Integration · Databricks

    Every question to Databricks under your own name.

    Datahub asks Databricks your question with your own identity. Unity Catalog's row and column rules therefore apply per person, and Databricks' audit log shows who asked.

    Revenue per region this month?Rights per personAs SanneNorth€84,20018%East€61,90021%As JorisSouth€73,500••••
    • Per person

      Your rights, not a service account's

      Questions run under the user's own identity. You only see the data you hold Unity Catalog rights for.

    • Audit

      Who asked is in Databricks

      Databricks' audit log attributes every query to the user who asked it.

    • Read-only

      Genie and dashboards read

      Genie questions and dashboards only read. Datahub creates or changes nothing in Databricks.

    100+ hours
    In practice

    of finance work a week now runs automatically every night

    Hudig & Veder works with Databricks and dbt.

    Hudig & Veder
    What you can ask

    Ask in plain language. Genie looks it up.

    • “What was the revenue per region this month?”

    • “And only for the new customers?”

      Follow-up
    • “How is churn developing this year?”

    • “Which products sell best per channel?”

    • “Who owns the orders table?”

      From the catalog
    • “What does active customer mean here?”

      From the catalog
    How it works

    Consent once, then a short-lived ticket per question.

    Every user consents once per connection. Datahub keeps that encrypted in the key vault and makes a short-lived access token for each question. Consenting again is only needed when access is revoked.

    ConsentonceThis questionExpired

    Every question its own ticket, which then expires.

    • Genie

      Questions about your data

      A question about figures goes to a Genie space, which runs the query. The answer appears in the chat.

      • Follow-ups keep the earlier filters
      • 50 rows by default, up to 1,000
      • Administrators sync and describe the spaces
    • Lakeview

      Dashboards in Datahub

      Pick published Lakeview dashboards and place them in Datahub, per group.

      • Readers only see their groups' dashboards
      • A Genie chat beside the dashboard
      • Each view carries the viewer's email for audit
    • Good to know

      Where the shared identity stays

      Embedded dashboards and batch jobs run under the shared service identity. Per-person row rules do not apply there.

      • Existing token and service principal connections keep working
      • Per-user sign-in is added alongside them
    Where you use it

    One connection, three places.

    • Questions

      Genie in the chat

      Questions about figures go to Genie, questions about terms and owners to Datahub's catalog.

    • Dashboards

      Lakeview in one place

      The dashboards already in Databricks open in Datahub next to your other insights.

    • Insights

      Next to your other sources

      Databricks SQL feeds insights in Datahub, next to Fabric and the catalog.

    How to connect Databricks

    From Databricks to your first answer in four steps.

    1. 01

      Connect the workspace

      An administrator connects the Databricks workspace under Connected apps, with per-user sign-in.

    2. 02

      Fetch the Genie spaces

      Sync the Genie spaces and add to their descriptions where needed, so every question reaches the right space.

    3. 03

      Pick dashboards

      Import the published dashboards you want to show and assign them to groups.

    4. 04

      Consent once

      Every user consents once. After that Genie answers under their own name.

    Frequently asked questions

    What data teams ask before they connect.

    Next step

    Ask Databricks your first question.

    An intro call maps which workspaces and Genie spaces you run and what it takes to connect them.