I have a predictive model that I am wrapping in a REST endpoint. Under the hood, the /forecast endpoint chooses one of hundreds of models to make the prediction based on the GET param feature_name.
We have a use case to get all predictions from all models with a single HTTP request. Should we:
- Make
feature_nameoptional, and if nothing is passed, return predictions for all models? - Make users pass a special word like
'all'tofeature_nameif they want all the predictions? - Maintain two endpoints,
forecastandforecast_all? - Something I haven't thought of?
Aucun commentaire:
Enregistrer un commentaire