HTTP API Reference¶
The Dory platform exposes a small HTTP surface for operating processors — not for sending them work. Application traffic flows over the event mesh; the endpoints here are how the platform tracks processor and edge-node health and how operators wire up health checks and monitoring.
Audience
This page is for edge operators and platform operators, not for application developers consuming the platform. If you are building a processor, you don't call these endpoints — the SDK handles all of it for you (see Observability). The endpoints are documented here so operators understand what the platform exposes and how it decides a processor is healthy.
Edge operations¶
These endpoints support edge deployments — processors running on your own hardware out at the edge. They are used by Dory's own components inside the cluster, are not reachable by outside callers, and require authentication. They are summarized here so you understand the lifecycle, not as a public integration surface.
| Operation | Purpose |
|---|---|
| Heartbeat | Edge processors periodically report that they are alive and in good health. In each reply the platform tells the processor to either keep running or shut down (for example, when it has been retired). These reports are what drive failover decisions. |
| Register an edge node | Bring a new edge node under the platform's management so processors can be scheduled onto it. |
| Decommission an edge node | Retire an edge node, which stops the processors running on it. |
For how heartbeats translate into health and failover, see Edge Failover.
Health and monitoring endpoints¶
These are standard, unauthenticated endpoints that operators and infrastructure use to check that the platform's control plane is up and to collect its metrics.
| Endpoint | Tells you |
|---|---|
| Liveness | The control plane process is running. Used to detect a hung process that should be restarted. |
| Readiness | The control plane is fully started and its dependencies (database, cluster API) are reachable, so it's ready to do work. Used to decide when it can start receiving traffic. |
| Health | A roll-up of the control plane's internal checks (database, cluster connectivity, configuration watcher, and the health poller that checks running processors). Reports healthy, degraded, or unhealthy. |
| Metrics | Prometheus-format metrics about the platform and the processors it manages. See Metrics. |
The platform's own deployment uses the liveness and readiness endpoints to keep the control plane healthy, and your monitoring stack scrapes the metrics endpoint. None of these require credentials.