"""
Utility functions for agentic forensic controller.
"""

from typing import Dict


def convert_to_evidence_dict(features: Dict,
                            prediction_score: float,
                            prediction_label: int,
                            thresholds: Dict) -> Dict:
    """
    Convert raw features and prediction to evidence dict format.
    
    Args:
        features: Raw feature dictionary
        prediction_score: Prediction score (0-1)
        prediction_label: 0 for REAL, 1 for FAKE
        thresholds: Threshold dictionary for each feature
    
    Returns:
        Evidence dict with forensic_cues and feature_contributions
    """
    pred_label = 'FAKE' if prediction_label == 1 else 'REAL'
    
    # Build forensic cues
    forensic_cues = {}
    
    # Noiseprint mismatch
    if 'noiseprint_mismatch' in features:
        mismatch = features['noiseprint_mismatch']
        threshold = thresholds.get('noiseprint_mismatch', 2.5)
        forensic_cues['noiseprint_mismatch'] = {
            'value': mismatch,
            'threshold': threshold,
            'supports': 'fake' if mismatch > threshold else 'real'
        }
    
    # Residual energy
    if 'residual_energy_p95' in features:
        energy = features['residual_energy_p95']
        threshold = thresholds.get('residual_energy_p95', 0.08)
        forensic_cues['residual_energy_p95'] = {
            'value': energy,
            'threshold': threshold,
            'supports': 'fake' if energy > threshold else 'real'
        }
    
    # FFT peakiness
    if 'fft_peakiness' in features:
        peakiness = features['fft_peakiness']
        threshold = thresholds.get('fft_peakiness', 3.0)
        forensic_cues['fft_peakiness'] = {
            'value': peakiness,
            'threshold': threshold,
            'supports': 'fake' if peakiness > threshold else 'real'
        }
    
    # Feature contributions (simplified - could be more sophisticated)
    feature_contributions = {}
    for feat_name, feat_value in features.items():
        if feat_name in thresholds:
            threshold = thresholds[feat_name]
            # Contribution is how far from threshold (normalized)
            contribution = abs(feat_value - threshold) / max(threshold, 1.0)
            feature_contributions[feat_name] = contribution
    
    return {
        'forensic_cues': forensic_cues,
        'feature_contributions': feature_contributions,
        'prediction': {
            'label': pred_label,
            'score': prediction_score
        }
    }