app.py

import os
import streamlit as st
import pdfplumber
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from transformers import pipeline

# Set up the page configuration
st.set_page_config(page_title="RAG-based PDF Chat", layout="centered", page_icon="📄")

# Load the summarization pipeline model
@st.cache_resource
def load_summarization_pipeline():
    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
    return summarizer

summarizer = load_summarization_pipeline()

# Function to preprocess PDFs and store embeddings
def preprocess_pdfs(folder_path, save_vectorstore_path):
    all_text = ""
    pdf_files = [os.path.join(folder_path, filename) for filename in os.listdir(folder_path) if filename.endswith('.pdf')]

    for file_path in pdf_files:
        with pdfplumber.open(file_path) as pdf:
            for page in pdf.pages:
                page_text = page.extract_text()
                if page_text:
                    all_text += page_text

    if all_text:
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=10000, chunk_overlap=1000)
        text_chunks = text_splitter.split_text(all_text)
        embedding_function = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
        vector_store = FAISS.from_texts(text_chunks, embedding=embedding_function)
        vector_store.save_local(save_vectorstore_path)
        st.success("Data preprocessing and vector store creation completed!")

# Load pre-trained FAISS vector store
@st.cache_resource
def load_vector_store(save_vectorstore_path):
    embedding_function = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
    return FAISS.load_local(save_vectorstore_path, embedding_function, allow_dangerous_deserialization=True)



# Generate summary based on the retrieved text
def generate_summary_with_huggingface(query, retrieved_text):
    summarization_input = f"{query} Related information:{retrieved_text}"
    max_input_length = 1024
    summarization_input = summarization_input[:max_input_length]
    summary = summarizer(summarization_input, max_length=500, min_length=50, do_sample=False)
    return summary[0]["summary_text"]

# Generate response for user query
def user_input(user_question, vector_store):
    docs = vector_store.similarity_search(user_question)
    context_text = " ".join([doc.page_content for doc in docs])
    return generate_summary_with_huggingface(user_question, context_text)

# Main function to run the Streamlit app
def main():
    st.title("📄 Gen AI Lawyers Guide")
    data_folder = 'documents1'  # Folder where your PDFs are located
    vectorstore_path = 'vector_store_data/faiss_vectorstore'  # Folder to save the vector store

    # Uncomment this line for initial preprocessing only. Once done, comment it out.
    # preprocess_pdfs(data_folder, vectorstore_path)

    # Load the pre-trained vector store
    vector_store = load_vector_store(vectorstore_path)

    user_question = st.text_input("Ask a Question:", placeholder="Type your question here...")

    if st.button("Get Response"):
        if not user_question:
            st.warning("Please enter a question before submitting.")
        else:
            with st.spinner("Generating response..."):
                answer = user_input(user_question, vector_store)
                st.markdown(f"**🤖 AI:** {answer}")

if __name__ == "__main__":
    main()