app.py

import gradio as gr
import spaces
from PIL import Image
from gradio_client import Client, handle_file
import logging
import tempfile
import os

# ==========================================================
# Setup
# ==========================================================
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

vton_client = None


# ==========================================================
# Init Remote Try-On Model (API Endpoint)
# ==========================================================
def init_client():
    global vton_client
    try:
        logger.info("🔗 Connecting to IDM-VTON API...")
        # ✅ Use the maintained API Space (not model repo)
        vton_client = Client("yisol-idm-vton-api.hf.space")
        logger.info("✅ Connected to yisol-idm-vton-api.hf.space")
        return True
    except Exception as e:
        logger.error(f"❌ Connection failed: {e}")
        vton_client = None
        return False


model_ready = init_client()


# ==========================================================
# GPU Accelerated Try-On Function
# ==========================================================
@spaces.GPU(duration=180)
def virtual_tryon(person_image, garment_image, progress=gr.Progress()):
    global vton_client

    try:
        if not person_image or not garment_image:
            raise gr.Error("Please upload both images!")

        if not vton_client:
            if not init_client():
                raise gr.Error("⚠️ Model API unavailable. Try again later.")

        logger.info("🎯 Starting try-on pipeline...")
        progress(0.2, desc="📸 Preparing images...")

        # Save temporary input files
        person_tmp = tempfile.NamedTemporaryFile(delete=False, suffix='.jpg')
        garment_tmp = tempfile.NamedTemporaryFile(delete=False, suffix='.jpg')

        person_image.convert('RGB').save(person_tmp.name, 'JPEG', quality=95)
        garment_image.convert('RGB').save(garment_tmp.name, 'JPEG', quality=95)

        progress(0.5, desc="🪄 Running AI model (60-90s)...")

        # Call remote API Space
        result = vton_client.predict(
            handle_file(person_tmp.name),
            handle_file(garment_tmp.name),
            "universal garment",   # 👗 supports saree, kurta, dress, etc.
            True,
            True,
            30,
            42,
            api_name="/tryon"
        )

        progress(0.9, desc="🎨 Generating result...")

        result_path = result[0] if isinstance(result, (tuple, list)) else result
        result_image = Image.open(result_path)

        # Cleanup temp files
        for tmp in [person_tmp.name, garment_tmp.name]:
            try:
                os.unlink(tmp)
            except:
                pass

        progress(1.0, desc="✅ Done!")
        logger.info("✅ Try-on complete!")
        return result_image

    except Exception as e:
        if "ZeroGPU" in str(e):
            raise gr.Error(
                "⚠️ GPU quota exceeded. Log in to Hugging Face or retry later."
            )
        logger.error(f"❌ Error: {e}")
        raise gr.Error(f"Failed: {str(e)}")


# ==========================================================
# Gradio Interface
# ==========================================================
with gr.Blocks(title="Mirro Virtual Try-On", theme=gr.themes.Soft()) as demo:

    gr.Markdown("# 👗 Mirro Virtual Try-On\n### 🆓 AI-Powered Outfit Fitting (All Garments)")

    if model_ready:
        gr.Markdown("🟢 **Model ready!** Upload your photos below.")
    else:
        gr.Markdown("🟡 **Connecting...** Try again in a moment.")

    with gr.Row():
        with gr.Column():
            gr.Markdown("### 📸 Upload Images")
            person_input = gr.Image(label="👤 Person Photo", type="pil", sources=["upload", "webcam"])
            garment_input = gr.Image(label="👔 Garment Photo", type="pil", sources=["upload", "webcam"])
            btn = gr.Button("✨ Generate Try-On", variant="primary")

        with gr.Column():
            gr.Markdown("### 🎯 Result")
            output_image = gr.Image(label="Virtual Try-On Result", type="pil")

    with gr.Accordion("💡 Tips for Best Results", open=False):
        gr.Markdown("""
        - Clear, front-facing full-body person photo  
        - Plain background recommended  
        - Any garment type: saree, kurta, dress, jeans, jacket, etc.  
        - Processing takes ~60-90 seconds on ZeroGPU  
        """)

    btn.click(
        fn=virtual_tryon,
        inputs=[person_input, garment_input],
        outputs=output_image,
        api_name="predict"
    )

    gr.Markdown("""
    ---
    <div style="text-align:center; color:#666; padding:15px;">
        <p>🆓 Powered by Hugging Face ZeroGPU + IDM-VTON API</p>
        <p style="font-size:0.9em;">Supports all garment categories — production ready for iOS integration</p>
    </div>
    """)

if __name__ == "__main__":
    demo.queue(max_size=20)
    demo.launch(server_name="0.0.0.0", server_port=7860)