Module worker 

Canonical worker types and identity.

This module defines the single source of truth for worker identity, type enums, and core configuration. These types are shared across API request/response boundaries and internal runtime state.

Structs

EngineAggregateMetricsSnapshot

EngineMemoryMetricsSnapshot

EngineQueueMetricsSnapshot

FlushCacheResult

Result from flush cache operations across workers

HealthCheckConfig

Health check configuration shared across protocol and runtime layers.

HealthCheckUpdate

Partial health check config for PATCH-style updates.

HttpPoolConfig

Per-worker HTTP connection pool configuration. All fields optional — None means “use router/global default”.

JobStatus

Job status for async control plane operations

ListWorkersQuery

Query parameters for GET /workers.

OverloadUpdate

Per-worker absolute overload threshold overrides. All fields optional — None means “use gateway default”. Either field set enables overload protection for this worker even when the gateway leaves it off. Mirrors HealthCheckUpdate pattern for PATCH-style config.

ProfileOptions

Options for starting a profiling run on workers.

ProfileResult

Result from profile start/stop operations across workers

ResilienceUpdate

Per-worker resilience overrides (retry + circuit breaker). All fields optional — None means “use router default”. Mirrors HealthCheckUpdate pattern for PATCH-style config.

SchedulerLoadSnapshot

Per-DP-rank load snapshot from a backend.

StartProfileRequest

Request body for the gateway /start_profile route: profile options plus an optional worker URL to target a single worker (e.g. one PD-disaggregation role). All workers are profiled when url is unset.

StopProfileRequest

Request body for the gateway /stop_profile route: optional worker URL to target a single worker.

WorkerApiResponse

Generic API response

WorkerErrorResponse

Error response

WorkerGroupKey

Composite key identifying a group of workers with the same characteristics.

WorkerInfo

Worker information for API responses.

WorkerListResponse

Worker list response

WorkerLoadInfo

Individual worker load information

WorkerLoadResponse

Full load response for a single worker across all DP ranks.

WorkerLoadsResult

Result from getting worker loads

WorkerSpec

Core worker identity and configuration.

WorkerStats

Worker statistics

WorkerTypeStats

Worker statistics by type

WorkerUpdateRequest

Worker update request

Enums

ConnectionMode

Connection mode for worker communication.

ProviderType

Provider type for external API transformations.

RuntimeType

Runtime implementation type for workers.

TransportMode

Multimodal tensor transport mode for large payloads.

WorkerModels

Models configuration for a worker.

WorkerStatus

Worker lifecycle status, modeled after Kubernetes pod conditions.

WorkerType

Worker type classification.

Constants

DEFAULT_WORKER_COST

DEFAULT_WORKER_PRIORITY