""" Tests for the integrated GAIA agent. These tests verify that all components of the integrated agent work correctly and that the agent correctly handles all types of questions in the GAIA assessment. """ import os import sys import unittest from unittest.mock import patch, MagicMock import logging # Add parent directory to path to import the module sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) # Import the integrated agent from agent_integrated import GAIAIntegratedAgent # Disable logging for tests logging.disable(logging.CRITICAL) class TestIntegratedAgent(unittest.TestCase): """Test cases for the integrated GAIA agent.""" def setUp(self): """Set up the test environment.""" # Use a minimal configuration for testing self.test_config = { "verbose": True, "memory": { "use_supabase": False, "cache_enabled": True } } # Create the agent self.agent = GAIAIntegratedAgent(self.test_config) def tearDown(self): """Clean up after tests.""" if hasattr(self, 'agent'): self.agent.reset() def test_agent_initialization(self): """Test that the agent initializes correctly.""" self.assertTrue(self.agent.state["initialized"]) self.assertIsNotNone(self.agent.multimodal_processor) self.assertIsNotNone(self.agent.search_manager) self.assertIsNotNone(self.agent.memory_manager) self.assertIsNotNone(self.agent.text_analyzer) self.assertIsNotNone(self.agent.tools_registry) def test_reversed_text_questions(self): """Test that the agent correctly handles reversed text questions.""" questions = [ "What does this reversed text say: ELPPA", "Decode the following: .ANANAB si sihT" ] for question in questions: with self.subTest(question=question): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) # Check for expected phrases in the answer if "ELPPA" in question: self.assertIn("APPLE", answer.upper()) if "ANANAB" in question: self.assertIn("BANANA", answer.upper()) def test_unscramble_word_questions(self): """Test that the agent correctly handles word unscrambling questions.""" questions = [ "Unscramble this word: ELPPA", "What is the correct spelling of ANANAB?" ] for question in questions: with self.subTest(question=question): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) # Check for expected phrases in the answer if "ELPPA" in question: self.assertIn("APPLE", answer.upper()) if "ANANAB" in question: self.assertIn("BANANA", answer.upper()) @patch('src.gaia.agent.components.video_analyzer.VideoAnalyzer.analyze_video_content') def test_youtube_video_questions(self, mock_analyze): """Test that the agent correctly handles YouTube video questions.""" # Mock the video analyzer to return a known result mock_analyze.return_value = { "success": True, "content": "The video shows 3 bird species.", "metadata": { "title": "Bird Video", "duration": 120 } } questions = [ "In this YouTube video https://www.youtube.com/watch?v=dQw4w9WgXcQ, how many bird species are shown?", "What is the content of this video: https://youtu.be/dQw4w9WgXcQ" ] for question in questions: with self.subTest(question=question): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) self.assertIn("3", answer) @patch('src.gaia.agent.components.search_manager.SearchManager.search') def test_factual_questions(self, mock_search): """Test that the agent correctly handles factual questions.""" # Mock the search manager to return a known result mock_search.return_value = { "success": True, "answer": "Paris is the capital of France.", "sources": ["Wikipedia"] } questions = [ "What is the capital of France?", "Who wrote the novel '1984'?" ] for question in questions: with self.subTest(question=question): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) if "France" in question: self.assertIn("Paris", answer) @patch('src.gaia.agent.graph.run_agent_graph') def test_complex_questions(self, mock_graph): """Test that the agent correctly handles complex questions using LangGraph.""" # Mock the graph to return a known result mock_graph.return_value = { "answer": "The answer requires multiple steps of reasoning.", "reasoning": "Step 1...\nStep 2...\nStep 3...", "tool_results": [] } questions = [ "Compare and contrast renewable and non-renewable energy sources.", "What are the implications of quantum computing for cryptography?" ] for question in questions: with self.subTest(question=question): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) @patch('src.gaia.agent.components.image_analyzer.ImageAnalyzer.process_image') def test_image_questions(self, mock_process): """Test that the agent correctly handles image analysis questions.""" # Mock the image analyzer to return a known result mock_process.return_value = { "success": True, "description": "The image shows a mountain landscape.", "elements": ["mountain", "sky", "trees"], "analysis_type": "general" } questions = [ "Analyze this image: /path/to/image.jpg", "What is shown in this picture: /path/to/photo.png?" ] for question in questions: with self.subTest(question=question): # Patch os.path.exists to return True for our test path with patch('os.path.exists', return_value=True): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) self.assertIn("mountain", answer.lower()) @patch('src.gaia.agent.components.audio_analyzer.AudioAnalyzer.process_audio') def test_audio_questions(self, mock_process): """Test that the agent correctly handles audio analysis questions.""" # Mock the audio analyzer to return a known result mock_process.return_value = { "success": True, "transcription": "This is a test audio recording.", "audio_type": "speech", "speakers": ["Speaker 1"] } questions = [ "Transcribe this audio file: /path/to/recording.mp3", "What is said in this audio: /path/to/audio.wav?" ] for question in questions: with self.subTest(question=question): # Patch os.path.exists to return True for our test path with patch('os.path.exists', return_value=True): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) self.assertIn("test audio", answer.lower()) def test_error_handling(self): """Test that the agent correctly handles errors.""" # Create a scenario that would cause an error in processing with patch.object(self.agent, '_initialize_components', side_effect=RuntimeError("Test error")): with self.assertRaises(RuntimeError): agent = GAIAIntegratedAgent(self.test_config) # Test graceful error handling in process_question with patch.object(self.agent.text_analyzer, 'process_text_question', side_effect=Exception("Test error")): answer = self.agent.process_question("What does this reversed text say: ELPPA") self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) # Check that we got an error message (since verbose=True in test_config) self.assertIn("Error", answer) def test_memory_caching(self): """Test that the agent correctly caches and retrieves answers.""" # First process a question to cache the answer question = "What is 2+2?" # Mock the memory manager's methods with patch.object(self.agent.memory_manager, 'get_cached_answer', return_value=None): with patch.object(self.agent.memory_manager, 'cache_question_answer') as mock_cache: answer = self.agent.process_question(question) self.assertIsNotNone(answer) # Verify the cache was called mock_cache.assert_called_once() # Now process the same question again and verify it uses the cache with patch.object(self.agent.memory_manager, 'get_cached_answer', return_value="The answer is 4.") as mock_get: answer = self.agent.process_question(question) self.assertEqual(answer, "The answer is 4.") # Verify the cache was checked mock_get.assert_called_once() def test_evaluation_questions(self): """Test the agent against mock evaluation questions.""" # Sample evaluation questions evaluation_questions = [ "What is the capital of France?", # Factual "What does this reversed text say: ELPPA", # Reversed text "Unscramble this word: ANANAB", # Word unscrambling "In this YouTube video https://www.youtube.com/watch?v=dQw4w9WgXcQ, what's happening?", # Video "How many bird species are visible in this YouTube video: https://youtu.be/dQw4w9WgXcQ?", # Assessment question "How many studio albums did Mercedes Sosa release between 2000 and 2009?", # Assessment question "What's shown in this image: /path/to/image.jpg?", # Image analysis "What is the content of this audio file: /path/to/audio.mp3?", # Audio analysis "What is the position evaluation of this chess position: /path/to/chess.png?", # Chess analysis "Compare and contrast renewable and non-renewable energy sources." # Complex reasoning ] # Setup necessary mocks with patch('src.gaia.agent.components.video_analyzer.VideoAnalyzer.analyze_video_content', return_value={"success": True, "content": "Birds flying"}), \ patch('src.gaia.agent.components.image_analyzer.ImageAnalyzer.process_image', return_value={"success": True, "description": "An image"}), \ patch('src.gaia.agent.components.audio_analyzer.AudioAnalyzer.process_audio', return_value={"success": True, "transcription": "Audio content"}), \ patch('os.path.exists', return_value=True), \ patch('src.gaia.agent.components.search_manager.SearchManager.search', return_value={"success": True, "answer": "A factual answer"}), \ patch('src.gaia.agent.graph.run_agent_graph', return_value={"answer": "A complex answer"}): # Test each question for question in evaluation_questions: with self.subTest(question=question): answer = self.agent.process_question(question) self.assertIsNotNone(answer) self.assertTrue(len(answer) > 0) # Check for specific content in assessment questions if "Mercedes Sosa" in question: self.assertIn("7", answer) if "bird species" in question and "YouTube" in question: self.assertIn("3", answer) if __name__ == "__main__": unittest.main()