I had a quick look at the library. I think this is an awesome project and something very needed in the community. I have some some remarks, though.

Why are floats used for money? This is a complete no-go for me since floating point arithmetic is inherently inacurate. I haven’t looked in depth if this is actually the case, but that’s what is presented in the examples.

Why is the project using GenServers in the middle? It should be perfectly enough to have “just” modules and functions. The worker process does not use the state in any way, so the only thing it does, is creating a bottleneck in the system.

There are a few libs that provide Money functions. money, ex_money and monetized (the first 3 on hex.pm) each seem to represent Money as a struct with a currency which is an ISO 4217 code and amount which is a Decimal.

It would be useful if Gringotts could support an abstraction for money types so that integration would be straight forward given than there’s a reasonable likelihood that someone using Gringotts is probably also using a Money lib. Maybe a Money protocol?

(disclaimer: I’m the author or ex_money)

Hi @pkrawat1 and team, I would like to reinforce the point above.

Using a GenServer when you don’t need one is an anti-pattern and it will actually have severe applications on the system performance, as you put all operations besides a single process which will become a bottleneck. We cover this in detail in Adopting Elixir and there is a thread about this particular topic in the forum.

Even if you may end-up using the worker for configuration, I would ask you to revisit that, because it should not be the place of libraries to impose a stateful configuration mechanism. Simply ask the users to pass the configuration every time the gateway is called. If the users of your library need to store this configuration somewhere, they can easily do so by using an Agent, application config, GenServer, etc.