DSH Pdf Mineru
Yurzi/dsh-pdf-mineru · skills
dsh-pdf-mineru enables DeepSeek Harness to have powerful PDF and document intelligent parsing capabilities. It supports MinerU official cloud (v4) and privatized self-built service (v2), providing AI Agent with high-precision document layout analysis, formula and table extraction, image and text analysis, and native support for background asynchronous tasks and intelligent caching.
★ 5GitHub stars
1Forks
2026-09-05Last updated
TypeScriptLanguage
MITLicense
Key features
- High-quality structured extraction: accurately identify double-column typesetting, complex multi-level headings, LaTeX mathematical formulas, tables and illustrations, and output Markdown with elegant typesetting.
- After calling read_pdf or the background task async_parse_pdf, the delivery result contains reliable text and integrity judgments
- Let DeepSeek Harness have powerful PDF and document intelligent parsing capabilities
- Unified tool distributionThe synchronous request (read_pdf) initiated by the Agent directly returns the result, and the asynchronous long task (async_parse_pdf) is scheduled by DSH's native JobRegistry.
- Free cloud/local switchingSupports the out-of-box MinerU official cloud API (no local graphics card required) and privatized self-built services (data does not leave the intranet), and the unified tool interface allows seamless switching.
Requirements
- Important version notices and environment requirements: - Minimum supported DSH version: = 0.1.2-rc.1.
- Only supports RC versionThis plug-in will only officially support the RC (Release Candidate) version of DeepSeek Harness and subsequent official releases.
Install command
dsh plugin --profile web add github:Yurzi/dsh-pdf-mineru