Interface TurnDetectorOptions

interface TurnDetectorOptions {
    apiKey?: string;
    apiSecret?: string;
    backchannelThreshold?: number | Record<string, number>;
    baseUrl?: string;
    connOptions?: APIConnectOptions;
    executor?: InferenceExecutor;
    localFallback?: boolean;
    sampleRate?: number;
    unlikelyThreshold?: number | Record<string, number>;
    version?: TurnDetectorVersion;
}

Properties

apiKey? apiSecret? backchannelThreshold? baseUrl? connOptions? executor? localFallback? sampleRate? unlikelyThreshold? version?

Properties

Optional apiKey

apiKey?: string

Optional apiSecret

apiSecret?: string

Optional backchannelThreshold

backchannelThreshold?: number | Record<string, number>

Backchannel threshold(s): above this, a pause is a backchannel opportunity. Server-driven and cloud-only by default; this is an override seam. A scalar applies to every language; a map is layered over the server defaults.

Optional baseUrl

baseUrl?: string

Optional connOptions

connOptions?: APIConnectOptions

Optional executor

executor?: InferenceExecutor

Inference executor that runs the local turn-detector-v1-mini model in the shared inference process. Defaults to the current job's getJobContext().inferenceExecutor. undefined (no job context / binding unavailable) degrades the local model to a positive-default prediction. Mainly an override seam for tests.

Optional localFallback

localFallback?: boolean

Allow v1 to fall back to the local v1-mini model. Defaults to true. When false, prediction timeouts keep the cloud transport active. A terminal cloud failure disables detection for the stream, so turns commit on the endpointing delay. Has no effect when v1-mini is selected. Set LIVEKIT_AGENTS_PRELOAD_LOCAL_INFERENCE=0 separately to skip preloading the local model weights.

Optional sampleRate

sampleRate?: number

Sample rate (Hz). Defaults to 16000.

Optional unlikelyThreshold

unlikelyThreshold?: number | Record<string, number>

Optional version

version?: TurnDetectorVersion

Which turn-detector version to run. 'v1' is the full cloud model (served over the inference gateway; model name 'turn-detector-v1'); 'v1-mini' is the local in-process model ('turn-detector-v1-mini'). When omitted, auto-selects 'v1' on hosted/dev environments (falling back to 'v1-mini' if cloud creds are missing) and 'v1-mini' otherwise. The worker still preloads the mini weights unless LIVEKIT_AGENTS_PRELOAD_LOCAL_INFERENCE=0 is set.

Settings

Member Visibility

Theme

On This Page

apiKeyapiSecretbackchannelThresholdbaseUrlconnOptionsexecutorlocalFallbacksampleRateunlikelyThresholdversion