app.py

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

# Global variable to track active requests
active_requests = {}

# Predefined assets
ASSETS = {
    "top": {
        "reference_image": "pullover.png",
        "prompt": "black pullover with the logo on the chest"
    },
    "bottom": {
        "reference_image": "sweatpants.png",
        "prompt": "black sweatpants with the silver logo on it"
    },
    "logo": "logo.png"
}

def needs_resizing(image_path, max_short_edge=1280):
    # Check if image needs resizing (returns True if short edge > max_short_edge)
    with Image.open(image_path) as img:
        width, height = img.size
        return min(width, height) > max_short_edge

def resize_image_to_short_edge(image_path, max_short_edge=1280):
    # Resize image so the short edge is at most max_short_edge pixels
    with Image.open(image_path) as img:
        width, height = img.size
        format = img.format
        
        # Determine scaling factor
        if width < height:
            scaling_factor = max_short_edge / width
        else:
            scaling_factor = max_short_edge / height
        
        # Calculate new dimensions and resize
        new_width = int(width * scaling_factor)
        new_height = int(height * scaling_factor)
        resized_img = img.resize((new_width, new_height), Image.LANCZOS)
        
        # Save to bytes buffer
        buffer = io.BytesIO()
        resized_img.save(buffer, format=format)
        buffer.seek(0)
        return buffer

def validate_assets():
    # Check if all required asset files exist
    missing = []
    for asset_type in ["top", "bottom"]:
        if not os.path.exists(ASSETS[asset_type]["reference_image"]):
            missing.append(ASSETS[asset_type]["reference_image"])
    if not os.path.exists(ASSETS["logo"]):
        missing.append(ASSETS["logo"])
    if missing:
        raise FileNotFoundError(f"Missing required asset files: {', '.join(missing)}")

def generate_and_wait_for_image(
    api_key: str,
    input_image: str,
    garment_type: str,
    progress=gr.Progress()
):
    # Make POST request and automatically poll for the result image
    # Validate assets first
    try:
        validate_assets()
    except FileNotFoundError as e:
        return None, str(e)
    
    # Create a unique ID for this request
    request_id = str(time.time())
    active_requests[request_id] = True
    
    try:
        # Start the job
        post_url = "https://api.stability.ai/private/alo/v1/vto-acolade"
        headers = {
            "Authorization": f"Bearer {api_key}",
            "Accept": "application/json"
        }
        
        files = {}
        
        try:
            progress(0, desc="Starting image generation...")
            
            # Prepare all required files
            files = {
                'logo_image': open(ASSETS["logo"], 'rb'),
                'reference_image': open(ASSETS[garment_type]["reference_image"], 'rb')
            }
            
            # Only resize input_image if needed
            if needs_resizing(input_image):
                files['input_image'] = resize_image_to_short_edge(input_image)
            else:
                files['input_image'] = open(input_image, 'rb')

            data = {
                'reference_image_type': (f'{garment_type}'),
                'output_format': (None, "png")  # Hardcoded to PNG
            }
            
            # Submit the job with timeout
            print(headers)
            print(files)
            response = requests.post(post_url, headers=headers, files=files, data=data, timeout=10)
            response.raise_for_status()
            job_data = response.json()
            job_id = job_data.get('id')
            
            if not job_id:
                return None, "Error: No job ID received in response"
            
            # Now poll for results with optimized timing
            get_url = f"https://api.stability.ai/private/alo/v1/results/{job_id}"
            headers = {
                "authorization": f"{api_key}",
                "accept": "*/*"
            }
            
            progress(0.3, desc="Processing your image...")
            
            # Optimized polling strategy
            max_attempts = 20
            initial_delay = 1.0
            max_delay = 5.0
            current_delay = initial_delay
            
            for attempt in range(max_attempts):
                if not active_requests.get(request_id, False):
                    return None, "Request cancelled by user"
                
                time.sleep(current_delay)
                progress(0.3 + (0.7 * attempt/max_attempts), 
                        desc=f"Checking status (attempt {attempt + 1}/{max_attempts})")
                
                try:
                    with requests.Session() as session:
                        response = session.get(get_url, headers=headers, timeout=10)
                    
                    if response.status_code == 200:
                        if 'image' in response.headers.get('Content-Type', ''):
                            img = Image.open(io.BytesIO(response.content))
                            progress(1.0, desc="Done!")
                            return img, f"Success! Job ID: {job_id}"
                        else:
                            json_response = response.json()
                            if json_response.get('status') == 'processing':
                                current_delay = min(current_delay * 1.5, max_delay)
                                continue
                            return None, f"API response: {json_response}"
                    
                    elif response.status_code == 202:
                        current_delay = min(current_delay * 1.5, max_delay)
                        continue
                    
                    else:
                        response.raise_for_status()
                        
                except requests.exceptions.RequestException:
                    current_delay = min(current_delay * 1.5, max_delay)
                    continue
                    
            return None, f"Timeout after {max_attempts} attempts. Job ID: {job_id}"
                
        except Exception as e:
            return None, f"Error: {str(e)}"
        finally:
            # Clean up file handles
            for key in files:
                if hasattr(files[key], 'close'):
                    files[key].close()
    
    finally:
        # Clean up the request tracking
        active_requests.pop(request_id, None)

def cancel_request():
    """Function to cancel active requests"""
    for req_id in list(active_requests.keys()):
        active_requests[req_id] = False

with gr.Blocks(title="Virtual Try-On Demo") as demo:
    gr.Markdown("""
    # Virtual Try-On Demo v1
    Upload your photo and select garment type to generate your VTon image.
    """)
    
    with gr.Row():
        with gr.Column():
            api_key = gr.Textbox(
                label="API Key",
                value="",
                type="password"
            )
            
            input_image = gr.Image(
                label="Upload Your Photo",
                type="filepath",
                sources=["upload"],
                height=300
            )
            
            garment_type = gr.Dropdown(
                label="Garment Type",
                choices=["top", "bottom"],
                value="top"
            )
            
            with gr.Row():
                submit_btn = gr.Button("Generate Image", variant="primary")
                cancel_btn = gr.Button("Cancel Request")
        
        with gr.Column():
            output_image = gr.Image(
                label="Generated Result",
                interactive=False,
                height=400
            )
            status_output = gr.Textbox(
                label="Status",
                interactive=False
            )
    
    submit_btn.click(
        fn=generate_and_wait_for_image,
        inputs=[api_key, input_image, garment_type],
        outputs=[output_image, status_output]
    )
    
    cancel_btn.click(
        fn=cancel_request,
        inputs=None,
        outputs=None,
        queue=False
    )
    gr.Markdown("""
    # Note:\n
    The image will be resized to 1280px on the short edge.
    """)

if __name__ == "__main__":
    # Verify required files exist before launching
    try:
        validate_assets()
        demo.launch()
    except FileNotFoundError as e:
        print(f"Error: {str(e)}")
        print("Please make sure these files exist in the same directory:")
        print("- logo.png")
        print("- pullover.png (for top selection)")
        print("- sweatpants.png (for bottom selection)")