Apologies for a delayed response Aaron, not sure how I missed this thread.
Picroft is a fantastic DIY kit but it does have its challenges. Audio issues are of course top of that list. With the wide range of audio devices it is not possible for us to support them all. Even with well established hardware like the Google AIY Kit we sometimes run into trouble, and because these rely on both Mycroft and 3rd party software, we cannot control if something is going to break.
I’ve been thinking about how we can improve our documentation to better direct people to the guides and fixes that the community have posted. As you said, sifting through the many threads here and in Chat can be a hair puller, it would be good to draw them out and incorporate them into our main docs or at least link directly to them from a central source.
In terms of stability and breaking changes, I can confidently say that from 19.02 onward, Mycroft will be far more stable now that we have a much more mature core. I know Steve is also very keen to get a new 19.02 Stable Picroft image out as soon as he can find a few spare moments.
Particularly through it’s early days, Mycroft has been evolving extremely rapidly and this does at times mean that certain breaking changes must be made. We try where ever possible to provide backwards compatibility and flag things that will be deprecated six months in advance to give time for people to make necessary changes, but this can never work 100% of the time and requires skill authors to put work into something they may have not thought about for a while. One of the great things about our model is that currently all Skills in the Marketplace must have an open license, so if something does stop working, even if an author is no longer around, anyone can fork the Skill and re-purpose it how they want. This is one of the strengths but also one of the challenges of open source - with great power comes great responsibility…
A few more specific things that got mentioned:
- Hey Mycroft is likely to be the best performing wake word as it has the most number of people contributing to the Open Dataset and tagger samples. This may vary as the team tries various optimisations but should become more consistent over time.
- For default audio levels, you can add a default level to your volume skill via mycroft.conf though I’m wondering if we should add it as a setting through home.mycroft.ai
Overall I’m glad that people can raise these frustrations openly here, and hopefully we can find some better ways to handle the pain points that do arise.