Databricks Apps

The on-behalf-of-user authorization mode for Databricks Apps has reached General Availability (GA).

The benefits of on-behalf-of-user authorization include security by default (inheriting the user's least privilege), development simplification (no need to repeatedly implement authorization logic), and clear auditing (data access is traceable to specific users).

After adopting on-behalf-of-user authorization, the content visible to users in the application is strictly limited to the permission scope granted to them in Unity Catalog.

In the traditional deployment model, Databricks Apps accesses data resources uniformly under the identity of a service principal, with all users sharing the same set of data read permissions.

Completed agents can be deployed as Databricks Apps and leverage Lakebase for memory capabilities to maintain context and state across multi-turn interactions.

Databricks Apps serves as the application layer, handling both frontend display and backend logic simultaneously, without the need to set up an independent web service.

The real-time IoT architecture consists of three core components: Zerobus Ingest, Lakebase Postgres, and Databricks Apps.

Under the centralized platform model, users essentially hold a usage license rather than ownership of their accounts.

Databricks Apps provides a managed web application runtime environment where developers can build frontends using Python frameworks (such as Gradio and Streamlit) and deploy them directly within the Databricks platform.