feat(assets): a bundle carries the images its sources declare (0.10.0)

Until now no reader in this package fetched, named, described or copied a
single image. `<img>`'s attributes were never read, a NISO-STS `<graphic>`
was walked past, a PDF was opened for its text alone, the converter's
markdown writer dropped every picture, and the only writer into a bundle
took `content: str`. The two lossiness warnings said so on every run, which
made the loss honest and did not make it smaller.

Measured on R761 Prosesskoden:2025, published as a 701-page PDF and as a
NISO-STS delivery: the process text is carried in full while 12 `Tabell N-N`
and 9 `Figur N-N` captions stand over nothing, because that publisher ships
those tables as raster pictures in both. Process 84's "toleranseklasse ...
er gitt i tabell 84-2" points at empty space.

THE GATE WAS WRITTEN FIRST AND RED. `tests/test_asset_gate.py` reads its
denominator out of the source (`page.images`, `word/media/`, `ppt/media/`,
`<img`, `<graphic`), never from a constant here. Measured at 332961a, built
from `git archive` and not from the editable tree: carried 0 of 8 local
images across 5 documents (9 declared), and no `assets/` at all. After: 8 of
8, with the ninth a remote source carried as a pointer without a file.

FIVE READERS PLACE, ONE MODULE DECIDES. `assets.py` owns what an image is
(sniffed from the bytes, never from the claimed extension), what it is
called (`<sha256[:12]>-<the source's own basename>`) and how it is pointed
at (one two-line block, one regex). `.xlsx` is deliberately not a row: a
block inside its pipe tables would break the `source_rows` locator, and 0 of
4 K2 workbooks hold media.

A PDF stream that is already a file is carried VERBATIM (29 of R761's 50
objects are DCTDecode); raw samples are encoded to PNG with stdlib zlib, so
no new dependency. Rendering the page region was the alternative and was
felled on determinism: a rasterised crop's bytes, and therefore the asset's
content-addressed name and the bundle's digest, would depend on the
installed rasteriser. What the encoder cannot express exactly is refused
with a code and counted, never approximated.

NO SIZE FLOOR, and that is a measurement: over the 4 828 image objects of
the K2 corpus the size distribution is a broad spread with no gap, unlike
OCR_CID_SHARE's bimodal one, so a threshold would be a number we chose.

ON BY DEFAULT, AND THE CONTROL IS TWO WHOLE BUILDS. The 43-document
reference corpus at 332961a versus rebuilt at HEAD with `--no-assets`:
865 files on both sides, `diff -rq` reports ONE difference, the added
`Images: NOT CARRIED` line in log.md. Every concept byte-identical.
Against the default: 453 -> 454 concepts, 865 -> 867 md, 0 -> 2 964 assets
(2 964 carried of 3 145 found, 4 622 pointers), 4.7 MB -> 115 MB, 2 414 s ->
3 088 s, peak RSS 6.26 -> 8.74 GB, 422 of 865 md files differ. The one new
concept has a measured cause: 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.

THE IMAGE BYTES ARE NOT SCREENED. The guard is text-only, the pointer block
passes the gate as body text, the picture beside it passes nothing, and
log.md says so on every run.

Also fixed, both found by measuring rather than by reading:

- a markdown image is no longer read as a cross-reference. `structure._LINK`
  never looked at the character in front of the bracket, so every pointer
  would have arrived in the index as an edge to a concept that cannot exist.
- Door C carries the assets its merged concepts point at. Before this,
  importing a bundle built with `--assets` merged 6 of 6 concepts and wrote
  no `assets/` at all, so every pointer named a missing file.

Report: docs/2026-09-17-bilder-i-bundlen-trinn1.md
Spec proposal: docs/plan/okf-assets-section-6-4.md
Suite 1 955 passed / 1 skipped (from 1 896), ruff and mypy --strict clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/llm_ingestion_okf/errors.py

@@ -108,6 +108,31 @@ class ExtractionError(IngestError):
       `extractor_extra_missing`: the `[extract]` extra can be fully installed
       and the document parsed, with only the OCR engine absent, and one code
       for both would send an operator to reinstall what they already have
+
+    ASSET codes (0.10.0). None of these fails a document: an image a reader
+    could not carry is a ROW in the run log and a line in the concept saying
+    what was there, because one unreadable picture must not cost the three
+    thousand concepts of text around it.
+
+    - `asset_type_unknown` — the bytes behind a pointer are not an image format
+      this package recognises. Sniffed from the bytes, never from the claimed
+      extension: a `.jpg` that is really a PNG would otherwise be written under
+      a name whose extension lies
+    - `asset_unresolved` — the file a document points at was not found beside
+      it, or an inline data URI could not be decoded. Containment is the
+      document's own directory, so a reference above it lands here rather than
+      being followed
+    - `asset_remote` — the source is off this machine. Extraction opens no
+      socket: network access is an explicit per-run opt-in and extraction is not
+      on that path, so a remote image is carried as a pointer and never as bytes
+    - `asset_pdf_unsupported` — a PDF image whose samples this encoder does not
+      express: a stencil mask, a `Decode` array, a colour space with no exact
+      PNG form, a soft mask that cannot be carried, or anything but 8-bit
+      samples. Refused rather than approximated, because a picture that is
+      plausibly the wrong colour is wrong in a way no consumer can detect
+    - `asset_samples_invalid` — the sample buffer does not fit the dimensions
+      the image dictionary declares. Refused rather than padded: a short buffer
+      means the dictionary was read wrong
     """
 
 
@@ -130,6 +155,12 @@ class MaterializationError(IngestError):
     - `ingested_at_invalid` — ingested_at is not ISO-8601 UTC with a Z suffix
     - `collision_unstamped` — the §3 collision gate: a generated name is
       occupied by a file without the ingest stamp
+    - `asset_collision` — two different images reduce to one asset name in one
+      run, or a name in `assets/` is occupied by different bytes (Doors B and
+      C, 0.10.0). An asset name carries the digest of its own contents, so this
+      is a `sha256[:12]` collision; refused rather than resolved, because
+      resolving it silently means one of two pictures is lost and every pointer
+      to it shows the other
     - `source_reference_unquotable` — a manifest source's id or locator
       contains a character that would restructure the `sources` flow mapping
       (Door A, v0.2 profiles); refused rather than emitted, because the