app.py

from PIL import Image, ImageDraw, ImageFont
import gradio as gr
from smolagents import CodeAgent, InferenceClientModel, DuckDuckGoSearchTool, Tool
from gradio_client import Client

#%% Tool Wrapper for the Hugging Face Space
class TextToImageTool(Tool):
    name = "text_to_image"
    description = "Generate an image from a text prompt using m-ric/text-to-image."
    
    # `inputs` is a dictionary: argument_name -> type
    #inputs = {"prompt": str}
    #output_type = "image" 
    #outputs = {"image": "image"}  # Optional, type can be a string or class

    inputs = {
        "prompt": {"type": "str"}  # Must be a dict, not just a type
    }
    output_type = "image"
    outputs = {
        "image": "image"
    }
    
    def __init__(self):
        super().__init__()
        self.client = Client("m-ric/text-to-image")

    def run(self, prompt: str):
        return self.client.predict(prompt, api_name="/predict")  # Returns PIL.Image


#%% Utility functions
def add_label_to_image(image, label):
    draw = ImageDraw.Draw(image)
    font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
    font_size = 30
    try:
        font = ImageFont.truetype(font_path, font_size)
    except:
        font = ImageFont.load_default()
    text_bbox = draw.textbbox((0, 0), label, font=font)
    text_width = text_bbox[2] - text_bbox[0]
    text_height = text_bbox[3] - text_bbox[1]
    position = (image.width - text_width - 20, image.height - text_height - 20)
    rect_margin = 10
    rect_position = [
        position[0] - rect_margin, position[1] - rect_margin,
        position[0] + text_width + rect_margin, position[1] + text_height + rect_margin
    ]
    draw.rectangle(rect_position, fill=(0, 0, 0, 128))
    draw.text(position, label, fill="white", font=font)
    return image

def plot_and_save_agent_image(image, label, save_path=None):
    labeled_image = add_label_to_image(image, label)
    labeled_image.show()
    if save_path:
        labeled_image.save(save_path)
        print(f"Image saved to {save_path}")

def generate_prompts_for_object(object_name):
    return {
        "past": f"Show an old version of a {object_name} from its early days.",
        "present": f"Show a {object_name} with current features/design/technology.",
        "future": f"Show a futuristic version of a {object_name}, by predicting advanced features and futuristic design."
    }

def generate_object_history(object_name):
    prompts = generate_prompts_for_object(object_name)
    labels = {
        "past": f"{object_name} - Past",
        "present": f"{object_name} - Present",
        "future": f"{object_name} - Future"
    }

    images = []
    for time_period, prompt in prompts.items():
        print(f"Generating {time_period} frame: {prompt}")
        result = agent.run(prompt)  # Runs tool
        if hasattr(result, "to_raw"):  # If wrapped output
            result = result.to_raw()
        images.append(result)
        plot_and_save_agent_image(result, labels[time_period], save_path=f"{object_name}_{time_period}.png")

    gif_path = f"{object_name}_evolution.gif"
    images[0].save(gif_path, save_all=True, append_images=images[1:], duration=1000, loop=0)
    return images, gif_path

#%% Tool & Agent Setup
image_generation_tool = TextToImageTool()
search_tool = DuckDuckGoSearchTool()
llm_engine = InferenceClientModel("Qwen/Qwen2.5-72B-Instruct")

agent = CodeAgent(tools=[image_generation_tool, search_tool], model=llm_engine)

#%% Gradio Interface
def create_gradio_interface():
    with gr.Blocks() as demo:
        gr.Markdown("# TimeMetamorphy: an object Evolution Generator")
        gr.Markdown("""
        ## Unlocking the secrets of time!
        Enter an object name (like bicycle or smartphone), and this app will generate its visual evolution.
        """)

        default_images = [
            ("car_past.png", "Car - Past"),
            ("car_present.png", "Car - Present"),
            ("car_future.png", "Car - Future")
        ]
        default_gif_path = "car_evolution.gif"

        with gr.Row():
            with gr.Column():
                object_name_input = gr.Textbox(label="Enter an object name", placeholder="e.g., bicycle, phone")
                generate_button = gr.Button("Generate Evolution")
                image_gallery = gr.Gallery(label="Generated Images", columns=3, rows=1, value=default_images)
                gif_output = gr.Image(label="Generated GIF", value=default_gif_path)

        generate_button.click(fn=generate_object_history, inputs=[object_name_input], outputs=[image_gallery, gif_output])
    return demo

# Launch app
demo = create_gradio_interface()
demo.launch(share=True)