backend/recommend.py

import pickle
import numpy as np
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity

# Load model embeddings
with open("../models/product_embeddings.pkl", "rb") as f:
    data = pickle.load(f)

# Load transformer model
model = SentenceTransformer('all-MiniLM-L6-v2')

def recommend_products(user_query, top_n=5):
    """Find similar products based on user search"""
    query_embedding = model.encode(user_query).reshape(1, -1)
    
    # Compute similarity scores
    similarities = cosine_similarity(query_embedding, data["embeddings"])
    top_indices = np.argsort(similarities[0])[-top_n:][::-1]  # Get top matches

    recommendations = []
    for i in top_indices:
        recommendations.append({
            "search_query": data["search_queries"][i],
            "product": data["product_names"][i],
            "score": float(similarities[0][i])
        })

    return recommendations

# Example test
if __name__ == "__main__":
    query = "gaming laptop"
    results = recommend_products(query)
    for r in results:
        print(f"🔹 {r['product']} (Score: {r['score']:.2f})")