Single applications and umbrellas both have their own merits and pros/cons. We don’t generate an umbrella by default for Phoenix because it may be overkill and it introduces more decisions and overhead for folks to know. Umbrellas exist to house multiple isolated applications under a single code repo, that can be started and stopped as a whole or independently. This is a great for a number of reasons, but also introduces complexity depending on your needs. For example, for the mix phx.new --umbrella case you now have three places to configure things, three places to add new deps, three mix.exs files, etc. Our goal w/ the new structure for single applications where you are focusing on boundaries and isolation is the moment you need to extract the web interface or other parts of your single app to a separate umbrella app, it’s just a matter of cd apps/ && mix new whatever then moving the code around. Little else should need refactored provided your boundaries are well established. Does that help?
