core/predictor.py

# FILE: ai-service/core/predictor.py (REPLACE EVERYTHING IN YOUR FILE WITH THIS)

import joblib
import pandas as pd
from typing import List, Dict

print(">>> Loading ai-service/core/predictor.py (Version: FINAL, COMPLETE)")

try:
    # Model ko load karna
    influencer_pipeline = joblib.load('models/influencer_matcher_v1.joblib')
    print("--- Predictor: Influencer Matcher model loaded successfully. ---")
except FileNotFoundError as e:
    print(f"--- Predictor FATAL ERROR: Model file not found: {e}. Predictions will fail. ---")
    influencer_pipeline = None

# Performance predictor ko bhi yahan theek se load karte hain
try:
    performance_pipeline = joblib.load('models/performance_predictor_v1.joblib')
    print("--- Predictor: Performance Predictor model loaded successfully. ---")
except FileNotFoundError:
    performance_pipeline = None

def rank_influencers_by_match(influencers: List[Dict], campaign_details: Dict, top_n: int = 5) -> List[Dict]:
    """
    Influencers ko rank karta hai, model ko saari zaroori details dekar.
    """
    print(f"--- Predictor Skill: Ranking {len(influencers)} influencers...")
    
    if not influencers or influencer_pipeline is None:
        return []
        
    try:
        # Step 1: Influencers ki list se DataFrame banayein
        influencer_df = pd.DataFrame(influencers)
        
        # === ✨ YEH AAKHRI AUR SABSE ZAROORI FIX HAI ✨ ===
        # Model ko woh saari jaankari de rahe hain jo use chahiye.
        
        # 1. Influencer ki taraf se aane wali jaankari
        features = influencer_df[['category', 'bio']].copy()
        
        # 2. Campaign ki taraf se aane wali jaankari (jo har influencer ke liye same hogi)
        # Hum default values bhi de rahe hain agar backend se data na aaye
        features['niche'] = campaign_details.get('category', '') # Assume campaign category maps to niche
        features['country'] = campaign_details.get('location', 'USA')
        features['followers'] = campaign_details.get('followers', 10000)
        features['engagement_rate'] = campaign_details.get('engagement_rate', 0.03)

        print(f"--- Predictor Skill: Preparing features for model: {features.columns.to_list()}")
        
        # Step 2: Sahi features ke saath predict karein
        match_scores = influencer_pipeline.predict(features)
        influencer_df['match_score'] = match_scores
        
        # Step 3: Score ke hisaab se sort karke top influencers nikalein
        top_influencers_df = influencer_df.sort_values(by='match_score', ascending=False).head(top_n)

        # Step 4: Jawab ko saaf-suthre format mein wapas bhejein
        result_cols = ['id', 'name', 'handle', 'followers', 'category', 'bio']
        # Jo columns exist nahi karte, unhein a gracefully handle karein
        final_cols = [col for col in result_cols if col in top_influencers_df.columns]
        results = top_influencers_df[final_cols].to_dict(orient='records')
        
        print(f"--- Predictor Skill: Successfully ranked and returning top {len(results)} influencers.")
        return results

    except Exception as e:
        print(f"--- Predictor Skill ERROR: Failed during prediction process. Error: {e}")
        import traceback
        traceback.print_exc()
        return []

def predict_performance(data: dict) -> int:
    """
    Campaign details ke aadhar par 'final_reach' predict karta hai.
    """
    if performance_pipeline is None:
        return 800000 # Fallback value agar model load na ho

    df = pd.DataFrame(data, index=[0])
    prediction = performance_pipeline.predict(df)
    return int(prediction[0])