app_1.py

import streamlit as st
import tempfile
import os
import shutil
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.text_splitter import CharacterTextSplitter
from langchain.vectorstores import FAISS
from langchain_community.document_loaders import WebBaseLoader
from langchain.chains.question_answering import load_qa_chain
# from langchain.llms import OpenAI
from langchain_openai import ChatOpenAI

# Streamlit UI
st.title("🔍 Chat with Any Website")

# User inputs
openai_api_key = st.text_input("Enter OpenAI API Key", type="password")
website_url = st.text_input("Enter Website URL")

# Temporary directory to store FAISS index
temp_dir = tempfile.gettempdir()
faiss_db_path = os.path.join(temp_dir, "faiss_index_dir")

# Ensure FAISS directory exists
if not os.path.exists(faiss_db_path):
    os.makedirs(faiss_db_path)

# Load embeddings if already created
if os.path.exists(os.path.join(faiss_db_path, "index.faiss")):
    docsearch = FAISS.load_local(faiss_db_path, OpenAIEmbeddings(), allow_dangerous_deserialization=True)
else:
    docsearch = None

if st.button("Build Embeddings") and openai_api_key and website_url:
    st.info("Fetching website data...")
    os.environ['OPENAI_API_KEY'] = openai_api_key
    
    # Load website data
    loader = WebBaseLoader(website_url)
    raw_text = loader.load()
    
    # Chunking the fetched text
    text_splitter = CharacterTextSplitter(separator='\n', chunk_size=500, chunk_overlap=50)
    docs = text_splitter.split_documents(raw_text)
    
    # Creating embeddings
    embeddings = OpenAIEmbeddings()
    docsearch = FAISS.from_documents(docs, embeddings)
    
    # Save FAISS index
    if os.path.exists(faiss_db_path):
        shutil.rmtree(faiss_db_path)
    os.makedirs(faiss_db_path)
    docsearch.save_local(faiss_db_path)
    
    st.success("Embeddings built and saved successfully!")

# Chat section
if docsearch:
    st.subheader("💬 Chat with Website")
    user_query = st.text_input("Enter your question")
    if st.button("Get Answer") and user_query:
        # chain = load_qa_chain(OpenAI(), chain_type="stuff")
        chain = load_qa_chain(ChatOpenAI(model="gpt-4o"), chain_type="stuff")
        docs = docsearch.similarity_search(user_query)
        response = chain.run(input_documents=docs, question=user_query)
        st.write("**Response:**", response)