fix: reduce error noise when converting speech to text

bot/linkai/link_ai_bot.py

@@ -47,10 +47,10 @@ def _chat(self, query, context, retry_count=0) -> Reply:
         :param retry_count: 当前递归重试次数
         :return: 回复
         """
-        if retry_count >= 2:
+        if retry_count > 2:
             # exit from retry 2 times
             logger.warn("[LINKAI] failed after maximum number of retry times")
-            return Reply(ReplyType.ERROR, "请再问我一次吧")
+            return Reply(ReplyType.TEXT, "请再问我一次吧")
 
         try:
             # load config
@@ -118,7 +118,7 @@ def _chat(self, query, context, retry_count=0) -> Reply:
                     logger.warn(f"[LINKAI] do retry, times={retry_count}")
                     return self._chat(query, context, retry_count + 1)
 
-                return Reply(ReplyType.ERROR, "提问太快啦，请休息一下再问我吧")
+                return Reply(ReplyType.TEXT, "提问太快啦，请休息一下再问我吧")
 
         except Exception as e:
             logger.exception(e)

bridge/bridge.py

@@ -18,7 +18,7 @@ def __init__(self):
             "text_to_voice": conf().get("text_to_voice", "google"),
             "translate": conf().get("translate", "baidu"),
         }
-        model_type = conf().get("model")
+        model_type = conf().get("model") or const.GPT35
         if model_type in ["text-davinci-003"]:
             self.btype["chat"] = const.OPEN_AI
         if conf().get("use_azure_chatgpt", False):

common/const.py

@@ -8,6 +8,7 @@
 CLAUDEAI = "claude"
 
 # model
+GPT35 = "gpt-3.5-turbo"
 GPT4 = "gpt-4"
 GPT4_TURBO_PREVIEW = "gpt-4-1106-preview"
 GPT4_VISION_PREVIEW = "gpt-4-vision-preview"

config-template.json

@@ -29,7 +29,7 @@
   "group_speech_recognition": false,
   "voice_reply_voice": false,
   "tts_voice_id": "alloy",
-  "conversation_max_tokens": 1000,
+  "conversation_max_tokens": 2500,
   "expires_in_seconds": 3600,
   "character_desc": "你是ChatGPT, 一个由OpenAI训练的大型语言模型, 你旨在回答并解决人们的任何问题，并且可以使用多种语言与人交流。",
   "temperature": 0.7,

plugins/godcmd/godcmd.py

@@ -266,14 +266,16 @@ def on_handle_context(self, e_context: EventContext):
                     if not isadmin and not self.is_admin_in_group(e_context["context"]):
                         ok, result = False, "需要管理员权限执行"
                     elif len(args) == 0:
-                        ok, result = True, "当前模型为: " + str(conf().get("model"))
+                        model = conf().get("model") or const.GPT35
+                        ok, result = True, "当前模型为: " + str(model)
                     elif len(args) == 1:
                         if args[0] not in const.MODEL_LIST:
                             ok, result = False, "模型名称不存在"
                         else:
                             conf()["model"] = self.model_mapping(args[0])
                             Bridge().reset_bot()
-                            ok, result = True, "模型设置为: " + str(conf().get("model"))
+                            model = conf().get("model") or const.GPT35
+                            ok, result = True, "模型设置为: " + str(model)
                 elif cmd == "id":
                     ok, result = True, user
                 elif cmd == "set_openai_api_key":

voice/linkai/linkai_voice.py

@@ -25,9 +25,12 @@ def voiceToText(self, voice_file):
             if not conf().get("text_to_voice") or conf().get("voice_to_text") == "openai":
                 model = const.WHISPER_1
             if voice_file.endswith(".amr"):
-                mp3_file = os.path.splitext(voice_file)[0] + ".mp3"
-                audio_convert.any_to_mp3(voice_file, mp3_file)
-                voice_file = mp3_file
+                try:
+                    mp3_file = os.path.splitext(voice_file)[0] + ".mp3"
+                    audio_convert.any_to_mp3(voice_file, mp3_file)
+                    voice_file = mp3_file
+                except Exception as e:
+                    logger.warn(f"[LinkVoice] amr file transfer failed, directly send amr voice file: {format(e)}")
             file = open(voice_file, "rb")
             file_body = {
                 "file": file
@@ -46,7 +49,7 @@ def voiceToText(self, voice_file):
             logger.info(f"[LinkVoice] voiceToText success, text={text}, file name={voice_file}")
         except Exception as e:
             logger.error(e)
-            reply = Reply(ReplyType.ERROR, "我暂时还无法听清您的语音，请稍后再试吧~")
+            return None
         return reply
 
     def textToVoice(self, text):
@@ -75,5 +78,5 @@ def textToVoice(self, text):
                 return None
         except Exception as e:
             logger.error(e)
-            reply = Reply(ReplyType.ERROR, "遇到了一点小问题，请稍后再问我吧")
-        return reply
+            # reply = Reply(ReplyType.ERROR, "遇到了一点小问题，请稍后再问我吧")
+            return None