Deployment

systemd service

Create /etc/systemd/system/hivemind-voice-sat.service:

[Unit]
Description=HiveMind Voice Satellite
After=network-online.target sound.target
Wants=network-online.target

[Service]
Type=simple
User=pi
Environment="PYTHONUNBUFFERED=1"
ExecStart=/home/pi/.venv/bin/hivemind-voice-sat \
    --host wss://your-hive-host:5678 \
    --key YOUR_ACCESS_KEY \
    --password YOUR_PASSWORD \
    --selfsigned
Restart=on-failure
RestartSec=10

[Install]
WantedBy=multi-user.target

Adjust User and the hivemind-voice-sat path to match your installation. If you stored credentials in the identity file, drop --key, --password, and --host.

Enable and start:

sudo systemctl daemon-reload
sudo systemctl enable --now hivemind-voice-sat
sudo journalctl -u hivemind-voice-sat -f

---

Using the identity file instead of CLI flags

Storing credentials in the service unit exposes them in systemctl status output. A cleaner approach is to populate the identity file once and omit the flags:

hivemind-client set-identity \
  --host wss://your-hive-host:5678 \
  --key YOUR_ACCESS_KEY \
  --password YOUR_PASSWORD

Then the ExecStart line simplifies to:

ExecStart=/home/pi/.venv/bin/hivemind-voice-sat

---

Raspberry Pi notes

Pi 4 (4 GB) minimum. Pi 5 is comfortable with mid-size models.

Audio group

The running user needs access to audio devices:

sudo usermod -aG audio pi

Log out and back in (or reboot) for the group change to take effect.

Microphone selection

List devices:

arecord -l          # ALSA devices
python3 -c "import sounddevice; print(sounddevice.query_devices())"

Set the device in ~/.config/mycroft/mycroft.conf:

{
  "microphone": {
    "module": "ovos-microphone-plugin-alsa",
    "ovos-microphone-plugin-alsa": {
      "device_name": "hw:1,0"
    }
  }
}

Or with SoundDevice:

{
  "microphone": {
    "module": "ovos-microphone-plugin-sounddevice",
    "ovos-microphone-plugin-sounddevice": {
      "device": 1
    }
  }
}

Speaker selection

ovos-audio uses the system default ALSA/PulseAudio output. To override:

# check current default
aplay -l
# set default card in ~/.asoundrc
pcm.!default { type hw; card 1; }
ctl.!default { type hw; card 1; }

Recommended local plugins for Pi

Role	Plugin	Notes
VAD	ovos-vad-plugin-silero	Ships with voice-sat, fast on CPU
Wakeword	ovos-ww-plugin-vosk	Ships with voice-sat, small model
STT	ovos-stt-plugin-faster-whisper	pip install ovos-stt-plugin-faster-whisper, use tiny or base model on Pi 4
TTS	ovos-tts-plugin-piper	pip install ovos-tts-plugin-piper, fast neural TTS, no GPU needed

Reduce latency on Pi

• Use a wired USB microphone rather than USB + sound card dongle.
• Prefer the tiny Whisper model for STT on Pi 4. Upgrade to small on Pi 5.
• Set "continuous_listen": false (default) so VAD + wakeword gate STT inference.

---

Autostart without systemd (user session)

# ~/.config/autostart/hivemind-voice-sat.desktop
[Desktop Entry]
Type=Application
Name=HiveMind Voice Satellite
Exec=/home/pi/.venv/bin/hivemind-voice-sat
Hidden=false
X-GNOME-Autostart-enabled=true

---

← Architecture · Home · Troubleshooting →