app.py

# ================================================================
# AI IMAGE GENERATOR - FINAL FIXED SOLUTION
# ================================================================
import logging
import torch
import gradio as gr
import numpy as np
import random
from diffusers import AutoPipelineForText2Image
from PIL import Image

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

# Constants
MODEL_ID = "stabilityai/sdxl-turbo"
MAX_SEED = np.iinfo(np.int32).max
DEFAULT_SIZE = 512
DEFAULT_STEPS = 1
DEFAULT_GUIDANCE = 0.0
NUM_IMAGES = 4

# Set up device
device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if device == "cuda" else torch.float32
logger.info(f"Using device: {device}, dtype: {dtype}")

# Load model
try:
    logger.info("⚙️ Loading model...")
    pipe = AutoPipelineForText2Image.from_pretrained(
        MODEL_ID,
        torch_dtype=dtype,
        use_safetensors=True
    )
    pipe = pipe.to(device)
    logger.info("✅ Model loaded successfully!")
except Exception as e:
    logger.error(f"Model loading failed: {str(e)}")
    gr.Error(f"Model loading failed: {str(e)}")

def generate_images(
    prompt,
    seed,
    randomize_seed,
    width,
    height,
    guidance_scale,
    num_inference_steps,
    num_images
):
    """Generate images with robust error handling"""
    # Input validation
    if not prompt.strip():
        return [Image.new('RGB', (512, 512), (255, 0, 0))] * num_images, "ERROR: Empty prompt"
    
    try:
        images = []
        seeds_used = []
        negative_prompt = ""  # Fixed: Negative prompt as empty string
        
        for i in range(num_images):
            # Seed handling
            if randomize_seed:
                current_seed = random.randint(0, MAX_SEED)
            else:
                current_seed = seed + i
                
            generator = torch.Generator(device=device).manual_seed(current_seed)
            
            # Generate image
            image = pipe(
                prompt=prompt,
                negative_prompt=negative_prompt,
                num_inference_steps=int(num_inference_steps),
                guidance_scale=guidance_scale,
                width=int(width),
                height=int(height),
                generator=generator,
            ).images[0]
            
            images.append(image)
            seeds_used.append(current_seed)
        
        return images, ", ".join(map(str, seeds_used))
    
    except torch.cuda.OutOfMemoryError:
        return [Image.new('RGB', (512, 512), (255, 165, 0))] * num_images, "ERROR: Out of GPU memory"
    except Exception as e:
        logger.error(f"Generation error: {str(e)}")
        return [Image.new('RGB', (512, 512), (255, 0, 0))] * num_images, f"ERROR: {str(e)}"

# UI Components
css = """
/* ... আপনার CSS কোড অপরিবর্তিত থাকবে ... */
"""

# Create interface
with gr.Blocks(css=css, title="AI Image Generator") as demo:
    with gr.Column(elem_id="main-container"):
        # Header
        gr.Markdown("# 🎨 AI Image Generator", elem_classes="header")
        gr.Markdown("Create stunning visuals from text prompts", elem_classes="header")
        
        # Inputs
        with gr.Row():
            prompt = gr.Textbox(label="Prompt", placeholder="Describe what you want to see...", lines=2)
        
        # Generate button
        generate_btn = gr.Button("✨ Generate Images", elem_classes="btn-generate")
        
        # Gallery output
        gallery = gr.Gallery(label="Generated Images", columns=2, height="auto", elem_classes="gallery-grid")
        seed_info = gr.Textbox(label="Seeds Used", interactive=False)
        
        # Advanced settings
        with gr.Accordion("⚙️ Advanced Settings", open=False):
            with gr.Row():
                num_images = gr.Slider(1, 4, value=2, step=1, label="Number of Images")
                width = gr.Slider(256, 768, value=DEFAULT_SIZE, step=32, label="Width")
                height = gr.Slider(256, 768, value=DEFAULT_SIZE, step=32, label="Height")
            with gr.Row():
                guidance_scale = gr.Slider(0.0, 10.0, value=DEFAULT_GUIDANCE, step=0.1, label="Guidance Scale")
                num_inference_steps = gr.Slider(1, 4, value=DEFAULT_STEPS, step=1, label="Steps")
            with gr.Row():
                seed = gr.Slider(0, MAX_SEED, value=0, step=1, label="Seed")
                randomize_seed = gr.Checkbox(value=True, label="Randomize Seed")
        
        # Examples
        gr.Markdown("### ✨ Example Prompts:")
        gr.Examples(
            examples=[
                "A red apple on a wooden table",
                "Mountain landscape at sunset",
                "Cyberpunk city street at night"
            ],
            inputs=prompt
        )
    
    # Event handling - FINAL FIX
    generate_btn.click(
        fn=generate_images,
        inputs=[
            prompt,
            seed,
            randomize_seed,
            width,
            height,
            guidance_scale,
            num_inference_steps,
            num_images
        ],
        outputs=[gallery, seed_info]
    )

# Launch the app
if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)