Build applications
Region availability
Lakebase Autoscaling is the latest version of Lakebase, with autoscaling compute, scale-to-zero, branching, and instant restore. For supported regions, see Region availability. If you are a Lakebase Provisioned user, see Lakebase Provisioned.
Lakebase Autoscaling gives your application a serverless Postgres database with native Databricks integration. Use Databricks Apps (recommended) for managed identity, automatic credentials, and built-in deployment. Use external applications when you must run in existing infrastructure or with a specific runtime outside Databricks.
Pick where your app runs
-
- Databricks Apps
- Recommended. The platform manages credentials and deployment. Best for new apps, dashboards, and internal tools.
-
- Manual OAuth tutorial
- Use when you need explicit control over OAuth token rotation, or your app code already manages its own connection lifecycle.
-
- External apps with SDK
- Use when your app runs outside Databricks in Python, Java, or Go. SDK handles workspace token rotation.
-
- External apps with API
- Use when your app runs outside Databricks in Node.js, Ruby, PHP, or any other language. You manage token rotation yourself.
Databricks Apps vs external applications
Aspect | Databricks Apps | External applications |
|---|---|---|
Deployment | Runs on Databricks platform | Runs on external infrastructure |
Identity | Platform provides app service principal automatically | Create and manage service principal manually |
Setup | Create Postgres role; configure connection in app.yaml | Create service principal, Postgres role, and configure connection parameters |
Authentication | WorkspaceClient uses app identity automatically (no credentials in code) | WorkspaceClient requires explicit service principal credentials |
Best for | New apps, rapid prototyping, internal tools | Existing infrastructure, specific frameworks, systems outside Databricks |
Lightweight data access
If your application doesn't need a full server, the Data API exposes your Postgres tables as REST endpoints. This works for serverless functions, mobile apps, or environments without a Postgres driver.