Skip to main content

Genie spaces with dashboards

Preview

This feature is in Public Preview.

When you create a dashboard, Databricks automatically creates a draft Genie space. A Genie space is a no-code interface that allows business users to explore data using natural language.

  • The draft Genie space stays in sync with your draft dashboard as you add and remove datasets and visualizations.
  • When you publish your dashboard, you can publish the associated Genie space or link to an existing one.
  • Publishing adds an Ask Genie button to the dashboard, allowing viewers to chat with the data instead of relying only on predefined visualizations. See What is an AI/BI Genie space.

For more information about Genie spaces, see What is an AI/BI Genie space.

Publish a Genie space with your dashboard

To publish the associated Genie space:

  1. Click Publish.
  2. The Enable Genie toggle is off by default. Turn the toggle on to allow dashboard viewers to use the associated Genie space.
  3. (Optional) To link an existing Genie space, select the Link existing Genie space radio button and paste in the associated URL. If this option is not selected, Databricks automatically generates a new Genie space based on your dashboard.
  4. Click Publish in the Publish dialog.

GIF showing a Genie chat window opening on a published dashboard. The user asks a question and Genie responds with a results table.

important

When viewers use Genie, their questions are answered from the full dataset query results. If a dataset query includes sensitive columns or rows, that data becomes available to Genie even if it is not shown in visualizations. Define your datasets to include only the fields and rows that should be accessible. See Permissions for linked Genie spaces.

note

The draft and published Genie spaces are independent, much like the draft and published versions of your dashboard. You must publish the changes in your draft dashboard for them to appear in the published Genie space.

Draft and published Genie spaces are companions to your dashboard. They do not appear in your file browser or the Genie listing page.

Open and preview Genie spaces

To open the published Genie space in a new window:

  1. Click the Kebab menu icon. kebab menu in the upper-right corner of the chat window
  2. Click Open Genie Space.

To preview your draft Genie space:

  1. Click the Kebab menu icon. kebab menu in the upper-right corner of your dashboard.
  2. Click Open the draft Genie space.
note

You cannot edit the instructions for this Genie space from the Genie space UI. Your draft Genie space only changes when you make changes to the draft dashboard.

Ask Genie about a specific chart

To ask Genie about a specific dashboard chart:

  1. Click Ask Genie from a Genie-enabled published dashboard.
  2. Click a dashboard chart to select it. Genie displays the selected chart title in the chat window.
  3. Ask questions about the data represented in the selected chart.

Permissions for linked Genie spaces

Permissions on a Genie space mirror the permissions of its dashboard. For auto-generated Genie spaces, this also applies to query execution. For published dashboards with embedded credentials, Genie uses the publisher’s credentials instead of the viewer’s to access data and compute. As a result, viewers can interact with Genie even if they do not have direct access to the underlying data. Viewers must be members of the workspace to interact with Genie spaces. To learn more about access types, see Manage dashboard and Genie access.

Incompatible SQL expressions

Linked Genie spaces cannot use datasets that contain the following SQL expressions:

  • Identifier clause with a non-constant string, as in the following example.

    SQL
    SELECT * FROM identifier(:param)
  • Temporary views or functions in multi-statement datasets, as in the following example.

    SQL
    CREATE OR REPLACE TEMPORARY VIEW taxis AS SELECT * FROM samples.nyctaxis.trips;

    SELECT * FROM taxis;
  • Referencing a volume in the SQL that defines the dataset, as in the following example.

    SQL
    SELECT * FROM csv.`/Volumes/catalog/schema/volume/my-data.csv`