Service status
Real-time operational status for HOUJIAYAN API and supporting infrastructure.
Auto-refreshes every 30 seconds
Components
- API Gateway Primary API endpoint (api.houjiayan.com) — NewAPI + nginx + Cloudflare Tunnel— Checking…
- Main Website Marketing site and documentation (houjiayan.com)— Checking…
- API Status API Public /api/status endpoint with channel and model info— Checking…
- DeepSeek upstream Chat and reasoning models routed via gateway— Checking…
- Qwen upstream Alibaba Qwen chat models routed via gateway— Checking…
- MiniMax upstream MiniMax multimodal models routed via gateway— Checking…
- Volcengine Ark upstream ByteDance Doubao chat + image models routed via gateway— Checking…
- Moonshot Kimi upstream Moonshot AI Kimi chat + coding models routed via gateway— Checking…
Service uptime
Live rolling uptime for each component over the last 30 days, 7 days, and 24 hours. Polled every 5 minutes by the gateway health monitor (see /__status.json). 90 days of history will be published as the dataset matures.
Aggregate across all monitored components. Raw values available at /api/__status.json.
Incident history (last 90 days)
Past incidents and resolutions. Incidents affecting a single upstream provider are still considered fully resolved (auto-failover handled user traffic).
Brief Qwen-compatible-mode latency spike
Some Qwen calls via the OpenAI-compatible mode returned 30s+ response times between 14:23 and 14:31 UTC. We routed traffic to the Qwen native type=17 endpoint as a workaround. Native endpoint latency stayed normal throughout. Root cause: Aliyun load-balancer hot-spot in the compatible-mode pool. No customer-visible failures; only tail latency. Resolved when Aliyun rebalanced the pool.
Cloudflare Tunnel reconnect blip
Tunnel daemon reconnected after a transient outbound network drop on the VPS. The 47s window showed 'Bad Gateway' on the public endpoint. No requests lost; in-flight requests were retried by client SDKs.
DeepSeek upstream rate-limit event
DeepSeek returned 429 on ~30% of requests between 03:00 and 03:14 UTC. Our gateway auto-retried, but customers using raw openai SDK saw 429 spikes. We documented the incident and added a per-user rate-limit gate (60 RPM, 100K TPM) to absorb upstream bursts. DeepSeek's status was 'elevated errors' for the same window.
Scheduled NewAPI upgrade (v0.5 → v0.6)
Planned upgrade of the NewAPI Docker image. Service was put behind maintenance mode for 3 minutes; no requests lost. Post-upgrade verified all 11 models returned 200.
Scheduled maintenance
Planned maintenance windows. We aim to give 48 hours' notice. During maintenance, the affected component will be marked 'Degraded' in advance.
No scheduled maintenance at this time.
Get notified
Email support@houjiayan.com to receive status updates for planned maintenance and incidents.