Skip to content
For Self-Managed AI Platform users running v11.1, see the self-managed platform documentation
Applications > Create

Create applications

Premium

Application upload from a pre-built image is a premium feature. Contact your DataRobot representative or administrator for information on enabling the feature.

Create applications in DataRobot to share machine learning projects using web applications—including Streamlit, Dash, and Plotly—from an image created in Docker. Once you create a custom machine learning app in Docker, you can upload it as an application in Registry > Application sources and deploy it with secure data access and controls. Alternatively, you can use the DRApps command line tool to create your app code and push it to DataRobot, building the image automatically.

The table below describes each method for creating an application:

Method Description
Upload an app Upload an application saved as a tar, gz, or tgz archive.
Create an app from an application source Build an application from an existing application source.

Persistent storage in applications

DataRobot uses the key-value store API and file storage to provide persistent storage in applications. This can include user settings, preferences, and permissions to specific resources, as well as chat history, monitoring usage, and data caching for large data frames.

Paused applications

Applications are paused after a period of inactivity. The first time you access a paused application, a loading screen appears while it restarts.

Supported frameworks

The following application frameworks are currently supported in DataRobot when creating an application from a Docker image:

  • Streamlit
  • Dash
  • Aiohttp
  • Plotly
  • Flask
  • React
  • FastAPI
  • JavaScript

Upload an application

To upload a custom application to DataRobot, first, you must create an app image in Docker:

  1. Install Docker.

  2. Create an app (see examples for Streamlit, Flask, and Aiohttp).

  3. Expose port 8080 in your Dockerfile for HTTP requests.

  4. Build your image with docker build [PATH] | [URL] --tag [IMAGE NAME].

  5. Test your app image locally with docker run --publish 8080:8080 [IMAGE NAME].

When you are ready to upload your app image to DataRobot, create a new build and then export it with docker save [IMAGE NAME] --output [PATH]. Once you have your app (exported as a tar, gz, or tgz archive), upload the image to the Applications page.

Once you have an application tar, gz, or tgz archive, you can upload the image in the DataRobot Registry:

  1. In Registry, go to the Application sources tile. Then, click the + Add new application source dropdown and select Upload application.

  2. In the Create new custom application panel, configure the following:

    • In the Name field, enter a descriptive name for your application.
    • In the Add application image group box, click Choose file, select the tar, gz, or tgz archive containing the image, and then click Open.
    • Choose a Resource bundle from the dropdown.

  3. Once the application is uploaded, click Create application. The application is added to the Applications page with a status of Initializing. After it builds, click Open to view the application.

    Note

    Click the Actions menu next to an application to Share or Delete the application.

Build an app from an application source

Note

The storage component of applications is not persistent. You can only write to the /tmp directory, and the directory's contents won't be persisted across sessions.

If you have configured an application source, you can use it to build an application:

  1. In Registry > Application sources, click on the application source you want to use, and then click Build application. Allow time for DataRobot to initialize the app.

  2. The application is added to the list of apps built from that source. After it builds, click Open to view the application.

    Note

    On the Applications page, click the Actions menu next to an application to Share or Delete the app.