/** Voice ids supported by the OpenAI Realtime model. */
export declare const SUPPORTED_VOICES: readonly ["alloy", "ash", "ballad", "breeze", "cedar", "cinnamon", "coral", "echo", "ember", "juniper", "marin", "sage", "shimmer", "verse"];
/** Server-side voice-activity turn-detection settings. Pass `null` to disable. */
export type TurnDetectionConfig = {
    /** Detection mode. Default `'server_vad'`. */
    type?: string;
    /** Speech-probability threshold in [0, 1]. Default `0.5`. */
    threshold?: number;
    /** Audio padding (ms) prepended before detected speech. Default `300`. */
    prefixPaddingMs?: number;
    /** Trailing silence (ms) that ends a turn. Default `200`. */
    silenceDurationMs?: number;
    /** Auto-create a model response when a turn ends. Default `true`. */
    createResponse?: boolean;
    /** Allow user speech to interrupt the model's response. Default `true`. */
    interruptResponse?: boolean;
};
/** Input-audio transcription settings. Pass `null` to disable transcription. */
export type InputAudioTranscriptionConfig = {
    /** Transcription model id. Default `'gpt-4o-mini-transcribe'`. */
    model?: string;
};
/** Configuration for {@link OpenAIRealtime}. */
export type OpenAIRealtimeConfig = {
    /** Realtime model id. Default `'gpt-4o-realtime-preview'`. */
    model?: string;
    /** Voice id (see {@link SUPPORTED_VOICES}). Default `'alloy'`. */
    voice?: string;
    /** Response modalities. Default `['text', 'audio']`. */
    modalities?: string[];
    /** Sampling temperature. Default `0.8`. */
    temperature?: number;
    /**
     * Max output tokens per response. Default the string `'inf'` (unlimited);
     * `null` omits the param.
     */
    maxResponseOutputTokens?: number | string | null;
    /** Turn-detection settings, or `null` to disable. Defaults applied when omitted. */
    turnDetection?: TurnDetectionConfig | null;
    /** Input-audio transcription settings, or `null` to disable. Default model applied when omitted. */
    inputAudioTranscription?: InputAudioTranscriptionConfig | null;
    /** Tool-selection strategy. Default `'auto'`. */
    toolChoice?: string;
};
interface ResolvedConfig {
    model: string;
    voice: string;
    modalities: string[];
    temperature: number;
    maxResponseOutputTokens: number | string | null;
    turnDetection: Required<TurnDetectionConfig> | null;
    inputAudioTranscription: Required<InputAudioTranscriptionConfig> | null;
    toolChoice: string;
}
declare class OpenAIRealtimeImpl {
    static readonly displayName = "OpenAIRealtime";
    _providerName: string;
    _isRealtimeModel: boolean;
    apiKey?: string;
    config: ResolvedConfig;
    model: string;
    voice: string;
    params: Record<string, string>;
    constructor(opts?: {
        apiKey?: string;
        config?: OpenAIRealtimeConfig | null;
        [key: string]: any;
    });
    close(): Promise<void>;
    cleanup(): Promise<void>;
    getRuntimeConfig(): Record<string, any>;
}
export type OpenAIRealtime = OpenAIRealtimeImpl;
/**
 * OpenAI Realtime speech-to-speech (S2S) model.
 *
 * @param opts.apiKey OpenAI API key; defaults to the `OPENAI_API_KEY` env var.
 * @param opts.config Realtime configuration; see {@link OpenAIRealtimeConfig}.
 */
export declare const OpenAIRealtime: (opts?: {
    apiKey?: string;
    config?: OpenAIRealtimeConfig | null;
    [key: string]: any;
}) => OpenAIRealtime;
export {};