Files

ethernet 712734436e fix(pm): make bootstrap and bundle ownership explicit

Finish bootstrap uv before PM replaces its store entry. Keep failure
receipts stdlib-only and align the cryptography requirement and override
with the locked version.

Let bundle builders declare launch paths and update ownership. Remove
payload discovery, Store probing, and the unused develop command.
Derive Nix Python from the PM lock and share its provenance stamp.

Document setup, activation, optional dependencies, and distribution
ownership. Targeted Windows tests, relocated runtime launches, Electron
bundling, and bilingual docs builds pass. Native Nix and signed-package
acceptance remain CI gates.

2026-09-08 00:24:51 -04:00

hey_hermes.tflite

…

README.md

fix(pm): make bootstrap and bundle ownership explicit

2026-09-08 00:24:51 -04:00

README.md

Bundled wake-word models

hey_hermes.tflite — the on-device "Hey Hermes" hotword model. This is the default detector for the wake word feature (see website/docs/user-guide/features/wake-word.md); no training or setup is required to say "hey hermes".

Engine: pyopen-wakeword (rhasspy's maintained fork of openWakeWord; Apache-2.0). Runs TFLite via a bundled tensorflowlite_c library — no onnx, no runtime download.
Provenance: trained with the openWakeWord training pipeline (synthetic TTS-generated speech), which produces the .tflite artifact. Redistribution is permitted under the openWakeWord license.
Label: the model registers as hey_hermes (matches the filename).
Runtime: the pyopen-wakeword wheel includes the shared melspectrogram and embedding models. Starting this engine requires no model download. Identical model files alone do not establish identical scores across inference engines or platforms.

To use a different phrase, point wake_word.openwakeword.model at an absolute path to a compatible .tflite model. Hermes does not download models by name, and this engine does not load .onnx files. See the wake-word docs for the training guide and platform limits.