Appendix: Deploying a Sphinx project online

When you are ready to show your documentation project to the world, there are many options available to do so. Since the HTML generated by Sphinx is static, you can decouple the process of building your HTML documentation from hosting such files in the platform of your choice. You will not need a sophisticated server running Python: virtually every web hosting service will suffice.

Therefore, the challenge is less how or where to serve the static HTML, but rather how to pick a workflow that automatically updates the deployed documentation every time there is a change in the source files.

The following sections describe some of the available options to deploy your online documentation, and give some background information. If you want to go directly to the practical part, you can skip to Publishing your documentation sources.

Sphinx-friendly deployment options

There are several possible options you have to host your Sphinx documentation. Some of them are:

Read the Docs

Read the Docs is an online service specialized in hosting technical documentation written in Sphinx, as well as MkDocs. They have a number of extra features, such as versioned documentation, traffic and search analytics, custom domains, user-defined redirects, and more.

GitHub Pages

GitHub Pages is a simple static web hosting tightly integrated with GitHub: static HTML is served from one of the branches of a project, and usually sources are stored in another branch so that the output can be updated every time the sources change (for example using GitHub Actions). It is free to use and supports custom domains.

GitLab Pages

GitLab Pages is a similar concept to GitHub Pages, integrated with GitLab and usually automated with GitLab CI instead.

Netlify

Netlify is a sophisticated hosting for static sites enhanced by client-side web technologies like JavaScript (so-called “Jamstack”). They offer support for headless content management systems and serverless computing.

Your own server

You can always use your own web server to host Sphinx HTML documentation. It is the option that gives more flexibility, but also more complexity.

All these options have zero cost, with the option of paying for extra features.

Embracing the “Docs as Code” philosophy

The free offerings of most of the options listed above require your documentation sources to be publicly available. Moreover, these services expect you to use a Version Control System, a technology that tracks the evolution of a collection of files as a series of snapshots (“commits”). The practice of writing documentation in plain text files with the same tools as the ones used for software development is commonly known as “Docs as Code”.

The most popular Version Control System nowadays is Git, a free and open source tool that is the backbone of services like GitHub and GitLab. Since both Read the Docs and Netlify have integrations with GitHub and GitLab, and both GitHub and GitLab have an integrated Pages product, the most effective way of automatically build your documentation online is to upload your sources to either of these Git hosting services.

Publishing your documentation sources

GitHub

The quickest way to upload an existing project to GitHub is to:

  1. Sign up for a GitHub account.

  2. Create a new repository.

  3. Open the “Upload files” page of your new repository.

  4. Select the files on your operating system file browser (in your case README.rst, lumache.py, the makefiles under the docs directory, and everything under docs/source) and drag them to the GitHub interface to upload them all.

  5. Click on the Commit changes button.

Note

Make sure you don’t upload the docs/build directory, as it contains the output generated by Sphinx and it will change every time you change the sources, complicating your workflow.

These steps do not require access to the command line or installing any additional software. To learn more, you can:

GitLab

Similarly to GitHub, the fastest way to upload your project to GitLab is using the web interface:

  1. Sign up for a GitLab account.

  2. Create a new blank project.

  3. Upload the project files (in your case README.rst, lumache.py, the makefiles under the docs directory, and everything under docs/source) one by one using the Upload File button 1.

Again, these steps do not require additional software on your computer. To learn more, you can:

Note

Make sure you don’t upload the docs/build directory, as it contains the output generated by Sphinx and it will change every time you change the sources, complicating your workflow.

1

At the time of writing, uploading whole directories to GitLab using only the web interface is not yet implemented.

Publishing your HTML documentation

Read the Docs

Read the Docs offers integration with both GitHub and GitLab. The quickest way of getting started is to follow the RTD tutorial, which is loosely based on this one. You can publish your sources on GitHub as explained in the previous section, then skip directly to Sign up for Read the Docs. If you choose GitLab instead, the process is similar.

GitHub Pages

GitHub Pages requires you to publish your sources on GitHub. After that, you will need an automated process that performs the make html step every time the sources change. That can be achieved using GitHub Actions.

After you have published your sources on GitHub, create a file named .github/workflows/sphinx.yml in your repository with the following contents:

.github/workflows/
name: Sphinx build

on: push

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v2
    - name: Build HTML
      uses: ammaraskar/sphinx-action@0.4
    - name: Upload artifacts
      uses: actions/upload-artifact@v1
      with:
        name: html-docs
        path: docs/build/html/
    - name: Deploy
      uses: peaceiris/actions-gh-pages@v3
      if: github.ref == 'refs/heads/main'
      with:
        github_token: ${{ secrets.GITHUB_TOKEN }}
        publish_dir: docs/build/html

This contains a GitHub Actions workflow with a single job of four steps:

  1. Checkout the code.

  2. Build the HTML documentation using Sphinx.

  3. Attach the HTML output the artifacts to the GitHub Actions job, for easier inspection.

  4. If the change happens on the default branch, take the contents of docs/build/html and push it to the gh-pages branch.

Next, you need to specify the dependencies for the make html step to be successful. For that, create a file docs/requirements.txt and add the following contents:

docs/requirements.txt
furo==2021.11.16

And finally, you are ready to enable GitHub Pages on your repository. For that, go to Settings, then Pages on the left sidebar, select the gh-pages branch in the “Source” dropdown menu, and click Save. After a few minutes, you should be able to see your HTML at the designated URL.

GitLab Pages

GitLab Pages, on the other hand, requires you to publish your sources on GitLab. When you are ready, you can automate the process of running make html using GitLab CI.

After you have published your sources on GitLab, create a file named .gitlab-ci.yml in your repository with these contents:

.gitlab-ci.yml
stages:
  - deploy

pages:
  stage: deploy
  image: python:3.9-slim
  before_script:
    - apt-get update && apt-get install make --no-install-recommends -y
    - python -m pip install sphinx furo
  script:
    - cd docs && make html
  after_script:
    - mv docs/build/html/ ./public/
  artifacts:
    paths:
    - public
  rules:
    - if: $CI_COMMIT_REF_NAME == $CI_DEFAULT_BRANCH

This contains a GitLab CI workflow with one job of several steps:

  1. Install the necessary dependencies.

  2. Build the HTML documentation using Sphinx.

  3. Move the output to a known artifacts location.

Note

You will need to validate your account by entering a payment method (you will be charged a small amount that will then be reimbursed).

After that, if the pipeline is successful, you should be able to see your HTML at the designated URL.