it would help a lot with the database-storage refactor if we just went ahead and merge the custom model interface and the native predictors interfaces and started differentiating between the two predictor types with an is_custom key we add to their data (both in get_models and in get_model_data.
1 or 2 functions might have some slightly more complicated signature, but nothing significant, and this encourages us to handle native vs custom differences at the interface level, which (hopefully) means custom models will start being better supported in all parts of the codebase.
it would help a lot with the database-storage refactor if we just went ahead and merge the custom model interface and the native predictors interfaces and started differentiating between the two predictor types with an
is_customkey we add to their data (both inget_modelsand inget_model_data.1 or 2 functions might have some slightly more complicated signature, but nothing significant, and this encourages us to handle native vs custom differences at the interface level, which (hopefully) means custom models will start being better supported in all parts of the codebase.