Starstruk Logo MarkStarstruk

mineru-4.0.9-released v17.13.0

medium risk

Gerstep/cybosSep 28, 2026 → Sep 28, 20260 days between releases

What's New

Switched PDF/A validation to pikepdf and changed hyperlink behavior in force-ocr mode

Breaking Changes

  • `--force-ocr` now keeps hyperlinks (use `--mode force-ocr-no-links` for old behavior)
  • `--force-ocr`, `--deskew`, and `--clean-final` now keep hyperlinks
  • Options like `--pdfa-image-compression` are now errors with `internal` backend

New Features

  • New `--pdfa-backend` option (auto, ghostscript, internal)
  • New `--mode force-ocr-no-links` mode
  • New `PageInfo.has_visible_text` property

Improvements

  • PDF/A validation switched from veraPDF to pikepdf for ghostscript-free paths
  • Speculative conversion now supports PDF/A-1b
  • Ghostscript font substitution uses `-dNONATIVEFONTMAP`
  • Faster output file checking with progress bar

Bug Fixes

  • Numerous bugs related to OCR, Ghostscript, and metadata handling

Upgrade Notes

Hyperlinks are now preserved by default in `--force-ocr`, `--deskew`, and `--clean-final`. Use `--mode force-ocr-no-links` to disable them. Options like `--pdfa-image-compression` are now errors with the `internal` backend.

The Two Releases

mineru-4.0.9-released

MinerU 4.0.9

Sep 28, 2026· 24 days ago

Focuses on improving PDF text extraction accuracy by fixing word spacing issues and optimizing the PDF rendering subprocess for better concurrency handling.

Full release notes
v17.13.0

v17.13.0

Sep 28, 2026· 24 days ago

This update improves PDF/A conversion by switching validation from veraPDF to pikepdf and introduces a new backend option, while fixing numerous bugs related to OCR, Ghostscript, and metadata handling.

Full release notes