@kairox_ai/tastekit-voice

Pre-1.0 voice-based onboarding for TasteKit — speak your preferences instead of typing them.

Providers

| Provider | Type | Requires | |----------|------|----------| | ElevenLabs Scribe v2 | STT (cloud) | ELEVENLABS_API_KEY | | ElevenLabs Streaming | TTS (cloud) | ELEVENLABS_API_KEY | | Whisper.cpp | STT (local) | whisper-cpp binary | | Piper | TTS (local) | piper binary |

Automatic 3-layer fallback: ElevenLabs → local binaries → text input.

Install

npm install @kairox_ai/tastekit-voice

Usage

# Via CLI (automatically loads voice package)
tastekit onboard --voice

// Programmatic
import { createVoiceIO } from '@kairox_ai/tastekit-voice';

const voiceIO = await createVoiceIO();
const input = await voiceIO.getUserInput();      // Records mic → STT
await voiceIO.onInterviewerMessage('Hello!');      // TTS → speaker
await voiceIO.dispose();

Requirements

• sox — required for microphone recording (brew install sox)
• ElevenLabs API key — for cloud STT/TTS (optional, falls back to local)
• whisper-cpp — for local STT (optional)
• piper — for local TTS (optional)

Part of TasteKit

• CLI: @kairox_ai/tastekit-cli
• Core: @kairox_ai/tastekit-core

License

MIT