There are some configuration values that are required at compile-time. This is primarily the adapter - ecto inlines some functions in the repo, based on the adapter, in order to speed-up dome hot paths.

Most of the other configuration values can be passed to the repo’s start_link function (or as options in the supervisor).

There’s a list of compile-time options and runtime options in the documentation: https://hexdocs.pm/ecto/Ecto.Adapters.Postgres.html

Ecto does not allow dynamic configuration of the DB at runtime. Your only option is to directly use the DB-Adapter to go through connection and query.

Thats at least what I understand when reading github issue #1183

Can you tell us a bit more broadly what your end result should be? I mean, I suspect you may want to do multi-tenancy by connecting to multiple databases? If this is the case then Ecto has sort of different approach, by using namespaces/prefixes rather than connecting to multiple databases.