#Issue Setting up Voice Assistant
1 messages ยท Page 1 of 1 (latest)
If it helps, every time I click "Set up voice assistant," the LED ring goes dark and doesn't come back until I unplug/replug the device.
**Update: **after a couple of tries, it did update, but now every time I try to set it up, I get "hello" spinning for a while, the LED ring turns off, then I get "unable to connect." If I hit "retry," the LED ring spin a very bright white for a moment, then off, then I'm back to square one.
The device is working; it lights up when I say "ok Nabu," so apparently it's only the "Set up Voice Assistant" that's not working?
what type of home assistant install do you have?
is your network a standard setup or do you have anything complex going on? (e.g. vlans)
the server is running in a Hyper-V container on my laptop. I just installed that today via these instructions.
There's nothing out of the ordinary going on. I'm able to access it via the web and everything through Home Assistant Cloud as well.
(I'm gonna move the server to a NUC at some point, though, if that matters)
moving to a nuc in the future is a good plan. but for now running on hyper-v shoud be mostly fine. I ask because the VPE requires connections to be able to be established both ways between HA and the VPE. and just trying to work out the network layout as that is one of thee common issues.
can you access the HA locally fine from other systems other than the hyper-v host?
the HA server? yeah, I can get to it via the local address on my local machine. The device itself? No idea. HA can; I can change settings like the LED lights and such.
I think the issue has to do with outputting audio. The wake word works, but it doesn't say anything when I give a command.
your local machine is not the hyper-v host? i am checking another system on the network can access HA via local network (if so then the vpe probably can)
when the wakeword is heard. does it "listen" (leds slow clockwise spin) or do the leds just go off?
trying to work out exactly where the issue is in the flow.
It is. It's a bridged connection.
I don't know it's it's listening, actually. The logs aren't saying anything about what it picks up, so maybe not?
The LED ring does spin one direction when I say the wake word and the other when I speak to it.
could you record a short video demo of you trying to use it? in a way i can see the led patterns throughout.
This is what I see in the logs, actually. And yeah, I'll drop something.
It outputs sound, but not voice, I guess.
ok, go to settings->voice assistants
then on your voice pipeline, hit the 3 dots and go to debug
stage: done
run:
pipeline: 01ke8ndv98xyn5pdvtbwv66fjv
language: en
conversation_id: 01KE8RHDQH2FTJQ8ZYQP0RHP57
satellite_id: assist_satellite.home_assistant_voice_0a1023_assist_satellite
tts_output:
token: 11AbWT1wHvlH5eSzbGaaxA.flac
url: /api/tts_proxy/11AbWT1wHvlH5eSzbGaaxA.flac
mime_type: audio/flac
stream_response: false
events:
- type: run-start
data:
pipeline: 01ke8ndv98xyn5pdvtbwv66fjv
language: en
conversation_id: 01KE8RHDQH2FTJQ8ZYQP0RHP57
satellite_id: assist_satellite.home_assistant_voice_0a1023_assist_satellite
tts_output:
token: 11AbWT1wHvlH5eSzbGaaxA.flac
url: /api/tts_proxy/11AbWT1wHvlH5eSzbGaaxA.flac
mime_type: audio/flac
stream_response: false
timestamp: "2026-01-06T04:22:53.792943+00:00"
- type: stt-start
data:
engine: stt.home_assistant_cloud
metadata:
language: en-US
format: wav
codec: pcm
bit_rate: 16
sample_rate: 16000
channel: 1
timestamp: "2026-01-06T04:22:53.793123+00:00"
- type: stt-vad-start
data:
timestamp: 1400
timestamp: "2026-01-06T04:22:55.219843+00:00"
- type: stt-vad-end
data:
timestamp: 2650
timestamp: "2026-01-06T04:22:56.470141+00:00"
- type: stt-end
data:
stt_output:
text: Never mind.
timestamp: "2026-01-06T04:22:56.616348+00:00"
- type: intent-start
data:
engine: conversation.home_assistant
language: en
intent_input: Never mind.
conversation_id: 01KE8RHDQH2FTJQ8ZYQP0RHP57
device_id: 8ae13ce2cff136873b8cd790c73a6e99
satellite_id: assist_satellite.home_assistant_voice_0a1023_assist_satellite
prefer_local_intents: false
timestamp: "2026-01-06T04:22:56.616612+00:00"
- type: intent-end
data:
processed_locally: true
intent_output:
response:
speech: {}
card: {}
language: en
response_type: action_done
data:
targets: []
success: []
failed: []
conversation_id: 01KE8RHDQH2FTJQ8ZYQP0RHP57
continue_conversation: false
timestamp: "2026-01-06T04:22:56.624829+00:00"
- type: run-end
data: null
timestamp: "2026-01-06T04:22:56.624972+00:00"
started: 2026-01-06T04:22:53.792Z
stt:
engine: stt.home_assistant_cloud
metadata:
language: en-US
format: wav
codec: pcm
bit_rate: 16
sample_rate: 16000
channel: 1
done: true
stt_output:
text: Never mind.
intent:
engine: conversation.home_assistant
language: en
intent_input: Never mind.
conversation_id: 01KE8RHDQH2FTJQ8ZYQP0RHP57
device_id: 8ae13ce2cff136873b8cd790c73a6e99
satellite_id: assist_satellite.home_assistant_voice_0a1023_assist_satellite
prefer_local_intents: false
done: true
processed_locally: true
intent_output:
response:
speech: {}
card: {}
language: en
response_type: action_done
data:
targets: []
success: []
failed: []
conversation_id: 01KE8RHDQH2FTJQ8ZYQP0RHP57
continue_conversation: false
finished: 2026-01-06T04:22:56.624Z
ok so you dont have any TTS specified
That would help ๐
This?
yup
After clicking "Update," this happens:
stage: done
run:
pipeline: 01ke8s3jt5b3y1p69mj116f0q3
language: en
conversation_id: 01KEAB0EF00Y8QWT8CD9XDK1V4
satellite_id: assist_satellite.home_assistant_voice_0a1023_assist_satellite
tts_output:
token: D9IhwHHC9Q3mwnxZODx5NA.flac
url: /api/tts_proxy/D9IhwHHC9Q3mwnxZODx5NA.flac
mime_type: audio/flac
stream_response: false
events:
- type: run-start
data:
pipeline: 01ke8s3jt5b3y1p69mj116f0q3
language: en
conversation_id: 01KEAB0EF00Y8QWT8CD9XDK1V4
satellite_id: assist_satellite.home_assistant_voice_0a1023_assist_satellite
tts_output:
token: D9IhwHHC9Q3mwnxZODx5NA.flac
url: /api/tts_proxy/D9IhwHHC9Q3mwnxZODx5NA.flac
mime_type: audio/flac
stream_response: false
timestamp: "2026-01-06T19:05:41.936981+00:00"
- type: stt-start
data:
engine: stt.home_assistant_cloud
metadata:
language: en-US
format: wav
codec: pcm
bit_rate: 16
sample_rate: 16000
channel: 1
timestamp: "2026-01-06T19:05:41.937305+00:00"
- type: error
data:
code: stt-stream-failed
message: Unexpected error during speech-to-text
timestamp: "2026-01-06T19:05:43.365160+00:00"
- type: run-end
data: null
timestamp: "2026-01-06T19:05:43.365879+00:00"
started: 2026-01-06T19:05:41.936Z
stt:
engine: stt.home_assistant_cloud
metadata:
language: en-US
format: wav
codec: pcm
bit_rate: 16
sample_rate: 16000
channel: 1
done: false
finished: 2026-01-06T19:05:43.365Z
error:
code: stt-stream-failed
message: Unexpected error during speech-to-text
ok, is there a "play audio" button on the debug screen?
oh wait, it doesnt get that far
so, i suspect that its failing to start the pipline because the VPE is not able to establish a connection to the ha server (this ability is needed for tts)
So the HA server can't reach the internet, you mean?
no
for tts works. HA server needs to be able to host a file which the vpe them connects to and streams
Can I set that up locally instead?
I'd rather do it that way, anyway.
Are there docs on that?
we can possibly solve the issue ๐
Well yeah, working is better right now ๐
so the HA server is virtualised via a bridged connection so it get its own ip on your network correct?
what are the options on your "default switch"?
FYI: its been a while since i touched hyper-v
me, too.
I'd rather just do it in Docker, but that seemed like more of a pain that it was worth right now.
vm is connected to the virtual switch but how the virtaul switch is connected to the network is the question
I'm figuring out how to find that ๐ซฅ
I made another "External" switch. Checking now.
SUCCESS!
This should be updated. Do you want a PR? ๐
That's not super clear, even for me.
yeah "bridged" is how virtualbox calls an external connection so the language is a bit off perhaps.
you could pr if you want to, or just feedback. stuff for it is at the bottom of the page.
tbh, hyper-v setups are not that common
Yeah, and I'm gonna have to reset everything when I move and get it all on a NUC anyway.
New wifi and all
with an internal connection it was making a nat most likely which meant that the vpe couldnt connect in. so when switching to external it was then on the same real network so could see it
depending on spec of nuc, going virtualised can be good but should go with something like proxmox
Also, is there a way to do custom wake words?
so... technically yes...
how important is it for you?
i ask because its going to get easier in the future so if it can wait then its best to leave it for now.
I'll wait then. I'm willing to go through the hassle but it'll be a month before I can get around to it anyway
a lot of guides and documentation around custom wake words are out of date btw so dont be falled by some of it.
the current best way of making a model is using THIS project but then you have to host it then take control and customise the firmware of your VPE to pull the model and compile it in.
there is a plan to make models loadable from the HA server but its not ready yet
Oh sweet
the feature is in the backlog repo - https://github.com/OHF-Voice/backlog-issues/issues/54
some of the groundwork has been done but its not working yet
can track progress from that issue. but realistically you will probably see it in patch notes when its ready anyway.
Thanks so much for the help