Member Junction
    Preparing search index...

    Interface MJAIPromptModelEntity_RealtimeTurnDetectionSettings

    MJ-normalized turn-detection settings for realtime (speech-to-speech) models. Every field is optional; absent fields contribute nothing. Provider profiles translate this vocabulary to their native wire block, and an unsupported value is diagnostic-logged and falls back to the profile default — a wrong inherited value degrades safely, it never rejects a session.

    interface MJAIPromptModelEntity_RealtimeTurnDetectionSettings {
        Eagerness?: "auto" | "low" | "high";
        Mode?: "native" | "default" | "serverVad" | "semanticVad";
        SilenceDurationMs?: number;
        Threshold?: number;
    }
    Index

    Properties

    Eagerness?: "auto" | "low" | "high"

    Semantic-VAD aggressiveness (OpenAI eagerness); ignored without a mapping.

    Mode?: "native" | "default" | "serverVad" | "semanticVad"
    • 'default' — let the provider profile decide (today's behavior).
    • 'serverVad' — classic silence-based server VAD.
    • 'semanticVad' — semantic end-of-utterance detection (OpenAI semantic_vad).
    • 'native' — this model's smartest documented turn/duplex mode, whatever the profile maps it to; the forward slot for full-duplex reasoning voice models.
    SilenceDurationMs?: number

    Server-VAD trailing-silence duration in ms; ignored without a mapping.

    Threshold?: number

    Server-VAD activation threshold (0–1); ignored without a mapping.