OptionalEndEpoch-millisecond timestamp at which this segment's speech ENDED. Used to measure the silence
window since the last human speech. When omitted, the policy treats the clock's now() as the
end time (i.e. no measurable silence yet).
OptionalIsWhether the speaker of this segment is an agent bot (excluded from "addressed me" detection).
OptionalSpeakerThe diarized speaker label, when the provider supports it (e.g. a participant id or name).
The transcribed text of this segment.
A single diarized transcript segment fed into the policy.
Produced from the bridge's inbound transcript stream.
SpeakerLabel/IsAgentcome from the provider's diarization;EndMslets the policy compute silence windows deterministically.