Module request_plane 

Request-plane metrics for AddressedPushRouter. Used to pinpoint serialization vs transport roundtrip latency.

Statics

REQUEST_PLANE_INFLIGHT

Currently in-flight requests (incremented at generate() entry, decremented on stream complete).

REQUEST_PLANE_QUEUE_SECONDS

Time from generate() entry to send_request() (serialization + encoding + control message).

REQUEST_PLANE_ROUNDTRIP_TTFT_SECONDS

Time from send_request() to first response item (transport roundtrip TTFT).

REQUEST_PLANE_SEND_SECONDS

Time for send_request() to complete (frontend view: network + queue + ack).

Functions

ensure_request_plane_metrics_registered

Register request-plane metrics with the given registry. Idempotent; only the first call registers.

ensure_request_plane_metrics_registered_prometheus

Register request-plane metrics with a raw Prometheus registry (e.g. for LLM HTTP service /metrics). Idempotent; only the first call registers. Call this when the service exposes /metrics from its own registry.