There are two ways of making predictions. The recommended way is to first upload your
dataset to the project, and then using the corresponding datasetId, predict against
that dataset. To follow that pattern, send the json request body.
Note that requesting prediction intervals will automatically trigger backtesting if
backtests were not already completed for this model.
The legacy method which is deprecated is to send the file
directly with the predictions request. If you need to predict against a file 10MB in
size or larger, you will be required to use the above workflow for uploaded datasets.
However, the following multipart/form-data can be used with small files:
:form file: a dataset to make predictions on
:form modelId: the model to use to make predictions
.. note:: If using the legacy method of uploading data to this endpoint, a new dataset
will be created behind the scenes. For performance reasons, it would be much better
to utilize the workflow of creating the dataset first and using the supported method
of making predictions of this endpoint. However, to preserve the functionality of
existing workflows, the legacy method still exists.