Snowflake (Beta)

Select the Snowflake option in the deploy screen to deploy your model to Snowflake

Deployment Settings

Once selected, there are several options to select for a successful deployment.

Snowflake Predictions Source Dataset

Select the table against which you will be generating predictions. Do this by entering the schema and table name, then clicking "Verify"

NOTE: The credentials used to train your Project will be used, i.e. the Account Identifier, Warehouse, Username, Password, and Role

To create a new role for a user and grant read-only access, run SQL similar to:

-- Assumes a pre-existing user named MY_USER

CREATE ROLE AKKIO_ROLE;

GRANT ROLE AKKIO_ROLE TO USER MY_USER;

GRANT USAGE ON WAREHOUSE MY_WH TO ROLE AKKIO_ROLE;

GRANT USAGE ON DATABASE MY_DB TO ROLE AKKIO_ROLE;

GRANT USAGE ON SCHEMA MY_DB.MY_SCHEMA TO ROLE AKKIO_ROLE;

GRANT SELECT ON TABLE MY_DB.MY_SCHEMA.MY_TABLE TO ROLE AKKIO_ROLE;

GRANT CREATE TABLE ON SCHEMA MY_DB.MY_SCHEMA TO ROLE AKKIO_ROLE;

GRANT CREATE FILE FORMAT ON SCHEMA MY_DB.MY_SCHEMATO ROLE AKKIO_ROLE;

NOTE: Predictions will be written to a new table, named <your_table>_akkio_predictions in the same schema as your source table. Your user must have access to create a table in this schema.

Map Fields

Map the fields betwen your training dataset and your predictions dataset. You only need to map the relevant fields, and can fields can be named differently

Other Options

• Scheduler - Select how often the prediction will run.

• Run on Deploy - Select whether it will run immediately upon deployment.

• Run Rules - Run against either all rows or rows without predictions. This allows you to input and output the same type of Salesforce field.

• Map Fields - Map the fields existing in the project to the fields on the prediction data. These will often be the same but can be manually mapped if they are different.

• Apply Data Prep - If you applied any data prep steps in the project, those transformations can be used to your prediction data before running the prediction

Preview and Deploy

Once you are happy with your settings, you can click to preview and then deploy your model.