OTHER · STATUS
Status
Current state of every public Hoonify surface. The live page at status.hoonify.dev updates every 60 seconds and is the source of truth during incidents — this page is a 90-day rollup.
All systems operational
Services
- Inference (chat completions)Operational99.97%
- Inference (embeddings)Operational99.99%
- Compute / instances APIOperational99.94%
- Webhook deliveryOperational99.92%
- Pool · NAOperational99.97%
- Pool · EUOperational99.95%
- Pool · APACOperational99.91%
- Dashboard / customer portalOperational99.99%
Recent incidents
- 2026-04-1927 min
EU pool — elevated p99 on Llama-3.3-70B
Two operators in the EU pool hit a kernel-driver issue that raised p99 latency from ~120ms to ~480ms on Llama-3.3-70B. Hoonify drained both within 12 minutes; remaining capacity absorbed traffic. No requests dropped. Mitigated permanently with a driver pin in the operator base image.
- 2026-04-029 min
Webhook deliveries delayed in APAC
A capacity issue in our Singapore egress pool delayed webhook deliveries by 2-7 minutes for events originating from APAC compute. Eventual delivery succeeded. Mitigated by failover to the Tokyo egress pool; permanent capacity increase landed the same day.
- 2026-03-1114 min
Models API returned partial lists
GET /v1/models intermittently returned a subset of models (Qwen and DeepSeek families missing) due to a stale catalog cache after a deploy. Inference itself was unaffected — all models continued to serve. Cache TTL reduced and a startup warmup added.
SLAs
Target uptime is 99.9% for every public service, measured monthly. Credit-back applies automatically on any month under target — no support ticket required, no cap on the credit per incident. Reservation customers get an extended SLA at 99.95% with the same credit-back mechanic.
Subscribe
Get notified on incidents via:
- RSS ·
https://status.hoonify.dev/feed - Webhook · subscribe to
pool.degradedandpool.recoveredin Settings · Webhooks - Email · per-org distribution list under Settings · Security