Operator's 2026-09-23 log: embed_image taking 107-246s each, ~49 slots in
flight by Little's law, and the daily CCIP sweep dying on its 1800s soft
limit in a numpy matmul. The billiard/pool.py frame in that traceback is
the soft-timeout signal handler, not a pool fault.
Two causes, both mine.
1. `derived_ceiling` computed the ML lane from MEMORY ALONE. Meanwhile
`embedder.py` carried `_INTRA_OP_THREADS = 4` beside a comment reading
"keep N_replicas x this within the cores allotted to ML" — a constraint
stated where nothing could act on it. A large-memory host offered ~49
slots, the operator took what the dial offered, and the lane asked the
box for ~200 torch threads.
The number moves onto the lane as `threads_per_slot`, the embedder
reads it rather than restating it, and the ceiling is now the smaller
of the two bounds. They fail differently on purpose: too little memory
is honestly zero, because the first task would OOM the container; too
few cores is merely slow, so it floors at one rather than making the
lane unreachable on a small box.
2. `scheduled_ccip_auto_apply` scored one image per matmul, over every
image in the library, on every daily run — ~119k products each too
small to pay for its own BLAS setup. `char_maxima` does the same
arithmetic in blocks bounded by elements, so its memory stays flat as
either axis grows.
Batching changes no arithmetic: a character's score for an image is a
max over that image's figures AND that character's prototypes, and max
does not care how it is grouped. Pinned against the old loop written
out longhand, and against itself with the blocking forced to split
every row.
The UI copy said the ML ceiling came from memory; it says cores or
memory, whichever runs out first.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LVjrnpQjRgHdvq95rASoiR
Completes "no self-training": unconfirmed auto-applied character tags no longer
seed CCIP references — character_references + the prototype builder
(_current_fingerprints/_rebuild_one) gain a shared _positive_char_tag filter
(human-applied OR operator-confirmed), mirroring the head-positive exclusion.
Confirming a tag also has to move the change-detectors, or an incremental
refresh/Retrain right after a confirm wouldn't fold the tag in (only the nightly
full pass would): the CCIP global gate now counts character confirmations, and
the head training fingerprint counts confirmations. Test for the CCIP path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM
match_image now sources character references from character_prototype via a
per-character in-process cache (_load_prototypes) that reloads ONLY the
characters whose ccip_prototype_state.updated_at advanced — no request-path
rebuild, so the per-accept ~4s stall is gone once the store is populated. Cold
start (store empty pre-first-refresh) falls back to the legacy on-the-fly
reference build, so character suggestions work immediately post-deploy and the
background refresh populates the store within ~15 min. Match math + grounding
are unchanged; existing tests exercise the legacy fallback, and a new test
covers matching from the populated prototype store.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM
match_image now tracks WHICH query figure produced the winning cosine per
character (argmax over the per-figure best-reference sim) and attaches its bbox as
grounding {bbox,kind:'figure',detector}. SuggestionService carries it: a CCIP-only
character hit grounds to its figure; a 'both' hit keeps the head's localized crop
if it had one, else falls back to the CCIP figure — so corroborated characters
stay grounded. Test: a character match carries the matched figure's bbox+kind.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM
Step 2 of milestone #128. _hygiene_excluded_ids (training_data.py) is the
one shared predicate: images carrying any system tag are dropped from
every OTHER concepts head training — not positives (a rough wip tagged
as a character drags the head toward generic-sketch) and not rejection
or sampled negatives (a wip OF character X is not evidence against X).
A system tags own head trains on them unfiltered; that is what makes
auto-flagging banners work. Selection is split out of train_head as the
sklearn-free head_training_ids so CI (no sklearn) can pin the behavior.
CCIP: reference prototypes skip hygiene-tagged images — a faceless wip
figure region must never become an identity reference — and the ref
cache signature now counts hygiene applications, since tagging an image
wip changes the reference set without touching character/region counts.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM
Works through the optional CCIP ideas + the "keep moving even if I forget" ask:
AUTOMATION (no button needed):
- Hourly beat auto-enqueues CCIP backfill — new images get embedded (and errored
ones retried) on their own; the queue never goes idle waiting for a click.
- CCIP auto-apply: a daily sweep tags confident matches (source='ccip_auto') so
identity tags keep flowing. ON by default (opt-out, like head auto-apply);
ml_settings.ccip_auto_apply_enabled + _threshold (0.92, above the suggest cut),
migration 0064. Vectorized (one matmul + reduceat per image), reversible, skips
already-applied/rejected. Switch + threshold in the GPU agent card; GET/PATCH
/api/ml/settings; auto_applied count in /api/ccip/overview.
REFERENCE QUALITY (the over-fire root cause):
- character_references now draws ONLY from single-character images — on a
multi-character image the tag is image-level, so every figure would otherwise
pollute each character's prototypes (a 2-char image tagged 'Velma' made
Daphne's figure a Velma reference). This is the contamination behind residual
over-firing.
- Cached on a cheap signature (char-tag count + ccip-region count/max-id) so the
reference load isn't redone on every modal open.
Tests: multi-character image not used as a reference; auto-apply tags a confident
match as ccip_auto.
NEXT (not done, confirmed): comic-panel cropping + SigLIP concept crops ("spot
interesting content").
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
Live data showed the v1 flat 0.75 cosine over-fired — ~64% of matched images got
3-10 character guesses dominated by the most-referenced characters (a 27-ref
character clears a low bar on many images). A sweep showed 0.85 collapses the
noise (noisy multi-matches 47→3) while keeping the confident single-character
matches.
- ml_settings.ccip_match_threshold (migration 0063, default 0.85); match_image
reads it (override still accepted). DEFAULT_SIM_THRESHOLD fallback 0.75→0.85.
- Exposed in GET/PATCH /api/ml/settings (validated 0.5–0.999).
- Slider in the GPU agent card ("Character-match strictness") — tune live, no
redeploy, same observe-and-tune loop as auto-apply.
Test: a ~0.9-cosine figure matches at 0.85, dropped at 0.95.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
The server-side brain that turns stored CCIP vectors into character suggestions
— no GPU. character_references() gathers each character tag's prototype vectors
(figure/face-region CCIP embeddings on images carrying that tag); match_image()
cosine-matches an image's figure vectors against every character (multi-
prototype: best over a character's examples), surfacing those above a tunable
threshold as {tag_id, name, category:'character', score, source:'ccip'},
excluding already-applied characters. v1 = cosine on raw CCIP vectors; the exact
CCIP metric/threshold gets validated against the model in the hands-on eval.
Tests (synthetic vectors): same-character match across images, no-match for an
orthogonal figure, already-applied exclusion, no-figure-vectors empty.
NEXT: merge CCIP character suggestions into the rail; the agent container that
actually produces the vectors (hands-on, GPU — not CI-verifiable).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
bvandeusenandClaude Opus 5