broswer-automation/agent-livekit/test_realtime_form_discovery.py

#!/usr/bin/env python3
"""
Test script for REAL-TIME form discovery capabilities.

This script tests the enhanced form filling system that:
1. NEVER uses cached selectors
2. Always uses real-time MCP tools for discovery
3. Gets fresh selectors on every request
4. Uses chrome_get_interactive_elements and chrome_get_content_web_form
"""

import asyncio
import logging
import sys
import os

# Add the current directory to the path so we can import our modules
sys.path.append(os.path.dirname(os.path.abspath(__file__)))

from mcp_chrome_client import MCPChromeClient

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

async def test_realtime_discovery():
    """Test the real-time form discovery capabilities"""
    
    # Initialize MCP Chrome client
    client = MCPChromeClient(
        server_type="http",
        server_url="http://127.0.0.1:12306/mcp"
    )
    
    try:
        # Connect to MCP server
        logger.info("Connecting to MCP server...")
        await client.connect()
        logger.info("Connected successfully!")
        
        # Test 1: Navigate to Google (fresh page)
        logger.info("=== Test 1: Navigate to Google ===")
        result = await client._navigate_mcp("https://www.google.com")
        logger.info(f"Navigation result: {result}")
        await asyncio.sleep(3)  # Wait for page to load
        
        # Test 2: Real-time discovery for search field (NO CACHE)
        logger.info("=== Test 2: Real-time discovery for search field ===")
        discovery_result = await client._discover_form_fields_dynamically("search", "python programming")
        logger.info(f"Real-time discovery result: {discovery_result}")
        
        # Test 3: Fill field using ONLY real-time discovery
        logger.info("=== Test 3: Fill field using ONLY real-time discovery ===")
        fill_result = await client.fill_field_by_name("search", "machine learning")
        logger.info(f"Real-time fill result: {fill_result}")
        
        # Test 4: Direct MCP element search
        logger.info("=== Test 4: Direct MCP element search ===")
        direct_result = await client._direct_mcp_element_search("search", "artificial intelligence")
        logger.info(f"Direct search result: {direct_result}")
        
        # Test 5: Navigate to different site and test real-time discovery
        logger.info("=== Test 5: Test real-time discovery on GitHub ===")
        result = await client._navigate_mcp("https://www.github.com")
        logger.info(f"GitHub navigation result: {result}")
        await asyncio.sleep(3)
        
        # Real-time discovery on GitHub
        github_discovery = await client._discover_form_fields_dynamically("search", "python")
        logger.info(f"GitHub real-time discovery: {github_discovery}")
        
        # Test 6: Test very flexible matching
        logger.info("=== Test 6: Test very flexible matching ===")
        flexible_result = await client._direct_mcp_element_search("query", "test search")
        logger.info(f"Flexible matching result: {flexible_result}")
        
        # Test 7: Test common selectors generation
        logger.info("=== Test 7: Test common selectors generation ===")
        common_selectors = client._generate_common_selectors("search")
        logger.info(f"Generated common selectors: {common_selectors[:10]}")  # Show first 10
        
        # Test 8: Navigate to a form-heavy site
        logger.info("=== Test 8: Test on form-heavy site ===")
        result = await client._navigate_mcp("https://httpbin.org/forms/post")
        logger.info(f"Form site navigation result: {result}")
        await asyncio.sleep(3)
        
        # Test real-time discovery on form fields
        form_fields = ["email", "password", "comment"]
        for field in form_fields:
            logger.info(f"Testing real-time discovery for field: {field}")
            field_result = await client._discover_form_fields_dynamically(field, f"test_{field}")
            logger.info(f"Field '{field}' discovery: {field_result}")
        
        logger.info("=== All real-time discovery tests completed! ===")
        
    except Exception as e:
        logger.error(f"Test failed with error: {e}")
        import traceback
        traceback.print_exc()
    
    finally:
        # Disconnect from MCP server
        try:
            await client.disconnect()
            logger.info("Disconnected from MCP server")
        except Exception as e:
            logger.error(f"Error disconnecting: {e}")

async def test_mcp_tools_directly():
    """Test MCP tools directly to verify real-time capabilities"""
    logger.info("=== Testing MCP tools directly ===")
    
    client = MCPChromeClient(server_type="http", server_url="http://127.0.0.1:12306/mcp")
    
    try:
        await client.connect()
        
        # Navigate to Google
        await client._navigate_mcp("https://www.google.com")
        await asyncio.sleep(3)
        
        # Test chrome_get_interactive_elements directly
        logger.info("Testing chrome_get_interactive_elements...")
        interactive_result = await client._call_mcp_tool("chrome_get_interactive_elements", {
            "types": ["input", "textarea", "select"]
        })
        
        if interactive_result and "elements" in interactive_result:
            elements = interactive_result["elements"]
            logger.info(f"Found {len(elements)} interactive elements")
            
            for i, element in enumerate(elements[:5]):  # Show first 5
                attrs = element.get("attributes", {})
                logger.info(f"Element {i+1}: {element.get('tagName')} - name: {attrs.get('name')}, id: {attrs.get('id')}, type: {attrs.get('type')}")
        
        # Test chrome_get_content_web_form directly
        logger.info("Testing chrome_get_content_web_form...")
        form_result = await client._call_mcp_tool("chrome_get_content_web_form", {})
        
        if form_result:
            logger.info(f"Form content result: {str(form_result)[:200]}...")  # Show first 200 chars
        
        # Test chrome_get_web_content for all inputs
        logger.info("Testing chrome_get_web_content for all inputs...")
        content_result = await client._call_mcp_tool("chrome_get_web_content", {
            "selector": "input, textarea, select",
            "textOnly": False
        })
        
        if content_result:
            logger.info(f"Web content result: {str(content_result)[:200]}...")  # Show first 200 chars
        
    except Exception as e:
        logger.error(f"Direct MCP tool test failed: {e}")
        import traceback
        traceback.print_exc()
    
    finally:
        try:
            await client.disconnect()
        except Exception:
            pass

async def test_field_matching_algorithms():
    """Test the field matching algorithms"""
    logger.info("=== Testing field matching algorithms ===")
    
    client = MCPChromeClient(server_type="http", server_url="http://127.0.0.1:12306/mcp")
    
    # Test elements (simulated)
    test_elements = [
        {
            "tagName": "input",
            "attributes": {
                "name": "q",
                "type": "search",
                "placeholder": "Search Google or type a URL",
                "aria-label": "Search"
            }
        },
        {
            "tagName": "input",
            "attributes": {
                "name": "email",
                "type": "email",
                "placeholder": "Enter your email address"
            }
        },
        {
            "tagName": "input",
            "attributes": {
                "name": "user_password",
                "type": "password",
                "placeholder": "Password"
            }
        },
        {
            "tagName": "textarea",
            "attributes": {
                "name": "message",
                "placeholder": "Type your message here",
                "aria-label": "Message"
            }
        }
    ]
    
    test_field_names = [
        "search", "query", "q",
        "email", "mail", "e-mail",
        "password", "pass", "user password",
        "message", "comment", "text"
    ]
    
    logger.info("Testing standard field matching...")
    for field_name in test_field_names:
        logger.info(f"\nTesting field name: '{field_name}'")
        for i, element in enumerate(test_elements):
            is_match = client._is_field_match(element, field_name.lower())
            selector = client._extract_best_selector(element)
            logger.info(f"  Element {i+1} ({element['tagName']}): Match={is_match}, Selector={selector}")
    
    logger.info("\nTesting very flexible matching...")
    for field_name in test_field_names:
        logger.info(f"\nTesting flexible field name: '{field_name}'")
        for i, element in enumerate(test_elements):
            is_match = client._is_very_flexible_match(element, field_name.lower())
            logger.info(f"  Element {i+1} ({element['tagName']}): Flexible Match={is_match}")

def main():
    """Main function to run the tests"""
    logger.info("Starting REAL-TIME form discovery tests...")
    
    # Check if MCP server is likely running
    import socket
    try:
        sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        sock.settimeout(1)
        result = sock.connect_ex(('127.0.0.1', 12306))
        sock.close()
        if result != 0:
            logger.warning("MCP server doesn't appear to be running on port 12306")
            logger.warning("Please start the MCP server before running this test")
            return
    except Exception as e:
        logger.warning(f"Could not check MCP server status: {e}")
    
    # Run the tests
    asyncio.run(test_field_matching_algorithms())
    asyncio.run(test_mcp_tools_directly())
    asyncio.run(test_realtime_discovery())

if __name__ == "__main__":
    main()