K3-21 D, red. The NISO-STS reader clips `depth` to 6 once and writes that
one value into BOTH the ATX heading, where markdown has six levels, and the
`OutlineMark` the declared route builds its plan from, where nothing clips.
On one standard 9 of 2 761 titled sections sit at depth 7; the plan read them
at 6, and `--shell-parent` gave its two depth-7 shells the ancestor one level
too high.
`sts-deep.xml` (invented setting) reproduces it: the two depth-7 shells point
at the depth-5 section on e717b1c's code and must point at the depth-6 one.
Held: the mark carries depth 7; the heading stays `######` and the extracted
headings do not move one character; the shells point at their depth-6
ancestor.
2 of 3 red on 5970369; the green one is the text guard.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Red. A process code states its lettered points once, on the section that
owns them, and every section nested below inherits them; built faithfully,
the nested section is a concept whose body is one heading line (710 of 2 761
on one measured standard), and the two-level directory tree does not carry
the parent either.
Behind `--shell-parent`, off by default: a plan entry whose span holds only
its heading gets `parent_id` naming the nearest preceding entry at a smaller
level whose own span holds text. An empty ancestor is passed over; a shell
with no ancestor holding text gets no parent. The rule reads the plan's level
and order, never the row, so the same outline through the bookmark arm's
route reaches the same parents. The fixture is hand-written.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Red first (K3-19 a). A NISO-STS delivery built with `okf build` put every
concept under a directory named for the delivery path's file name -- a UUID
that occurs 0 times in the document -- and every `sources` entry named that
file twice, while the document carries exactly one <std-ident> with a
<doc-number> and one <title-wrap>, neither of which this package read.
11 of the 14 tests are red: `extract.declared_identity` does not exist, and
the build still names the directory and the address title from the file.
The 3 that pass today pin the fallback layer that must survive the change: no
declaration keeps the file name, a declared title carrying a flow terminator
falls to the file name, and a name another document's file already holds is
not taken.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RED. Nine sections declared by `sts-container.xml`, a hand-written fixture that
carries no sentence from any source: Arm B delivers 8 (the orphan check takes
the container chapter) and the shipped build defaults deliver 3 (Arm F folds
every level below the shallowest repeated one). That is 2 761 -> 2 051 -> 23 on
R761 in miniature, measured.
`find_candidates` already skips both steps for `outline_marks`, which is why
the PDF bookmark arm reaches 2 762 of 2 761. These tests pin the same route for
the one row whose reader wrote the heading itself, under its OWN rule name, and
pin that no other row reaches it: the same markdown from a `.md` file is a
guess and keeps `rule:heading`.
7 failed, 1 passed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RED. Measured on R761 after the reader landed: non-whitespace preservation was
1 283 393 characters against the source's 1 283 395. Two characters, and a
percentage would have let them pass -- a ratio of 0.999998 reads like rounding.
Located by a prefix/suffix scan rather than a diff: the whole loss is one `x)`
marking a `<sec>` that carries a label and nothing else.
The mechanism is this reader's own. A label-only section holds its label as a
PREFIX for the body line beneath it, the way `li` is handled in the HTML
reader; when no body line follows -- the next thing emitted is the following
section's heading -- the pending prefix is overwritten by the next one and the
label never reaches the text.
`sts-empty-label.xml` is the case, hand-written and small: a lettered point
with a body, then one with none, then the next titled section.
pytest -q: 1 failed, 1566 passed, 1 skipped.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RED, 11 failing. Reproduced first, before any code: `okf build` on the folder
a publisher's own viewer delivers gives **110 of 110 unreadable, 0 plans, exit
2**, with `no extractor is registered for file extension '.xml'`. The
conservation identity `merged + coded rejections == N` is never written at all
-- the run aborts earlier on `FAILED - no segmentation plans` -- so the hole
was not even visible as a count.
The one xml file IS the whole product: R761 Prosesskoden:2025, the document
round 12 met as a 701-page PDF, in NISO-STS form. Measured on the file with
`xml.etree.ElementTree`: 7 715 `<sec>`, 2 761 with a `<title>`, 4 954 with a
`<label>` and no title, 10 `<table-wrap>`, 12 528 `<p>`, root `<standard>`, 0
`<!DOCTYPE` and 0 `<!ENTITY`. Its `<sec>`-nesting depths over the titled
sections are 28/118/500/1141/868/97/9 -- row for row the fasit's own
distribution. The ceiling is therefore structural rather than computed.
FOUR HAND-WRITTEN FIXTURES, none through `make_fixtures.py` and none
serialised by `ElementTree`: a library that writes and reads its own format
proves only that it agrees with itself. A known-positive STS mini, generic
non-STS xml, a `<!DOCTYPE` with an entity expansion, and a malformed file.
Two assertions that already existed are extended rather than duplicated: the
converter fence, because a file routed to the converter is read by a second
parser that never sees this reader's DTD refusal, and the evidence table,
because a row without a class is the failure that test exists for.
pytest -q: 11 failed, 1554 passed, 1 skipped.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RED. Measured on R761: 2 763 /Outlines nodes entered the bridge, 2 762 marks
came out, and `unresolved` was 0 -- so one node left no trace in the return
value at all. The lost one is `SVV - Forside`, sharing line 0 with the tree's
own root node `R761 Prosesskoden`. The cause is structural, not an edge value:
the marks are collected in a dict keyed on the destination line index, so a
second bookmark on a line is discarded by `setdefault` in silence. One node
today; a document with several bookmarks per line loses several and reports
none of them.
`outline-collision.pdf` is the fixture no existing one could stand in for --
every other outline fixture puts one bookmark on one line, which is exactly
the case where the defect cannot appear. It is laid out by hand through
`make_fixtures.py` like the other PDF fixtures, and regenerating that file
left every other fixture byte-identical.
The known-negative is in the same commit: `outline-collided.pdf` aside, the
two existing trees must report `collided == 0`, so the counter can be measured
at zero and is not green for an unstated reason.
pytest -q: 2 failed, 1554 passed, 1 skipped.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`okf build` recovers a PDF's boundaries from the shape of its page text and
never opens the file's own `/Outlines` bookmark tree. On a 701-page process
code whose publisher also ships a NISO-STS structure for it, measured outside
this repository and reproduced here exactly: the shipped default finds 1967 of
2761 titled sections, 0 of its 28 chapters, and 794 of 794 misses have their
heading text PRESENT in the extracted text. The line was read; the boundary
was never opened. The same file's bookmark tree matches 2761 of 2761 of those
titles exactly after normalisation.
`--pdf-outline`, OFF, cuts a PDF at the boundaries its tree declares.
boundaries 1967 of 2761 -> 2759 of 2761 (gate was 2700)
depth 1 0 of 28 -> 28 of 28
titles identical to source -- -> 2761 of 2761
false positives 163 of 2182 -> 3 of 2762
directories with two files 132 of 2050 -> 2 of 2738
front-matter concepts 72 of 2182 -> 2 of 2762
consumption fasit present 4 of 7 -> 7 of 7
hit@1 / hit@8 / hit@50 1/6 2/6 4/6 -> 3/6 5/6 6/6
It is a SEGMENTATION arm, not a reader option: the extracted text is byte for
byte the same either way. A PDF with no tree builds byte-identically with the
flag on -- `diff -r` empty across the pre-change tree, the arm off and the arm
on. An unresolvable `/Dest` is dropped and COUNTED, never fabricated into a
boundary and never a refusal of the file.
The bridge from (page, y) to a line index is the whole risk, so both routes
are measured. `extract_text_lines` splits lines identically to `extract_text`
on 701 of 701 pages, and is CHECKED per page rather than assumed. The y route
and the title route disagree on 0 of 2762 nodes, flat from a 0pt tolerance to
8pt and collapsing at 12pt, so the rule ships with no tolerance constant. The
naive "nearest line" rule was wrong on 1840 of 2762, one line early every time.
The orphan check is not applied to a bookmark mark: it asks whether anything
stands under a candidate's first line, which is the right question for a
heuristic's guess and the wrong one for a publisher's declaration. 683 of 2762
marks are container sections; applying it scores 2079 instead of 2759.
No new dependency and no second parse of the pages: `pdfminer.six` already
ships under `pdfplumber` in `[extract]`. 119.22s -> 183.31s wall, peak RSS
3252 -> 3251 MiB. The default does not move; 1 of the 8 reference PDFs carries
a usable tree at all.
`.pdf` also gains its `_EVIDENCE` row, as `measured` -- it was the row with the
most measurement behind it and no entry in the table.
Report: docs/2026-09-10-k3-runde12-pdf-outlines.md
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three of round 9's four measured holes, each closed with a rule chosen on a
measurement rather than named as a limit.
`rtf` GIVES 0 SEGMENTS -> 6 of 6 AUTHORED TITLES over N = 4. The container has
no heading style, so the author's title is bold text. The grammar is markdown,
not `rtf`: the converter already writes that title as `**...**` in the same
output every office row produces, so no `rtf`-only heading form exists. Three
parameters were swept over 47 readable documents and ONE carried -- refusing a
line that ends in terminal punctuation takes false-positive lines from 9-12 to
1-2. A maximum title length (unlimited/40/60/80/120) and a
must-stand-between-blank-lines clause are both FLAT, so neither is in the rule.
The last false positive is closed by G1, the principle `_gate_outline` already
carries: recovery yields to declaration. False positives are then 0 of the 31
declaring documents by construction, and 0 of 27 on the corpus. Reach: 2 of 39
corpus documents, both `docx`, 0 of 33 `pdf` and 0 of 2 `xlsx`. Behind
`--bold-title`, default OFF pending the hit@8 measurement; the default bundle
is byte-identical without it.
BOTH ALTERNATIVES THE ORDER NAMED WERE MEASURED AND FELLED. A fourth hand-laid
fixture DECLARES heading styles in a stylesheet and the converter discards
them, emitting the same bold line -- so "read the declared headings out of the
markdown" has nothing to read. `rtf` -> `docx` -> markdown yields 0 ATX
headings on that same document, because the loss is in the `rtf` READER before
any writer sees the style. Fixtures are hand-laid in `make_k2_office.py` with
the fasit written first; they live in their own directory because Door B walks
a drop directory recursively and `k2-office/` reads its N off the listing.
THE PREFIX OVER-MATCH: THREE CANDIDATES MEASURED, ALL THREE FAILED ON ONE ROW.
Re-measured on the pinned 453-concept bundle with the control run first:
`under` occurs 79 times by equality and matches 172 by prefix, `undersjoisk` 0
and 172, `bilateral` 0 and 400 of 453, `standhaftig` 0 and 219. The two extra
known-negatives were FOUND, not chosen -- every 4-character prefix ranked by
document frequency, then a real word taken from the widest. A longer floor
(5-8), a coverage share (0.5-0.8) and a long-words-only floor (>= 8) each cost
row 1 its rank on the default bundle and the whole row on Arm B. Decomposed:
row 1's token `prisene` reaches its gold document through
`pris|sammenstilling` on four characters -- 0.57 of one word and 0.22 of the
other -- so the over-match and the wanted match are one mechanism.
THE FOURTH CANDIDATE IS THE ANSWER: the shared prefix must be a WORD the bundle
uses. `pris` is; `bila` and `stan` are not. `bilateral` 400 -> 0 and 512 -> 0,
`standhaftig` 219 -> 56 and 235 -> 33, every hit@8 row keeping rank 1 on BOTH
bundles. `undersjoisk` stops at 162 because `under` IS a word here -- a genuine
Norwegian morpheme, so that residual is a different answer, not a ceiling. ON
by default (`--no-stem-prefix`), pinned with its own known-negative on the
shipped bytes.
THE SHIM: a path importer holds the object `module_from_spec` made, and
`sys.modules[__name__] = _impl` never reaches it. Measured under both counting
methods -- 3 of 76 public names by `vars()`. One line copies the public names
into this file's globals; the dunder filter is load-bearing, because an
unfiltered copy overwrites `__name__` before the next line uses it as the alias
key. It restores attribute ACCESS and not patch-through, which is why the alias
stays. A CHANGELOG note under 0.7.0 and a shim docstring line say so, since
what the consumer asked for was the note.
Suite 1515 -> 1535.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
K3 round 5. Three questions, three answers, and two of them correct a figure
this repository published.
RETRIEVAL FIRST, because it could have reversed a default. hit@8 over the six
questions on BOTH K2 bundles -- Arm B at 629 concepts and the shipped default
at 492 -- is 5 of 6 with ranks 1,1,1,1,1,- on each, so 0 of 6 rows lost. The
order's rule reverses `--unit-fold` at >= 2 of 6; it does not fire, and the
default stands. The gold sets shrink (49->26, 20->17, 43->36, 11->18) while
every rank holds at 1, which is the fold merging concepts rather than removing
a document from the top.
TWO PUBLISHED NUMBERS CORRECTED, both ours. The S7 candidate ranks 96 of 629
and 159 of 492 were measured with the cost vocabulary passed to
`concept_scores` and NOT to `document_scores`, while `build_payload` passes it
to both; scored the way the shipped payload scores it, the same concept is 10
of 629 and 19 of 492. And round 4 attributed its non-delivery to the default
move -- measured here, it is not delivered on the Arm B bundle either, for a
different reason (knapsack eviction at 68 046 bytes of a 120 000 budget, versus
`below_k`). That column had been inherited from round 3's own build, never
re-measured.
`--pdf-headings font-reserve`, OFF, and the hypothesis behind it is falsified
by its own condition rather than by a score: position 7, the one position the
flag exists for, has THREE outline runs, so the reserve is silent there at
every minimum. It changes 0 of 12 cells on the reference and reaches 4 of 39
corpus documents, none of them rated. Built anyway because it was authorised
and because the condition is now measured rather than assumed. The predicate
lives in one place (`propose.heading_reserve_applies`) and the door receives it
as a callable, like `gate`: a plan indexes the exact string it was proposed
against, so a reserve firing on one side only would make every document it
touches a coded rejection.
The `xlsx` re-reading is confirmed on the artifact -- 11 `rule:sheet-section`
units plus 1 `rule:table-block` ingress -- but the number alone makes the cell
worse (distance 1 -> 2), because the criterion counts that ingress as a table
that should have been merged. A hit needs both halves ratified, and the
reference is the operator's.
`--sheet-section-rows` as a default: three cells better and none worse on the
twelve positions, but the K2 control moves -- row 1's gold document splits 1 ->
12 concepts and its best concept ranks 2 instead of 1. Condition not met,
default not moved.
Default build byte-identical before and after (`diff -r`, 30 md files).
Suite 1441 -> 1449; three of the eight were red first.
Report: docs/2026-09-08-k3-runde5-hitat8-og-skriftakse.md
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A PDF carries no notion of a heading -- a heading in a PDF is a typographic
fact -- so the text stream `pdfplumber` hands the segment proposer has already
thrown away the only evidence there was. The `docx` path never had that problem:
the converter emits ATX headings and `_ATX` cuts on them. Two readers close the
gap, and both are OFF.
`--pdf-headings font` infers a heading from the conjunction this repository
already measured (size above the document's character-weighted body median AND
a bold font name, recall 1.000 / precision 0.846) and emits it as ATX in the
SAME markdown the office path produces, so `_ATX` applies unchanged and no
PDF-only heading grammar exists.
It stays off BY MEASUREMENT, and the measurement is the point of the round:
against the operator's unit worksheet it takes `pdf` from 2 of 8 to 0 of 8,
losing two exact matches. The mechanism of the loss is stated rather than
guessed -- on those documents the outline rule already recovers the document's
own numbered chapters, so a second heading source can only add. Whole-corpus
screen: 25 of 32 `pdf` change, 0 of 5 `docx`, 0 of 2 `xlsx`. The default bundle
is byte-identical before and after this commit (`diff -r`, exit 0).
`--ocr` reads a page as an image when its own text never arrived: empty, or
`(cid:N)` placeholder codes at or above a threshold READ OFF a measured
distribution -- 834 pages over 32 files, 818 at exactly 0.0 and 16 at 0.93 or
above, nothing in between. On the one corpus document with the failure: 95.07 %
cid to 0 %, 44 to 2561 words of four or more letters, 17 to 18 pages with text.
Its engine is an optional dependency group and never a runtime dependency; a
packaging test pins both halves, and without the group every affected file is a
coded rejection (`extractor_ocr_group_missing`) rather than a crash.
Also corrects two stale published facts found while measuring: the README still
said two segmentation rules were on by default after `f6fea13` made it three,
and CLAUDE.md's K2 digest named the round-3 default. The current default is
492 concepts / 944 files, `bdefa679...`.
Report: docs/2026-09-08-k3-runde4-pdf-skrift-og-ocr.md
Co-Authored-By: Claude <claude-opus-5>
The pre-pass delivered the right concept and the answer could not name it.
Measured by portfolio-optimiser 2026-09-08 over three paid arms: the gold
concept came back at rank 1 of 8 on 3 of 3 bundles, and the model answered
correctly on 1 of 3, because a delivered excerpt carried `concept_id`, body
text and nothing the document is known by. The previous session measured the
same gap from the other side: the provenance it had just written into every
concept did not reach the payload at all.
`excerpt_for` now carries `title` unconditionally, and `req_number`, the SPEC
5.1 address `sources` and each locator key (`source_pages`, `source_sheet`,
`source_rows`, `source_lines`, `source_offset`) when the concept has them. A key
the producer did not write stays absent: an empty value would assert that they
wrote an empty one, which is the contract's 6.4 failure.
`sources` is read in BOTH YAML forms, on a measurement rather than a taste. K2
writes the flow form on 629 of 629 concepts; the largest N-bundle writes the
block form on 270 of 270 and carries no locator key at all, so a flow-only
reader delivers that bundle with no address whatsoever. Reading the block form
is not a licence to write it - the emission rule is untouched, because the
line-oriented parser still cannot round-trip a block list. A `sources` value
this reader cannot decode is named (`sources_unreadable`), never dropped into
the same silence as an absent one.
Contract 8 gains the requirement and the checker gains its code
(`excerpt_unnamed`, 15 rules now, was 14): an excerpt a reader cannot name is
one an answer cannot cite, whatever its rank. `req_number`, `sources` and the
locators are SHOULD, not MUST - they are conditional on the producer, and a
bundle whose concepts carry no identifier cannot deliver one.
K2 controls, same question and same k, before against a frozen copy of the tool
at b6a8c8b: the RANKING does not move - the same 8 concept ids in the same
order, identical `text_sha256`, identical `withheld`, identical denominators
(629 = 621 + 8). The FIELD is what moved: payload 108 877 -> 111 744 B
(+2.63 %), budget spent 18 606 -> 20 907 (+287.6 B per excerpt), excerpt
members 9 -> 15, 83 changed lines. The contract document's own bytes moved with
8, so the budget instrument's known-positive moves with it: 10 349 -> 12 049
measured, 10 060 -> 11 719 raw, delta 289 -> 330.
New fixture `tests/fixtures/consume-provenance`: the two address forms and a
concept carrying neither address nor identifier. Purpose-built, because the two
real bundles are complementary and neither exercises both forms.
Suite 1347 (1339 before), ruff clean, mypy src clean.
Co-Authored-By: Claude <claude-opus-5>
A concept named its source file by basename and, when segmented, carried a
`source_offset` into the text THIS LIBRARY extracted. Following that pointer
needed the corpus directory, the extractor and its exact transitive version --
none of which the bundle carries. Hand-walked on a real K2 concept: six steps,
four of them requiring knowledge from outside the bundle, to learn that a
requirement sits on pages 12-13 of a 20-page document.
The address is spec's: `sources: [{ resource, title }]`, where `resource` is
the dropped file's inbox-relative path (SPEC v0.2 5.1:303-306 -- "an absolute
URL, a bundle-relative path, or a path into a `references/` subdirectory").
The locator is ours, and it has to be: 5.1 has no field for a place within a
resource, and the pinned guard (1.3.0) rejects every route to putting one
inside a `sources` entry -- a non-allowlisted key by name, a nested flow list
as "scalar leaves only", and quoting as an unsupported form. So the locator is
top-level keys shaped like `source_offset`, and a path carrying a flow
terminator is refused fail-fast rather than mangled.
The unit table is built AT EXTRACTION, where the extracted text and the
original's structure are known to agree: pdf -> `source_pages` from
pdfplumber's own page numbers (a page that yielded no text does not renumber
the ones after it), xlsx -> `source_sheet` + `source_rows`, everything else ->
`source_lines`. `source_offset` stays.
Two measurements changed the design before it shipped. A `paragraphs` key for
docx would name a number the document does not have: `<w:p>` counts of
108/27/65/176/57 against converted-markdown lines of 75/33/67/144/63, not one
pair agreeing -- so the key is `source_lines` and says what it indexes. And an
empty spreadsheet row renders exactly like a table separator: the content-based
rule ate 8 empty rows on the K2 price sheet and reported its last row as 92
against a workbook that says 100. The separator is now found by position, and
`tomrad.xlsx` keeps that red.
One profile moves. `provenance` is a policy object, `None` everywhere but
`SEGMENTED_OKF_V0_2`; the other five shipped profiles are byte-identical.
K2 rebuilt from a frozen src copy: 629 concepts, 1108 files, name set identical,
0 ids moved, 479 files byte-identical, 629 changed and 0 lines removed anywhere.
629/629 now carry an address and a locator. New ref
`sha256-tree:665563a2f74423fcbcc8e4f0b0954ee73b73985ac0418de4f6987bd162a1f7c8`;
`2f82fcfe...` is stale. The pre-pass payload does not grow by one byte
(209 092 B before and after, 18 changed lines: the ref and eight per-concept
digests) -- because an excerpt carries the body, not the frontmatter, which is
also why the consumer still cannot cite "file X page 12" from a payload alone.
Report: docs/2026-09-08-proveniens-k2.md. 1339 tests, ruff and mypy clean.
Co-Authored-By: Claude <claude-opus-5>
Two producer-side findings from the consumer's S7c acid test (ordre 20260908T063454Z-3648220855-from-.claude), both measured on K2 before and after, both with the corpus rebuilt from scratch.
FUNN 3 -- THE FORM. The converter's default markdown writer emits simple tables, which pad every cell out to the width of the widest cell in its column. Measured on the tender's price sheet: one 594-character prose cell produced a 67 244-character whitespace carpet with runs of up to 887 characters between a label and its amount, 19 integral amounts carrying a converter `.0`, and a header row naming one column. The bytes reached a live model in 2 of 11 prompts and 0 of 11 answers. The spreadsheet row now writes pipe tables with `--columns=1` (load-bearing: the pipe writer pads to a width computed from it, so at the default 72 a narrow table gains runs of up to 45). Same sheet after: 11 048 characters, longest run 2, one row per line, 0 artificial `.0`. Spreadsheet-only, and the scoping is pinned by three digests -- the same change moves the odt fixture 1366 -> 1105, so it can fail.
The `.0` rewrite is bounded twice: to a cell whose whole content is such a number, anchored between unescaped pipes, and skipped when the literal is in the workbook's shared string table -- the converter renders the number 92 and the TEXT "92.0" identically, so the output alone cannot tell them apart. Read with zipfile and xml.etree; no new dependency.
FUNN 2 -- THE LOG LINK. `link_log_in_root_index` (95eb271) is removed. Consumption contract SS 9.2 forbids a consumer from enumerating the bundle directory unless the profile says the index is derived, so the index tree is the entire map a consumer may use and everything it links is a document: their navigator returned 630 where our pre-pass counts 629, and a corpus run's own log was citable as content. The log is still written to the bundle root (SPEC section 9); `tools/okf_consume.py` keeps its exclusion for the bundles already built with the link.
K2 rebuilt twice. BEFORE reproduces the consumer's ref exactly (`sha256-tree:f14872a0...c8a92a`, 629 concepts) and their three consume figures to the token (57 289 / 62 149 / 58 401). AFTER: 629 concepts, `merged + coded rejections = 43 = N`, new ref `sha256-tree:c26eed6a...e3261f`, 627 of 629 concepts byte-identical, 1104 of 1108 files identical to the delivered bundle.
ONE REGRESSION, MEASURED AND NOT FIXED: on the mandate-shaped question with the vocabulary bridge the priced concept moves from candidate rank 10 to 19, so `--k 12` withholds it `below_k`; `--cost-vocabulary --k 20` delivers it at 65 912 o200k. The cause is measured rather than argued -- restoring only the concept's title on the new short body ranks it 10 again. The chain ends at the orphan check (`propose.py:461`), which drops the sheet heading once a table block opens two lines below it. That is the already-reported orphan gate, and changing it is a default-ON segmentation rule affecting every document type. The specific question is unaffected: rank 1 before and after. The priced excerpt's budget share falls from 56.5 % to 9.7 %.
11 new tests (RED first), 8 mutations, 8 red, with an unmutated control green each time. One mutation survived twice before the fixture could make it fire, and both survivals are written down. 1279 -> 1287 tests. mypy --strict clean on 28 files. ruff clean. Both proposer goldens byte-unchanged. One frozen literal moved with the fix and is reported rather than hidden.
Report: docs/2026-09-08-prisform-og-loggen-k2.md
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The K3 Arm E round (order 20260907T075834Z-18584396-from-.claude) measures a
rule that has not been written yet. This commit pins what the code does TODAY,
so the later claim "the default did not move" rests on a committed artifact
rather than on reading a diff.
Three characterization tests and one new golden fixture. Every count they
assert was measured against unmodified code before the test was written:
GRID_TABLE yields 3 candidates, TWO_TABLES_BLANK_SEPARATED yields 3,
TWO_TABLES_RULE_SEPARATED 2, PIPE_TABLE 1, PROSE_WITH_A_STRAY_RULE 1.
The second golden is not duplication. `propose-golden-default.json` is taken
over `OUTLINE_DOCUMENT`, which contains no `|` row and no `+` rule line, so no
table rule -- present or future -- can move its bytes. It is a guard that
cannot fire. `propose-golden-grid-default.json` is taken over a document that
has a pandoc grid table, and it is generated here from PRE-Arm-E code: a
fixture recorded after the change would pin post-change bytes and prove
nothing about the default.
Two fixture shapes are load bearing and say so in the band's comment. The
second table in TWO_TABLES_BLANK_SEPARATED has exactly ONE row group, so it
cannot be joined -- that is the only shape that catches a join whose pending
state is never cleared. And `propose()` is excluded from this band: it pops
`source` and discards every other kwarg, so an Arm E test written through it
would silently measure the default and pass.
Tests first: 1 red (the golden did not exist), then green. 1230 -> 1233.
ruff check: exit 0. ruff format --check: exit 0. pytest -q: exit 0.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
hit@8 = 5 of 6, every hit at rank 1, against a chance baseline of 1.35 of 6
over a denominator of 629 concepts per question. Wall time 0.51-0.56 s per
question; spent 17 970 - 74 838 bytes against a 120 000 limit.
Two things this measurement did NOT establish, both in the report:
- BOTH known-negative controls FAILED. A question the bundle has no answer to
still returns eight excerpts, because no natural Norwegian question is
lexically disjoint from a 629-concept corpus under a four-character
shared-prefix rule -- measured per token, the interrogative `hvor` reaches 40
concepts, `brukes` 83. So `no_lexical_match` works per concept and not as a
whole-question gate: an empty excerpt list is evidence of absence, a full one
is not evidence of presence. The fix is named (rarity weighting) and NOT
built, because this step's fence freezes the instrument before it is measured.
- The question texts were written during execution, after the ranker existed.
The plan recorded the gold documents' SIZE profile -- its per-row baselines
sum to 1.35 and the sizes used here reproduce that exactly, which is an
independent check that this is the set the plan profiled -- but it recorded no
question texts, and three of six gold documents could not be pinned uniquely
from the sizes. Not a blind evaluation, and the report says so.
The scorer is a tool, not a script in a document: `tools/okf_consume_measure.py`
takes the gold set as an INPUT because it is tracked in a public repository and
an answer key names a consumer's documents. hit_rank, both chance baselines and
the document-size census are unit-tested; the corpus run is a measurement.
Public-file rule, checked with a pattern DERIVED from the corpus's own 39
document names rather than hand-picked, and shown able to find first (67 hits on
the bundle's own index): zero corpus document names in any tracked file in this
repository. One leak was found and removed on the way -- a corpus concept name
in a code comment and a hardcoded corpus path in a test.
Suite run after git add: 1230 passed, mypy --strict clean on 27 files,
ruff clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`docs/2026-09-04-k2-pptx-odt-rtf.md` measured the corpus denominator for
these three office rows and found it ZERO: `K2/trinn1` holds 43 files and
not one of them is a `pptx`, an `odt` or an `rtf`. So `extract._EVIDENCE`
calls those rows `unmeasured` in the strongest sense available -- they
work by construction and had never met a document at all.
This is the smallest thing that changes that without inventing a corpus.
One authored document -- a title, an intro, a 20-row label/value table, a
caption and a 4x4 grid -- laid out three times in three containers, so
the container and its reader are the only variable between the three
measurements. `k2-office-fasit.json` carries the hand count taken from
the AUTHORED content rather than from any converter's output: 56 cells,
20 pairs, 59 distinct strings, shared by all three. It is committed here,
before the measurement runs, because a fasit written afterwards is a
description of a result rather than a denominator for it.
No converter wrote any of these files. `make_k2_office.py` lays every
part by hand, for the reason `make_fixtures.py` already states and this
set inherits: a file written by the converter and then read by the
converter proves only that the converter agrees with itself, and stays
green through any conversion defect that is symmetric. The commissioning
order offered pandoc as one generator option; the committed fixture
policy forbids it and the policy wins.
Two converter behaviours were measured while laying the RTF out, both of
them structurally plausible input read silently wrong, exit code 0 and no
warning. Without `\pard\intbl` on cell paragraphs, consecutive
`\trowd...\row` rows come back as each row NESTED inside the previous
one: five label/value rows read as five levels of nested table, 2076
characters where 117 were expected. And the `\uN?` unicode escape -- the
form Word emits -- loses the character after it: `A\u248?BC` reads back
as `AoC` with the `B` gone, `A\u248?xBC` reads back as `AoBC`. The
fixture writes `\uN ?` with an explicit space, which round-trips. Neither
is worked around anywhere in `src/`.
The generator and the fasit live one level above `k2-office/` and that is
not tidiness: Door B walks its drop directory recursively, so anything
parked beside the three documents would enter the run and N would stop
being 3.
Three synthetic documents in one house style are not a corpus. The rows
stay `unmeasured` and the suite asserts that they do.
Suite 1132 passed (1127 + 5), `ruff check` and `ruff format --check`
clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three hand-laid OOXML containers, every part written out by hand and zipped
with a fixed date_time so they are byte-reproducible. No converter output
anywhere in them: a .docx written by the converter and read by the converter
proves only that the converter agrees with itself, and would stay green through
any conversion defect that is symmetric -- which is most of them.
two-line-krav.docx heading + label/value on one line (the docx mirror of
the PDF fixture)
no-styles-krav.docx the SAME document without word/styles.xml
two-line-krav.xlsx sheet name as heading + label/value on one row
THE FIXTURES FOUND A REAL DEFECT IN THE SEAM THEY WERE MEANT TO PIN. The
converter call used pypandoc's TEXT entry point, which takes an `encoding`
because it treats its source as text -- and that corrupts a zip. The xlsx
fixture failed with `Failed to unpack XLSX archive: not enough bytes` while
reading correctly from disk with the same binary. The docx of the same shape
happened to survive, which is the part worth writing down: the defect is silent
for some inputs and fatal for others, so "it worked on the file I tried" was
never evidence. Input now goes through a temporary file.
Two measurements while building, both the same shape -- structurally valid
input, silently reduced output, exit code 0, no warning:
- Without word/styles.xml the docx extracts as flat prose with no heading. A
fixture lacking that part would pin the body and pin nothing about structure.
Committed as a negative control that RUNS rather than a sentence in a README.
- With inline strings rather than a shared string table, the xlsx extracts with
the sheet name intact and every cell value gone. The fixture uses a dimension
element and a shared string table instead.
The frozen literals are pinned to a NAMED converter version, asserted beside
them: a frozen literal without one says "these bytes" without saying what
produced them.
Suite 908 -> 913. Fixtures regenerate byte-identically.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Order G2a. Populates the optional `[extract]` extra for the first time with
one parser, `pdfplumber>=0.11.10,<0.12` (MIT), and wires `pdf` through it.
The default install is untouched: exactly one runtime dependency, stdlib
otherwise, enforced by test_packaging.py.
The gate for `pdf` becomes an import probe rather than a frozenset membership
test, exactly as extract.py's docstring had promised. The rejection does not
change: without the extra, `pdf` still raises `extractor_extra_missing` with
the same message. That behaviour is asserted UNCONDITIONALLY via a sys.modules
monkeypatch, so it holds on machines where the parser is installed too — a
skip would have preserved nothing there. Verified in a clean venv without the
extra: 589 passed, 7 skipped; with it, 596 passed.
`docx`/`xlsx` are unchanged and still fail fast — the extra names exactly what
it ships.
The parser choice was forced by measurement, not preference (b73dd9d,
docs/2026-08-21-g2-pdf-extraction-measurement.md): on a real requirement table
pdfplumber keeps 4 of 4 rows with label and value on one line, where pypdf,
pdfminer.six and pymupdf each keep 0 of 4. pymupdf is additionally out on
licence (AGPL-3.0), which an MIT package must not push onto a consumer.
Three facts from that measurement are now carried in code rather than in a
report:
- Extracted text is pinned to an exact transitive parser version
(pdfplumber pins pdfminer.six==20260107; date-stamped, no stability
contract). tests/test_extract.py freezes the expected text of a committed
hand-written fixture so a parser upgrade breaks something visible instead of
drifting silently. Reasoning at the declaration site and in
tests/fixtures/README.md.
- Determinism within a version is now held by a test, not only measured once.
- Drawn content does not survive extraction. Every pdf extraction emits the
new `ExtractionWarning`: figures have no text to recover, so a bundle built
from drawn documents is incomplete by construction. Stated categorically
rather than detected — deciding "is there a figure here" is the layout
heuristic G2b declined.
Two new error codes, both mirroring existing patterns: `extractor_empty_pdf`
(a scanned/image-only PDF, refused rather than persisted as an empty concept)
and `extractor_pdf_error` (parser failure wrapped, never leaked).
Structured table recovery (G2b) is NOT implemented and is documented as out of
scope: two independent parsers return the same wrong shape, so the breakage is
document geometry, not a library choice. PDFs enter as prose.
Also corrects an install promise this change would otherwise have published:
the README no longer presents a bare `pip install 'llm-ingestion-okf[extract]'`
as working, because the package is not on an index.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HtNhsdHnMGtMi7U2mvMU8z
