import gradio as gr
import torch
import numpy as np
from PIL import Image
from transformers import AutoProcessor, AutoModelForVision2Seq

# Load the model & processor
model_name = "Murasajo/Llama-3.2-VL-Finetuned-on-HandwrittenText"
processor = AutoProcessor.from_pretrained(model_name)
model = AutoModelForVision2Seq.from_pretrained(model_name)

# Function to extract handwritten text
def extract_text(image):
    try:
        # Convert input to PIL Image
        if isinstance(image, np.ndarray):  
            if len(image.shape) == 2:  # If grayscale (H, W), add channels
                image = np.stack([image] * 3, axis=-1)
            image = Image.fromarray(image)
        else:
            image = Image.open(image).convert("RGB")

        # Process image through model
        pixel_values = processor(images=image, return_tensors="pt").pixel_values
        generated_ids = model.generate(pixel_values)
        extracted_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

        return extracted_text

    except Exception as e:
        return f"Error: {str(e)}"

# Gradio Interface
iface = gr.Interface(
    fn=extract_text,
    inputs="image",
    outputs="text",
    title="Handwritten Text OCR",
    description="Upload a handwritten document and extract text using AI.",
)

# Run the app
iface.launch()