import sys import numpy as np from environment import DocumentClassificationEnv from grading import AgentGrader from models import Observation, Action, Reward, EnvironmentState from agent import TicketAgent def test_environment_creation(): print("Testing environment creation...") for difficulty in ["easy", "medium", "hard"]: try: env = DocumentClassificationEnv(task_difficulty=difficulty) obs, info = env.reset() print(f" ✓ {difficulty} environment created successfully") assert "features" in obs, f"Missing features in {difficulty} observation" assert "content" in obs, f"Missing content in {difficulty} observation" except Exception as e: print(f" ✗ Failed to create {difficulty} environment: {e}") return False return True def test_step_function(): print("\nTesting step function...") env = DocumentClassificationEnv(task_difficulty="easy") obs, _ = env.reset() for i in range(5): action = i % env.num_categories obs, reward, done, truncated, info = env.step(action) assert isinstance(reward, float), f"Reward should be float, got {type(reward)}" assert "is_correct" in info, "Missing is_correct in info" print(f" ✓ Step {i+1}: reward={reward:.3f}, accuracy={info.get('episode_accuracy', 0):.3f}") if done: break return True def test_ticket_agent_loop(): print("\nTesting TicketAgent interaction loop...") agent = TicketAgent(model_dir=".") env = DocumentClassificationEnv(task_difficulty="easy") state = agent.run_episode(env) print(f" ✓ Agent finished episode. Accuracy: {state['accuracy']:.2%}, Total Reward: {state['total_reward']:.2f}") assert state['total_documents'] > 0 assert 'accuracy' in state return True def test_pydantic_models(): print("\nTesting Pydantic typed models...") env = DocumentClassificationEnv(task_difficulty="easy") obs_raw, _ = env.reset() obs_typed = Observation.from_gym_obs(obs_raw) assert obs_typed.document_id != "" print(f" ✓ Observation model: doc_id={obs_typed.document_id}") state_raw = env.state() state_typed = EnvironmentState.from_state_dict(state_raw) assert state_typed.difficulty == "easy" print(f" ✓ EnvironmentState model: difficulty={state_typed.difficulty}, accuracy={state_typed.accuracy:.3f}") return True def run_all_tests(): print("=" * 60) print("Document Classification Environment - Fixed Test Suite") print("=" * 60) tests = [ ("Environment Creation", test_environment_creation), ("Step Function", test_step_function), ("TicketAgent Loop", test_ticket_agent_loop), ("Pydantic Models", test_pydantic_models), ] results = [] for name, func in tests: try: res = func() results.append((name, res)) except Exception as e: print(f" ✗ {name} failed: {e}") import traceback traceback.print_exc() results.append((name, False)) print("\nSUMMARY") passed = sum(1 for _, r in results if r) for name, r in results: print(f" [{'PASS' if r else 'FAIL'}] {name}") print(f"\nTotal: {passed}/{len(tests)} passed") return passed == len(tests) if __name__ == "__main__": success = run_all_tests() sys.exit(0 if success else 1)