I’m new to the HA world, but have given myself a mission to replace the Alexa devices in the house with something have more control over.
I currently have a Pi4 running HAOS, and Piper and Whisper running in containers on my NAS (which happens to also be a giant bucket of x86 cores). Voice is working fine with an Atom Echo running ESPHome, but I’d like to migrate to Rhasspy with a couple of satellites so I get wakeword detection. I have the Rhasspy3 addon installed, and can get a satellite pointed at it (and wakeword detection is working), but I’m a bit lost as to how to configure Rhasspy3 to use the Wyoming servers for TTS and STT.
Could someone point me at a config snippet (or doc that I’ve missed) that shows how to do this?
HA Voice Assist is essentially Rhasspy v3 (well, when I upgraded to HA Voice Assist Chapter 5 last month the URLs were the rhasspy GIThub repository ), and Mike’s focus has been mostly on the HA implementation.
Are you wanting local wakeword detection on an Atom Echo ? My understanding is that the Echo just doesn’t have enough power to do that well enough - so wakeword detection is done on the HA server. It is maybe a little slower, and sends all audio over the wi-fi - but does the job.
Or which hardware device are you wanting to use as a satellite ? You didn’t say. The S3 Box and Raspberry Pi already can run openwakeword locally.
Or are you confusing Rhasspy (the voice assistant software) with Raspberry Pi (the hardware device) ?
If you are wanting to use Rhasspy with Home Assistant, I suggest that Voice Assist in the HA forum will give you more and better information.
I haven’t touched this project since last September, but now wouldn’t be a bad time to pick it up.
I’m not particularly set on hardware. I’d set up the Atom Echo as an easy PoC, but would be perfectly happy if the devices in the living space had to be Pis of some variety, and I have a ‘Voice Bonnet’ I was going to use for testing.
I happened to set up Piper and Whisper in containers on the NAS because my HA machine is only a Pi4, but the NAS has a pair of E5-2690V4s.
Given the hardware described above, what would you recommend as a clean-sheet setup for voice control?
For anyone else wanting to setup a Voice Assist satellite, I think the RasPi with Mic HAT option is more expensive and the drivers for the voice HATs do not provide the level of benefit implied by the hardware.
For anyone wanting a new satellite I currently would recommend the ESP-32-S3 BOX 3 … but I would wait and see if Year of the Voice Chapter 6 makes any better recommendation.
I personally upgraded my HA machine from a RasPi4 to a virtual machine on a used business micro-PC. I installed HAOS because it has best support. HAOS already uses Docker, and sets up the various add-ons in Docker - so I get the benefits of using Docker without having to learn how to drive it. And I can still setup other VMs as i need.
I am very impressed with HA Voice Assist (aka rhasspy 3 integrated with HA), though there are still a few things i’m hesitant about - mainly that there is no clear recommended satellite, that Jarvis recognises my commands less than half the time, and that I haven’t yet looked into linking my Voice Assist into node-RED where i have all my automations
@twin My Home Assistant is still using Rhasspy 2.5, and I have added Piper and Whisper Addons which gives Voice Assist (aka rhasspy 3) in my study. Not quite ready to switch the rest of the house over to Jarvis yet, though
Hello,
I know, this is not really a rhasspy problem, but maybe someone could help.
state now
rhasspy
I’m using HA with rhasspy 2.5, AppDaemon and several satellites.
After a satellite has recognized a wakeword via mqtt (hermes/dialogueManager/sessionStarted) my TV is muted, so the following command is recognized very well.
After AppDaemonhas finished processing the command, it is answering via mqtt (hermes/dialogueManager/endSession)
HA assist
I would like to change to HA assist and wyoming protokol.
Right now I managed this:
sentences.yaml → intent_script.yaml → firing event with data → appDaemon → processing
I’m not sure, if this is the best way to react on assist with appDaemon
Request
Is this way to react on assist with appdaemon the right way, or can appDaemon directly react (without intent_script.yaml)?
Is there a way to answer to a satellite from AppDaemon (wyoming??) ?
), and Mike’s focus has been mostly on the HA implementation.