Commit graph

12 commits

Author SHA1 Message Date
changukyu
4f140efd9b feat: 실사용 투입 전 T0 스탬프 4종 + 임베딩 정책 개정
두 갈래의 작업이 한 커밋에 들어 있다. llm_assistance.py 에서 T0-4 가 앞선
작업의 프롬프트 수정을 대체했기 때문에(프롬프트 리터럴이 llm_prompts.py 로
이동), 분리하려면 실제로 존재한 적 없는 중간 상태를 만들어야 한다.

## (1) 앞서 진행 중이던 review-intelligence 작업

- decision_stats(): 위험밴드 x 실제판정 교차표 + 캘리브레이션 지표
- cleared_reference 지식엔트리 타입: 이전에 클리어된 이미지 매칭은 +80 이
  아니라 0점 긍정 신호
- LLM 요약 본문 영속화(_evidence_payload 에 summary/sourceUrls) — 그전까지
  요약이 저장되지 않아 운영자에게 한 번도 보인 적이 없었다
- 운영자 GUI 반영

## (2) 실사용 투입 전 T0 — 첫 판정 이후에는 소급 불가한 것들

- T0-1 initialRiskScore/initialRiskBand/scorerVersion 각인.
  _rescore_submission 이 12곳에서 riskScore 를 덮어쓰므로 저장된 점수는 늘
  왕복 점수였고, 그것을 재채점과 비교하는 테스트는 항상 통과하는 빈 테스트다.
  레거시 행은 scorerVersion="legacy" 로 표시만 하고 점수를 날조하지 않는다.
- T0-2 얼굴검출 Haar -> YuNet, 인물 등급(identifiable/background) 도입.
  기존의 평평한 +35 는 사람이 찍힌 모든 사진을 medium 문턱 위로 올려 밴드
  분포를 붕괴시켰다(실측 high 4 / medium 2 / low 0). 등급은 payload 라운드
  트립으로 보존한다 — 잃으면 재채점마다 조용히 +35 로 되돌아간다.
  가중치는 models/(gitignore) 에 두고 offline-install.md 에 sha256 반입 절차.
  가중치가 없으면 Haar 로 강등하고 1회만 경고한다.
- T0-3 판정 시점 피처 스냅샷. 큐가 riskScore 내림차순이고 화면이 점수를 먼저
  보여준 뒤 판정을 받으므로, 스냅샷 없이 모은 라벨은 "이 근거가 반려를
  예측한다"가 아니라 "운영자가 내가 보여준 점수에 동의했다"를 기록한다.
- T0-4 모델/프롬프트 버전 + latency_ms 스탬프. latency_ms 는 _evidence_id
  해시에서 제외한다 — 포함하면 같은 요약이 재생성될 때마다 id 가 바뀐다.

## (3) 임베딩 저장 금지 정책 개정

금지의 취지는 생체 식별이지 이미지 유사도가 아니므로 범위를 좁힌다.
장면 임베딩(전체 이미지 의미 벡터)을 이미지 유사도 판정에 한해 허용하되
얼굴 마스킹 / 인물 지배 프레임 제외 / 목적 제한 / 별도 저장 클래스 /
연쇄 삭제 5조건을 assert_scene_embedding_allowed 로 강제한다. 마스킹 증명이
없으면 거부한다 — 누락은 마스킹을 건너뛴 것과 구별할 수 없다.
얼굴 임베딩 금지는 landmarks/face_descriptor/face_encoding 추가로 강화.

pytest 502 passed (착수 전 463).
2026-07-25 23:08:19 +09:00
changukyu
8b5b3f5cf9 docs: 2026-07-24 review report, implementation checklist, spec revision
- Add the 2026-07-24 multi-agent review report (43 confirmed findings, runtime
  smoke, feature proposals with judge scores) and the implementation checklist
  tracking what was done vs. deliberately deferred (with reasons).
- Record the batch-decision decision: the v1 "no bulk approve/reject" GUI-spec
  restriction is revised, not silently overridden, and lists the guardrails that
  made lifting it acceptable.
- Fix the README `python -m rights_filter.server` invocation (needs
  PYTHONPATH=src) and start a docs/INDEX.md continuity log.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E6GD9gxUtfUiVB3JHejGpU
2026-07-24 14:18:27 +09:00
changukyu
2c607e0df6 docs: 2026-07-11 multi-agent review report, remediation checklist, offline extras
- Add the 2026-07-11 multi-agent project-review report and the remediation
  checklist (per-item status, with deferred items and their rationale).
- offline-install: document Ollama model and Playwright Chromium transfer, a
  hash-verified lock, --only-binary, and correct the egress note (image
  comparison fetches arbitrary SSRF-guarded public URLs, not only API hosts).
2026-07-11 16:08:04 +09:00
changukyu
d008e35c2e docs: add 2026-06-22 multi-agent project review report
66 verified findings (0 CRITICAL, 6 HIGH, 43 MEDIUM, 17 LOW) with a
prioritized remediation roadmap. Generated from an adversarially-verified
multi-agent review.
2026-06-22 14:00:45 +09:00
유창욱
2eb7bd3b8b docs: mark god-file split complete in remediation plan 2026-06-20 22:36:02 +09:00
유창욱
bd35cf6f3f docs: record remediation implementation status in plan 2026-06-20 20:57:02 +09:00
유창욱
8958dd1b83 chore: pin runtime dependencies for offline air-gapped install
Add requirements.txt (numpy/opencv-python-headless/pillow — the only
third-party runtime imports) and requirements-dev.txt, plus an offline
install runbook. Ignore .coverage and wheelhouse/.
2026-06-20 18:19:08 +09:00
유창욱
62c13faafa docs: implementation plan for project-review remediation 2026-06-20 18:18:54 +09:00
유창욱
63bbf0d755 docs: implementation plan for operator workbench efficiency (F1-F5) 2026-06-12 17:40:34 +09:00
유창욱
b4b8f4b5d8 docs: operator workbench efficiency design (F1-F5) 2026-06-11 16:03:15 +09:00
유창욱
ed701bd436 feat: clean review-instrument restyle with bundled Pretendard font
- Bundle Pretendard Variable woff2 locally (air-gapped safe, no CDN)
  and switch UI/stamp font stacks to it; preload in index.html
- Replace the forensic-dossier paper theme with a flat neutral cool
  palette: single teal accent, white cards, no noise texture, and
  zero linear/radial gradients (per design contract)
- Restore the product-purpose top-bar block and its CSS, drop the
  unused global search form, and strip the stray UTF-8 BOM
- Re-skin queue hover/selection, eyebrows, nav rail, chips, and
  empty states to the neutral palette; tabular numerals for numbers
- Regenerate ui-overhaul final audit artifacts: zero horizontal
  overflow across 8 views at 1440x900 and 390x844, Pretendard active

Design spec: docs/superpowers/specs/2026-06-11-operator-console-clean-review-ui-design.md
Plan: docs/plans/2026-06-11-001-feat-operator-console-clean-review-ui-plan.md
Tests: 358 passed (full suite incl. browser smoke)
2026-06-11 10:31:16 +09:00
유창욱
3f7b3a9cf2 chore: initial commit of copyrighter (rights_filter)
Image rights / copyright detection system: SQLite store, HTTP app,
search integrations (Naver, Google Custom Search, Google Cloud Vision
web detection), image analysis (fingerprints, face/person detection,
evidence enrichment, risk scoring), an admin/review layer, governance
and retention policies, batch jobs, and a browser-based operator GUI.

This baseline incorporates a full code-review remediation pass
(46 fixes; 358 tests passing). Highlights:

CRITICAL
- Prevent evidence cascade-delete during the schema-constraint
  migration by disabling FK enforcement around the table rebuild.

Security
- Sandbox served media (neutralize stored XSS from uploaded/collected
  SVGs) via CSP + nosniff on the untrusted media routes.
- Strip embedded EXIF/GPS from external image derivatives before they
  are sent to third-party APIs.
- Return a clean 404 (not an uncaught StopIteration) for PATCH on an
  unknown provider.

Correctness
- LLM-summary failures no longer add +30 to the risk score.
- Decode only explicit JS escapes so Korean image URLs are not mangled.
- Consume search quota only after a successful request.
- Naver/Google adapters map responses inside the failure boundary, so a
  malformed response degrades to evidence instead of crashing enrichment.
- Domain-aware provider attribution; face-box IoU de-duplication; count
  searches (not result items); per-box crop isolation; clamp evidence
  confidence and Google CSE num; real submittedEpoch; and more.

Robustness
- Offline LLM connect fast-fails (short connect timeout) so seed/reload
  requests are not stalled; full read timeout preserved for generation.
- Malformed numeric env vars fall back to defaults instead of crashing
  startup.

Performance
- Per-submission evidence reads (no full-table scan per rescore),
  audit-log LIMIT, lazy active-store lookup, hoisted timestamps.

Tests
- ~24 regression tests added pinning the above fixes.

Runtime data (data/, outputs/, *.sqlite3, *.log), secrets (.env), and
node_modules are gitignored.
2026-06-09 09:50:31 +09:00