app_gradio.py

import gradio as gr
import os
import json
from pathlib import Path
from smart_ocr_pipeline_final import main as process_invoice

# Set page title and description
title = "🧠 Smart OCR Pipeline - Full AI Processing"
description = """
**Advanced Invoice OCR with AI Post-Processing**

This service uses:
- DocTR for text extraction
- GPT-4o-mini Vision for structured data extraction (with image)
- Advanced validation and error correction
- Math verification and auto-correction

**Cost:** ~$0.01-$0.05 per invoice
**Best for:** Complex invoices, highest accuracy needed
"""

def process_invoice_gradio(image):
    """Process invoice image and return structured data"""
    if image is None:
        return "Please upload an image first."
    
    try:
        # Save uploaded image temporarily
        temp_dir = "temp_uploads"
        Path(temp_dir).mkdir(exist_ok=True)
        
        temp_path = os.path.join(temp_dir, "temp_invoice.jpg")
        image.save(temp_path)
        
        # Process with OCR pipeline
        result = process_invoice(temp_path, temp_dir)
        
        # Format output as JSON
        output = json.dumps(result, indent=2, ensure_ascii=False)
        
        return output
        
    except Exception as e:
        return f"Error processing invoice: {str(e)}"

# Create Gradio interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown(f"# {title}")
    gr.Markdown(description)
    
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(
                type="pil", 
                label="Upload Invoice Image",
                sources=["upload", "clipboard"]
            )
            submit_btn = gr.Button("Process Invoice", variant="primary")
        
        with gr.Column():
            output = gr.Textbox(
                label="Extracted Data (JSON)",
                lines=20,
                max_lines=30
            )
    
    # Examples
    gr.Markdown("### 📋 Features:")
    gr.Markdown("""
    - ✅ Image preprocessing (deskew, denoise, enhance)
    - ✅ DocTR OCR extraction
    - ✅ GPT-4o-mini Vision post-processing
    - ✅ Automatic validation and error correction
    - ✅ Math verification
    - ✅ Structured JSON output
    """)
    
    # Event handler
    submit_btn.click(
        fn=process_invoice_gradio,
        inputs=image_input,
        outputs=output
    )

# Launch with authentication (optional)
if __name__ == "__main__":
    demo.launch(
        share=False,
        server_name="0.0.0.0",
        server_port=7860
    )