🤖 AI Browser Agent

Intelligent web automation agent that can understand, navigate, and interact with web pages autonomously.

!Python !License !Status

🎯 What This Agent Does

This AI agent can:

• 🌐 Open and navigate to any website automatically
• 🔍 Intelligently detect form elements and interactive components
• ✍️ Fill forms with provided data
• 📤 Submit forms autonomously
• 🧠 Adapt strategies when initial approaches fail
• 📸 Take screenshots for verification

✨ Features

• Multi-Strategy Element Detection: Uses both DOM service and CSS selectors
• Smart Fallback Logic: Automatically tries alternative approaches
• Real Browser Automation: Uses Playwright with Chromium
• Local Profile Management: Avoids permission issues with local directories
• Comprehensive Logging: Detailed progress tracking
• Error Recovery: Handles failures gracefully

🚀 Quick Start

Prerequisites

# Python 3.11+
# Conda environment (recommended)

Installation

1. Clone the repository:

git clone https://github.com/yourusername/browser-ai-agent.git
cd browser-ai-agent

2. Install dependencies:

pip install -r requirements.txt

3. Run the demo:

python examples/google_form_demo.py

📖 Usage Examples

Basic Form Filling

from ai_browser_agent import DirectFormFiller

async def fill_my_form():
    filler = DirectFormFiller()
    try:
        await filler.fill_google_form()
    finally:
        await filler.close()

# Run the agent
import asyncio
asyncio.run(fill_my_form())

Custom Form Data

# Modify the form data in the script
form_data = {
    'name': 'Your Name',
    'email': 'your.email@domain.com',
    'organization': 'Your Organization',
    'message': 'Your custom message'
}

🎬 Demo

The included demo successfully:

1. Opens Google Form - "Agentic Form" (live example)
2. Detects 4 text inputs using CSS selectors
3. Fills form fields:
   • Name: "AI Agent Demo"
   • Email: "demo@tsai.in"
   • Organization: "The School of AI"
   • Message: Custom automation message
4. Submits form automatically
5. Takes screenshot for verification

🛠️ Technical Architecture

Core Components

• DirectFormFiller: Main agent class
• BrowserSession: Manages browser lifecycle
• DomService: Handles page element detection
• CSS Selectors: Fallback element targeting
• Action Registry: Available browser actions

Key Technologies

• Playwright: Browser automation engine
• Chromium: Headless/headed browser
• AsyncIO: Asynchronous execution
• CSS Selectors: Robust element targeting

🔧 Configuration

Browser Settings

# Local profile directory (avoids permission issues)
local_profile_dir = Path(__file__).parent / "browser_profiles" / "default"
local_downloads_dir = Path(__file__).parent / "browser_profiles" / "downloads"

profile = BrowserProfile(
    headless=False,  # Set to True for headless mode
    user_data_dir=local_profile_dir,
    downloads_dir=local_downloads_dir
)

Element Detection Strategy

1. Primary: DOM Service with highlight elements
2. Fallback: CSS selectors with multiple strategies
3. Recovery: Index-based element targeting

📁 Project Structure

browser-ai-agent/
├── README.md                 # This file
├── requirements.txt          # Python dependencies
├── examples/
│   └── google_form_demo.py   # Working form filling demo
├── docs/
│   └── setup.md             # Detailed setup guide
├── browser_profiles/        # Local browser data (git ignored)
└── .gitignore              # Git ignore rules

🐛 Troubleshooting

Permission Issues

• Uses local browser_profiles/ directory
• No system config directory dependencies

Element Detection Failures

• Multiple fallback strategies implemented
• CSS selectors as robust backup
• Index-based targeting as final fallback

Browser Issues

• Automatically manages browser lifecycle
• Local profile prevents conflicts
• Detailed logging for debugging

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Test your changes thoroughly
4. Submit a pull request

📄 License

MIT License - see LICENSE file for details.

🎉 Success Story

This agent successfully demonstrated:

• Intelligent adaptation when MCP HTTP protocol failed
• Robust element detection using multiple strategies
• Real-world automation on live Google Forms
• Error recovery and graceful fallbacks

Result: 100% successful form completion with zero manual intervention! 🎯

🔗 Links

• Demo Video
• Documentation
• Issues
• Contributing Guide

---

Built with ❤️ for intelligent web automation