Currently Mnemo is structured as a Phoenix monolith - the intention is eventually to split portions out into services if need be.
The architecture laid out in the "arch" documents is still preserved, both in
terms of file structure within lib, as well as in the naming of modules
(with slight adjustments to accommodate elixir conventions).
There is a pre-commit git hook configured to check formatting/linting via the
elixir formatter and to run tests via mix test. Failure of any of these
two commands will stop a commit.
On pushing to main, a CircleCI pipeline is configured, which runs two jobs:
The first runs tests in a test env, imitating what happens locally on commits.
The second builds a release and runs e2e tests with playwright, imitating
the configuration of our fly deployment. If these tests succeeds, the build
is launched to fly.
The flyctl auth token is kept as a secret in circleci.
Once deployed, the server reads envs from the fly environment, most importantly
the SECRET_KEY_BASE and DATABASE_URL.
The Dockerfile found in the mnemo src folder was created by flyctl, and
modified slightly to include relevant dependencies. It is not used in the
CircleCI flow, only as a deployment container by fly.
The Postgres DB for dev & testing is a docker-compose project that runs a
single PG container with a few envs set, and includes a shell script to
start and to stop the service locally. This can be found in resources.
TODO:
- Split up logic for svelte components in neater fashion (store for every component)
- Consider switch to Vue
- Tests for manager, prioritize over unit tests.
- Tests and specs for functions we created.