Injected clock (epoch ms). Defaults to Date.now.
RMS threshold. Defaults to HUMAN_SPEECH_RMS_THRESHOLD.
Inspects one inbound PCM16 frame from a HUMAN participant.
The frame's samples.
true when the caller should report human speech activity to the coordinator now (the frame
is speech-like AND the report interval has elapsed); false otherwise.
Energy-based detector of human speech in inbound room audio. Deliberately simple: it only has to be right enough to tell the coordinator "a person is talking" so agents yield; the room's own transcripts remain the record of what was said. It throttles its reports so the coordinator is not called per frame.