createResponseStream

Streams a Responses API call as semantic server-sent events.

Unlike the chat stream, which repeats the whole envelope in every chunk, the Responses API narrates the response as it is assembled — see ResponseStreamEvent. There is no [DONE] marker: the flow ends after ResponseCompletedEvent, ResponseIncompleteEvent or ResponseFailedEvent, each carrying the finished ModelResponse with its usage.

Example:

val request = responseStreamParams { model = ChatModel.DEEPSEEK_FLASH }
    .createRequest("Write a haiku about Kotlin")

client.createResponseStream(request).collect { event ->
    if (event is OutputTextDeltaEvent) print(event.delta)
}

Parameters

Throws