SA Dashboard — 359bbec6

Role: SA · Window: last 24h · Generated: 2026-05-20T23:42:31.144Z · Auto-refresh: 60s
Workers tracked
19
Total requests (24h)
3.6k
Overall error rate
0.03%
AI Gateway cost (24h)
$0.0000
How to read this: Summary stats give you the shape of the day. The error rate is the first thing to scan — anything over 5% on a Worker handling real traffic is worth a click-through. AI Gateway cost trending up faster than request count usually means a model swap or a cache-miss pattern; investigate.

Workers — by request volume (24h)

WorkerRequestsErrorsErr %CPU p50CPU p95
princess-polka-dottie 584 0 0.00% 333.00ms 945.00ms
daemon 490 0 0.00% 369.00ms 701.00ms
velovigil-fleet 463 0 0.00% 777.00ms 1616.00ms
perplexity-worker 299 0 0.00% 910.00ms 2815.00ms
mirror-worker 293 0 0.00% 1294.00ms 1906.00ms
gemini-worker 291 0 0.00% 914.00ms 1823.00ms
fleet-dashboard 289 0 0.00% 2234.00ms 5381.00ms
mistral-worker 285 0 0.00% 920.00ms 1939.00ms
fleet-relay 283 0 0.00% 470.00ms 780.00ms
overnight-builder 94 0 0.00% 699.00ms 1016.00ms
mycelia-api 93 0 0.00% 5065.00ms 10167.00ms
northwoods-daemon-network 58 0 0.00% 9566.00ms 20019.00ms
brook 24 0 0.00% 580.00ms 1129.00ms
mirror-app 23 0 0.00% 2136.00ms 2635.00ms
rhizome 15 0 0.00% 464.00ms 898.00ms
lookout 7 0 0.00% 3707.00ms 5399.00ms
ingest-cisa-kev 6 0 0.00% 496.00ms 779.00ms
quiettable 3 0 0.00% 596.00ms 734.00ms
fleet-operator-thinker 2 1 50.00% 1348.00ms 135936.00ms
Pattern to watch: p95 climbing while p50 stays flat is usually downstream contention (D1, KV, external API). p50 and p95 both climbing means the Worker itself is slowing down. CPU time is wall-clock for the Worker only; downstream wait time isn't here, which is why latency anomalies can hide in plain sight.