Currently there are three modules: decision-api, decision-common and typedvalue-api.
The latter is better to left untouched. But the former contains code that (mostly) could be shared by different models; at the same time, even inside decision-common there may be something common. The goal of this task is to keep/move "common" code inside decision-api (renaming it to "tracing-api") and put decision-specific code inside decision-common.
Having done that, implementation of further models (process, prediction) should involve creating only one module for each.
Keep refactoring/abstraction at bare minimum to avoid over-engineering.