utzspreQuiet when healthy.
Clear when it isn’t.
Scheduled checks, recorded evidence, and reviewable incident triage for Autzspre.
Retrieving recorded observations…
The latest check.
A healthy check is one successful sample, not an uptime guarantee. After eight healthy samples, response time is compared with the median of up to 24 prior healthy observations. Alert threshold: above 3× median and 2.5 seconds. AI checks verify configuration only, not model inference.
Keep the evidence.
Every observation is reviewable.
Built in stages. Clearly labelled.
Evidence first. Human review next.
Up to 12 new free AI drafts are allowed per day across check reviews and incident triage. Clicking an AI review shares only the public check observations with OpenRouter and its model provider. New recorded incidents automatically request one free AI triage draft, subject to daily capacity. Incident triage is a text-only AI draft based on these check results. It cannot inspect repositories, execute code, launch a sandbox, create a PR, or fix a deployment. Those workflows need separate repository and execution-provider authorization.
Existing provider keys remain server-side. No secrets are sent in the triage prompt. A sandbox credential broker and application-managed AES-256-GCM vault have not been implemented; no such claim is made. Public account signup remains disabled.
Slack notifications send a check name, failure summary and dashboard link when a check changes state, plus recovery notices. The connection is active only after a channel webhook is authorized and securely configured. No routine healthy-status spam.