Python: Cannot include kernel kwarg in semantic_function

[tjprescott]

Describe the bug
Following the documentation here:
https://learn.microsoft.com/en-us/semantic-kernel/concepts/ai-services/chat-completion/function-calling/?pivots=programming-language-python#reserved-parameter-names-for-auto-function-calling

I'm trying to include the kernel argument so that it will automatically be passed into my @kernel_function.

    @kernel_function(
        name="search_guidelines",
        description="Semantic search against the guidelines with natural language query.",
    )
    def search_guidelines(
        self,
        query: str,
        programming_language: str,
        kernel: Kernel
    ) -> str:

However, whenever I do this, the following code blows up:

        result = await chat_completion.get_chat_message_content(
            chat_history=history,
            settings=execution_settings,
            kernel=kernel,
        )

I get the following errors:

Traceback (most recent call last):
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\semantic_kernel\connectors\ai\open_ai\services\open_ai_handler.py", line 87, in _send_completion_request
    response = await self.client.chat.completions.create(**settings_dict)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\openai\resources\chat\completions\completions.py", line 2000, in create
    return await self._post(
           ^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\openai\_base_client.py", line 1767, in post
    return await self.request(cast_to, opts, stream=stream, stream_cls=stream_cls)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\openai\_base_client.py", line 1461, in request
    return await self._request(
           ^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\openai\_base_client.py", line 1492, in _request
    request = self._build_request(options, retries_taken=retries_taken)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\openai\lib\azure.py", line 67, in _build_request
    return super()._build_request(options, retries_taken=retries_taken)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\openai\_base_client.py", line 506, in _build_request
    return self._client.build_request(  # pyright: ignore[reportUnknownMemberType]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\httpx\_client.py", line 378, in build_request
    return Request(
           ^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\httpx\_models.py", line 408, in __init__
    headers, stream = encode_request(
                      ^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\httpx\_content.py", line 216, in encode_request
    return encode_json(json)
           ^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\site-packages\httpx\_content.py", line 177, in encode_json
    body = json_dumps(
           ^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\json\__init__.py", line 238, in dumps
    **kw).encode(obj)
          ^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\json\encoder.py", line 200, in encode
    chunks = self.iterencode(o, _one_shot=True)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\json\encoder.py", line 258, in iterencode
    return _iterencode(o, 0)
           ^^^^^^^^^^^^^^^^^
  File "C:\Users\trpresco\.pyenv\pyenv-win\versions\3.12.9\Lib\json\encoder.py", line 180, in default
    raise TypeError(f'Object of type {o.__class__.__name__} '
TypeError: Object of type _PydanticGeneralMetadata is not JSON serializable

According to the documentation, this is supposed to "just work".

Expected behavior
Should pass the kernel object into my @kernel_function

Platform

• Language: Python
• Source: semantic-kernel 1.27.2
• AI model: Azure OpenAI: GPT-4o-mini
• IDE: VS Code
• OS: Windows

[moonbox3]

@tjprescott, I will look into this. In the meantime, you can create a member variable on your plugin class where you pass in your kernel instance during plugin instantiation like this:

class SamplePlugin:
    def __init__(self, kernel: Kernel):
        self.kernel = kernel

    @kernel_function
    async def get_weather(self, location: str) -> str:
        # Simulate a weather plugin function
        # You have access to self.kernel here...
        return f"The weather in {location} is sunny with a chance of rain."

kernel = Kernel()
kernel.add_plugin(SamplePlugin(kernel), plugin_name="SamplePlugin")

The current issue, without digging further, is we're trying to create a JSON schema to represent the kernel, which is not JSON serializable. I've got an issue around this (#11143) to make sure we actually filter out kernel function params that are related to the reserved params as you show in the learn docs.

[moonbox3]

@tjprescott, we have some further handling for this, if you choose not to use the way I suggested above: https://learn.microsoft.com/en-us/semantic-kernel/concepts/ai-services/chat-completion/function-calling/?pivots=programming-language-python#custom-reserved-parameter-names-for-auto-function-calling

The idea is to add the metadata to the Annotated type hint:

class SimplePlugin:
    @kernel_function(name="GetWeather", description="Get the weather for a location.")
    async def get_the_weather(self, location: str, special_arg: Annotated[str, {"include_in_function_choices": False}]) -> str:
        # The 'special_arg' parameter is reserved and you need to ensure it either has a default value or gets passed.

This will make sure that the special_arg (or kernel in your case) does is not configured as required therefore the model won't try and send an argument back for it.