supportsStreaming

Check if true reactive streaming is supported by the underlying LLM model. Always check this before calling streaming() to avoid exceptions.

The answer is established by streaming a short prompt at the model the first time it is asked for, then remembered, so repeated checks on a configured model cost nothing. A model built per request, as bring-your-own-key setups do, is new each time and so is probed each time. A false may therefore be a model that cannot stream, or a provider that could not be reached on this attempt; the latter is logged and re-checked on the next call.