7503a441c2
kg: graph build, traversal queries, neo4j export - kg_ocr.graph builds a networkx graph (docs, chunks, entities, citations, co-occurrence) from chunk markdown - analyzer for summaries, top entities/citations, anomaly checks - traversal: chunks_for_entity/citation, related_entities, expand_context - export: JSON round-trip, GraphML, batched MERGE into neo4j - new CLI: ocr-pipeline kg build|stats|query|export - lazy kg_ocr imports, networkx/neo4j behind extras - dropped dead watch.py shim, added KgConfig stub - trimmed README, updated TODO
tests / core (macos-latest, py3.11) (push) Has been cancelled
tests / core (macos-latest, py3.12) (push) Has been cancelled
tests / core (macos-latest, py3.13) (push) Has been cancelled
tests / core (ubuntu-latest, py3.11) (push) Has been cancelled
tests / core (ubuntu-latest, py3.12) (push) Has been cancelled
tests / core (ubuntu-latest, py3.13) (push) Has been cancelled
tests / doctor CLI smoke test (push) Has been cancelled
2026-07-20 19:50:38 +02:00
39655fc35f
add doctor command, fix ISBN false positives, wire up CI
...
tests / core (macos-latest, py3.11) (push) Has been cancelled
tests / core (macos-latest, py3.12) (push) Has been cancelled
tests / core (macos-latest, py3.13) (push) Has been cancelled
tests / core (ubuntu-latest, py3.11) (push) Has been cancelled
tests / core (ubuntu-latest, py3.12) (push) Has been cancelled
tests / core (ubuntu-latest, py3.13) (push) Has been cancelled
tests / doctor CLI smoke test (push) Has been cancelled
ISBN regex now requires an explicit prefix, so it stops matching
years, URLs, job IDs, and dilution ratios.
doctor checks the environment end to end: Python, Tesseract,
PaddleOCR, the PyTorch backend, table/figure detectors, scispaCy,
config, DB, output.
CI runs on 2 OS x 3 Python versions plus a doctor smoke job.
Also added a changelog
2026-07-20 12:37:26 +02:00
a0211155fa
refactor: solidify startup UX and engine-aware preprocessing
...
- Slim core deps: move ML stack to optional extras (paddle/tables/figures/scientific/full)
- Lazy settings proxy with config search paths (env var, cwd, user dir)
- New commands: init, demo, setup [basic|full], first-run guard on run/watch
- Engine-aware preprocessing: Paddle gets original image (fixes dark mode 0.83->0.95)
- Results table shows Skipped count; lazy run-dir creation
- kg_ocr marked experimental with extra, Docker defaults with OCR_PIPELINE_CONFIG
- 25/25 tests, ruff clean
2026-07-19 22:14:54 +02:00
deab02610b
update README pyproject & config
2026-07-19 20:01:32 +02:00
1cf1c6eeab
improve the OCR pipeline processing and outputs formatting
2026-07-19 19:54:53 +02:00
d375db47c3
platform agnosticity
2026-07-19 11:45:08 +02:00
4fe9e76cad
add ImagePreprocessor, compute_image_hash, and OutputMetadata
2026-07-19 08:46:53 +00:00
bf9e8b8a8f
add ImagePreprocessor and compute_image_hash to ocr module
2026-07-19 08:43:20 +00:00
a32b7508c7
OCR-to-RAG pipeline for life science screenshots
...
PaddleOCR with preprocessing, scispaCy NER, figure/table detection,
citation extraction, and chunked Markdown output with frontmatter.
Includes watch mode and notebook reprocessing.
2026-07-18 19:35:21 +00:00
012549b4bc
added llm and rag queries
2026-03-24 20:15:22 +01:00
cbfcf1e315
added scaffold struct for the prj
2026-03-24 14:58:58 +01:00
95655686f9
RAG init using litellm
2026-03-21 19:31:11 +01:00
37fc80931f
add txtai, miniLM
2026-03-19 14:29:38 +01:00
4b25500600
prj init
2025-12-09 22:53:41 +01:00
