app.py

import os
import streamlit as st
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_chroma import Chroma
from langchain_community.chat_message_histories import ChatMessageHistory
from langchain_core.chat_history import BaseChatMessageHistory
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_groq import ChatGroq
from langchain_core.runnables.history import RunnableWithMessageHistory
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import PyPDFLoader, WebBaseLoader
from chromadb.config import Settings

# Load secrets from environment (set in Hugging Face Space Settings > Secrets)
groq_api_key = os.environ.get("GROQ_API_KEY")
hf_token = os.environ.get("HUGGINGFACE_API_KEY")

# Initialize LLM and embeddings
embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
llm = ChatGroq(groq_api_key=groq_api_key, model_name="llama3-70b-8192")

# Streamlit UI
st.title("Conversational RAG | PDF + Website")
st.write("Upload PDFs or enter a website URL to chat with their content.")

session_id = st.text_input("Session ID:", value="default_session")

# Initialize session state
if 'store' not in st.session_state:
    st.session_state.store = {}
if 'vectorstore' not in st.session_state:
    st.session_state.vectorstore = None
if 'document_source' not in st.session_state:
    st.session_state.document_source = ""
if 'documents' not in st.session_state:
    st.session_state.documents = []

# Input Mode Selector
input_mode = st.radio("Select Input Type:", ("PDF Upload", "Website URL"))

# --- Load PDF ---
if input_mode == "PDF Upload":
    uploaded_files = st.file_uploader("Upload PDF(s)", type="pdf", accept_multiple_files=True)
    if uploaded_files:
        for uploaded_file in uploaded_files:
            with open("temp.pdf", "wb") as f:
                f.write(uploaded_file.getvalue())
            loader = PyPDFLoader("temp.pdf")
            st.session_state.documents.extend(loader.load())
        st.session_state.document_source = "PDF"

# --- Load Website ---
elif input_mode == "Website URL":
    url = st.text_input("Enter Website URL")
    if st.button("Load Website") and url:
        loader = WebBaseLoader(web_paths=[url])
        st.session_state.documents = loader.load()
        st.session_state.document_source = url

# --- Build Vector Store and Chain ---
if st.session_state.documents:
    splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    splits = splitter.split_documents(st.session_state.documents)

    vectorstore = Chroma.from_documents(
        documents=splits,
        embedding=embeddings,
        collection_name="rag_collection",
        persist_directory=None,
        client_settings=Settings(anonymized_telemetry=False)
    )
    st.session_state.vectorstore = vectorstore
    retriever = vectorstore.as_retriever()

    # Prompts
    contextualize_prompt = ChatPromptTemplate.from_messages([
        ("system", "Given a chat history and the latest user question which might reference context in the chat history, formulate a standalone question."),
        MessagesPlaceholder("chat_history"),
        ("human", "{input}")
    ])

    qa_prompt = ChatPromptTemplate.from_messages([
        ("system", "Use the following context to answer the question in 3 sentences or less.{context}"),
        MessagesPlaceholder("chat_history"),
        ("human", "{input}")
    ])

    # Chains
    history_aware_retriever = create_history_aware_retriever(llm, retriever, contextualize_prompt)
    qa_chain = create_stuff_documents_chain(llm, qa_prompt)
    rag_chain = create_retrieval_chain(history_aware_retriever, qa_chain)

    def get_session_history(session_id: str) -> BaseChatMessageHistory:
        if session_id not in st.session_state.store:
            st.session_state.store[session_id] = ChatMessageHistory()
        return st.session_state.store[session_id]

    conversational_rag_chain = RunnableWithMessageHistory(
        rag_chain,
        get_session_history,
        input_messages_key="input",
        history_messages_key="chat_history",
        output_messages_key="answer"
    )

    st.divider()
    st.write(f"You are chatting with: **{st.session_state.document_source}**")

    user_input = st.text_input("Your question:")
    if user_input:
        session_history = get_session_history(session_id)
        response = conversational_rag_chain.invoke(
            {"input": user_input},
            config={"configurable": {"session_id": session_id}}
        )
        st.write("**Assistant:**", response['answer'])

        with st.expander("📜 Chat History"):
            for msg in session_history.messages:
                role = "You" if msg.type == "human" else "Assistant"
                st.markdown(f"**{role}:** {msg.content}")

elif not groq_api_key:
    st.warning("❗ Please set your GROQ_API_KEY in the Space secrets.")