MinerU PDF Parser

**Judgement:** Strong, specialized document-to-Markdown parser skill, well-suited for OpenClaw workflows that need high‑quality extraction of PDFs/Office files, especially for papers and technical docs. **What it’s good at** - High-fidelity parsing: preserves formulas as LaTeX, handles complex tables, and performs layout analysis (multi-column, mixed images/text). - Broad format support: PDF (including scans), DOCX, PPTX, JPG/PNG with OCR, plus options for different model backends (speed vs accuracy vs HTML-style output). - Workflow fit: clear curl-based examples, environment variable (`MINERU_TOKEN`) usage, and an explicit paper-reading workflow make it easy to plug into OpenClaw automations. **Key risks / limitations** - **External SaaS dependency:** All documents are sent to MinerU servers; not appropriate for highly sensitive or regulated data, and uptime/rate limits are outside your control. - **API & quota limits:** File size (≤200 MB), page count (≤600), and concurrency depend on the MinerU plan. Heavy/batch users must design around quotas and possible throttling. - **Latency for large/complex docs:** Asynchronous task model plus polling; `vlm` mode trades speed for accuracy and may be relatively slow on big or complex documents. **Recommended scenarios** - Research and engineering teams wanting reliable parsing of papers, technical PDFs, and slide decks into Markdown/JSON for RAG, summarization, or note-taking. - Knowledge base or documentation migrations where preserving formulas, tables, and general layout quality is more important than offline or on-prem processing. - OpenClaw users building automated paper-reading or bulk document ingestion pipelines who are comfortable using a third‑party cloud API and managing an API key.