Repository navigation
Release Details
This release is a major rewrite of the table extraction pipeline, with a focus on performance and extraction quality.
- Added Python 3.14 support.
- Reduced runtime dependencies and moved performance-critical code to compiled extensions.
- Added RapidOCR support.
- Added concurrent PDF page processing via
max_workers. - Improved bordered table detection.
- Added normalized bounding boxes through
bbox.relative. - Reworked native PDF text extraction for better accuracy.
- Added a new borderless table pipeline based on layout segmentation and structural scoring.
Full Changelog: 1.4.2...2.0.0