Closed joelis10 closed 4 months ago
Hi , thanks for looking at this. the delay was to just eliminate the screen going back to idle before the response had finished playing, this is purely a display trigger and shouldn't affect the playing of the media response. However this is a better way to handle the display so I will merge that in. Thank you for taking the time to look into this. Cheers Rob
Hi!
I try :)
Thanks
Hello, the display conversation now remains! That's pretty cool but unfortunately the announcement keeps stopping :/
Could you please elaborate on what you mean by the conversation remaining? Does the "listening" page stay and not change until human interaction? And unfortunately I think that cut off might not be related to this project, could you please (if you're able) try flashing the stock Voice Assistant firmware and see if it still happens?
By this I mean the display conversation i.e. the smiling house after the change in yaml. remains in place. But the speech output from ChatGPT continues to stop. In fact, I'm coming from the original firmware from the ESP32 for Home Assistant. Everything ran smoothly here and without any problems, but the speaker was too quiet. That's why I switched to this firmware.
thanks for the update, this looks like it may be something going on with the adf-pipeline , i will ask the question to the dev of that component .
In your YAML, you hardcode a delay of 5 seconds before the listening image disappears in favour of the homepage. Doing so obviously cuts off longer responses, specifically of those from an LLM. This PR should fix that by moving the switch to the homepage to when Assist actually finishes talking, or at least a second after. Also this is dipper998 from Discord by the way and I hope this isn't my first PR here, should have another one at some point with those substitutions I was talking about.
@Unreality81 (if this mentions you) - could you see if this small change fixes your issue (#27)?