Notebooks Hub Compose

This section documents the local deployment for the Notebooks Hub application using Docker Compose, including the mongo database, the API, the UI, and Jupyterhub.

Launch Mode	Description
node	install node modules in preparation for dev mode
conda	install default jupyter kernel for LMOD
runtime	launch notebooks hub from remote dockerhub
dev	launch notebooks hub from local filesystem with file watching

Launch Mode - Runtime

This mode runs the the Notebooks Hub ecosystem with latest API and UI images pulled from our PolusAI Docker Hub. If necessary, a specific application version can be run by modifying the following environment variables in the .env file.

• UI_IMAGE_VERSION

• API_IMAGE_VERSION

Pre-Launch Steps

First, ensure that Docker Docker Desktop is running. Then verify Dockerhub access to the polusai organization, and login to Docker using docker login or via the Docker Desktop sign in button.

Install default Jupyter kernel using conda via launch script (Optional). Collecting the initial package data takes 2-5 minutes before logging in the console begins.

./launch-notebooks-hub.sh -m conda

Launch Action

launch notebooks hub in runtime mode

./launch-notebooks-hub.sh -m runtime

Launch Mode - Dev

This mode runs the the Notebooks Hub ecosystem with API and UI images built from local files and include a file watch that update the application automatically as code is modified.

Pre-Launch Steps

Set necessary environment variables in the .env file.

Set the Npm authentication token for local development with the automation type.

NPM_AUTOMATION_TOKEN=

Install node modules via launch script.

./launch-notebooks-hub.sh -m node

Install default Jupyter kernel using conda via launch script (Optional). Collecting the initial package data takes 2-5 minutes before logging in the console begins.

./launch-notebooks-hub.sh -m conda

Launch Action

Launch notebooks hub in file watch developer mode.

./launch-notebooks-hub.sh -m dev

Sync the shared modules by running the following curl request in your terminal.

curl --location --request PUT 'http://127.0.0.1:8002/modules' --data ''

Note: The launch takes a moment. Please wait until you see the following in your terminal:

“Angular Live Development Server is listening on 0.0.0.0:4200, open your browser on http://localhost:4200/”

Post-Launch Steps

The localhost ports for each application can be referenced below:

• Jupyterhub

• Notebooks Hub UI

• Notebooks Hub API

• Login with the default local user (polus@example.com) in Jupyterhub, configurable in the .env file via the JUPYTERHUB_USER variable. No password is required.

• Login to Polus Notebooks Hub

Stop the application by navigating to the Compose folder and running the following command

docker compose -p notebooks-hub down

General FAQ

Question	Answer
How to auto-install a new module?	A new module can be auto-installed by changing the file name from default-env.yaml to your name of choice, while also changing the internal module name and modifying the dependencies before re-launching with the launch script.
How does authentication work for this implementation?	This installation implements a no-auth-flow and can be run fully offline when the necessary.
What does the down command do?	When running the docker compose -p notebooks-hub down command to remove the notebooks hub deployment, some named Jupyter containers may have to be manually deleted after being orphaned, and external volumes are not deleted until the next launch.
Are there configurable environment variables?	Yes, you can navigate to the .env file to update variables such as the Jupyter User.