microsoft/semantic-kernel · error · ServiceInvalidResponseError

No message content found in response part.

Error message

No message content found in response part.

What it means

Raised as ServiceInvalidResponseError in `_create_streaming_chat_message_content` (the raw Mapping/dict branch of streaming chat) when a streamed part dict has no `'message'` key (`part.get('message', None)` is falsy). The streaming counterpart of error 1117: the part was a dict (passed the 1116 check) but lacks the message envelope.

Source

Thrown at python/semantic_kernel/connectors/ai/ollama/services/ollama_chat_completion.py:312

                        name=tool_call.get("function").get("name"),
                        arguments=tool_call.get("function").get("arguments"),
                    )
                )

        return ChatMessageContent(
            role=AuthorRole.ASSISTANT,
            items=items,
            inner_content=response,
            metadata=self._get_metadata_from_response(response),
        )

    def _create_streaming_chat_message_content(
        self, part: Mapping[str, Any], function_invoke_attempt: int
    ) -> StreamingChatMessageContent:
        """Create a streaming chat message content from the response part."""
        items: list[STREAMING_ITEM_TYPES] = []
        if not (message := part.get("message", None)):
            raise ServiceInvalidResponseError("No message content found in response part.")

        if content := message.get("content", None):
            items.append(
                StreamingTextContent(
                    choice_index=0,
                    text=content,
                    inner_content=message,
                )
            )
        if tool_calls := message.get("tool_calls", None):
            for tool_call in tool_calls:
                items.append(
                    FunctionCallContent(
                        inner_content=tool_call,
                        ai_model_id=self.ai_model_id,
                        name=tool_call.get("function").get("name"),
                        arguments=tool_call.get("function").get("arguments"),
                    )

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Log/inspect the offending part (it is the inner_content) to see if it is an error or control chunk.
2. Ensure the Ollama server and SDK versions match the connector's expected streaming schema.
3. If some chunks legitimately lack 'message', filter them in a custom client wrapper before yielding.
4. Verify the model is loaded and healthy (`ollama ps`).

Example fix

# before - server yields a final {'done': True, ...} chunk with no 'message'

# after - wrap the client to skip message-less parts
async def chat(self, **kw):
    async for part in self._real.chat(**kw):
        if isinstance(part, Mapping) and not part.get('message'):
            continue
        yield part

# In a smoke test, ensure streaming parts carry a message
async for part in await svc.client.chat(model=svc.ai_model_id, messages=[...], stream=True):
    if isinstance(part, dict):
        # allow message-less parts only if they are clearly terminal/error
        if not part.get('message'):
            assert part.get('done') or part.get('error'), f'unexpected part: {part}'

from collections.abc import Mapping

def stream_part_has_message(part) -> bool:
    return isinstance(part, Mapping) and bool(part.get('message'))

from semantic_kernel.exceptions import ServiceInvalidResponseError
try:
    async for chunk in svc._inner_get_streaming_chat_message_contents(chat_history, settings):
        ...
except ServiceInvalidResponseError as e:
    if 'No message content found in response part' in str(e):
        raise RuntimeError('a streaming chunk lacked a message envelope; check server/SDK version') from e
    raise

When it happens

Trigger: A streaming chunk dict from Ollama that omits 'message' - e.g. a final/usage chunk, a control/error chunk, or an Ollama streaming schema variant that nests content differently in some chunks.

Common situations: Ollama server emits a trailing chunk without a message field; an error chunk mid-stream; an API/SDK version whose streaming schema differs for the last chunk.

Related errors

• Invalid response type from Ollama streaming chat completion.
• Invalid response type from Ollama streaming chat completion.
• No message content found in response.
• Invalid response type from Ollama chat completion. Expected
• The _inner_get_streaming_chat_message_contents method is not

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/625ba7fba5b2fe6d. Report an issue: GitHub.