Skip to content

Databricks Custom Apps

Private Preview Documentation

🚧 This documentation is in a very early state. Lots still missing. And by all means, feel free to contribute! 🚧

Introduction

Databricks Custom Apps are the fastest way to deploy data and AI apps with the full security and governance of Databricks.

The v1 release will support web apps written in Python using any framework of your choice, including Gradio, Dash, Streamlit, Shiny, etc. Apps are hosted in Databricks' serverless platform using a compute specific for apps.

  • Apps interact with Databricks via REST APIs, SQL, or Spark.
  • Authentication, routing, monitoring, and much more are all provided out-of-the-box.
  • Apps have their own URL and can be easily shared to workspace users or all account users.

What is special about Databricks Custom Apps?

Apps are portable

Fundamentally, we strive to avoid magic. If you can run your app in your laptop you should be able run your app in Databricks. Consequently, if you deploy your app in workspace A, you should be able to deploy it without modifications in Workspace B. And of course, if your use case grows to more than what we can offer, the same app can be deployed with one of our many application hosting partners.

Apps are sandboxed

Apps are configured with the least privileges they need to run. Every app gets their own service principal and use industry-standard OAuth2 authorization controls. There is no support for Personal Access Tokens, or auto-injecting permissions that you have little control over. Every access token the app has access to is short-lived.

Fundamentals: Learn more on authentication, runtime, installing packages, etc.

Developing Apps: Tips on developing using your favorite framework, how to test, etc.

Resources: Useful resources such as common questions, talks, roadmap, etc.

Examples: A list of app examples showcasing a variety of use-cases and frameworks.

Part of the Lakehouse Apps Project.