Support python machine learning backend

[danpoltawski]

Ideally this would be in a separate container..

[scara]

Hi @danpoltawski, still thinking at a separate container as the preferred way of doing and not inline with the compose (i.e. via build)?

I'm thinking at unoconv for example and a nice container is missing there (I'd like to contribute for it - want to file an issue).

TIA, Matteo

[danpoltawski]

It could be via build, my point was that it ideally it shouldn't be just inside the apache container (this may be unavoidable).

Note, I found using build quite unmanageable to generate seperate images published on docker hub (thats why i switched in 824742285913fcfd76a75f2471c0b87ccdd0796a). I also quite like having the separately tested and isolated projects. For sure those images are useful without the entire compose infrastructure

[dmonllao]

Pull request in #113

[stronk7]

Just noting that, since mins ago, moodle-mlbackend-python docker images are being auto-build by DockerHub (by tag):

https://hub.docker.com/repository/docker/moodlehq/moodle-mlbackend-python/builds

So, whenever we release a new version @ https://github.com/moodlehq/moodle-docker-mlbackend-python, bump the corresponding image becomes built.

Note there is https://github.com/moodlehq/moodle-docker-mlbackend-python/issues/1 about any further improvement that we can we apply to the auto-building process.

[stronk7]

As far as it seems that #113 is were we are going to introduce support to run the python mlbackend unit tests (via server, aka, another container), ... let's continue there.