This guide explains how to set up, monitor, and manage your background prediction jobs using both the dashboard and the .
Table of Contents
The Setup Process
Using the Setup Configurator (Recommended)
Most users should use the Job Setup guide located in the dashboard. This multi-step process automates the complex task of creating a Source, Version, and Job in one atomic transaction.
Manual Backend Setup
If you are building your own automation, you can use these individual endpoints:
1. Register a Source: POST /api/v1/sources (Defines the model metadata).
2. Define a Version: POST /api/v1/source_versions (Defines the input/output roles).
3. Finalize Job: POST /api/v1/jobs/setup (Binds the roles to physical hardware).
Managing Active Jobs
Once a job is running, you can manage it via the Jobs Dashboard or the following API endpoints:
List Jobs
GET /api/v1/jobs (or /jobs/customer/{org_id})
Returns all configured jobs for your organisation, including full binding and target matrices.
Update Metadata
PUT /api/v1/jobs/{id}
Updates descriptive metadata like name or notes.
Toggle Active Status
PATCH /api/v1/jobs/{id}/active
A lightweight toggle to start or stop a job without deleting it. Note: Inactive jobs are ignored by prediction workers.
Deleting a Job
DELETE /api/v1/jobs/{id}
Permanently removes the job and its role bindings.
Associated Virtual Sensors are NOT deleted during this process, allowing you to keep your historical prediction data for analytics.
Core Concepts & Logic
Understanding the underlying logic helps in advanced troubleshooting or complex integration scenarios.
- Source: A conceptual prediction engine (e.g., "Yield Predictor").
- Source Version: An immutable data contract defining Input Roles (metric needs) and Output Roles (prediction metrics produced).
- Job Role Binding: The "glue" that maps an abstract model requirement to a physical sensor UUID.
- Job Target: A destination for results. The system creates a Virtual Sensor for each target automatically.
The Internal Data Flow
1. Ingestion: Local workers fetch your configuration to see which hardware IDs they should query. 2. Execution: The prediction script processes the live telemetry. 3. Storage: Results are pushed to the Virtual Sensor targets via the ingestion pipeline.
Related Guides
- Job Setup — Comprehensive guide to the setup process.
- Predictions Guide — How to implement a worker that processes these jobs.