Publishing a plugin

1. Choose a name

The naming convention for AiiDA plugins is aiida-mycode for the plugin and aiida_mycode for the corresponding python package, leading to the following folder structure:


This marks your plugin as an AiiDA package and makes it easy to find on package indices like PyPI.

Note: Python packages cannot contain dashes, thus the underscore.

2. Get Your Plugin Listed

AiiDA plugins should be listed on the AiiDA plugin registry to avoid name-clashes with other plugins.

If you wish to get your plugin listed on the official registry for AiiDA plugins, you will provide the following keyword arguments as key-value pairs in a setup.json or setup.yaml. It is recommended to have read the keyword arguments from that file:

   setup.json       # or setup.yaml
  • name
  • author
  • author_email
  • description
  • url
  • license
  • classifiers (optional)
  • version
  • install_requires
  • entry_points
  • scripts (optional)

Now, fork the plugin registry repository, fill in your plugin’s information in the same fashion as the plugins already registered, and create a pull request. The registry will allow users to discover your plugin using verdi plugin search (note: the latter verdi command is not yet implemented in AiiDA).

3. Get Your Plugin On PyPI

For packaging and distributing AiiDA plugins, we recommend to follow existing guidelines for packaging python, which include making the plugin available on the python package index. This makes it possible for users to simply pip install aiida-myplugin.

Our suggested layout:

aiida-compute/       top-folder containing you package and additional files
   aiida_compute/    The package that is to be installed
      ...       (optional) lists non-python files to be installed
   README.rst        (optional) description to be used by github etc and PyPI          installation script
   setup.json        contains requirements, metainformation, etc

Note: In principle, aiida-compute could contain and install multiple packages.

Incidentally a distribution can contain and install more than one package at a time.

The most user-friendly way to distribute a package is to create such a distribution and uploading it to PyPI. Users then can simply install the package(s) by running pip <distribution-name>.