Live status

Live numbers for this very page render plus a one-click load tester for the demo's endpoints. The bench loops in the browser — stats animate as requests complete. No node-side agent, just fetch with timing.

This response

Rendered server-side at 2026-05-05T10:52:30.355Z. The Server-Timing response header on this page carries compile / render / total in milliseconds — the panel below pulls them out of the current document so refresh re-paints.

Load test

Pick an endpoint and a duration. The harness fires fetch() in a tight loop with N concurrent requests. Numbers animate live — TPS is a 1-second rolling window, latencies are p50 / p99 over the run so far.

Note: latencies here are browser→server→browser round-trip and include your network RTT. Click Ping first to see the floor — at concurrency C, TPS is bounded above by C / ping_seconds, so a 350 ms RTT puts a hard ceiling at ~46 TPS for c=16 even when the server can do 4k+. oha (or any same-LAN client) will report the true backend throughput.