app.py

import streamlit as st 
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.document_loaders import PyPDFLoader
from transformers import T5Tokenizer, T5ForConditionalGeneration
from transformers import pipeline
import base64
from huggingface_hub import login
import torch
import fitz  # PyMuPDF


# model and tokenizer loading
checkpoint = "MBZUAI/LaMini-Flan-T5-248M"
# checkpoint = "google/flan-t5-base"
tokenizer = T5Tokenizer.from_pretrained(checkpoint)
base_model = T5ForConditionalGeneration.from_pretrained(checkpoint, device_map='auto', torch_dtype=torch.float32)

# LLM pipeline
def llm_pipeline(pdf_contents):
    # Extract text from the PDF contents
    pdf_document = fitz.open(stream=pdf_contents, filetype="pdf")
    pdf_text = ""
    for page_num in range(pdf_document.page_count):
        page = pdf_document.load_page(page_num)
        pdf_text += page.get_text()

    # Use the pipeline to generate the summary
    pipe_sum = pipeline(
        'summarization',
        model=base_model,
        tokenizer=tokenizer,
        max_length=500,
        min_length=50
    )

    result = pipe_sum(pdf_text)
    summary = result[0]['summary_text']
    return summary

# Streamlit code
st.set_page_config(layout="wide")

def main():
    st.title("Document Summarization App using Language Model")

    uploaded_file = st.file_uploader("Upload your PDF file", type=['pdf'])

    if uploaded_file is not None:
        if st.button("Summarize"):
            summary = llm_pipeline(uploaded_file.read())

            # Display the summary
            st.info("Summarization Complete")
            st.success(summary)

if __name__ == "__main__":
    main()