False wake word detections:
Sadly even “Ok, Nabu” isn’t nearly as reliable on the VPE as Alexa on Amazon Hardware (which in fact doesn’t show any false triggering anymore, as this improved vastly over the last years).
At least if you use the VPE outside your playground in your study like e.g. the living room with all its noise from conversations, radio and tv.
So, I added this to my prompt and it helps a lot.
Important note about mistakenly triggered voice input:
-------------------------------------------------
Most of the time we communicate to you
through audio devices with microphone and speaker.
This sometimes leads to wrong wake word detections.
So, if you get text as input that doesn't make sense
or sounds like we didn't want to talk to you,
cancel the conversation.
Simply use a single space charater as response in this case.
Don't reply with any text or questions.
edit: I missed to explain the LLM in a detailed way how to differentiate between false and correct wake work detections.
This made the LLM often respond in a weird way as it wasn’t sure if the question should be answered or not.
Better use this prompt to get more reliable results:
The VPE simply ends the conversation with a short red blinking ring and no voice response.
(And more important: Not trying to make any sence of the text, followed by starting crazy actions in your smart home. My favorite so far was turning off the lights and starting the vacuum robot …) 