app.py

import gradio as gr
import requests
import base64
import io
import time
import os
from PIL import Image


SR_API = os.getenv("sr_api")
ENHANCE_API = os.getenv("enhance_api")
ZERODCE_API = os.getenv("zerodce_api")
ZERODCE_PLUS_API = os.getenv("zerodce_plus_api")

# Get API endpoints from environment
API_ENDPOINTS = {
    "SR API": SR_API,
    "ENHANCE API": ENHANCE_API,
    "ZERODCE API": ZERODCE_API,
    "ZERODCE++ API": ZERODCE_PLUS_API
}

def apply_super_resolution(input_image, scale_factor, tile_size, api_choice):
    """Apply super-resolution to input image"""
    if input_image is None:
        return None, "❌ Please upload an image first", gr.update(visible=False)
    
    # Select API endpoint based on user choice
    api_endpoint = API_ENDPOINTS[api_choice]
    
    if not api_endpoint:
        return None, f"❌ {api_choice} endpoint not configured", gr.update(visible=False)
    
    try:
        # Convert PIL image to bytes
        img_buffer = io.BytesIO()
        input_image.save(img_buffer, format='PNG')
        img_bytes = img_buffer.getvalue()
        
        # Call super-resolution API
        response = requests.post(
            f"{api_endpoint}/invocations",
            headers={"Content-Type": "application/octet-stream"},
            data=img_bytes,
            params={
                "model_name": "RealESRGAN_x4plus",
                "outscale": scale_factor,
                "tile": tile_size,
                "fp32": False
            },
            timeout=300,
            verify=False  # Disable SSL verification for testing
        )
        
        if response.status_code == 200:
            result = response.json()
            # Decode base64 result
            enhanced_data = base64.b64decode(result["prediction"])
            enhanced_image = Image.open(io.BytesIO(enhanced_data))

            # Create download file
            timestamp = int(time.time())
            download_path = f"enhanced_image_{timestamp}.png"
            enhanced_image.save(download_path, format='PNG')
            
            status = f"✅ Enhancement successful using {api_choice}!\n"
            status += f"Model: {result.get('model', 'RealESRGAN_x4plus')}\n"
            status += f"Scale: {result.get('outscale', 4.0)}x\n"
            status += f"Input: {result.get('input_img_width', 0)}x{result.get('input_img_height', 0)}\n"
            status += f"Output: {result.get('output_img_width', 0)}x{result.get('output_img_height', 0)}"
            status += f"process_time: {result.get('upscaling_time', 0)}"
            
            # Return tuple for ImageSlider: [original, enhanced]
            return (input_image, enhanced_image), status, gr.update(visible=True, value=download_path)
        else:
            return None, f"❌ API Error: {response.status_code}\n{response.text}", gr.update(visible=False)
            
    except Exception as e:
        return None, f"❌ Error: {str(e)}", gr.update(visible=False)

def main():
    with gr.Blocks(title="Image Enhancement App") as demo:
        gr.Markdown("# 🚀 Image Enhancement App")
        gr.Markdown("Upload an image and enhance it with AI-powered super-resolution")
        
        # Row 1: Upload and Controls
        with gr.Row():
            with gr.Column(scale=3):
                input_image = gr.Image(
                    label="📤 Upload Image", 
                    type="pil",
                    height=300
                )
            
            with gr.Column(scale=1):
                gr.Markdown("### Enhancement Settings")
                
                api_dropdown = gr.Dropdown(
                    choices=["SR API", "ENHANCE API", "ZERODCE API", "ZERODCE++ API"], 
                    value="SR API", 
                    label="API Choice",
                    info="Choose which enhancement API to use"
                )
                
                scale_dropdown = gr.Dropdown(
                    choices=[1, 2, 4], 
                    value=4, 
                    label="Scale Factor",
                    info="How much to upscale the image"
                )

                tile_size = gr.Number(
                    value=0,
                    label="Tile Size",
                    info="Tile size for the image"
                )
                
                enhance_button = gr.Button(
                    "✨ Enhance Image", 
                    variant="primary",
                    size="lg"
                )
                
                status_text = gr.Textbox(
                    label="Status",
                    lines=6,
                    value="Ready to enhance images!",
                    interactive=False
                )
        
        # Row 2: Before/After Comparison with Image Slider
        with gr.Row():
            gr.Markdown("### 📊 Before vs After Comparison")
        
        with gr.Row():
            image_slider = gr.ImageSlider(
                label="Original vs Enhanced",
                height=500,
                interactive=False
            )
        
        # Download button
        with gr.Row():
            download_button = gr.DownloadButton(
                "📥 Download Enhanced Image",
                visible=False,
                size="lg"
            )
        
        # Event handlers
        enhance_button.click(
            fn=apply_super_resolution,
            inputs=[input_image, scale_dropdown, tile_size, api_dropdown],
            outputs=[image_slider, status_text, download_button],
            show_progress=True
        )
        
        # Clear results when new image is uploaded
        input_image.change(
            fn=lambda: (None, "Image uploaded! Ready to enhance.", gr.update(visible=False)),
            outputs=[image_slider, status_text, download_button]
        )

    # Launch the app
    demo.queue(default_concurrency_limit=3, max_size=10).launch()

if __name__ == "__main__":
    main()