The tool is moving into the jray-project submodule to be shared by
scene-actor-extraction (C++/Python), jRay (C#) and JRay-public-server
(Rust). Two constants blocked that: LOCAL_TYPES and SOURCE_SUFFIXES were
hardcoded to this repo, so either sibling parsed zero requirements and
scanned zero files. Both, plus the register path, scan roots, system-spec
path, exclude list and CI-executable tier set, are now configuration.
One implementation, parameterised. A second copy for "the other language"
is how two implementations start drifting apart, so there is exactly one -
the same code path now produces:
scene-actor-extraction AR/DP/IR/GR/VR 59 defined 0 tagged 0.0%
jRay JR 46 defined 24 covered 52.2%
JRay-public-server UR/DR 32 defined 23 covered 71.9%
Configuration is traceability.toml at the component repo root, CLI flags,
or both (flags win). Its directory defines the repo root, so the gate works
from any subdirectory. `--print-example-config` emits the annotated schema.
The JSON report echoes the settings it ran with, since a shared tool's
output is otherwise ambiguous about which repo it describes.
The refusal behaviour is kept and sharpened, because parameterising is
exactly what makes it easy to point a repo at the wrong prefixes or the
wrong suffixes. Zero requirements parsed or zero files scanned is still a
hard failure, and the message now names the setting that is wrong rather
than printing a plausible 0%. Config errors exit 2, not 1: a broken config
is not a coverage failure, and conflating them makes CI logs lie about why
the job went red.
Also fixed while adapting to the sibling registers, which are read but not
modified here:
* escaped `\|` inside a markdown cell no longer shifts every later column
(the server's register contains `small-\|M\|`);
* a tag above an attribute-decorated declaration attributes to the
declaration, not to `[HttpGet(...)]` or `#[derive(...)]` - the gap
jRay's register calls out;
* Rust and C# declaration patterns for context extraction;
* the missing-tier warning is suppressed for a register that assigns no
tiers at all, rather than listing every requirement in it.
The workflow is now component-agnostic too: the changed-file check reads
its extension list out of the report the gate just wrote, so the definition
of "source file" lives in one place.
79 tests, still fixture-based, now including the cross-repo cases: the same
parser over JR and UR/DR registers, the same scanner over Rust and C#, and
both misconfigurations failing loudly.
Ports JellyTau's traceability tooling, rewritten in stdlib Python because
this repo is C++/Python and adding a bun/node toolchain to check source
comments would be a worse trade than writing the scanner.
scripts/traceability/extract_traces.py scans .cpp/.hpp/.py under src, tests,
scripts, experiments and eval for the house tag format
/// TRACES: AR-012, AR-013 | SR-002
and reports EXCEPTION tags separately. An exception is a recorded decision to
depart from an invariant, so folding it into coverage would invert its
meaning; it is listed with its reason, and a missing reason is flagged.
Two rules carried over from JellyTau's gate repair:
* Denominators are parsed out of docs/requirements.md at run time. A
requirement is defined only by a row in a table whose header is
`| ID | Requirement | ... |`, so references in the Traces to column, in
prose, and in the verification-plan table do not inflate the count.
Adding a register row lowers coverage until it is traced - the property
that dies the moment a denominator is frozen.
* Coverage above 100% is a hard failure. It cannot happen through the
intersection, which is the point: if it ever does, the arithmetic is
broken and the run must not be reported as a pass.
One rule specific to this repo: CI is an Intel N100 with no discrete GPU. The
extractor reads each requirement's verification tier from requirements.md and
reports T4/GPU-only requirements as tagged but unexecuted, never as covered.
Counting a test that can never run is the same failure mode as the 158% bug.
MIN_COVERAGE starts at 0 because almost nothing is tagged yet - tags are added
as the pipeline is built. That is not a gate that cannot fail: orphan tags, a
>100% ratio, a register that parses to nothing, and an empty source scan are
all hard failures from day one. The threshold lives in traceability-gate.sh
alone, never duplicated into the workflow YAML.
53 tests over fixture strings, so their meaning does not drift as requirements
are added.
The phase structure encoded ordering assumptions that stopped being true as the
design changed, and its Phase 2 still described retuning constants that are now
withdrawn. Ordering is now derived from per-requirement dependencies instead:
anything with no unmet dependency is startable.
Carries over the TrackRegistry design (now keyed to AR-012/AR-013) and records
what was withdrawn from the old plan, including the --presence-mode flag —
comparison against old behaviour uses recorded reference output rather than a
second live code path.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds the requirements baseline for the pipeline redesign:
- SPEC.md — software requirements with Current/Gap deltas per item, so the
document doubles as a work list.
- requirements.md — stable flat IDs (AR/DP/IR/GR/VR) with parent traces,
priorities, statuses, and a per-requirement verification plan. Replaces the
thematic A1..E8 scheme, which had already produced an A1a and an out-of-order
E6; IDs are now permanent and never reused.
- IMPLEMENTATION-PLAN.md — phased work.
The central change is AR-012: presence follows track extent rather than
per-frame recognition, so a window starts when an actor appears rather than
when the recogniser first succeeded. anneal_sec and extinction_sec are
withdrawn rather than retuned — a track that survives its own gaps leaves them
nothing to do.
Verification is shaped by CI running on an N100 with no dGPU: the existing
HDF5 dump makes everything downstream of embedding replayable on CPU, which
covers the bulk of the redesign.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Her bbox is frozen at identical coordinates for t=2450 and t=2451; the
dump's own per-frame detections show only one real face at t=2451, and
it matches the Chloë Sevigny box (IoU 1.0), not hers. The frame is one
ghost overlapping one fresh misidentification, not two competing fresh
identities as previously written.
Replaces narrative claims with verified numbers across all report pages:
- Cross-model held-out validation (LVFace/mbf/r18, all 5 held-out
films): LVFace wins every film outright, not just "consistent with"
the training-set pick. r50 dropped from the detailed comparison
(gallery has ~30% fewer reference images per actor than the other
three models on identical source photos).
- Per-film training breakdown: LVFace does not win every training
film (mbf beats it on Lord of War); the 75.3% macro figure hides a
10.7pp spread.
- Gallery coverage computed per film (20.3%-78.6%) instead of one
flat 67%-missing average.
- Found and fixed a real scoring bug in optimize.py: a candidate
whose hardest film's replay timed out was averaged over survivors
instead of penalized, silently rewarding partial coverage. Affected
3 of 16 training combos; corrected throughout, and optimize.py now
scores an incomplete evaluation f1=0.0 instead of averaging over
whichever films happened to finish.
- Every FPI frame in the deep dive now comes from the proper montage
renderer (Onscreen/Offscreen panel, ghosts never drawn as boxes),
never the bare-box debug overlay used earlier.
- Every distinct out-of-cast name across all 9 films gets its own
frame at its first appearance (9 names, 4 films), not a
single-example spot check: 2 ground-truth gaps, 1 photograph
misread as a person, 6 genuine lookalike confusions.
- New methodology.md: the scene-level-vs-per-second scoring mismatch
that the rest of the report assumes, written out once.
- Cut the deadlock/gdb debugging narrative from the experiment log;
kept the one fact that matters (KPN's node/network split lets the
expensive GPU stage run once and the cheap stage replay against
cached embeddings).
- Plain declarative style throughout, no em dashes, no blog voice.
- switch report frames to the scene best/worst montage renderer
(Onscreen/Offscreen panels + TPI/FPI/FN legend): perfect-second hero,
wedding couple, funeral 19-of-20, polygraph bridging, crew-scene FN
ceiling, Robert Patrick ground-truth gap, rapid-cut double label,
Herbie Hancock on an in-fiction screen
- deep dive restructured: extinction bridging framed as designed
behavior with a measurable cost (debug overlay draws the boxes; the
shipped output is presence windows), plus the face-vs-presence
ceiling and two X-Ray-is-wrong exhibits
- Material polish: light/dark palette toggle, landing-page grid cards,
figure/caption CSS, how-to-read admonition; site_url set so 404 links
resolve under the Pages subpath
- README: perfect-second and screen-call frames committed (gitignore
exceptions), readme_example.jpg retired
- build_site.sh: stage_frame helper downscales montage frames to 1920px
and pulls any missing montage-frames packages
- experiment_charts.py generates 4 figures from experiments/ artifacts:
held-out per-film F1, 16-combo ranking, DE search landscape, and the
Downton detector-vs-tracker ghost timeline (replaces the blank
title-card screenshot)
- new frames: 19-correct wedding shot (success case), Many Saints
ghost-vs-unknown frame (three error classes in one image)
- rename rep4-optimizer-results.md -> model-bakeoff.md; rep4 kept only
as the on-disk artifact prefix, explained once
- repo file references are now links via https://REPOLINK/<path>
placeholders; build_site.sh pins them to the HEAD commit's raw URLs
and fails the build if a linked path doesn't exist at HEAD
- drop references to removed scripts (scene_score.py, score_config.py)
and to session-memory names; mark artifact-registry paths with their
pull commands
- commit readme_example.jpg + pipeline_topology.svg so README renders
on the plain Gitea repo view
- deploy_pages.sh: push built site/ to the gitea-pages branch
Splits the rep4 write-up's key findings into their own linkable pages:
- best-model.md: calibration curves first (discriminative power, independent
of any threshold), then F1 on the benchmark — LVFace-B Glint360K wins both.
- gallery-scope.md: whole vs. cast-restricted gallery, isolated from model and
expansion choice — restriction wins on every axis, but isn't a shipped
runtime feature yet.
- pose-expansion.md: the training-set expand_gallery effect, and the held-out
replication attempt that found it doesn't reproduce (5 films, 2 models,
after catching and fixing a replay-timeout truncation bug and a bbox
first-match-instead-of-best-match bug in the comparison harness itself). An
honest null result, with the methodology errors documented since they're
exactly the kind that manufacture a false "it works!" finding.
- lvface-deep-dive.md: the winning model's held-out generalization gap, its
two failure modes (frozen-bbox ghost tracks), and a verified case (cross-
checked against Jellyfin's independent cast metadata) where LVFace
correctly identified an actor that X-Ray's ground truth failed to credit.
Adds a "report-highlights" artifact-registry package (scripts/artifacts/
push_artifacts.sh, pull_artifacts.sh) for hand-picked illustrative frames that
aren't reproducible via the automated best/worst montage selection, and wires
pulling it into scripts/docs/build_site.sh.
docs/rep4-optimizer-results.md is the main deliverable: the model bake-off +
threshold re-tune experiment log, including the ROCm teardown deadlock root
cause and fix, DE concurrency tuning, the 16-combo results table, held-out
validation against 5 films never seen by the optimizer (macro F1 67.4% vs.
75.3% training — a real generalization gap), the frozen-bbox "ghost track"
failure mode found via annotated frame evidence, calibration curves per model,
and an isolated-effects breakdown of gallery scope vs. pose expansion.
MkDocs site (mkdocs.yml, docs/index.md) renders docs/*.md; scripts/docs/
pulls referenced images from the artifact registry and generates the
calibration chart at build time (see the tooling commit) rather than
committing images to the repo.
experiments/ now keeps only scripts + README + SESSION_STATE.md in git — every
data artifact (galleries, dumps, X-Ray corpus, montage frames, trajectories,
manifests, results) moved to the Gitea package registry. film-lut.template.json
is the committed placeholder for the gitignored file-lut.json (real local
movie paths, never shared — some source filenames carry scene-release tags).
Adds models/transnetv2.onnx (via Git LFS, matching the other ONNX models) for
the new scene-detection path.
dtourolle
