A comprehensive Bun-based CLI and Web application for OCR processing using multiple AI models.

• Dual Mode Operation: Works as both a CLI tool and a web application
• Multiple AI Models: Support for Mistral AI OCR and Gemini Pro 2.5
• PDF Processing: Converts PDF files to images and performs OCR
• Real-time Progress: WebSocket-based progress updates in web mode
• Persistent Output: Web mode saves results to permanent folders
• Modern UI: React frontend with Material-UI components

• Bun runtime installed

1. Install dependencies:

   bun install

2. Set up API keys:

   cp .env.example .env
   # Edit .env with your API keys

3. Build the application:

   bun run build:full

4. Run in web mode:

   bun run dev
   # Opens http://localhost:3001 automatically

# Basic usage
bun run dev --pdf input.pdf --model mistral-ai-ocr --api-key YOUR_KEY --output ./output

# With Gemini model and custom prompt
bun run dev --pdf input.pdf --model gemini-pro-2.5 --api-key YOUR_KEY --prompt "Extract all text with formatting" --output ./output

# Extract images separately
bun run dev --pdf input.pdf --model mistral-ai-ocr --api-key YOUR_KEY --get-images --output ./output

# Start web server (opens browser automatically)
bun run dev

# Or build and run production version
bun run build:full
bun run start

The web interface provides:

• Model selection (Mistral AI OCR, Gemini Pro 2.5)
• API key management (stored in browser)
• Custom prompt input
• Drag-and-drop PDF upload
• Real-time progress updates
• Side-by-side PDF viewer and markdown results
• Copy markdown button
• Open output folder button

bun-ai-ocr-tester/
├── public/                 # Built React app and static assets
│   ├── index.html          # Main HTML file
│   ├── index.js            # Bundled React app
│   └── pdf.worker.min.js   # PDF.js worker
├── src/
│   ├── index.ts           # Main entry point (CLI/Web router)
│   ├── cli.ts             # CLI mode handler
│   ├── server/            # Web server components
│   │   ├── index.ts       # Express server setup
│   │   └── ws.ts          # WebSocket handler
│   ├── models/            # OCR model implementations
│   │   ├── modelInterface.ts
│   │   ├── mistralOCR.ts
│   │   └── geminiOCR.ts
│   ├── utils/             # Utility functions
│   │   ├── apiKeyManager.ts
│   │   ├── outputManager.ts
│   │   ├── pdfProcessor.ts
│   │   └── progressLogger.ts
│   ├── web/               # React frontend source
│   │   ├── src/
│   │   │   ├── App.tsx
│   │   │   ├── components/
│   │   │   ├── hooks/
│   │   │   └── services/
│   │   └── public/
│   └── types/             # TypeScript definitions
├── ocr-outputs/           # Web mode output directory
├── package.json
├── tsconfig.json
└── README.md

1. Visit https://console.mistral.ai/
2. Create an account and get your API key
3. Use model: mistral-ai-ocr

1. Visit https://makersuite.google.com/app/apikey
2. Create an account and get your API key
3. Use model: gemini-pro-2.5

• MISTRAL_API_KEY: Default API key for Mistral AI
• GEMINI_API_KEY: Default API key for Gemini Pro
• PORT: Custom port for web server (default: 3001)

• bun run build:full - Clean and full build
• bun run build:frontend - Build React app only
• bun run build:backend - Build backend only
• bun run clean - Remove build artifacts

# Install dependencies
bun install

# Start in development mode (web)
bun run dev

# Test CLI mode
bun run dev --pdf sample.pdf --model mistral-ai-ocr --api-key YOUR_KEY

# Build for production
bun run build:full

• Process PDFs with command-line arguments
• Support for both Mistral and Gemini models
• Configurable output directory
• Optional image extraction
• Console progress logging

• Modern React interface with Material-UI
• Real-time WebSocket progress updates
• Persistent browser settings (API keys, prompts)
• Side-by-side PDF viewer and results
• Permanent output folders with "Open Folder" feature
• Copy markdown to clipboard

• API keys stored locally (browser localStorage or environment)
• Path validation for file serving
• Input sanitization
• CORS protection

MIT