Introducing apps and dev environments

Andrey Cheptsov's photo
Andrey Cheptsov
·Jun 28, 2022·

2 min read

Introducing apps and dev environments
Play this article

Over the last few months, we've been working on something cool. Today, we’re excited to share with you the details, and invite you to give it a try!

Streamlit, Gradio, and FastAPI apps

With the new update, it’s now possible not only to prepare data and train models but also to build AI apps and share them with anyone.

You’re free to use your favorite application frameworks, such as Streamlit, Gradio, and FastAPI.

Building and running an AI app is as easy as running any other workflow:

Now, you can run your app right from the CLI:

Once the app is up and running, you’ll be able to open it through the user interface or the CLI.

For more examples and documentation, go to the GitHub pages of the streamlit, gradio, and fastapi providers.

Run providers from the CLI

One more thing that the update brings is the ability to run providers right from the CLI, without defining a workflow.

For example, it’s now possible to run the app from above this way:

You can pass any parameters, including tag dependencies, hardware resources, output artifacts, and anything else the provider supports.

All providers, including python and docker, can now run from the CLI. Check their GitHub pages for examples and documentation.

VS Code and JupyterLab dev environments

Training models without an interactive development environment is hardly possible.

This is why we’ve added the providers that allow launching VS Code, JupyterLab, and Jupyter applications.

Launching an online VS Code application with the required hardware resources and access to your data is now super easy:

Once the app is up and running, go ahead and continue working with your code from the started VS Code.

You can stop any application when you don’t need it from the user interface or from the CLI.

Check the GitHub pages of the code, lab, and notebook providers for more examples.

It’s not all of course. The update adds many things and fixes quite a few bugs.

In case you’ve tried dstack a long time ago, it’s the perfect time to try it once again! You’re very welcome to build your first AI app with dstack.

Note, that the new providers are still in beta. Found a bug? Please report it to our issue tracker. Have a question? Join our Slack channel.

In the meantime… Give us a star on GitHub! 🤗

Share this