app.py

import os
import streamlit as st
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.vectorstores import Chroma
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chat_models import ChatOpenAI
from langchain.chains import ConversationalRetrievalChain, ConversationChain
from langchain.memory import ConversationBufferMemory
from langchain.document_loaders import PyPDFLoader
import time

# Initialize session state variables
if "messages" not in st.session_state:
    st.session_state.messages = []
if "chain" not in st.session_state:
    st.session_state.chain = None
if "processed_pdfs" not in st.session_state:
    st.session_state.processed_pdfs = False
if "waiting_for_answer" not in st.session_state:
    st.session_state.waiting_for_answer = False

def create_sidebar():
    with st.sidebar:
        st.title("PDF Chat")
        st.markdown("### Quick Demo of RAG")
        
        api_key = st.text_input("OpenAI API Key:", type="password")
        
        st.markdown("""
        ### Tools Used
        - OpenAI
        - LangChain
        - ChromaDB
        
        ### Steps
        1. Add API key
        2. Upload PDF
        3. Chat!
        """)
        
        return api_key

def save_uploaded_file(uploaded_file, path='./uploads/'):
    os.makedirs(path, exist_ok=True)
    file_path = os.path.join(path, uploaded_file.name)
    with open(file_path, "wb") as f:
        f.write(uploaded_file.getbuffer())
    return file_path

def load_texts_from_papers(papers):
    all_texts = []
    for paper in papers:
        try:
            file_path = save_uploaded_file(paper)
            loader = PyPDFLoader(file_path)
            documents = loader.load()
            text_splitter = RecursiveCharacterTextSplitter(
                chunk_size=1000,
                chunk_overlap=200,
                length_function=len,
                is_separator_regex=False,
            )
            texts = text_splitter.split_documents(documents)
            all_texts.extend(texts)
            os.remove(file_path)
        except Exception as e:
            st.error(f"Error processing {paper.name}: {str(e)}")
    return all_texts

def initialize_vectorstore(api_key):
    embedding = OpenAIEmbeddings(openai_api_key=api_key)
    vectorstore = Chroma(embedding_function=embedding, persist_directory="db")
    return vectorstore

def process_pdfs(papers, api_key):
    if papers and not st.session_state.processed_pdfs:
        with st.spinner("Processing PDFs..."):
            texts = load_texts_from_papers(papers)
            if texts:
                vectorstore = initialize_vectorstore(api_key)
                vectorstore.add_documents(texts)
                st.session_state.chain = ConversationalRetrievalChain.from_llm(
                    ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo", openai_api_key=api_key),
                    vectorstore.as_retriever(),
                    memory=ConversationBufferMemory(
                        memory_key="chat_history",
                        return_messages=True
                    )
                )
                st.session_state.processed_pdfs = True
                st.success("PDFs processed successfully!")
            return texts
    return []

def get_assistant_response(prompt, texts):
    try:
        if texts or st.session_state.processed_pdfs:
            result = st.session_state.chain({"question": prompt})
            return result["answer"]
        else:
            return "Please upload a PDF first."
    except Exception as e:
        return f"Error: {str(e)}"

def main():
    st.set_page_config(page_title="PDF Chat", layout="wide")
    
    api_key = create_sidebar()
    
    if not api_key:
        st.warning("Please enter your OpenAI API key")
        return

    st.title("Chat with PDF")
    
    # File uploader
    papers = st.file_uploader("Upload PDFs", type=["pdf"], accept_multiple_files=True)
    
    # Process PDFs
    texts = process_pdfs(papers, api_key)
    
    # Chat interface
    chat_container = st.container()
    
    with chat_container:
        # Display existing chat messages
        for message in st.session_state.messages:
            with st.chat_message(message["role"]):
                st.markdown(message["content"])
        
        # Get user input
        if prompt := st.chat_input("Ask about your PDFs"):
            # Add user message immediately
            st.session_state.messages.append({"role": "user", "content": prompt})
            st.chat_message("user").markdown(prompt)
            
            # Get assistant response with a loading indicator
            with st.chat_message("assistant"):
                with st.spinner("Thinking..."):
                    response = get_assistant_response(prompt, texts)
                st.markdown(response)
            
            # Add assistant response to messages
            st.session_state.messages.append({"role": "assistant", "content": response})

if __name__ == "__main__":
    main()