Can staging stay out of the on-call channel?

Filter on env so only production spikes page anyone; staging can route to a quieter place or nowhere.

How is severity decided?

classify weighs the error class and count against your prompt and returns a typed severity the pipeline branches on.

Will one bad minute spam the channel?

You emit one spike event per window, so a burst is summarized as a count rather than streamed error by error.

Monitor error spikes in Webhook

Turn a sudden burst of one error into a single, classified alert — production only, severity already decided — instead of a wall of duplicate noise.

01source

sourcesdk.eventTypeScript SDK

matcherror.spike

02pipeline · 2 steps

01CTLfilter.matchenv = prod only
02ENRclassifyseverity → page | notify | ignore

03destinations · 1

towebhook.outWebhook
urlhttps://api.acme.com/hooks

the event

You emit error.spike with this shape. The TypeScript SDK keeps the call type-safe, and the event is stored whole — so every field below is available to the pipeline by name.

servicestring
errorstringerror class
countnumberin window
windowstringe.g. 5m
envstringprod | staging

emit it

From your code with the TypeScript SDK — or any language over the REST endpoint and signed webhook ingress.

emit error.spike

import { ingest } from "@ingestlayer/sdk";

await ingest("error.spike", {
  service: "checkout",
  error:   err.name,
  count:   windowCount,
  window:  "5m",
  env:     process.env.NODE_ENV,
});

route it to Webhook

POST the processed event as JSON to any HTTPS endpoint you control.

01
set the URL
Any HTTPS endpoint. The processed event is delivered as a JSON body on POST.
02
choose auth
None, a bearer token, or HMAC signing. Signed requests carry an X-Ingestlayer-Signature header you verify with your shared secret.
03
confirm receipt
Return a 2xx within the timeout. Non-2xx responses trigger retries with exponential backoff before the delivery dead-letters.

in webhookdelivered

POST /hooks HTTP/1.1
Host: api.acme.com
Content-Type: application/json
X-Ingestlayer-Signature: t=1717000000,v1=9f86d08…

{
  "type": "user.signed_up",
  "payload": { "email": "ada@acme.com", "plan": "pro" }
}

notes

Endpoints must respond within 10 seconds; slower responses are treated as failures and retried.
Retries use exponential backoff for several attempts before dead-lettering — make your handler idempotent.
Verify the HMAC signature before trusting a payload; the raw body is signed, so compute the digest before JSON parsing.

questions

Can staging stay out of the on-call channel?: Filter on env so only production spikes page anyone; staging can route to a quieter place or nowhere.
How is severity decided?: classify weighs the error class and count against your prompt and returns a typed severity the pipeline branches on.
Will one bad minute spam the channel?: You emit one spike event per window, so a burst is summarized as a count rather than streamed error by error.

build this pipeline or read the quickstart →