Plugin Components

A plugin is made of a number of components. Each component is a single kind of object in Dataiku DSS, such as a dataset, recipe, or webapp. Each component adds an additional functionality to a plugin. A plugin bundles components together as a single consistent whole.

The (non-exhaustive) list of plugin components includes:

• Dataset
• Recipe
• Format extractors
• Format exporters
• Filesystem providers
• Macros
• Metric probes and checks
• Scenario triggers and steps
• Preparation script processors
• Shared code to import in recipes or notebooks
• Webapps
• Prediction algorithm
• Custom Fields
• Parameter set
• Custom Policy Hooks

The most up-to-date list of possible components can be found in the product. In the plugin editor, click + Add > Create component. The resulting dialog shows the list of possible components. When you add a component, Dataiku DSS automatically makes the appropriate additions to the plugin directory structure, and adds some starter code to help you get started.

Structure of the plugin

The elements of a plugin are contained within a top-level directory that identifies the plugin. When you create a plugin from scratch (called, for example, myplugin), that top-level directory contains:

• A python-lib subdirectory with a myplugin subdirectory that has an __init__.py file that is empty of code. The python-lib directory is a good place to put functions that will be reused throughout the plugin.
• A plugin.json file that describes the plugin as a whole. As a best practice, the id element of this JSON file should be the same as the name of the top-level directory.

As you add components to the plugin, this adds to the plugin folder structure. Each type of component has a subdirectory under the top-level directory. Each individual component has a directory that contains a JSON file that describes the component as a whole, and code files that define what the component does.

Each component generally has some configuration parameters.