Fix the issue of Arch-Intent

model_server/app/commons/constants.py

@@ -24,11 +24,7 @@ Provide your tool assessment for ONLY THE LAST USER MESSAGE in the above convers
 
 
 ARCH_INTENT_GENERATION_CONFIG = {
-    "generation_params": {
-        "stop_token_ids": [151645],
-        "max_tokens": 1,
-        "guided_choice": ["Yes", "No"],
-    }
+    "generation_params": {"max_tokens": 1, "stop_token_ids": [151645]}
 }
 
 

model_server/app/model_handler/base_handler.py

@@ -28,8 +28,8 @@ class ChatCompletionResponse(BaseModel):
     id: Optional[int] = 0
     object: Optional[str] = "chat_completion"
     created: Optional[str] = ""
-    model: str
     choices: List[Choice]
+    model: str
 
 
 class ArchBaseHandler:
@@ -124,7 +124,7 @@ class ArchBaseHandler:
 
         if tools:
             processed_messages.append(
-                {"role": "system", "content": self._format_system(tools)}
+                {"role": "system", "content": self._format_system_prompt(tools)}
             )
 
         for message in messages:

model_server/app/model_handler/function_calling.py

@@ -108,7 +108,9 @@ class ArchIntentHandler(ArchBaseHandler):
             extra_body=self.generation_params,
         )
 
-        model_response = Message(content=model_response, tool_calls=[])
+        model_response = Message(
+            content=model_response.choices[0].message.content, tool_calls=[]
+        )
 
         chat_completion_response = ChatCompletionResponse(
             choices=[Choice(message=model_response)], model=self.model_name

model_server/app/tests/test_function_calling.py

@@ -86,7 +86,7 @@ async def test_chat_completion(mock_hanlder, mock_client):
     mock_client.chat.completions.create.return_value = mock_response
 
     # Mock the tool formatter
-    mock_hanlder._format_system.return_value = "<formatted_tools>"
+    mock_hanlder._format_system_prompt.return_value = "<formatted_tools>"
 
     response = Response()
     chat_response = await chat_completion(request, response)