Update app.py

app.py

@@ -206,8 +206,18 @@ def point(img, object):
     )
 
 @spaces.GPU(duration=10)
-def localized_query(ann_image, question):
-    yield repr(ann_image), {"text": "..."}
+def localized_query(img, x, y, question):
+    w, h = img.size
+    x, y = x * w, y * h
+    img_clone = img.copy()
+    draw = ImageDraw.Draw(img_clone)
+    draw.ellipse(
+        (x - 3, y - 3, x + 3, y + 3),
+        fill="red",
+        outline="red",
+    )
+    
+    yield "", {"text": "..."}, gr.update(visible=True, value=img_clone)
 
 
 js = ""
@@ -328,8 +338,11 @@ with gr.Blocks(title="moondream vl (new)", css=css, js=js) as demo:
                         img = gr.Image(type="pil", label="Upload an Image")
                         x_slider = gr.Slider(label="x", minimum=0, maximum=1)
                         y_slider = gr.Slider(label="y", minimum=0, maximum=1)
-                    submit.click(localized_query, [img, prompt], [output, thought])
-                    prompt.submit(localized_query, [img, prompt], [output, thought])
+                    submit.click(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
+                    prompt.submit(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
+                    x_slider.change(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
+                    y_slider.change(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
+                    img.change(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
                 else:
                     gr.Markdown("Coming soon!")