llm-ingestion-okf/docs
Kjell Tore Guttormsen 17c49fc04b feat(consume): give every excerpt the name and the address an answer must cite
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>
2026-09-08 15:01:41 +02:00
..
plan fix(corpus): the root index links the bundle's own log 2026-09-03 20:23:56 +02:00
2026-08-21-g2-pdf-extraction-measurement.md docs(extract): measure one Vegnormalene PDF page against the extraction registry 2026-08-21 19:16:15 +02:00
2026-08-29-konverteringsarmen-arm-a.md docs(measure): run the conversion arm on K2 - pandoc wins both office formats 2026-08-30 00:41:48 +02:00
2026-08-29-telling-offentlig-prosjektkorpus.md docs(measure): count the public project corpus - K2 measured, K1 blocked 2026-08-29 07:03:50 +02:00
2026-09-02-k3-k4-k5-baseline.md docs(measurement): K3, K4 and K5 numbers against the committed method 2026-09-02 15:55:19 +02:00
2026-09-02-k3-k4-k5-metode.md docs(measurement): K3, K4 and K5 method with n declared before any verdict 2026-09-02 15:46:03 +02:00
2026-09-02-korpuskjoring-k1-k2.md docs(measurement): K1 and K2 over the K2 corpus 2026-09-02 15:43:45 +02:00
2026-09-03-k2-bundle-rebuild.md test(corpus): a rebuild with log.md present reproduces the bundle exactly 2026-09-03 05:16:25 +02:00
2026-09-04-cid-andel-k2.md docs(measure): CID-glyph share across K2 -- Bilag 9.1 alone, not a K1 arm 2026-09-04 20:23:41 +02:00
2026-09-04-k2-pptx-odt-rtf.md docs(measure): K2 on pptx/odt/rtf -- denominator 0 in both named corpora 2026-09-04 21:12:41 +02:00
2026-09-04-k3-arm-c.md docs(measure): K3 with Arm C beside the baseline, two numbers, no threshold 2026-09-04 17:58:50 +02:00
2026-09-07-k2-pptx-odt-rtf-fixtures.md test(fidelity): K2 on pptx, odt and rtf -- 59/59 and 20/20, on a set we wrote 2026-09-07 05:22:12 +02:00
2026-09-07-k3-arm-d.md docs(measure): K3 Arm D beside a re-rated Arm B -- reach and no threshold 2026-09-07 03:04:01 +02:00
2026-09-07-k3-arm-e.md docs(measure): K3 Arm E beside Arm B and Arm D -- three rows, no threshold 2026-09-07 11:48:58 +02:00
2026-09-07-k3-enhetsark-metode.md docs(measure): K3 unit worksheet delivered for operator, no code changed 2026-09-07 16:36:16 +02:00
2026-09-07-okf-konsumskill-maaling.md docs(consume): refutation sweep — smoothed-IDF rarity weighting cannot zero both known-negatives 2026-09-07 09:56:36 +02:00
2026-09-08-blindsone-below-k-k2.md feat(consume): measure the below-k blind spot, add one flag-gated vocabulary bridge 2026-09-08 04:55:45 +02:00
2026-09-08-blindsone-laas2-budsjett-k2.md feat(consume): measure the budget lock, add one flag-gated top-rank reservation 2026-09-08 05:18:30 +02:00
2026-09-08-eksakt-oppslag.md fix(consume): match an identifier by equality, deliver the concept a question names 2026-09-08 13:27:59 +02:00
2026-09-08-kravnummer-tokenisering.md fix(consume,propose): hold an identifier number as one token, give an orphaned heading's name to its table 2026-09-08 11:43:39 +02:00
2026-09-08-prisform-og-loggen-k2.md fix(extract,build): write a spreadsheet as pipe tables, stop linking the run log from the index 2026-09-08 10:06:58 +02:00
2026-09-08-proveniens-k2.md feat(inbox): point every concept at the document it came from, with a locator per format 2026-09-08 14:39:24 +02:00
2026-09-08-sjeldenhetsvekt.md feat(consume): weight a lexical hit by its rarity, off by default and measured 2026-09-08 12:36:57 +02:00
consumption-contract.md feat(consume): give every excerpt the name and the address an answer must cite 2026-09-08 15:01:41 +02:00
okf-nokkelinventar.md build(deps): pin llm-ingestion-guard v1.3.0 so the gate reads our own goldens 2026-09-03 20:41:03 +02:00
phase-3-split-table.md docs(phase-3): keep the index-entry grammar on one line 2026-08-04 12:37:02 +02:00
upstream-okf-upgrade-runbook.md feat(watch): weekly OKF upstream watch that can prove it found nothing 2026-08-23 20:38:37 +02:00