Job Setup Guide

Create a prediction job from Jobs → Create job. Progress is saved automatically on the current device, so you can safely leave and return.

Before you start

You need an organisation and a clear idea of the data the model consumes and produces. You do not need to collect UUIDs in advance: the setup flow searches the devices, sensors, locations, and virtual sensors available to the selected organisation.

1. Job

Choose the organisation, enter a recognisable name, and decide whether the job should start active. Manual organisation IDs are available under Advanced for integration and support cases.

2. Model and release

A Model is called a source by the API. A model release is an immutable input/output contract.

• Choose an existing model and release when its contract already matches your prediction worker.
• Create a release when the model is familiar but its contract changed.
• Create a model when you are introducing a new algorithm.

New models and releases are defined inline; the job setup never needs to send you to another page.

3. Input connections

Each required input must be connected to at least one concrete data provider. Compatible sensors are shown first.

You can connect multiple providers to one input. When you do, select how their values should be combined: Average, Sum, Minimum, or Maximum.

Optional inputs may be left unconnected. Manual UUID entry remains available under Advanced.

4. Prediction outputs

Every output creates a virtual sensor. Give it a clear display name and add tags when other Yookr applications should be able to group or find it.

5. Review and create

The review step validates required connections and summarises every resource that will be used or created. If validation fails, the interface returns to the relevant step.

After creation

Save the generated service password immediately; it is only shown once. The success screen also provides the service ID, target IDs, and starter API requests for authentication and prediction ingestion.