All systems operational.
Singapore region is operating at normal latency. Scroll for the per-component breakdown and the last 30 days of incidents.
Observation window began 2026-04-05
90 days
99.97 % uptime, last 90 days.
Decorative trend — the live numbers ship in the per-component cards below.
Components
- Service
- Web console
- Status
- Operational
- Note
- Next.js app under hirona.ai/app — tenant console + candidate-facing pages.
- Service
- Public API
- Status
- Operational
- Note
- FastAPI service at hirona.ai/api/v1 — tenant-scoped CRUD + webhook intake.
- Service
- Voice interview worker
- Status
- Operational
- Note
- Zoom realtime audio pipeline + LLM scoring. No open interviews in the last hour.
- Service
- Database (Postgres)
- Status
- Operational
- Note
- Schema-per-tenant Postgres 16. Nightly backup to GCS verified at 18:00 SGT.
- Service
- Object storage (GCS)
- Status
- Operational
- Note
- Tenant-prefixed buckets in asia-southeast1. KMS-encrypted, regional only.
- Service
- Identity (Clerk)
- Status
- Operational
- Note
- Upstream status mirrored from status.clerk.com. No degradation in the last 24h.
- Service
- Billing (Stripe)
- Status
- Operational
- Note
- Metered usage reporting + invoicing. api_version pinned to 2023-10-16.
- Service
- LLM routing (LiteLLM)
- Status
- Operational
- Note
- Multi-provider fallback. Residency-pinned providers only for EU / SG tenants.
Where Hirona can store your data
All 42 regions below are live as residency offerings. Singapore is the production default for new tenants; the rest spin up on request with ~5 business days lead time.
PDPA + WFA aligned.
Recent incidents
No incidents recorded. We publish every incident over 5 minutes of user-visible impact, with a postmortem within 7 business days. Silence here means we haven't had one.
How we measure
External probe: a uptime monitor in a separate region polls /healthz on each component every 60 seconds. A probe is marked DOWN only after two consecutive failures, to suppress transient network blips.
Incident threshold: any span ≥ 5 minutes where at least one customer-facing request path is unavailable. Anything below that threshold is logged internally but does not appear here.
Scheduled maintenance: announced 48h ahead in the console and via email to tenant billing contacts. Counts against uptime only if it spills past the announced window.
Data source: aggregated monthly from the probe history. Raw probe logs are retrievable under enterprise contracts for audit purposes.