Set the deployment backend's health (the emergency kill switch)

false = emergency kill (backend excluded from routing entirely); true = clear the kill

True while an operator has frozen the load-driven autoscaler. The min/max band is kept as bounds and reconcile still converges the fleet to replicasDesired; the autoscaler just stops moving replicasDesired on its own. An operator can still hand-set replicasDesired within the band through /scale. Absent when unset.

Live catalog routing weight of this deployment's backend. Only present when backendRegistered is true. May differ from backendWeight (the row mirror) while a rollout ramp is in flight.

Live catalog health of this deployment's backend (the kill switch). Only present when backendRegistered is true.

Whether this deployment's pool:// backend is currently present on the catalog model row. Absent when the live catalog state could not be read; false until the first replica passes readiness.

The rollout tag this deployment's backend reports to the gateway as the version metric label on ig_model_requests_total and ig_stream_ttft_seconds. Use it to separate the two sides of a rollout in a dashboard query. Only present when backendRegistered is true.

Routing weight of the catalog backend. 0 = standby (blue/green starts here)

Operator-supplied reason for the revision that created this deployment

Inference engine, e.g. vllm or sglang

Engine container image override; absent = the engine's default

1 for a hand-created deployment, parent+1 for each revision. History ordering only: the controller never reads it, because rollout adoption is the explicit rolloutRole opt-in and nothing else.

True when replicas and their private endpoint pool are for direct operator testing only. Isolated deployments never register a catalog backend, never participate in blue/green rollout, and cannot receive gateway traffic.

Exclusive text-request byte ceiling for routing eligibility; 0 or absent = unbounded. Routing hint only, not a context limit.

Rolling only: how many replicas may be simultaneously not-ready (absolute count; surfaces show the percentage of desired).

Inclusive text-request byte floor for routing eligibility; 0 or absent = unbounded. Routing hint only, not a context limit.

Stable logical capacity lane within the public model. Blue/green only replaces versions in the same node pool; empty preserves legacy model-wide grouping.

Org that owns the managed replicas (and their metering)

The deployment this one was revised from. Set only by /revise, so an absent value means this deployment was created by hand.

Endpoint pool the deployment's replicas register in

Catalog model this deployment serves capacity for

How each replica is placed. Absent or "" or "vm" = a QEMU/VFIO VM on the baremetal fleet (the legacy shape; every pre-existing deployment reads as this). "pod" = a hardened GPU pod on a provider-onboarded runc node. One deployment is one runtime; a public model mixes runtimes through lanes.

True while an in-place rolling rollout is in flight on this deployment.

True while an operator has paused the ramp. Traffic is untouched: the deployment keeps the weight it holds and keeps serving, and the controller takes no rollout action until resume. The pause is also on rollout.paused; this raw column is what a surface falls back to when the lane could not be read at all.

Opt-in blue/green role. Empty (default) = an INDEPENDENT weighted capacity member of its lane, never blue or green. "green" = the lane's successor generation: the controller ramps it against the older members under a capacity guard and cuts them over once it reaches capacity parity. Rollouts are never implicit.

How a version bump reaches the fleet. blue_green (default) mints a ramped sibling; rolling replaces this deployment's replicas in place one at a time.

Model id the engine serves; must equal the catalog backend ref

Human-readable cause for degraded/failed states

Generation marker for blue/green within one public model and node pool; the target a rolling rollout replaces replicas to

Staged-checkpoint reference; absent = the engine loads from its own source