Sphinx AutoAPI

Documentation Status https://travis-ci.org/rtfd/sphinx-autoapi.svg?branch=master https://ci.appveyor.com/api/projects/status/5nd33gp2eq7411t1?svg=true


This is a pre-release version. Some or all features might not work yet.

Sphinx AutoAPI aims to provide “autodoc” or “javadoc” style documentation for Sphinx. The aim is to support all programming languages, be easy to use, and not require much configuration.

AutoAPI is a parse-only solution for both static and dynamic languages. This is in contrast to the traditional Sphinx autodoc, which is Python-only and uses code imports.

Full documentation can be found on Read the Docs.

Basic Workflow

Sphinx AutoAPI has the following structure:

  • Configure directory to look for source files
  • Serialize those source files, using language-specific tooling
  • Map the serialized data into standard AutoAPI Python objects
  • Generate RST through Jinja2 templates from those Python objects

This basic framework should be easy to implement in your language of choice. All you need to do is be able to generate a JSON structure that includes your API and docs for those classes, functions, etc.


First you need to install autoapi:

pip install sphinx-autoapi

Then add it to your Sphinx project’s conf.py:

extensions = ['autoapi.extension']

# Document Python Code
autoapi_type = 'python'
autoapi_dirs = ['path/to/python/files', 'path/to/more/python/files']

# Or, Document Go Code
autoapi_type = 'go'
autoapi_dirs = 'path/to/go/files'

AutoAPI will automatically add itself to the last TOCTree in your top-level index.rst.

This is needed because we will be outputting rst files into the autoapi directory. This adds it into the global TOCTree for your project, so that it appears in the menus.

We hope to be able to dynamically add items into the TOCTree, and remove this step. However, it is currently required.

See all available configuration options in AutoAPI Configuration.



The .NET mapping utilizes the tool docfx. To install docfx, first you’ll need to install a .NET runtime on your system.

The docfx tool can be installed with:

dnu commands install docfx

By default, docfx will output metadata files into the _api path. You can configure which path to output files into by setting the path in your docfx configuration file in your project’s repository. For example:

  "metadata": [{
    "dest": "docs/_api",


The dest configuration option is required to output to the docs/ path, where autoapi knows to search for these files.

With a working docfx toolchain, you can now add the configuration options to enable the .NET autoapi mapper. In your conf.py:

extensions = ['autoapi.extension']
autoapi_type = 'dotnet'
autoapi_dirs = ['..']

This configuration assumes your conf.py is in a docs/ path, and will use your parent path (‘..’) to search for files to pass to docfx. Unless you specify a custom pattern, using the autoapi_patterns option, sphinx-autoapi will assume a list of file names to search.

First, a docfx.json file will be searched for. If this file exists, it will be used, regardless of whether you have other file patterns listed. Otherwise, any file matching ['project.json', 'csproj', 'vsproj'] will be searched for.


Install go domain extension for sphinx.

pip install go sphinxcontrib_golangdomain

Install the go environment (from https://golang.org/dl/).

Install a git client for your environment (e.g. from https://desktop.github.com/).

Install our godocjson tool (preprocess godoc output to JSON, in a way similar to jsdoc -X).

go get github.com/rtfd/godocjson

Add go domain in your conf.py.

extensions = [ 'sphinxcontrib.golangdomain',...

Before running building your doc, make sure the godocjson executable is in your path.


Requires jsdoc, which in turn requires nodejs to be installed.

Install nodejs on your platform.

Install jsdoc using npm.

npm install jsdoc -g

Before building your doc, make sure the jsdoc executable is in your path.


All of the pages that AutoAPI generates are templated with Jinja2 templates. You can fully customize how pages are displayed on a per-object basis. Read more about it in Templates.


Read more about the deisgn in our Sphinx AutoAPI Reference Design.

Currently Implemented

  • Python (2 only – Epydoc doesn’t support Python 3)
  • .NET
  • Go
  • Javascript

Adding a new language

Adding a new language should only take a couple of hours, assuming your language has a tool to generate JSON from API documentation.

The steps to follow:

  • Add a new Mapper file in mappers/. It’s probably easiest to copy an existing one, like the Javascript or Python mappers.
  • Implement the create_class() and read_file() methods on the SphinxMapperBase.
  • Implement all appropriate object types on the PythonMapperBase
  • Add a test in the tests/test_integration.py, along with an example project for the testing.
  • Include it in the class mapping in mappers/base.py and extension.py