Surf AI - AI Browser Automation 🌐

Surf AI is a powerful browser automation agent built with LangGraph and Vision-capable LLMs (via Groq). It allows you to control a web browser using natural language, seeing the screen just like a human does.

Features 🚀

• Vision-Based Navigation: The agent "sees" the screen using screenshots to interact with elements accurately.
• Natural Language Control: Tell the agent what to do (e.g., "Search for LangGraph on Google") and watch it work.
• Real-Time Live View: Watch the browser's actions in real-time through the React-based dashboard.

Project Structure 📂

• client/: Modern Vite + React frontend dashboard.
• src/: Core agent logic, tools, and browser management.
• server.py: FastAPI backend with WebSocket support for real-time streaming.
• Dockerfile: Production-ready container configuration.

Setup & Installation 🛠️

1. Backend Setup

1. Clone the repository.
2. Install dependencies:

   uv sync

3. Create a .env file with your API keys:

   GROQ_API_KEY=your_key_here
   MODEL=meta-llama/llama-4-scout-17b-16e-instruct
   ALLOWED_ORIGINS=http://localhost:5173

4. Start the server:

   uv run python server.py

2. Frontend Setup

1. Navigate to the client directory:

   cd client

2. Install dependencies:

   npm install

3. (Optional) Set your backend URL in .env:

   VITE_WS_BACKEND_URL=localhost:8000

4. Start the dev server:

   npm run dev

Deployment 🚢

Backend (Railway)

1. Link your GitHub repo to Railway.
2. Railway will automatically detect the Dockerfile.
3. Add your GROQ_API_KEY and other variables in the Railway dashboard.

Frontend

1. Deploy the client/ folder to any static hosting service.
2. Ensure you set VITE_WS_BACKEND_URL to your live backend domain.

License 📜

MIT