Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

repo2docker JupyterLab

too long, didn't read... here is the example!

docker run -it --name repo2docker -p 8888:8888 vanessa/repo2docker-jupyterlab jupyter lab --ip


Now build your own, with your own code to share, a la containershare! Thanks to the project jupyter team and @blink1073 for the team effort on this container.

Getting Started

This is a continuous build template based on the jupyterlab binder example, meaning that it shows how you can install Python dependencies in a jupyterlab notebook to a container deployed automatically with Docker Hub. For repo2docker, we detect files automatically by default given presence of the binder directory. The url is accessed by appending ?urlpath=lab to the jupyter lab notebook. That's it! Within the binder folder:

  • the requirements.txt defines what will be installed in the notebooks.
  • the postBuild script installs jupyerLab extensions and widgets.

The build->test->deploy setup will have integration for a containershare registry. In a nutshell, when you add the configuration files here to your repository and connect to continuous integration, you will get a container, Github pages to serve metadata with complete manifest and inspection of your containers. It's a completely free, transparent, and open source way to practice reproducible science. See here for a demo of JupyterLab.


Note that we are currently using jupyterlab 0.34 and the geojson extension version 0.17.

Examples of URLs

Point to jupyterlab

  • localhost:<port>?urlpath=lab

Point to jupyterlab particular file

  • localhost:<port>?urlpath=lab/tree/path/to/my/notebook.ipynb


<@vsoch:Stanford Research Computing Center> <@binder-examples:Binder Examples from the Jupyter team>


This template will let you do the following:

  1. Add your binder notebook files and dependencies for to a Github repository
  2. Connect the respository to CircleCI and Docker Hub to deploy a container along with metadata on Github Pages
  3. Submit the container to a containershare registry, or just share the repository with your collaborators.

This usage is adopted from the continuous-build repository. The software has been extended to deploy metadata and a web interface to explore it back to the Github repository. As described above, you will be able to push a Docker container to Docker Hub directly from a Github repository that has a Jupyter notebook.

Reproducible Content Generated

  • container you generate a Docker Hub container with your notebook and dependencies, ready for use wherever you can run the container. The container always provides a latest, but also a tagged version coinciding with the first 10 characters of each merged commit.
  • maniests along with Docker Hub manifests, the Continuous Integration step deploys a table of updated container manifests to the Github repository Github pages, which are available programatically and from an interactive web interface.
  • inspection using the container-diff tool, every tag of a container has apt, pip, and files recorded. Yes, this means that you can programatically find containers based on software. @vsoch will be developing a command line tool to help with this.

Importantly, both manifests and the inspections are generated by the continuous integration, and served with the same repository that bullds the container.

Getting Started

Today you will be doing the following:

  1. Fork and clone the continuous-build Github repository to obtain the hidden .circleci folder.
  2. creating an image repository on Docker Hub
  3. connecting your repository to CircleCI
  4. creating a Github Machine User account to deploy back to Github Pages
  5. push, commit, or create a pull request to trigger a build.

You don't need to install any dependencies on your host to build the container, it will be done on a continuous integration server, and the container built and available to you to pull from Docker Hub. If you add a Machine user (step 4) your container

Step 1. Clone the Repository

First, fork the repo2docker-jupyterlab Github repository to your account, and clone the branch.

git clone<username>/repo2docker-jupyterlab
git clone<username>/repo2docker-jupyterlab.git

Step 2. Configuration

The hidden folder .circleci/config.yml has instructions for CircleCI to automatically discover and build your repo2docker jupyterlab notebook container. There is also a template.html file that is used as a template for your Github pages. The first does most of the steps required for build and deploy, including:

  1. clone of the repository with the notebook that you specify
  2. build
  3. push to Docker Hub
  4. generate manifests and inspections
  5. generate Github Pages content, and push back to Github pages

Thus, if you have forked the repository and cloned your fork, you should be able to use the files that are pulled. And if you are an advanced user, you could even customize if you please.

Step 3. Docker Hub

Go to Docker Hub, log in, and click the big blue button that says "create repository" (not an automated build). Choose an organization and name that you like (in the traditional format <ORG>/<NAME>), and remember it! We will be adding it, along with your Docker credentials, to be encrypted.

Step 4. Connect to CircleCI

If you do not already have a Circle CI account, head here and create one, and add your project to your Circle CI account. Here are instructions if you've never done this before.

Once you have an account, if you navigate to the main app page you should be able to click "Add Projects" and then select your repository. If you don't see it on the list, then select a different organization in the top left. Once you find the repository, you can click the button to "Start Building" adn accept the defaults.

Before you push or trigger a build, let's set up the following environment variables. Also in the project interface on CirleCi, click the gears icon next to the project name to get to your project settings. Under settings, click on the "Environment Variables" tab. In this section, you want to define the following:

  1. CONTAINER_NAME should be the name of the Docker Hub repository you just created.
  2. DOCKER_TAG is the tag you want to use. If not defined, will use first 10 characters of commit.
  3. DOCKER_USER and DOCKER_PASS should be your credentials (to allowing pushing)
  4. GITHUB_USER and GITHUB_EMAIL should be your machine user Github account
  5. REPO_NAME should be the full Github url (or other) of the repository with the notebook. This doesn't have to coincide with the repository you are using to do the build (e.g., "myrepo" in our example).

If you don't define the CONTAINER_NAME it will default to be the repository where it is building from, which you should only do if the Docker Hub repository is named equivalently. If you don't define either of the variables from step 3. for the Docker credentials, your image will build but not be pushed to Docker Hub. Finally, if you don't define the REPO_NAME it will again use the name of the repository defined for the CONTAINER_NAME. As a quick sanity check, here are the variables you should have defined in your CircleCI settings.

Step 5. Github Machine User

If you want to deploy the manifests back to Github pages, the easiest option (and one that doesn't put your entire Github account under risk) is to create a machine user. This comes down to creating a second Github user account (with a different email) and then giving the account permission to the repository, and generate an ssh key for it. You won't need to worry about how the deploy is done - this is handled in the circleCI recipe included with the template. Here are instructions for setting up credentials, derived from this great resource.

Why do I need to do this?

Pushing content back to Github pages requires a deploy key. Although Circle will generate a deploy key for you, it only has read access. We need to generate a machine user with write access. Read more about machine user keys keys

Instructions If you haven't done this before, you can follow the instructions here to generate a new ssh key. The steps to add it to your project are a little weird, but I'll try to be specific:

  1. Open a second browser so you can stay logged into your main Github account in one browser, and create a new Github account there. You will basically need another email address, and a creative username.
  2. In your main Github account (the primary browser) add this user as a collaborator to your repository. They will need push access.
  3. Accept the invitation in the second browser, or the emali sent to you.
  4. In the second browser, again log in to Circle CI with your new Github account. Make sure you log in via your Github machine user account, and that you have accepted the invitation.
  5. Click on "Add Projects", and select your regular Github username under "Choose Organization". This is the owner of the project. Then click "Follow Project" next to the repository name on the left of the menu. If you don't see the project there, copy paste the link to it directly.
  6. This is important! Once followed, go to the Project Settings -> "Checkout SSH keys", and click on the button to "Authorize with GitHub." You will be taken back to Github, signed in as the machine user, and you should click "Authorize Application." Finally, click the Create and add machine user github name key button on the same page.

Step 6. Push and Deploy!

Once the environment variables are set up, you can push or issue a pull request to see circle build the workflow. Don't forget that you will need to change the default in settings to also build pull requests. Remember that you only need the .circleci/config.yml, .circleci/template.html and not any other files in the repository. If your notebook is hosted in the same repo, you might want to add these, along with your requirements.txt, etc.


How do I customize the build or template?

The circle configuration file is the entire workflow that does build, test, and deploy. This literally means you can edit this text file and change any or all behavior. This could be as simple as changing some of the text output, to adding an additional set of testing or deployment options, or more complex like adding entire new steps in the workflow. The template.html is the same! You can tweak it, completely change it, or throw it out and push something entirely different back to Github pages. This deployment is open and completely transparent, as it should be.

How do I customize the jupyterlab install

The dependencies are managed in the requirements.txt files in the binder folder. You can add modules here to customize your jupyterlab container.

How do I run builds for pull requests?

By default, new builds on CircleCI will not build for pull requests and you can change this default in the settings. You can easily add filters (or other criteria and actions) to be performed during or after the build by editing the .circleci/config.yml file in your repository.

How do I use my container?

For a repo2docker container, you should expect a jupyter notebook. Here is an example of how to pull and run this container:

docker pull <ORG>/<NAME>:<TAG>
docker run -it --name repo2docker -p 8888:8888 <ORG>/<NAME>:<TAG> jupyter notebook --ip

For a pre-built working example, try the following to run jupyterlab:

docker pull vanessa/repo2docker-jupyterlab
docker run -it --name repo2docker -p 8888:8888 vanessa/repo2docker-jupyterlab jupyter lab --ip

You can then enter the url and token provided in the browser to access your notebook. When you are done and need to stop and remove the container:

docker stop repo2docker 
docker rm repo2docker

or just don't name it, and have it be removed automatically:

docker run -it --rm -p 8888:8888 vanessa/repo2docker-jupyterlab jupyter lab --ip


If you want to get help please post an issue!.


a build->test->deploy containershare template for a jupyterlab container




BSD-3-Clause, MIT licenses found

Licenses found






No releases published


No packages published