Skip to main content

Build a Data App

A data app connects an external data source to the platform — a machinery FMIS, a regional data catalogue, a weather service, a soil database. The job is to map the external schema onto the AgriFoodData data model and to keep them in sync.

Decision tree​

  • One-way read of an external system? → Connector pattern.
  • Bidirectional sync (writes back to the FMIS)? → Adapter pattern with a clearing-house contract.
  • Bulk historical import? → ETL pattern, runs once or on a schedule.

What you need​

  • Read access to the external system (API key, OAuth client, etc.).
  • A target organisation in AgriFoodData.
  • Knowledge of Farm and Activity APIs.

Step-by-step (read path)​

  1. Map fields. Match the external system's field / parcel / crop notion to the Farm API's Field + AgroVoc tags.
  2. Pull boundaries. Translate geometries to GeoJSON (RFC 7946).
  3. Sync activities. For every operation in the external system, write an Activity (with timestamps, inputs, outputs).
  4. Run on a schedule. Use the Service Registry's scheduling primitives.

Reference implementations​

The concept paper points to three real-world examples that should land here as full tutorials in later phases:

  • John Deere Operations Center read path (delivered by the NaLamKI consortium)
  • Agri-Gaia EDC connector (Gaia-X data space)
  • ADEX in Telangana

Common patterns​

  • AgroVoc translation table committed to the connector repo — make the mapping inspectable.
  • Watermarks for incremental sync (last-seen timestamps per resource).
  • Idempotent writes — re-running an import does not duplicate.