debug_prediction.py

#!/usr/bin/env python3
"""
Quick test to see what the model is actually predicting
"""

import requests
import json

BASE_URL = "http://localhost:8000"

def test_single_prediction():
    """Test a single prediction to see raw output"""
    
    # Test obvious hallucination
    payload = {
        "prompt": "Question: What is the capital of France?",
        "response": "London is the capital of France",
        "question": "What is the capital of France?"
    }
    
    print("🧪 Testing obvious hallucination:")
    print(f"Question: {payload['question']}")
    print(f"Response: {payload['response']}")
    print(f"Expected: HALLUCINATION")
    print()
    
    try:
        response = requests.post(f"{BASE_URL}/api/debug-predict", json=payload, timeout=10)
        if response.status_code == 200:
            result = response.json()
            print("📊 Debug Results:")
            print(f"Raw Model Output: '{result.get('raw_model_output', '')}'")
            print(f"Extracted Prediction: '{result.get('extracted_prediction', '')}'")
            print(f"Is Hallucination: {result.get('is_hallucination', False)}")
            print(f"Confidence Score: {result.get('confidence_score', 0):.3f}")
            print()
            
            # Also test regular API
            response2 = requests.post(f"{BASE_URL}/api/predict", json=payload, timeout=10)
            if response2.status_code == 200:
                result2 = response2.json()
                print("📊 Regular API Results:")
                print(f"Is Hallucination: {result2.get('is_hallucination', False)}")
                print(f"Confidence Score: {result2.get('confidence_score', 0):.3f}")
                print(f"Raw Prediction: '{result2.get('raw_prediction', '')}'")
        else:
            print(f"❌ Error: {response.status_code}")
            print(response.text)
            
    except Exception as e:
        print(f"❌ Error: {e}")

if __name__ == "__main__":
    test_single_prediction()