app.py

import os
from flask import Flask, request, jsonify
import logging
import openai
from llama_index.core import Settings
from llama_index.llms.openai import OpenAI
from llama_index.vector_stores.pinecone import PineconeVectorStore
from llama_index.core import VectorStoreIndex, StorageContext
from dotenv import load_dotenv
from pinecone import Pinecone
from llama_index.embeddings.openai import OpenAIEmbedding

load_dotenv()

app = Flask(__name__)
app.debug = True

logging.basicConfig(level=logging.DEBUG)

openai.api_key = os.getenv('OPENAI_API_KEY')

pc = Pinecone(
    api_key=os.getenv('PINECONE_API_KEY')
)

PINECONE_INDEX_NAME = os.getenv('PINECONE_INDEX')

pinecone_index = pc.Index(PINECONE_INDEX_NAME)

# After initializing the Pinecone index
stats = pinecone_index.describe_index_stats()
app.logger.debug(f"Pinecone index stats: {stats}")

# Set up LlamaIndex global settings
Settings.llm = OpenAI(
    model=os.getenv('OPENAI_MODEL', 'gpt-3.5-turbo'),
    temperature=0
)

vector_store = PineconeVectorStore(
    pinecone_index=pinecone_index,
    namespace="ai"
)

# Create Storage Context with the Vector Store
storage_context = StorageContext.from_defaults(vector_store=vector_store)

# Initialize LlamaIndex with the existing Pinecone vector store
index = VectorStoreIndex.from_vector_store(
    vector_store=vector_store,
    storage_context=storage_context
)

@app.route('/run/predict', methods=['POST'])
def predict():
    try:
        data = request.json
        app.logger.debug(f"Received data: {data}") # data => {'query': 'What is LangChain?'}

        if not data:
            app.logger.error("No data provided in the request.")
            return jsonify({'error': 'No data provided.'}), 400

        user_query = data.get('query')

        if not user_query:
            app.logger.error("No query provided in the request.")
            return jsonify({'error': 'No query provided.'}), 400

        # Log Pinecone query details
        app.logger.debug(f"Querying Pinecone index: {PINECONE_INDEX_NAME}")
        app.logger.debug(f"Query: {user_query}")

        # Perform the query using LlamaIndex
        query_engine = index.as_query_engine(similarity_top_k=5)
        app.logger.debug(f"Query engine: {query_engine}")
        
        response = query_engine.query(user_query)
        app.logger.debug(f"Raw response object: {response}")
        app.logger.debug(f"Response type: {type(response)}")
        
        if hasattr(response, 'source_nodes'):
            app.logger.debug(f"Number of source nodes: {len(response.source_nodes)}")
            for i, node in enumerate(response.source_nodes):
                app.logger.debug(f"Source node {i}: {node.node.text[:100]}...")  # Log first 100 chars of each source node
        else:
            app.logger.warning("No source nodes found in the response")

        if hasattr(response, 'response'):
            response_text = response.response
        else:
            response_text = str(response)
        
        app.logger.debug(f"Response text: {response_text}")

        return jsonify({'response': response_text})

    except Exception as e:
        app.logger.error(f"Error processing request: {e}", exc_info=True)
        return jsonify({"error": "An error occurred while processing the request"}), 500

@app.route('/empty-datastore', methods=['DELETE'])
def empty_datastore():
    try:
        # Attempt to delete all vectors in the default namespace
        delete_response = pinecone_index.delete(delete_all=True, namespace="")
        app.logger.debug(f"Delete response: {delete_response}")
        
        # Verify the index is empty
        stats = pinecone_index.describe_index_stats()
        app.logger.debug(f"Index stats after deletion: {stats}")
        
        if stats['total_vector_count'] == 0:
            app.logger.info("Datastore emptied successfully.")
            return jsonify({'message': 'Datastore emptied successfully'}), 200
        else:
            app.logger.warning("Datastore not fully emptied.")
            return jsonify({'message': 'Datastore not fully emptied'}), 500
    
    except Exception as e:
        app.logger.error(f"Error emptying datastore: {e}")
        return jsonify({'error': f'An error occurred while emptying the datastore: {str(e)}'}), 500

if __name__ == '__main__':
    from os import environ
    app.run(host='0.0.0.0', port=int(environ.get('PORT', 7860)))