app.py

import streamlit as st
from langchain.document_loaders import PyPDFDirectoryLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import Chroma
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain import hub
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_google_genai import ChatGoogleGenerativeAI
import os

# Set up the directories for data and vector DB
DATA_DIR = "/content/MyData"
DB_DIR = "/content/VectorDB"

# Create directories if they don't exist
os.makedirs(DATA_DIR, exist_ok=True)
os.makedirs(DB_DIR, exist_ok=True)

# Initialize the embeddings model
embeddings_model = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")

# Load and process PDF documents
def load_data():
    loader = PyPDFDirectoryLoader(DATA_DIR)
    data_on_pdf = loader.load()
    text_splitter = RecursiveCharacterTextSplitter(
        separators=["\n\n", "\n", ". ", " ", ""],
        chunk_size=1000,
        chunk_overlap=200
    )
    splits = text_splitter.split_documents(data_on_pdf)
    vectorstore = Chroma.from_documents(documents=splits, embedding=embeddings_model, persist_directory=DB_DIR)
    return vectorstore

# Set up the generative AI model
llm = ChatGoogleGenerativeAI(model="gemini-1.5-pro", google_api_key="YOUR_GOOGLE_API_KEY")

# Load vector store
vectorstore = load_data()

# Streamlit interface
st.title("RAG App: Question-Answering with PDFs")

# File uploader for PDF documents
uploaded_files = st.file_uploader("Upload PDF files", accept_multiple_files=True, type=["pdf"])

if uploaded_files:
    for uploaded_file in uploaded_files:
        with open(os.path.join(DATA_DIR, uploaded_file.name), "wb") as f:
            f.write(uploaded_file.getbuffer())
    st.success("PDF files uploaded successfully!")

    # Reload vector store after uploading new files
    vectorstore = load_data()

# User input for question
question = st.text_input("Ask a question about the documents:")

if st.button("Submit"):
    if question:
        retriever = vectorstore.as_retriever()
        prompt = hub.pull("rlm/rag-prompt")
        
        def format_docs(docs):
            return "\n\n".join(doc.page_content for doc in docs)

        rag_chain = (
            {"context": retriever | format_docs, "question": RunnablePassthrough()}
            | prompt
            | llm
            | StrOutputParser()
        )

        response = rag_chain.invoke(question)
        st.markdown(response)
    else:
        st.warning("Please enter a question.")