app.py

import gradio as gr
from PIL import Image
from typing import Tuple, Optional
import os
import shutil

# Import the core logic functions
from models import generate_remixed_image

# --- Setup Dummy Examples for immediate runnability ---
EXAMPLE_DIR = "examples"
if not os.path.exists(EXAMPLE_DIR):
    os.makedirs(EXAMPLE_DIR)

# Create dummy images for examples if they don't exist
def create_dummy_image(filename, color):
    path = os.path.join(EXAMPLE_DIR, filename)
    if not os.path.exists(path):
        dummy_img = Image.new('RGB', (100, 100), color=color)
        dummy_img.save(path)

# Ensure example files exist (using descriptive names for the demo)
create_dummy_image("eiffel_tower.jpg", "blue")
create_dummy_image("sunset.jpg", "orange")
create_dummy_image("painting.jpg", "green")

# --- Component Definitions ---

# Note: The component type should be "filepath" for external API interaction 
# to ensure we pass a local file path.
IMAGE_INPUT_PROPS = {
    "type": "filepath",
    "image_mode": "RGB",
    "height": 250,
    "label": "Drag Image Here",
    "sources": ["upload"],
    "interactive": True
}

# --- Interface Function Wrapper ---

def remixer_wrapper(
    model_choice: str,
    prompt: str,
    img1_path: Optional[str],
    img2_path: Optional[str],
    img3_path: Optional[str],
    progress: gr.Progress = gr.Progress()
) -> Tuple[str, Image.Image]:
    """
    Wrapper function to handle Gradio inputs and call the main generation logic.
    """
    if not prompt:
        raise gr.Error("Please provide a creative prompt.")
        
    progress(0.1, desc=f"Analyzing inputs using {model_choice}...")
    
    # We pass the paths directly, the model logic handles validation and API interaction
    result_prompt, result_image = generate_remixed_image(
        model_choice,
        prompt,
        img1_path,
        img2_path,
        img3_path
    )
    
    progress(1.0, desc="Generation complete.")
    
    return result_prompt, result_image

# --- Gradio Blocks Layout ---

with gr.Blocks(title="Image Remixer AI", theme=gr.themes.Soft()) as demo:
    gr.HTML("""
        <div style="text-align: center; max-width: 900px; margin: 0 auto; padding: 20px;">
            <h1 style="font-size: 2.5em; margin-bottom: 10px;">🎨 Multimodal Image Remixer</h1>
            <p style="margin-bottom: 20px;">Upload 3 images and provide a prompt to fuse them into a new creative result using advanced AI models.</p>
            <p>Built with <a href="https://huggingface.co/spaces/akhaliq/anycoder" target="_blank" style="color: blue;">anycoder</a></p>
        </div>
    """)

    with gr.Row():
        image_input_1 = gr.Image(**IMAGE_INPUT_PROPS)
        image_input_2 = gr.Image(**IMAGE_INPUT_PROPS)
        image_input_3 = gr.Image(**IMAGE_INPUT_PROPS)
        
    prompt_input = gr.Textbox(
        label="Creative Prompt/Instructions",
        placeholder="Describe the desired fusion style, mood, and content (e.g., 'A dramatic digital painting, blending these objects into a high-tech cityscape').",
        lines=2
    )

    with gr.Row():
        model_selector = gr.Radio(
            choices=["gemini-2", "gpt image-1"],
            value="gpt image-1",
            label="Analysis Model (Used to create a detailed prompt for DALL-E 3 Generation)",
            info="Choose the underlying multimodal model for creative analysis: Gemini 2.5 Flash Live ('gemini-2') or GPT-4o ('gpt image-1').",
            scale=1
        )
        submit_btn = gr.Button("Remix Images (Requires API Keys)", variant="primary", scale=1)

    with gr.Column():
        generated_prompt_output = gr.Textbox(label="AI Generated DALL-E Prompt", interactive=False)
        remixed_image_output = gr.Image(label="Remixed Output Image (DALL-E 3)", interactive=False)

    
    # --- Event Binding ---
    submit_btn.click(
        fn=remixer_wrapper,
        inputs=[
            model_selector,
            prompt_input,
            image_input_1,
            image_input_2,
            image_input_3
        ],
        outputs=[generated_prompt_output, remixed_image_output],
    )

    # --- Examples ---
    gr.Examples(
        examples=[
            [
                "A hyper-realistic oil painting of a mythical creature created by combining the style and features of the inputs.",
                "gpt image-1",
                os.path.join(EXAMPLE_DIR, "eiffel_tower.jpg"),
                os.path.join(EXAMPLE_DIR, "sunset.jpg"),
                os.path.join(EXAMPLE_DIR, "painting.jpg"),
            ]
        ],
        inputs=[prompt_input, model_selector, image_input_1, image_input_2, image_input_3],
        cache_examples=False,
    )

if __name__ == "__main__":
    demo.launch()