app.py

import gradio as gr
import requests
import os

# Function to query the Qwen model for OCR
def query_ocr(image):
    # Get the Hugging Face access token from secrets
    hf_token = os.getenv("secret")
    headers = {"Authorization": f"Bearer {hf_token}"}

    # API endpoint for the model
    API_URL = "/static-proxy?url=https%3A%2F%2Fapi-inference.huggingface.co%2Fmodels%2FQwen%2FQwen-VL%26quot%3B%3C%2Fspan%3E%3C!-- HTML_TAG_END -->

    # Prepare the request
    files = {"file": image}
    response = requests.post(API_URL, headers=headers, files=files)

    # Raise an error if the request fails
    response.raise_for_status()
    
    return response.json()  # Assuming the model returns a JSON with the extracted text

# Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# OCR Application using Qwen Model")
    gr.Markdown("Upload an image to extract text using OCR.")

    image_input = gr.Image(type="file", label="Upload Image")
    output_text = gr.Textbox(label="Extracted Text", lines=10)

    submit_btn = gr.Button("Extract Text")
    
    submit_btn.click(fn=query_ocr, inputs=image_input, outputs=output_text)

# Launch the app
demo.launch()