Validate llm_config passed to ConversableAgent (issue #1522)

autogen/agentchat/assistant_agent.py

@@ -33,7 +33,7 @@ def __init__(
         self,
         name: str,
         system_message: Optional[str] = DEFAULT_SYSTEM_MESSAGE,
-        llm_config: Optional[Union[Dict, Literal[False]]] = None,
+        llm_config: Optional[Union[Dict, Literal[False]]] = False,
         is_termination_msg: Optional[Callable[[Dict], bool]] = None,
         max_consecutive_auto_reply: Optional[int] = None,
         human_input_mode: Optional[str] = "NEVER",

autogen/agentchat/conversable_agent.py

@@ -78,8 +78,8 @@ def __init__(
         human_input_mode: Optional[str] = "TERMINATE",
         function_map: Optional[Dict[str, Callable]] = None,
         code_execution_config: Union[Dict, Literal[False]] = False,
-        llm_config: Optional[Union[Dict, Literal[False]]] = None,
-        default_auto_reply: Optional[Union[str, Dict, None]] = "",
+        llm_config: Optional[Union[Dict, Literal[False]]] = False,
+        default_auto_reply: Union[str, Dict] = "",
         description: Optional[str] = None,
     ):
         """
@@ -121,7 +121,7 @@ def __init__(
                 Please refer to [OpenAIWrapper.create](/docs/reference/oai/client#create)
                 for available options.
                 To disable llm-based auto reply, set to False.
-            default_auto_reply (str or dict or None): default auto reply when no code execution or llm-based reply is generated.
+            default_auto_reply (str or dict): default auto reply when no code execution or llm-based reply is generated.
             description (str): a short description of the agent. This description is used by other agents
                 (e.g. the GroupChatManager) to decide when to call upon this agent. (Default: system_message)
         """
@@ -143,6 +143,11 @@ def __init__(
             self.llm_config = self.DEFAULT_CONFIG.copy()
             if isinstance(llm_config, dict):
                 self.llm_config.update(llm_config)
+            # We still have a default `llm_config` because the user didn't
+            # specify anything. This won't work, so raise an error to avoid
+            # an obscure message from the OpenAI service.
+            if self.llm_config == self.DEFAULT_CONFIG:
+                raise ValueError("Please specify the value for 'llm_config'.")
             self.client = OpenAIWrapper(**self.llm_config)
 
         # Initialize standalone client cache object.

autogen/oai/client.py

@@ -345,15 +345,27 @@ def __init__(self, *, config_list: Optional[List[Dict[str, Any]]] = None, **base
                 and additional kwargs.
         """
         openai_config, extra_kwargs = self._separate_openai_config(base_config)
+        # This *may* work if the `llm_config` has specified the `model` attribute,
+        # so just warn here.
         if type(config_list) is list and len(config_list) == 0:
-            logger.warning("openai client was provided with an empty config_list, which may not be intended.")
+            logger.warning("OpenAI client was provided with an empty config_list, which may not be intended.")
+            # If the `llm_config` has no `model` then the call will fail. Abort now.
+            if "model" not in extra_kwargs:
+                raise ValueError("Please specify a value for the 'model' in 'llm_config'.")
 
         self._clients: List[ModelClient] = []
         self._config_list: List[Dict[str, Any]] = []
 
         if config_list:
             config_list = [config.copy() for config in config_list]  # make a copy before modifying
             for config in config_list:
+                # We require that each element of `config_list` has a non-empty value
+                # for `model` specified.
+                model = None
+                if "model" in config:
+                    model = config["model"]
+                if model is None or len(model) == 0:
+                    raise ValueError("Please specify a non-empty 'model' value for every item in 'config_list'.")
                 self._register_default_client(config, openai_config)  # could modify the config
                 self._config_list.append(
                     {**extra_kwargs, **{k: v for k, v in config.items() if k not in self.openai_kwargs}}

test/agentchat/contrib/test_web_surfer.py

@@ -51,7 +51,9 @@ def test_web_surfer() -> None:
         mp.setenv("OPENAI_API_KEY", "mock")
         page_size = 4096
         web_surfer = WebSurferAgent(
-            "web_surfer", llm_config={"config_list": []}, browser_config={"viewport_size": page_size}
+            "web_surfer",
+            llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]},
+            browser_config={"viewport_size": page_size},
         )
 
         # Sneak a peak at the function map, allowing us to call the functions for testing here

test/agentchat/test_conversable_agent.py

@@ -474,7 +474,7 @@ async def test_a_generate_reply_raises_on_messages_and_sender_none(conversable_a
 def test_update_function_signature_and_register_functions() -> None:
     with pytest.MonkeyPatch.context() as mp:
         mp.setenv("OPENAI_API_KEY", "mock")
-        agent = ConversableAgent(name="agent", llm_config={})
+        agent = ConversableAgent(name="agent", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
 
         def exec_python(cell: str) -> None:
             pass
@@ -618,9 +618,9 @@ def get_origin(d: Dict[str, Callable[..., Any]]) -> Dict[str, Callable[..., Any]
 def test_register_for_llm():
     with pytest.MonkeyPatch.context() as mp:
         mp.setenv("OPENAI_API_KEY", "mock")
-        agent3 = ConversableAgent(name="agent3", llm_config={"config_list": []})
-        agent2 = ConversableAgent(name="agent2", llm_config={"config_list": []})
-        agent1 = ConversableAgent(name="agent1", llm_config={"config_list": []})
+        agent3 = ConversableAgent(name="agent3", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
+        agent2 = ConversableAgent(name="agent2", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
+        agent1 = ConversableAgent(name="agent1", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
 
         @agent3.register_for_llm()
         @agent2.register_for_llm(name="python")
@@ -691,9 +691,9 @@ async def exec_sh(script: Annotated[str, "Valid shell script to execute."]) -> s
 def test_register_for_llm_api_style_function():
     with pytest.MonkeyPatch.context() as mp:
         mp.setenv("OPENAI_API_KEY", "mock")
-        agent3 = ConversableAgent(name="agent3", llm_config={"config_list": []})
-        agent2 = ConversableAgent(name="agent2", llm_config={"config_list": []})
-        agent1 = ConversableAgent(name="agent1", llm_config={"config_list": []})
+        agent3 = ConversableAgent(name="agent3", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
+        agent2 = ConversableAgent(name="agent2", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
+        agent1 = ConversableAgent(name="agent1", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
 
         @agent3.register_for_llm(api_style="function")
         @agent2.register_for_llm(name="python", api_style="function")
@@ -762,7 +762,7 @@ async def exec_sh(script: Annotated[str, "Valid shell script to execute."]) -> s
 def test_register_for_llm_without_description():
     with pytest.MonkeyPatch.context() as mp:
         mp.setenv("OPENAI_API_KEY", "mock")
-        agent = ConversableAgent(name="agent", llm_config={})
+        agent = ConversableAgent(name="agent", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
 
         with pytest.raises(ValueError) as e:
 
@@ -774,25 +774,33 @@ def exec_python(cell: Annotated[str, "Valid Python cell to execute."]) -> str:
 
 
 def test_register_for_llm_without_LLM():
-    with pytest.MonkeyPatch.context() as mp:
-        mp.setenv("OPENAI_API_KEY", "mock")
-        agent = ConversableAgent(name="agent", llm_config=None)
-        agent.llm_config = None
-        assert agent.llm_config is None
+    try:
+        ConversableAgent(name="agent", llm_config=None)
+        assert False, "Expected ConversableAgent to throw ValueError."
+    except ValueError as e:
+        assert e.args[0] == "Please specify the value for 'llm_config'."
 
-        with pytest.raises(RuntimeError) as e:
 
-            @agent.register_for_llm(description="run cell in ipython and return the execution result.")
-            def exec_python(cell: Annotated[str, "Valid Python cell to execute."]) -> str:
-                pass
+def test_register_for_llm_without_configuration():
+    try:
+        ConversableAgent(name="agent", llm_config={"config_list": []})
+        assert False, "Expected ConversableAgent to throw ValueError."
+    except ValueError as e:
+        assert e.args[0] == "Please specify a value for the 'model' in 'llm_config'."
+
 
-        assert e.value.args[0] == "LLM config must be setup before registering a function for LLM."
+def test_register_for_llm_without_model_name():
+    try:
+        ConversableAgent(name="agent", llm_config={"config_list": [{"model": "", "api_key": ""}]})
+        assert False, "Expected ConversableAgent to throw ValueError."
+    except ValueError as e:
+        assert e.args[0] == "Please specify a non-empty 'model' value for every item in 'config_list'."
 
 
 def test_register_for_execution():
     with pytest.MonkeyPatch.context() as mp:
         mp.setenv("OPENAI_API_KEY", "mock")
-        agent = ConversableAgent(name="agent", llm_config={"config_list": []})
+        agent = ConversableAgent(name="agent", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
         user_proxy_1 = UserProxyAgent(name="user_proxy_1")
         user_proxy_2 = UserProxyAgent(name="user_proxy_2")
 
@@ -827,7 +835,7 @@ async def exec_sh(script: Annotated[str, "Valid shell script to execute."]):
 def test_register_functions():
     with pytest.MonkeyPatch.context() as mp:
         mp.setenv("OPENAI_API_KEY", "mock")
-        agent = ConversableAgent(name="agent", llm_config={"config_list": []})
+        agent = ConversableAgent(name="agent", llm_config={"config_list": [{"model": "gpt-4", "api_key": ""}]})
         user_proxy = UserProxyAgent(name="user_proxy")
 
         def exec_python(cell: Annotated[str, "Valid Python cell to execute."]) -> str: