app_simple.py

import json
import ast
from PIL import Image, ImageDraw, ImageFont
from openai import OpenAI
import os
import base64
import gradio as gr
import tempfile

def plot_bounding_boxes(image, bounding_boxes):
    """Simple bounding box plotter."""
    if isinstance(image, str):
        img = Image.open(image)
    else:
        img = image.copy()
    
    width, height = img.size
    draw = ImageDraw.Draw(img)
    
    # Parse JSON
    lines = bounding_boxes.splitlines()
    for i, line in enumerate(lines):
        if line == "```json":
            bounding_boxes = "\n".join(lines[i+1:])
            bounding_boxes = bounding_boxes.split("```")[0]
            break
    
    # Try to load font with CJK support
    font = None
    font_paths = [
        # Noto CJK fonts (installed via packages.txt)
        "/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
        "/usr/share/fonts/truetype/noto-cjk/NotoSansCJK-Regular.ttc",
        # Fallback fonts
        "/usr/share/fonts/truetype/liberation/LiberationSans-Regular.ttf",
        "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
    ]
    
    for font_path in font_paths:
        try:
            font = ImageFont.truetype(font_path, size=14)
            break
        except:
            continue
    
    if font is None:
        font = ImageFont.load_default()
    
    # Parse and plot
    try:
        bbox_list = ast.literal_eval(bounding_boxes)
        if not isinstance(bbox_list, list):
            bbox_list = [bbox_list]
            
        for bbox in bbox_list:
            coords = bbox.get("bbox_2d", [])
            text = bbox.get("text_content", "")
            if len(coords) < 4:
                continue
                
            x1, y1, x2, y2 = coords[0], coords[1], coords[2], coords[3]
            
            # Ensure order
            if x1 > x2:
                x1, x2 = x2, x1
            if y1 > y2:
                y1, y2 = y2, y1
            
            # Clamp to image bounds
            x1 = max(0, min(x1, width - 1))
            y1 = max(0, min(y1, height - 1))
            x2 = max(0, min(x2, width - 1))
            y2 = max(0, min(y2, height - 1))
            
            # Draw box
            draw.rectangle(((x1, y1), (x2, y2)), outline='green', width=2)
            
            # Draw text label if we have font
            if text and font:
                text_x = x1 + 2
                text_y = y2 + 2
                draw.text((text_x, text_y), text, fill='green', font=font)
            
    except Exception as e:
        print(f"Error plotting boxes: {e}")
    
    return img


def process_image(image, prompt):
    """Process image using API."""
    
    if image is None:
        return None, "Please upload an image"
    
    # Get API key from environment variable (HF Secret)
    api_key = os.environ.get("QWEN_API_KEY")
    if not api_key:
        return None, "API key not configured. Please set QWEN_API_KEY in Space secrets."
    
    # Fixed model
    model = "qwen2.5-vl-7b-instruct"
    
    try:
        # Save image temporarily
        with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as tmp:
            image.save(tmp.name, format='JPEG', quality=95)
            temp_path = tmp.name
        
        # Encode image
        with open(temp_path, "rb") as image_file:
            base64_image = base64.b64encode(image_file.read()).decode("utf-8")
        
        # Call API
        client = OpenAI(
            api_key=api_key,
            base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
        )
        
        messages = [
            {
                "role": "system",
                "content": [{"type": "text", "text": "You are a helpful assistant."}]
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "image_url",
                        "image_url": {"url": f"data:image/jpeg;base64,{base64_image}"}
                    },
                    {"type": "text", "text": prompt}
                ]
            }
        ]
        
        completion = client.chat.completions.create(
            model=model,
            messages=messages,
        )
        
        response = completion.choices[0].message.content
        
        # Plot boxes
        annotated_image = plot_bounding_boxes(image, response)
        
        # Clean up
        os.unlink(temp_path)
        
        return annotated_image, response
        
    except Exception as e:
        return None, f"Error: {str(e)}"


# Create interface  
with gr.Blocks(title="安全なう - 行為規制") as demo:
    # Simple text-based logo
    gr.Markdown(
        """
        <h2 style='color: #15803d; font-family: monospace; margin: 10px 0;'>
        ◆ <span style='background: #15803d; color: white; padding: 2px 6px;'>ANZEN</span><span style='color: #15803d;'>NOW</span>
        </h2>
        """
    )
    gr.Markdown("# 行為規制違反の検出")
    gr.Markdown("画像をアップロードしますと、行為規制違反の検出を行います。")
    
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(type="pil", label="Upload Image")
            prompt_input = gr.Textbox(
                value="Spotting all the text in the image with line-level, and output in JSON format.",
                label="Prompt",
                lines=2
            )
            submit_btn = gr.Button("🔍 行為規制違反事項を検出", variant="primary", size="lg")
        
        with gr.Column():
            output_image = gr.Image(label="Annotated Result")
            output_text = gr.Textbox(label="JSON Output", lines=12)
    
    submit_btn.click(
        fn=process_image,
        inputs=[image_input, prompt_input],
        outputs=[output_image, output_text]
    )

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)