What’s on the page
- Current status overview — health of the main components (API gateway, inference fleet, storage, auth, dashboard).
- Incidents — full incident history with timelines, impact summaries, root-cause when available, and resolution notes.
- Scheduled events — upcoming maintenance windows and automated notices about planned changes. These are created from our maintenance scheduler; users do not need to publish anything manually.
How to read statuses
- Operational / Degraded / Partial Outage / Major Outage — check impact scope (single tenant, region, or global) before taking action.
- Incident timeline — shows when detection, mitigation, and resolution occurred; check the timeline to find when your requests were impacted.
- Severity / impact notes — explains which user-visible features were affected (e.g., model deployment, inference, billing).
Automated publishing — what that means for you
- Monitoring alerts and on-call responders automatically create and update incidents; timelines and messages are programmatically generated from telemetry and on-call notes.
- Because this is automated, the status page is the single source of truth for platform health — always check it first before opening a support ticket.
Suggested actions during an incident
- Confirm the incident scope on the status page (global vs. specific component).
- If the incident affects only one of your models, check that model’s Logs tab (dashboard) for deployment-specific diagnostics.
- Retry idempotent operations after the incident moves to Degraded or Operational. For active inference sessions, expect possible timeouts—reconnect logic is recommended.
- If your systems depend on strict availability, implement retry/backoff and circuit-breaker logic and monitor the status page for recovery.
When to contact support
- If the status page shows no incident but you still experience failures, or if an incident’s impact on your account differs from the public description, contact support. When contacting support, include the request timestamp, model ID, and a short description of observed behavior.
Subscription & notifications
- Visit the status page to subscribe for updates using an email address. Use subscriptions to receive automated notifications for incidents and scheduled events.