ovos-ww-plugin-wakeforge

An OVOS wake-word plugin for custom models trained with wakeforge.

Wakeforge trains a wake-word detector from a single phrase and exports a two-file ONNX pipeline: a feature extractor and a classifier head. This plugin loads that pipeline and runs it as an always-on hotword engine. The runtime uses onnxruntime and numpy only. It does not need PyTorch.

Install

pip install ovos-ww-plugin-wakeforge

Train a model

pip install wakeforge
wakeforge-quickstart "hey jarvis" ./hey_jarvis
# → ./hey_jarvis/best_f1_featurizer.onnx  +  ./hey_jarvis/best_f1.onnx

Configure

In mycroft.conf, point a hotword at the two ONNX files. Use local paths or URLs.

{
  "listener": {
    "wake_word": "hey_jarvis"
  },
  "hotwords": {
    "hey_jarvis": {
      "module": "ovos-ww-plugin-wakeforge",
      "listen": true,
      "featurizer": "~/hey_jarvis/best_f1_featurizer.onnx",
      "model": "~/hey_jarvis/best_f1.onnx",
      "threshold": 0.5
    }
  }
}

Config keys

URL models are cached under ${XDG_DATA_HOME}/wakeforge/.

Related projects

• TigreGotico/wakeforge trains the ONNX models this plugin loads.
• OpenVoiceOS/ovos-plugin-manager loads and manages this plugin alongside other STT, TTS, and wake-word plugins.
• OpenVoiceOS/ovos-ww-plugin-precise-onnx is a sibling wake-word plugin that runs Precise models through ONNX.

Credits

Developed by TigreGótico for OpenVoiceOS.

This project was funded through the NGI0 Commons Fund, a fund established by NLnet with financial support from the European Commission's Next Generation Internet programme, under the aegis of DG Communications Networks, Content and Technology under grant agreement No 101135429.

License

Apache-2.0