app.py

import os
import streamlit as st
from langchain.chains import create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_together import ChatTogether
from langchain_community.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_core.prompts import ChatPromptTemplate

st.set_page_config(page_title="Chat with Website")

st.title("Chat with Website")

# User input for website URL
url_input = st.text_input("Enter a website URL:")

# Initialize variables for documents and retrieval chain
docs = []
documents = []
retriever = None
result = None

# Load documents from the user-provided web URL if provided
if url_input:
    try:
        loader = WebBaseLoader(url_input)
        docs = loader.load()
        
        # Split documents into manageable pieces using recursive character splitter
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=50)  # You can adjust these parameters
        documents = text_splitter.split_documents(docs)
        
        # Create embeddings with Hugging Face and store in FAISS index
        embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
        vector_store = FAISS.from_documents(documents, embeddings)

        # Set up the retriever
        retriever = vector_store.as_retriever()

    except Exception as e:
        st.error(f"Failed to load documents from the specified URL: {str(e)}")
        docs = []
        documents = []

# Set up the Together chat model with environment variable for API key
llm = ChatTogether(
    together_api_key=os.environ.get("TOGETHER_API_KEY"),
    model="meta-llama/Llama-3-70b-chat-hf"
)

# Streamlit interface for user query about the website content
user_query = st.text_area("Ask a question about the website content:", height=200)

# Process the input when the button is clicked
if st.button("Submit"):
    if documents and user_query and retriever:
        # Prepare the context with the retriever
        context = retriever.retrieve(user_query)

        # Create the retrieval chain if documents are available
        prompt_template = """Answer the following question briefly based only on the provided context:
        <context>
        {context}
        </context>
        Question: {input}"""
        
        prompt = ChatPromptTemplate.from_template(prompt_template).render(context=context, input=user_query)

        # Invoke the Together model with context
        result = llm.invoke(prompt)
        
        # Correctly access the 'content' from the 'AIMessage' object
        st.write(result.content)
    else:
        st.warning("Please enter a valid URL and a question to proceed.")