<aside> 💡 A step-by-step guide to setting up Databricks integration in the Trustible platform.

</aside>

The Trustible platforms supports a Databricks integration that allows transferring data into the Inventory. This guide covers how to set-up the integration and transfer data into the platform.

Note: Currently we only support transferring MLFlow metrics/parameters into specific Model Card fields.The Trustible platforms supports a Databricks integration that allows transferring data into the Inventory. This guide covers how to set-up the integration and transfer data into the platform.

Note: Currently we only support transferring MLFlow metrics/parameters into specific Model Card field

Set-up [Admin-Only]

To connect to Databricks, you will need to set-up a Service Principal and grant it access to the relevant models/notebooks/files in your Workspace. A service principal is a separate identity for connecting to external services - this will give you fine-grained control over what Trustible can access. You can read in depth instructions for generating tokens in the Databricks Documentation.

To add a token into the platform:

1. Create a personal access tokens for the Service Principle following the Databricks Documentation. It will look like dapixxxx
2. Go to https://app.trustible.ai/integrations/databricks/create-token/ and enter info for this token:
   1. Unique Key: An identifier for tracking keys in Trustible
   2. Databricks Url: Link to workspace (https://dbc-{XXX}.cloud.databricks.com)
   3. Databricks Access Token: Token from Step 1

MLFlow Data Transfer

The MLFlow Integration is currently available for specific fields on the Model Card.

To use the integration, the following conditions must be met:

1. The “Registered Model” has a Run associated with it

   

2. The Service Principle has access to the Registered Model and Run.

   

Copying over the data:

1. If the Databricks Integration is available, then the “Databricks” button will be available in Edit mode for the field:

1. Data can be accessed by:

   1. Registered Model: Will identify the Run associated with the latest version of the model

      

   2. Specific Run ID

2. The next view will show you all the available Metrics/Parameters. Select the ones that should be copied over and add custom names/comments. This data will be stored as a special JSON on the field. When you are done, click Copy to Form.