Update app.py

app.py

@@ -19,7 +19,10 @@ def get_model_response(client, messages, max_tokens, temperature, top_p):
         for message in response:
             if stop_event.is_set():
                 break
-            token = message.choices[0].delta.content if hasattr(message.choices[0], 'delta') else message.choices[0].text
+            if hasattr(message.choices[0], 'delta'):
+                token = message.choices[0].delta.content
+            else:
+                token = message.choices[0].text
             if token:
                 full_response += token
                 yield full_response
@@ -35,10 +38,9 @@ def respond(message, history, system_message, max_tokens, temperature, top_p, se
         messages.extend([{"role": "user" if i % 2 == 0 else "assistant", "content": m} for h in history for i, m in enumerate(h) if m])
         messages.append({"role": "user", "content": message})
         
-        response = ""
-        for partial_response in get_model_response(client, messages, max_tokens, temperature, top_p):
-            response = partial_response
-            history.append((message, response))
+        history.append((message, ""))
+        for response in get_model_response(client, messages, max_tokens, temperature, top_p):
+            history[-1] = (message, response)
             yield "", history
         
     except Exception as e:
@@ -59,16 +61,13 @@ def continue_writing(history, system_message, max_tokens, temperature, top_p, mo
         messages.extend([{"role": "user" if i % 2 == 0 else "assistant", "content": m} for h in history for i, m in enumerate(h)])
         messages.append({"role": "user", "content": prompt})
         
-        response = ""
-        for partial_response in get_model_response(client, messages, max_tokens, temperature, top_p):
-            response = partial_response
-        
-        continued_response = last_assistant_message + " " + response
-        history[-1] = (last_user_message, continued_response)
-        return "", history
+        for response in get_model_response(client, messages, max_tokens, temperature, top_p):
+            continued_response = last_assistant_message + " " + response
+            history[-1] = (last_user_message, continued_response)
+            yield "", history
     except Exception as e:
         history.append(("시스템", f"계속 작성 중 오류 발생: {str(e)}"))
-        return "", history
+        yield "", history
 
 def stop_generation():
     stop_event.set()
@@ -112,7 +111,6 @@ with gr.Blocks() as demo:
         top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.90, step=0.05, label="Top-p (핵 샘플링)")
         model = gr.Radio(list(models.keys()), value=list(models.keys())[0], label="언어 모델 선택", info="사용할 언어 모델을 선택하세요")
 
-    # JavaScript를 사용하여 키 입력 동작 커스터마이즈
     msg.javascript = """
     (x) => {
         const textbox = document.querySelector("#component-3");
@@ -130,7 +128,6 @@ with gr.Blocks() as demo:
     }
     """
 
-    # 이벤트 핸들러 설정
     send.click(respond, inputs=[msg, chatbot, system_message, max_tokens, temperature, top_p, model], outputs=[msg, chatbot])
     msg.submit(respond, inputs=[msg, chatbot, system_message, max_tokens, temperature, top_p, model], outputs=[msg, chatbot])
     continue_btn.click(continue_writing, inputs=[chatbot, system_message, max_tokens, temperature, top_p, model], outputs=[msg, chatbot])