app.py

# Streamlit application
import streamlit as st
import os
from getpass import getpass
from transformers import pipeline

from llama_index.node_parser import SemanticSplitterNodeParser
from llama_index.embeddings import OpenAIEmbedding
from llama_index.ingestion import IngestionPipeline
from pinecone.grpc import PineconeGRPC
from pinecone import ServerlessSpec
from llama_index.vector_stores import PineconeVectorStore
from llama_index import VectorStoreIndex
from llama_index.retrievers import VectorIndexRetriever
from llama_index.query_engine import RetrieverQueryEngine

# Function to initialize the Pinecone and LlamaIndex setup
def initialize_pipeline():
 pinecone_api_key = os.getenv("PINECONE_API_KEY") 
 openai_api_key = os.getenv("OPENAI_API_KEY") 

 embed_model = OpenAIEmbedding(api_key=openai_api_key)
 pipeline = IngestionPipeline(
 transformations=[
 SemanticSplitterNodeParser(
 buffer_size=1,
 breakpoint_percentile_threshold=95,
 embed_model=embed_model,
 ),
 embed_model,
 ],
 )

 pc = PineconeGRPC(api_key=pinecone_api_key)
 index_name = "anualreport"
 pinecone_index = pc.Index(index_name)
 vector_store = PineconeVectorStore(pinecone_index=pinecone_index)
 pinecone_index.describe_index_stats()

 if not os.getenv('OPENAI_API_KEY'):
 os.environ['OPENAI_API_KEY'] = openai_api_key

 vector_index = VectorStoreIndex.from_vector_store(vector_store=vector_store)
 retriever = VectorIndexRetriever(index=vector_index, similarity_top_k=5)
 query_engine = RetrieverQueryEngine(retriever=retriever)
 
 return query_engine

# Streamlit UI
st.title("Chat with Annual Reports")

# Initialize the query engine
query_engine = initialize_pipeline()

# Conversation model using Hugging Face transformers
conversation_pipeline = pipeline("conversational", model="microsoft/DialoGPT-medium")

# User input
user_input = st.text_input("You: ", "")

if user_input:
 # Query the vector DB
 llm_query = query_engine.query(user_input)
 response = llm_query.response

 # Generate response using Hugging Face conversation model
 conversation = conversation_pipeline([user_input, response])
 bot_response = conversation[-1]["generated_text"]

 # Display response
 st.text_area("Bot: ", bot_response, height=200)