opendatalab/MinerU

Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.

What it solves

MinerU is a comprehensive document parsing tool designed to convert complex documents—including PDFs, scanned images, and various Office formats—into clean, structured, and machine-readable content. It solves the problem of losing structural information (like tables, formulas, and layout) when extracting text from diverse file formats, making it ideal for AI agents and RAG pipelines that require high-fidelity document understanding.

How it works

MinerU employs a multi-tiered parsing approach to balance speed and quality:

  • Four Parsing Tiers: It offers Flash (fast previews/indexing), Basic (OCR and model-based), and Standard/Advanced (high-quality layout analysis).
  • Broad Format Support: It handles PDF, images, DOC/DOCX, PPT/PPTX, XLS/XLSX, RTF, ODT/ODS/ODP, EPUB, OFD, HTML/MHTML, and CSV/TSV.
  • Flexible Runtimes: It supports various inference engines, including ONNX and Torch for small models, and llama.cpp, vLLM, or LMDeploy for Vision Language Models (VLMs).
  • Document Library: It manages a local library that caches results, allows for content searching, and provides stable locators for citing specific pages or blocks.

Who it’s for

  • AI Agent Developers: Who need a reliable way for agents to read, cite, and navigate long documents without overloading the context window.
  • Data Engineers: Who need to convert large volumes of heterogeneous documents into Markdown, LaTeX, or HTML for training or analysis.
  • RAG Pipeline Builders: Who require precise document parsing to ensure the quality of the retrieved context.

Highlights

  • High-Fidelity Extraction: Specialized in handling tables, formulas, and complex page layouts.
  • Agent-Centric Design: Provides stable locators (doc:id/tier/page/block) for precise citations and progressive reading.
  • Multi-Target Rendering: Can export results to Markdown, HTML, LaTeX, DOCX, EPUB, PDF, and structured content.
  • Local-First: Operates locally by default, ensuring privacy and data security.

Related

  • Project
    magicyuan876/mineru-tianshu

    Tianshu (天枢) is an enterprise‑grade, self‑hosted AI data‑pre‑processing platform. It ingests multimodal files (PDF, Office docs, images, audio, video, bio‑informatics formats, encrypted archives), runs GPU‑accelerated OCR, VLM captioning, speech‑to‑text, watermark removal, and outputs structured Markdown and JSON. Features include JWT role‑based auth, audit logs, webhook callbacks, MCP protocol integration, Docker/offline deployment, and extensible format plugins. Built with Vue 3 front‑end, FastAPI + LitServe back‑end, and a suite of open‑source AI engines.

  • Project
    docling-project/docling

    Docling is an open‑source Python library/CLI that parses many document types (PDF, Word, slides, spreadsheets, HTML, e‑books, audio, video, email, etc.) into a unified, structured representation. It extracts layout, tables, code, formulas, images, and speech‑to‑text, then exports to Markdown, HTML, JSON, DocLang, DocTags, etc. Designed for generative‑AI workflows, it offers plug‑and‑play integrations with LangChain, LlamaIndex, Crew AI, Haystack, and supports local execution for privacy‑sensitive data.

  • Project
    pymupdf/PyMuPDF

    A high-performance Python library for extracting, analyzing, and converting PDF and Office documents, featuring LLM-ready Markdown and JSON output for AI pipelines.

  • Project
    xunbu/docutranslate

    DocuTranslate is a Python tool that translates many document formats locally using LLM APIs. It parses PDFs via MinerU, preserves formatting for DOCX/XLSX, supports async batch jobs, offers a web UI, REST API, and an MCP server, and can be run via pip, Docker, or portable binaries.

  • Project
    datalab-to/marker

    A high-accuracy document conversion tool that transforms PDFs and other files into structured Markdown, JSON, or HTML, specializing in tables and equations.