app.py

import gradio as gr
from diffusers import AutoPipelineForText2Image
from generate_propmts import generate_prompt
from concurrent.futures import ThreadPoolExecutor, as_completed
from PIL import Image
import traceback

# Load the model once outside of the function
model = AutoPipelineForText2Image.from_pretrained("stabilityai/sdxl-turbo")

# Create a thread-local storage for step indices
scheduler_step_storage = threading.local()

def generate_image(prompt):
    try:
        # Initialize step index per thread if not already set
        if not hasattr(scheduler_step_storage, 'step'):
            scheduler_step_storage.step = 0
        
        # Use the thread-local step index
        output = model(
            prompt=prompt, 
            num_inference_steps=1,  # Add a sensible default for inference steps
            guidance_scale=0.0, 
            output_type="pil"  # Directly get PIL Image objects
        )

        # Increment the step index after generating the image
        scheduler_step_storage.step += 1  
        
        # Check for output validity and return
        if output.images:
            return output.images[0]
        else:
            raise Exception("No images returned by the model.")
    
    except Exception as e:
        print(f"Error generating image: {e}")
        traceback.print_exc()
        return None  # Return None on error to handle it gracefully in the UI


def inference(sentence_mapping, character_dict, selected_style):
    images = []
    print(f'sentence_mapping: {sentence_mapping}, character_dict: {character_dict}, selected_style: {selected_style}')
    prompts = []

    # Generate prompts for each paragraph
    for paragraph_number, sentences in sentence_mapping.items():
        combined_sentence = " ".join(sentences)
        prompt = generate_prompt(combined_sentence, sentence_mapping, character_dict, selected_style)
        prompts.append(prompt)
        print(f"Generated prompt for paragraph {paragraph_number}: {prompt}")

    with ThreadPoolExecutor() as executor:
        futures = [executor.submit(generate_image, prompt) for prompt in prompts]

        for future in as_completed(futures):
            try:
                image = future.result()
                if image:
                    images.append(image)
            except Exception as e:
                print(f"Error processing prompt: {e}")
                traceback.print_exc()

    return images

gradio_interface = gr.Interface(
    fn=inference,
    inputs=[
        gr.JSON(label="Sentence Mapping"),
        gr.JSON(label="Character Dict"),
        gr.Dropdown(["oil painting", "sketch", "watercolor"], label="Selected Style")
    ],
    outputs=gr.Gallery(label="Generated Images")
)

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