Is your feature request related to a problem? Please describe.
skpro does not yet support online regressors that can update but only do point prediction, not full probabilistic output.
Describe the solution you'd like
Add an internal RiverRegressor adapter for River models. It should support learn_one and learn_many, do point prediction only, and wrap River models automatically when passed to update-capable meta-estimators.
Describe alternatives you've considered
Users wrapping River models themselves, or making the adapter probabilistic - not preferred for now.
Additional context
Full design decisions are in the docs: Design Doc
Is your feature request related to a problem? Please describe.
skpro does not yet support online regressors that can update but only do point prediction, not full probabilistic output.
Describe the solution you'd like
Add an internal RiverRegressor adapter for River models. It should support learn_one and learn_many, do point prediction only, and wrap River models automatically when passed to update-capable meta-estimators.
Describe alternatives you've considered
Users wrapping River models themselves, or making the adapter probabilistic - not preferred for now.
Additional context
Full design decisions are in the docs: Design Doc