So i have a pretty nice(i7-6850k, 32Gb ram, GTX 1080, 16 TB HDD) server (Ubuntu 17.04) set up here that basically just being used as a NAS. while the picroft is fine i was thinking of making my home server my main mycroft System and trying to make raspberry pi nodes throughout the house, i would imagine this could help latency and give it some more ram to work with. I got this idea from the echo dots. since most of the computational power comes from the pi, maybe a beefier computer could help this out… OR im completely wrong and thats fine too.
This is something I would love to see Mycroft do. I think at this point, there’s no option for a back end other than what you find in the web portal (at least not without doing some of your own coding to make it work), but I would love to be able to run the back end server stuff out of my house. I realize that this requires a lot more computing power than your average Raspberry Pi or even desktop computer, but still, it would be a nice option. Maybe Mycroft could partner with someone like Pogo Linux to produce a rack mounted server to run the back end processes. This would be in line with the privacy focus Mycroft has and reduce load for Mycroft’s own servers. If the end user is willing to front the cost of that kind of hardware, it would be an amazing option to have available. You’re not going to see Google or Amazon doing that.
mycroft in the cloud available by https request? - check
mycroft in the cloud available by websocket? - check
mycroft in the cloud available by javascript? - check
routes answers only to source of question? - check
mycroft to mycroft communication? - check
nothing goes to mycroft-home? - check
tested, debugged and finished? - WIP
personal “mycroft home” backend? - WIP
Ive seen some people use the Jarvis AI and make it an in home system. it would be pretty cool
Is it possible to, in the future, be able to place a RPI in each room that is just a remote speaker/mic off of the Full Mycroft install?
Supposedly you can point a sub instance at the main’s websocket, and then config the sub to do tts/stt/etc on the main host. I haven’t done that ever, but…maybe you can.
There’s a lot that’s possible with Mycroft - one of the many benefits of having an open design and architecture.
The specific challenges that would need to be overcome here would be:
- Some sort of system to take a recording from a remote speaker and pass to a base unit securely
- A way to reduce any latency this introduces
- A way to tell the base unit that the output (ie TTS) should be redirected to the remote unit