aixblock-voice-ai-deepgram/dist/types/transcription.d.ts

A React component for real-time transcription and voice agent interactions using Deepgram APIs

/**
 * Types related to Deepgram's transcription API
 */
/**
 * Configuration for the Deepgram transcription API
 * Based on Deepgram /v1/listen query parameters
 */
export interface TranscriptionOptions {
    /**
     * Transcription model to use
     * e.g., "nova-2"
     */
    model?: string;
    /**
     * Language to transcribe
     * e.g., "en-US"
     */
    language?: string;
    /**
     * Enable speaker identification
     */
    diarize?: boolean;
    /**
     * Enable smart formatting of numbers, dates, etc.
     */
    smart_format?: boolean;
    /**
     * Add punctuation to the transcript
     */
    punctuate?: boolean;
    /**
     * Enable automatic endpoint detection
     * Can be boolean or milliseconds of silence
     */
    endpointing?: boolean | number;
    /**
     * Return interim (non-final) results
     */
    interim_results?: boolean;
    /**
     * Enable voice activity detection events
     */
    vad_events?: boolean;
    /**
     * Keywords to detect in the audio.
     * @see https://developers.deepgram.com/docs/keywords
     */
    keywords?: string[];
    /**
     * Keyterms to boost recognition for (Nova-3 English only).
     * Each string in the array will be sent as a separate 'keyterm' parameter.
     * Phrases with spaces are handled correctly.
     * @see https://developers.deepgram.com/docs/keyterm
     */
    keyterm?: string[];
    /**
     * Any other parameters supported by Deepgram API
     */
    [key: string]: any;
}
/**
 * Word object in a transcript
 */
export interface TranscriptWord {
    /**
     * The transcribed word
     */
    word: string;
    /**
     * Start time in seconds
     */
    start: number;
    /**
     * End time in seconds
     */
    end: number;
    /**
     * Confidence score (0-1)
     */
    confidence: number;
    /**
     * Speaker ID (if diarization is enabled)
     */
    speaker?: number;
}
/**
 * Alternative transcript
 */
export interface TranscriptAlternative {
    /**
     * The transcript text
     */
    transcript: string;
    /**
     * Confidence score (0-1)
     */
    confidence: number;
    /**
     * Words with timing information
     */
    words: TranscriptWord[];
}
/**
 * Transcript response from Deepgram
 */
export interface TranscriptResponse {
    /**
     * Indicates this is a transcript message
     */
    type: 'transcript';
    /**
     * Audio channel index
     */
    channel: number;
    /**
     * Whether this is a final result
     */
    is_final: boolean;
    /**
     * Whether speech has ended
     */
    speech_final: boolean;
    /**
     * Channel indexes for multi-channel audio
     */
    channel_index: number[];
    /**
     * Start time of this segment in seconds
     */
    start: number;
    /**
     * Duration of this segment in seconds
     */
    duration: number;
    /**
     * Alternative transcriptions, ordered by confidence
     */
    alternatives: TranscriptAlternative[];
    /**
     * Additional metadata
     */
    metadata?: any;
}
/**
 * Voice activity detection event
 */
export interface VADEvent {
    /**
     * Indicates this is a VAD event
     */
    type: 'vad';
    /**
     * Start time in seconds
     */
    start: number;
    /**
     * End time in seconds
     */
    end: number;
    /**
     * Whether speech is present
     */
    speech_detected: boolean;
}
/**
 * Union type for all transcription messages
 */
export type TranscriptionMessage = TranscriptResponse | VADEvent;