test

@step
def evaluate_tla(specs: dict) -> dict:
    project_root = Path(__file__).resolve().parent.parent
    generated_dir = project_root / "outputs" / "generated"
    eval_output_dir = project_root / "outputs" / "evaluations"
    eval_output_dir.mkdir(parents=True, exist_ok=True)

    results = {}

    for model_name, tla_code in specs.items():
        print(f"\n--- Evaluating {model_name} ---")

        # Write the generated spec to file
        tla_path = generated_dir / f"{model_name}.generated.tla"
        tla_path.write_text(tla_code.strip())

        # Run TLC without a config file
        result = subprocess.run(
            ["tlc", "-nowarning", str(tla_path)],
            capture_output=True,
            text=True
        )

        # Write logs
        log_path = eval_output_dir / f"{model_name}.tlc.log"
        log_path.write_text(result.stdout + "\n=== STDERR ===\n" + result.stderr)

        # Evaluate result based on TLC output
        if "The specification is correct" in result.stdout:
            results[model_name] = "PASS"
        elif "TLC threw an unexpected exception" in result.stdout:
            results[model_name] = "ERROR"
        else:
            results[model_name] = "FAIL"

    return results