eyadgaran
Namespace library packages
Separate libraries into independent namespace packages (e.g. simpleml-sklearn). Isolates dependencies to keep install lightweight and allows a la carte versioning for extras
Library list:
- SimpleML-SQLAlchemy-ORM
- SimpleML-MLFlow
- SimpleML-Scikit-Learn
- SimpleML-Tensorflow
- SimpleML-Dask
- SimpleML-Pandas
For now keep core interfaces and persistence package internal.
[Thought] use common naming and registries to enable third-party extension. Flask has some good examples of ecosystem libraries using standard naming conventions.
Open question - create a virtual namespace package: e.g. simpleml.ecosystem.{library}... that all ecosystem libraries install into or keep them completely isolated with a required active registration (package "does not exist" until first imported).
active registration puts more overhead on users to configure their environments in each shell (or globally in a config) but is less error prone compared to passive behavior and conflicts