microsoft/autogen · error · ValueError

No response was generated

Error message

No response was generated

What it means

OpenAIAgent.on_messages consumes on_messages_stream and expects exactly one Response event to emerge. It initializes response=None and iterates the stream collecting non-Response events as inner messages; if the stream completes without ever yielding a Response, it raises ValueError('No response was generated'). This is a stream-protocol/lifecycle failure, not a model error — the agent's stream ended cleanly but never emitted its final result.

Source

Thrown at python/packages/autogen-ext/src/autogen_ext/agents/openai/_openai_agent.py:535

        return api_params

    async def on_messages(
        self: "OpenAIAgent", messages: Sequence[BaseChatMessage], cancellation_token: CancellationToken
    ) -> Response:
        response = None
        inner_messages: List[
            Union[AgentEvent, TextMessage, MultiModalMessage, StopMessage, ToolCallSummaryMessage, HandoffMessage]
        ] = []

        async for msg in self.on_messages_stream(messages, cancellation_token):
            if isinstance(msg, Response):
                response = msg
            # ModelClientStreamingChunkEvent does not exist in this version, so skip this check
            else:
                inner_messages.append(msg)

        if response is None:
            raise ValueError("No response was generated")

        if response.inner_messages is None:
            response.inner_messages = []

        for msg in inner_messages:
            if msg not in response.inner_messages:
                response.inner_messages = list(response.inner_messages) + [msg]

        return response

    async def on_messages_stream(
        self: "OpenAIAgent", messages: Sequence[BaseChatMessage], cancellation_token: CancellationToken
    ) -> AsyncGenerator[
        Union[
            AgentEvent, TextMessage, MultiModalMessage, StopMessage, ToolCallSummaryMessage, HandoffMessage, Response
        ],
        None,
    ]:

View on GitHub (pinned to 027ecf0a37)

Solutions

  1. Ensure you pass a non-empty, well-formed message sequence to on_messages.
  2. If you subclass or wrap OpenAIAgent, make sure on_messages_stream always yields a Response as its final event.
  3. Upgrade autogen-ext — stream lifecycle edge cases around empty message lists have been fixed across releases.
  4. As a defensive measure in caller code, consume on_messages_stream directly and handle the no-Response case explicitly (log + retry with the message re-sent).

Example fix

# before
response = await agent.on_messages([], cancellation_token)

# after
from autogen_core import CancellationToken
response = await agent.on_messages(
    [TextMessage(source="user", content="Hello")], CancellationToken()
)
Defensive patterns

Strategy: try-catch

Validate before calling

if not messages:
    raise ValueError("on_messages requires at least one message")

Try / catch

try:
    response = await agent.on_messages(msgs, ct)
except ValueError as e:
    if "No response was generated" in str(e):
        response = await agent.on_messages(msgs, ct)  # single retry with same input
    else:
        raise

Prevention

When it happens

Trigger: The on_messages_stream implementation exiting before yielding Response (e.g. an empty message sequence, an internal early-return, or a subclass/custom stream that only yields chunk events); consuming a stream that was cancelled or drained elsewhere.

Common situations: Calling on_messages([]) or with only system messages in code paths where the underlying implementation skips the completion; custom subclasses of OpenAIAgent overriding on_messages_stream incorrectly; race conditions where the stream's final Response was already consumed.

Related errors

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/654a48999aaddcc2. Report an issue: GitHub.