PDF Content Extractor & Translator

Turn static PDFs into living, editable data without your files ever leaving your machine.

This is a feature-rich, privacy-focused platform that converts Documents into structured formats (Word, CSV, ODT), translates them using local neural networks, and allows you to chat with your files using local LLMs.

---

📚 Documentation

• Technical Manual 👈 Start Here for deep dives
  • System Architecture & Data Flow
  • Detailed Usage Guide (GUI & CLI)
  • API Reference
• Development Guide - For contributors.
• MCP Server - Connect this tool to AI assistants like Claude.

---

🚀 Key Features

• 🛡️ Privacy First: 100% offline processing. No external APIs, no cloud uploads.
• 📄 Structural Extraction: Converts PDFs to .docx while preserving massive layout details (tables, headings, lists) using Docling.
• 🌍 Offline Translation: Translate full documents to 9+ languages (English, Spanish, French, German, etc.) preserving layout.
• ✏️ Rich PDF Editor: Annotate, Redact, Sign, Merge, Split, and Compress PDFs in the browser.
• 🤖 Local AI Chat: Ask questions about your documents using completely local LLMs (via Ollama).
• 🔌 MCP Support: Expose these tools to your AI agent (Claude, etc.) via the Model Context Protocol.

---

🏁 Quick Start

🐳 Option A: Docker (Recommended)

The fastest way to run the full stack (App + Redis + OCR + Translation Models).

git clone https://github.com/matis-dev/pdf-extractor-translator.git
cd pdf-extractor-translator
docker-compose up --build

Access at: http://localhost:5000

🐍 Option B: Manual Setup

For developers who want to run it natively on Linux/macOS.

Prerequisites: redis-server, tesseract-ocr, pandoc.

# 1. Setup Environment
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# 2. Start Services
./scripts/start.sh

---

🏗️ Architecture Overview

graph TD
    User((User)) --> WebUI[Flask Web App]
    User --> CLI[Python CLI Scripts]
    
    WebUI --> Tasks[Celery Tasks]
    Tasks --> Docling[Docling Engine]
    Tasks --> Translation[Argos Translate]
    
    Docling --> PDF[PDF Processing]
    Translation --> LocalModels[Local NLP Models]
    
    subgraph Storage
        Redis[(Redis Queue)]
        Uploads[Uploads/Outputs Folder]
        Chroma[(ChromaDB)]
    end
    
    Tasks <--> Redis
    WebUI <--> Uploads
    Tasks <--> Uploads
    WebUI <--> Chroma

Loading

---

📸 Screenshots

Home Page - Library & Upload

Editor Page - Annotation & Manipulation Tools

---

📂 Project Structure

src/
├── app.py                  # Main Flask Application
├── tasks.py                # Celery Background Workers
├── routes/                 # API Controllers
│   ├── ai_routes.py        # LLM & RAG Endpoints
│   ├── convert_routes.py   # Unified Conversion API
│   └── pdf_routes.py       # PDF Manipulation API
├── static/                 # Frontend (Vanilla JS Modules)
├── templates/              # Jinja2 HTML Templates
├── extract_*.py            # Standalone CLI Scripts
└── mcp_server.py           # Model Context Protocol Server

---

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for details on how to submit pull requests, report issues, and request features.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.