Missing extra_create_kwargs in AssistantAgent

[gagb]

What happened?

Model client might need them, but assistant agent in agentchat never provides an option to set them or pass them to the client.

I discovered this while trying to use the skadapter with assistant agent + deepseek r1.

(python) gaganbansal@MacBook-Pro ~ % python ~/Downloads/ollamask.py
Error processing publish message for assistant/d88e2d75-f480-4e3a-990d-692ca482e085
Traceback (most recent call last):
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-core/src/autogen_core/_single_threaded_agent_runtime.py", line 409, in _on_message
    return await agent.on_message(
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-core/src/autogen_core/_base_agent.py", line 113, in on_message
    return await self.on_message_impl(message, ctx)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-agentchat/src/autogen_agentchat/teams/_group_chat/_sequential_routed_agent.py", line 48, in on_message_impl
    return await super().on_message_impl(message, ctx)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-core/src/autogen_core/_routed_agent.py", line 485, in on_message_impl
    return await h(self, message, ctx)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-core/src/autogen_core/_routed_agent.py", line 268, in wrapper
    return_value = await func(self, message, ctx)  # type: ignore
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-agentchat/src/autogen_agentchat/teams/_group_chat/_chat_agent_container.py", line 53, in handle_request
    async for msg in self._agent.on_messages_stream(self._message_buffer, ctx.cancellation_token):
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-agentchat/src/autogen_agentchat/agents/_assistant_agent.py", line 386, in on_messages_stream
    model_result = await self._model_client.create(
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-ext/src/autogen_ext/models/semantic_kernel/_sk_chat_completion_adapter.py", line 342, in create
    kernel = self._get_kernel(extra_create_args)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gaganbansal/workspace/autogen/python/packages/autogen-ext/src/autogen_ext/models/semantic_kernel/_sk_chat_completion_adapter.py", line 301, in _get_kernel
    raise ValueError("kernel must be provided either in constructor or extra_create_args")
ValueError: kernel must be provided either in constructor or extra_create_args

@lspinheiro @ekzhu

What did you expect to happen?

There should be a way to pass the kernel or extra kwargs to the client from the agent. I think the same is true for json output etc.

How can we reproduce it (as minimally and precisely as possible)?

import asyncio

from semantic_kernel import Kernel
from semantic_kernel.memory.null_memory import NullMemory
from semantic_kernel.connectors.ai.ollama import OllamaChatCompletion, OllamaChatPromptExecutionSettings

from autogen_agentchat.agents import AssistantAgent, CodeExecutorAgent
from autogen_agentchat.conditions import TextMentionTermination, MaxMessageTermination
from autogen_agentchat.ui import Console
from autogen_agentchat.teams import RoundRobinGroupChat

from autogen_ext.models.semantic_kernel import SKChatCompletionAdapter
from autogen_ext.code_executors.local import LocalCommandLineCodeExecutor


async def main():
    # Step 1: Initialize the Semantic Kernel
    kernel = Kernel(memory=NullMemory())

    # Step 2: Configure the Ollama client
    ollama_client = OllamaChatCompletion(
        service_id="ollama",
        host="http://localhost:11434",  # Local Ollama server
        ai_model_id="deepseek-r1",        # Replace with your desired Ollama model
    )
    model_client = SKChatCompletionAdapter(sk_client=ollama_client)

    assistant = AssistantAgent(
        name="assistant",
        system_message="You are a helpful assistant. Write all code in python. If the correct output is generated by code execution by user reply 'TERMINATE'. Code must be returned using ``` ```",
        model_client=model_client
    )

    code_executor = CodeExecutorAgent(
        name="code_executor",
        code_executor=LocalCommandLineCodeExecutor(work_dir="coding"),
    )

    # The termination condition is a combination of text termination and max message termination, either of which will cause the chat to terminate.
    termination = TextMentionTermination("TERMINATE") | MaxMessageTermination(10)

    # The group chat will alternate between the assistant and the code executor.
    group_chat = RoundRobinGroupChat([assistant, code_executor], termination_condition=termination)

    # `run_stream` returns an async generator to stream the intermediate messages.
    stream = group_chat.run_stream(task="Print hello world.")
    # `Console` is a simple UI to display the stream.
    await Console(stream)

asyncio.run(main())

AutoGen version

main

Which package was this bug in

AgentChat

Model used

deepseek-r1

Python version

No response

Operating system

No response

Any additional info you think would be helpful for fixing this bug

No response

[ekzhu]

@gagb I believe the SK client cannot be used by agent at this moment. #5134 is a way to address this issue, see discussion here: #4741