Fix ZeroGPU recursion: split click handler from GPU function

app.py

@@ -130,11 +130,22 @@ def segment_with_text(image: np.ndarray, text_prompt: str):
         return image, None, f"❌ Error: {e}"
 
 
+def handle_click(image, evt: gr.SelectData):
+    """Handle click event and extract coordinates"""
+    if image is None or evt is None:
+        return None, None, None, "❌ Click on an image first"
+    # Store coordinates and pass to GPU function
+    x, y = evt.index[0], evt.index[1]
+    return image, x, y, "Processing..."
+
+
 @spaces.GPU(duration=60)
-def segment_with_click(image: np.ndarray, evt: gr.SelectData):
-    """Segment object at clicked point with SAM3"""
+def segment_with_point(image: np.ndarray, x: int, y: int):
+    """Segment object at point with SAM3"""
     if image is None:
         return None, None, "❌ No image provided"
+    if x is None or y is None:
+        return None, None, "❌ No point selected"
 
     try:
         from PIL import Image as PILImage
@@ -146,12 +157,9 @@ def segment_with_click(image: np.ndarray, evt: gr.SelectData):
         else:
             pil_image = image
 
-        # Get click coordinates
-        point = [evt.index[0], evt.index[1]]
-
         # Run SAM3 with point prompt
         state = processor.set_image(pil_image)
-        output = processor.set_point_prompt(state=state, points=[point], labels=[1])
+        output = processor.set_point_prompt(state=state, points=[[x, y]], labels=[1])
 
         if output is None or "masks" not in output:
             return image, None, "⚠️ No object found"
@@ -243,6 +251,8 @@ with gr.Blocks(title="SAM 3D Objects MCP") as demo:
     """)
 
     mask_state = gr.State(None)
+    click_x = gr.State(None)
+    click_y = gr.State(None)
 
     with gr.Row():
         with gr.Column():
@@ -278,9 +288,14 @@ with gr.Blocks(title="SAM 3D Objects MCP") as demo:
         outputs=[preview, mask_state, status]
     )
 
+    # Click handler: first extract coordinates (no GPU), then segment (GPU)
     input_image.select(
-        segment_with_click,
+        handle_click,
         inputs=[input_image],
+        outputs=[input_image, click_x, click_y, status]
+    ).then(
+        segment_with_point,
+        inputs=[input_image, click_x, click_y],
         outputs=[preview, mask_state, status]
     )