My plan is to run this on a local device (probably a Raspberry Pi) and add server-side functionality to get around the browser limits. The rationale is that I can write my own commands that better fit my needs and which have access to local devices without giving all of my personal data to the ad giants.
If you're wondering why it's browser-based in the first place, it's because the browser solves the problem of UI, speech API, and won't require any installation/updates on clients (just the server).
> All of the speech recognition and command dispatching is done client-side so that no data is sent to any server unless an external service is required
How do you do the speech recognition in the browser client-side? Afaik the only browser with in-build speech recognition is Google Chrome, which sends the audio stream to Google for recognition.
Wow! I didn't realize that Chrome sent the audio stream to Google. Thank's for the heads-up.
Edit: This means that I may have to look into adding the recognition server-side locally. FireFox had a working experimental API at one point but it's gone now: https://bugzilla.mozilla.org/show_bug.cgi?id=1248897
If you're wondering why it's browser-based in the first place, it's because the browser solves the problem of UI, speech API, and won't require any installation/updates on clients (just the server).