I would leave them out for the moment.
For celery, rabbitmq and consumers:
by default tasks are run sync and locally. See this default setting. We will do some more testing on making things async, but for the moment I qould keep things simple, given the kind of "automated" deployment. If we employ celery workers a supervisor should be installed and configured to monitor workers, etc. Too many volatile things in this moment...
Docker compose (version==2) with the following features: