Getting Started
July 31, 2026 · View on GitHub
Prerequisites
Hardware
- A microphone accessible to the OS (USB, 3.5 mm, or built-in).
- A speaker or headphone output for TTS playback.
- Enough compute to run STT and TTS models locally. A Raspberry Pi 4 (4 GB) is the practical minimum; Pi 5 or any x86 machine is comfortable.
Software
- Python 3.10 or later.
- A running HiveMind-core instance (the "hive") reachable from this device.
Install
pip install HiveMind-voice-sat
Linux — adds ALSA and SoundDevice microphone backends:
pip install "HiveMind-voice-sat[linux]"
macOS:
pip install "HiveMind-voice-sat[mac]"
Default plugins installed alongside the core package:
| Role | Default plugin |
|---|---|
| VAD | ovos-vad-plugin-silero |
| Wakeword | ovos-ww-plugin-vosk |
| STT | ovos-stt-plugin-server |
| TTS | ovos-tts-plugin-server |
You can swap any of these — see Configuration.
Pairing: get an access key from the hive
On the machine running HiveMind-core, add a client entry and note the credentials it prints:
hivemind-core add-client --name my-voice-sat
Example output:
Access Key : abc123...
Password : def456...
These two values identify this satellite to the hive. Keep them. The server does not store them after display, but you can re-list them with hivemind-core list-clients.
First run
Pass the credentials and host directly on the command line:
hivemind-voice-sat \
--host <hive-host-or-ip> \
--key abc123... \
--password def456... \
--port 5678
Or store the identity once so you do not repeat the flags:
hivemind-client set-identity \
--host <hive-host-or-ip> \
--key abc123... \
--password def456...
hivemind-voice-sat # reads ~/.config/mycroft/identity2.json
If neither credentials nor an identity file are present, the satellite falls back to GGWave. It listens for an audio-encoded identity broadcast from the hive (see hivemind-ggwave).
Verify it works
-
Watch the log output. A successful connection prints:
HiveMind Voice Satellite alive. HiveMind Voice Satellite is ready. -
Say your wakeword (default: Hey Mycroft). You should hear a chime.
-
Ask a question: "What time is it?" The hive processes the utterance and the satellite speaks the response.
If you see connection errors, consult Troubleshooting.
Next steps
- Swap STT/TTS/wakeword plugins: Configuration
- Run as a system service: Deployment
- Understand the data flow: Architecture