"""
Visual Question Answering with Hugging Face API
"""

import gradio as gr
import requests
from PIL import Image
import io
import base64
import os

# Your Hugging Face token
HF_TOKEN = os.environ.get("HF_TOKEN", "")

def answer_question(image, question):
    if image is None:
        return "Please upload an image first."
    if not question.strip():
        return "Please ask a question."
    
    try:
        # Convert image to base64
        buffered = io.BytesIO()
        image.save(buffered, format="PNG")
        img_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
        
        # API request
        headers = {
            "Authorization": f"Bearer {HF_TOKEN}",
            "Content-Type": "application/json"
        }
        
        payload = {
            "inputs": {
                "image": img_base64,
                "question": question
            }
        }
        
        response = requests.post(
            "https://api-inference.huggingface.co/models/Salesforce/blip-vqa-base",
            headers=headers,
            json=payload,
            timeout=60
        )
        
        if response.status_code == 200:
            result = response.json()
            if isinstance(result, list):
                return result[0].get("answer", "No answer")
            return result.get("answer", "No answer")
        else:
            return f"Model is loading (or error). Please try again. Error: {response.status_code}"
    
    except Exception as e:
        return f"Error: {str(e)[:100]}"

with gr.Blocks() as demo:
    gr.Markdown("# 🖼️ Visual Question Answering (API)")
    
    with gr.Row():
        with gr.Column():
            img = gr.Image(type="pil", label="Upload Image", height=300)
            q = gr.Textbox(label="Question", placeholder="What's in this image?")
            btn = gr.Button("Ask", variant="primary")
        with gr.Column():
            out = gr.Textbox(label="Answer", lines=4)
    
    btn.click(answer_question, [img, q], out)

demo.launch()