Authentication - REST Interface

If used with Databricks the MLFlow REST API interfaces uses the Databricks Unified Authentication approach as per the the rest of the Databricks interface. That is a .databrickscfg and other details as specified in the Documentation/Authentication.md of that package.

If not using Databricks the REST interface is authenticated using an authentication file. Any authentication needed in the fluent interface will be taken care of by the underlying Python libraries, see also: working with a remote tracking server in the fluent interface.

The authentication file for the REST interface can be either:

• ~/.mlflow-connect in the users home directory

• mlflow.json somewhere on the MATLAB path

If this is used with a standalone mlflow server, a simple file can be created (either the .mlflow-connect or the .mlflow.json mentioned above).

In its simplest form it can look like this:

{
  "host":"http://localhost:5000"
}

A bearer token, if available, can be added as the field token. A template file is provided: ``Software/MATLAB/config/mlflow.json.template`.