tests/test_evaluators/test_empathy_evaluators.py

"""
Test script for Empathy Evaluators (ER, IP, EX)

Usage:
    python tests/test_evaluators/test_empathy_evaluators.py
"""
import sys
import os

# Add parent directory to path for imports
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '../..')))

from evaluators import create_evaluator
from custom_types import Utterance


# Test conversation with seeker-response pairs
test_conversation: list[Utterance] = [
    {
        "speaker": "Patient",
        "text": "I've been feeling really anxious lately and can't sleep."
    },
    {
        "speaker": "Therapist",
        "text": "I understand how difficult that must be. Have you tried any relaxation techniques?"
    },
    {
        "speaker": "Patient",
        "text": "I'm struggling with depression."
    },
    {
        "speaker": "Therapist",
        "text": "Just think positive thoughts!"
    },
    {
        "speaker": "Patient",
        "text": "I feel like nobody understands me."
    },
    {
        "speaker": "Therapist",
        "text": "It sounds like you're feeling very isolated. Can you tell me more about what makes you feel that way?"
    }
]


def test_single_evaluator(metric_name: str, label: str):
    """Test a single empathy evaluator."""
    print(f"\n{'='*80}")
    print(f"Testing {label}")
    print(f"{'='*80}")
    
    try:
        # Create evaluator
        print(f"Creating {metric_name} evaluator...")
        evaluator = create_evaluator(metric_name)
        
        if not evaluator:
            print(f"❌ Failed to create evaluator for {metric_name}")
            return False
        
        print(f"✓ Evaluator created successfully")
        print(f"  Model: {evaluator.MODEL_NAME}")
        
        # Execute evaluation
        print(f"\nEvaluating conversation ({len(test_conversation)} utterances)...")
        result = evaluator.execute(test_conversation)
        
        # Check result structure
        assert result["granularity"] == "utterance", "Expected utterance-level granularity"
        assert result["per_utterance"] is not None, "Expected per_utterance results"
        assert len(result["per_utterance"]) == len(test_conversation), "Mismatch in result count"
        
        print(f"✓ Evaluation complete")
        
        # Display results
        print(f"\nResults:")
        for i, (utt, utt_result) in enumerate(zip(test_conversation, result["per_utterance"])):
            print(f"\n  Utterance {i+1}:")
            print(f"    Speaker: {utt['speaker']}")
            print(f"    Text: {utt['text'][:60]}{'...' if len(utt['text']) > 60 else ''}")
            
            if metric_name in utt_result["metrics"]:
                score = utt_result["metrics"][metric_name]
                print(f"    {metric_name}: {score['label']} (confidence: {score['confidence']:.3f})")
            else:
                print(f"    {metric_name}: (not evaluated)")
        
        print(f"\n✅ {label} test passed!")
        return True
        
    except Exception as e:
        print(f"\n❌ Error testing {label}: {e}")
        import traceback
        traceback.print_exc()
        return False


def test_all_empathy_evaluators():
    """Test all three empathy evaluators."""
    print("\n" + "="*80)
    print("Testing All Empathy Evaluators")
    print("="*80)
    
    evaluators = [
        ("empathy_er", "Empathy ER (Emotional Reaction)"),
        ("empathy_ip", "Empathy IP (Interpretation)"),
        ("empathy_ex", "Empathy EX (Exploration)")
    ]
    
    results = {}
    for metric_name, label in evaluators:
        results[metric_name] = test_single_evaluator(metric_name, label)
    
    # Summary
    print(f"\n{'='*80}")
    print("Test Summary")
    print(f"{'='*80}")
    
    for metric_name, label in evaluators:
        status = "✅ PASSED" if results[metric_name] else "❌ FAILED"
        print(f"  {label}: {status}")
    
    total = len(evaluators)
    passed = sum(results.values())
    print(f"\nTotal: {passed}/{total} tests passed")
    print(f"{'='*80}\n")
    
    return all(results.values())


if __name__ == "__main__":
    success = test_all_empathy_evaluators()
    sys.exit(0 if success else 1)