User Guide

This guide is for a modeller who runs climate-informed disease forecasts in DHIS2 after CAPS is installed. Installing the engine and the app is the Deployment Guide. Changing a handler or the app is the Development Guide.

A forecast needs a fixed order of operations: case data present and current in DHIS2, climate data obtained, a prediction run, and the forecast written back. The forecast then appears as the case numbers expected in the months ahead. That ordered set of operations is a workflow. CAPS holds the workflow as a pipeline: a named, reusable sequence of steps. Each step performs one operation through a handler. CAPS runs the steps in order, on demand or on a schedule.

CAPS coordinates three services:

• DHIS2, which holds the routine health data and receives the climate covariates and forecast results.
• The Open Climate Service, which provides climate datasets and aggregates them onto organisation units for a period.
• The Climate Health Analytics Platform (CHAP), which runs the prediction models.

CAPS is installed into a DHIS2 instance and uses that instance’s interface. It reads setups already configured in the services it connects to. It does not build models or define predictions.

How a run works

You create a pipeline as an ordered list of steps and assign each step a handler. When you trigger the pipeline, or when a schedule fires, CAPS places each step on its queue and runs the steps in order, passing the output of one step forward.

Many steps hand work to a service that does not finish immediately. For those, a later step polls the existing job at a configured interval until it completes or the configured limits are reached. Polling checks a job that is already running. Retry runs a failed step again. Each step has its own maximum retries and retry delay.

Every run is an execution with its own status, a step-by-step record, and logs.

What you can run

• Import climate datasets from the Open Climate Service and keep them current as new periods become available.
• Pull routine data from other DHIS2 instances into the instance used for modelling.
• Trigger a prediction against an approved CHAP evaluation and write the forecast back to DHIS2.
• Run a complete forecasting workflow on a monthly or weekly cycle.

Next

1. Open CAPS.
2. Prepare the data.
3. Create a pipeline, then schedule it or read a run.