Interface AgentTurnLatency

Per-turn latency metrics (all values in milliseconds).

interface AgentTurnLatency {
    transcriber_latency?: number;
    turn_detection_latency?: number;
    model_latency?: number;
    voice_latency?: number;
    preflight?: AgentPreflightMetrics;
}

Properties

transcriber_latency?: number

STT processing latency: stop talking to final transcript received.

turn_detection_latency?: number

Additional wait after transcript for end-of-turn detection.

model_latency?: number

LLM latency: wait for first token after ready to prompt. Absent on preflight hit.

voice_latency?: number

TTS engine latency: first text sent to TTS until first audio received.

Early generation metrics.