broswer-automation/agent-livekit/test_field_workflow.py

#!/usr/bin/env python3
"""
Test script for the enhanced field workflow functionality.

This script demonstrates how to use the new execute_field_workflow method
to handle missing webpage fields with automatic MCP-based detection.
"""

import asyncio
import logging
import json
from mcp_chrome_client import MCPChromeClient

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)


async def test_field_workflow():
    """Test the enhanced field workflow with various scenarios."""
    
    # Initialize MCP Chrome client
    chrome_config = {
        'mcp_server_type': 'chrome_extension',
        'mcp_server_url': 'http://localhost:3000',
        'mcp_server_command': '',
        'mcp_server_args': []
    }
    
    client = MCPChromeClient(chrome_config)
    
    try:
        # Test scenarios
        test_scenarios = [
            {
                "name": "Google Search Workflow",
                "url": "https://www.google.com",
                "field_name": "search",
                "field_value": "LiveKit agent automation",
                "actions": [
                    {"type": "keyboard", "target": "Enter"}
                ]
            },
            {
                "name": "Login Form Workflow",
                "url": "https://example.com/login",
                "field_name": "email",
                "field_value": "test@example.com",
                "actions": [
                    {"type": "wait", "target": "1"},
                    {"type": "click", "target": "input[name='password']"},
                    {"type": "wait", "target": "0.5"},
                    {"type": "submit"}
                ]
            },
            {
                "name": "Contact Form Workflow",
                "url": "https://example.com/contact",
                "field_name": "message",
                "field_value": "Hello, this is a test message from the LiveKit agent.",
                "actions": [
                    {"type": "click", "target": "button[type='submit']"}
                ]
            }
        ]
        
        for scenario in test_scenarios:
            logger.info(f"\n{'='*50}")
            logger.info(f"Testing: {scenario['name']}")
            logger.info(f"{'='*50}")
            
            # Navigate to the test URL
            logger.info(f"Navigating to: {scenario['url']}")
            nav_result = await client._navigate_mcp(scenario['url'])
            logger.info(f"Navigation result: {nav_result}")
            
            # Wait for page to load
            await asyncio.sleep(3)
            
            # Execute the field workflow
            logger.info(f"Executing workflow for field: {scenario['field_name']}")
            workflow_result = await client.execute_field_workflow(
                field_name=scenario['field_name'],
                field_value=scenario['field_value'],
                actions=scenario['actions'],
                max_retries=3
            )
            
            # Display results
            logger.info("Workflow Results:")
            logger.info(f"  Success: {workflow_result['success']}")
            logger.info(f"  Field Filled: {workflow_result['field_filled']}")
            logger.info(f"  Detection Method: {workflow_result.get('detection_method', 'N/A')}")
            logger.info(f"  Execution Time: {workflow_result['execution_time']:.2f}s")
            
            if workflow_result['field_selector']:
                logger.info(f"  Field Selector: {workflow_result['field_selector']}")
            
            if workflow_result['actions_executed']:
                logger.info(f"  Actions Executed: {len(workflow_result['actions_executed'])}")
                for i, action in enumerate(workflow_result['actions_executed']):
                    status = "✓" if action['success'] else "✗"
                    logger.info(f"    {i+1}. {status} {action['action_type']}: {action.get('target', 'N/A')}")
            
            if workflow_result['errors']:
                logger.warning("  Errors:")
                for error in workflow_result['errors']:
                    logger.warning(f"    - {error}")
            
            # Wait between tests
            await asyncio.sleep(2)
            
    except Exception as e:
        logger.error(f"Test execution error: {e}")
    finally:
        # Cleanup
        logger.info("Test completed")


async def test_workflow_with_json_actions():
    """Test the workflow with JSON-formatted actions (as used by the LiveKit agent)."""
    
    chrome_config = {
        'mcp_server_type': 'chrome_extension',
        'mcp_server_url': 'http://localhost:3000',
        'mcp_server_command': '',
        'mcp_server_args': []
    }
    
    client = MCPChromeClient(chrome_config)
    
    try:
        # Navigate to Google
        await client._navigate_mcp("https://www.google.com")
        await asyncio.sleep(3)
        
        # Test with JSON actions (simulating LiveKit agent call)
        actions_json = json.dumps([
            {"type": "keyboard", "target": "Enter", "delay": 0.5}
        ])
        
        # This simulates how the LiveKit agent would call the workflow
        logger.info("Testing workflow with JSON actions...")
        
        # Parse actions (as done in the LiveKit agent)
        parsed_actions = json.loads(actions_json)
        
        result = await client.execute_field_workflow(
            field_name="search",
            field_value="MCP Chrome automation",
            actions=parsed_actions,
            max_retries=3
        )
        
        logger.info(f"Workflow result: {json.dumps(result, indent=2)}")
        
    except Exception as e:
        logger.error(f"JSON actions test error: {e}")


if __name__ == "__main__":
    logger.info("Starting enhanced field workflow tests...")
    
    # Run the tests
    asyncio.run(test_field_workflow())
    
    logger.info("\nTesting JSON actions format...")
    asyncio.run(test_workflow_with_json_actions())
    
    logger.info("All tests completed!")