Device prompt/memories

[TheBest-Name]

When I speak to my assistant, it doesn't seem to know what player it is (Without prior context). In my home assistant I have music assistant, so I turned off the visibility of the media player that it was built with and turned it on in Music Assistant. It would be nice to be able to control a system prompt per device so I can tell it what it is. For example, I can tell it when I ask to play music, it will know that I'm talking about its music assistant media player. Also, I would like to be able to tell it that if it doesn't know what it's talking about in a conversation, and it sounds like it's just a conversation between two people, to just ignore it and stop. That way it doesn't just listen in and keep annoying people.

Another thing I would like is basically the same thing, but have memories per device so it can remember previous conversations you had on that device. I think that would just be a little bit better and give a little bit more personality to the assistant.
Another thing is if #240 gets implemented, then memories per person, per device, or even not per device would actually be a whole lot better, because then it would give the assistant a whole bunch more personality, and it would even give it personality per person. It'd almost be like it would be a different agent for each person, just based off what it would remember for that person.

[nielsnl68]

do you have an ESPHOME device with Voice Assistant possibilities?
When so could you test this on that device as well?

[TheBest-Name]

I'm not actually sure what you were saying. If you are saying if I have this Linux Voice Assistant, then I have it on a Raspberry Pi 0 W 2. Is that what you were saying?

[nielsnl68]

Sorry, indeed it was a bit unclear; what i mend was do you have a spare ESP32 device where you could hook an I2S audio device on it. Or a combi device like a M5stack Atom echo something alike.
And use ESPHome to config it the way you are using the PI setup.

[TheBest-Name]

No, I'm sorry. I only recently got the Pi 0, so I am kind of just using that with this installed using docker. I wouldn't really want to have to pay extra to have a new edition. Does what you're making have to have an ESP device?

[TheBest-Name]

Would I need a device to attach to have memories and stuff added to the LLM input?

[nielsnl68]

Would I need a device to attach to have memories and stuff added to the LLM input?

No not at all, i just wanted to know of the same issue assist with an original setup so we could compare the results and see where the issue you described is coming from.

[TheBest-Name]

Yes, I only have the one assistant right now or one device at least, I can still ask through the web interface.

[TheBest-Name]

What would you like me to do now? I'm not sure why this got turned into a discussion instead of an issue. It makes it a little bit harder to navigate now because I'm not sure where I should make a new post or whether I should reply to an existing one. But I would like to continue this if possible because I would still like this feature.

[TheBest-Name]

Another thing I would like to do is to put in the system prompt if I could that if it would get the impression that it should stop, then it should just go ahead and stop. Because sometimes it will start and I was it just happened just a while ago that it started and I wasn't in the room and it just kept repeating and repeating trying to get a response.

[TheBest-Name]

Does each device already have a built-in system prompt? Is it not able to be customized if it does?