app.py

import gradio as gr
import requests
import fitz  # PyMuPDF
from fpdf import FPDF
import os
import tempfile
import math

# --- CONFIGURATION ---
MODEL_ID = "sshleifer/distilbart-cnn-12-6"
API_URL = f"https://api-inference.huggingface.co/models/{MODEL_ID}"
HF_API_TOKEN = os.environ.get("HF_API_TOKEN")

# --- HELPER FUNCTIONS ---

def extract_text_from_pdf(pdf_file):
    """Extract text from uploaded PDF file."""
    try:
        doc = fitz.open(stream=pdf_file.read(), filetype="pdf")
        full_text = ""
        for page in doc:
            full_text += page.get_text()
        doc.close()
        return full_text.strip()
    except Exception as e:
        raise gr.Error(f"Failed to read PDF. Is it valid? Error: {e}")

def chunk_text(text, max_tokens=1000):
    """Split text into chunks of approximately max_tokens words."""
    words = text.split()
    for i in range(0, len(words), max_tokens):
        yield " ".join(words[i:i+max_tokens])

def summarize_text(text_to_summarize):
    """Send text to Hugging Face API for summarization, chunking if too long."""
    if not HF_API_TOKEN:
        raise gr.Error("Hugging Face API token is not set. Add it as an environment variable.")

    headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
    final_summary = []

    for chunk in chunk_text(text_to_summarize, max_tokens=500):
        payload = {
            "inputs": chunk,
            "parameters": {"min_length": 50, "max_length": 250, "do_sample": False}
        }
        response = requests.post(API_URL, headers=headers, json=payload)
        if response.status_code == 200:
            final_summary.append(response.json()[0]["summary_text"])
        else:
            error_details = response.json().get('error', response.text)
            raise gr.Error(f"Model API Error: {error_details}")
    
    return " ".join(final_summary)

def save_text_to_pdf(text):
    """Save summarized text to a new PDF and return its path."""
    pdf = FPDF()
    pdf.add_page()
    pdf.set_font("Arial", "B", 16)
    pdf.cell(0, 10, "Tweaked Document Summary", ln=1, align='C')
    pdf.ln(10)

    pdf.set_font("Arial", size=12)
    cleaned_text = text.encode('latin-1', 'replace').decode('latin-1')  # FPDF limitation
    pdf.multi_cell(0, 10, cleaned_text)

    tmp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf")
    pdf.output(tmp_file.name)
    tmp_file.close()
    return tmp_file.name

# --- MAIN WORKFLOW ---

def tweak_pdf_workflow(uploaded_pdf):
    if uploaded_pdf is None:
        raise gr.Error("Please upload a PDF first.")
    
    # Step 1: Extract
    original_text = extract_text_from_pdf(uploaded_pdf)
    if not original_text.strip():
        raise gr.Error("PDF contains no extractable text.")

    # Step 2: Summarize
    tweaked_text = summarize_text(original_text)

    # Step 3: Save
    output_pdf_path = save_text_to_pdf(tweaked_text)
    return output_pdf_path

# --- GRADIO INTERFACE ---

with gr.Blocks(theme=gr.themes.Soft()) as iface:
    gr.Markdown(
        """
        # 📄 PDF Document Tweaker (TL;DR)
        Upload a PDF and get a summarized, tweaked PDF using Hugging Face's `distilbart-cnn-12-6`.
        """
    )
    with gr.Row():
        pdf_input = gr.File(label="Upload Your PDF", file_types=[".pdf"])
        pdf_output = gr.File(label="Download Tweaked PDF")
    submit_button = gr.Button("Tweak My Document!", variant="primary")
    submit_button.click(fn=tweak_pdf_workflow, inputs=pdf_input, outputs=pdf_output)

if __name__ == "__main__":
    iface.launch(server_name="0.0.0.0", server_port=7860)