v1.23.0 v2.130.0

medium risk

tanchongmin/strictjsonSep 22, 2026 → Sep 22, 20261 day between releases

What's New

Adds Mineru 2.5 Pro VLM support and improves Markdown/DOCX/HTML parsing capabilities

New Features

  • Adds support for Mineru 2.5 Pro VLM
  • Adds recognition for .markdown files
  • Emits table captions in HTML output
  • Defer docling-parse import in PDF backend for performance
  • Fixes VLM end token stripping logic

Improvements

  • Honors backslash-escaped pipes in Markdown tables
  • Omit only unset arguments in service client calls
  • Harden LibreOffice profile handling and timeout kills
  • Improves DOCX list indentation handling
  • Improves PDF backend import deferral

Bug Fixes

  • Removes double preset registration of chart extraction
  • Fixes VLM end token stripping
  • Keeps DOCX list items when starting above indent level 0
  • Fixes service client argument handling (omits unset)

Upgrade Notes

No breaking changes were reported. Users should be aware that Mineru 2.5 Pro is now supported as a VLM and that Markdown files are now recognized by default. Configuration for LibreOffice timeouts may need adjustment if the new hardening behavior affects workflows.

The Two Releases

v1.23.0

1.23.0 — Text-layer repairs

Sep 22, 2026· 1 month ago

Focuses on text-layer repairs, handling malformed Form XObjects and invisible text layers under images.

Full release notes
v2.130.0

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 notes