MinerU

Best for: anyone building a RAG pipeline who needs PDFs and Office documents converted into clean, structured text an LLM can actually use.

MinerU, from OpenDataLab, is a document parsing engine that converts PDF, DOCX, PPTX, XLSX, images and web pages into structured Markdown or JSON. It uses a dual VLM-plus-OCR engine, supports 109 languages, and integrates natively with LangChain, Dify and FastGPT. An MCP server ships with it, so agents can call it directly.

What makes it useful: it handles layout, which is where naive PDF extraction usually fails. Multi-column academic papers, tables and mixed text-and-image pages come out with structure intact rather than as scrambled text — the difference between a RAG pipeline that works and one that quietly returns nonsense.

Where it falls short: accurate parsing is computationally expensive, and the VLM path benefits substantially from a GPU. If you only need to read simple single-column PDFs, a lighter library will be faster to run and easier to deploy. The license is non-standard (NOASSERTION), so check terms before commercial use.

Details: Python. Zero-install web version and desktop client also available.