Are pipelines not enough?

event
|> rule_a()
|> rule_b()
|> rule_c()

Or like you want to have a file that defines rules dynamically?

I think you could even have your rule modules, then a list of rule names that mapped to those modules and then dynamically built that pipeline based on the input of rule names from somewhere (a file, a database record, etc).

I have used a camunda rules modeler to create executable xml’s within a bpms. The interpreter for the rules I wrote in elixir (for one of the ruletypes). But camunda has it’s own interpreter written in java also. I never tried it, but here is a link https://camunda.com/products/dmn-engine/

Here is an intro to camunda dmn https://camunda.com/dmn/
The wysiwyg modeler is easy to integrate in your phoenix application.
There is drools also https://www.drools.org/ (eclipse ide plugin for development so no webfrontent I suspect), less lightweight (but mind you: “Drools Fusion (complex event processing features)”!), I do not know if it’s easy to integrate.
About different rules languages: https://stackoverflow.com/questions/52299097/drools-with-dmn-vs-drl