Welcome to Anemoi’s plugin documentation!

The Anemoi packages can be extended with plugins. This documentation provides examples of how to create plugins.

When to develop a plugin?

Anemoi plugins are intended to be developed for features that are specific to an organisation. For example, if you want to create datasets from your own archived data, encoded using an in-house bespoke format, then you should implement an anemoi-datasets source plugin.

If you intend to develop an anemoi-transform filter that could be used by many users, you should consider contributing it to the main repository.

What are plugins?

Plugins are Python classes that extend the functionality of the Anemoi packages, but are developed and maintained separately. Anemoi will look for plugins that have been installed in the Python environment and add them to the available functionality:

% pip install anemoi-dataset
% pip install my-dataset-plugin

Getting started

The following packages can be extended with plugins:

• anemoi-transform

  • Sources

  • Filters

• anemoi-datasets (create)

  • Sources

  • Filters

• anemoi-inference

  • Inputs

  • Outputs

  • Pre-processors

  • Post-processors

  • Runners

To get started with creating plugins, it is suggested that you install this package and run the anemoi-plugins new command to create a new plugin project.

% pip install anemoi-plugins
% anemoi-plugins new datasets.create.source --name my-source

Then, you can follow the instructions in the User Guide to create your plugin.

Other Anemoi packages

• anemoi-utils

• anemoi-transform

• anemoi-datasets

• anemoi-models

• anemoi-graphs

• anemoi-training

• anemoi-inference

• anemoi-registry

• anemoi-plugins

License

Anemoi is available under the open source Apache License.