Workflow’s Guide For AiiDA

Creating new workflows

New user specific workflows should be put in aiida/workflows/user. If the workflow is general enough to be of interest for the community, the best is to create a git repository (e.g. on GitHub) and clone the content of the repository in a subfolder of aiida/workflows/user, e.g. in aiida/workflows/user/epfl_theos for workflows from the group THEOS at EPFL.

Put __init__.py files in all subdirectories of aiida/workflows/user to be able to import any workflows. Also, it may be a good idea to create a specific workflow factory to load easily workflows of the subdirectory. To do so place in your __init__.py file in the main workflow directory (e.g. in aiida/workflows/user/epfl_theos/__init__.py in the example above):

from aiida.orm.workflow import Workflow

def TheosWorkflowFactory(module):
    """
    Return a suitable Workflow subclass for the workflows defined here.
    """
    from aiida.common.pluginloader import BaseFactory

    return BaseFactory(module, Workflow, "aiida.workflows.user.epfl_theos")

In this example, a workflow located in e.g. aiida/workflows/user/epfl_theos/quantumespresso/pw.py can be loaded simply by typing:

TheosWorkflowFactory('quantumespresso.pw')

Note

The class name of the workflow should be compliant with the BaseFactory syntax. In the above example, it should be called PwWorkflow otherwise the workflow factory won’t work.

You can also customize your verdi shell by adding this function to the modules to be loaded automatically – see here for more information.