Update app.py

app.py

@@ -126,11 +126,17 @@ def convert_to_entities(text, coords):
 
 
 @spaces.GPU(duration=30)
-def answer_question(img, prompt):
+def answer_question(img, prompt, reasoning):
     buffer = ""
+    resp = moondream.query(img, prompt, stream=True, reasoning=reasoning)
+    reasoning_text = resp["reasoning"]["text"] if reasoning else "[reasoning disabled]"
+    entities = [
+        {"start": g["start_idx"], "end": g["end_idx"], "entity": json.dumps(g["points"])}
+        for g in resp["reasoning"]["grounding"]
+    ] if reasoning else []
     for new_text in moondream.query(img, prompt, stream=True)["answer"]:
         buffer += new_text
-        yield buffer.strip(), {"text": "Thinking...", "entities": []}
+        yield buffer.strip(), {"text": reasoning_text, "entities": entities}
 
 
 @spaces.GPU(duration=10)
@@ -272,10 +278,12 @@ with gr.Blocks(title="moondream vl (new)", css=css, js=js) as demo:
                                 scale=4,
                             )
                             submit = gr.Button("Submit")
+                            reasoning = gr.Checkbox(label="Enable reasoning")
                         img = gr.Image(type="pil", label="Upload an Image")
-                    submit.click(answer_question, [img, prompt], [output, thought])
-                    prompt.submit(answer_question, [img, prompt], [output, thought])
-                    img.change(answer_question, [img, prompt], [output, thought])
+                    submit.click(answer_question, [img, prompt, reasoning], [output, thought])
+                    prompt.submit(answer_question, [img, prompt, reasoning], [output, thought])
+                    reasoning.change(answer_question, [img, prompt, reasoning], [output, thought])
+                    img.change(answer_question, [img, prompt, reasoning], [output, thought])
                     img.change(lambda img: img, [img], [input_image])
                 elif mode == "Caption":
                     with gr.Group():