Interface STTV3Options

Options specific to saaras:v3 (recommended).

See

Sarvam STT API docs

interface STTV3Options {
    apiKey?: string;
    firstTurnMinSpeechFrames?: number;
    flushSignal?: boolean;
    highVadSensitivity?: boolean;
    interruptMinSpeechFrames?: number;
    languageCode?: string;
    minSpeechFrames?: number;
    mode?: string;
    model?: "saaras:v3";
    negativeFramesCount?: number;
    negativeFramesWindow?: number;
    negativeSpeechThreshold?: number;
    numInitialIgnoredFrames?: number;
    positiveSpeechThreshold?: number;
    preSpeechPadFrames?: number;
    prompt?: string;
    startSpeechVolumeThreshold?: number;
    streaming?: boolean;
    withTimestamps?: boolean;
}

Hierarchy

  • STTBaseOptions
    • STTV3Options

Properties

apiKey? firstTurnMinSpeechFrames? flushSignal? highVadSensitivity? interruptMinSpeechFrames? languageCode? minSpeechFrames? mode? model? negativeFramesCount? negativeFramesWindow? negativeSpeechThreshold? numInitialIgnoredFrames? positiveSpeechThreshold? preSpeechPadFrames? prompt? startSpeechVolumeThreshold? streaming? withTimestamps?

Properties

Optional apiKey

apiKey?: string

Sarvam API key. Defaults to $SARVAM_API_KEY

Optional firstTurnMinSpeechFrames

firstTurnMinSpeechFrames?: number

Fine-grained VAD first-turn minimum speech frames (WS only).

Optional flushSignal

flushSignal?: boolean

Enable flush signal events from server (WS only). Maps to flush_signal query param.

Optional highVadSensitivity

highVadSensitivity?: boolean

Increase VAD sensitivity (WS only). Maps to high_vad_sensitivity query param.

Optional interruptMinSpeechFrames

interruptMinSpeechFrames?: number

Fine-grained VAD interrupt minimum speech frames (WS only).

Optional languageCode

languageCode?: string

Language code (BCP-47). Default: 'en-IN'. Set to 'unknown' for auto-detection.

Optional minSpeechFrames

minSpeechFrames?: number

Fine-grained VAD minimum speech frames (WS only).

Optional mode

mode?: string

Transcription mode (v3 only). Default: 'transcribe'

Optional model

model?: "saaras:v3"

Optional negativeFramesCount

negativeFramesCount?: number

Fine-grained VAD negative frames count (WS only).

Optional negativeFramesWindow

negativeFramesWindow?: number

Fine-grained VAD negative frames window (WS only).

Optional negativeSpeechThreshold

negativeSpeechThreshold?: number

Fine-grained VAD negative speech threshold (WS only).

Optional numInitialIgnoredFrames

numInitialIgnoredFrames?: number

Fine-grained VAD initial ignored frames (WS only).

Optional positiveSpeechThreshold

positiveSpeechThreshold?: number

Fine-grained VAD positive speech threshold (WS only).

Optional preSpeechPadFrames

preSpeechPadFrames?: number

Fine-grained VAD pre-speech padding frames (WS only).

Optional prompt

prompt?: string

Conversation context to boost model accuracy

Optional startSpeechVolumeThreshold

startSpeechVolumeThreshold?: number

Fine-grained VAD start speech volume threshold (WS only).

Optional streaming

streaming?: boolean

Whether to use native WebSocket streaming for stream(). Set to false to prefer non-streaming REST recognition (used by Agent via StreamAdapter + VAD). Default: true.

Optional withTimestamps

withTimestamps?: boolean

Return chunk-level timestamps in REST response

Settings

Member Visibility

Theme

On This Page

apiKeyfirstTurnMinSpeechFramesflushSignalhighVadSensitivityinterruptMinSpeechFrameslanguageCodeminSpeechFramesmodemodelnegativeFramesCountnegativeFramesWindownegativeSpeechThresholdnumInitialIgnoredFramespositiveSpeechThresholdpreSpeechPadFramespromptstartSpeechVolumeThresholdstreamingwithTimestamps