handle_message_chunk(
self,
payload: Any,
) -> AsyncIterator[Any]Handle a payload from stream_mode="messages".
The payload carries either one token chunk of a streaming chat model
response or a whole :class:AIMessage (non-streaming LLM call, or a
node that built the message itself). Payloads sharing a message id
accumulate into a single message output item; a different id
closes the open item and starts a new one.