AiOpenAIFinishReason
OpenAI Chat Completions streaming shapes. toOpenAIChatCompletionStream maps the engine's transport-agnostic ChatEvent stream onto these chunks so a host can expose an OpenAI-compatible POST /v1/chat/completions (stream: true) endpoint backed by the same chat pipeline as the in-app widget. Only the subset of fields the engine can populate is emitted; everything else an OpenAI client tolerates as absent.
Possible values:
stoplengthtool_callscontent_filternull