Update app.py

app.py

@@ -39,6 +39,13 @@ class MyClient(discord.Client):
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
         self.is_processing = False
+        self.all_embeddings = None
+        self.initialize_embeddings()
+
+    def initialize_embeddings(self):
+        global all_datasets, model
+        text_data = all_datasets['text'].fillna('').astype(str).tolist()
+        self.all_embeddings = model.encode(text_data, convert_to_tensor=True)
 
     async def on_ready(self):
         logging.info(f'{self.user}로 로그인되었습니다!')
@@ -70,7 +77,7 @@ async def generate_response(message):
     user_mention = message.author.mention
     
     # 유사한 데이터 찾기
-    most_similar_data = find_most_similar_data(user_input)
+    most_similar_data = find_most_similar_data(user_input, message.client)
     
     system_message = f"{user_mention}, DISCORD에서 사용자들의 광고 카피라이팅 요청에 답하는 어시스턴트입니다."
     system_prefix = """
@@ -110,11 +117,10 @@ async def generate_response(message):
     conversation_history.append({"role": "assistant", "content": full_response_text})
     return f"{user_mention}, {full_response_text}"
 
-def find_most_similar_data(query):
+def find_most_similar_data(query, client):
     query_embedding = model.encode(query, convert_to_tensor=True)
-    all_embeddings = model.encode(all_datasets['text'].tolist(), convert_to_tensor=True)
     
-    cosine_scores = util.pytorch_cos_sim(query_embedding, all_embeddings)
+    cosine_scores = util.pytorch_cos_sim(query_embedding, client.all_embeddings)
     best_match_index = torch.argmax(cosine_scores).item()
     
     if cosine_scores[0][best_match_index] > 0.5:  # 유사도 임계값 설정