Introduction

This guide is intended to provide a quick and easy way to get started with developing, deploying and running custom extensions for Kaapana.

It consists of three parts. This part gives a brief overview over the technologies and needed preparations for the rest of the guide. The part Developing Workflow Extensions focuses on the integration of custom processing workflows into the platform. These can be as simple as collecting metadata from DICOM images or as complex as a full-fledged machine learning pipeline. This section explains how to write Airflow-DAGs and how to integrate custom workflows as extensions into the Kaapana technology stack. The last section Developing Applications provides instructions on deploying a general-purpose web application inside Kaapana.

Important

The reader should be familiar with basic concepts of the Kaapana platform. If you are not, please refer to the User Guide for a general introduction to the concept of the platform and its components.

Terminology

List of Kaapana specific terms used in this guide. Note that the reader is also expected to know about containers, images and registries, which are not explained here.

  • Extension: A component that can be installed and run on the Kaapana platform. List of stable extensions can be found in the Extensions section of the documentation.

  • Application Extension: These extensions deploy standalone web applications, usually accessible via subpaths in the platform (e.g. JupyterLab, MITKWorkbench)

  • Workflow Extension: These extensions deploy workflows, which are executed via Airflow. Every processing pipeline inside Kaapana is considered a workflow (e.g. nnunet-training, totalsegmentator, collect-metadata)

  • DAG: Airflow Directed Acyclic Graph, defines a collection of operators that are executed in a specific order. Every workflow inside Kaapana defines a DAG that is managed by Airflow.

  • Operator: A single task inside a DAG. Kaapana utilizes distributed operators, where each operator is linked to a specific container that is executed in a Kubernetes pod. The scheduler of Airflow is responsible for executing the operators in the correct order.

  • Chart: A Kaapana extension is technically a Helm package which contains its configuration. It references the container images of the extension, and defines the necessary Kubernetes resources around them.

  • Microk8s: Kubernetes distribution used in Kaapana. It contains the container runtime called containerd, abbreviated as ctr, which is used to run containers inside the platform.

  • EDK: Extension Development Kit, available for versions >= 0.4.0. It provides a development environment inside Kaapana for building and deploying containers and charts. For more information Extension Development Kit (EDK)

List of Technologies

The following technologies will be referenced throughout the guide:

  • Docker or Podman: Necessary for building containers on a local machine

  • Airflow: Workflow management system used in Kaapana

  • Kubernetes: Container orchestration system, every component in Kaapana is inside a Kubernetes cluster. We deploy the platform with microk8s .

  • Helm: Package manager for Kubernetes. Necessary when you want to build helm packages on your local machine