llm-ingestion-okf

Turn a folder of documents into a bundle a model can answer from with a source on every claim — offline, deterministic, no model call anywhere in the run path. Thirteen file types are read; Supported file types lists each one with the evidence behind it.

Install

Python 3.10+ and uv. One line:

uv tool install "llm-ingestion-okf[extract] @ git+https://git.fromaitochitta.com/open/llm-ingestion-okf.git@v0.10.1"

Use it

okf project ~/my-documents   # folder in: bundle + Claude Code skill, in this directory
claude                       # start Claude Code here

Then ask in plain language. Three shapes of request work, and the skill states the rules for each:

• a question — "hva er kravene til pris?"
• a hypothesis — "stemmer det at leverandoeren baerer risikoen for grunnforhold?" Answered per premise as confirmed / refuted / undecidable-from-bundle.
• a task whose answer is a document — "lag krav-pris.md med alle krav til pris, ett avsnitt per krav, med dokument og kravnummer." Every claim in the written file carries its source; a paragraph with no ground is written and marked, never dropped.

Everything below is detail: Consume in Claude Code for the same thing in steps and with several bundles at once, Build for the flags, Requirements for the pip fallback and the guard pairing.

What this library is

Status: phases 1–3 are implemented. Phase 1 (spec-based ingestion) covers manifest validation, the file/sql/http connectors, deterministic materialization, index generation, and the golden fixture suite under examples/. Phase 2 adds the bundle inbox (process_inbox) and external-bundle import (import_bundle), both against an injected persist gate, with llm_ingestion_okf.guard_adapter wiring that gate to the real guard (see below). Phase 3 makes the bundle contract configurable, so types, layers, frontmatter sets, index shape, and reserved-file policy are carried by a profile rather than by constants (see Upstream OKF versions). Binary extraction runs behind the optional [extract] extra: pdf through a PDF parser, and five office formats through a vendored document converter. Three of those five office rows are constructed rather than measured — see Binary extraction. Phase 4 (the Node half) is planned (see docs/plan/).

Supported file types

Thirteen extensions are read. The table below is pinned to the extraction registry by tests/test_docs_promises.py — row for row, and cell for cell on the evidence class the code records — so a type added without a row here turns that test red.

File type	Read by	Dependency	Evidence	Note
.md	_extract_passthrough	core	stdlib, no corpus class	The decoded bytes are the concept text, so source_lines are the original's own lines.
.txt	_extract_passthrough	core	stdlib, no corpus class	As .md. A document with no headings yields no segments, which is a failed build rather than a flat bundle.
.csv	_extract_csv	core	stdlib, no corpus class	Parsed with the stdlib reader and rendered as one markdown table; a file with no header row is refused.
.json	_extract_json	core	stdlib, no corpus class	Fenced verbatim. No structure is derived from the keys.
.html	_extract_html	core	measured	Block tags open their own lines and h1–h6 carry the ATX marker for their own level. It does not go through the converter, although the converter reads HTML: that route would add CVE-2025-51591 (SSRF via an iframe in HTML input) and buy nothing. The denominator is 828 files — one product, one format, one publisher.
.htm	_extract_html	core	stdlib, no corpus class	The same reader as .html. The 828-file class is recorded for .html alone, and this row does not borrow it.
.xml	_extract_xml	core	measured	A NISO-STS document (<standard> root, or any <sec>) becomes one heading per titled section at the section's own nesting depth; any other XML keeps its text in document order and gets no invented structure. A <!DOCTYPE is refused unparsed. The denominator is one file, one publisher, one schema — 2 761 titled sections.
.pdf	_extract_pdf	[extract]: pdfplumber	measured	Eight corpus documents with a hand-counted fasit, plus a 701-page process code whose publisher also ships its structure. Prose only: drawn content has no text to recover, and every extraction warns. OCR lives here as a reading mode for a PDF page whose own text never arrived (--ocr, OCR_CID_SHARE), never as an entry for image files.
.docx	_extract_office	[extract]: pypandoc-binary	measured	Five corpus documents. source_lines index the extracted text and not the original's paragraphs — the two counts agree on none of the five.
.xlsx	_extract_office	[extract]: pypandoc-binary	measured	Written as pipe tables, one source row per line; a sheet name becomes a heading, and a row is located by source_sheet and source_rows.
.pptx	_extract_office	[extract]: pypandoc-binary	constructed	N = 2 decks. 2 of 2 slide titles recovered on a deck that declares them, 0 of 2 on a deck that does not, where the converter writes Slide 1 / Slide 2 because it has no title to use.
.odt	_extract_office	[extract]: pypandoc-binary	constructed	N = 1 document. 1 of 1 declared headings recovered, 1 concept, 0 characters in no segment.
.rtf	_extract_office	[extract]: pypandoc-binary	constructed	N = 1 document, and the weakest row here. 0 declared headings: the container carries no heading style, so the author's title is bold text and the document lands as one concept — content preserved, structure zero. The --bold-title flag reads that bold line and is off by default.

Since 0.10.0 five of those rows also carry IMAGES — .pdf, .docx, .pptx, .html/.htm and .xml. The bytes go to assets/ at the bundle root under a content-addressed name, and the concept carries a pointer where the picture stood. .xlsx is deliberately not among them: its converter writes one pipe table per sheet and a two-line block inside one would break the row locator source_rows is read back out of; measured 2026-09-16, 0 of 4 K2 workbooks hold any media at all, so the row is a stated limit and not a loss taken. .csv, .json, .md, .txt, .odt and .rtf are absent because nothing has measured an image reaching them.

The three classes are the code's own and are not interchangeable. measured means real documents someone wrote for their own purposes, counted against a fasit written before the lookup. constructed means the row has been through end to end on a hand-built document with a hand-written fasit and has met no document anyone else wrote. stdlib, no corpus class means the code records no class for the row at all. The office rows are set out in full under Binary extraction.

Not read today

Facts about the registry as it stands, not a queue — nothing here is planned.

• .doc (Word 97) — the converter does not read it.
• .epub — the converter does read it, and the row is deliberately absent: it would buy nothing today, the same half of the reason that keeps HTML out of the converter.
• .eml and .msg — no reader. An email file arrives as an unregistered extension.
• Image files — no reader, and OCR is a PDF page mode rather than an entry for them.
• Source files (.py, .ts, and the rest) — no reader. The text rows are .md, .txt, .csv and .json.
• .one (OneNote) and .vsd (Visio) — no reader.

An unregistered extension is a coded rejection (extractor_unknown), never a silent skip.

Install in detail

Neither this package nor the guard it depends on is on a package index yet, so both install by direct reference. With uv, one command resolves both:

uv pip install "llm-ingestion-okf @ git+https://git.fromaitochitta.com/open/llm-ingestion-okf.git@v0.10.1"

uv resolves the guard on its own, because it reads the [tool.uv.sources] entry in the pyproject.toml of the tag it is installing, and v0.10.1 points that entry at llm-ingestion-guard v1.4.0. Use uv tool install instead of uv pip install when you want the okf command on PATH without an active virtualenv — that is the form the first screen shows.

With plain pip, the transitive git dependency does not resolve on its own — install the guard first, or installing this package fails with No matching distribution found for llm-ingestion-guard:

pip install "llm-ingestion-guard @ git+https://git.fromaitochitta.com/open/llm-ingestion-pipeline-security.git@v1.4.0"
pip install "llm-ingestion-okf @ git+https://git.fromaitochitta.com/open/llm-ingestion-okf.git@v0.10.1"

The guard tag is paired to the okf tag, not to this branch. v0.10.1 declares llm-ingestion-guard>=1.2,<2.0, which v1.4.0 satisfies; the pairing above is read off that tag's own [tool.uv.sources], not off this branch. Reading a pin off main and installing it against an older okf tag is the one combination that fails.

Earlier tags, as history

These are not install lines. They record what each earlier tag was, so a reader who meets one in an older document knows what they are looking at.

• v0.10.1 — the current tag: the image path of v0.10.0, with the two defects an independent review found in it closed. A remote <img src> or xlink:href is inert text with the address in one code span, never a live markdown image link, and no longer loses the figure's caption. An image is bounded in three places rather than one: the size a container DECLARES, the size a carried file has, and — new in this tag — what the stream behind a PDF image actually DECOMPRESSES to, which is an independent number. A declared size that is not positive is refused with its own code, asset_size_invalid, before the stream is read.
• v0.10.0 — a bundle carries the IMAGES its sources declare. Five readers place them (pdf, the converter's office rows, html, xml), assets/ at the bundle root holds the bytes under a content-addressed name, and the concept carries a two-line pointer where the picture stood. ON by default; --no-assets reproduces the pre-0.10.0 bytes, measured on the 43-document reference corpus as a one-line difference in log.md. The image bytes are not screened — the gate reads text — and the log says so. Door C carries the assets its merged concepts point at.
• v0.9.0 — okf quality <bundle>, a per-file-type verdict on a bundle with the denominator on every line. Three verdicts (PASS / FAIL / UNMEASURED) and a type with no measured threshold is never PASS; two bars ship, .pdf 8/32 and .docx 2/5, both regression bars against the pinned 43-document reference rather than quality claims. okf check is untouched and still has seventeen rules.
• v0.8.5 — a block sources: sequence is decoded by all three of this library's flat frontmatter readers, where they returned the key with an empty value. okf.parse_frontmatter is public API, so this changes what an outside caller reads: it returns a flow string where it returned an empty one. That string is a READING projection — PyYAML reads it back on 0 of the 4 605 block files measured, because the ? opening a query string in the source URL ends the flow scalar — and the emitter still writes flow, so no bundle bytes move. okf consume also stops scoring the door's own Enclosing section: link line, which is now the default reading; the excerpt still carries the line, so what moves is order and never an excerpt's bytes. No new functionality; okf check has seventeen rules.
• v0.8.4 — frontmatter this library writes is YAML a YAML reader reads back the same (a block scalar it would misread is written double-quoted; a flow leaf with no form both readers accept is refused), and okf consume resolves a concept's parent: pointer into the excerpt (parent: { concept_id, title }) with one enclosing-section link in a heading-only body, a pointer the index now resolves too. okf consume --follow-parent and okf build --shell-parent ship off; okf check has seventeen rules (parent_unfollowable). The guard pin moves to v1.4.0, which reads a flow sequence of plain scalars (source_offset: [1, 24]) that the previous pin refused.
• v0.8.3 — a NISO-STS document's own <doc-number> names its directory and titles its sources entry, okf build --frontmatter KEY=VALUE stamps a key on every concept of a run, and an STS section's description comes from its own first spec point. okf consume keeps a leading directory every concept id shares out of its first fusion signal, which is what keeps that identity from costing the answering section its rank. okf build --shell-parent ships off; okf check still has sixteen rules.
• v0.8.2 — okf check refuses a skill and a payload that name different bundles (bundle_mismatch, the checker's sixteenth rule; v0.8.1 has fifteen), and a covered title no longer rises above a title that answers more of the question. No new command or flag; the shipped skills/okf-consume/ is regenerated so it passes that check.
• v0.8.1 — a question that accounts for a concept's WHOLE title reads that concept first (--title-covered, on by default, opt out with --no-title-covered). A ranking fix, no new functionality: on one publisher's 2 761-concept bundle the answering section was delivered at rank 1 on 3 of 6 scored questions before it and 6 of 6 after, and no other measured bundle's payload changed one byte.
• v0.8.0 — .xml is a core file type, read as NISO-STS through the stdlib parser, and a section the source DECLARES takes the declared-structure route — one publisher's process code segments at 2 761 of 2 761 of its own declared sections at the shipped defaults. No other file type changes one byte, measured on the bytes.
• v0.7.0 — okf project builds the bundle okf build builds (they were one flag apart before it), and the generated skill states the question, hypothesis and document-task modes with relative paths.
• v0.6.0 — the first tag carrying the okf project, okf consume, okf check and okf skill subcommands.
• v0.5.0a2 — a pre-release for the named OKF v0.2 pilot set only.
• v0.4.0 — the last tag before the OKF v0.2 work; it declares llm-ingestion-guard>=0.2,<0.3, which only guard v0.2.0 satisfies.

No tag yet makes OKF v0.2 generally available. OKF_LATEST is unchanged and still points at DEFAULT; flipping that alias is the GA event and none of the tags above is it (see Upstream OKF versions).

Build

Installing the package installs one command. A folder of documents in, an OKF bundle out:

okf build ./documents --bundle ./bundle --bundle-id my-bundle --okf-version 0.2

It walks the folder recursively, proposes a segmentation for each document with the mechanical rules, replays those proposals through the bundle inbox, writes the bundle and its log.md, and prints the run's numbers. Every proposal is marked PROPOSED and adjudicated: false — the command segments nothing a human has approved, and says so in the artifact.

The last line that matters is the conservation identity: merged + coded rejections == N, where N is the folder's file count read at run time. The run exits non-zero when it does not hold, and names the unaccounted files, so a pipeline cannot mistake a partial bundle for a complete one.

Flags worth knowing: --segments off ingests each document as one concept and asks for no root values; --plans-dir keeps the proposals instead of discarding them; --report writes the full report to a file as well as stdout. --ingested-at and --proposed-at default to 1970-01-01T00:00:00Z rather than the clock, so two builds of the same folder are byte-identical — a wall-clock default would break rebuild-equals-incremental for every caller who did not pass them.

Images: what the source draws, carried (0.10.0)

okf build carries the images its sources declare. The bytes go to assets/ at the bundle root, named <sha256[:12]>-<the source's own base name>, and the concept carries a two-line pointer where the picture stood:

![Tabell 84-2 Toleranseklasser](/assets/e54e5f5da0e8-tabell-84-2.png)
Image: graphics/tabell-84-2.png (120x90 px) -- Tabell 84-2 Toleranseklasser

The first line is markdown, so a reader that renders the concept sees the picture. The second states what the first cannot — the name the SOURCE gave the file and the size in pixels — which are the two facts a person checking the bundle against the original needs.

Why it exists. Measured on R761 Prosesskoden:2025, a process code published as a 701-page PDF and as a NISO-STS delivery: the process text is carried in full, and 12 Tabell N-N and 9 Figur N-N captions stand over nothing, because the publisher ships those tables as raster images in both deliveries. Process 84 says "toleranseklasse ... er gitt i tabell 84-2" and table 84-2 is a JPEG. A bundle like that reads as complete and is not.

What it costs, measured on the 43-document reference corpus (K2/trinn1, two builds of the same commit, 2026-09-17):

	--no-assets	default
concept files	453	454
markdown files	865	867
assets	0	2 964
bundle size	4.7 MB	115 MB
wall time	2 414 s	3 088 s
peak RSS	6.26 GB	8.74 GB

2 964 carried of 3 145 found; 4 622 pointers, so content de-duplication folds 1 658 repeats into the files they already are. 422 of the 865 markdown files differ. One concept appears, and the mechanism is measured rather than guessed: the pointers are body text, so a section holding 146 of that document's images grew from 19.0 % to 30.6 % of the extracted text and crossed --outline-gate's 0.20 share clause.

--no-assets reproduces the pre-0.10.0 bytes, and log.md then says NOT CARRIED rather than falling silent — a bundle nobody looked for figures in must not read like a bundle of documents that had none.

The image bytes are not screened. The gate reads text; a picture is not text. The pointer block passes the gate like any other body line, and the file beside it passes nothing. log.md says so on every run that carries one.

Every carried image is one a model can be SHOWN. A bundle that holds a picture nothing can read is worse than one that says the picture is missing: the count reports that it arrived. Measured over the frozen R761 delivery's own assets/ (denominator 50): 29 JPEG, 2 PNG and 19 RLE8 BMP — correct files that no model decodes. Every asset's type is read off its bytes and tested against the viewable set; a BMP is converted losslessly to PNG (8-bit uncompressed, 8-bit RLE8, 24-bit uncompressed), and anything else outside the set is refused with asset_not_viewable and a line in the concept saying what stood there. A BMP variant this reader does not express — RLE4, BITFIELDS, 16- or 32-bit samples, a 12-byte BITMAPCOREHEADER — is asset_bmp_unsupported, a different fact about the document and a different thing to go and fix.

The reader is stdlib and adds no dependency. Pillow, which this tree already carries transitively under pdfplumber, was measured first and rejected on two counts: images are carried on the CORE path, where .html and .xml need no [extract] extra, and an asset's name is its content digest — encoding through an installed library would make a bundle's identity move with that library's version, which is the property page rasterisation was felled over. Pillow is the independent decoder in the tests instead, and against it 19 of 19 of R761's real RLE8 assets convert with identical RGB, 2 366 365 pixels compared.

A converted asset is ONE asset: one file in assets/, one pointer, one row in the accounting. The pointer's second line — where the source's own file name and the size in pixels already live — states the original media type, the original sha256 in full and the new one, so a reader can take the original delivery, run shasum -a 256 and find the row.

The cost is measured per image rather than per bundle, with a committed script (tools/okf_asset_census.py) run from two pinned trees over 9 714 image rows: exactly 35 rows moved. Nineteen are the BMPs, now PNG. The other 16 are JPEG 2000 objects carried out of PDF streams — a format no model decodes either, and one no stdlib route converts, so they are refused with asset_not_viewable and stated in the concept instead of being carried unreadably. 9 321 of 9 321 JPEG and PNG rows are byte-identical across the move.

A size CEILING, read off the same corpora (0.10.1). An image over MAX_IMAGE_PIXELS (40 000 000 pixels) or MAX_IMAGE_BYTES (256 MiB) is refused with asset_too_large, counted like every other refusal. The largest image in the 43-document reference corpus is 4 515 x 4 128 (18.6 MP) and the largest of R761's 109 pictures is 2 072 x 656 (1.4 MP), so the bound is an order of magnitude above anything measured.

It exists because a few kilobytes can declare an enormous picture: a 9.6 KB PDF declaring 3 000 x 3 000 grayscale zeros took 83 MB of peak RSS, a 63 KB one declaring 8 000 x 8 000 took 276 MB, and the cost is linear in the pixel count, so one document could take a whole batch build with it — before any gate, because the guard never sees image bytes.

Three numbers are bounded, not one, because a claim is not a cost. What a container DECLARES (/Width x /Height, a PNG header, a data: payload's encoded length) is read before anything is decoded. What a carried FILE measures is read the same way — this package never decodes such a file, so it pays nothing for it, but writing a 7 000 x 7 000 PNG of 47 705 bytes into a bundle would hand the consumer the same bomb with 7000x7000 px printed beside it. And what a PDF image's STREAM decompresses to is measured before it is held, a chunk at a time and discarded, because /Length is the COMPRESSED length and a dictionary declaring 1x1 may hang 400 MB of deflated zeros off it. Measured: that document is 408 516 bytes and cost 892 MB of peak RSS with only the declared size bounded; with the stream bounded it is refused at 54 MB, and a three-times-larger bomb costs 62 MB rather than 2 436 MB.

Every link of the filter chain is bounded, not only the first. A PDF decodes a stream through a list of filters, and /Filter [/FlateDecode /FlateDecode] puts the whole expansion in the second one: measured, 1 636 bytes of file cost 886 554 624 bytes of peak RSS when only the first link was measured (52 367 360 with every link measured), and the picture was still refused at the end — after the memory had been spent. An encrypted stream is deciphered first and then measured like any other.

And what a link COSTS is bounded, not the size of its output. Bounding every FlateDecode was still not a bound, because ASCII85Decode had been classed as safe "because it shrinks" and it does not: z is that encoding's shorthand for four zero bytes, so the filter quadruples its input, and base64.a85decode holds about a hundred bytes of memory per byte of input. Measured in paired subprocesses on an idle machine, the document built once and read from a file so the fixture is not what is measured: a 33 475-byte PDF decoding through [/FlateDecode /ASCII85Decode] cost 3 261 599 744 bytes of peak RSS and the picture was CARRIED; bounded it is 42 070 016 and asset_too_large. Doubling the run of z takes the old cost to 6 461 558 784 and the bounded one to 40 280 064 — the cost no longer follows the bomb. So FlateDecode is measured a chunk at a time as it is paid, and every other permitted filter carries a MEASURED worst-case cost per byte of input which is checked against the budget before its decoder is called. Any other filter — LZWDecode, RunLengthDecode, CCITTFaxDecode, /Crypt, anything written after this — has no measured ratio, so an image behind one is refused UNREAD with its own code, asset_pdf_unbounded, rather than decoded to find out what it costs. The cap that falls out for ASCII85Decode is read off the corpora the way the pixel bound is: over the 9 668 image objects of the 77 PDFs measured, 16 decode through such a link and the largest input to one is 450 739 bytes, more than ten times under the cap. A property test runs every chain of length 1–3 over the ten filters pdfminer decodes — 1 110 of them — and requires each to be delivered under the bound or refused with a published code, never paid for on the way.

What the stream bound does NOT reach, stated because the difference matters: a stream something else has already decoded, where the memory was spent before this package was asked. That one is caught by a check on the decoded length AFTER the decode, which makes it a counted refusal rather than a bounded one — the picture is dropped by count, not by bound.

A declared size that is not a size — a zero or negative /Width or /Height — is refused with its own code, asset_size_invalid, before the stream is read. Distinct from asset_too_large on purpose: one is a publisher shipping a picture bigger than this package carries, the other is a dictionary written wrong or written to be read wrong.

A remote reference is INERT (0.10.1). <img src="https://..."> and an STS xlink:href to an address off this machine are written as text with the address in one code span, never as ![...](https://...) and never as a bare URL a linkifying renderer would autolink. Extraction opens no socket, but a markdown renderer or an agent that fetches what it renders does, which would turn "this bundle was opened" into a beacon to whoever wrote the document. The address is still stated, and so is the figure's caption, because a reader has to know what stood there.

images: N in a concept counts POINTER BLOCKS, not unique pictures. One image referenced twelve times in one concept is images: 12 and one file in assets/. The key is a count of the places a picture stands, and dedup is on content.

No size floor, and that is a measurement too. The obvious filter is "ignore anything under N pixels", and the distribution offers no N: over the 4 828 image objects in that corpus, 149 declare no size, 162 are under 32x32, 92 under 64x64, 406 under 128x128, 498 under 256x256, 590 under 512x512 and 2 931 are larger — a broad spread with no gap, unlike OCR_CID_SHARE's, which is bimodal with nothing between the modes. A threshold read off no gap is a number this package chose, and it would silently drop somebody's small table.

--accounting PATH accounts for the CONTENT, not only the files. Before extraction, every source document is inventoried in a per-format element vocabulary: headings, paragraphs, tables, cells, images, pages, and so on. After the run, each element gets exactly one fate:

• carried: all of its text is in the concepts written for the document, or its image is in assets/;
• pointer: a remote image, which is never fetched, or a markdown image reference kept verbatim;
• a coded rejection: the gate's or the reader's code.

The result goes to PATH as JSON and into log.md. The build exits 1 when an element has no fate or has two. A document the gate refuses is logged as <file>: <M> elements found in the source, 0 carried: document rejected ``, and the Images bullet then counts what the sources declare.

"Carried" means the text is present, not that it is in the right place. A short element such as a section label can be found elsewhere in the same document. The judge is tools/okf_accounting_gate.py, which compares the inventory against an independent witness.

Two operator decisions, 2026-09-17. The accounting stays OPT-IN until the losses it reports on the reference corpus are fixed, because a default-on door would fail builds that pass today. And of the three exceptions the gate proposed, only the PDF one is approved: a PDF without a structure tree declares no heading, paragraph or table, so no witness can count them. An image in a workbook, or in md, txt, csv, json, odt or rtf, stays unaccounted and therefore stays red.

A run that refused a document whole says so in both places: the accounting carries refused and each document's own status, and log.md carries R of D document(s) refused whole. The exit code does not move for it — it belongs to the whole run, and a corpus holding one unreadable file among many is ordinary — so the count is what keeps a partial refusal from being silent. The judge treats such a document as never clean, with its elements in their own refused column: every one of them is booked honestly as a coded rejection, so u and d both stay 0 and nothing else could see the loss.

The soft hyphen is removed before the persist gate, and counted (operator decision 2026-09-18). U+00AD is in llm-ingestion-guard's zero-width set, and output:zero-width-present is an any-tier carrier: a document carrying one is fail_secure at every trust level. Measured on R761 Prosesskoden:2025 — 71 U+00AD, and 0 of U+200B, U+200C, U+200D, U+FEFF and U+2060 — those 71 are Norwegian hyphenation points inside words (ar[SHY]beider, bitu[SHY]men), so a 701-page process code was unreadable for the whole chain over typography. extract.normalise_extracted removes that one character from every extracted text and reports the count as normalised_soft_hyphen, per document and for the run, in the accounting JSON and in a **Normalisation** bullet in log.md. The guard is not touched and the other four characters are not touched: they carry no typographic job in running text, so removing one would be a decision about what the guard screens for, taken in the wrong repository. U+00A0 NBSP is not in the guard's set and is not touched either. Reach, measured 2026-09-19: 0 of the 78 readable documents of the reference corpus carry any of the six characters, so no bundle measured here moves.

Two things hold with or without the flag:

• okf build exits 1 when it extracted at least one document and persisted none.
• An image file beside a document is counted once. If a persisted document carried it, it is in the conservation identity's own column (merged + files carried through a document + coded rejections = N); otherwise it is a coded rejection.

--frontmatter KEY=VALUE (repeatable) stamps a key on every concept of the run, for what the operator knows and the document does not say — an edition, a publisher's address. It splits on the first = and writes the value verbatim on one line, so 'sources=[{ resource: <url>, title: <t> }]' survives whole. It adds any key and replaces only sources and description, the two with a layer below them (what the document declares, else the file name); every other key the door writes itself is refused before anything is read.

--shell-parent (off; opt out explicitly with --no-shell-parent) gives a concept whose body is its heading alone a parent: naming the segment_id of the nearest ancestor that holds text — the nearest preceding plan entry at a smaller level, passing over an ancestor that is empty too. It copies no text and moves no boundary. It exists for a document that states its points once and lets every nested section inherit them: measured on one process code, 710 of 2 761 concepts are heading-only, and the plan's level and order name the same ancestor as the document's own nesting on 710 of 710 since K3-21 (708 before: the two others sit at depth 7, and the reader clipped their level to 6 in the plan as well as in the markdown heading, so they pointed one level too high). It was off because okf consume did not read parent. Since K3-21 it does: an excerpt carries parent as the enclosing concept's concept_id and title (resolved inside the concept's own document, never the raw segment id), and a heading-only body gains one line, Enclosing section: [title](/path), in the bundle-relative form SPEC § 6.1 recommends. okf consume --follow-parent carries the enclosing section's text inside parent as well, with that concept's own sha256, and only from the room the cut left — so it never displaces an excerpt. Whether either default moves is measured separately. The ranking does not score that line (K3-25). It reaches the excerpt and the file exactly as before; consume.DEFAULT_LINK_IN_SIGNAL is False, so the body signal and the stem vocabulary read a heading-only body without it. The measurement: of the newcomers the line ever added a question token to, 39 of 39 gained it from the bundle-absolute PATH and 0 of 39 from the link's title, every such token being a segment of the document's own directory — so a bundle built with --shell-parent now delivers what the unflagged build delivers, on 16 of 16 rows. It cost no shipped bundle a byte: 0 of 5 carry the line, so 5 of 5 payloads are byte-identical across the move. The index resolves the pointer too: a segment id is looked up among the concepts of its own document, so it renders parent: p1 when the concept it names is in the bundle and keeps the ? only when nothing answers to it.

The segmentation flags

Nine rules are reachable from okf build, and since 2026-09-11 all nine are ON by default — --outline-run 3, --table-grid and --unit-fold since 2026-09-08, --drop-wrapped-outline and --outline-gate since 2026-09-09, --sheet-section-rows, --keep-table-heading and --first-span-from-zero since 2026-09-10, and --close-span-gaps since 2026-09-11 — each an operator decision, and each with an explicit opt-out: --outline-run 0, --no-table-grid, --no-unit-fold, --keep-wrapped-outline, --no-outline-gate, --no-sheet-section-rows, --no-keep-table-heading, --no-first-span-from-zero, --no-close-span-gaps. Passing all nine reproduces the pre-2026-09-08 bytes exactly, and the last three reproduce the pre-2026-09-10 bundle byte for byte — measured with diff -rq, 0 differences, not asserted. Each line below carries the number it was measured at, and nothing beyond it.

An eleventh flag, --pdf-outline, is off by default, and it is the one rule here that does not read the extracted text at all: it cuts a PDF at the boundaries the file's own /Outlines bookmark tree declares. It is not --outline-run under another name — that one is a text heuristic over numbered lines in the extracted text, while this one opens a structure index the PDF already carries and the build had never looked at.

It is a segmentation arm and not a reader option: the extracted text is byte for byte the same either way, and a PDF that carries no bookmark tree builds byte-identically with the flag on. Two consequences come free with it. The concept title comes from the BOOKMARK, so it is not cut short where the page wrapped the heading across two lines; and a page that lies before the first bookmark destination is the contents listing rather than a second copy of the body, so a contents entry and the section it lists stop landing as two concepts under one id.

The measurement is one 701-page process code whose publisher also ships a NISO-STS structure for it, so the fasit is the publisher's own. Under the shipped default that document gives 1967 of 2761 boundaries, none of its 28 chapters, and 794 of 794 misses have their heading text present in the text the build read. With the arm it gives 2759 of 2761 and 28 of 28. The flag stays off because reach is the open question, not quality: 1 of the 8 reference PDFs in this repository's own sample carries a usable tree, and a bookmark tree is the publisher's claim about its own structure — a stale or wrongly pointing one carries that error straight into the segmentation.

A tenth flag, --bold-title, is off by default. It is the rule for the type whose container declares nothing: rtf has no heading style, so an author's title is bold text, and the row was measured at 0 of 0 declared headings, 0 concepts and 1368 of 1368 characters in no segment. The grammar is markdown, not rtf — the converter already writes that title as **…** in the same output every office row produces — and it is gated by the principle --outline-gate already carries: recovery yields to declaration. Measured over 47 readable documents, three parameters were swept and one carried; false positives are 0 of the 31 documents that declare, and the rule reaches 2 of 39 corpus documents, both docx, 0 of 33 pdf and 0 of 2 xlsx. On the rtf fixture set it recovers 6 of 6 authored titles over N = 4 with 0 false titles and 0 of 1994 characters in no segment. On a five-document folder it moves 26 → 27 concepts, replacing a mechanical tabell-linje-30 with two named concepts.

A re-run is what this costs a consumer, and it is not a small one: on the 43-document reference corpus the default bundle goes from 629 concepts in 1108 files (the 2026-09-03 tree) to 492 in 944 after the 2026-09-08 move, to 425 in 810 after the 2026-09-09 one and to 436 in 832 after the 2026-09-10 one (digest 8dff8a8e6c15d2f7…). On a five-document folder the last move is 15 concepts in 30 files → 26 in 52. The proposer's own defaults (tools/okf_propose_segments.py) did NOT move, so every published reproduction block still runs as written.

What the 2026-09-09 move had to clear, stated because it is the bar every later move is held to: the twelve-position reference improves (pdf 2 of 8 → 7 of 8, the sheet 5 of 12 → 10 of 12, docx unchanged at 3 of 3) AND hit@8 on a K2 bundle built with it holds 5 of 6 at ranks 1,1,1,1,1,–, with no row losing rank 1. A configuration that improved the reference and cost a rank was measured in the same session and did NOT ship; see docs/2026-09-09-k3-runde6-outline-gaten-og-prioren.md § 9.1.

flag	what it does	measured
--outline-run N (default 3)	also propose a boundary where the document's own bare-integer numbering sustains an ascending run of at least N; 0 is this arm's opt-out	a tender PDF whose headings are bare integers: no boundary at 0, 9 concepts at 3, against a reference of 9
--table-grid (on by default; opt out with --no-table-grid)	a pandoc grid-table rule line no longer closes an open table block, so one grid table is one concept	a .docx experience list: 21 → 6 concepts
--unit-fold (on by default; opt out with --no-unit-fold)	discard a contents-list run, fold a deeper heading into its parent, fold a table into the shorter heading that introduces it. Adds no boundary, so it can only reduce a plan	on a 12-document sample scored against an operator's unit worksheet: 5 of 12 match — but that figure was measured with --table-grid ON, and the shipped default does not include it. Measured without it the same sample scores 2 of 12, docx 0 of 3, because the fold's table clause has no joined table to fold
--keep-table-heading (on by default since 2026-09-10; opt out with --no-keep-table-heading)	keep a heading whose body is empty only because a table opens under it, and absorb that table into its span	the two spreadsheets in that corpus, and 0 of 32 pdf and 0 of 5 docx: the concept count does not move (1 → 1), its first byte does — the concept gains the heading line it was missing
--sheet-section-rows (on by default since 2026-09-10; opt out with --no-sheet-section-rows)	cut an open table block at the rows that label its sections — a run of at least three rows whose first cell is a bare numeric label. The opposite direction from --table-grid, which decides how far a block extends	a tender price sheet whose whole body is one table block: 1 → 12 concepts, against a reference of 11 cost groups plus the sheet's preamble. Whole corpus: 1 of 39 readable documents changes, 0 of 32 pdf, 0 of 5 docx, 1 of 2 xlsx. It reached 11 of 12 on the reference two rounds before it shipped, and was held back both times by a RETRIEVAL cost that turned out not to be its own: on a bundle built with it the gold document splits 1 → 12 concepts and row 1 of the hit@8 set fell rank 1 → 2. The repair is on the reading side (--tie-shared-rank, now the default), and with it in place the sheet reaches 11 of 12 with hit@8 holding 5 of 6 at ranks 1,1,1,1,1,–
--drop-wrapped-outline (on by default since 2026-09-09; opt out with --keep-wrapped-outline)	do not admit an --outline-run candidate whose line continues onto the next one. Judges recovered candidates only, never a heading the document declares	quoted regulation text, whose numbered paragraphs match the outline grammar exactly: 4 → 1 concepts, the reference. Whole corpus: 5 of 39, all pdf; on the 12-document sample 8 of 34 outline candidates wrap, and none of the 26 the operator kept. On the reference it carries pdf from 5 of 8 to 6 of 8 together with the gate below, and neither reaches 7 of 8 without the other
--outline-gate (on by default since 2026-09-09; opt out with --no-outline-gate)	admit --outline-run's RECOVERED headings only where the document declares none of its own, plus any one recovered heading whose span covers OUTLINE_SHARE (0.20) of the text. Applied at admission, before spans are closed, so the text a removed mark opened is carried by the mark above it rather than lost	on the 12-document sample: pdf 2 of 8 → 5 of 8 alone and 7 of 8 with the rule above, docx unchanged at 3 of 3. Whole corpus: it fires on 25 of 39 readable documents, changes the plan in 15 of 39, and removes 64 of 485 proposed entries. No plan disappears (32 → 32)
--first-span-from-zero (on by default since 2026-09-10; opt out with --no-first-span-from-zero)	start the first concept at character 0, so the text above it belongs to a segment instead of to none. Adds no boundary and removes none	Measured over the 39-document corpus, the default before this rule left 207 435 characters — 11.92 % — in no segment at all: 163 804 above the first entry (in 32 of the 32 documents that get a plan), 26 041 between entries and 17 590 after the last. This rule closes the first part entirely, 79 % of the whole, leaving 43 631 characters (2.51 %) over 8 of 32 documents with two named mechanisms of their own. It adds no boundary and the K2 concept count is identical with and without it (425 = 425); on the 12-position reference it changes not one cell, and hit@8 on a K2 bundle built with it holds 5 of 6 at ranks 1,1,1,1,1,– under both tie-breaks
--close-span-gaps (on by default since 2026-09-11; opt out with --no-close-span-gaps)	close a concept's span against the next SURVIVING concept, and the last against the end of the text. Three steps remove a candidate AFTER its neighbour's span was already closed against it — the orphan check, and fold_units clause 1 both between entries and on the last run — and the removed mark's text then belongs to no segment. Adds no boundary and removes none; only spans' ends move	It closes the whole remainder the rule above left: 43 631 characters, 2.51 % of the corpus over 8 of the 32 documents with a plan, to 0 — both the 26 041 between entries and the 17 590 after the last. Decomposed: orphan check 18 527 over 15 of 39 documents, clause 1 7 514 between entries, clause 1 on the last run all 17 590 of the tail (with unit_fold=False the corpus tail gap is 0). The entry count is identical (429 = 429 on the corpus, 436 = 436 concepts on K2, 52 = 52 md on a five-document folder); on the 12-position reference it changes not one cell (11 of 12 under `
--pdf-outline (off; opt out is the default, opt in with the flag)	cut a PDF at the boundaries its own /Outlines bookmark tree declares, instead of at the ones the text rules recover. A SEGMENTATION arm, not a reader option: the extracted text is byte for byte the same either way, and a PDF that carries no tree builds byte-identically with the flag on. The title comes from the BOOKMARK, so it is not cut short at the page's line break, and a page before the first bookmark destination is the contents listing rather than a second copy of the body	one 701-page process code whose publisher also ships a NISO-STS structure for it, so the fasit is the publisher's own: boundaries 1967 of 2761 → 2759 of 2761, chapter level 0 of 28 → 28 of 28, concept titles identical to the source title after normalisation 2761 of 2761, false positives 163 of 2182 → 3 of 2762, directories carrying two concept files 132 of 2050 → 2 of 2738, contents-copy pairs 65 → 0. Consumption on the same eight questions: fasit present in the bundle 4 of 7 → 7 of 7, hit@1/8/50 1/6 · 2/6 · 4/6 → 3/6 · 5/6 · 6/6. Cost 119.22 s → 183.31 s wall, peak RSS 3252 → 3251 MiB, no new dependency and no second parse of the pages. Off, and the reach is why: 1 of the 8 reference PDFs carries a usable tree at all, and a bookmark tree is the publisher's CLAIM about its own structure — a stale or wrongly pointing one carries that error straight into the segmentation

They compose, and the order above is the order they apply in. Measured on a five-document tender folder (2 pdf, 2 docx, 1 xlsx), concepts per document:

document	default	--outline-run 3	+ --table-grid	+ --unit-fold	+ --keep-table-heading	+ --sheet-section-rows --drop-wrapped-outline
tender PDF, technical requirements	1 (no boundary)	9	9	9	9	9
tender PDF, technical layout	1 (no boundary)	1	1	1	1	1
price sheet .xlsx	1	1	1	1	1	12
experience list .docx	21	21	6	3	3	3
agreement .docx	2	2	1	1	1	1
markdown files in the bundle	31	49	33	30	30	52

Every column merged 5 of 5 with 0 rejections. The reference for the first row is 9, so the default is a full arm behind what the proposer can do on that document — which is a statement about the default, not a licence to change it here.

The two PDF reader flags

Separate from the six above, and they sit before every one of them: a segmentation flag changes how the proposer cuts a text, these change what the text says. All three are off by default.

flag	what it does	measured
--pdf-headings font	a PDF carries no heading markup, so one is inferred from typography — a line whose dominant font size is above the document's character-weighted median AND whose dominant font name says bold — and emitted as an ATX heading in the same markdown the office path produces, so the existing heading rule reads it	on a tender PDF: 9 of 9 numbered chapters found, plus 4 extra candidates. Whole corpus: 25 of 32 pdf change, 0 of 5 docx, 0 of 2 xlsx. Off by measurement: against the operator's unit worksheet it takes pdf from 2 of 8 to 0 of 8, losing two exact matches, because on those documents the outline rule already found the chapters and a second heading source can only add
--pdf-headings font-reserve	the same typographic rule, applied ONLY to a document whose own numbering the outline arm finds no run of — typography as a second heading source where there is no first one, never on top of one. Three values of one option (none, font, font-reserve), so no caller can ask for two at once	reaches 4 of 39 readable corpus documents (10 of 32 pdf admit no outline run; 4 of those render differently at all). Off by measurement, and the measurement is that it changes nothing measurable: on the operator's twelve-position unit worksheet it alters not one cell — the five positions where it fires are one PDF whose glyphs carry no ToUnicode mapping and four office documents the PDF reader never touches. The position it was built for numbers its own chapters, so the reserve is silent there by construction
--ocr	read a PDF page as an image when its own text never arrived: the page extracts empty, or as (cid:N) placeholder codes at or above 10 % of its characters. Needs the optional ocr group	on the one corpus document with the failure: 95.07 % → 0 % cid, 44 → 2561 words of four or more letters, 17 → 18 pages with text, 3.6 s/page. Whole corpus: 16 of 834 pages qualify, in 1 of 32 files

pip install "llm-ingestion-okf[extract,ocr]"

Without that group --ocr is a typed refusal (extractor_ocr_group_missing) per file, never a crash, and the corpus run still reports merged + coded rejections == N. OCR text is a model's reading of an image: it is reproducible against the model version and rendering resolution it was produced with, and no dependency pin can promise more. The full measurement, including the per-page distribution the 10 % threshold was read off, is docs/2026-09-08-k3-runde4-pdf-skrift-og-ocr.md.

Measured 2026-09-08 on a 43-file corpus (33 pdf, 5 docx, 2 xlsx, and three files no reader accepts), one okf build invocation replacing the shell loop over tools/ that produced the same corpus's bundle on 2026-09-03:

figure	value
N (folder file count, computed)	43
merged	39/43
coded rejections	4/43 (extractor_unknown 3, extractor_empty_pdf 1)
K1b	39 + 4 = 43 = N, exit 0
files written	1108
identical to the 2026-09-03 bundle	1104/1108
wall time	842.82 s total, 19.600 s per file (re-measured 2026-09-08)

The six files that differ are all in the corpus's two spreadsheet documents, and they are the change reported in docs/2026-09-08-prisform-og-loggen-k2.md: a spreadsheet's tables are now written as pipe tables, so each row is one line with its cells delimited rather than padded out to the widest cell in the column. Two concept files are renamed by it, two are removed under their old names, and the two documents' own index.md follow. The root index.md is identical to the stored one again, because this library no longer links the bundle's log.md from it. The command's own byte-identity test compares it against the two scripts at the current commit, where the two agree over the whole tree.

Consume

The other direction: a bundle plus one question in, one bounded, contract-shaped payload out.

python3 tools/okf_consume.py ./bundle --question "your question" --out payload.json

tools/okf_consume.py is the pre-pass docs/consumption-contract.md § 1 defines — the deterministic program that reads the bundle, ranks its concepts, cuts them to a bounded set and emits one payload. It decides nothing about the question; the skill that reads the payload does the judgement. It calls no model, opens no socket, imports nothing outside the standard library and this package, and takes no clock: the same bundle bytes and the same (question, k, limit, cost_vocabulary, reserve_top_rank, rarity_weight) produce byte-identical output.

--cost-vocabulary is off by default and widens one question class: it lets a declared list of cost/price/quantity terms bridge a question and a document that name money with different words. The gate is the question — one naming no such term gets byte-identical bytes either way — and what it does and does not close is measured in docs/2026-09-08-blindsone-below-k-k2.md.

--reserve-top-rank is off by default and answers a different objection: the budget is packed by an exact knapsack, which maximises a SUM of scores and therefore has no opinion about rank, so a top-ranked excerpt costing a large share of the budget is out-summed by many small ones. Measured, that made --k a dial that could EVICT the concept a question was asked about. The flag gives rank one its bytes before the pack runs — after the over_budget_alone pre-exclusion, never before — and the payload then declares budget.reserved. On a 629-concept corpus it changed the delivered set in 2 of 24 measured combinations, both of them that eviction: docs/2026-09-08-blindsone-laas2-budsjett-k2.md.

--source-quota N is on by default at 2 since 2026-09-10 (opt out with --no-source-quota), and it is the third widening here that alters a payload with no bundle changing. It caps how many DELIVERED places one source document may take, cutting where the shortlist is cut so the freed place goes to the next candidate and k is still delivered in full. The defect it repairs was measured outside this repository on a 3206-concept bundle of a published handbook: the code's own process overview contributes 28 of 3206 concepts (0.87 %) and 8.0 % of the source characters, and took 8 of 8 delivered places on one question and 7 of 8 on the known-positive, which was not delivered at all. Identical at 343 and 1651 concepts, so the cause is the corpus's COMPOSITION — that it holds its own table of contents — and not its size; any corpus with a contents list, a project overview or a summary document has the same property. Swept over {2, 3, 4, off} on three bundles: at 2 and 3 hit@8 goes 5 of 6 to 6 of 6 on both K2 bundles with all five standing rank-1 rows unmoved, and at 4 and off it stays 5 of 6. On the handbook bundle hit@8 goes 2 of 6 to 4 of 6 and the dominant document's share of delivered places 8 of 8 to 2 of 8. 2 rather than 3 on rank: the recovered rows come in at 5 and 4 rather than 7 and 5. What the gain is NOT: hit@8 asks whether the gold DOCUMENT was delivered, and a document quota directly raises how many distinct documents a payload holds, so that metric is not neutral with respect to this rule — the five rows that were already rank 1 are, and they did not move. The adverse case is named rather than found later: a bundle built from ONE document has one source_file on every concept, so the quota would deliver 2 excerpts instead of k; the shortlist is topped back up from the best-ranked over-quota candidates, which makes such a bundle byte-identical to the quota being off. --rarity-weight was measured against the same defect in the same session and does not repair it: on the handbook bundle it leaves the dominant document at 8 of 8 places on the question it floods and delivers neither that answer nor the known-positive.

--rarity-weight is off by default and weights each lexical hit by log(N/df) over the bundle's own concepts instead of counting it as one, so a requirement number is not worth what a common verb is worth. The default being off is a measurement rather than a preference: on four corpora it took one gold concept from withheld to delivered and a priced sheet from candidate rank 10 to 2, left one gold rank unmoved, and cost another seven rank positions — because the four-character prefix matcher makes a unique identifier read as 135-of-446 common on that bundle. Where it cannot help is decomposed rather than guessed: RRF fuses RANKS, so a weight moves nothing on a signal the gold already leads. docs/2026-09-08-sjeldenhetsvekt.md. Its published figures were measured under the pre-2026-09-10 tie-break and are not re-measured.

--stem-prefix is on by default since 2026-09-09 (opt out with --no-stem-prefix), and like --tie-shared-rank below it alters a payload with no bundle changing. MIN_SHARED_PREFIX = 4 exists for Norwegian compounding, and it also matches four characters that are not a stem: measured on the pinned 453-concept bundle with the control run first, under occurs 79 times by equality and matches 172 concepts by prefix, while bilateral occurs 0 times and matched 400 of 453 through bilag, and standhaftig 0 and 219 through standard. Three repairs were measured and all three failed on the same row — 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, because row 1's token prisene reaches its gold document through pris|sammenstilling on four characters. The rule that works asks whether the shared prefix is a WORD the bundle uses: bilateral 400 → 0 and 512 → 0, standhaftig 219 → 56 and 235 → 33, every hit@8 row keeping rank 1 on both bundles. undersjøisk stops at 162 because under is a word here — a genuine Norwegian morpheme, so the residual is a different answer and not a ceiling.

--tie-shared-rank is on by default since 2026-09-10 (opt out with --no-tie-shared-rank), and it is the one change in this library that alters a payload with no bundle changing — a consumer pinned to the previous excerpt order needs the opt-out. RRF emits a rank for every concept in every signal, including a signal that scored them all the same, and the declared tie-break then orders that group by concept_id; the fusion reads alphabetical order as if it were a measurement. Shared ranks make a signal that separates nothing contribute the same constant to each concept in the group. What it buys is general rather than cosmetic: a document the segmenter splits from 1 concept into 12 fills that signal's whole top tie group with its own concepts, so the one that leads the body signal takes position 11 instead of 1 and the document loses fused rank 1 to a single-concept competitor leading nothing — the fusion was punishing fine-graining for being fine-grained, which put the segmentation side and the retrieval side in competition over one number.

It shipped OFF on 2026-09-08 because hit@8 fell 5 of 6 to 4 of 6, and that figure is real and conditional: swept over 2 document-prior exponents x 3 bundles x 6 rows, the lost row is lost only at exponent 1.0. The exponent moved to 0.5 on 2026-09-09 for an unrelated reason, correctly reported as moving no hit@8 row, and nobody measured the pair — so a rule sat behind a published number that had stopped being true in the same commit. A flag's "off by measurement" is a measurement of a configuration, not a property of the flag. docs/2026-09-10-k3-runde7-forste-spenn-og-rangeringen.md.

--title-covered is on by default since 2026-09-10 (opt out with --no-title-covered), and it is the fourth widening here that alters a payload with no bundle changing. A concept whose EVERY title token is a token of the question is read before the concepts the fusion ranked above it. The defect it repairs is that both lexical signals are unnormalised COVERAGE COUNTS: they measure how much of the question a candidate answers, and nothing measures how much of the CANDIDATE the question accounts for, so a section titled with the question's subject alone scores exactly what a narrower section titled with that subject plus a qualifier scores — and then loses on the body count, because a longer title and a longer body can only reach more of the question. Measured on a 2 761-concept bundle of one standard, where none of the six flags above moved the number at all: hit@1/8/50 3 of 6 · 5 of 6 · 5 of 6 → 6 of 6 · 6 of 6 · 6 of 6 at default k, the same 6 of 6 at --k 50, the known-positive holding rank 1 at both and the known-negative still not a hit. The three recovered rows go 4 → 1, not-delivered → 1 and 3 → 1.

It is a PARTITION and not a fourth RRF signal, and the arithmetic is the reason: RRF consumes ranks alone, so with shared ranks a rule whose positive group has m members is worth 1/61 − 1/(61 + m), and a rule firing on ONE concept of 2 761 is worth 0.00026 against a body-signal gap of 0.0029 — a precise rule is worth least under this fusion, backwards from what precision is for. Measured as a signal the same predicate moves hit@1 not at all; as a partition it reaches 6 of 6. lookup_hits is the same shape for the same measured reason, and it still wins: this partition lands below it. The title is read by EQUALITY, never by shared prefix — four shared leading characters take the group from 1 to 6 on one question and 9 to 31 on another, with the answering section falling to candidate rank 6 and the known-positive to 2.

Two candidate repairs were measured and felled first. Pivoted length normalisation of the body signal collapses at every value swept (b = 0.25, 0.5, 0.75, 1.0 give hit@8 3, 1, 1 and 0 of 6, and at 1.0 the known-positive falls to rank 49), because the median concept holds 22 tokens against a mean of 60 — so length normalisation promotes thousands of tiny concepts over the section that treats the subject. Title precision as a signal reaches candidate hit@1 5 of 6 and takes the known-positive from rank 1 to 4 every time it does. The reach of what shipped is narrow and stated as such: it fires on 6 of 8 questions on that bundle and on 0 of 21 measured cells across the pinned K2 bundle, Arm B and the three N bundles, whose payloads are byte-identical either way. docs/2026-09-10-k3-runde16-hele-tittelen-tar-ruten.md.

Since round 17 the partition is bounded by RECALL, under the same flag. A covered title states its PRECISION — it says nothing the question did not ask — and round 16 let that claim override the fusion even against a title answering MORE of the question: on a 26-concept bundle of five tender documents, a question naming a section by three of its title tokens also held a neighbour's whole one-token title, and the neighbour took rank 1. A covered concept now rises through the fusion's order and stops beneath the first concept whose title answers more question tokens, by equality, than it holds. On the standard's bundle nothing stood above a covered concept answering more, so its 8 payloads are byte-identical at both k; the 27 payloads where the rule never fires are byte-identical too. A minimum title length, a share of the question, an order inside the group and a stop list were measured beside it and not taken. 0.8.1's unbounded order is reproducible by no flag; --no-title-covered still gives the pre-0.8.1 order. docs/2026-09-11-k3-runde17-dekningen-stopper-ved-en-bredere-tittel.md.

It emits the § 8 shape — contract, bundle (bundle_id plus a sha256-tree: content identity), budget (unit, instrument, limit, spent and a validated known-positive), denominators, excerpts and withheld — and every withheld concept names the rule that dropped it, from a closed set of seven.

Every excerpt carries the concept's title, and — when the producer wrote them — req_number, the SPEC § 5.1 address sources, and every top-level source_* key, by prefix rather than by allowlist: a fixed list names the locators its author thought of, and one real bundle locates by source_element_id on 269 of its 274 concepts. A key the producer did not write stays absent rather than arriving empty, and an address this reader cannot decode is named (sources_unreadable) rather than dropped into the same silence. A concept naming the section that encloses it delivers parent — that concept's concept_id and title — or parent_unresolved when the pointer lands nowhere; okf check (seventeen rules) refuses a parent a reader could not follow. The reason is a measurement: with concept_id and body text alone, a delivered gold concept at rank 1 still left the answer unable to name the document it was quoting. considered == withheld + delivered closes by construction, and the payload is refused rather than reported when it does not.

Three exit codes, not two: 0 a payload was written, 1 the run happened and refused (the budget admitted none of the concepts that answered the question, or an asserted --ref contradicted the bytes), 2 the run did not happen. Collapsing 2 into 1 would report an unread bundle as a failed cut. --ref is an assertion, never an override — the identity is always computed from the bytes, because labelling a payload with an identity its bytes do not have is the one thing § 3.3 exists to prevent.

Check any payload against the skill that will read it:

python3 tools/okf_contract_check.py --skill skills/okf-consume/SKILL.md --payload payload.json

skills/okf-consume/ is an instantiated consumption skill generated by okf skill from the golden bundle this repository ships, examples/ingest-golden-segmented-okf-v0-2/expected-bundle, together with the payload that proves it: okf check on the pair is conformant with 0 findings. It is regenerated with the command in skills/okf-consume/references/README.md rather than edited, and a test holds the shipped bytes to the generator's. Its numbers describe that three-concept bundle and nothing larger. The measurement behind the consumption skill as a form — hit@8 5 of 6 questions at rank 1 on a 629-concept bundle against a chance baseline of 1.35 of 6, with one control that failed — is docs/2026-09-07-okf-konsumskill-maaling.md; the copy filled by hand for that corpus cannot ship and is no longer the one here.

Judge a bundle: okf quality

okf check is a contract check — it asks whether a payload carries what a claim must rest on. It is not a quality gate, and that is measured rather than conceded: on 2026-09-10 three arms over one corpus all returned 0 findings and exit 0 while their hit@k ranged from 6 of 6 to 0 of 6.

okf quality <bundle> asks the other question, per file type and with the denominator on every line:

okf quality .okf/my-bundle

Three verdicts and no fourth — PASS, FAIL, UNMEASURED — and a type with no measured threshold is never PASS. Exit codes: 0 judged and clean, 1 at least one FAIL, 2 the run did not happen, 3 nothing could be judged (every row UNMEASURED), because exit 0 over a table of unmeasured rows would be the silent pass this command exists to stop.

Two thresholds exist today, both structure_null_share — the share of a type's documents that yielded exactly one concept — read off the pinned 43-document reference bundle:

file type	threshold	N
.pdf	8/32	32 documents
.docx	2/5	5 documents
every type	0 concepts with an empty body	definitional

Every other type is UNMEASURED, including .xlsx (2 documents), .xml (1 document) and .html (no bundle measured here). A threshold needs at least five documents on both sides — the bundle's and its own — because a 1/1 is not a rate.

What a PASS is not. It is a regression bar against a pinned artifact, not a claim that the cut found the document's own structure. hit@k needs a question set and is not asked here; the measurement that says so, the corpora behind each number, and three candidate metrics that were measured and not shipped are in docs/2026-09-12-g37-terskler.md.

Boundary recall: --fasit

Give the command the boundaries the source itself declares and it adds one whole-bundle row, boundary_share — the share of them that became a concept:

okf quality .okf/my-bundle --fasit declared-sections.json

The fasit is a JSON list whose rows carry title and the key they are matched on, norm (all whitespace stripped, lowercased). An unreadable one exits 2, never UNMEASURED. A boundary counts as recovered in either of two forms — a concept whose normalised title equals norm, or the pair of the concept's own directory and its residual title — because the numbering token a publisher glues into a heading lands in the concept id on one route and in the title on another: measured on one 2 761-section standard, the first form alone reaches 22 of 2 761 where the two together reach 2 759.

metric	threshold	N
boundary_share	2 759/2 761	2 761 declared boundaries, 1 corpus

--fasit is an assertion, the way okf consume --ref is: it says this bundle is a build of the document the fasit describes. A bundle of another product scores near zero and reads FAIL — that is the assertion being wrong, not the bundle. The bar itself rests on one product, which the output says on every run. Both facts, the arm it separates (1 148 of 2 761 against 2 759 of 2 761) and the interval any bar could sit in are in docs/2026-09-12-g37-terskler.md § 7.

Judge the retrieval: python3 tools/okf_retrieval_gate.py

A separate question from okf quality, and a separate command: quality asks what a bundle looks like, this asks whether the payload for a question carries the fasit — and whether the payload says so when it does not know.

python3 tools/okf_retrieval_gate.py          # nine rows, one exit code
python3 tools/okf_retrieval_gate.py --json   # the same rows as JSON

Nine rows, exit 0 only when every one is green, 1 otherwise, 2 for wrong input. Rows 1–4, 6 and 7 run against a synthetic corpus this repository generates and six question sets it ships, pinned by sha256: no network, no private corpus, no clock. A question set is always an input — sha256 is checked before a byte is measured and a mismatch is exit 2 — because a gold set names a consumer's documents and this repository is public. The corpus is pinned the same way (SPECS_SHA256): every row counts against those documents, so moving them without moving the pin is exit 2.

It is RED today, on rows 3, 4, 5, 7, 8 and 9, and each of those is a finding rather than a defect in the gate:

row	what it asks	today
1	hit@payload, one fasit entry = one unit	10 of 10
2	every miss carries exactly one class, each forced by its own fixture	7 of 7
3	the rule the payload prints for a withheld fasit is the true one	2 of 5
4	an uncovered question comes back marked, a covered one does not	3 of 6
5	a hold-out set, frozen and with its threshold written first	0 of 1
6	every delivery confirmed against the bundle's own bytes	10 of 10
7	mechanical mutants of the ranking and the cut, felled	12 of 14
8	the three real sets, from path + sha256	44 of 64 questions
9	K2	0 of 6, no gold set exists

Row 3 is the one to read first: in a bundle built from ONE source document, every concept past the first two carries that document's source_file, so a concept the RANK had already lost is withheld as source_quota_exceeded. The gate decides the truth with the same cut run without the quota, and a consumer reading rule today gets the wrong reason. Row 4 is the second: the payload has no key a consumer can read as "this bundle does not answer that", so an uncovered question comes back with excerpts and no statement — the only honest case today is the one where nothing matched at all and nothing was delivered.

Row 7 reports two survivors with what they moved rather than with a shrug: killing the document prior and flattening the fusion (RRF_K) each moved 0 ranks and 0 deliveries on these fixtures. Both have a mechanism — a question that names its document reaches it through the title-and-id signal as well, and 1/(K+r) is strictly decreasing in r for every K.

Rows 8 and 9 are never green by leaving something out, and since 2026-09-19 that is enforced rather than stated: row 8 requires all three named sets (wiki-20, r761-sk2, vegnormal-32) and is NOT RUN until it has them, whatever the ones that ran scored — one set of three used to read 6 of 6 GREEN. The sets live in other repositories and are read, never written: --real wiki <set.json> <sha256> <bundle> runs one, and --real vegnormal <set.json> <sha256> "N100=<bundle>,N200=<bundle>" runs one that spans bundles. Row 9 takes --k2 <set.json> <sha256> <bundle> in this gate's own set shape; without one it stays RED against its recorded denominator of six.

Granularity is stated on every line and the two forms are never summed: a set naming a citation is measured at citation granularity, a set naming only a section is measured at concept granularity. Row 8's own headline is therefore at QUESTION granularity — the one unit all three sets share — with the two unit totals printed below it, each with its own denominator. Measured 2026-09-19 against the three real sets on one machine: 44 of 64 questions, and below it 7 of 29 at citation granularity, 38 of 50 at concept granularity.

Consume in Claude Code

A folder of documents to an answer a model can cite, in three lines. You do not need this repository — the first line installs the command, the second builds the bundle and writes a skill beside it, the third asks.

uv tool install "llm-ingestion-okf[extract] @ git+https://git.fromaitochitta.com/open/llm-ingestion-okf.git@v0.10.1"
okf project ~/my-documents
claude

okf project writes the bundle to .okf/<id>/ and a skill to .claude/skills/<id>-consume/ in the current directory, then prints what it read, what it wrote, and which documents a question cannot reach. Start claude in that directory and ask in plain language; the generated skill runs the pre-pass and the contract check itself and marks every claim with its source.

Running non-interactively. In print mode the skill needs its tools named explicitly, or the model answers without ever reading the bundle and marks every premise undecidable-from-bundle:

claude -p --allowedTools=Bash,Read,Grep,Glob "<the question>"

Add Write,Edit for the mode that produces a document. --permission-mode acceptEdits alone does not do it — measured 2026-09-09 over four runs, the okf consume call is refused without the explicit tool list. The interactive claude above needs none of this.

<id> is the folder's name reduced to [a-z0-9-]. Run it once per folder with --id <name> to have several bundles reachable at once — each skill carries its own bundle_id, which is what lets a model pick between them. --out <dir> puts the project somewhere other than the current directory.

Measured 2026-09-09 from a fresh uv tool install with this repository nowhere on the path: 5 documents in, 26 concepts out, a skill carrying 0 paths into any checkout, and okf check conformant on its own payload (17 rules, 0 findings; the count was 15 until bundle_mismatch landed 2026-09-10 and 16 until parent_unfollowable landed 2026-09-11, and the pair was re-measured each time rather than carried over -- the last time from a frozen export of this repository, with 26 concepts and 0 checkout paths again). The 2026-09-08 run of the same measurement reported 15 concepts, and that number was the defect rather than the result: okf project was calling build() as a function and reading its signature's defaults, which disagreed with argparse's on two flags. Two tests now hold the two default sets equal. Before that day the same result took a PYTHONPATH, a snapshot of a clone, and a generated skill that named that clone by absolute path on four lines — so it could not be moved, shared, or run by anyone else.

The same thing in steps, if you want to see the payload

okf build ./documents --bundle ./bundle --bundle-id my-bundle --okf-version 0.2
okf skill ./bundle --out ./project/.claude/skills/my-bundle-consume
okf consume ./bundle --question "your question" --out /tmp/payload.json
okf check --skill ./project/.claude/skills/my-bundle-consume/SKILL.md --payload /tmp/payload.json

A bundle you only have read access to is fine — the generator only reads it.

The honest limits

Measured on four questions across two bundles, which is a demonstration and not a hit rate. The ranking is lexical, and one of the four found a topic the bundle does cover and did not rank it into the cut — the skill then said so with its denominator instead of answering, which is the behaviour the contract asks for, but a miss is still a miss. docs/2026-09-08-claude-code-skill-vilkaarlig-bundle.md has the runs.

Two more things the summary tells you and this paragraph will not repeat: a document that landed whole (no heading, table or numbered outline to cut it on) comes back as one excerpt, which the budget often refuses and which often does not carry the answer at the place you asked about; and a document that is in the folder but not in the bundle cannot be quoted at all. Both cases are answered [sourced-not-sufficient], and okf project names the documents.

The skill that runs this for you

skills/okf-prosjekt/ in this repository is a Claude Code skill (Norwegian) that wraps the command above: it takes a folder, runs okf project, and reads the summary back. Install it for your user account after cloning:

mkdir -p ~/.claude/skills && cp -R skills/okf-prosjekt ~/.claude/skills/

Serve a bundle over MCP: okf mcp

Two shapes, one implementation, and the difference is what an agent has to be told in advance.

okf mcp --bundle .okf/my-bundle          # one server, one bundle
okf mcp --root ~/bundles --root ./.okf   # one server, every bundle under the roots

--bundle serves exactly one bundle, fixed at startup; its tools take no bundle argument, because there is nothing to choose. --root (repeatable) serves every bundle found under the given directories and knows none of them by name: it discovers them per call, so a bundle you add, remove or rebuild while the server is running is picked up by the next call. No restart, no configuration edit, no code change.

Four tools, and each one's description says why it exists:

tool	what it answers
okf_list	which bundles are reachable right now, with each one's content identity and concept count (multi-bundle servers only)
okf_describe	what one bundle is: id, ref, concept count, source documents, and how many concepts carry each conditionally-written field
okf_ask	one question, one bounded payload of excerpts, each with its bundle id, concept id, title and provenance locators. Omitting bundle_id on a multi-bundle server asks them all and splits the budget
okf_fetch	one named concept, verbatim, with its frontmatter and locators

Nothing is cached between calls, and that is the design. Every call re-reads the directories and recomputes the bundle's content identity, so the identity in an answer is a fact about the bytes at the moment of the call rather than at startup — a server that answered from yesterday's bundle is the one failure you cannot see from the outside. The cost is real and is paid per call: on a 2 756-concept bundle the identity is a 0.75 s hash of the whole concept tree, and one okf_ask is 5.6 s.

Refusals are loud. A path climbing out of the bundle, a symlink leaving the served root, a bundle id nobody answers to, a directory whose manifest cannot be read, and a concept above the server's size ceiling each come back as an error with a code — never as a plausible-looking empty answer. A concept over the ceiling is refused whole rather than truncated: a truncated concept read as whole is a wrong answer that looks right. A directory that cannot be read as a bundle is reported in okf_list's unreadable, not skipped.

The protocol is written with the standard library only. An MCP SDK would be this package's second runtime dependency on the default install path, for four JSON-RPC methods and a newline framing — see Requirements.

tools/okf_mcp_gate.py is the eval: it starts the server as a subprocess, speaks real stdio to it, and measures six rows. It was written red before the server existed, and it is red today on row 2. The measurements, the update drill and the limits are in docs/2026-09-20-mcp-to-varianter.md.

One skill for every bundle: okf card and okf skill --generic

okf skill <bundle> writes a consumption skill for that bundle, with its identity and its numbers measured into the text — which is what makes the file stale the moment the bundle is rebuilt. okf skill --generic writes one installable skill for any bundle instead:

okf skill --generic --out ~/.claude/skills/okf-consume-any
okf card .okf/my-bundle          # the per-bundle numbers, as JSON, on demand

The generic skill carries no bundle's id, no ref and no count; it tells its reader to run okf card <bundle> first. The card is derived on every run and never written into the bundle, so there is no second artefact that can disagree with the bytes beside it.

Measured on two unrelated bundles: two per-bundle skills are identical on 280 of 312 and 310 lines. The 62 lines that differ are exactly identity, concept count, the conditional-field table, the whole-bundle cost and the breaking point — the five things a rebuild invalidates.

Implemented scope (v1)

The library provides three entry points for getting content into an OKF bundle:

1. Spec-based ingestion. An implementation of the normative ingest specification owned by portfolio-optimiser-commons: manifest → file/sql/http connector → deterministic materialization of ingest-{id}.md concept files → index generation. Zero model calls in the run path; output is reproducible byte-for-byte against golden fixtures.

2. Bundle inbox. A drop directory where common file types are converted to OKF concept files. All file-type→text extraction lives in this library: md, txt, csv, json, html and xml are handled by the stdlib core; pdf and the five office formats (docx, xlsx, pptx, odt, rtf) require the optional [extract] extra and are rejected fail-fast without it. Extracted text passes the security gate before anything is persisted. An xml file that declares NISO-STS structure (<standard> root, or any <sec>) becomes one heading per titled section, at the section's own nesting depth, with the section's <label> and <title> on one line; a <sec> carrying only a label is a body line and never a heading, and a <table-wrap> becomes one markdown table. Any other XML keeps its text in document order and gets no invented structure. XML carrying a <!DOCTYPE is refused unparsed. A NISO-STS document that states who it is names its own directory: okf build takes the directory from the document's one <std-ident> <doc-number> (reduced to the id grammar) instead of the file's stem, and the sources title from <doc-number> + <year>, then <title-wrap>, then the file name, whichever is the first that can be written verbatim. Stated more than once, or claimed by a second document in the same run, a declared name is not used and the file name stays. Each titled section's description is its own first spec point — the first <p> of its first direct-child <sec sec-type="spec">, whole — and a section with none gets no description at all; nothing is derived from the title. A point a YAML reader could not read verbatim (a : inside it) is left out rather than quoted or cleaned up. Measurements: docs/2026-09-11-k3-runde19-dokumentidentitet-og-frontmatter.md. The drop directory is walked recursively, in sorted relative-path order: a file at any depth is ingested and records its path relative to the inbox root as its source_file, while dot-directories and a bundle directory sitting inside the inbox are skipped with a reported code.

   Under the segmented v0.2 profile a concept also points back at the document it was extracted from, so an agent citing it can open the original at the right place: sources: [{ resource, title }] in the spec's own §5.1 form, where resource is the inbox-relative path, plus a locator per format — source_pages for a PDF, source_sheet and source_rows for a spreadsheet, source_lines otherwise. The locator keys are this library's own, because §5.1 has no field for a place within a resource; the line numbers index the extracted text and say so. Measurements: docs/2026-09-08-proveniens-k2.md.

3. External bundle import. Import and merge of third-party OKF bundles: each concept is assessed via the security gate, and only concepts that pass are merged, materialized, and linked into the index.

Boundary: security is delegated

Security is owned by the sibling package llm-ingestion-guard (pinned >=1.2,<2.0). The division is strict:

• guard answers "is this content safe to persist?" — scan, sanitize, quarantine, fail-secure, provenance stamping.
• this library does the plumbing — connect a source, materialize a deterministic OKF bundle, generate the index.

No security functionality is reimplemented here.

What is gated today: read this before trusting a door

• Door A (materialize_bundle) is ungated. It calls nothing before writing to disk and writes what it is given. A caller materializing untrusted content is responsible for gating it.

• Doors B and C (process_inbox, import_bundle) gate through an adapter you pass in. Each takes a gate argument; the flow hands it the content and obeys the verdict, refusing to persist anything that does not clear the guard's non-blocking floor — including a disposition it does not recognise, and (at Door C) a concept the gate returned no verdict for. What it cannot do is check that your adapter is a real guard: a permissive stub approves everything, and the flow will believe it.

• okf build runs a real guard by default, and names it in the bundle. --gate takes guard-trusted-source (the default), guard-user-upload or none, and the name is written into the bundle's log.md either way, so a consumer holding a bundle can tell a screened one from an unscreened one without asking. okf project has no such flag and takes the default.

  That paragraph is new, and the sentence above it was true of our own command until 2026-09-15: okf build injected a permissive stub and no argument anywhere in the package named a gate, so the only path most people use screened nothing while the guard sat in pyproject.toml as a mandatory runtime dependency. It was reported from outside, reproduced here, and the cost of each tier was measured before the default was chosen — over the 453 concept bodies of the pinned reference bundle, guard-trusted-source returns the persist disposition on 453 of 453 and guard-user-upload holds 1 of them. Neither waves anything through: an invisible carrier and a CRITICAL finding fail secure at both.

• --gate none is still reachable, by name. The corpus harness reproduces published numbers with it, and a caller measuring segmentation alone has a legitimate reason to take the gate out of the picture. What changed is that asking for it is an act, and the log says NOTHING WAS SCREENED.

llm_ingestion_okf.guard_adapter is the adapter over the real guard, and the only module here that imports it — importing the package itself does not:

from llm_ingestion_okf import process_inbox
from llm_ingestion_okf.guard_adapter import inbox_gate

result = process_inbox(inbox_dir, bundle_dir, "2026-07-25T12:00:00Z",
                       okf_type="reference", gate=inbox_gate)

Two properties of that adapter are worth knowing before you rely on it. It screens the exact bytes it persists — the guard's prepare_input bookend prepares text for a model call, which this library never makes, so only screen_output is used and the screened string is the written string. And it refuses rather than repairs: a file carrying an invisible zero-width or bidi character is rejected, not silently stripped and written. Door B screens under the untrusted-upload policy, so any finding at all is held back rather than persisted.

This section is stated plainly because earlier wording ("calls the guard at every persist gate") described the intended end state in the present tense, and a consumer reasonably read it as safe-by-default.

Roadmap

The library is built in four phases so that every known OKF surface in the ecosystem is eventually covered. Each phase has a detailed plan with verification criteria:

1. Spec-based ingestion (Python) with byte-exact golden fixtures — plan.
2. Bundle inbox and external-bundle import (Python), guard-gated — plan.
3. Configurable bundle contract (types, layers, frontmatter sets, index shape, and reserved-file policy as configuration), enabling stricter bundle profiles such as strict-v1 — plan.
4. A node/ half: a zero-dependency Node/ESM package (importable and CLI-invokable, vendored per consumer) providing bundle checking, index generation, inbox processing, and document conversion for the OKF second-brain plugin ecosystem. The Python and Node halves share the OKF contract and fixture suite, not code — plan.

Upstream OKF versions

The library targets the current latest version of Google's OKF. Support is additive — a new upstream version arrives as a new profile, never as a migration of an existing one — so an upstream release does not change the bytes an existing profile emits.

That guarantee is about upstream, and one profile tracks a second contract as well. DEFAULT states the ingest-spec owned by portfolio-optimiser-commons, so when they change that spec, DEFAULT follows them. It happened on 2026-08-09: generated moved from true to { by: process:okf-ingest, at: <ingested_at> }, one changed line per generated file. Upgrading across it costs a re-run and nothing more — a profile still recognises bundles stamped by earlier versions, so re-running writes in place instead of refusing. DEFAULT remains OKF v0.1 on every axis upstream owns.

Profile	Contract	Status
DEFAULT	commons' ingest-spec layer (OKF v0.1 semantics)	stable
STRICT_V1	a consumer's ratified v0.1 contract	stable
OKF_V0_2	OKF v0.2	provisional, pre-release only
STRUCTURED_V1	DEFAULT plus a faceted, derived index	stable
OKF_LATEST	alias for the latest version supported as stable	currently DEFAULT

STRUCTURED_V1 is DEFAULT in every respect but the index. Under it, Door B derives each dropped document's title, number, hierarchy and cross-references, writes them into the concept's own frontmatter, and carries them into the index entry — so a consumer can reason over the bundle rather than only look things up in it. Every inferred field is named in a derived list, because an unmarked heuristic is worse than no heuristic: the consumer cannot know when to doubt it. A pointer to a document not dropped yet is rendered N200? rather than omitted, since a bundle is built up over several drops and an absence that leaves no trace is the dangerous kind. Carrying the metadata costs index characters — roughly 3x to 6x the flat index, depending on how many facets the profile names — and the facet key set is the dial. Design record and measurements: docs/plan/structure-derivation.md.

OKF_V0_2 ships first as a pre-release to a named pilot set and may change on their feedback without a deprecation cycle. Pin the versioned constant rather than OKF_LATEST unless you have explicitly opted into tracking; OKF_LATEST moves at general availability, which is a deliberate release event rather than a side effect of an upgrade.

Selecting a profile is keyword-only, so existing call sites are unaffected:

materialize_bundle(manifest, bundle_dir, ingested_at, profile=OKF_V0_2)

A bundle may declare the version it targets. OKF v0.2 §12 makes this a MAY, and puts the declaration in the bundle-root index.md's frontmatter block. The profile names the key; the caller supplies the value, because that value tracks the upstream version and is not this library's to decide:

materialize_bundle(
    manifest, bundle_dir, ingested_at,
    profile=OKF_V0_2,
    root_frontmatter_values={"okf_version": "0.2"},
)

Omit the argument and no frontmatter block is written. Offering a key the profile does not name is refused before anything is written to disk.

Attested computations (v0.2 §10)

OKF_V0_2 supports the Attested Computation type as a format: its five contract fields — runtime, parameters, computation, executor, attester — are emitted in canonical position, judged, and round-tripped. runtime is required for that type and for no other, which the profile expresses through FrontmatterSchema.required_by_type; a type the mapping does not name carries no extra requirement, because §14 forbids a consumer to reject on an unknown type.

Nothing here executes a computation or checks an attestation. Upstream defers the receipt and verdict wire formats, so there is no contract to implement, and the question an attestation answers — was this value produced the sanctioned way — is not this library's. It re-enters scope when upstream specifies the protocol.

On the import side, a third-party concept may name an executor or attester resource pointing at executable code. Door C imports the pointer and never the code — it writes concepts verbatim and skips every non-.md file — so such a reference may not resolve, or may resolve to a file the destination tree already holds under that path. Each one is reported in ImportResult.unverified_references; the concept still merges, because §14 forbids rejecting a bundle over a broken cross-link while §10.5 asks a consumer to surface rather than silently drop. The report names the pointer key, not the resource it points at: recovering the resource needs the structured reader.

One limit worth knowing before you write such a concept: §10.2 presents executor and attester as nested block mappings, and this library's frontmatter parser is line-oriented. Which form it reads depends on the KEY, and the distinction is worth stating precisely, because this paragraph used to blur it:

• sources in either form. consume.read_sources has always read both, and since 2026-09-12 (K3-24) so do all three copies of the flat grammar (materialize.parse_frontmatter, which is public API, and the two internal ones). A block sequence of mappings is decoded into the flow rendering those readers round-trip; the entries never enter the document's key namespace. Measured against PyYAML 6.0.3 and the pinned guard 1.4.0 on 4 609 of 4 609 concept files carrying a block sources, all three readers return the same entries both references do.
• Every other key, flow only. executor, attester and any other block mapping are still skipped rather than parsed: two block mappings that both carry a resource would collapse into one namespace and the first would be lost, so they are refused instead. Reading them needs the structured reader (D1b).

Write the flow form, and know its limit: it is valid YAML only while every plain value inside it avoids what ends a flow scalar — ,, [, ], {, }, and for PyYAML also ? — as well as ": ", " #", a trailing : and a leading YAML indicator. Within that limit a YAML consumer recovers the same structure from either form. Beyond it no flow form works: quoting satisfies PyYAML, but the guard refuses a quoted value inside a flow mapping, so this library refuses such a value rather than write frontmatter a reader cannot parse. An earlier version of this paragraph said "both are valid YAML" without that limit; it was measured false for an unquoted URL with a query string. That refusal is also why K3-24 is a reading change only: this library still emits flow, because a named downstream consumer classifies a block sequence as unreadable provenance even though the guard reads it.

Non-goals

• Verdict/feedback machinery from the method specification (stays in the consuming repositories).
• Embedding- or retrieval-layer functionality.
• Security functionality, in either runtime — that is always llm-ingestion-guard's domain.

Requirements

Python 3.10+, and exactly one runtime dependency — the security boundary, llm-ingestion-guard>=1.2,<2.0. Everything else is stdlib. The commands are under Install; what follows is why they look the way they do.

From a checkout, the test suite runs with:

.venv/bin/python -m pytest

The suite is the verification surface for everything above: 1827 tests collected, 1826 passed and 1 skipped, run on 2026-09-12 with the [extract] extra installed. Both numbers are given because they are two measurements: the figure published before the v0.8.2 release was the PASSED count, and pytest --collect-only -q reported one more. (The figure stood at 596 until 2026-09-09 — measured 2026-08-21 and never updated as the suite grew — at 1515 until the v0.8.0 release, at 1575 through it, at 1582 through the v0.8.1 release, at 1602 through the v0.8.2 release, at 1658 after K3-19, at 1667 after K3-20 and through the v0.8.3 release, and at 1783 through the v0.8.4 release, after K3-22 and K3-21; the figure above is the v0.8.5 release's, after K3-23, K3-24 and K3-25: a count is a measurement with a date on it.) Without the extra the same suite skips the tests covering the parser path; that split was last counted on 2026-08-21 as 589 passed and 7 skipped and has not been re-measured since. The tests holding the fail-fast rejection for an uninstalled extra run in both. The suite is not shipped in an installed distribution — tests/ lives at the repository root, so this command needs a clone rather than a pip install.

The lint acceptance is BOTH of ruff's gates, and the rule set is declared. ruff check src tests tools and ruff format --check ., both named in a report with the version they ran under. Neither was true before 2026-09-09: the formatter gate was not in the acceptance and had gone red unseen, and [tool.ruff] set only line-length and target-version, so the acceptance was whatever ruff's default happened to be — which is why the tree read green only as long as uv.lock froze ruff at 0.15.22. Under 0.16.6 the same untouched code reported 148 findings, all of them new rules rather than new defects, because 0.16 widened the default set to whole families. select is now written down (E4, E7, E9, F, I, RUF100), the dev pin is ruff>=0.16.6,<0.17, and the wider families are a separate decision with 148 as its starting number. S is measured out rather than assumed out: it reports 2657 S101 on a suite whose every assertion is an assert. Add --extra extract to the sync or mypy src cannot find pdfplumber.

Markdown is excluded from ruff format. ruff 0.16 formats fenced Python inside markdown, and the two files it would change here are records rather than source — a README call example and a published measurement's quotation of COST_VOCABULARY as it stood when that measurement was taken. Reformatting a quotation makes it stop being one.

A git URL is a PEP 508 direct reference and pins one exact tag, so it is an install-time channel, not the pin: the range above stays the declared dependency — a wheel built from this branch carries Requires-Dist: llm-ingestion-guard<2.0,>=1.2, measured 2026-08-23 — and resolves normally once the package index exists. A wheel built from a tag carries that tag's range instead, which is why the install commands pair tag with tag.

Binary extraction

The optional [extract] extra ships two things: pdfplumber (MIT) for pdf, and pypandoc-binary for five office formats. It is opt-in because it pulls binary wheels, which the default install must never do — the single runtime dependency rule covers the default install and this extra sits outside it.

The converter binary travels inside the wheel and is resolved by path rather than found on PATH, with its version asserted against a pin. A host carrying a different converter is refused, not silently used: extraction is deterministic within a converter version and not across one.

These six rows have their reader and their evidence class in the one table this README carries: Supported file types, which is pinned to the extraction registry cell by cell. A second table here would be a copy nothing checks, and a copy of an evidence class is exactly the thing that goes false quietly.

constructed means what it says, and it is a weaker word than measured on purpose. The corpus this work was measured on contains zero pptx, odt and rtf files. Until 2026-09-09 those three rows were unmeasured — they worked by construction and had never been checked against a document anyone wrote. They have now each been put through end to end on a hand-built document with a hand-written fasit, which is more than nothing and is not a corpus:

• odt — 1 of 1 declared headings recovered, 1 concept, 0 characters in no segment. N = 1 document.
• pptx — 2 of 2 declared slide titles recovered on a deck that declares them (a real <p:ph type="title"/> placeholder); 0 of 2 on a deck that does not, where the converter writes Slide 1 / Slide 2 because it has no title to use. That is the converter naming an unnamed slide, not a segmentation failure. N = 2 decks.
• rtf — 0 declared headings, because the container has no heading style and the author's title is bold text. The proposer therefore proposes nothing, and the document reaches the bundle inbox as one whole concept: content preserved, structure zero. N = 1 document. This is the one open finding of the three.

They are not known to be broken; a single constructed document is not a denominator, and the distinction is the point.

What stays out. .doc (Word 97) is not supported — the converter does not read it. Rastered or scanned PDFs are refused rather than persisted as empty concepts, because this library does not do OCR. Drawn content — figures, diagrams, shapes — does not survive extraction in any format here, and every extraction says so with a warning. Structured table recovery is out of scope.

Request it by appending [extract] to the package name in whichever install command from Install you are using — this package is not on an index, so a bare pip install 'llm-ingestion-okf[extract]' does not work today, and the error message naming that command is written for the day it does. The extra is unreleased: it reaches a consumer through a tag that contains it, and no such tag exists yet.

Two properties of the extra are worth knowing before depending on its output:

• Extracted text is pinned to an exact parser version. pdfplumber pins pdfminer.six==20260107 exactly, and pdfminer.six ships date-stamped releases with no stability contract. Extraction is deterministic within a parser version and not guaranteed across one, so a golden fixture built on extracted PDF text is a fixture migration away from any parser upgrade.
• Text extraction recovers text, and nothing that is drawn. Figures, diagrams and images have no text to recover — only their captions survive — so a bundle built from drawn documents is incomplete by construction. The library says so itself: every pdf extraction emits an ExtractionWarning. Structured table recovery is separately out of scope; PDFs enter as prose.

The planned Node half targets Node/ESM with zero npm dependencies.

License

MIT — see LICENSE.