Strata monitor
Live Strata inference server status in Paseo: sidebar prefill/decode tok/s and a detail screen
Shows live status of a local Strata inference server. A sidebar row is always visible and shows decode and prefill tokens per second (the last request's speeds when idle); clicking it opens a screen with a speed graph, request state and queue, GPU and memory use, and recent requests.
The daemon reads Strata's /metrics endpoint once per poll and shares that result between open clients. Polling runs only while the screen is open, every second while a request is running and every three seconds otherwise.
By default the daemon tries http://127.0.0.1:8080 and then http://127.0.0.1:8093. Set STRATA_URL in the daemon environment to use another address, and STRATA_API_KEY if the server needs a bearer token. Only the Strata server's own metrics are read; nothing is sent elsewhere.