import gradio as gr
import torch
from PIL import Image
from transformers import AutoModelForCausalLM, AutoProcessor
from starvector.data.util import process_and_rasterize_svg

# Load model and processor
model = AutoModelForCausalLM.from_pretrained(
    "starvector/starvector-8b-im2svg", 
    trust_remote_code=True,
    torch_dtype=torch.float16
).cuda()
processor = AutoProcessor.from_pretrained("starvector/starvector-8b-im2svg")

def generate_svg(input_data, input_type):
    if input_type == "image":
        # Process image input
        image = processor(input_data, return_tensors="pt")['pixel_values'].cuda()
        raw_svg = model.generate_im2svg({"image": image}, max_length=4000)[0]
    else:
        # Process text input
        raw_svg = model.generate_text2svg(input_data, max_length=4000)[0]
        
    svg_code, raster_image = process_and_rasterize_svg(raw_svg)
    return svg_code, raster_image

with gr.Blocks() as demo:
    gr.Markdown("# 💫 StarVector SVG Generator")
    
    with gr.Tab("Image to SVG"):
        gr.Markdown("Upload an image to convert to SVG")
        with gr.Row():
            image_input = gr.Image(type="pil", label="Input Image")
            image_output = gr.Image(label="SVG Preview")
        svg_code = gr.Code(label="Generated SVG Code")
        image_button = gr.Button("Convert to SVG")

    with gr.Tab("Text to SVG"):
        gr.Markdown("Enter text to generate SVG")
        with gr.Row():
            text_input = gr.Textbox(label="Text Prompt")
            text_output = gr.Image(label="SVG Preview")
        text_svg_code = gr.Code(label="Generated SVG Code")
        text_button = gr.Button("Generate SVG")

    image_button.click(
        fn=lambda x: generate_svg(x, "image"),
        inputs=image_input,
        outputs=[svg_code, image_output]
    )
    text_button.click(
        fn=lambda x: generate_svg(x, "text"),
        inputs=text_input,
        outputs=[text_svg_code, text_output]
    )

demo.launch()