Live Sphinx Documentation Preview January 30, 2013

Update (2015-06-11): There is now a nice installable package to make this even easier, sphinx-autobuild. All you have to do is

$ pip install sphinx-autobuild
$ sphinx-autobuild . _build/html

Update (2013-04-23): livereload now has a ‘shell’ compiler, which simplifies the Guardfile. The post has been modified to reflect this.

The big idea

I’ve been working with Middleman quite a bit, and really enjoyed their integration with livereload.js. When you save a change to a document you’re working on, the content is regenerated, and the browser is pretty seamlessly reloaded (including preserving the point you’re scrolled to.)

I’m now in the process of writing a lot of technical documentation with Sphinx, including sequence diagrams with PlantUML and graphs with Graphviz, so doing:

  1. Save file in editor
  2. make html
  3. ⌘-Tab to the browser
  4. ⌘ R

… was getting old when I’d been used to Middleman, where all you need to do is step 1.

I wanted a Python tool for this, since I didn’t want to make other people working on these documents to need to have both working Python and Ruby environments. Thankfully, there’s python-livereload. The only problem was that a Sphinx documentation tree looks like


and by default, python-livereload wanted to watch and serve from only the current working directory. One small pull request later and the problem was fixed.

Using it

It’s as easy as:

  1. Install python-livereload
  2. Create a Guardfile
  3. Tweak the Sphinx Makefile

Install livereload

This part’s easy.

    $ pip install livereload

(or however else you like to install python modules.)

If it’s already installed, upgrade it – newer releases have some required features.

Create a Guardfile

Create this file in the root of your Sphinx documentation tree. (Next to the Makefile.)

#!/usr/bin/env python
from livereload.task import Task
from livereload.compiler import shell

# You may have a different path, e.g. _source/
Task.add('source/', shell('make html'))

Run livereload

I chose to do this from the Makefile, so I get a new target: make livehtml. You have to make changes in 3 different spots in the file:

# Add the new target to the other .PHONY targets.
.PHONY: help clean html dirhtml singlehtml pickle json htmlhelp qthelp devhelp epub latex latexpdf text man changes linkcheck doctest gettext livehtml

# Add a note to the help
@echo "  livehtml   build html and point livereload server at them"

# And now the target; depends on 'make html', since livereload will only build if things change.
# We need to make sure the docs are current with any existing changes
livehtml: html
    livereload -b $(BUILDDIR)/html

So all you have to do is

$ make livehtml

and start editing.

The -b option was another pull request; it will actually open the browser and point it at localhost:<port>, which is a feature I’d never seen in other livereload implementations, but seemed quite convenient.


You still need to navigate to the page that you’re editing, though I discussed some ways to support ‘jump to the last file I edited’ with the python-livereload author. But after that, sphinx will rebuild, and your content will be updated, every time you save a file.