updated_rag.json

import json
import torch
import torch.nn.functional as F
from transformers import AutoTokenizer, AutoModel
import chromadb
from chromadb.utils import embedding_functions

def retrieve_relevant_context(user_query):
    # Load the query-answer data
    with open('prompt_sql_query.json', 'r') as f:
        sql_query_data = json.load(f)
    
    user_prompts = [item['prompt'] for item in sql_query_data]
    best_sql_queries = [item['sql_query'] for item in sql_query_data]
    relevant_tables = [item['redshift_tables'] for item in sql_query_data]
    
    # Load a pre-trained embedding model
    model_name = "sentence-transformers/all-MiniLM-L6-v2"
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModel.from_pretrained(model_name)
    
    def encode(texts):
        """Encode a list of texts into embeddings."""
        inputs = tokenizer(texts, padding=True, truncation=True, return_tensors="pt")
        with torch.no_grad():
            embeddings = model(**inputs).last_hidden_state[:, 0]
            embeddings = F.normalize(embeddings, p=2, dim=1)
        return embeddings.numpy().tolist()
    
    # Initialize ChromaDB client
    client = chromadb.PersistentClient(path="./chromadb_store")
    collection = client.get_or_create_collection(name="sql_queries")
    
    # Store embeddings in ChromaDB if not already stored
    if collection.count() == 0:
        for idx, prompt in enumerate(user_prompts):
            embedding = encode([prompt])[0]
            collection.add(embeddings=[embedding], documents=[prompt], ids=[str(idx)])
    
    # Query ChromaDB
    query_embedding = encode([user_query])[0]
    results = collection.query(query_embeddings=[query_embedding], n_results=3)
    
    retrieved_prompts = results['documents'][0]
    retrieved_indices = results['ids'][0]
    
    retrieved_sql_queries = [best_sql_queries[int(idx)] for idx in retrieved_indices]
    retrieved_tables = [relevant_tables[int(idx)] for idx in retrieved_indices]
    
    retrieved_prompt_sql_pairs = "Sample prompt - sql query pairs:\n"
    for idx, prompt in enumerate(retrieved_prompts):
        retrieved_prompt_sql_pairs += f"{idx + 1}. Prompt: {prompt}\n  SQL Query: {retrieved_sql_queries[idx]}\n\n"
    
    print(retrieved_prompt_sql_pairs)
    
    query_tables = list(set([item for sublist in retrieved_tables for item in sublist]))
    with open('redshift_tables.json', 'r') as f:
        schema_data = json.load(f)
    
    retrieved_tables_info = ""
    for index, table in enumerate(query_tables, start=1):
        retrieved_tables_info += f"{index}. {schema_data[0].get(table, 'Table schema not found')}\n\n"
    
    print(retrieved_tables_info)
    return retrieved_prompt_sql_pairs, retrieved_tables_info

# Example usage
# retrieve_relevant_context("What is the total IM device sold for DMI in first quarter of 2024?")