microsoft/semantic-kernel · error · NotImplementedError

The AutoGenConversableAgent does not support streaming.

Error message

The AutoGenConversableAgent does not support streaming.

What it means

Raised unconditionally by AutoGenConversableAgent.invoke_stream, which overrides the streaming entry point to signal that AutoGen 0.2 ConversableAgent does not support streaming responses. Any caller that routes through invoke_stream (e.g. generic streaming code paths) hits this immediately.

Source

Thrown at python/semantic_kernel/agents/autogen/autogen_conversable_agent.py:266

                messages=[message.to_dict() async for message in thread.get_messages()],
            )

            logger.info("Called AutoGenConversableAgent.a_generate_reply.")

            yield await self._create_reply_content(reply, thread)

    @override
    def invoke_stream(
        self,
        messages: str | ChatMessageContent | list[str | ChatMessageContent] | None = None,
        *,
        thread: AgentThread | None = None,
        kernel: "Kernel | None" = None,
        arguments: KernelArguments | None = None,
        **kwargs: Any,
    ) -> AsyncIterable[AgentResponseItem["StreamingChatMessageContent"]]:
        """Invoke the agent with a stream of messages."""
        raise NotImplementedError("The AutoGenConversableAgent does not support streaming.")

    @staticmethod
    def _to_chat_message_content(message: dict[str, Any]) -> ChatMessageContent:
        """Translate an AutoGen message to a Semantic Kernel ChatMessageContent."""
        items: list[TextContent | FunctionCallContent | FunctionResultContent] = []
        role = AuthorRole(message.get("role"))
        name: str = message.get("name", "")

        content = message.get("content")
        if content is not None:
            text = TextContent(text=content)
            items.append(text)

        if role == AuthorRole.ASSISTANT:
            tool_calls = message.get("tool_calls")
            if tool_calls is not None:
                for tool_call in tool_calls:
                    items.append(

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Use the non-streaming `agent.invoke(...)` (or get_response) for AutoGenConversableAgent.
2. Branch in your runner: check `isinstance(agent, AutoGenConversableAgent)` and call invoke instead of invoke_stream.
3. Switch to an agent type that supports streaming (e.g. ChatCompletionAgent) if streaming is mandatory.

Example fix

# before
async for chunk in agent.invoke_stream(messages='hi', thread=thread):
    ...

# after
async for response in agent.invoke(messages='hi', thread=thread):
    ...

from semantic_kernel.agents.autogen import AutoGenConversableAgent
streaming = not isinstance(agent, AutoGenConversableAgent)

from semantic_kernel.agents.autogen import AutoGenConversableAgent
def supports_streaming(agent) -> bool:
    return not isinstance(agent, AutoGenConversableAgent)

from semantic_kernel.agents.autogen import AutoGenConversableAgent
if isinstance(agent, AutoGenConversableAgent):
    responses = [r async for r in agent.invoke(messages=msg, thread=thread)]
else:
    async for chunk in agent.invoke_stream(messages=msg, thread=thread):
        ...

When it happens

Trigger: Calling `async for chunk in agent.invoke_stream(...)` on an AutoGenConversableAgent; a framework/orchestrator that always uses the streaming API regardless of agent type.

Common situations: Generic agent runner that picks invoke_stream for all agents; migrating from a streaming-capable agent to AutoGenConversableAgent without switching to non-streaming invoke.

Related errors

• Streaming text contents not implemented.
• Streaming chat message contents not implemented.
• Unable to get channel keys. Channel type not configured.
• Unable to create channel. Channel type not configured.
• Cannot retrieve chat history, since the thread has been dele

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/d2e463f3f31bce01. Report an issue: GitHub.