Platform status & performance
Current health of every VerifyAI component, plus the performance characteristics you can design against. This is a snapshot, not a live uptime monitor — component status reflects our last review and any active incident.
Status as of June 2026. Incident updates are posted on this page and sent to affected accounts by email.
Components
As of June 2026Verification API
Hosted photo-verification endpoints (verify, sessions, exceptions). Accepting and processing requests normally.
Dashboard
Web dashboard, analytics, exception triage queue, and account management. Loading and authenticating normally.
Webhooks
Outbound event delivery with the 1m / 5m / 30m retry schedule. Delivering and retrying on schedule.
SDK delivery / OTA
npm + pub.dev SDK distribution and server-side policy/prompt rollout to mobile clients. Publishing and serving updates.
Vision providers
Cloud VLM inference across OpenAI, Anthropic, and Gemini with automatic fallback. Provider health is reviewed in each status pass.
Performance characteristics
These are the latency envelopes we design and operate to. Real numbers vary with policy complexity, image size, network path, and the vision provider serving a given request.
End-to-end hosted verification typically resolves in well under a second on a standard parking or POD policy. Heavier policies (damage grading, OCR) add provider inference time.
Where the on-device ML path is enabled, the local detector runs an initial pass in under 200ms before anything leaves the device — fast enough for an in-flow capture experience.
The per-frame Streams API for fixed cameras returns a synchronous verification result in under 800ms, suitable for gated drive-through lanes.
Edge vs cloud
VerifyAI runs a hybrid pipeline. When the on-device ML path is enabled, a local detector does an initial pass on the device in under 200ms — fast enough to guide capture and screen obvious cases without a round trip. The hosted cloud pipeline then applies the full policy with semantic scene understanding and, where configured, damage grading, OCR, and report generation. You get an immediate on-device signal plus the depth and centralized control of server-side verification.
For the data behind these numbers, see our methodology and benchmarks.
Reliability & architecture
The platform is designed so that no single vision vendor is a hard dependency for a verification result.
Multi-provider VLM fallback
Cloud inference routes across OpenAI, Anthropic, and Gemini. If a provider is degraded, slow, or rate-limited, requests fail over to a healthy provider so verification results keep flowing.
Fail-soft delivery
Webhooks and integration deliveries use a retry queue (1m / 5m / 30m, then terminal). Optional vendor calls such as repair-cost estimation are fail-soft — the verification still returns if a downstream vendor times out.
Managed, compliant infrastructure
Hosted on SOC 2 / ISO 27001 certified infrastructure with TLS in transit. SDK delivery rides npm and pub.dev, and policy and prompt updates roll out server-side without forcing app releases.
Get notified about incidents
We post incident updates on this page during active incidents and email affected accounts directly. There's no separate signup — if your organization is impacted, the contacts on your account hear from us.
- Email alerts to affected accounts during active incidents
- Incident timeline and post-incident notes posted here
- Need a status SLA or a dedicated channel? Talk to us.
Currently reporting
No active incidents as of June 2026. All components are operational.