LangChain 1.0 + OCR Multimodal Document Analysis System

Integrates MinerU, PaddleOCR‑VL, and DeepSeekOCR — top-performing OCR parsing projects

English | 中文

⚡ Overview

Deploy the industry's leading OCR parsing projects via the vLLM inference framework — MinerU, DeepSeek‑OCR, and PaddleOCR‑VL — and build a multimodal data analysis system with a unified parsing service interface. The project includes optimizations and wrappers for DeepSeek‑OCR and MinerU service interfaces, making it ready for enterprise use.

redpandacompress_.2025-11-09_164002_924.mp4

For installation and detailed instructions for MinerU, PaddleOCR‑VL, and DeepSeekOCR, see the tutorial.

🎯 Key Features

• Unified parsing interface: pluggable selection of MinerU, PaddleOCR‑VL, and DeepSeek‑OCR
• Batch parsing: supports batch processing for PDFs and images; auto-splits multi‑page documents
• High performance: powered by the vLLM inference framework
• Multimodal support: extract text, tables, formulas, images, and more
• Standardized outputs: unified format with Markdown/JSON and image exports

🚀 Quick Start

For MinerU, PaddleOCR‑VL, and DeepSeekOCR installation and detailed guidance, refer to the tutorial.

Configure Backend Environment

Edit backend/.env:

# Server Configuration
PORT=8000
HOST=0.0.0.0
DEBUG=True

# MinerU Configuration - Using Direct API
MINERU_API_URL=http://192.168.130.4:50000/file_parse
VLLM_SERVER_URL=http://192.168.130.4:40000
MINERU_BACKEND=vlm-vllm-async-engine
MINERU_TIMEOUT=600
MINERU_VIZ_DIR=/home/MuyuWorkSpace/05_OcrProject/backend/mineru_visualizations

# DeepSeek OCR Configuration
DEEPSEEK_OCR_API_URL=http://192.168.130.4:8797/ocr

# PaddleOCR Configuration
PADDLEOCR_API_URL=http://192.168.130.4:10800/layout-parsing

# File Upload Limits
MAX_FILE_SIZE=10485760
ALLOWED_FILE_TYPES=application/pdf,image/png,image/jpeg,image/jpg,image/webp

# Storage Paths
UPLOAD_DIR=./uploads
EXPORT_DIR=./exports
TEMP_DIR=./temp

# Processing Timeout (seconds)
OCR_TIMEOUT=300

# CORS Settings
ALLOWED_ORIGINS=http://localhost:3000,http://localhost:5173

Start Backend Service

    cd backend

    # Create and activate virtual environment
    python -m venv venv
    source venv/bin/activate  # Linux/Mac
    # or
    venv\Scripts\activate    # Windows

    # Install dependencies
    pip install -r requirements.txt

    # Start server
    python main.py

Start Frontend Service

    cd frontend

    # Install dependencies
    npm install

    # Start dev server
    npm run dev

🙈 Contributing

Contributions via GitHub PRs or issues are welcome. We appreciate any form of contribution, including feature improvements, bug fixes, and documentation.

😎 Community

Explore our tech community 👉 Large Model Tech Community | Fanfan Space

Scan to add the contact and reply "OCR" to join the technical group and learn with other members.