Skip to content

Long response generation leads to an error, help me figure it out. #84

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mitrokun opened this issue Mar 11, 2025 · 1 comment
Open
Labels

Comments

@mitrokun
Copy link

mitrokun commented Mar 11, 2025

I'm experimenting with non-stop conversation on the standard esp32 and I noticed this feature: if the response text is generated longer than a certain time, then an error appears in the logs, and the entire logic of the device breaks down.

Several examples with piper:

  • Weak Intel processor in mini PC gives an error after a couple of sentences.
  • AMD 5900x stops working after 2000 characters.
  • Piper on the GPU has not yet caused errors.

I've also encountered the identical problem with cloud TTS.

If you look at the time in the logs, the error occurs 5 seconds after the start of the response mode

[05:09:55][D][voice_assistant:642]: Event Type: 8
[05:09:55][D][voice_assistant:720]: Response URL: "http://192.168.1.162:8123/api/tts_proxy/FhgG4XTYwqaqKyzIxmGtkQ.wav"
[05:09:55][D][voice_assistant:505]: State changed from AWAITING_RESPONSE to STREAMING_RESPONSE
[05:09:55][D][voice_assistant:512]: Desired state set to STREAMING_RESPONSE
[05:09:55][D][media_player:073]: 'esp32VA' - Setting
[05:09:55][D][media_player:080]:   Media URL: http://192.168.1.162:8123/api/tts_proxy/FhgG4XTYwqaqKyzIxmGtkQ.wav
[05:09:55][D][media_player:086]:  Announcement: yes
[05:09:55][D][speaker_media_player:420]: State changed to ANNOUNCING
[05:09:55][D][voice_assistant:642]: Event Type: 2
[05:09:55][D][voice_assistant:736]: Assist Pipeline ended
[05:10:00][D][esp-idf:000][ann_read]: W (3021499) HTTP_CLIENT: Connection timed out before data was ready!

[05:10:00][E][speaker_media_player.pipeline:112]: Media reader encountered an error: ESP_FAIL

I used to use a wyoming satellite and it doesn't have such a strict timeout.

And now please help with advice. How to realize the assistant reset after this error.
The speaker_media_player error has nothing to do with the on_error trigger, so the wakeword is not started

code

p.s.
Hopefully in the future Assist will move to streaming responses to prevent situations like this from happening.

Copy link

As there has been no activity on this issue for 30 days, I am marking it as stale. If you think this is a mistake, please comment below and I will remove the stale label.

@github-actions github-actions bot added the stale label Apr 30, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

1 participant