Numen uses dotool and the Vosk speech recognition library.
For reference, these are the build scripts for Alpine:
And these are the build scripts for Void:
Thank you!