Update app.py

app.py

@@ -42,7 +42,7 @@ async def respond(
         ):
             token = message.choices[0].text
             response += token
-            yield response
+        return response
     except APIError as e:
         error_details = e.body
         error_type = error_details.get("type")
@@ -56,33 +56,36 @@ async def respond(
             error_str = "An error occurred during streaming"
 
         print(f"Error: {error_str}")
-        yield error_str
+        return error_str
     except Exception as e:
         print(f"Error: {e}")
-        yield "Error occurred. Please try again."
+        return "Error occurred. Please try again."
 
 def launch_app():
     try:
-        demo = gr.Interface(
-            respond,
-            [
-                gr.Textbox(label="Message"),
-                gr.Dataframe(label="History", headers=["User", "Assistant"]),
-                gr.Textbox(label="System message"),
-                gr.Slider(minimum=1, maximum=2048, value=2048, step=1, label="Max new tokens"),
-                gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-                gr.Slider(
-                    minimum=0.1,
-                    maximum=1.0,
-                    value=0.95,
-                    step=0.05,
-                    label="Top-P",
-                ),
-            ],
-            gr.Text(label="Response"),
-            title="Chatbot",
-            description="A chatbot that responds to user input",
-        )
+        demo = gr.Blocks()
+        with demo:
+            gr.Markdown("# Chatbot")
+            message = gr.Textbox(label="Message")
+            history = gr.State([["", ""]])
+            system_message = gr.Textbox(label="System message")
+            max_tokens = gr.Slider(minimum=1, maximum=2048, value=2048, step=1, label="Max new tokens")
+            temperature = gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature")
+            top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-P")
+            response = gr.Text(label="Response")
+
+            def generate_response(message, history, system_message, max_tokens, temperature, top_p):
+                new_history = history + [[message, ""]]
+                response = asyncio.run(respond(message, history, system_message, max_tokens, temperature, top_p))
+                new_history[-1][1] = response
+                return response, new_history
+
+            gr.Button("Generate Response").click(
+                generate_response,
+                inputs=[message, history, system_message, max_tokens, temperature, top_p],
+                outputs=[response, history],
+                show_progress=False,
+            )
         demo.launch(show_error=True)
     except KeyError as e:
         print(f"Error: {e}")