rag_utils.py

import json
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np

class RAGEngine:
    def __init__(self, json_path):
        self.embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
        with open(json_path, 'r') as f:
            self.data = json.load(f)
        self.texts = []
        self.build_corpus()
        self.build_index()

    def build_corpus(self):
        # Combine multiple fields for better context
        self.texts = [
            f"Product: {item['product_name']}\nCategory: {item['category']}\nPolicy: {item['return_policy']}\nReason: {item['return_reason']}"
            for item in self.data
        ]

    def build_index(self):
        embeddings = self.embedder.encode(self.texts, convert_to_numpy=True)
        dim = embeddings.shape[1]
        self.index = faiss.IndexFlatL2(dim)
        self.index.add(embeddings)

    def retrieve(self, query, top_k=3):
        query_emb = self.embedder.encode([query], convert_to_numpy=True)
        distances, indices = self.index.search(query_emb, top_k)
        results = [self.texts[idx] for idx in indices[0] if idx != -1]
        return results