Skip to main content

Getting Started

Get Kith running in under 5 minutes.

Prerequisites​

  • Bun >= 1.3 (also works with Node.js 18+)
  • Python 3.11+ with uv (for the Pipecat sidecar)
  • An ElevenLabs API key

Install​

bun add @kithjs/core @kithjs/runtime-pipecat @kithjs/voice-router

Or with npm:

npm install @kithjs/core @kithjs/runtime-pipecat @kithjs/voice-router

Optional packages:

bun add @kithjs/observability    # traces + dup-send guards
bun add @kithjs/runtime-livekit  # LiveKit adapter (mock mode in v0.1)

Set Up the Python Sidecar​

The Pipecat runtime needs a Python virtualenv with the sidecar installed:

cd node_modules/@kithjs/runtime-pipecat/python
uv venv --python 3.11
source .venv/bin/activate
uv pip install -e .

Minimal Example​

import { PipecatRuntime } from "@kithjs/runtime-pipecat";
import { VoiceRouter, DEFAULT_ENGLISH_SLANG } from "@kithjs/voice-router";

const runtime = new PipecatRuntime({
  pythonPath: "node_modules/@kithjs/runtime-pipecat/python/.venv/bin/python",
  cwd: "node_modules/@kithjs/runtime-pipecat/python",
  config: {
    pipeline: "elevenlabs",
    apiKey: process.env.ELEVENLABS_API_KEY,
    voiceId: process.env.ELEVENLABS_VOICE_ID,
    modelId: "eleven_v3",
  },
});

await runtime.connect({ sessionId: "hello-kith" });

const voice = new VoiceRouter({
  runtime,
  slang: DEFAULT_ENGLISH_SLANG,
});

voice.on((event) => {
  if (event.type === "tts_audio_chunk") {
    console.log("Got audio chunk:", event.audioB64.length, "bytes");
  }
});

await voice.speak("Hello from Kith! This is sentence-aware TTS.");
await runtime.disconnect();

Next Steps​

  • Tutorial — Build a full companion with character profiles, emoji reactions, and laugh tags in 15 minutes
  • Architecture — Understand how the pieces fit together
  • Voice Characters — Define your companion's voice personality