A few of you have helped me here and there over the last year as I’ve been planning the details around the tech in my new house.
A big “The House Home Assistant is Building” post is in the works but we’re talking tablet panels instead of light switches in every room, all Hue bulbs, over 100 Aoetec multi-sensors, powered shades, Unifi networking hardware, his and her smart mirrors in the master bath, over a dozen cameras, etc., etc., etc, all powered by HA. My last count for CAT6 drops was 250 (lots of PoE!). Hosting it all on a powerful NUC (or two) with a list of docker containers as long as my arm. The geek in me can barely sleep thinking about it all. And it’s been a blast setting up a small test area in my home. Not surprising to this group, I’m sure, my priorities are local control, privacy, future-proofing, and all the other values embodied in HA.
Super duper exciting, with one big issue: voice. Until VERY recently, I believed that the best option for voice control was to stick a Google Home Mini in every ceiling. Nothing about that says local control, privacy, future proofing, or anything else. But what’s a guy to do?
Then I read about Ada and Almond. Then I read about Rhasspy. Then I installed Rhasspy. And now I see the future. I see a manageable voice setup where I’m in complete control, the hardware is disconnected from the software, and I can make smart choices now (when the walls are still open!) that allow me to upgrade into the reasonable future.
I have about 6 months (June 2020) before I have to commit to the hard details (we just broke ground and it’s a big build) but, specifically, based on the information I have available right now, it looks like this to me:
- ReSpeaker 6-Mic Circular Array + Raspberry Pi (PoE) in every room/area (36 will make every square foot of the house easily in range to a microphone)
- Ceiling mounts for each ReSpeaker+Pi
- Dedicated NUC for Rhasspy and other voice related Docker containers
- Node-RED for all voice automations
- TTS output on the closest tablet panel to the voice input (I plan on using HACC for the display which I see becoming something really amazing)
Here are the issues I see right now:
- Rhasspy does not yet support more than one mic (and I need 36!), but there is a ticket in for this improvement (thanks @synesthesiam)
- There is no ceiling mount for a ReSpeaker+Pi combo
- HACC is still in its infancy and I don’t believe there is, as of yet, a roadmap for supporting TTS output
I see the single-satellite issue with Rhasspy being solved in the near future.
I have no ideas for a ceiling mount outside of 3D printing my own (big learning curve there).
I need to buy the HACC guy a LOT of coffees.
Anyone else see any additional holes in this plan? Any advice for improvement?