Management
System management endpoints
Get system version
Get system version
System health check
System health check
Scheduler cost/delay distribution
Returns an aggregate, low-cardinality distribution of the scheduler's per-job cost_ewma_ms and delay_ewma_ms across all check_jobs, plus fast/slow job counts at a candidate threshold. Read-only and super-admin guarded. Used to decide whether fast/slow scheduling lanes are warranted.
Rate-limit and concurrency introspection
Returns the server's rate-limit and concurrency configuration together with the calling client's live counters (remaining tokens, in-flight requests), the resolved caller IP and which bucket ('ip' or 'token') the caller's traffic is accounted against. Public, no authentication.
Runtime memory snapshot
Returns a JSON snapshot of runtime memory (heap / stack / goroutines / GC and the runtime/metrics memory classes), the process RSS breakdown from /proc (anon / file / shmem split, peak, threads, smaps_rollup Pss), the container's own cgroup accounting, the derived off-heap gap, suspect-subsystem cardinalities (DEK cache, rate-limit entries, event listeners) and build facts (cgo, SQLite driver, Go version). Linux-only sections report `present: false` on other platforms rather than failing. Super-admin only; the raw pprof surface stays on the localhost-bound profiler server.