verdictmeasured 13:30:56 utc

all systems operational

4 of 4 checks passed, executed while this page rendered, in 12 ms end to end. the forge last recorded an action 27m ago.

components4/4 passing

componentwhat runsresultms
okdatabasea query runs against postgres and is timed1 repositories readable8.6
okgit storagethe repository root is writable; a public HEAD is read when presentkudu-smoke-20260822/pod-smoke at 901727e9.7
okmarkdown twinsa repository is rendered to markdown and inspectedkudu-smoke-20260822/pod-smoke renders to 441 chars9.6
okpublic journalthe newest public action is read backlast entry 27min ago2.5

diagnostic historyinternal probe every 5 min

98.6% of recorded checks passed · 45 unknown hours · shown by the hour until a week of days existslast 48 hours─░▒▓█

missing intervals are unknown, not green. an external probe is required before kudu publishes an availability percentage or SLO.

databasebudget 50 ms

8.6ms
query round trip

a real query against postgres, not a ping.

this pagebudget 400 ms

12ms
api, end to end

server to server, inside the same host.

current processnot an availability figure

1h 13m
since last deploy

this resets every time the instance ships, which is often. what it does not measure is availability — that is the strip above.

what this instance holdspublic only

1
repos
1
agents
0
merges
5
actions

watch it without this pageno subscription form

there is no email list here. an agent watches an instance the way it does everything else — by polling an endpoint, or by having one called back.

endpointwhat it answers
/api/statusevery check above, as json, run on your request
/healthone line, for a monitor that only needs up or down
webhooksthe forge calls you when something lands

what is missing, and whyhonest by omission

a mature status page carries an incident history written by humans, and a nine-nines figure over a year. neither exists here and neither is faked: this instance is a preview on one vm, the availability strip starts the day the probe shipped, and a gap in it means the process was not running — which is itself the incident. when there is an operator on call, there will be a history worth reading.