Mistral OCR

A simple, self-hosted web application for optical character recognition (OCR) powered by Mistral AI's OCR API.

Features

• Document Processing: Upload PDF files or images (JPG, PNG, GIF, BMP, TIFF, WebP) up to 20MB
• Secure API Key Storage: Your Mistral API key is encrypted locally in your browser using AES-256-GCM with PBKDF2 key derivation
• Multiple Export Formats: Download results as Markdown, JSON, and extract embedded images
• Privacy-Focused: All processing happens through your own proxy server - no data stored on external servers
• Choice of models: proxy server and frontend are prepared for choice between two models (currently Mistral OCR 2 and 3), As long as Mistral keeps the same API structure (/v1/ocr endpoint with the same request format), you can update the model identifiers in proxy-server.js: Line 90: const validModels = ['mistral-ocr-2505', 'mistral-ocr-2512'];
• Simple Setup: Just two files - an HTML frontend and a Node.js proxy server

Prerequisites

• Node.js (v14 or higher)
• A Mistral AI API key

Installation

1. Clone the repository:

   git clone https://github.com/PetrAPConsulting/Mistral-OCR.git
   cd mistral-ocr

2. Install dependencies:

   npm install

Usage

1. Start the proxy server:

   npm start

2. Open web browser and open

   http://localhost:3001

3. Enter your Mistral API key and a password to encrypt it (the key is stored encrypted locally in your browser)

4. Select model for OCR, MistralOCR 2 is default model with API endpoint 'mistral-ocr-2505'

5. Upload a document and click "Process with OCR"

6. Download results in your preferred format

How It Works

┌─────────────────┐      ┌─────────────────┐      ┌─────────────────┐
│                 │      │                 │      │                 │
│  Browser (HTML) │ ───► │  Proxy Server   │ ───► │  Mistral API    │
│                 │      │  (localhost)    │      │                 │
└─────────────────┘      └─────────────────┘      └─────────────────┘

The proxy server is necessary because browsers block direct API calls to Mistral due to CORS restrictions. The proxy:

• Receives file uploads from the browser
• Forwards requests to Mistral's API
• Returns OCR results to the browser

API Endpoints

Endpoint	Method	Description
/health	GET	Health check - returns server status
/api/upload	POST	Upload a file to Mistral for processing
/api/ocr	POST	Process uploaded file with OCR

Security Considerations

• API Key Encryption: Your API key is encrypted using AES-256-GCM before being stored in localStorage. A password-derived key (PBKDF2, 100,000 iterations) is used for encryption.
• Local Processing: The proxy server runs locally - your documents are not stored anywhere.
• Session-Based: API keys are only decrypted into memory for the current session.

File Structure

mistral-ocr/
├── index.html          # Frontend
├── Assets              # Frontend components
      └── favicon.png
      └── images.png
├── proxy-server.js     # Node.js proxy server
├── package.json        # Dependencies configuration
├── node_modules        # Installed dependencies
└── README.md           # This file

Dependencies

• Express - Web server framework
• Multer - File upload handling
• CORS - Cross-origin resource sharing
• node-fetch - HTTP client for API calls
• form-data - Multipart form data handling

License

MIT License - see LICENSE for details.

Version

ver. 1.2.0 December 2025

Acknowledgments

• Powered by Mistral AI OCR API
• Developed by AP Consulting