import torch
import spaces
import gradio as gr
from diffusers import FluxInpaintPipeline
import random
import numpy as np

MARKDOWN = """
# FLUX.1 Inpainting 🎨
Shoutout to [Black Forest Labs](https://huggingface.co/black-forest-labs) team for 
creating this amazing model, and a big thanks to [Gothos](https://github.com/Gothos) 
for taking it to the next level by enabling inpainting with the FLUX.
"""

MAX_SEED = np.iinfo(np.int32).max
DEVICE = "cuda" #if torch.cuda.is_available() else "cpu"

pipe = FluxInpaintPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16).to(DEVICE)

@spaces.GPU()
def process(input_image_editor, input_text, strength, seed, randomize_seed, num_inference_steps, guidance_scale=3.5, progress=gr.Progress(track_tqdm=True)):
    if not input_text:
        raise gr.Error("Please enter a text prompt.")

    image = input_image_editor['background']

    if not image:
        raise gr.Error("Please upload an image.")

    width, height = image.size

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
        
    generator = torch.Generator(device=DEVICE).manual_seed(seed)

    result = pipe(prompt=input_text, image=image, mask_image=mask_image, width=width, height=height,
                  strength=strength, num_inference_steps=num_inference_steps, generator=generator,
                  guidance_scale=guidance_scale).images[0]

    return result, mask_image, seed

with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown(MARKDOWN)
    with gr.Row():
        with gr.Column(scale=1):
            input_image_component = gr.ImageEditor(
                label='Image',
                type='pil',
                sources=["upload", "webcam"],
                image_mode='RGB',
                layers=False,
                brush=gr.Brush(colors=["#FFFFFF"], color_mode="fixed"))
            input_text_component = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            with gr.Accordion("Advanced Settings", open=False):
                strength_slider = gr.Slider(
                    minimum=0.0,
                    maximum=1.0,
                    value=0.7,
                    step=0.01,
                    label="Strength"
                )
                num_inference_steps = gr.Slider(
                    minimum=1,
                    maximum=100,
                    value=30,
                    step=1,
                    label="Number of inference steps"
                )
                guidance_scale = gr.Slider(
                    label="Guidance Scale",
                    minimum=1,
                    maximum=15,
                    step=0.1,
                    value=3.5,
                )
                seed_number = gr.Number(
                    label="Seed", 
                    value=42,
                    precision=0
                )
                randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
            with gr.Accordion("Upload a mask", open=False):
                uploaded_mask_component = gr.Image(label="Already made mask (black pixels will be preserved, white pixels will be redrawn)", sources=["upload"], type="pil")
            submit_button_component = gr.Button(
                value='Inpaint', variant='primary')
        with gr.Column(scale=1):
            output_image_component = gr.Image(
                type='pil', image_mode='RGB', label='Generated Image')
            output_mask_component = gr.Image(
                type='pil', image_mode='RGB', label='Generated Mask')
            with gr.Accordion("Debug Info", open=False):
                output_seed = gr.Number(label="Used Seed")

    submit_button_component.click(
        fn=process,
        inputs=[input_image_component, input_text_component, strength_slider, seed_number, randomize_seed, num_inference_steps, guidance_scale],
        outputs=[output_image_component, output_mask_component, output_seed]
    )

demo.launch(debug=False, show_error=True)