Your signal comes in.
Uh ohs come out.
Catch the uh oh before the oh no.
Spot unusual patterns sooner, so your team can investigate before small issues grow.
To get access to the private beta, contact @scottshebby on X.
Mostly business as usual. Then a spike at 215 and a dip at 379, around 42% and 74% through the signal. Classic uh oh behavior.
SYNTHETIC SIGNALWaiting for your signal.
Give it a signal to get a verdict.
Three steps.
Zero poker faces.
One signal, or an ongoing stream with a rolling window.
Two ways to uh ohGot a signal? POST it.
Sensor readings, latency, or your machine’s mysterious little numbers.
{ "signal": [42, 41, 85, …] }We find the funny bits.
We look for the moments that don’t fit the pattern. The spikes. The surprises.
﹏﹏╱╲﹏﹏Get a face. Take action.
Get the sample, the score, and the severity. Plus a face that says it all.
{ "emoji": "😬", "severity": 3 }Not all uh ohs are created equal.
From a raised eyebrow to a full existential moment.
Eight faces. One very expressive signal.
😬 Well, this is awkward. That definitely looks unusual.
One POST.
A whole lot of 😮.
No dashboard scavenger hunt. Just a small JSON response that fits right into your alerts, automations, and “hey, look at this” channels.
Bring your own TCN. This public demo uses rolling statistics. Create a token to run your saved model on complete signals or ongoing streams.
# Your signal, please.
curl -X POST https://uhohface.com/demo/analyze \
-H "Content-Type: application/json" \
-d '{"signal": [42, 41, 43, 42, 41, 43, 42, 85]}'{
"uhohs": [{
"index": 7,
"value": 85,
"severity": 8,
"emoji": "😱",
"name": "Maximum uh oh"
}]
}Time-series anomaly detection
What is uhohface?
uhohface is a time-series anomaly detection API for sensor readings, service metrics, and other measurements. It returns where an anomaly occurred and how unusual it is.
Can I try it without an account?
Yes. Try the public demo with a built-in example or paste a JSON array of your own measurements.
Can I monitor a live stream?
Yes. In the private beta, send timestamp-ordered samples to your stream endpoint. See how ongoing analysis works.
Your data has a tell.
Let’s find its uh oh face.