Browser Use Agent

A contextual web navigation agent built with Google's Agent Development Kit (ADK) that can navigate the web using contextual clues to achieve end objectives. The agent supports both cloud-based (Gemini) and local LLM models.

Features

• Contextual Web Navigation: Expert at finding and following contextual clues on web pages
• Flexible LLM Support: Works with both cloud-based (Gemini) and local models (Ollama)
• Browser Automation: Uses Playwright MCP tools for web interaction
• Adaptive Approach: Interprets ambiguous instructions and adapts strategies to reach goals

Prerequisites

• Python 3.11 or higher
• Node.js and npm (for Playwright MCP tools)
• Ollama (optional, for local LLM usage)

Installation

1. Clone the Repository

git clone <repository-url>
cd "Whatsapp Reader"

2. Set Up Python Environment

# Create virtual environment
python -m venv venv

# Activate virtual environment
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

3. Install Python Dependencies

pip install google-adk litellm python-dotenv

4. Install Node.js Dependencies

# Install Playwright MCP tools globally
npm install -g @playwright/mcp

5. Set Up Environment Variables

Create a .env file in the project root:

# For Google Gemini (cloud-based)
GOOGLE_API_KEY="your_google_api_key_here"

# For local LLM (optional)
OLLAMA_API_BASE="http://localhost:11434"

Usage

Using Cloud-Based Gemini Model (Default)

The agent is configured by default to use Google's Gemini 2.5 Flash model:

# In agent.py - already configured
model="gemini-2.5-flash"

Using Local LLM with Ollama

1. Install and Start Ollama:

   # Install Ollama (follow instructions at https://ollama.com/)
   # Pull a model (example with Gemma 3)
   ollama pull gemma3:12b
   # Start Ollama server
   ollama serve

2. Modify the Agent Configuration:

   # In agent.py, uncomment the local model line and comment out the Gemini line:
   # model="gemini-2.5-flash",
   model=LiteLlm(model="ollama_chat/gemma3:12b"),

3. Set Environment Variable:

   export OLLAMA_API_BASE="http://localhost:11434"

Running the Agent

Option 1: Using ADK Web Interface

adk web

Option 2: Using ADK CLI

adk run Whatsapp_agent.agent

Option 3: Programmatic Usage

from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai import types
from Whatsapp_agent.agent import root_agent

# Set up session and runner
session_service = InMemorySessionService()
session = session_service.create_session(
    app_name="whatsapp_reader",
    user_id="user_1",
    session_id="session_1"
)

runner = Runner(
    agent=root_agent,
    app_name="whatsapp_reader",
    session_service=session_service
)

# Run the agent
content = types.Content(role='user', parts=[types.Part(text="Navigate to example.com and find the contact information")])
events = runner.run(user_id="user_1", session_id="session_1", new_message=content)

for event in events:
    if event.is_final_response():
        print("Agent Response:", event.content.parts[0].text)

Configuration

Available Models

Cloud-Based Models

• gemini-2.5-flash (default) - Fast and efficient
• gemini-2.5-pro - More powerful for complex tasks
• gemini-2.0-flash - Stable version

Local Models (via Ollama)

• ollama_chat/gemma3:12b - Google's open-source model
• ollama_chat/llama3:8b - Meta's Llama 3
• ollama_chat/mistral:7b - Mistral AI's model
• ollama_chat/codellama:7b - Specialized for coding tasks

Tool Configuration

The agent uses Playwright MCP tools for web automation:

• browser_navigate - Navigate to web pages
• browser_screenshot - Take screenshots
• browser_fill - Fill forms
• browser_click - Click elements

You can filter specific tools by uncommenting the tool_filter line in the agent configuration.

Troubleshooting

Common Issues

1. Ollama Connection Issues:

   • Ensure Ollama is running: ollama serve
   • Check if the model is downloaded: ollama list
   • Verify API endpoint: curl http://localhost:11434/api/tags

2. Playwright MCP Issues:

   • Ensure Node.js and npm are installed
   • Reinstall Playwright MCP: npm install -g @playwright/mcp@latest

3. Python UTF-8 Issues (Windows):

   $env:PYTHONUTF8 = "1"

4. Tool Calling Issues with Ollama:

   • Use ollama_chat/ provider instead of ollama/
   • Ensure you're using the latest version of litellm

Debug Mode

The agent includes debug logging for LiteLLM. To see detailed request/response logs, the debug mode is already enabled in the code:

import litellm
litellm._turn_on_debug()

Project Structure

Whatsapp Reader/
├── Whatsapp_agent/
│   ├── __init__.py
│   └── agent.py          # Main agent configuration
├── README.md             # This file
└── .env                  # Environment variables (create this)

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test with both cloud and local models
5. Submit a pull request

License

[Add your license information here]

Support

For issues and questions:

• Check the Google ADK documentation
• Review LiteLLM documentation
• Check Ollama documentation