test_free_gaia.py

"""
Test script for Free GAIA Agent
Run this to verify everything works before processing the full dataset
"""

import sys
import requests
import wikipedia as wiki

def test_internet_connection():
    """Test internet connectivity"""
    print("🌐 Testing internet connection...")
    try:
        response = requests.get("https://www.google.com", timeout=5)
        print("✅ Internet connection working")
        return True
    except Exception as e:
        print(f"❌ Internet connection failed: {e}")
        return False

def test_duckduckgo_api():
    """Test DuckDuckGo API"""
    print("🦆 Testing DuckDuckGo API...")
    try:
        response = requests.get("https://api.duckduckgo.com/", params={
            "q": "capital of France",
            "format": "json"
        }, timeout=10)
        
        if response.status_code == 200:
            data = response.json()
            print("✅ DuckDuckGo API working")
            if data.get("Answer"):
                print(f"   Sample result: {data['Answer']}")
            return True
        else:
            print(f"⚠️  DuckDuckGo API returned status {response.status_code}")
            return False
    except Exception as e:
        print(f"❌ DuckDuckGo API failed: {e}")
        return False

def test_wikipedia_api():
    """Test Wikipedia API"""
    print("📚 Testing Wikipedia API...")
    try:
        # Search for a page
        results = wiki.search("Paris", results=1)
        if results:
            page = wiki.page(results[0])
            print("✅ Wikipedia API working")
            print(f"   Sample result: {page.title} - {page.content[:100]}...")
            return True
        else:
            print("⚠️  Wikipedia search returned no results")
            return False
    except Exception as e:
        print(f"❌ Wikipedia API failed: {e}")
        return False

def test_gaia_dataset():
    """Test GAIA dataset access"""
    print("📊 Testing GAIA dataset access...")
    try:
        from datasets import load_dataset
        dataset = load_dataset("gaia-benchmark/GAIA", "2023_all")
        print("✅ GAIA dataset access successful")
        print(f"   Validation questions: {len(dataset['validation'])}")
        print(f"   Test questions: {len(dataset['test'])}")
        
        # Show sample question
        sample = dataset["validation"][0]
        print(f"   Sample question: {sample['Question'][:100]}...")
        return True
    except Exception as e:
        print(f"❌ GAIA dataset access failed: {e}")
        print("💡 You may need to request access at:")
        print("   https://huggingface.co/datasets/gaia-benchmark/GAIA")
        return False

def test_free_agent():
    """Test the free agent with a simple question"""
    print("🤖 Testing Free GAIA Agent...")
    try:
        from free_gaia_agent import FreeGAIAAgent
        
        agent = FreeGAIAAgent()
        
        # Test simple math question
        test_question = "What is 15 + 27?"
        result = agent.solve_question(test_question, "test_001", "1")
        print(f"✅ Agent working - Test answer: {result['model_answer']}")
        
        # Test capital question
        test_question2 = "What is the capital of France?"
        result2 = agent.solve_question(test_question2, "test_002", "1")
        print(f"✅ Agent working - Capital answer: {result2['model_answer']}")
        
        return True
    except Exception as e:
        print(f"❌ Agent test failed: {e}")
        return False

def main():
    """Run all tests"""
    print("🧪 Free GAIA Agent Pre-flight Tests")
    print("=" * 40)
    
    tests = [
        ("Internet Connection", test_internet_connection),
        ("DuckDuckGo API", test_duckduckgo_api),
        ("Wikipedia API", test_wikipedia_api),
        ("GAIA Dataset", test_gaia_dataset),
        ("Free Agent", test_free_agent)
    ]
    
    results = []
    
    for test_name, test_func in tests:
        print(f"\n🔍 {test_name}")
        print("-" * 20)
        try:
            result = test_func()
            results.append(result)
        except Exception as e:
            print(f"❌ Test failed with exception: {e}")
            results.append(False)
    
    # Summary
    print("\n" + "=" * 40)
    print("📊 TEST SUMMARY")
    print("=" * 40)
    
    passed = sum(results)
    total = len(results)
    
    for i, (test_name, _) in enumerate(tests):
        status = "✅ PASS" if results[i] else "❌ FAIL"
        print(f"{test_name:20} {status}")
    
    print("-" * 40)
    print(f"Total: {passed}/{total} tests passed")
    
    if passed == total:
        print("\n🎉 All tests passed! Ready to run Free GAIA Agent!")
        print("📋 Next step: python3 free_gaia_agent.py")
    else:
        print(f"\n⚠️  {total - passed} test(s) failed. Please fix issues before proceeding.")
        print("\n💡 Common issues:")
        print("   - No internet connection")
        print("   - GAIA dataset access not granted")
        print("   - Missing Python packages")

if __name__ == "__main__":
    main()