The OpenAI provider profile — the defaults every OpenAI-compatible subclass overrides from.
A raw audio key in the Config bag merges into the built audio block (see MergeAudioOverride),
so these native keys are reachable: audio.input.turn_detection.{ prefix_padding_ms, idle_timeout_ms, interrupt_response, create_response }. create_response only takes effect
outside meeting mode (disableAutoResponse), where the floor create_response: false always wins,
and a live Reconfigure sets it from its own DisableAutoResponse. An explicit null is sent as
null: turn_detection: null turns VAD off (the client triggers responses).
The OpenAI provider profile — the defaults every OpenAI-compatible subclass overrides from.
A raw
audiokey in the Config bag merges into the built audio block (seeMergeAudioOverride), so these native keys are reachable:audio.input.turn_detection.{ prefix_padding_ms, idle_timeout_ms, interrupt_response, create_response }.create_responseonly takes effect outside meeting mode (disableAutoResponse), where the floorcreate_response: falsealways wins, and a liveReconfiguresets it from its ownDisableAutoResponse. An explicitnullis sent asnull:turn_detection: nullturns VAD off (the client triggers responses).