app.py

import streamlit as st
from dotenv import load_dotenv
from langchain_community.document_loaders.url import UnstructuredURLLoader
from langchain_huggingface.embeddings import HuggingFaceEndpointEmbeddings
from langchain_community.vectorstores.faiss import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
import os
import time
from langchain_together import ChatTogether
from langchain.chains.qa_with_sources.retrieval import RetrievalQAWithSourcesChain

# Load environment variables
load_dotenv()

# Set Streamlit app title
st.title("News Research Tool 📈")
st.sidebar.title("News Article URLs")

# Get URLs from user input
urls = []
for i in range(3):
    url = st.sidebar.text_input(f"URL {i+1}")
    urls.append(url)

# Button to process URLs
process_url_clicked = st.sidebar.button("Process URLs")
faiss_index_path = "faiss_index"

# Placeholder for main content
main_placeholder = st.empty()

# Initialize the OpenAI LLM
llm = ChatTogether(model="meta-llama/Llama-3.3-70B-Instruct-Turbo-Free", api_key=os.getenv("OPENAI_API_KEY"))


def save_faiss_index(vectorstore, path):
    # Save FAISS index and documents separately
    vectorstore.save_local(path)

def load_faiss_index(path, embeddings):
    # Load FAISS index and recreate vectorstore
    return FAISS.load_local(path, embeddings, allow_dangerous_deserialization=True)

if process_url_clicked:
    # Load data from URLs
    loader = UnstructuredURLLoader(urls=urls)
    main_placeholder.text("Data Loading...Started...✅✅✅")
    data = loader.load()

    # Split data into chunks
    text_splitter = RecursiveCharacterTextSplitter(
        separators=['\n\n', '\n', '.', ','],
        chunk_size=1000
    )
    main_placeholder.text("Text Splitter...Started...✅✅✅")
    docs = text_splitter.split_documents(data)

    # Create embeddings and save them to FAISS index
    embeddings = HuggingFaceEndpointEmbeddings(huggingfacehub_api_token=os.getenv("HUGGINGFACEHUB_API_TOKEN"))
    vectorstore_openai = FAISS.from_documents(docs, embeddings)
    main_placeholder.text("Embedding Vector Started Building...✅✅✅")
    time.sleep(2)

    # Save the FAISS index to disk
    save_faiss_index(vectorstore_openai, faiss_index_path)

# Get query from user input
query = main_placeholder.text_input("Question: ")
if query:
    if os.path.exists(faiss_index_path):
        embeddings = HuggingFaceEndpointEmbeddings(huggingfacehub_api_token=os.getenv("HUGGINGFACEHUB_API_TOKEN"))
        vectorstore = load_faiss_index(faiss_index_path, embeddings)
        chain = RetrievalQAWithSourcesChain.from_llm(llm=llm, retriever=vectorstore.as_retriever())
        result = chain({"question": query}, return_only_outputs=True)

        # Display the answer
        st.header("Answer")
        st.write(result["answer"])

        # Display sources, if available
        sources = result.get("sources", "")
        if sources:
            st.subheader("Sources:")
            sources_list = sources.split("\n")
            for source in sources_list:
                st.write(source)