Label-bound wrapper around the raw Prometheus metrics.
Metric collectors use the vllm_omni: prefix, distinct from the upstream vllm:* families.
observe_tokens
observe_tokens(
prompt_tokens: int, generation_tokens: int
) -> None
request_succeeded
request_succeeded(
e2e_seconds: float, finished_reason: str = "stop"
) -> None
set_running
set_running(n: int) -> None
set_waiting
set_waiting(n: int) -> None