I sometimes have a similar issue with Travis CI, if I can't use one of their built in integrations or something. A simple workaround is to just upload an artifact to S3 after a successful build. Then on another system scan the bucket with a cron or listen for an S3 event notification for further processing. You can put a lifecycle policy on the bucket so it cleans up after a while too.

It's sort of annoying to have to add on another system to the pipeline, but I think it's generally reasonable. Managing artifacts shouldn't really be part of the CI system anyway, I feel it's better to handle that stuff separately.

I'm still using the old approach, which is adding a local directory as a docker volume, where all my artefacts are stored.

nginx then serves them under a unique path.

I would very much like to see gitlab offer the ability to get artifacts via some API so I can remove all that hackery.

We're working on an artifact browser for 8.4. I assume an API for that will be trivial after that has landed.

shouldn't your build tool do that? If your build creates artifacts, I'd imagine you'd want to store them in an artifact manager (Nexus, Artifactory, etc) or deploy them.