Update app.py

app.py

@@ -30,20 +30,25 @@ def chat_with_gemma(user_input, temperature, top_p, frequency_penalty, presence_
 
     return response["choices"][0]["message"]["content"].strip()
 
-# Set up the Gradio interface
-demo = gr.Interface(
-    fn=chat_with_gemma,
-    inputs=[
-        gr.Textbox(label="Enter your message to Gemma"),
-        gr.Slider(0.0, 1.5, value=0.7, step=0.05, label="Temperature"),
-        gr.Slider(0.0, 1.0, value=0.9, step=0.05, label="Top-p (Nucleus Sampling)"),
-        gr.Slider(0.0, 2.0, value=0.4, step=0.1, label="Frequency Penalty"),
-        gr.Slider(0.0, 2.0, value=0.2, step=0.1, label="Presence Penalty")
-    ],
-    outputs=gr.Textbox(label="Gemma's Response", lines=8),
-    title="Talk to Gemma",
-    description="Generate short responses using Google's Gemma model with adjustable settings. If you want to use this space via api, duplicate it, then look at the end of the app.py file to find a commented out line that you can renable to turn on api useage and protect it with a username and password so trolls don't ruin your api :)"
-)
+with gr.Blocks() as demo:
+    user_input = gr.Textbox(label="Enter your message to Gemma")
+    temperature = gr.Slider(0.0, 1.5, value=0.7, step=0.05, label="Temperature")
+    top_p = gr.Slider(0.0, 1.0, value=0.9, step=0.05, label="Top-p (Nucleus Sampling)")
+    freq_penalty = gr.Slider(0.0, 2.0, value=0.4, step=0.1, label="Frequency Penalty")
+    pres_penalty = gr.Slider(0.0, 2.0, value=0.2, step=0.1, label="Presence Penalty")
+    output = gr.Textbox(label="Gemma's Response", lines=8)
+
+    submit_button = gr.Button("Submit")
+
+    submit_button.click(
+        chat_with_gemma,
+        inputs=[user_input, temperature, top_p, freq_penalty, pres_penalty],
+        outputs=output,
+        api_name=False  # <---- This disables API for this endpoint
+    )
+
+demo.launch(share=True)
+
 
 # Launch the ap
 demo.launch(share=True, enable_api=False)