Automating Jupyter Slides with Travis CI and gh-pages.

3 minute read

First off, if you use Python and you don’t use Jupyter Notebooks then you need to start right now. For that matter I would say the same thing about R users (although I’m not that familiar with R-markdown so maybe that’s ok). Jupyter notebooks are changing the way we work with code, instead of writing obscure “scripts” or re-writing the same code blocks over and over in Ipython or Rstudio, we now are writing computational essays in which the code, comments, notes, and figures all live in the same place ensuring more robust and reproducible research.

Ok, now that my evangelizing is done, lets get down to business. Lets say you have been working on a beautiful Jupyter notebook and now you have to present your results at a conference or business meeting. The standard way of doing this would be to gather up all of the figures, some text, and maybe some code from the notebook and put it into a PowerPoint/Keynote/Beamer presentation. If you then want to share this presentation more widely you may host them on Speaker Deck or just throw them on a Github repository somewhere. Actually, if you have already done this, then you are ahead of the game in terms of open-source science. But you did so much work on that notebook, wouldn’t it be better if you didn’t then have to spend all that extra valuable time making slides too? Well, we can do that with Jupyter notebooks and with a few quick hacks you can go from notebook to web-published slides.

The Notebook

I will assume a familiarity with Jupyter notebooks in general but not with the slideshow features. Lets say you have a notebook called nb.ipynb with some content. To go into slideshow mode click View -> Cell Toolbar -> Slideshow, this will now add a toolbar to each cell in the notebook from which you can choose what kind of slide you want each cell to be. You can choose Slide for a standard slide, Sub-Slide for a continuation slide, Fragment for revealing bullet points or other elements within a single slide, Notes to add slide notes, Skip to not include that cell in the slideshow. Here is an example notebook in case you don’t already have one ready. Lastly, if you include any extra images, place them in an img/ folder and commit that to your Github repo. You can view your notebook in the web browser with jupyter nbconvert nb.ipynb --to slides --post serve. Now that we have the notebook, lets do this!


Much of this setup is inspired by Dan Foreman-Mackey’s awesome article on continuous integration of academic papers and the first few steps are nearly identical to what is done there.

First, create a Github repo and commit your notebook. To get started with Travis, create a .travis.yml file and log in to Travis with your Github account and enable builds for your repository. You will also need to create a personal access token. Copy this token and go to the settings page of your repo on travis and add set

  1. GITHUB_API_KEY: set to the personal access token you just created.
  2. GITHUB_USER: set to your Github username.

OK, now that we have Travis set up lets edit the .travis.yml file:

language: python
cache: pip

    - "2.7"

# install jupyter and get reveal.js as we will need it to build the website
# from Travis
    - pip install jupyter
    - wget
    - unzip
    - mv reveal.js-master reveal.js

    - jupyter nbconvert nb.ipynb --to slides --reveal-prefix=reveal.js

after_success: |
    if [ -n "$GITHUB_API_KEY" ]; then
    git checkout --orphan gh-pages
    git rm -rf --cached .
    mv nb.slides.html index.html
    git add -f --ignore-errors index.html img reveal.js
    git -c'travis' -c'travis' commit -m init
    git push -f -q https://$GITHUB_USER:$$TRAVIS_REPO_SLUG gh-pages

Push this to your Github repository and if all goes well then your web-presentation should be available at If this worked, any further changes that you make to your Jupyter notebook will then be automatically built and displayed in the web-presentation. Here is my example Github repo and presentation.

Final Thoughts

Now you can make nice automated Jupyter notebook slides; however, I only showed the very basics of Jupyter Notebook slideshow options and themes. Things are pretty customizable if you aren’t opposed to some hacking. Lastly, I don’t think Jupyter notebooks are good for all presentations but it is great for code demos and talks that have code attached to them.


Leave a Comment