Commit History
-
feat(documents): PaddleOCR-VL parser for v2 processing, server toggle
Rifqi Hafizuddin +1: Claude Opus 5.5POST /api/v2/document/process now parses with PaddleOCR-VL-1.6 (Baidu AI Studio job API) by default; DOCUMENT_PARSER_BACKEND=paddleocr|mistral picks the parser server-side. New paddle_ocr.py (multipart submit -> poll state -> JSONL; figures downloaded at parse time) and normalize_paddle.py (same ParsedDocument; a title before a table becomes its caption; figures matched by label+bbox since the job API leaves image blocks empty). DOCX is refused on PaddleOCR with a readable error. Rows record their parser; chart images reach image-token. Knowledge pipeline default unchanged (Mistral). Live 2026-09-24: 64-page Komatsu manual -> completed, 80 chunks, 27 figures, signed figure URL 200. Needs PADDLEOCR_ACCESS_TOKEN. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>7ee2b45
-
/feat knowledge management (#20)
- [NOTICKET] docs: knowledge pipeline test integration — context, audited task checklist, DEV_PLAN §0.8 (8e69dc6c638509f4dd7b6775d520fb4aa4efbd3e) - [NOTICKET] docs: knowledge pipeline — reframe kex as prototype, not port target (7cce5ffc9ab88791ac139dbef8e88e124a3d049e) - [NOTICKET] feat(eval): knowledge-extraction gold set, scorer + frozen prototype baseline (60cf4a885ef31a8aa64916f5a70ac5a0e8062c5f) - [NOTICKET] chore: add missing planner and readiness eval results (bd6659b3bf523597e4482b04ab1fcff5e1284d25) - [NOTICKET] feat(knowledge_extraction): free stages — models, seam adapter, filters, cluster, ranking (024c30abd1189119d33508eded334fe5a3ed5011) - [NOTICKET] fix(knowledge_extraction): move runner into the package; fix two ranking defects (72605ff6876a7936ab685d417c766ce820f6214b) - [NOTICKET] feat(knowledge_extraction): span filter — E1 recall 0.8537, matches the frozen baseline (3a016343398753cf730d50548c0e92ff47c6716c) - [NOTICKET] feat(knowledge_extraction): paid extraction …b68816f
