API reference · Schema
EndpointReplica
objectstringrequiredAlways `endpoint_replica`: the type of this object
One of
endpoint_replicaidstringrequiredUnique ID (`rpl_…`)
ID with prefix rpl_
deployment_idstring | nullrequiredID of the deployment
statestringrequired`starting`: downloading weights and loading (not billed). `ready`: passed the engine health check, taking traffic, billed per second. `draining`: leaving (no new requests).
One of
startingreadyunhealthydrainingterminatedhardwareobjectrequiredThe hardware the replica runs on
ShowHide 1 field of
hardwareskustringrequiredHardware type code
node_idstring | nullrequiredThe node the replica is placed on (null until it is placed, and once removed)
ready_atstring (date-time) | nullrequiredTime ready (ISO 8601, UTC)
health_messagestring | nullrequiredThe engine's health check failure that made the replica `unhealthy`; null once it is ready again, or if it never failed one
drain_reasonstring | nullrequiredWhy the replica is draining: `stopped` (the endpoint was stopped or deleted, or its deployment removed) or `scale_down` (demand fell); null when not draining
termination_reasonstring | nullrequiredWhy the replica was removed, for example `drained`, `startup_timeout` (not ready within the startup limit) or `unhealthy` (failing health checks for too long); null while it is live
created_atstring (date-time)requiredWhen it was created (ISO 8601, UTC)
terminated_atstring (date-time) | nullrequiredTime terminated (ISO 8601, UTC)
{
"object": "endpoint_replica",
"id": "rpl_01jc5zq4m8k3v6x9t2w7r0b1e4",
"deployment_id": "string",
"state": "starting",
"hardware": {
"sku": "string"
},
"node_id": "string",
"ready_at": "2026-01-01T00:00:00.000Z",
"health_message": "string",
"drain_reason": "string",
"termination_reason": "string",
"created_at": "2026-01-01T00:00:00.000Z",
"terminated_at": "2026-01-01T00:00:00.000Z"
}Used by
- An endpoint's replicas, newest first
GET /v1/endpoints/{dedicated_endpoint_id}/replicas