Liveness
Each service is probed over HTTPS on a short interval. Catches hard-down, process death and TLS failure — but only proves a server answered.
Independent availability monitoring
duck detector!
We watch the ducks. Somebody has to.
What this is?
Duck Detector is the watchdog watchduck for the qwack! suite — the text, email and voice
platform behind twack!, smack! and qwack!. It exists to answer one question honestly, at any hour,
without us being asked: is the platform actually working right now?
Most status pages are marketing. They are hosted on the same infrastructure they claim to monitor, which means when the platform goes down, the page that would have told you goes down with it. Duck Detector is deliberately built the other way around: it runs off our cloud entirely, on infrastructure that shares no server, no network and no account with the systems it watches. If we go dark, this page stays lit — and tells you so.
...but who's watching the watchers? 🧐
The design principle is simple: every watcher is itself watched, and no alert path shares a failure domain with the thing it monitors. Checks that quietly stop running are the most dangerous kind of failure, so the watchers report in even when everything is fine. Silence is treated as a fault, not as good news.
Everything published here is measured, not asserted. We do not round up, and we do not quietly redefine "down." When we eventually commit to a service level, the number will come from the history below — not from a brochure.
Live status
Current state of each component of the platform. Probed continuously from outside our infrastructure.
Loading components…
Upstream dependencies. These are third-party carriers and platforms in the message path. We report their state for transparency, but we do not control them — and they are excluded from our own service-level measurement.
Measured uptime
One bar per day. Colour reflects the worst state observed that day. Hover any bar for detail.
—
Awaiting data
90 days agoToday
Operational
Degraded
Outage
No data
Incident history
Every incident is published in the same three parts: what we detected, what we fixed, and what we changed so it cannot recur. We publish the prevention step because it is the only part that actually matters to you.
No incidents recorded in the current measurement window.
How we measure
Transparency means showing the method, not just the score. Here is exactly what is being checked and from where.
Synthetic transaction
A real message is sent end-to-end through the live carrier path on a schedule and matched by a unique nonce on arrival. This is the check that proves the product actually works, not merely that a process is listening.
Off-cloud watcher
The probes run on independent infrastructure with no shared server, network or account with the monitored systems — so a total loss of our cloud cannot suppress the alarm.
Dead-man's switch
The watchers publish on a fixed heartbeat whether or not anything is wrong. A missing heartbeat is itself an alert: it means the monitor died, and we treat that as an outage of the monitor.
Append-only ledger
Every status write is committed to a public, timestamped, append-only history. Past results cannot be quietly rewritten to flatter a number.
What we don't publish
Raw logs, internal hostnames, IPs and file paths are never exposed here. Curated status is deliberate: an operational detail page is a reconnaissance surface.