Loading...
Open-source Python CLI for auditable OCR and document-extraction runs, with page-level provenance, quality signals, budgets, review queues, and reproducible reruns.
OCR output is only useful when researchers can explain where it came from and which pages need review. PageLedger records that evidence as plain files that can be inspected, cited, and reconstructed without a service or database.
Auditable page-level run ledger for OCR and document extraction
PageLedger makes document extraction defensible: every page keeps its route, engine, quality evidence, cost basis, review status, and rerun lineage.
Text, born-digital PDF, and local PDF OCR.
Run, classify, inspect, verify, rerun, compare, align, diagnose, and initialize.
The documented synthetic run processes about 2,100 pages per second.
Each run writes a manifest, provenance, quality warnings, cost evidence, review queue, and an executable rerun plan as ordinary files.
Flagged pages can move to a stronger extractor without repeating the entire collection, while cross-run comparison preserves lineage and quality changes.