Deployment Patterns

Three production patterns, each appropriate to a different operational reality.

Cloud-native

A single Kubernetes cluster runs all components. Suitable for centralised state-level or national deployments.

• Helm chart: agrifooddata/agrifooddata
• Persistence: managed Postgres + managed object store
• Ingress: gateway behind a managed load balancer

Edge

Compute moves to the field — a gateway on the farm runs a slim agent that buffers observations and synchronises when connectivity is available.

• Edge agent runs on Raspberry Pi 4+ / generic ARM64
• Durable local queue for sensor observations
• Forward to a regional or cloud deployment on a schedule
• See Build · Sensor Integration

Hybrid

The most common pattern for ACRAT-style deployments: a regional cloud deployment handles APIs and persistence; edge gateways feed sensor data.

• Same Helm chart on the regional side
• Edge gateways treated as Sensor-Things clients
• Clearing-house contracts at the regional boundary