contextual-agent-sdk/dist/core/ModalityRouter.d.ts

SDK for building AI agents with seamless voice-text context switching

import { Modality, Message } from '../types';
export interface SpeechToTextProvider {
    transcribe(audioInput: any, options?: {
        language?: string;
        model?: string;
        temperature?: number;
        responseFormat?: 'json' | 'text' | 'verbose_json';
        prompt?: string;
        [key: string]: any;
    }): Promise<{
        text: string;
        confidence?: number;
        language?: string;
        duration?: number;
        segments?: Array<{
            start: number;
            end: number;
            text: string;
            confidence?: number;
        }>;
        [key: string]: any;
    }>;
}
export interface TextToSpeechProvider {
    synthesize(text: string, options?: {
        voice?: string;
        language?: string;
        speed?: number;
        pitch?: number;
        volume?: number;
        format?: 'mp3' | 'wav' | 'ogg' | 'flac';
        sampleRate?: number;
        [key: string]: any;
    }): Promise<{
        audioData: any;
        duration?: number;
        format?: string;
        sampleRate?: number;
        [key: string]: any;
    }>;
}
export interface ModalityRouterConfig {
    speechToText?: SpeechToTextProvider;
    textToSpeech?: TextToSpeechProvider;
    useMockWhenUnavailable?: boolean;
    defaultSTTOptions?: {
        language?: string;
        model?: string;
        temperature?: number;
        responseFormat?: 'json' | 'text' | 'verbose_json';
        prompt?: string;
        [key: string]: any;
    };
    defaultTTSOptions?: {
        voice?: string;
        language?: string;
        speed?: number;
        pitch?: number;
        volume?: number;
        format?: 'mp3' | 'wav' | 'ogg' | 'flac';
        [key: string]: any;
    };
}
export declare class ModalityRouter {
    private isProcessing;
    private config;
    constructor(config?: ModalityRouterConfig);
    detectModality(input: any): Modality;
    processMessage(input: any, modality: Modality, sessionId: string): Promise<Message>;
    private processVoiceMessage;
    private processTextMessage;
    prepareResponse(content: string, targetModality: Modality, sessionId: string): Promise<Message>;
    private prepareVoiceResponse;
    setSpeechToTextProvider(provider: SpeechToTextProvider): void;
    setTextToSpeechProvider(provider: TextToSpeechProvider): void;
    setDefaultSTTOptions(options: NonNullable<ModalityRouterConfig['defaultSTTOptions']>): void;
    setDefaultTTSOptions(options: NonNullable<ModalityRouterConfig['defaultTTSOptions']>): void;
    transcribeWithOptions(audioInput: any, options?: Parameters<SpeechToTextProvider['transcribe']>[1]): Promise<ReturnType<SpeechToTextProvider['transcribe']>>;
    synthesizeWithOptions(text: string, options?: Parameters<TextToSpeechProvider['synthesize']>[1]): Promise<ReturnType<TextToSpeechProvider['synthesize']>>;
    hasSpeechToText(): boolean;
    hasTextToSpeech(): boolean;
    private isAudioInput;
    private mockSpeechToText;
    private getAudioDuration;
    private estimateVoiceDuration;
    private generateMessageId;
    private delay;
    isCurrentlyProcessing(): boolean;
    isModalitySupported(modality: Modality): boolean;
    getCapabilities(): {
        voice: boolean;
        text: boolean;
        speechToText: boolean;
        textToSpeech: boolean;
        usingMocks: boolean;
    };
}
//# sourceMappingURL=ModalityRouter.d.ts.map