Files
manga-recap-pipeline/decisions/CLAUDE.md
T
kami de7e34bc0f File the over-merge findings: two tracklet fixes, and cosine is not identity
The user checked the lead's crops. LIM SEONHO's 36 assignments cover at least
six different men, a chibi and a cat, so 36 of 68 is a bug. 22 came from
native resolves and 14 from reconcile merges.

Two fixes are written in the orchestrator repo, tested, not deployed and not
run on a GPU. `link_tracklets` caps a tracklet's panel span, because `window`
bounded each pair while transitivity was unbounded and the lead's 22 native
assignments came from tracklets spanning 22 and 30 panels. And one shared
appearance tokenizer drops generic words, because whole chains hung on the
word `short` and one pair on the word `hair`.

The obvious third fix is ruled out by measurement. Over all 22 crop
embeddings, the cat scores up to 0.82 against men, two different men score
0.93, and the highest pair is 0.96. No threshold separates them, so
crop-to-crop cosine is not a link signal.

Item 1 of the agreed plan, sending the live cast instead of a cosine top-k
gallery, is not started.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-12 23:35:15 +04:00

4.4 KiB

decisions/

One entry per settled question about this pipeline. An entry names the claim, the evidence, the date, and what it forbids or permits next.

No entry is a plan. NEXT.md holds the plan, ROADMAP.md holds the horizon, and a limitation that is still live belongs in caveats/.

Rules for this directory

  • One file per topic, one ## section per question, so the index can link an anchor.
  • An entry cites a file and line, a commit, or a test. No citation means it is an opinion, and an opinion gets deleted rather than defended.
  • Closed means acting on it is safe. Open means investigated and undecided. Void means the evidence turned out to be invalid and the claim must not be cited again.
  • Rewrite an entry when something contradicts it, and say what changed.

Index

decision state
A dialogue row's speaker is speaker_ref, not speaker closed
Script verification must not fail on sentence-initial capitals or short quotes closed
Tracklet links are transitive over similarity, never across a hard constraint closed
A detected face takes an identity only when it falls inside that character's box closed
A 409 from the GPU mutex is a queue signal, not a stale lease closed
Correctness flags resolve by flag id, and never block autonomous TTS closed
An out-of-range resolver index is unresolved, never a new character closed
The session manager holds no lock across a model load closed
One bucket per artifact class, not everything under manga closed
The orchestrator creates missing buckets at startup closed
RustFS is staged, not adopted open
A model guess is never labelled tail closed
The model's speaker answer is resolved against what the prompt showed closed
An unpaired set-of-mark label grounds nothing closed
An interjection is not a name and not a misquote closed
Cast names enter the verifier tokenized closed
Clearing a stage strips the vision blob it wrote closed
The shipped 72.7s gap was a stream copy across mixed frame rates closed
xfade offsets come from min(video, audio), never format=duration closed
A vision bbox is gemma's 0-1000 grid, converted to pixels at /vision closed
Assembly verifies its own output instead of trusting ffmpeg's exit code closed
The single-item passthrough is not the assembly bug void
A stage result proves nothing until the worker is newer than the edit closed
A detection with no detected face never enrolls or binds closed
A resolver NONE mints an anonymous character, it does not clear the crop closed
The extras gate runs at enrollment and at narration, not at the speaker prompt closed
A roster name is a guess, so it never reaches detection closed
merged_into is exactly one hop deep closed
_bbox_to_pixels orders the corners, because the model sometimes swaps them closed
A tracklet is bounded by span, not only by pairwise distance closed, GPU pending
A generic word is not identity evidence, and one tokenizer serves both consumers closed, GPU pending