Repository navigation

Feat/txt chapter detection - #15

Merged

chaxus merged 54 commits into

mainfrom

feat/txt-chapter-detection

Jul 25, 2026

Merged

Feat/txt chapter detection#15
chaxus merged 54 commits into
mainfrom
feat/txt-chapter-detection

Commits on Jul 16, 2026
- feat: add chapter detection plan and pnpm workspace configuration
  chaxus
  committed
  7b0a40c
- feat: add rule-based txt chapter detection (P1)
  
  chaxus
  and
  claude
  committed
  c986090
- feat: add browser NLP inference infrastructure (P2)
  
  chaxus
  and
  claude
  committed
  10f8f4a
Commits on Jul 17, 2026
- feat: chapter cache persistence and detection confidence (P4 partial)
  
  chaxus
  and
  claude
  committed
  15280b6
- fix: catalogue first entry jumped past the opening page
  
  chaxus
  and
  claude
  committed
  9dd70f0
- feat: add ml/ training pipeline for chapter-title model (P3)
  
  chaxus
  and
  claude
  committed
  9aaed40
- chore: add train smoke-test flag and pin python 3.11/3.12
  
  chaxus
  and
  claude
  committed
  fbc4d0a
- feat: negative subsampling on train split to fix class imbalance
  
  chaxus
  and
  claude
  committed
  ece69b5
- fix: stabilize DeBERTa-v3 training (batch 16, tunable lr, eval subsampling)
  
  chaxus
  and
  claude
  committed
  b447efc
Commits on Jul 18, 2026
- feat: English txt weak-supervision + threshold sweep tool
  
  chaxus
  and
  claude
  committed
  d622507
- feat: format features for title detection (shared Python/JS impl)
  
  chaxus
  and
  claude
  committed
  0ba557a
- test: independent generalization eval on untrained Chinese txts
  
  chaxus
  and
  claude
  committed
  3fd8fb6
- docs: chapter detection journey — process, pitfalls, lessons
  
  chaxus
  and
  claude
  committed
  feca845
- fix: correct formatting inconsistencies in chapter detection documentation
  chaxus
  committed
  f5e8a44
- docs: expand chapter-detection journey (decision narrative + engineering)
  
  chaxus
  and
  claude
  committed
  2b43d1f
- docs: fill independent-test result (recall 0.988 on zero-contact books)
  
  chaxus
  and
  claude
  committed
  076b552
- docs: browser deployment plan for the chapter model (lazy-load/chunk/cache)
  
  chaxus
  and
  claude
  committed
  4766d1a
- fix: charset detection + anchored literary-phrase weak labeling
  
  chaxus
  and
  claude
  committed
  d9be44b
- docs: v2->v3 investigation — aggregate metrics hid flagship failure
  
  chaxus
  and
  claude
  committed
  ef92b28
- docs: v3 milestone — best model, honest capability boundary
  
  chaxus
  and
  claude
  committed
  6416405
- feat: wire v3 model into browser worker; hit DeBERTa-v3 quant wall
  
  chaxus
  and
  claude
  committed
  ec72f13
- docs: record per-language model direction; update all docs to current state
  
  chaxus
  and
  claude
  committed
  3cfbffa
- feat: train.py --base for per-language models
  
  chaxus
  and
  claude
  committed
  70646aa
- docs: per-language architecture validated — Chinese model int8 works
  
  chaxus
  and
  claude
  committed
  5bf14a1
- feat: wire per-language model into chapter detection (browser e2e proven)
  
  chaxus
  and
  claude
  committed
  00807b9
- feat: AI-enhance catalogue UI — confidence-gated model enhancement
  
  chaxus
  and
  claude
  committed
  b90ec93
- feat: English chapter-title model (distilbert int8 67MB), register en
  
  chaxus
  and
  claude
  committed
  46047db
- docs: record per-language model lessons (Chinese + English)
  
  chaxus
  and
  claude
  committed
  bb1e83b
- fix: correct spacing in chapter detection documentation and deployment plan
  chaxus
  committed
  e244380
Commits on Jul 19, 2026

Commits

Commits on Jul 16, 2026

Commits on Jul 17, 2026

Commits on Jul 18, 2026

Commits on Jul 19, 2026