Feature
- Add the ebcdic backend (#3926) (
9bdb2af) - service datamodel: Expose PDF heading-level inference in the service API (#3874) (
a47f76a) - skills: Add agent skills for using docling (#3913) (
5dc8c19) - Refactor RapidOcrModel to support all PP-OCR languages via version/backbone resolution (#3863) (
f7eb9b4)
Fix
- docx: Stop dropping section header/footer content in Docx (#3843) (
2ec4f64) - docx: Resume ordered-list numbering after an intervening list (#3896) (#3897) (
62f6a51) - docx: Keep body text after a blank spacer inside a list (#3898) (#3902) (
7883aaa) - pdf: Rewind stream inputs in docling-parse (#3895) (
995ae2a) - odf: Preserve text after inline elements (#3850) (
955d579) - reading-order: Dehyphenate hard continuations (#3888) (
52d8a6f) - pptx: Process slide shapes in visual reading order (#3393) (
91fa745)