update: 增加了智谱大模型，增加了修改max_tokens接口

app/abstract.py

@@ -5,7 +5,7 @@ class ASR(ABC):
     def __init__(self):
         self.is_slience = False
     @abstractmethod
-    async def stream_recognize(self, chunk):
+    async def stream_recognize(self, assistant, chunk):
         pass
 
 class LLM(ABC):