This is experimental, to work with jupyter_kernel_mgmt.

You will need docker set up. Build a tagged docker image with the Dockerfile in this repo, by running:

docker build -t jupyter-kernel-eg .

Create a file ~/.jupyter/docker_kernels.toml with contents like this:

[kernels.python]
image = "jupyter-kernel-eg"
language = "python"
cwd = "/home/takluyver/scratch"

Then test with python3 -m jupyter_docker_kernels. It should start a kernel in a docker container, connect to it, get kernel info, and shut it down cleanly.

To use another docker image, ensure that it is built to start the kernel with a connection file /connect/kernel.json and a working directory /working.