/**
 * Speech-to-text provider surface. fabric-harness ships Deepgram and
 * Cartesia implementations; bring-your-own for Whisper, AssemblyAI, etc.
 * Used by `PipelineVoiceProvider` to drive the audio-in side of
 * STT→LLM→TTS voice agents.
 */
import type { VoiceAudioFormat } from './voice.js';
export interface SttSessionOptions {
    /** Audio format the caller will feed via `send()`. Default `pcm16`. */
    audioFormat?: VoiceAudioFormat;
    /** Sample rate in Hz. Default 16000 (Deepgram-recommended) for `pcm16`. */
    sampleRate?: number;
    /** BCP-47 language tag, e.g. `en-US`, `es-MX`, or `multi` for autodetect. */
    language?: string;
    /** Provider-specific model id. */
    model?: string;
    /**
     * Voice-activity detection mode. `endpointing` lets the provider mark
     * end-of-utterance via silence detection; `none` returns continuous
     * transcripts and leaves turn detection to the caller.
     */
    vad?: 'endpointing' | 'none';
    /**
     * Optional ms of trailing silence to treat as end-of-utterance.
     * Provider-specific default when omitted.
     */
    endpointingMs?: number;
    /** Abort signal — terminates the session cleanly. */
    signal?: AbortSignal;
}
export type SttEvent = {
    type: 'transcript';
    /** Final transcript for an utterance — safe to send to LLM. */
    text: string;
    /** Whether this is the final transcript for the current utterance. */
    isFinal: true;
} | {
    type: 'partial';
    /** Interim transcript — useful for live UI, NOT for LLM input. */
    text: string;
    isFinal: false;
} | {
    type: 'speech_started';
} | {
    type: 'speech_ended';
} | {
    type: 'error';
    message: string;
};
export interface SttSessionUsage {
    /** Audio seconds processed (used for billing rollup). */
    seconds: number;
    /** Wall-clock duration in milliseconds. */
    durationMs: number;
}
export interface SttSession {
    /** Send a raw audio frame in the configured format. */
    send(frame: Uint8Array): void;
    /** Async iterable of recognition events. Closes when the session ends. */
    events(): AsyncIterable<SttEvent>;
    /** Tear down the underlying connection. */
    close(): Promise<void>;
    /** Usage rollup — populated as audio is processed. */
    usage(): SttSessionUsage;
}
export interface SttProvider {
    /** Stable provider id used in cost telemetry, e.g. `deepgram`, `cartesia`. */
    readonly name: string;
    /** Open a streaming recognition session. */
    open(opts?: SttSessionOptions): Promise<SttSession>;
}
//# sourceMappingURL=voice-stt.d.ts.map