import os
import asyncio
from io import BytesIO
from diffusers import AutoPipelineForText2Image
import gradio as gr
from generate_prompts import generate_prompt

# Initialize model
model = AutoPipelineForText2Image.from_pretrained("stabilityai/sdxl-turbo")

async def generate_image(prompt, prompt_name):
    """
    Generates an image based on the provided prompt.
    Parameters:
        - prompt (str): The input text for image generation.
        - prompt_name (str): A name for the prompt, used for logging.
    Returns:
        bytes: The generated image data in bytes format, or None if generation fails.
    """
    try:
        print(f"Generating image for {prompt_name}")
        output = await model(prompt=prompt, num_inference_steps=50, guidance_scale=7.5)
        if isinstance(output.images, list) and len(output.images) > 0:
            image = output.images[0]
            buffered = BytesIO()
            image.save(buffered, format="JPEG")
            image_bytes = buffered.getvalue()
            return image_bytes
        else:
            return None
    except Exception as e:
        print(f"An error occurred while generating image for {prompt_name}: {e}")
        return None

async def queue_api_calls(sentence_mapping, character_dict, selected_style):
    """
    Generates images for all provided prompts in parallel using asyncio.
    Parameters:
        - sentence_mapping (dict): Mapping between paragraph numbers and sentences.
        - character_dict (dict): Dictionary mapping characters to their descriptions.
        - selected_style (str): Selected illustration style.
    Returns:
        dict: A dictionary where keys are paragraph numbers and values are image data in bytes format.
    """
    prompts = []
    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((paragraph_number, prompt))

    tasks = [generate_image(prompt, f"Prompt {paragraph_number}") for paragraph_number, prompt in prompts]
    responses = await asyncio.gather(*tasks)
    
    images = {paragraph_number: response for (paragraph_number, _), response in zip(prompts, responses)}
    return images

def process_prompt(sentence_mapping, character_dict, selected_style):
    """
    Processes the provided prompts and generates images.
    Parameters:
        - sentence_mapping (dict): Mapping between paragraph numbers and sentences.
        - character_dict (dict): Dictionary mapping characters to their descriptions.
        - selected_style (str): Selected illustration style.
    Returns:
        dict: A dictionary where keys are paragraph numbers and values are image data in bytes format.
    """
    try:
        loop = asyncio.get_running_loop()
    except RuntimeError:
        loop = asyncio.new_event_loop()
        asyncio.set_event_loop(loop)

    cmpt_return = loop.run_until_complete(queue_api_calls(sentence_mapping, character_dict, selected_style))
    return cmpt_return

gradio_interface = gr.Interface(
    fn=process_prompt,
    inputs=[gr.JSON(label="Sentence Mapping"), gr.JSON(label="Character Dict"), gr.Dropdown(["oil painting", "sketch", "watercolor"], label="Selected Style")],
    outputs="json"
).queue(default_concurrency_limit=20)  # Set concurrency limit if needed

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