ModelDeploymentStatus interface

The runtime status of a model deployment. All fields are read-only and populated once reconciliation has started.

Properties

currentReplicas

The current number of ready replicas serving traffic.

desiredReplicas

The desired replica count reported by the controller. Equals properties.scale.manual.replicas when manual scaling is used; current target replica count derived from autoscaler otherwise.

endpoint

The inference endpoint URL exposed by the deployment, once ready.

engine

The inference engine used to serve the model, e.g. "vllm".

engineVersion

The version of the inference engine, e.g. "0.17".

estimatedProvisionTimeSeconds

Estimated total time, in seconds, for the deployment to become ready end-to-end (GPU vm provisioning, image/weight pull, engine warm-up).

maxModelLen

The maximum model context length, in tokens, configured for this deployment.

peakTokensPerMinute

The peak tokens per minute measured by live stress test.

quantization

The quantization level applied to the model weights, e.g. "fp16", "awq-int4".

Property Details

currentReplicas

The current number of ready replicas serving traffic.

currentReplicas?: number

Property Value

number

desiredReplicas

The desired replica count reported by the controller. Equals properties.scale.manual.replicas when manual scaling is used; current target replica count derived from autoscaler otherwise.

desiredReplicas?: number

Property Value

number

endpoint

The inference endpoint URL exposed by the deployment, once ready.

endpoint?: string

Property Value

string

engine

The inference engine used to serve the model, e.g. "vllm".

engine?: string

Property Value

string

engineVersion

The version of the inference engine, e.g. "0.17".

engineVersion?: string

Property Value

string

estimatedProvisionTimeSeconds

Estimated total time, in seconds, for the deployment to become ready end-to-end (GPU vm provisioning, image/weight pull, engine warm-up).

estimatedProvisionTimeSeconds?: number

Property Value

number

maxModelLen

The maximum model context length, in tokens, configured for this deployment.

maxModelLen?: number

Property Value

number

peakTokensPerMinute

The peak tokens per minute measured by live stress test.

peakTokensPerMinute?: number

Property Value

number

quantization

The quantization level applied to the model weights, e.g. "fp16", "awq-int4".

quantization?: string

Property Value

string