Using other images on the JupyterHub

Other images on the hub

Use the dropdown to select a non-default image on the hub. There are a variety. You can learn about them on the NMFS Open Science container images repo.

Using other images not on the hub

The JupyterHub can run other images that are compatible with a JupyterHub, e.g. Binder images. When you start the hub, use the image dropdown to select “Other”:

You can add a url to a Docker image to this. For example, if you wanted to use the Pangeo notebook docker images image, you would paste one of these into the “Custom image” box.

From DockerHub: pangeo/pangeo-notebook From Quay.io quay.io/pangeo/pangeo-notebook

Other common data science images:

• Jupyter Docker Stacks
• NASA Openscapes python
• Rocker Binder image
• Pangeo
• geocompx
• GPU accelerated data science docker images

Using a GitHub repo

You can also create an environment with a MyBinder.org compatible GitHub repo. By selecting the “Build your own image” option.

Simple example for Python

There are two ways to do this. Either via a conda environment or a pip install.

conda example

1. Put an environment.yml file in your GitHub repo at the base level with your Python packages that you need.
2. Copy the url to your repo and paste that into the “Repository” box (above).

environment.yml

name: example-environment
channels:
  - conda-forge
dependencies:
  - python=3.11
  - numpy
  - psutil
  - toolz
  - matplotlib
  - dill
  - pandas
  - partd
  - bokeh
  - dask

pip install example

You will need requirements.txt for packages and runtime.txt for the Python version.

requirements.txt

numpy
matplotlib==3.*
seaborn==0.13.*
pandas

runtime.txt

python-3.10

Simple example for R

r example

You will need install.R for packages and runtime.txt for the R version.

install.R

install.packages("tidyverse")
install.packages("rmarkdown")
install.packages("httr")
install.packages("shinydashboard")
install.packages("leaflet")

runtime.txt

r-4.3.2-2024-01-10