Update app.py

app.py

@@ -4,7 +4,6 @@ import os
 from threading import Event
 
 hf_token = os.getenv("HF_TOKEN")
-
 stop_event = Event()
 
 def get_model_response(client, messages, max_tokens, temperature, top_p):
@@ -24,7 +23,6 @@ def get_model_response(client, messages, max_tokens, temperature, top_p):
             if token:
                 full_response += token
                 yield full_response
-        stop_event.clear()
     except Exception as e:
         yield f"모델 추론 실패: {str(e)}"
 
@@ -40,18 +38,16 @@ def respond(message, history, system_message, max_tokens, temperature, top_p, se
         response = ""
         for partial_response in get_model_response(client, messages, max_tokens, temperature, top_p):
             response = partial_response
-            yield "", history + [(message, response)]
+            history.append((message, response))
+            yield "", history
         
     except Exception as e:
-        yield "", history + [(message, f"오류 발생: {str(e)}")]
-
-def stop_generation():
-    stop_event.set()
-    return "생성이 중단되었습니다."
+        history.append((message, f"오류 발생: {str(e)}"))
+        yield "", history
 
 def continue_writing(history, system_message, max_tokens, temperature, top_p, model):
     if not history:
-        return history
+        return "", history
     last_user_message = history[-1][0]
     last_assistant_message = history[-1][1]
     
@@ -63,13 +59,20 @@ 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 = get_model_response(client, messages, max_tokens, temperature, top_p)
-        continued_response = last_assistant_message + " " + next(response)
+        response = ""
+        for partial_response in get_model_response(client, messages, max_tokens, temperature, top_p):
+            response = partial_response
         
-        new_history = history[:-1] + [(last_user_message, continued_response)]
-        return new_history
+        continued_response = last_assistant_message + " " + response
+        history[-1] = (last_user_message, continued_response)
+        return "", history
     except Exception as e:
-        return history + [("시스템", f"계속 작성 중 오류 발생: {str(e)}")]    
+        history.append(("시스템", f"계속 작성 중 오류 발생: {str(e)}"))
+        return "", history
+
+def stop_generation():
+    stop_event.set()
+    return "생성이 중단되었습니다."
 
 models = {
     "deepseek-ai/DeepSeek-Coder-V2-Instruct": "(한국회사)DeepSeek-Coder-V2-Instruct",
@@ -110,23 +113,27 @@ with gr.Blocks() as demo:
         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.js = """
-    function(e) {
-        if (e.key === 'Enter' && !e.shiftKey) {
-            e.preventDefault();
-            return;
-        }
-        if (e.key === 'Enter' && e.shiftKey) {
-            e.preventDefault();
-            document.getElementById('component-3').click();
-        }
+    msg.javascript = """
+    (x) => {
+        const textbox = document.querySelector("#component-3");
+        textbox.addEventListener("keydown", function(e) {
+            if (e.key === 'Enter' && !e.shiftKey) {
+                e.preventDefault();
+                return;
+            }
+            if (e.key === 'Enter' && e.shiftKey) {
+                e.preventDefault();
+                document.querySelector("#component-5").click();
+            }
+        });
+        return x;
     }
     """
 
     # 이벤트 핸들러 설정
-    msg.submit(lambda x: "", inputs=msg, outputs=msg)  # 엔터 키 입력 시 아무 동작 안 함
     send.click(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=[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])
     regenerate.click(lambda h, s, m, t, p, mod: respond(h[-1][0] if h else "", h[:-1], s, m, t, p, mod), inputs=[chatbot, system_message, max_tokens, temperature, top_p, model], outputs=[msg, chatbot])
     stop.click(stop_generation, inputs=[], outputs=[msg])
     clear.click(lambda: (None, None), outputs=[msg, chatbot])