Riven Status
LiveLive service health and incident history for every Riven surface and the gateway behind them.
- Service
- surfaces_status
Status is the first place to look when something is not working. It reports current health per surface and per upstream inference pool, plus a history of incidents with what happened and what changed as a result.
What is monitored#
- Each surface: Chat, Computer, Console, Studio, Discover, Docs.
- The gateway:
api.rivenai.ioavailability and latency percentiles. - Inference pools: on-prem fleet health and cloud upstream reachability, which is what determines whether requests are falling back.
- Auth: sign-in and token issuance through Riven Auth.
Reading a degraded state#
Degraded rarely means down. If the on-prem fleet is unhealthy but cloud upstreams are green, requests still succeed — they are being served by fallback, which you can confirm per response with the x-riven-upstream header. A 503 with UPSTREAMS_EXHAUSTED means every pool was unavailable; see Errors.
Status is deliberately hosted so that it can survive the platform it reports on. If everything else is unreachable, status.rivenai.io should still answer.
Health signals come from Riven Telemetry. For the underlying probe endpoint used by orchestration, see GET /health on each service.