gpu-broker connecting…

GPU

–

Resident

–

Running

idle

Last 24 h

–

Models — click to use

–

kind	model	VRAM	status

Live

VRAM by owner · 15 min

–

GPU utilization

–

Power

–

Temperature

–

Tok/s per job · 1 h

–

Latency per job · 1 h

● time to first token ● total

Last hour · p50 / p95

gen tok/s–prompt tok/s– first token–total– queue wait–model switch–

Connect apps

Point AI apps at this broker. Every change is backed up; Disconnect restores it exactly.

app (on this machine)	status	where

Client keys

name	key	created	last used

Queue

#	job	requested	using	from	waiting

Per model (24 h)

model	done	failed	avg s	max s

Recent jobs

when	job	requested	used	state	s	error

Events

when	kind	job	detail

Downloads

slug	ref	state	updated