Skip to content

First Run

Start Kiwi

source venv/bin/activate
python -m kiwi

.\start.ps1
# or
.\venv\Scripts\activate
python -m kiwi

source venv/Scripts/activate
python -m kiwi

You should see startup logs like:

[14:08:25.342] [INFO] [CORE] Kiwi Voice starting...
[14:08:25.500] [INFO] [STT] Loading Faster Whisper (small, cuda)...
[14:08:26.100] [INFO] [TTS] Kokoro ONNX initialized (14 voices)
[14:08:26.200] [INFO] [API] Dashboard at http://localhost:7789
[14:08:26.300] [INFO] [CORE] Listening...

Open the Dashboard

Navigate to http://localhost:7789 in your browser. You'll see the real-time dashboard with live status, event log, controls, and more.

Kiwi Voice Dashboard

Register Your Voice

Say the wake word followed by the registration command:

"Kiwi, remember my voice"

Kiwi will capture your voiceprint and register you as the Owner — the highest priority level with full access.

Important

Register your voice first. Until you do, all speakers are treated as guests and may need Telegram approval for certain commands.

Try a Command

Once registered, try:

"Kiwi, what time is it?"

The flow:

  1. Wake word "kiwi" is detected (text fuzzy match or ML model)
  2. Your speech is transcribed by Faster Whisper
  3. Speaker ID confirms you're the Owner
  4. Command is sent to OpenClaw via WebSocket
  5. LLM response streams back
  6. TTS speaks the answer

Use the Web Microphone

If you don't have a local microphone (or don't want to install pyaudio), click the microphone button in the dashboard to talk to Kiwi directly from the browser.

What's Next