Just my two cents, as a community developer who has been working on skills and Mycroft GUI development for a long time, I will not be porting any of my skills over to Dinkum. More accurately, I cannot port my skills over even if I wanted to because of Dinkum’s technical limitations:
Dinkum core does not fully adhere to the Mycroft GUI protocol. Most of the protocol level has been crippled down core side to only allow certain functionality needed to show some basic pages.
Dinkum provides a very limited, cut-down GUI skill API experience, which means that advanced skills and use cases of the GUI cannot be ported over.
Whatever Dinkum is trying to do with the GUI is likely to break when a skill developer tries to do something outside of its forced, cut-down GUI environment.
Unless Dinkum core decides to fix their internal GUI protocol and remove the artificial limitations they have put in place, there will be very little that any skill developer will be able to do with that GUI skill API implementation.
If you want a refined GUI development experience with freedom of creativity and no walled API, you are better off running OpenVoiceOS, Neon, or even the normal Mycroft-Core (even tho the GUI api is outdated in dev branch).
i have the reverse question, what is the future of classic core?
Lots of time has been invested in classic core and projects depend on it, is everyone who developed for classic core expected to migrate to dinkum? as said in reddit dinkum removed half of the code base, anything not needed by mk2 default skills is gone. Can existing stuff even be ported?
If classic core wont get updates, is all the 3rd party just outdated and waiting till core stops working for good? If so why would anyone maintain a project built for classic core?
Is mycroft maintaining 2 competing cores and for how long?
Any new developer aware of how classic core community was treated will be wary of developing for dinkum, will a new dinkum ecosystem be thrown away with no consideration for the community all over again in 1 year? The way this whole mk2/dinkum situation is being dealt with does not inspire confidence as a new platform devs want to be a part off
If dinkum is the way forward it seems to me community devs are in a situation of discarding all their projects, rewriting them, or maintaining/moving to a fork of classic core from now on. Between skills plugins and other companion projects this means hundreds of github repositories are now potentially obsolete without warning. Not long ago Mycroft boasted about having 60k developers from the community, if the number is legit then thats a lot of people to surprise with the dinkum move
You have a few options for the mk2, dinkum and sandbox classic core are the official images. then there is OVOS and Neon, those are in beta stage and not fully ready, but dinkum appears to be in pre-alpha when compared to those…
@jmillerv one thing that would aid my uptake of the process of conversion is if each of the skills in your repository had a link back to the classic version of the skill.
@AIIX I think your points are very valid, yet at the same time, @clintonthegeek and @jmillerv 's point also still stand. In all likelihood an expanded Dinkum (even though it appears it will be with new developers) will be the future of the official version of Mycroft, so I believe we must start and do what we can starting now to port and create Dinkum skills. It is very likely that the additional facilities you point out are missing will be made available in time. The alternative projects are fantastic, more power to everyone (that is a strength of FLOSS IMO), but we will continue to need the official Dinkum to achieve full robustness to achieve Mycroft overall project commercial viability.
Very much doubt it will happen, but given all the community feedback, wouldn’t it be good to at least consider ditching dinkum as a failed experiment and resume work on classic core?
They say its not maintainable and full of bugs, but the reality is that Neon was built on top of it, Chatterbox was built on top of it, OVOS was built on top of it… Neon in particular has extensive unittests to validate the code is working as intended. Thats 3 projects already that were founded on classic core.
If you include classic core itself with all the community feedback claiming it works better, thats 4 projects proving the rationale for dinkum is not fully accurate
Disclaimer: I may not be fully neutral in my opinion as I have worked or continue working very closely with all the above mentioned projects, including classic core
The dinkum dev showed up on reddit a few days ago. He basically said that the team could never get a production quality device out of classic-core which is why he was asked to build dinkum out. He then stripped out anything that wasn’t necessary to get the device to work with the basic skills.
I honestly think this is a dead project. I have spent a lot of time this week researching the state of affairs with Mycroft and it does not look good. I am going to move on with either Neon or OVOS and hope either they end up building a vibrant community or that Mycroft is able to pull themselves out of the mess they are in.
As I see it here are their issues:
The dinkum devs are gone.
The founders are gone.
The community of devs is beyond pissed because Mycroft stripped a lot of things they relied on for their skills for the Dinkum build.
The Mycroft dev community is shrinking and those that are staying are mostly not willing to work on porting the skills to Dinkum.
After reading many of the comments, I second, third, forth … the questions.
But I have a few of questions of my own.
First, I built my own mycroft using a Raspberry pi4, 7in touch screen, eyetoy 3, and using an aux to audio cable hooked up to a stereo for sound.
The software I am using is Ubuntu 20.04 as the base,
added the kde neon unstable repos and apt update/upgrade from their.
git clone and cd into the mycroft-core/gui and dev_setup from them and all is working fine.
Even have squeezelite loaded and running, and mimic3, and at one time also had the Squeezebox/LMS server and motion server running at the same time on the same pi.
Yes, sometimes, with all running, it was slow. (Took motion and LMS server off later after some advice from a friend).
My questions are,
Why not just build this setup and image it from their after all is up and running/working.
I’m not a developer, but I can follow directions, and google code when needed, but with the daughter board and camera, would these be too hard to load into Ubuntu?
As with the new dinkum system, will it be able to run on my current setup, if the mycroft-core becomes outdated?
Sorry, can not afford $500 for a mark II system, or I would test out updates, ports.
Also, from what I have read, isn’t the neon ai, ovos, and bigscreen plasma systems following the mycroft-core system, and isn’t the core already modular?
Just my thoughts. Like I said, I’m not a developer or programmer. I could be wrong in some of my assumptions. I only know html and css.
tldr, ovos-core is modular, mycroft-core is not, it could be but Dinkum was what we got instead of a backwards compatible refactor like in ovos
mycroft-core is a monolithic application, to run any of the components you need the full stack installed. For example to run the messagebus service, you need all the audio pipeline installed too. This can become quite messy if you want to treat components as optional, such as running a server without the speech client or in docker.
ovos-core untangled the code and made sure every component can be run standalone without the dependencies of the other components, ovos-core is also considered a python library (its on pypi!) you can use in your own projects.
Neon is built on top of ovos-core as a library, Plasma Bigscreen also adopted ovos-core recently instead of mycroft-core
What surprises me looking back over the past few days is that I’ve not seen a single comment from anyone at Mycroft and a fair number of comments from people from Neon and OVOS.
