"""MrrrMe Backend - Emotion Fusion Logic"""
import numpy as np
from ..config import EMOTION_MAP, FUSION_WEIGHTS

def adjust_fusion_weights(face_quality: float, voice_active: bool, text_length: int) -> dict:
    """
    Adjust fusion weights based on quality metrics
    
    Returns:
        Dict with adjusted weights and adjustment log
    """
    adjusted_weights = FUSION_WEIGHTS.copy()
    adjustments = []
    
    # Reduce face weight if quality is poor
    if face_quality < 0.5:
        adjusted_weights['face'] *= 0.7
        adjustments.append(f"Face weight reduced (low quality: {face_quality:.3f})")
    
    # Reduce voice weight if not active
    if not voice_active:
        adjusted_weights['voice'] *= 0.5
        adjustments.append("Voice weight reduced (no recent speech)")
    
    # Reduce text weight if very short
    if text_length < 10:
        adjusted_weights['text'] *= 0.7
        adjustments.append(f"Text weight reduced (short input: {text_length} chars)")
    
    # Normalize to sum to 1.0
    total = sum(adjusted_weights.values())
    final_weights = {k: v/total for k, v in adjusted_weights.items()}
    
    return {
        'weights': final_weights,
        'adjustments': adjustments
    }

def calculate_fusion(face_probs, voice_probs, text_probs, weights: dict):
    """Calculate weighted fusion of emotion probabilities"""
    fused_probs = (
        weights['face'] * face_probs +
        weights['voice'] * voice_probs +
        weights['text'] * text_probs
    )
    fused_probs = fused_probs / (np.sum(fused_probs) + 1e-8)
    
    fused_idx = int(np.argmax(fused_probs))
    emotions = ['Neutral', 'Happy', 'Sad', 'Angry']
    fused_emotion = emotions[fused_idx]
    intensity = float(np.max(fused_probs))
    
    return fused_emotion, intensity, fused_probs