P3
Composer router timeout (eu-west-1)
§ 01 / 05
Statusall systemslive
All systems operational.
Status updates land every 60s. RSS feed, email digest, and Slack webhook subscriptions are all live.
Subscribe via RSS for raw events, email for human-readable digests, or write directly to status@exai.cloud to be added to the on-call broadcast list. The page tells you, first, whether it’s us or you.
Overall posture
7 of 8 services green. Audit-log streaming on watch.
0.000%
0
0
0
§ 02 / 05
Service · breakdown
Service-by-service.
Reading the same dashboard SRE reads.
Eight surfaces. Each row reports current month uptime, status, and the last filed incident date. Same numbers our on-call engineer sees on the wall.
Surface
Status
Uptime · MTD
Last incident
01
Workspace
Operational
0.000%
Apr 14
02
Composer
Operational
0.00%
Mar 22
03
Builder
Operational
0.000%
Jan 16
04
Orchestrator
Operational
0.000%
Feb 02
05
API
Operational
0.000%
Mar 22
06
Webhook bus
Operational
0.00%
Mar 22
07
Audit log streaming
Degraded
0.00%
ongoing
08
Authentication (SAML / SCIM)
Operational
0.000%
Nov 28
§ 03 / 05
Incidents · active & recent
What’s on fire now.
And what just was.
One open incident, owner-attributed. Plus the four most recent filings inside the 90-day window — every entry links to its annotated post-mortem.
P3
Incident · INC-2026-0428-01
Audit log streaming · Splunk shipping lag (eu-west-1)
Customer-managed Splunk HEC endpoint in eu-west-1 returning backpressure. Audit events are buffering and being shipped via the S3 fallback path. No event loss; delivery delay is currently under 90 seconds.
Timeline · UTC
14:22detected— eu-west-1 Splunk HEC backpressure
14:38mitigation— switched to S3 fallback
14:45monitoring— primary recovering
14:45ongoing— next update in 15m
Recent incidents · last 90d
P2
SCIM webhook delivery delayed
P3
Prebuild cache eviction
P3
Audit log Splunk lag
§ 04 / 05
Regions · live posture
Five regions. Five realities.
Latency, prebuild warm time, and cold-start are measured per region from in-region synthetic probes — refreshed every 60s. Model provider availability rolls up Anthropic, OpenAI, and Google upstream status.
Region · 01
us-east-1
0ms
0ms
0ms
Region · 02
us-west-2
0ms
0ms
0ms
Region · 03
eu-west-1
0ms
0ms
0ms
Region · 04
eu-central-1
0ms
0ms
0ms
Region · 05
ap-southeast-1
0ms
0ms
0ms
§ 05 / 05
Maintenance · subscribe
Scheduled change.
No surprises.
Maintenance windows are advanced 14 days — signed by SRE on duty and broadcast on every subscription channel before they open.
- 01May 12 · 02:00–02:45 UTC·eu-west-1Postgres minor upgrade
- 02May 19 · 03:00–04:00 UTC·globalKMS rotation (90-day cadence)
- 03Jun 02 · 02:00–03:30 UTC·ap-southeast-1Firecracker host refresh
Subscribe · status events
Get told first. Quietly.
Three pipes for status events: a raw RSS feed for the on-call channel, a daily email digest for platform owners, and a Slack webhook for team-room broadcasts. Subscribe to one — or all three.
Page is published from the same release pipeline as Composer and Orchestrator — every change to it is itself a tagged release.
Cross-checked by an external probe network. If our page goes down, the probe network publishes via status.exai.cloud/external.
Enterprise tenants can page the on-duty SRE through their named-account hotline. Response objective · 5 min P1.