Shorter Responses

Know of anyway to make Mycroft have shorter responses for the workflow of it not understanding what you said? This is on a Mark 2 using Neon OS

I used to have Alexa setup to when it didn’t understand a command it would just make a single beep in a lower tone that would denote no success in processing or understanding.

Lately the Mycroft will go on about using a Language Model, and sorry about not understanding, etc etc after Saying Sorry I don’t understand.

I am on the newest stable version as well.

I tried a couple of commands to “Use Limited Responses” and “Talk to me Faster” but the device gets going and I happen to be in a meeting and it thinks it hears a wake word, so I have to unplug because it will talk for what seems like 30 - 60 seconds non stop over me.

I appreciate it.

in OVOS you would just uninstall the fallback-unknown skill to get the error sound on failure

under Neon I’m not sure how to best do it, seems the Neon LLM skill is answering your questions so maybe open an issue there

@JarbasAI is correct, if it starts talking about being a language model, it’s activated an LLM fallback skill. You can turn that off with the command “disable LLM fallback skill” and I’ll add your request for an even shorter response option to our community wish list. Perhaps we can do that in two steps, first increasing the shortness of the responses for “Use Limited Responses” in some skills, and the second to add an optional setting for a simple auditory feedback for failures.

Thank you for the detailed feedback and suggestions, I’m pleased to hear it’s getting good enough for you to use it regularly!

Rock on, I appreciate both responses.

When I saw JarbasAI’s response I went to the NeonOS website to look for a way to submit an issue request, but didn’t see anything that stuck out to me, sorry.

I have executed the command NeonClary recommended, and we’ll see if it is successful, will let you know for sure. It was funny because as soon as I did it, it went into a lengthy explanation of “I apologize if my previous response…” so on and so forth. Hopefully that will be the last time she apologizes.

Yep, it works well for regular use for me. I have about 6 of them that I use throughout and have been tweaking them based on the area they are in. Next is configuration of Home Assistant.

Only real two things that i can see now as non parody functions from the Echoes I used to use is identification and activation of only the one closest to you when you say the wake word and Plex music streaming integration.

Everything else like weather, time, measurements, news, and general questions seems to work well and the newest version seems to have really picked up the snappiness of responses.

I appreciate it again.

1 Like

@3rd if you successfully disable the LLM fallback skill, Neon will confirm by saying “Okay, I will not send requests to an LLM”.
If it didn’t say those words, it probably didn’t quite get the command interpreted correctly. and I’d try telling it “disable LLM fallback skill” again until you get that confirmation response.

(There isn’t yet a programmed way to have Neon report whether the LLM fallback skill is active or not, but it will give the same confirmation sentence if it is already disabled when you give the command.)

You’re the second person to mention the linked devices feature, and I agree it’s a useful one, but I think it’s going to take some significant additions to the system.

I did mention to the other person who had two devices that since there are two wake words now, you could change the wake word of one of the devices. You’d of course have to remember which was which, so that’s not super practical as it is.

However, we’ve been considering a custom wake word capability. It just occurred to me giving each its own wake word according to location might be another interesting path to that functionality - “Living Room Neon” or “Kitchen Neon” or “Downstairs Neon” would qualify as fairly distinctive wake words. What do you think of that, might it be a practical alternative solution?

Did you want just the audio from Plex, or the video as well?

Yeah I figured as much with the response state, as I would imagine each would need to be aware of the other, and they would do something like a compare and contrast so the one with the louder sound recording would be able to respond.

Being able to say Hey Kitchen Mycroft or Hey Living Room Mycroft may be a good alternative, a little long, but would potentially solve the issue, as long as I remember which is what :stuck_out_tongue:

For Plex it’s just music streaming I am mentioning. I don’t have any services like Spotify or anything, I use Plexamp for most of my music needs, since it also integrates with Android Auto and such. I think the Plex server can also be used as a DLNA streaming server also, for music specifically.

Honestly, streaming from any platform that hosts locally would be awesome, I can setup an Emby or Jellyfin if those are easier, I already have a Plex and Logitech Media Server running on the network, but could add anything additional if there was an option for streaming music locally.

The location aware response is a like to have but know that’s a tough one and probably right now not a large enough base to justify the development time to achieve, but the locally streaming music would be the really nice to have. With Bluetooth now though I can technically bluetooth connect my phone and then play through it to the Mycroft, but would be nice to be able to change songs and such with voice while in the Kitchen cooking or Living Room cleaning.

I appreciate it though, I’ll let you know if anything happens with the disable LLM. It seemed to take the command, cross fingers.

1 Like

I have a partially finished fork of a Plex music skill that also does video, although I wouldn’t recommend streaming video on a Mark 2 - I tried it, much too slow! GitHub - mikejgray/skill-plex at feat-minor-improvements

It’s a work in progress but I’ll happily accept feedback.