app.py

import streamlit as st
import torch
from transformers import BartForConditionalGeneration, BartTokenizer
import PyPDF2
import requests
from io import StringIO, BytesIO
import nltk
from nltk.tokenize import sent_tokenize
import spacy
import numpy as np
from typing import List, Tuple, Optional
import time
import re

# Download NLTK data
try:
    nltk.data.find('tokenizers/punkt')
except LookupError:
    nltk.download('punkt')

# Cache the BART model and tokenizer
@st.cache_resource
def load_bart_model(model_name: str = "facebook/bart-large-cnn"):
    """Load BART model and tokenizer"""
    try:
        device = "cuda" if torch.cuda.is_available() else "cpu"
        st.info(f"Using device: {device}")
        
        tokenizer = BartTokenizer.from_pretrained(model_name)
        model = BartForConditionalGeneration.from_pretrained(model_name)
        model = model.to(device)
        
        return model, tokenizer, device
    except Exception as e:
        st.error(f"Error loading model: {str(e)}")
        return None, None, None

# Text preprocessing functions
def clean_text(text: str) -> str:
    """Clean and preprocess text"""
    # Remove extra whitespace
    text = re.sub(r'\s+', ' ', text)
    # Remove special characters but keep basic punctuation
    text = re.sub(r'[^\w\s.,!?;:]', ' ', text)
    return text.strip()

def split_into_chunks(text: str, max_chunk_length: int = 1024) -> List[str]:
    """Split text into chunks for processing"""
    sentences = sent_tokenize(text)
    chunks = []
    current_chunk = ""
    
    for sentence in sentences:
        if len(current_chunk) + len(sentence) < max_chunk_length:
            current_chunk += " " + sentence
        else:
            if current_chunk:
                chunks.append(current_chunk.strip())
            current_chunk = sentence
    
    if current_chunk:
        chunks.append(current_chunk.strip())
    
    return chunks

# Text extraction functions
def extract_text_from_pdf(pdf_file) -> str:
    """Extract text from PDF file"""
    text = ""
    try:
        pdf_reader = PyPDF2.PdfReader(pdf_file)
        for page in pdf_reader.pages:
            text += page.extract_text() + "\n"
    except Exception as e:
        st.error(f"Error reading PDF: {str(e)}")
    return clean_text(text)

def extract_text_from_url(url: str) -> str:
    """Extract text from Wikipedia or other web pages"""
    try:
        headers = {
            'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
        }
        response = requests.get(url, headers=headers, timeout=10)
        
        if response.status_code == 200:
            # Simple HTML stripping
            text = re.sub(r'<[^>]+>', ' ', response.text)
            text = re.sub(r'\s+', ' ', text)
            return clean_text(text)
        else:
            st.error(f"Failed to fetch URL: Status {response.status_code}")
            return ""
    except Exception as e:
        st.error(f"Error fetching URL: {str(e)}")
        return ""

# Summarization functions
def summarize_with_bart(
    text: str, 
    model, 
    tokenizer, 
    device: str,
    max_length: int = 150,
    min_length: int = 40,
    do_sample: bool = False
) -> str:
    """Summarize text using BART model"""
    if not text or len(text.split()) < 10:
        return text  # Return original if too short
    
    try:
        # Split text into chunks if too long
        chunks = split_into_chunks(text, max_chunk_length=1000)
        summaries = []
        
        for chunk in chunks:
            inputs = tokenizer(
                chunk, 
                max_length=1024, 
                truncation=True, 
                return_tensors="pt"
            ).to(device)
            
            # Generate summary
            summary_ids = model.generate(
                inputs["input_ids"],
                max_length=max_length,
                min_length=min_length,
                length_penalty=2.0,
                num_beams=4,
                early_stopping=True,
                do_sample=do_sample
            )
            
            summary = tokenizer.decode(
                summary_ids[0], 
                skip_special_tokens=True
            )
            summaries.append(summary)
        
        # Combine chunk summaries
        combined_summary = " ".join(summaries)
        
        # If combined summary is still too long, summarize it again
        if len(combined_summary.split()) > 200:
            inputs = tokenizer(
                combined_summary, 
                max_length=1024, 
                truncation=True, 
                return_tensors="pt"
            ).to(device)
            
            final_summary_ids = model.generate(
                inputs["input_ids"],
                max_length=max_length,
                min_length=min_length,
                length_penalty=2.0,
                num_beams=4,
                early_stopping=True
            )
            
            final_summary = tokenizer.decode(
                final_summary_ids[0], 
                skip_special_tokens=True
            )
            return final_summary
        
        return combined_summary
        
    except Exception as e:
        st.error(f"Error during summarization: {str(e)}")
        return ""

# Streamlit UI
def main():
    st.set_page_config(
        page_title="BART Text Summarizer",
        page_icon="🤖",
        layout="wide",
        initial_sidebar_state="expanded"
    )
    
    # Custom CSS
    st.markdown("""
    <style>
    .main-header {
        font-size: 2.5rem;
        color: #4A90E2;
        text-align: center;
        margin-bottom: 1rem;
    }
    .sub-header {
        font-size: 1.2rem;
        color: #666;
        text-align: center;
        margin-bottom: 2rem;
    }
    .stats-card {
        background-color: #f0f2f6;
        padding: 1rem;
        border-radius: 10px;
        margin: 0.5rem 0;
    }
    .summary-box {
        border: 2px solid #4A90E2;
        border-radius: 10px;
        padding: 1rem;
        background-color: #f8f9fa;
    }
    </style>
    """, unsafe_allow_html=True)
    
    # Header
    st.markdown('<h1 class="main-header">🤖 BART Text Summarizer</h1>', unsafe_allow_html=True)
    st.markdown('<p class="sub-header">Powered by Facebook\'s BART-large-CNN model from Hugging Face</p>', unsafe_allow_html=True)
    st.markdown("---")
    
    # Sidebar
    with st.sidebar:
        st.header("⚙️ Configuration")
        
        # Model selection
        model_option = st.selectbox(
            "Choose BART model:",
            [
                "facebook/bart-large-cnn", 
                "facebook/bart-large-xsum",
                "sshleifer/distilbart-cnn-12-6"
            ],
            help="BART-large-cnn is best for general summarization"
        )
        
        # Summary length
        st.subheader("Summary Settings")
        max_length = st.slider(
            "Maximum summary length (words)",
            min_value=50,
            max_value=500,
            value=150,
            step=10
        )
        
        min_length = st.slider(
            "Minimum summary length (words)",
            min_value=10,
            max_value=100,
            value=40,
            step=5
        )
        
        # Advanced options
        with st.expander("Advanced Options"):
            do_sample = st.checkbox(
                "Use sampling (more creative)", 
                value=False,
                help="When enabled, uses sampling instead of beam search"
            )
            
            num_beams = st.slider(
                "Number of beams",
                min_value=1,
                max_value=8,
                value=4,
                help="Higher values produce better results but are slower"
            )
        
        st.markdown("---")
        
        # Model info
        st.info("""
        **Model Information:**
        - BART-large-CNN: Fine-tuned on CNN/Daily Mail
        - Parameters: 400 million
        - Best for: Article summarization
        """)
        
        # Load model button
        if st.button("🔄 Load Model", type="secondary"):
            with st.spinner("Loading BART model..."):
                model, tokenizer, device = load_bart_model(model_option)
                if model:
                    st.success(f"Model loaded successfully on {device}!")
    
    # Main content
    col1, col2 = st.columns([1, 1])
    
    with col1:
        st.subheader("📥 Input Text")
        
        # Input method selection
        input_method = st.radio(
            "Choose input method:",
            ["📝 Direct Text", "📄 Upload File", "🌐 Website URL"],
            horizontal=True
        )
        
        text_input = ""
        
        if input_method == "📝 Direct Text":
            text_input = st.text_area(
                "Enter your text here:",
                height=300,
                placeholder="Paste or type your text here...",
                help="Minimum 100 words for best results"
            )
            
        elif input_method == "📄 Upload File":
            uploaded_file = st.file_uploader(
                "Upload a file",
                type=['txt', 'pdf', 'docx'],
                help="Supports TXT, PDF, and DOCX files"
            )
            
            if uploaded_file:
                file_ext = uploaded_file.name.split('.')[-1].lower()
                
                if file_ext == 'pdf':
                    with st.spinner("Extracting text from PDF..."):
                        text_input = extract_text_from_pdf(uploaded_file)
                elif file_ext == 'txt':
                    stringio = StringIO(uploaded_file.getvalue().decode("utf-8"))
                    text_input = stringio.read()
                else:
                    st.warning("Please upload a PDF or TXT file")
        
        elif input_method == "🌐 Website URL":
            url = st.text_input(
                "Enter URL:",
                placeholder="https://en.wikipedia.org/wiki/...",
                help="Supports Wikipedia and other websites"
            )
            
            if url and st.button("Fetch Content", type="secondary"):
                with st.spinner("Fetching content from URL..."):
                    text_input = extract_text_from_url(url)
        
        # Display text stats
        if text_input:
            words = text_input.split()
            sentences = sent_tokenize(text_input)
            
            with st.expander("📊 Text Statistics", expanded=True):
                col_a, col_b, col_c = st.columns(3)
                with col_a:
                    st.metric("Words", len(words))
                with col_b:
                    st.metric("Sentences", len(sentences))
                with col_c:
                    st.metric("Characters", len(text_input))
            
            # Preview
            with st.expander("🔍 Preview Original Text"):
                preview_length = min(500, len(text_input))
                st.text(text_input[:preview_length] + "..." if len(text_input) > preview_length else text_input)
    
    with col2:
        st.subheader("📤 Generated Summary")
        
        if text_input and len(text_input.split()) >= 10:
            if st.button("🚀 Generate Summary", type="primary", use_container_width=True):
                with st.spinner("Generating summary with BART..."):
                    # Load model if not already loaded
                    model, tokenizer, device = load_bart_model(model_option)
                    
                    if model and tokenizer:
                        # Generate summary
                        start_time = time.time()
                        summary = summarize_with_bart(
                            text_input,
                            model,
                            tokenizer,
                            device,
                            max_length=max_length,
                            min_length=min_length,
                            do_sample=do_sample
                        )
                        processing_time = time.time() - start_time
                        
                        if summary:
                            # Display summary
                            st.markdown('<div class="summary-box">', unsafe_allow_html=True)
                            st.write(summary)
                            st.markdown('</div>', unsafe_allow_html=True)
                            
                            # Summary stats
                            st.markdown("### 📈 Summary Statistics")
                            col1_stat, col2_stat, col3_stat, col4_stat = st.columns(4)
                            
                            with col1_stat:
                                st.metric(
                                    "Summary Words", 
                                    len(summary.split()),
                                    delta=f"-{len(text_input.split()) - len(summary.split())}"
                                )
                            
                            with col2_stat:
                                reduction = ((len(text_input.split()) - len(summary.split())) / len(text_input.split()) * 100)
                                st.metric(
                                    "Reduction", 
                                    f"{reduction:.1f}%"
                                )
                            
                            with col3_stat:
                                st.metric(
                                    "Processing Time", 
                                    f"{processing_time:.2f}s"
                                )
                            
                            with col4_stat:
                                st.metric(
                                    "Compression Ratio", 
                                    f"1:{len(text_input.split())//len(summary.split()) if summary.split() else 0}"
                                )
                            
                            # Download button
                            st.download_button(
                                label="📥 Download Summary",
                                data=summary,
                                file_name="bart_summary.txt",
                                mime="text/plain",
                                use_container_width=True
                            )
                            
                            # Show sample comparison
                            with st.expander("🔍 Compare Original vs Summary"):
                                col_orig, col_sum = st.columns(2)
                                with col_orig:
                                    st.write("**Original (first 200 words):**")
                                    st.write(" ".join(text_input.split()[:200]) + "...")
                                with col_sum:
                                    st.write("**Summary:**")
                                    st.write(summary)
                        else:
                            st.error("Failed to generate summary. Please try again.")
                    else:
                        st.error("Failed to load model. Please check your internet connection.")
        elif text_input and len(text_input.split()) < 10:
            st.warning("Please enter at least 10 words for summarization")
        else:
            st.info("👈 Enter text on the left to generate a summary")
            
            # Example text
            with st.expander("📚 Try with Example Text"):
                example_text = """
                Artificial Intelligence (AI) is transforming industries across the globe. 
                From healthcare to finance, AI algorithms are being deployed to solve complex problems, 
                automate processes, and generate insights from massive datasets. Machine learning, 
                a subset of AI, enables computers to learn from data without being explicitly programmed. 
                Deep learning, powered by neural networks, has achieved remarkable success in areas like 
                image recognition, natural language processing, and autonomous vehicles. 
                However, AI also raises important ethical considerations around bias, privacy, 
                and job displacement. As AI continues to evolve, it's crucial to develop responsible 
                AI frameworks that ensure these technologies benefit society while mitigating potential risks. 
                The future of AI holds tremendous promise, but requires careful stewardship and collaboration 
                between technologists, policymakers, and the public.
                """
                if st.button("Load Example Text"):
                    st.session_state.example_loaded = example_text
                    st.rerun()
    
    # Footer
    st.markdown("---")
    
    col_footer1, col_footer2, col_footer3 = st.columns(3)
    
    with col_footer1:
        st.markdown("**Powered by:**")
        st.markdown("[![Hugging Face](https://huggingface.co/front/assets/huggingface_logo-noborder.svg)](https://huggingface.co)")
        
    with col_footer2:
        st.markdown("**Model:**")
        st.markdown("[BART-large-CNN](https://huggingface.co/facebook/bart-large-cnn)")
        
    with col_footer3:
        st.markdown("**Built with:**")
        st.markdown("[Streamlit](https://streamlit.io) | [PyTorch](https://pytorch.org)")
    
    st.caption("© 2024 BART Summarizer | Deploy your own on [Hugging Face Spaces](https://huggingface.co/spaces)")

if __name__ == "__main__":
    main()