Applications

User Guide (Latest)

Applications are software or user interfaces available to use in, or with, a Workbench Project.

Some Example Applications

  • JupyterLab: The default base images for NVIDIA AI Workbench have JupyterLab installed.

  • Visual Studio Code: If VS Code is installed on your local system, you can connect it to the Project container, similar to Dev Containers.

  • Gradio Apps: Some Projects have simple web apps like chat apps or RAG apps installed.

Types of Applications

  • Web App: an application that runs in the Project container with a browser-based UI.

    • A typical example is JupyterLab.

    • Web apps are installed in the Project container, and that is where they run.

  • Process: a command or script that runs to perform a task without a UI or exposing any network access.

    • A typical example is a shell script that downloads a model.

    • Processes usually rely on some code in the Project repository, and they run in the Project container.

  • Native App: an application launched outside of the Project container on your machine.

    • VS Code is the only supported native app at the moment.

    • Native apps run on your local system. They may not have access to the Project container.

dev_envs.png

Each Project has its own applications, and you can start and stop them on the Project window in the NVIDIA AI Workbench UI or using the CLI.

Default Applications and Built-In Applications

  • When you create a new Project in AI Workbench, each of the Default Bases has JupyterLab installed and configured. The Pytorch and Tensorflow bases also have Tensorboard installed.

  • In addition, if you are working on a Project that doesn’t have JupyterLab installed, AI Workbench has a feature that will install it for you.

Note

You can also install and configure applications yourself. You can also use NVIDIA’s example projects on GitHub as a reference.

Installing and Configuring a Web App

  • You install a web app by installing any dependencies in the container and making sure any relevant code is set up in the Project.

  • Basically, if you know how to run the application manually then you should be able to get it running in the Project.

  • In addition to installing it, you need to configure AI Workbench to manage the application. This involves some simple metadata and commands.

  • For example, configuring a web app requires information like

    • Application name, e.g. JupyterLab

    • Start and stop commands

    • A health check command

    • Port the app is available on

custom_app.png

Next steps

Previous Integrations
Next Reference Overview
© Copyright © 2024, NVIDIA Corporation. Last updated on Jul 2, 2024.