Update app.py

app.py

@@ -24,8 +24,7 @@ intents.guilds = True
 intents.guild_messages = True
 
 # 추론 API 클라이언트 설정
-hf_client_secondary = InferenceClient("CohereForAI/aya-23-35B", token=os.getenv("HF_TOKEN"))
-hf_client_primary = InferenceClient("CohereForAI/c4ai-command-r-plus", token=os.getenv("HF_TOKEN"))
+hf_client = InferenceClient("CohereForAI/c4ai-command-r-plus", token=os.getenv("HF_TOKEN"))
 
 # 수학 전문 LLM 파이프라인 설정
 math_pipe = pipeline("text-generation", model="AI-MO/NuminaMath-7B-TIR")
@@ -71,59 +70,65 @@ class MyClient(discord.Client):
         super().__init__(*args, **kwargs)
         self.is_processing = False
         self.math_pipe = math_pipe
-        self.current_client = "primary"
-        self.hf_client = hf_client_primary
-
-    def switch_client(self):
-        if self.current_client == "primary":
-            self.hf_client = hf_client_secondary
-            self.current_client = "secondary"
-            logging.info("Switched to secondary client (CohereForAI/aya-23-35B).")
-        else:
-            self.hf_client = hf_client_primary
-            self.current_client = "primary"
-            logging.info("Switched back to primary client (CohereForAI/c4ai-command-r-plus).")
-
-    async def retry_request(self, func, retries=5, delay=2):
-        for i in range(retries):
-            try:
-                return await func()
-            except Exception as e:
-                logging.error(f"Attempt {i+1}/{retries}: Error encountered: {type(e).__name__}: {str(e)}")
-                if isinstance(e, HTTPError) and getattr(e.response, 'status_code', None) == 503:
-                    logging.warning(f"503 error encountered. Switching client and retrying in {delay} seconds...")
-                    self.switch_client()
-                elif i < retries - 1:
-                    logging.warning(f"Error occurred. Retrying in {delay} seconds...")
-                await asyncio.sleep(delay)
-        
-        logging.error(f"All {retries} attempts failed.")
-        raise Exception("Max retries reached")
+
+    async def on_ready(self):
+        logging.info(f'{self.user}로 로그인되었습니다!')
+        subprocess.Popen(["python", "web.py"])
+        logging.info("Web.py server has been started.")
+
+    async def on_message(self, message):
+        if message.author == self.user:
+            return
+        if not self.is_message_in_specific_channel(message):
+            return
+        if self.is_processing:
+            return
+
+        self.is_processing = True
+        try:
+            # 새로운 스레드 생성
+            thread = await message.channel.create_thread(name=f"질문: {message.author.name}", message=message)
+            if self.is_math_question(message.content):
+                text_response = await self.handle_math_question(message.content)
+                await self.send_message_with_latex(thread, text_response)
+            else:
+                response = await self.generate_response(message)
+                await self.send_message_with_latex(thread, response)
+        finally:
+            self.is_processing = False
+
+    def is_message_in_specific_channel(self, message):
+        return message.channel.id == SPECIFIC_CHANNEL_ID or (
+            isinstance(message.channel, discord.Thread) and message.channel.parent_id == SPECIFIC_CHANNEL_ID
+        )
+
+    def is_math_question(self, content):
+        return bool(re.search(r'\b(solve|equation|calculate|math)\b', content, re.IGNORECASE))
 
     async def handle_math_question(self, question):
         loop = asyncio.get_event_loop()
         
+        # AI-MO/NuminaMath-7B-TIR 모델에게 수학 문제를 풀도록 요청
         math_response_future = loop.run_in_executor(None, lambda: self.math_pipe(question, max_new_tokens=2000))
         math_response = await math_response_future
         math_result = math_response[0]['generated_text']
         
         try:
-            cohere_response = await self.retry_request(lambda: self.hf_client.chat_completion(
+            # Cohere 모델에게 AI-MO/NuminaMath-7B-TIR 모델의 결과를 번역하도록 요청
+            cohere_response_future = loop.run_in_executor(None, lambda: hf_client.chat_completion(
                 [{"role": "system", "content": "다음 텍스트를 한글로 번역하십시오: "}, {"role": "user", "content": math_result}], max_tokens=1000))
         
+            cohere_response = await cohere_response_future
             cohere_result = ''.join([part.choices[0].delta.content for part in cohere_response if part.choices and part.choices[0].delta and part.choices[0].delta.content])
 
             combined_response = f"수학 선생님 답변: ```{cohere_result}```"
 
-        except Exception as e:
-            logging.error(f"Error in handle_math_question: {type(e).__name__}: {str(e)}")
+        except HTTPError as e:
+            logging.error(f"Hugging Face API error: {e}")
             combined_response = "An error occurred while processing the request."
 
         return combined_response
 
-
-
-
     async def generate_response(self, message):
         global conversation_history
         user_input = message.content
@@ -141,12 +146,12 @@ class MyClient(discord.Client):
         messages = [{"role": "system", "content": f"{system_prefix}"}] + conversation_history
 
         try:
-            response = await self.retry_request(lambda: self.hf_client.chat_completion(
+            response = await asyncio.get_event_loop().run_in_executor(None, lambda: hf_client.chat_completion(
                 messages, max_tokens=1000, stream=True, temperature=0.7, top_p=0.85))
             full_response = ''.join([part.choices[0].delta.content for part in response if part.choices and part.choices[0].delta and part.choices[0].delta.content])
             conversation_history.append({"role": "assistant", "content": full_response})
-        except Exception as e:
-            logging.error(f"Error in generate_response: {type(e).__name__}: {str(e)}")
+        except HTTPError as e:
+            logging.error(f"Hugging Face API error: {e}")
             full_response = "An error occurred while generating the response."
 
         return f"{user_mention}, {full_response}"
@@ -180,30 +185,6 @@ class MyClient(discord.Client):
             for part in parts:
                 await channel.send(part)
 
-    def switch_client(self):
-        if self.hf_client == hf_client_primary:
-            self.hf_client = hf_client_secondary
-            logging.info("Switched to secondary client (CohereForAI/aya-23-35B).")
-        else:
-            self.hf_client = hf_client_primary
-            logging.info("Switched back to primary client (CohereForAI/c4ai-command-r-plus).")
-
-    async def retry_request(self, func, retries=5, delay=2):
-        for i in range(retries):
-            try:
-                return await func()
-            except Exception as e:
-                logging.error(f"Error encountered: {type(e).__name__}: {str(e)}")
-                if isinstance(e, HTTPError) and e.response.status_code == 503:
-                    logging.warning(f"503 error encountered. Retrying in {delay} seconds...")
-                    self.switch_client()  # 클라이언트 전환
-                    await asyncio.sleep(delay)
-                elif i < retries - 1:
-                    logging.warning(f"Error occurred. Retrying in {delay} seconds...")
-                    await asyncio.sleep(delay)
-                else:
-                    raise
-
 if __name__ == "__main__":
     discord_client = MyClient(intents=intents)
-    discord_client.run(os.getenv('DISCORD_TOKEN'))
+    discord_client.run(os.getenv('DISCORD_TOKEN'))