app.py

import gradio as gr
import fitz
import torch
from transformers import pipeline
import time, logging, re
import matplotlib.pyplot as plt
import io

logging.basicConfig(level=logging.ERROR)
device = -1  # CPU-only
print("⚠️ CPU-only. Expect ~20–30s for 300,000 chars.")

try:
    summarizer = pipeline("summarization", model="t5-small", device=device, torch_dtype=torch.float32)
except Exception as e:
    print(f"❌ Model loading failed: {str(e)}")
    exit(1)

def visualize_chunk_status(chunk_data):
    status_colors = {'summarized': 'green', 'skipped': 'orange', 'error': 'red'}
    labels = [f"C{i['chunk']}" for i in chunk_data]
    colors = [status_colors.get(i['status'], 'gray') for i in chunk_data]
    times = [i.get('time', 0.1) for i in chunk_data]  # Avoid zero-time bars

    fig, ax = plt.subplots(figsize=(10, 2.5))
    ax.barh(labels, times, color=colors)
    ax.set_xlabel("Time (s)")
    ax.set_title("📊 Chunk Processing Status")
    plt.tight_layout()

    buf = io.BytesIO()
    plt.savefig(buf, format='png')
    buf.seek(0)
    return buf

def summarize_file(file_bytes):
    start = time.time()
    chunk_info = []

    try:
        doc = fitz.open(stream=file_bytes, filetype="pdf")
        text = "".join(page.get_text("text") for page in doc)
        text = re.sub(r"\$\s*([^$]+)\s*\$", r"\1", text)
        text = re.sub(r"\\cap", "intersection", text)
        text = re.sub(r"\s+", " ", text).strip()
        text = "".join(c for c in text if ord(c) < 128)
    except Exception as e:
        return f"❌ Text extraction failed: {str(e)}", None

    if not text.strip():
        return "❌ No text found", None

    text = text[:300000]
    chunks = [text[i:i+2000] for i in range(0, len(text), 2000)]
    summaries = []

    for i, chunk in enumerate(chunks):
        chunk_start = time.time()
        chunk_result = {'chunk': i+1, 'status': '', 'time': 0}

        if time.time() - start > 20:
            summaries.append("⚠️ Stopped early")
            break

        if sum(1 for c in chunk if not c.isalnum()) / len(chunk) > 0.5:
            summaries.append(f"**Chunk {i+1}**: Skipped (equation-heavy)")
            chunk_result['status'] = 'skipped'
        else:
            try:
                summary = summarizer(chunk, max_length=60, min_length=10, do_sample=False)[0]['summary_text']
                summaries.append(f"**Chunk {i+1}**:\n{summary}")
                chunk_result['status'] = 'summarized'
            except Exception as e:
                summaries.append(f"**Chunk {i+1}**: ❌ Error: {str(e)}")
                chunk_result['status'] = 'error'

        chunk_result['time'] = time.time() - chunk_start
        chunk_info.append(chunk_result)

    final_summary = f"**Chars**: {len(text)}\n**Time**: {time.time()-start:.2f}s\n\n" + "\n\n".join(summaries)
    image_buf = visualize_chunk_status(chunk_info)
    return final_summary, image_buf

demo = gr.Interface(
    fn=summarize_file,
    inputs=gr.File(label="📄 Upload PDF", type="binary"),
    outputs=[
        gr.Textbox(label="📝 Summarized Output"),
        gr.Image(label="📊 Visual Process Flow")
    ],
    title="AI-Powered PDF Summarizer",
    description="Summarizes long PDFs (up to 300,000 characters) and visualizes chunk-level automation status."
)

if __name__ == "__main__":
    try:
        demo.launch(share=False, server_port=7860)
    except Exception as e:
        print(f"❌ Gradio launch failed: {str(e)}")