Feature
- ocr: Detect page orientation for RapidOCR (#4709) (
5e72863) - api: Add InputDocument.release() (#3600) (
761d4cf)
Fix
- pipeline: Drop page backend references on unload (#5089) (
544f368) - jats: Stop rendering the literal text None in figure and table captions (#4291) (
dc4693f) - md: Strip leading YAML frontmatter instead of misparsing it (#4596) (
8f9a971) - asr: Avoid initializing MLX during preset imports (#5090) (
55c1da8) - jats: Keep the target of an ext-link in a mixed citation (#4813) (
d518808) - html: Keep rows stored in tfoot (#4323) (
ad03515) - vlm: Keep the page when a Chandra table is too large (#5054) (
7d62cae) - jats: Keep tables whose only row is a header row span (#5041) (
851dce7) - xlsx: Do not create empty cells while finding tables (#5024) (
8f7ae54) - Preserve native VLM inline math through exports (#4908) (
9051777) - service_client: Do not send default page range in request options (#5056) (
907dcf2) - pdf: Replace NUL from unmapped glyphs with U+FFFD (#5070) (
ca26ff1) - xlsx: Keep currency, percent, and date number formats (#4628) (
25cdd57) - vlm: Scale DocLang locations to the page size (#5051) (
6d1470d)
Documentation
- Correct Nemotron OCR installation instructions (#5083) (
52aceeb) - Document installing docling-slim without RapidOCR (#5058) (
38719d4)