app.py

import os
import numpy as np
import random
from pathlib import Path
from PIL import Image
import streamlit as st
from huggingface_hub import InferenceClient, AsyncInferenceClient
from gradio_client import Client, handle_file
import asyncio
from concurrent.futures import ThreadPoolExecutor

MAX_SEED = np.iinfo(np.int32).max
HF_TOKEN = os.environ.get("HF_TOKEN")
HF_TOKEN_UPSCALER = os.environ.get("HF_TOKEN_UPSCALER")
client = AsyncInferenceClient()
llm_client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
DATA_PATH = Path("./data")
DATA_PATH.mkdir(exist_ok=True)

def run_async(func):
    loop = asyncio.new_event_loop()
    asyncio.set_event_loop(loop)
    executor = ThreadPoolExecutor(max_workers=1)
    result = loop.run_in_executor(executor, func)
    return loop.run_until_complete(result)

def enable_lora(lora_add, basemodel):
    return lora_add if lora_add else basemodel

async def generate_image(combined_prompt, model, width, height, scales, steps, seed):
    try:
        if seed == -1:
            seed = random.randint(0, MAX_SEED)
        seed = int(seed)
        image = await client.text_to_image(
            prompt=combined_prompt, height=height, width=width, guidance_scale=scales,
            num_inference_steps=steps, model=model
        )
        return image, seed
    except Exception as e:
        return f"Error al generar imagen: {e}", None

def get_upscale_finegrain(prompt, img_path, upscale_factor):
    try:
        client = Client("finegrain/finegrain-image-enhancer", hf_token=HF_TOKEN_UPSCALER)
        result = client.predict(
            input_image=handle_file(img_path), prompt=prompt, upscale_factor=upscale_factor
        )
        return result[1] if isinstance(result, list) and len(result) > 1 else None
    except Exception as e:
        return None

def save_prompt(prompt_text, seed):
    try:
        prompt_file_path = DATA_PATH / f"prompt_{seed}.txt"
        with open(prompt_file_path, "w") as prompt_file:
            prompt_file.write(prompt_text)
        return prompt_file_path
    except Exception as e:
        st.error(f"Error al guardar el prompt: {e}")
        return None

def save_image(image, seed):
    image_path = DATA_PATH / f"generated_image_{seed}.jpg"
    image.save(image_path)
    return image_path

async def improve_prompt(prompt, language):
    try:
        instruction = (
            "With this idea, describe in English a detailed txt2img prompt in 500 characters at most, add illumination, atmosphere, cinematic elements, and characters..."
            if language == "English"
            else "Con esta idea, describe en español un prompt detallado de txt2img en un máximo de 500 caracteres, añadiendo iluminación, atmósfera, elementos cinematográficos y personajes..."
        )

        formatted_prompt = f"{prompt}: {instruction}"
        response = llm_client.text_generation(formatted_prompt, max_new_tokens=300)
        improved_text = response['generated_text'].strip() if 'generated_text' in response else response.strip()
        return improved_text[:300] if len(improved_text) > 300 else improved_text
    except Exception as e:
        return f"Error mejorando el prompt: {e}"

async def gen(prompt, basemodel, width, height, scales, steps, seed, upscale_factor, process_upscale, lora_model, process_lora, process_enhancer, prompt_language):
    model = enable_lora(lora_model, basemodel) if process_lora else basemodel
    combined_prompt = prompt  # Usar el prompt original por defecto

    if process_enhancer:
        improved_prompt = await improve_prompt(prompt, prompt_language)
        combined_prompt = f"{prompt} {improved_prompt}"
    
    if seed == -1:
        seed = random.randint(0, MAX_SEED)
    seed = int(seed)
    progress_bar = st.progress(0)
    image, seed = await generate_image(combined_prompt, model, width, height, scales, steps, seed)
    progress_bar.progress(50)

    if isinstance(image, str) and image.startswith("Error"):
        progress_bar.empty()
        return [image, None, combined_prompt]

    image_path = save_image(image, seed)
    prompt_file_path = save_prompt(combined_prompt, seed)

    if process_upscale:
        upscale_image_path = get_upscale_finegrain(combined_prompt, image_path, upscale_factor)
        if upscale_image_path:
            upscale_image = Image.open(upscale_image_path)
            upscale_image.save(DATA_PATH / f"upscale_image_{seed}.jpg", format="JPEG")
            progress_bar.progress(100)
            image_path.unlink()  
            return [str(DATA_PATH / f"upscale_image_{seed}.jpg"), str(prompt_file_path)]
        else:
            progress_bar.empty()
            return [str(image_path), str(prompt_file_path)]
    else:
        progress_bar.progress(100)
        return [str(image_path), str(prompt_file_path)]

def main():
    st.set_page_config(layout="wide")
    st.title("FLUX with enhancer and upscaler with LORA model training")
    
    prompt = st.sidebar.text_input("Descripción de la imagen", max_chars=200)
    process_enhancer = st.sidebar.checkbox("Mejorar Prompt", value=True)  
    prompt_language = st.sidebar.selectbox("Idioma para mejorar el prompt", ["English", "Spanish"])
    basemodel = st.sidebar.selectbox("Modelo Base", ["black-forest-labs/FLUX.1-schnell", "black-forest-labs/FLUX.1-DEV"])
    lora_model = st.sidebar.selectbox("LORA Realismo", ["Shakker-Labs/FLUX.1-dev-LoRA-add-details", "XLabs-AI/flux-RealismLora"])
    format_option = st.sidebar.selectbox("Formato", ["9:16", "16:9"])
    process_lora = st.sidebar.checkbox("Procesar LORA", value=True)
    process_upscale = st.sidebar.checkbox("Procesar Escalador", value=True)
    upscale_factor = st.sidebar.selectbox("Factor de Escala", [2, 4, 8], index=0)
    scales = st.sidebar.slider("Escalado", 1, 20, 10)
    steps = st.sidebar.slider("Pasos", 1, 100, 20)
    seed = st.sidebar.number_input("Semilla", value=-1)

    if format_option == "9:16":
        width = 720
        height = 1280
    else:
        width = 1280
        height = 720

    if st.sidebar.button("Generar Imagen"):
        with st.spinner("Mejorando y generando imagen..."):
            result = asyncio.run(gen(prompt, basemodel, width, height, scales, steps, seed, upscale_factor, process_upscale, lora_model, process_lora, process_enhancer, prompt_language))
            image_paths = result[0]
            prompt_file = result[1]

        st.write(f"Image paths: {image_paths}")

        if image_paths:
            if Path(image_paths).exists():
                st.image(image_paths, caption="Imagen Generada")
            else:
                st.error("El archivo de imagen no existe.")

            if prompt_file and Path(prompt_file).exists():
                prompt_text = Path(prompt_file).read_text()
                st.write(f"Prompt utilizado: {prompt_text}")
            else:
                st.write("El archivo del prompt no está disponible.")
        else:
            st.error("No se pudo generar la imagen.")

if __name__ == "__main__":
    main()