callysto/2i2c-image

Docker image for 2i2c based hub

An 2i2c Compatible Image for Callysto

This repository provides the environment to be run for callysto via 2i2c. Basically, it defines a docker image based on docker-stacks which includes the packages and configurations needed to run the Callysto notebooks. The default is to build from the juptyer/minimal-notebook:latest image, but if you want more reproducible builds, that repository typically includes date tags as well and you can override the base image with e.g.

$ make build-all DOCKER_BUILD_ARGS="--build-arg \
    BASE_CONTAINER=jupyter/minimal-notebook:2022-08-11"

Quick Build Instructions

A Makefile is provided which will help build, tag and push images similar to what is done in docker-stacks (that's where the Makefile originates). To build and tag the image, try

$ make build-all

As long as you have docker installed, that should build callysto/callysto-notebook:latest.

You can also build the image manually from the image directory, by running e.g.

$ docker build -t callysto-notebook:latest .

Package Management

As far as possible, package management is handled by mamba, which is a drop-in alternative to conda but with more efficient dependency resolution. Packages dependencies are listed in environment.yml which is processed by conda-lock to produce a lockfile for reproducible builds. A handful of other packages and utilities are installed by either apt or pip. The environment.yml file supports direct versioning where appropriate (e.g. python=3.8, but as far as possible, we prefer to follow the development of the underlying packages to keep the image up to date and relevant.

Assuming you have conda/mamba installed somewhere, you can generate an updated conda-lock.yml with

# Assuming miniconda is installed somewhere
$ conda install -c conda-forge conda-lock mamba
$ conda-lock --mamba -k explicit --file environment.yml -p linux-64

The Dockerfile will process that lock file and install those packages via mamba. Note that the underlying image for the docker build is usually jupyter/minimal-notebook:latest, but if you need something more reproducible you can try overriding the BASE_CONTAINER argument. That said, there are some apt-get update and apt-get install commands which might install slight different packages as the ubuntu repositories are updated.