app.py

import os
import streamlit as st
from transformers import pipeline
import torch
from PyPDF2 import PdfReader

# Disable tokenizers parallelism
os.environ["TOKENIZERS_PARALLELISM"] = "false"

# Setup for the model
device = 0 if torch.cuda.is_available() else -1
summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6", device=device)

def split_text(text, max_chunk_size=512):
    words = text.split()
    for i in range(0, len(words), max_chunk_size):
        yield " ".join(words[i:i + max_chunk_size])

def extract_text_from_pdf(pdf_file):
    reader = PdfReader(pdf_file)
    text = ""
    for page_num in range(len(reader.pages)):
        page = reader.pages[page_num]
        text += page.extract_text()
    return text

def summarize_text(text, summarizer):
    chunks = list(split_text(text))
    summaries = []
    for chunk in chunks:
        input_length = len(chunk.split())
        max_summary_length = max(10, int(input_length * 0.6))
        min_summary_length = max(5, int(input_length * 0.2))
        result = summarizer(chunk, max_length=max_summary_length, min_length=min_summary_length, do_sample=False)
        summaries.append(result[0]['summary_text'])
    return " ".join(summaries)

def extract_and_summarize_page_by_page(pdf_file, summarizer):
    reader = PdfReader(pdf_file)
    summaries = []
    for page_num in range(len(reader.pages)):
        page = reader.pages[page_num]
        text = page.extract_text()
        if text:
            page_summary = summarize_text(text, summarizer)
            summaries.append(page_summary)
        else:
            summaries.append(f"Page {page_num + 1}: No extractable text found.")
    return summaries

# Streamlit interface
st.subheader("Generate PDF Summary")
pdf_file = st.file_uploader("Upload a PDF", type=["pdf"])

if pdf_file:
    text = extract_text_from_pdf(pdf_file)
    if len(text) > 0:
        summaries = extract_and_summarize_page_by_page(pdf_file, summarizer)
        st.subheader("Summary")
        for i, summary in enumerate(summaries, 1):
            st.write(f"### Page {i}\n{summary}\n")
    else:
        st.warning("No extractable text found in the PDF.")