Manage your data to optimize Hex Magic
The Data manager gives Hex Admins one place to access and manage their workspace Data connections, metadata, and Magic usage settings.
Admins can use the Data manager to provide both Magic and users with additional context on data available in the workspace, and what data to include, exclude, or prioritize, which helps improve the response quality for Magic SQL generations.
For each workspace Data connection, Admins can add additional metadata to databases, schemas, tables, and columns. These additional metadata are used to help Magic identify the purpose of these various data objects. Additionally, they can be used to help provide Magic with more context about how to use the data in these data assets.
For example, adding a description to your
dim_users table that says “no internal employees are contained in the table” can help Magic and other Hex users understand that it’s better for queries related to external users. Another example is to provide context in the
AUTHORS column of our
BOOK_DATA table that explains Authors' names are formatted "Last name, First name."
These descriptions work in conjunction with warehouse-level metadata and dbt metadata, or on their own. All of these descriptions may be provided to Magic for the purpose of improving Magic generations.
Descriptions in the Data manager can also be utilized in workspaces with Magic disabled to share additional context for users viewing the Data browser in a project.
Include and Exclude
The Data manager provides the option to exclude databases, schemas, tables, or specific columns from Magic’s responses This may be because a data object is not the preferred path to address most questions. For cases like these, you can mark a data object as Excluded.
By default, all workspace Data connections are Included. If you exclude a data object, it’s children and all of their children are also automatically excluded.
The Data manager also allows Admins to more directly influence which data objecs are preferred by Magic. In the case where you have two data objects with somewhat similar data, it can be important to indicate what the preferred object is.
STG_ tables commonly contain similar dimensions to those downstream, but are not production-ready. Instead,
DIM_ tables are much more relevant and thus should be
Prioritized. This feature helps Magic understand these preferences so that users can trust Magic’s tendencies.
Databases, schemas, and tables can be prioritized in the top right of their pages in the Data Manager.