v2.130.0 v1.24.0
tanchongmin/strictjsonSep 22, 2026 → Sep 23, 20261 day between releases
What's New
Adds text paint info and metadata extraction; fixes PDF title decoding and layout issues.
New Features
- Text paint information extraction (fill/stroke colors, render mode).
- Expanded document metadata extraction (Author, Subject, Keywords, Creator, Producer, CreationDate, ModDate).
- PDF title decoding improvements (handles UTF-16LE and PDFDocEncoding).
- Extract text rendered with the `"` operator.
- Keep two lines of small type apart when baselines are close.
Improvements
- ActualText span weight calculation now uses the first painted glyph's paint.
Bug Fixes
- Decode PDF Title as a PDF text string to prevent Mojibake.
- Fix stacked table headers where baselines were too close.
- Extract text rendered with the `"` operator which was previously skipped.
- Take ActualText span weight from first painted glyph.
Upgrade Notes
TextItem now includes new properties: fill_color, stroke_color, and render_mode. Document metadata extraction now includes Author, Subject, Keywords, Creator, Producer, CreationDate, and ModDate. PDF titles are now decoded correctly for UTF-16LE and PDFDocEncoding.
The Two Releases
Sep 22, 2026· 30 days ago
Adds Mineru 2.5 Pro VLM support and fixes chart extraction double registration. Includes numerous improvements for Markdown table parsing, PDF, DOCX, and HTML backends.
Full release notes1.24.0 — Text paint and document information
Sep 23, 2026· 29 days ago
Adds text paint information (fill/stroke colors, render mode) to TextItem and expands document metadata extraction.
Full release notes