Niels Bantilan

Flyte™ Sandbox: Union-Hosted Sandbox

When I joined the Union.ai team in 2020, the time it took to execute an initial Flyte workflow must have been 30 minutes to an hour.

You’d have to pip install flytekit and make sure Docker was running on your local machine. After writing a few tasks in Python, you’d then need to execute some arcane Docker incantations to spin up a local Flyte cluster, register your tasks to it, then execute your workflows on the console UI. Sure, Flyte is a powerful and feature-rich platform, but that doesn’t mean it has to take this long to try it out!

The Union.ai team listened to your feedback and after a lot of hard work, I’m excited to announce the release of Flyte Sandbox 🎉, a completely self-contained, browser-based IDE that you can use to get familiar with Flyte quickly!

What is it, exactly?

Flyte Sandbox is a trial and prototyping environment that combines two things: a hosted Flyte Sandbox Cluster — powered by Union.ai — and a VSCode server, which enables you to:

  1. Write tasks and workflows
  2. Run workflows on a minimal Flyte Cluster
  3. Interact with Flyte through a Jupyter Notebook or Python REPL

To help familiarize yourself with Flyte, Flyte Sandbox ships with a walkthrough that guides you through the process of writing, executing, inspecting and iterating on your first workflow.

<a href="https://sandbox.union.ai/" class="button w-button" target="_blank">Try Flyte Sandbox</a>

After stepping through the walkthrough, try creating your own workflows! One thing to keep in mind: Flyte Sandbox is limited in its compute and memory resources. If you like what you see and want to unlock more of Flyte’s features, you can either:

Get started with your own Flyte cluster locally:

Join the Union Cloud waitlist so that you don’t have to manage a cluster yourself!

Article
Update