Starstruk Logo MarkStarstruk

v4.0.0a1

Pre-release

opendatalab/MinerUv4.0.0a1Jul 14, 2026by github-actions[bot]

AI Summary

Initial alpha release for MinerU 4.0.0, featuring major upgrades to PDF processing, image handling, and the removal of the VLM backend.

Key Highlights

  • Enhanced PDF processing and concurrency management
  • Added functionality to skip broken PDF pages
  • Integrated ImagePayloadCache for improved image handling
  • Removed VLM backend
  • Enhanced formula content handling with normalization

Breaking Changes

  • Removed VLM backend

New Features

  • PDF page skipping logic
  • ImagePayloadCache integration
  • Formula normalization and tagging
  • DOCX style circular inheritance prevention
  • Hybrid magic model INDEX block handling

Full Release Notes

## What's Changed
* master->dev by @myhloli in https://github.com/opendatalab/MinerU/pull/5042
* Enhance PDF processing and improve concurrency management by @myhloli in https://github.com/opendatalab/MinerU/pull/5062
* feat: add functionality to skip broken PDF pages during rewrite process by @myhloli in https://github.com/opendatalab/MinerU/pull/5064
* next with 3.2.2 by @myhloli in https://github.com/opendatalab/MinerU/pull/5174
* feat: integrate ImagePayloadCache for improved image handling and path management by @myhloli in https://github.com/opendatalab/MinerU/pull/5194
* Next with 3.3.1 by @myhloli in https://github.com/opendatalab/MinerU/pull/5196
* Next with 3.4.0 by @myhloli in https://github.com/opendatalab/MinerU/pull/5202
* feat: enhance formula content handling with normalization and tagging functions by @myhloli in https://github.com/opendatalab/MinerU/pull/5206
* feat: adjust block type handling for INDEX in hybrid magic model by @myhloli in https://github.com/opendatalab/MinerU/pull/5207
* feat: remove VLM backend by @myhloli in https://github.com/opendatalab/MinerU/pull/5214
* Next by @myhloli in https://github.com/opendatalab/MinerU/pull/5236
* Next by @myhloli in https://github.com/opendatalab/MinerU/pull/5241
* feat: prevent circular inheritance in DOCX styles by tracking visited styles by @myhloli in https://github.com/opendatalab/MinerU/pull/5249


**Full Changelog**: https://github.com/opendatalab/MinerU/compare/mineru-3.2.1-released...v4.0.0a1