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b",
|
||||
"notater.md": "3f6fca70b29b2c792aeafc30293768fd2c1e9155ded497adb67124ea5625927b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m13-spec11-type-missing",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-2-type",
|
||||
"mutation": "the type key removed",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "35314e80b6e94b07d42db9c9bb3c45d0153f09d073b01ea56e815d0b69b2df45",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m14-spec11-type-empty",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-2-type",
|
||||
"mutation": "the type key present but empty",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "5501bba10c1da38f77a5e86ef9e9cb4d8a00d36749957896250a99bacbf6ab80",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m15-spec11-index-frontmatter",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-3-index",
|
||||
"mutation": "index.md given frontmatter other than okf_version (SPEC section 8)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "235818b7240abcf4132c00593fb63480353e74875151af321aa9bd669f90cf62",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m16-spec11-index-no-heading",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-3-index",
|
||||
"mutation": "index.md with its entries but no section heading (SPEC section 8)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "9979350efead7543b8407740b9664216a4cb2ece41704efc3751076c7e26ac92",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m17-spec11-log-date-form",
|
||||
"kind": "mutant",
|
||||
"class": "spec11-3-log",
|
||||
"mutation": "log.md date heading not in YYYY-MM-DD form (SPEC section 9)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "159cb9b91e3c094b8462f7700731594c7be24fe96506a7405bc768d0c249d25e"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m18-timestamp-form",
|
||||
"kind": "mutant",
|
||||
"class": "timestamp-form",
|
||||
"mutation": "generated.at in the space-separated form a default PyYAML load turns into a datetime (format choice c)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "ab037c1c76fc9599d8dbc370d2f8f16821e1d6f943772e3ad5109f4c1ad9743c",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m19-broken-link-list",
|
||||
"kind": "mutant",
|
||||
"class": "broken-link",
|
||||
"mutation": "a link to a page that does not exist added to the related-concepts list",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "141a82b39078767e0c220aeda7bfb43f6051756ee03b7cc0d4c4cd2ee61184bc",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m20-broken-link-inline",
|
||||
"kind": "mutant",
|
||||
"class": "broken-link",
|
||||
"mutation": "a link to a page that does not exist inside a cited body sentence (the footnote stays)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "556992d5065ea9d7a92162fc0b6f64206984071c71166ed654fb9c3d8facb320",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m21-orphan-page",
|
||||
"kind": "mutant",
|
||||
"class": "orphan-page",
|
||||
"mutation": "acceptEdits-modus no longer linked from any page or index (the main page's related-concepts section, the link from bypassPermissions-modus and its index line removed; its log.md entry stays, because the log is history, not navigation)",
|
||||
"expected_extra": [
|
||||
"index-coverage",
|
||||
"no-main-page-link"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "94fec4bcf9b654f2e8b265401009057a688577e9681f17781f14dc0a1a6b67b1",
|
||||
"dontask-modus.md": "69268d8c456f9109ba880f659ed0db66bc8744a5bc122b9dba54da78ab3e7e6b",
|
||||
"index.md": "9f2b671195854b9d13996c1bd52d8e36c70caa59a08b09e586f9699f9659a362",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m22-index-missing-page",
|
||||
"kind": "mutant",
|
||||
"class": "index-coverage",
|
||||
"mutation": "bypassPermissions-modus exists and is linked from dontAsk-modus, but its index line is removed",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "d1d56579b553c9eb791bdbcdc43bc859eef5f85e89f29b6b4199bfe109fc9a39",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m23-index-line-no-file",
|
||||
"kind": "mutant",
|
||||
"class": "index-coverage",
|
||||
"mutation": "an index line for a page that does not exist",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "a3dc6a1c845863d9cc2c81e44007d5c48c75144dadde02b6988d1c6c0e322acf",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m24-no-main-link-only-from-index",
|
||||
"kind": "mutant",
|
||||
"class": "no-main-page-link",
|
||||
"mutation": "bypassPermissions-modus (non-main) is linked only from index.md: the main page's one link to it (the N6 sentence) unlinked to plain text and its log.md link unlinked",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "9ebcf466579bf06ddc9b7c00cfe7ce2d6f51b9733c42cb37e7e2bf2ad5120202",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "795007610823ec2843477445837f70172ec7736a7fbacb0b14120bb553d1d089"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m25-link-misplaced-in-code",
|
||||
"kind": "mutant",
|
||||
"class": "link-misplaced",
|
||||
"mutation": "the main page's one link to bypassPermissions-modus wrapped in inline code",
|
||||
"expected_extra": [
|
||||
"no-main-page-link"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "5c20e6cda4c9234e3b393c3b8690a7a610a8f7e6c717608246ef231315d88190",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m26-link-misplaced-in-heading",
|
||||
"kind": "mutant",
|
||||
"class": "link-misplaced",
|
||||
"mutation": "the main page's one link to bypassPermissions-modus moved into a heading; the sentence keeps its text and footnote",
|
||||
"expected_extra": [
|
||||
"no-main-page-link"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "93495b766cb7195d6e03da7d6512fb4c097f0bd17d418310360bc754e0fad67f",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m27-link-misplaced-in-footnote",
|
||||
"kind": "mutant",
|
||||
"class": "link-misplaced",
|
||||
"mutation": "the main page's one link to bypassPermissions-modus moved into a footnote definition; the sentence keeps its text and footnote",
|
||||
"expected_extra": [
|
||||
"no-main-page-link"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "5d8610879bb1722c63151cac41643cece1ee61350e1c996313799f17eacd59b8",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m28-no-main-link-only-self",
|
||||
"kind": "mutant",
|
||||
"class": "no-main-page-link",
|
||||
"mutation": "the main page's one link to bypassPermissions-modus unlinked; the only link to it left in a concept body is a self-link on its own page (index.md and log.md still link it)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "24be674fd6bf49b55b1b576c929efac130c6625df3955fb6baa5a4bcb0092aa9",
|
||||
"dontask-modus.md": "9ebcf466579bf06ddc9b7c00cfe7ce2d6f51b9733c42cb37e7e2bf2ad5120202",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m29-no-main-link-only-from-log",
|
||||
"kind": "mutant",
|
||||
"class": "no-main-page-link",
|
||||
"mutation": "bypassPermissions-modus is linked only from log.md: the main page's one link to it unlinked and its index line removed",
|
||||
"expected_extra": [
|
||||
"index-coverage",
|
||||
"orphan-page"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "9ebcf466579bf06ddc9b7c00cfe7ce2d6f51b9733c42cb37e7e2bf2ad5120202",
|
||||
"index.md": "d1d56579b553c9eb791bdbcdc43bc859eef5f85e89f29b6b4199bfe109fc9a39",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m30-no-main-link-only-from-non-main",
|
||||
"kind": "mutant",
|
||||
"class": "no-main-page-link",
|
||||
"mutation": "acceptEdits-modus is linked only from the other non-main page (bypassPermissions-modus), index.md and log.md: the main page's related-concepts section, its one link to it, removed",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "69268d8c456f9109ba880f659ed0db66bc8744a5bc122b9dba54da78ab3e7e6b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m31-main-page-missing",
|
||||
"kind": "mutant",
|
||||
"class": "main-page-missing",
|
||||
"mutation": "the main page loses its mark for one source: /kilde/docs-en-permissions.md removed from its x-main-for",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "b8b208220f289339116665f4b80d6e407c2b4ef32b78c77eaa19992436c26aad",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m32-main-page-duplicate",
|
||||
"kind": "mutant",
|
||||
"class": "main-page-duplicate",
|
||||
"mutation": "a non-main page marks a source the main page already marks: bypassPermissions-modus gets x-main-for [/kilde/docs-en-permission-modes.md], which it cites",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "53f251de532fdd1e013460ed1a18a7bfa985b7144c9a63e83b879b01ce466458",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m33-main-page-invalid-not-a-list",
|
||||
"kind": "mutant",
|
||||
"class": "main-page-invalid",
|
||||
"mutation": "x-main-for is not a list: bypassPermissions-modus gets the string /kilde/docs-en-permission-modes.md, which it cites (read as a list, it would be a second main page)",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "d349f931cdcb4c2a59d5d9508b80765f1771088c5874409afa98ab6070b22df8",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m34-main-page-invalid-not-own-source",
|
||||
"kind": "mutant",
|
||||
"class": "main-page-invalid",
|
||||
"mutation": "an x-main-for value that is not among the page's own sources[].resource: bypassPermissions-modus marks /kilde/docs-en-permissions.md, a main-topic source it does not cite",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "6dd007076d921df5a2e522bdb836e626043726a1664e6ecc3b07516cb40c04a5",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m35-main-page-invalid-release-note",
|
||||
"kind": "mutant",
|
||||
"class": "main-page-invalid",
|
||||
"mutation": "an x-main-for value that is in K1 but not main_topic: acceptEdits-modus marks the release note it cites, /kilde/release-v1.0.117.md",
|
||||
"expected_extra": [],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "667540a5a9be917a421c5fed7a6ddb491ce0712e6006509e9f577269aa3295da",
|
||||
"bypasspermissions-modus.md": "210c44745fa51ed430ab1b201fe99b6063225427a656ed6e49568cd9db2801e6",
|
||||
"dontask-modus.md": "22215ba30352895d80bd01f4dca40c547dc9ff0f9ecfa299848df2842c4eb59b",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m36-main-role-kept-despite-invalid-value",
|
||||
"kind": "mutant",
|
||||
"class": "main-page-duplicate",
|
||||
"mutation": "the main page keeps its role beside an invalid value: dontAsk-modus adds /kilde/release-v1.0.117.md (not among its own sources, not main_topic) to its x-main-for, and bypassPermissions-modus marks /kilde/docs-en-permission-modes.md, which it cites; an auditor that strips the role for one invalid value sees no duplicate",
|
||||
"expected_extra": [
|
||||
"main-page-invalid"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "53f251de532fdd1e013460ed1a18a7bfa985b7144c9a63e83b879b01ce466458",
|
||||
"dontask-modus.md": "b39ce8a5c1f5b8254f6f216aad1675ec745d3cb5df5c93c9307da6478553acad",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "m37-only-invalid-marks-not-a-main-page",
|
||||
"kind": "mutant",
|
||||
"class": "no-main-page-link",
|
||||
"mutation": "a page whose every value is invalid is not a main page: bypassPermissions-modus marks /kilde/docs-en-permissions.md (a main-topic source it does not cite), and the main page's one link to it is unlinked; an auditor that makes every page with the key a main page sees no missing link",
|
||||
"expected_extra": [
|
||||
"main-page-invalid"
|
||||
],
|
||||
"files": {
|
||||
"acceptedits-modus.md": "e8d9a420a91b0980dbfc0b06feb9c2e5eec62fb047a4ea7e37bd1a63a08b3182",
|
||||
"bypasspermissions-modus.md": "6dd007076d921df5a2e522bdb836e626043726a1664e6ecc3b07516cb40c04a5",
|
||||
"dontask-modus.md": "9ebcf466579bf06ddc9b7c00cfe7ce2d6f51b9733c42cb37e7e2bf2ad5120202",
|
||||
"index.md": "fd080296ad4a460afc83c7a4a590067e558942d684747feb539125d25e5584e4",
|
||||
"log.md": "8176033370edbe2d3b4c431dd28e62abe2a3c3256996fb4dcf21c183e16b7b3b"
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
170
gates/polarity.py
Normal file
170
gates/polarity.py
Normal file
|
|
@ -0,0 +1,170 @@
|
|||
"""Polarity keys (E2d): counts the polarity keys that accept an answer of the opposite polarity. No model call.
|
||||
|
||||
A polarity key is an answer key whose fact is carried by a negation or a restriction in its source ("not synced
|
||||
back" became the key `synced`). Which keys are polarity keys is given per set part as a list:
|
||||
|
||||
{"schema": "okf-wiki/polarity-keys/1", "keys": [{"id": "<question id>", "atoms": ["<key text>", ...]}]}
|
||||
|
||||
For each listed key the tool builds two answers mechanically from the key's own spans (the spans of the key's file;
|
||||
in a whole-document set, every span). The answer from the span is the first sentence where the first atom has a
|
||||
negation within NEGATION_WINDOW words before it, or else the first sentence that holds it. The opposite answer is
|
||||
that sentence with its polarity flipped: the negations within the window are taken out ("isn't" -> "is", "can't"
|
||||
-> "can", "without" -> "with", "no longer" and every other negation dropped), or, when there is none, "not" is put
|
||||
in front of the atom. A key accepts an answer when the gates' own key_holds holds for every atom.
|
||||
|
||||
python3 gates/polarity.py --set <set.json> --keys <keys.json> [--set <set.json> --keys <keys.json> ...] [--list]
|
||||
|
||||
Prints "polarity keys that accept the answer from their span: Y of N" and "polarity keys that accept the opposite
|
||||
answer: X of N". Exit 0 = PASS (N > 0, Y = N and X = 0); 1 = FAIL or could not run.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import sys
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from multidoc import (
|
||||
NEGATION_WINDOW,
|
||||
RunError,
|
||||
as_list,
|
||||
is_negation,
|
||||
key_holds,
|
||||
key_positions,
|
||||
load_json,
|
||||
negated_at,
|
||||
normalise_answer,
|
||||
sentences,
|
||||
)
|
||||
|
||||
KEYS_SCHEMA = "okf-wiki/polarity-keys/1"
|
||||
|
||||
|
||||
def unnegate(word: str) -> str:
|
||||
"""The word without its negation: "isn't" -> "is", "can't" and "cannot" -> "can", "won't" -> "will",
|
||||
"without" -> "with"; any other negation -> "" (dropped)."""
|
||||
w = word.replace("’", "'")
|
||||
special = {"cannot": "can", "can't": "can", "won't": "will", "shan't": "shall", "without": "with"}
|
||||
if w in special:
|
||||
return special[w]
|
||||
return w[: -len("n't")] if w.endswith("n't") else ""
|
||||
|
||||
|
||||
def flip(norm: str, pos: int) -> str:
|
||||
"""The normalised sentence with the polarity of the atom at pos flipped (see the module text)."""
|
||||
before = norm[:pos].split()
|
||||
cut = max(0, len(before) - NEGATION_WINDOW)
|
||||
head, window = before[:cut], before[cut:]
|
||||
if not any(is_negation(w) for w in window):
|
||||
return norm[:pos] + "not " + norm[pos:]
|
||||
kept: list[str] = []
|
||||
dropped_no = False
|
||||
for w in window:
|
||||
if dropped_no and w == "longer":
|
||||
dropped_no = False
|
||||
continue
|
||||
dropped_no = False
|
||||
if is_negation(w):
|
||||
base = unnegate(w)
|
||||
dropped_no = w == "no"
|
||||
if base:
|
||||
kept.append(base)
|
||||
else:
|
||||
kept.append(w)
|
||||
return " ".join(head + kept + [norm[pos:]])
|
||||
|
||||
|
||||
def answer_from_span(atom: str, spans: list[str]) -> tuple[str, int] | None:
|
||||
"""(normalised sentence, atom offset): the first sentence with the atom negated in the window, else the first
|
||||
sentence that holds the atom; None when no span holds it."""
|
||||
found: list[tuple[str, int]] = []
|
||||
for span in spans:
|
||||
for s in sentences(span):
|
||||
norm = normalise_answer(s)
|
||||
found += [(norm, p) for p in key_positions(atom, norm)]
|
||||
negated = [f for f in found if negated_at(*f)]
|
||||
chosen = negated or found
|
||||
return chosen[0] if chosen else None
|
||||
|
||||
|
||||
@dataclass
|
||||
class PolarityReport:
|
||||
keys: int = 0
|
||||
accept_span: int = 0
|
||||
accept_opposite: int = 0
|
||||
listed: list[str] = field(default_factory=list)
|
||||
|
||||
@property
|
||||
def passed(self) -> bool:
|
||||
return self.keys > 0 and self.accept_span == self.keys and self.accept_opposite == 0
|
||||
|
||||
|
||||
def count_polarity(questions: list[dict[str, Any]], entries: list[Any], label: str = "") -> PolarityReport:
|
||||
rep = PolarityReport()
|
||||
by_id = {q.get("id"): q for q in questions if isinstance(q, dict)}
|
||||
for e in entries:
|
||||
qid, atoms = (e.get("id"), e.get("atoms")) if isinstance(e, dict) else (None, None)
|
||||
q = by_id.get(qid)
|
||||
if q is None or not isinstance(atoms, list) or not atoms:
|
||||
raise RunError(f"{label}: listed key {e!r} names no question or no atoms")
|
||||
keys: list[dict[str, Any]] = []
|
||||
for atom in atoms:
|
||||
match = [k for k in as_list(q.get("answer_keys")) if isinstance(k, dict) and k.get("text") == atom]
|
||||
if len(match) != 1:
|
||||
raise RunError(f"{label} {qid}: atom {atom!r} is not exactly one answer key of the question")
|
||||
keys.append(match[0])
|
||||
name = keys[0].get("file", q.get("file"))
|
||||
spans = [ev["span"] for ev in as_list(q.get("evidence"))
|
||||
if isinstance(ev, dict) and isinstance(ev.get("span"), str) and ev.get("file", q.get("file")) == name]
|
||||
found = answer_from_span(atoms[0], spans)
|
||||
if found is None:
|
||||
raise RunError(f"{label} {qid}: atom {atoms[0]!r} stands in none of its spans")
|
||||
norm, pos = found
|
||||
opposite = flip(norm, pos)
|
||||
on_span = all(key_holds(k, norm) for k in keys)
|
||||
on_opposite = all(key_holds(k, opposite) for k in keys)
|
||||
rep.keys += 1
|
||||
rep.accept_span += on_span
|
||||
rep.accept_opposite += on_opposite
|
||||
rep.listed.append(f"{label} {qid} {atoms}: span {'accepted' if on_span else 'REFUSED'}, "
|
||||
f"opposite {'ACCEPTED' if on_opposite else 'refused'} | {opposite}")
|
||||
return rep
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
p.add_argument("--set", action="append", required=True)
|
||||
p.add_argument("--keys", action="append", required=True, help="the polarity keys of the set given at that place")
|
||||
p.add_argument("--list", action="store_true", help="list every key with its opposite answer")
|
||||
args = p.parse_args(argv)
|
||||
total = PolarityReport()
|
||||
try:
|
||||
if len(args.set) != len(args.keys):
|
||||
raise RunError("give one --keys per --set")
|
||||
for set_path, keys_path in zip(args.set, args.keys):
|
||||
data = load_json(Path(set_path), "set")
|
||||
listing = load_json(Path(keys_path), "polarity keys")
|
||||
if not isinstance(data, dict) or not isinstance(data.get("questions"), list):
|
||||
raise RunError(f"{set_path} is not a set")
|
||||
if not isinstance(listing, dict) or listing.get("schema") != KEYS_SCHEMA:
|
||||
raise RunError(f"{keys_path} is not a polarity key list with schema {KEYS_SCHEMA}")
|
||||
rep = count_polarity(data["questions"], as_list(listing.get("keys")), Path(set_path).name)
|
||||
total.keys += rep.keys
|
||||
total.accept_span += rep.accept_span
|
||||
total.accept_opposite += rep.accept_opposite
|
||||
total.listed += rep.listed
|
||||
except RunError as exc:
|
||||
print(f"FAIL: {exc}", file=sys.stderr)
|
||||
return 1
|
||||
if args.list:
|
||||
print("\n".join(total.listed))
|
||||
print(f"polarity keys that accept the answer from their span: {total.accept_span} of {total.keys}")
|
||||
print(f"polarity keys that accept the opposite answer: {total.accept_opposite} of {total.keys}")
|
||||
print("PASS" if total.passed else "FAIL")
|
||||
return 0 if total.passed else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
216
gates/quote_rule.py
Normal file
216
gates/quote_rule.py
Normal file
|
|
@ -0,0 +1,216 @@
|
|||
"""The quote rule for K's fragments, measured without a model (V1 Step 4, brief SC3, R5).
|
||||
|
||||
P1's compiler wrote numbers as words to slip past a rule that asked for a quote only when the text held a digit or a
|
||||
backticked name. Two rules close that path, and this module measures both on P1's fragments:
|
||||
|
||||
- Rule E: every fragment carries a quote.
|
||||
- Rule L: a fragment carries a quote when its text holds a digit, a backticked name, or a number word from the NARROW
|
||||
list (zero to twenty, hundred, thousand, million).
|
||||
|
||||
The answer key is made apart from rule L: a quoteless fragment whose text holds a word from the BROAD list (brief SC3).
|
||||
Its denominator is the number of hits. The key must find at least one fragment (the known-positive: P1 saw the compiler
|
||||
do this), or the list fails and the run exits 1.
|
||||
|
||||
E is chosen unless L lets 0 key fragments through AND E rejects at least 5 % of the fragments more than L does (33 of
|
||||
650); then L. E is the simplest rule and closes the path outright; L pays only if E forces noticeably more rewrites.
|
||||
|
||||
Two new reject rules are measured next to it, each adopted in Step 6 only if it rejects at most 5 % of the fragments:
|
||||
- short quote: a quote of fewer than 3 words;
|
||||
- id mismatch: an "existing" fragment whose index entry's title and aliases match its concept name neither by
|
||||
normalised form nor by Jaccard >= MERGE_JACCARD (or whose id is not in the index).
|
||||
|
||||
NEEDS_QUOTE, MERGE_JACCARD and norm are copies of the trial's (evals/p1/kpack.py), kept equal by a test that reads the
|
||||
trial at run time. They are copied, not imported, so that `mypy --strict okfwiki gates tests` never follows trial code.
|
||||
|
||||
measure FRAGMENTS_DIR INDEX
|
||||
Every fragment in FRAGMENTS_DIR/*.json ({"fragments": [...]}) against both rules and both reject rules.
|
||||
Prints one JSON object of counts only (never fragment text): the counts, the answer-key hits per file
|
||||
(to find the known-positive pack) and the choice.
|
||||
|
||||
Exit 0 = measured and a rule chosen, 1 = the answer key found no fragment (no choice), 2 = could not run.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import math
|
||||
import re
|
||||
import sys
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
NEEDS_QUOTE = re.compile(r"\d|`")
|
||||
MERGE_JACCARD = 0.6
|
||||
NARROW = ("zero", "one", "two", "three", "four", "five", "six", "seven", "eight", "nine", "ten", "eleven", "twelve",
|
||||
"thirteen", "fourteen", "fifteen", "sixteen", "seventeen", "eighteen", "nineteen", "twenty",
|
||||
"hundred", "thousand", "million")
|
||||
BROAD = NARROW + ("thirty", "forty", "fifty", "sixty", "seventy", "eighty", "ninety",
|
||||
"dozen", "dozens", "once", "twice",
|
||||
"first", "second", "third", "fourth", "fifth", "sixth", "seventh", "eighth", "ninth", "tenth",
|
||||
"half", "double", "pair", "couple", "several", "single")
|
||||
SHORT_QUOTE_WORDS = 3
|
||||
|
||||
|
||||
def _word_pattern(words: tuple[str, ...]) -> re.Pattern[str]:
|
||||
return re.compile(r"(?<![\w-])(" + "|".join(sorted(words, key=len, reverse=True)) + r")(?![\w-])", re.IGNORECASE)
|
||||
|
||||
|
||||
NARROW_WORD = _word_pattern(NARROW)
|
||||
BROAD_WORD = _word_pattern(BROAD)
|
||||
|
||||
|
||||
def norm(name: str) -> str:
|
||||
text = name.replace("`", "").lower()
|
||||
text = re.sub(r"[^\w\s/.\-]", " ", text)
|
||||
words = text.split()
|
||||
if words and words[0] == "the":
|
||||
words = words[1:]
|
||||
words = [w[:-1] if len(w) > 3 and w.endswith("s") and not w.endswith("ss") else w for w in words]
|
||||
return " ".join(w.strip(".") or w for w in words)
|
||||
|
||||
|
||||
def _text(fragment: dict[str, Any]) -> str:
|
||||
return str(fragment.get("text", ""))
|
||||
|
||||
|
||||
def has_quote(fragment: dict[str, Any]) -> bool:
|
||||
q = fragment.get("quote")
|
||||
return isinstance(q, str) and bool(q.strip())
|
||||
|
||||
|
||||
def rule_e(fragment: dict[str, Any]) -> bool:
|
||||
"""True when rule E rejects the fragment: it has no quote."""
|
||||
return not has_quote(fragment)
|
||||
|
||||
|
||||
def rule_l(fragment: dict[str, Any]) -> bool:
|
||||
"""True when rule L rejects the fragment: no quote, and a digit, a backticked name or a NARROW number word."""
|
||||
text = _text(fragment)
|
||||
return not has_quote(fragment) and bool(NEEDS_QUOTE.search(text) or NARROW_WORD.search(text))
|
||||
|
||||
|
||||
def number_words(fragment: dict[str, Any]) -> list[str]:
|
||||
"""The answer key: the BROAD-list words in a quoteless fragment's text, lower-cased and sorted ([] when quoted)."""
|
||||
if has_quote(fragment):
|
||||
return []
|
||||
return sorted({m.group(1).lower() for m in BROAD_WORD.finditer(_text(fragment))})
|
||||
|
||||
|
||||
def short_quote(fragment: dict[str, Any], min_words: int = SHORT_QUOTE_WORDS) -> bool:
|
||||
return has_quote(fragment) and len(str(fragment["quote"]).split()) < min_words
|
||||
|
||||
|
||||
def _jaccard(a: str, b: str) -> float:
|
||||
x, y = set(a.split()), set(b.split())
|
||||
return len(x & y) / len(x | y) if x | y else 0.0
|
||||
|
||||
|
||||
def id_mismatch(fragment: dict[str, Any], index: dict[str, Any]) -> bool:
|
||||
if fragment.get("status") != "existing":
|
||||
return False
|
||||
entry = next((e for e in index.get("entries", []) if e.get("id") == fragment.get("id")), None)
|
||||
if entry is None:
|
||||
return True
|
||||
names = {norm(str(n)) for n in [entry.get("title", ""), *entry.get("aliases", [])]} - {""}
|
||||
concept = norm(str(fragment.get("concept", "")))
|
||||
return concept not in names and not any(_jaccard(concept, n) >= MERGE_JACCARD for n in names)
|
||||
|
||||
|
||||
def five_percent(n: int) -> int:
|
||||
return math.ceil(0.05 * n)
|
||||
|
||||
|
||||
def choose(e_rejects: int, l_rejects: int, l_slips: int, n: int) -> str:
|
||||
return "L" if l_slips == 0 and e_rejects - l_rejects >= five_percent(n) else "E"
|
||||
|
||||
|
||||
class RunError(Exception):
|
||||
"""The measurement cannot run as asked."""
|
||||
|
||||
|
||||
def load_fragments(directory: Path) -> list[tuple[str, dict[str, Any]]]:
|
||||
"""(file name, fragment) for every fragment in directory/*.json."""
|
||||
files = sorted(directory.glob("*.json"))
|
||||
out: list[tuple[str, dict[str, Any]]] = []
|
||||
for f in files:
|
||||
try:
|
||||
data = json.loads(f.read_text(encoding="utf-8"))
|
||||
except (OSError, json.JSONDecodeError) as exc:
|
||||
raise RunError(f"cannot read {f}: {exc}") from exc
|
||||
items = data.get("fragments") if isinstance(data, dict) else None
|
||||
if not isinstance(items, list):
|
||||
raise RunError(f"{f} has no fragments list")
|
||||
out.extend((f.name, x) for x in items if isinstance(x, dict))
|
||||
if not out:
|
||||
raise RunError(f"no fragments under {directory}")
|
||||
return out
|
||||
|
||||
|
||||
def measure(located: list[tuple[str, dict[str, Any]]], index: dict[str, Any]) -> dict[str, Any]:
|
||||
fragments = [f for _, f in located]
|
||||
n = len(fragments)
|
||||
per_file: dict[str, int] = {}
|
||||
for name, f in located:
|
||||
if number_words(f):
|
||||
per_file[name] = per_file.get(name, 0) + 1
|
||||
key = [f for f in fragments if number_words(f)]
|
||||
e_rejects = sum(rule_e(f) for f in fragments)
|
||||
l_rejects = sum(rule_l(f) for f in fragments)
|
||||
short = sum(short_quote(f) for f in fragments)
|
||||
mismatch = sum(id_mismatch(f, index) for f in fragments)
|
||||
words: dict[str, int] = {}
|
||||
for f in key:
|
||||
for w in number_words(f):
|
||||
words[w] = words.get(w, 0) + 1
|
||||
result: dict[str, Any] = {
|
||||
"schema": "okf-wiki/quote-rule/1",
|
||||
"fragments": n,
|
||||
"quoteless": e_rejects,
|
||||
"e_rejects": e_rejects,
|
||||
"l_rejects": l_rejects,
|
||||
"key_hits": len(key),
|
||||
"key_hits_narrow": sum(bool(NARROW_WORD.search(_text(f))) for f in key),
|
||||
"key_words": dict(sorted(words.items())),
|
||||
"key_hits_per_file": dict(sorted(per_file.items())),
|
||||
"e_slips": sum(not rule_e(f) for f in key),
|
||||
"l_slips": sum(not rule_l(f) for f in key),
|
||||
"five_percent": five_percent(n),
|
||||
"short_quotes": short,
|
||||
"short_quote_adopt": short * 20 <= n,
|
||||
"id_mismatches": mismatch,
|
||||
"id_mismatch_adopt": mismatch * 20 <= n,
|
||||
"choice": None,
|
||||
}
|
||||
if key:
|
||||
result["choice"] = choose(e_rejects, l_rejects, result["l_slips"], n)
|
||||
return result
|
||||
|
||||
|
||||
def build_parser() -> argparse.ArgumentParser:
|
||||
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
sub = p.add_subparsers(dest="command", required=True)
|
||||
m = sub.add_parser("measure", help="measure rule E, rule L and the reject rules on a fragments directory")
|
||||
m.add_argument("fragments", type=Path)
|
||||
m.add_argument("index", type=Path)
|
||||
return p
|
||||
|
||||
|
||||
def main(argv: list[str] | None = None) -> int:
|
||||
args = build_parser().parse_args(argv)
|
||||
try:
|
||||
fragments = load_fragments(args.fragments)
|
||||
try:
|
||||
index = json.loads(args.index.read_text(encoding="utf-8"))
|
||||
except (OSError, json.JSONDecodeError) as exc:
|
||||
raise RunError(f"cannot read {args.index}: {exc}") from exc
|
||||
except RunError as exc:
|
||||
print(f"cannot run: {exc}", file=sys.stderr)
|
||||
return 2
|
||||
result = measure(fragments, index if isinstance(index, dict) else {})
|
||||
print(json.dumps(result, indent=1))
|
||||
return 0 if result["choice"] else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
sys.exit(main())
|
||||
2087
gates/repair.manifest.json
Normal file
2087
gates/repair.manifest.json
Normal file
File diff suppressed because it is too large
Load diff
7
gates/repair/agent-shared.md
Normal file
7
gates/repair/agent-shared.md
Normal file
|
|
@ -0,0 +1,7 @@
|
|||
---
|
||||
name: repair-gate-agent-shared
|
||||
description: An agent file that repeats a line of a source; task must refuse it.
|
||||
tools: Read, Write
|
||||
---
|
||||
|
||||
Read the parts your prompt names. For example: a frame holds wax comb where bees store honey and raise brood.
|
||||
8
gates/repair/agent.md
Normal file
8
gates/repair/agent.md
Normal file
|
|
@ -0,0 +1,8 @@
|
|||
---
|
||||
name: repair-gate-agent
|
||||
description: A thin agent file for the repair gate's task cases. It names no subject of its own.
|
||||
tools: Read, Write
|
||||
---
|
||||
|
||||
Read the parts your prompt names and nothing else. Follow the instructions in them, write one JSON file to the path
|
||||
your prompt names, and hand back only the word done.
|
||||
21
gates/repair/fragments/pack-1.json
Normal file
21
gates/repair/fragments/pack-1.json
Normal file
|
|
@ -0,0 +1,21 @@
|
|||
{"fragments": [
|
||||
{"concept": "Hive", "id": "hive", "status": "existing", "text": "A hive is the box where a bee colony lives.", "anchor": {"file": "docs-en-hive.md", "start": 3, "end": 4}, "quote": "the box where a bee colony lives", "provenance": "documented"},
|
||||
{"concept": "Hive", "id": "hive", "status": "existing", "text": "Open the hive only on a warm and calm day during an inspection.", "anchor": {"file": "docs-en-hive.md", "start": 14, "end": 17}, "quote": "only on a warm and calm day", "context": ["Inspections"], "provenance": "documented"},
|
||||
{"concept": "Hive", "id": "hive", "status": "existing", "text": "Light the smoker before you lift the lid.", "anchor": {"file": "docs-en-hive.md", "start": 14, "end": 17}, "quote": "Light the smoker before you lift the lid", "provenance": "documented"},
|
||||
{"concept": "Hive", "id": "hive", "status": "existing", "text": "Watch for signs that the colony will swarm.", "anchor": {"file": "docs-en-hive.md", "start": 14, "end": 17}, "quote": "signs that the colony is getting ready to swarm", "provenance": "inferred"},
|
||||
{"concept": "Frames", "id": "frames", "status": "existing", "text": "Frames can be lifted out one by one.", "anchor": {"file": "docs-en-hive.md", "start": 8, "end": 10}, "quote": "lifted out one by one", "provenance": "documented"},
|
||||
{"concept": "Frames", "id": "frames", "status": "existing", "text": "Each frame holds comb for honey and brood.", "anchor": {"file": "docs-en-hive.md", "start": 8, "end": 10}, "quote": "Each frame holds wax comb", "provenance": "documented"},
|
||||
{"concept": "Frames", "id": "frames", "status": "existing", "text": "Bees store honey and raise brood in the comb of a frame.", "anchor": {"file": "docs-en-hive.md", "start": 8, "end": 10}, "quote": "holds wax comb where bees store honey and raise brood", "provenance": "documented"},
|
||||
{"concept": "Brood", "id": "brood", "status": "existing", "text": "Brood is raised in the wax comb of the frames.", "anchor": {"file": "docs-en-hive.md", "start": 8, "end": 10}, "quote": "holds wax comb where bees store honey and raise brood", "provenance": "documented"},
|
||||
{"concept": "Brood", "id": "brood", "status": "existing", "text": "Young larvae are fed for about six days and then sealed for about twelve days.", "anchor": {"file": "docs-en-hive.md", "start": 21, "end": 23}, "quote": "feed the young larvae for about six days", "provenance": "merged"},
|
||||
{"concept": "Brood", "id": "brood", "status": "existing", "text": "The queen lays one egg in each empty cell.", "anchor": {"file": "docs-en-hive.md", "start": 21, "end": 23}, "quote": "The queen lays one egg in each empty cell", "provenance": "documented"},
|
||||
{"concept": "Smoker", "id": "smoker", "status": "existing", "text": "The smoker makes cool smoke that calms the bees.", "anchor": {"file": "docs-en-hive.md", "start": 27, "end": 27}, "quote": "makes cool smoke that calms the bees", "provenance": "documented"},
|
||||
{"concept": "Water", "id": "water", "status": "existing", "text": "Bees need a water source near the hive.", "anchor": {"file": "docs-en-hive.md", "start": 31, "end": 31}, "quote": "need a water source near the hive", "provenance": "documented"},
|
||||
{"concept": "Water", "id": "water", "status": "existing", "text": "Bees need a water source near the hive.", "anchor": {"file": "docs-en-honey.md", "start": 12, "end": 12}, "quote": "need a water source near the hive", "provenance": "documented"},
|
||||
{"concept": "Honey harvest", "id": "honey-harvest", "status": "existing", "text": "Harvest honey when most cells are capped with wax.", "anchor": {"file": "docs-en-honey.md", "start": 3, "end": 3}, "quote": "when most cells in the frame are capped with wax", "provenance": "documented"},
|
||||
{"concept": "Swarming", "id": "swarming", "status": "existing", "text": "A swarm is a large group of bees that leaves with the old queen.", "anchor": {"file": "docs-en-swarm.md", "start": 3, "end": 3}, "quote": "leaves the hive with the old queen", "provenance": "documented"},
|
||||
{"concept": "Queen cells", "id": "queen-cells", "status": "existing", "text": "Workers build queen cells along the lower edge of the comb. A queen cell looks like a peanut shell.", "anchor": {"file": "docs-en-swarm.md", "start": 7, "end": 9}, "quote": "build queen cells along the lower edge of the comb", "context": ["looks like a peanut shell"], "provenance": "documented"},
|
||||
{"concept": "Queen cells", "id": "queen-cells", "status": "existing", "text": "Remove queen cells early to keep the colony together.", "anchor": {"file": "docs-en-swarm.md", "start": 7, "end": 9}, "quote": "Remove queen cells early", "provenance": "documented"}
|
||||
],
|
||||
"main_pages": [{"source": "docs-en-hive.md", "concept": "Hive"}, {"source": "docs-en-honey.md", "concept": "Honey harvest"}, {"source": "docs-en-swarm.md", "concept": "Swarming"}],
|
||||
"homes": [{"concept": "Frames", "home": "hive"}, {"concept": "Brood", "home": "hive"}, {"concept": "Smoker", "home": "hive"}, {"concept": "Water", "home": "hive"}, {"concept": "Queen cells", "home": "swarming"}]}
|
||||
4
gates/repair/fragments/pack-2.json
Normal file
4
gates/repair/fragments/pack-2.json
Normal file
|
|
@ -0,0 +1,4 @@
|
|||
{"fragments": [
|
||||
{"concept": "Hive", "id": "hive", "status": "existing", "text": "The hive guide gained a part on lighting the smoker.", "anchor": {"file": "release-v2.1.0.md", "start": 3, "end": 3}, "quote": "how to light the smoker", "provenance": "documented"},
|
||||
{"concept": "Honey harvest", "id": "honey-harvest", "status": "existing", "text": "The honey guide gained a part on the extractor.", "anchor": {"file": "release-v2.1.0.md", "start": 4, "end": 4}, "quote": "a part on the spinning extractor", "provenance": "documented"}
|
||||
], "main_pages": [], "homes": []}
|
||||
31
gates/repair/kilder/kilde/docs-en-hive.md
Normal file
31
gates/repair/kilder/kilde/docs-en-hive.md
Normal file
|
|
@ -0,0 +1,31 @@
|
|||
# Hive
|
||||
|
||||
A hive is the box where a bee colony lives and stores its honey.
|
||||
Most hives stand on a low stand a little above the ground.
|
||||
|
||||
## Frames
|
||||
|
||||
Frames can be lifted out one by one during an inspection.
|
||||
Each frame holds wax comb where bees store honey and raise brood.
|
||||
A frame that is full of capped honey is heavy.
|
||||
|
||||
## Inspections
|
||||
|
||||
Open the hive only on a warm and calm day.
|
||||
Move slowly so the bees stay calm.
|
||||
Light the smoker before you lift the lid.
|
||||
Look for signs that the colony is getting ready to swarm.
|
||||
|
||||
## Brood
|
||||
|
||||
The queen lays one egg in each empty cell.
|
||||
Worker bees feed the young larvae for about six days.
|
||||
Capped brood stays sealed for about twelve days.
|
||||
|
||||
## Smoker
|
||||
|
||||
The smoker makes cool smoke that calms the bees.
|
||||
|
||||
## Water
|
||||
|
||||
Bees need a water source near the hive.
|
||||
12
gates/repair/kilder/kilde/docs-en-honey.md
Normal file
12
gates/repair/kilder/kilde/docs-en-honey.md
Normal file
|
|
@ -0,0 +1,12 @@
|
|||
# Honey harvest
|
||||
|
||||
Harvest honey when most cells in the frame are capped with wax.
|
||||
|
||||
## Equipment
|
||||
|
||||
An uncapping knife cuts the wax caps off the comb.
|
||||
A spinning extractor throws the honey out of the comb.
|
||||
|
||||
## Water
|
||||
|
||||
Bees need a water source near the hive.
|
||||
9
gates/repair/kilder/kilde/docs-en-swarm.md
Normal file
9
gates/repair/kilder/kilde/docs-en-swarm.md
Normal file
|
|
@ -0,0 +1,9 @@
|
|||
# Swarming
|
||||
|
||||
A swarm is a large group of bees that leaves the hive with the old queen.
|
||||
|
||||
## Queen cells
|
||||
|
||||
Before a swarm leaves, the workers build queen cells along the lower edge of the comb.
|
||||
A queen cell looks like a peanut shell hanging from the comb.
|
||||
Remove queen cells early if you want to keep the colony together.
|
||||
4
gates/repair/kilder/kilde/release-v2.1.0.md
Normal file
4
gates/repair/kilder/kilde/release-v2.1.0.md
Normal file
|
|
@ -0,0 +1,4 @@
|
|||
## 2.1.0
|
||||
|
||||
- The hive guide now explains how to light the smoker.
|
||||
- The honey guide gained a part on the spinning extractor.
|
||||
45
gates/repair/kilder/source-manifest.json
Normal file
45
gates/repair/kilder/source-manifest.json
Normal file
|
|
@ -0,0 +1,45 @@
|
|||
{
|
||||
"collection": "beekeeping",
|
||||
"frozen_at": "2026-01-01T00:00:00Z",
|
||||
"schema": "okf-wiki/source-manifest/1",
|
||||
"sources": [
|
||||
{
|
||||
"bytes": 779,
|
||||
"kind": "docs",
|
||||
"lines": 31,
|
||||
"main_topic": true,
|
||||
"path": "kilde/docs-en-hive.md",
|
||||
"sha256": "5ba8883c65cb0828c87dc80fe75b4dd3497603d9a0409ad709b7262bdb919025",
|
||||
"url": "https://example.org/bees/hive"
|
||||
},
|
||||
{
|
||||
"bytes": 253,
|
||||
"kind": "docs",
|
||||
"lines": 12,
|
||||
"main_topic": true,
|
||||
"path": "kilde/docs-en-honey.md",
|
||||
"sha256": "6c5e68d0748c2f7974a55400aca4937abc29988a6ce7623143683cb9053197b1",
|
||||
"url": "https://example.org/bees/honey"
|
||||
},
|
||||
{
|
||||
"bytes": 318,
|
||||
"kind": "docs",
|
||||
"lines": 9,
|
||||
"main_topic": true,
|
||||
"path": "kilde/docs-en-swarm.md",
|
||||
"sha256": "646fdf8dd7b2ea4c33fc521d9a8ea9883ec6c0b2e89ababa85e10ffd9aee2c1a",
|
||||
"url": "https://example.org/bees/swarm"
|
||||
},
|
||||
{
|
||||
"bytes": 124,
|
||||
"kind": "release-notes",
|
||||
"lines": 4,
|
||||
"main_topic": false,
|
||||
"path": "kilde/release-v2.1.0.md",
|
||||
"sha256": "388caad1b376e3e37c27772296ac411b55c2503c69a3935bfb168e6255c57a26",
|
||||
"url": "https://example.org/bees/changelog#2.1.0",
|
||||
"version": "2.1.0"
|
||||
}
|
||||
],
|
||||
"subset_of": null
|
||||
}
|
||||
Some files were not shown because too many files have changed in this diff Show more
Loading…
Add table
Add a link
Reference in a new issue