Ready to run docker images of the e2x project for various use cases.
All images are hosted on ghcr.io and quay.io.
All hub images contain JupyterHub, e2xhub, as well as authenticators.
Jupyter Notebook images for different use cases, defined in notebooks.toml.
External base images are the roots; each image builds on the one above it.
What an image adds is described in its folder under notebooks/images.
jupyter/minimal-notebook:notebook-6.5.3
└── bi-notebook [e2x]
jupyter/minimal-notebook:notebook-6.5.4
└── minimal-notebook [e2x]
├── bidata-notebook [e2x]
├── datascience-notebook [e2x]
│ ├── datascience-notebook-code [e2x]
│ ├── ml-notebook [e2x]
│ │ └── nlp-notebook [e2x]
│ ├── sql-notebook [e2x]
│ ├── tf-notebook [e2x]
│ └── wus-notebook [e2x]
├── desktop-notebook
│ └── dbmain-notebook [e2x]
└── java-notebook [e2x]
Every image marked [e2x] also comes in three flavors with e2xgrader
installed, named <image>-<flavor> (e.g. ml-notebook-exam):
-teacher: teacher mode, for instructors creating and grading assignments.-student: student mode, for students working on assignments.-exam: student_exam mode, for students working on exams, with a restricted notebook.
docker run -p 8888:8888 ghcr.io/digiklausur/docker-stacks/<image>:latest-
Registries:
ghcr.ioandquay.iohold the same images (same digests). -
Tags (defined as tag sets in
notebooks.toml):Branch Moving tag Per day (UTC) Per commit mainlatestYYYY-MM-DDsha-<commit>devdevdev-YYYY-MM-DDdev-sha-<commit>Per-commit tags never move, so they are the safest way to pin a course or an exam to a fixed set of images. A per-day tag is only updated by later builds on the same day.
All images are defined in notebooks.toml. The script in
scripts/image_builder reads it. It only needs Docker and Python ≥ 3.11,
with no packages to install.
# Show which image builds on which
python scripts/image_builder tree
# Build one image → ghcr.io/digiklausur/docker-stacks/ml-notebook:local
python scripts/image_builder build ml-notebook
# e2xgrader variants are images of their own
python scripts/image_builder build ml-notebook-student
# An image and everything below it, each on its fresh :local parent
python scripts/image_builder build datascience-notebook --with-dependents
# Every image, in build order
python scripts/image_builder build --all
# Build on another parent tag instead of `latest`: your own local build,
# or exactly the parent of a CI run (to reproduce a failure)
python scripts/image_builder build minimal-notebook-exam --parent-tag local
python scripts/image_builder build ml-notebook-exam --parent-tag dev-sha-a10dab6
# Only print the docker command(s)
python scripts/image_builder build ml-notebook --dry-run
# Try another e2xgrader version, or install it from a GitHub branch
python scripts/image_builder build ml-notebook-student --build-arg E2XGRADER_VERSION=0.4.0
python scripts/image_builder build ml-notebook-student \
--build-arg FROM_REPO=true --build-arg E2XGRADER_BRANCH=dev
# Run a locally built image
docker run -p 8888:8888 ghcr.io/digiklausur/docker-stacks/ml-notebook:local
# See what CI would build for the changes on your branch
python scripts/image_builder plan --tag-set dev --since origin/dev- Results are tagged
:localand never pushed. A parent built in the same call is used as:local; any other parent is pulled from the registry (taglatest). - To add an image, create
notebooks/images/<name>/and add an[images.<name>]entry tonotebooks.toml. Runpython scripts/image_builder treeto check it. - Tests:
pip install -r scripts/image_builder/requirements-dev.txt, thenpython -m pytest.
If you'd like to contribute to this repository, please feel free to open an issue or submit a pull request.
The images in this repository are licensed under MIT.