app.py

import gradio as gr
from transformers import pipeline
import docx2txt
import PyPDF2
from docx import Document
from fpdf import FPDF
import os
from io import BytesIO

# Load models
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
translator_hi_en = pipeline("translation", model="Helsinki-NLP/opus-mt-hi-en")
translator_mr_en = pipeline("translation", model="Helsinki-NLP/opus-mt-mr-en")
translator_en_hi = pipeline("translation", model="Helsinki-NLP/opus-mt-en-hi")
translator_en_mr = pipeline("translation", model="Helsinki-NLP/opus-mt-en-mr")

# Extract text based on file type
def extract_text(file):
    ext = file.name.split(".")[-1].lower()
    if ext == "txt":
        return file.read().decode("utf-8")
    elif ext == "pdf":
        reader = PyPDF2.PdfReader(file)
        return "\n".join(page.extract_text() for page in reader.pages if page.extract_text())
    elif ext == "docx":
        return docx2txt.process(file)
    else:
        return "Unsupported file type. Please upload a .pdf, .docx, or .txt file."

# Chunk long text for translation and summarization
def chunk_text(text, max_length=1000):
    paragraphs = text.split("\n")
    chunks = []
    current_chunk = ""
    for para in paragraphs:
        if len(current_chunk) + len(para) < max_length:
            current_chunk += para + "\n"
        else:
            chunks.append(current_chunk.strip())
            current_chunk = para + "\n"
    if current_chunk:
        chunks.append(current_chunk.strip())
    return chunks

# Translate to English from selected language
def translate_to_english(text, lang):
    if lang == "Hindi":
        return " ".join([translator_hi_en(chunk)[0]['translation_text'] for chunk in chunk_text(text, 500)])
    elif lang == "Marathi":
        return " ".join([translator_mr_en(chunk)[0]['translation_text'] for chunk in chunk_text(text, 500)])
    return text

# Translate from English to selected output language
def translate_from_english(text, lang):
    if lang == "Hindi":
        return " ".join([translator_en_hi(chunk)[0]['translation_text'] for chunk in chunk_text(text, 500)])
    elif lang == "Marathi":
        return " ".join([translator_en_mr(chunk)[0]['translation_text'] for chunk in chunk_text(text, 500)])
    return text

# Save summary to DOCX
def generate_docx(text):
    doc = Document()
    doc.add_heading("Summary", 0)
    doc.add_paragraph(text)
    buffer = BytesIO()
    doc.save(buffer)
    buffer.seek(0)
    return buffer

# Save summary to PDF
def generate_pdf(text):
    pdf = FPDF()
    pdf.add_page()
    pdf.set_font("Arial", size=12)
    for line in text.split("\n"):
        pdf.multi_cell(0, 10, line)
    buffer = BytesIO()
    pdf.output(buffer)
    buffer.seek(0)
    return buffer

# Main summarization function
def summarize_input(text, file, length, input_lang, output_lang):
    source_text = text.strip() if text.strip() else extract_text(file)
    if not source_text:
        return "", None, None

    # Translate to English if needed
    if input_lang != "English":
        source_text = translate_to_english(source_text, input_lang)

    # Set summary length
    if length == "Short (1–2 sentences)":
        min_len, max_len = 20, 60
    elif length == "Detailed (paragraph)":
        min_len, max_len = 80, 200
    else:
        min_len, max_len = 40, 130

    chunks = chunk_text(source_text)
    summaries = [summarizer(chunk, max_length=max_len, min_length=min_len, do_sample=False)[0]['summary_text'] for chunk in chunks]
    final_summary = "\n\n".join(summaries)

    # Translate from English to output language
    if output_lang != "English":
        final_summary = translate_from_english(final_summary, output_lang)

    docx_file = generate_docx(final_summary)
    pdf_file = generate_pdf(final_summary)
    return final_summary, ("summary.docx", docx_file), ("summary.pdf", pdf_file)

# Gradio interface
iface = gr.Interface(
    fn=summarize_input,
    inputs=[
        gr.Textbox(lines=8, label="Enter text (optional)"),
        gr.File(label="Upload file (.txt, .pdf, .docx)", file_types=[".pdf", ".docx", ".txt"]),
        gr.Radio([
            "Short (1–2 sentences)",
            "Medium (3–5 sentences)",
            "Detailed (paragraph)"
        ], label="Summary length", value="Medium (3–5 sentences)"),
        gr.Dropdown(["English", "Hindi", "Marathi"], label="Document Language", value="English"),
        gr.Dropdown(["English", "Hindi", "Marathi"], label="Summary Output Language", value="English")
    ],
    outputs=[
        gr.Textbox(label="Summary"),
        gr.File(label="Download as DOCX"),
        gr.File(label="Download as PDF")
    ],
    title="🌍 Multilingual Document Summarizer",
    description="Upload or paste a document in English, Hindi, or Marathi. App will translate if needed and summarize it into your chosen output language."
)

iface.launch()