magicyuan876/mineru-tianshu

天枢 - 企业级 AI 一站式数据预处理平台 | PDF/Office转Markdown | 支持MCP协议AI助手集成 | Vue3+FastAPI全栈方案 | 文档解析 | 多模态信息提取

What it solves

MinerU Tianshu is an enterprise-grade data preprocessing platform designed to convert unstructured data from various formats into structured, AI-ready formats like Markdown and JSON. It addresses the challenge of preparing high-quality training or RAG (Retrieval-Augmented Generation) data from complex sources including PDFs, images, audio, and video.

How it works

The platform uses a modular architecture combining a Vue 3 frontend, a FastAPI backend, and LitServe for GPU load balancing. It integrates several specialized engines to handle different modalities:

  • Documents & Images: Uses MinerU and MarkItDown for PDF, Word, Excel, and PPT parsing, with optional YOLO11x and LaMa for watermark removal.
  • Audio & Video: Employs SenseVoice for multi-language speech-to-text, speaker diarization, and emotion recognition, while using FFmpeg and OCR for video keyframe extraction.
  • Specialized Formats: Includes plugin-based engines for bioinformatics formats like FASTA and GenBank.
  • Infrastructure: Implements a task queue system with worker pools that pull tasks actively, supporting GPU isolation and multi-GPU scaling. It also integrates RustFS for S3-compatible object storage of extracted images.

Who it’s for

It is built for enterprises and developers who need to process large volumes of multimodal data for AI applications, specifically those requiring a scalable, self-hosted solution with user authentication, audit logs, and API integration.

Highlights

  • Multimodal Support: Processes documents, images, audio, video, and bioinformatics data.
  • Enterprise Features: Includes JWT authentication, role-based access control, API key management, and webhook notifications for task completion.
  • GPU Optimization: Features LitServe-based load balancing and support for NVIDIA CUDA and Apple Silicon (MPS/MLX).
  • AI Assistant Integration: Supports the Model Context Protocol (MCP), allowing AI agents like Claude Desktop to trigger document parsing directly.
  • Scalable Deployment: Offers one-click Docker deployment with options for GPU, CPU, and fully offline environments.

Related

  • Project
    opendatalab/MinerU

    A high-accuracy document parsing engine that converts PDF, DOCX, PPTX, XLSX, and images into structured Markdown or JSON for LLM and RAG workflows.

  • Project
    pymupdf/PyMuPDF

    A high-performance Python library for extracting and manipulating PDF and Office documents, featuring specialized tools for converting documents into LLM-ready Markdown for RAG pipelines.

  • Project
    HKUDS/RAG-Anything

    RAG‑Anything is a Python library that builds a multimodal Retrieval‑Augmented Generation system. It parses PDFs, Office files, images, tables, and equations using MinerU, creates a cross‑modal knowledge graph, and retrieves relevant pieces with a hybrid vector‑graph search before passing them to an LLM for answer generation. Install via `pip install raganything`, configure with `RAGAnythingConfig`, ingest documents, and query in natural language.

  • Project
    yuezhiai/jonex

    An AI multimodal domain knowledge engine that transforms raw documents, audio, and video into ontology-powered knowledge services for enterprise AI agents.

  • Project
    aiptimizer/TurboOCR

    A high-speed GPU document parser that converts images and PDFs into structured Markdown, including tables and formulas, using C++, CUDA, and TensorRT.