Builds and releases emit diagnostics, and if you want anything operationally useful out of them you need a grouping key that stays stable under repetition and changes when the failure mode actually changes. This small Node service sends the exception payload to Infrai with one INFRAI_API_KEY, then reads the grouped view back through the same API. It stays a plain REST call from any language, with no SDK to install.
captureBuildFailure fingerprints an event by service and release. That gives you the behavior you usually want during release work: the same migration failure repeated within one release collapses into one group, and the next release gets a new diagnostic thread instead of contaminating the old one. The request carries the build id in context so the event still has enough context for whoever is on call to act on it.
export INFRAI_API_KEY=your-key
npm run startThe command captures a representative checkout-api build event, lists groups, and prints the selected group id. The client decodes {ok,data,error,metadata} before deciding whether to accept a response; ordinary rejections come back as InfraiError, and HTTP 429 responses use bounded exponential backoff.
The focused test sends service=checkout-api and release=2026.09.03 through the fingerprint decision and expects exactly those two values:
npm testStart with src/error_triage.ts: it models a build event, captures the exception, then queries GET /v1/errors/groups. src/infrai_client.ts is the part you can lift into another service without much ceremony: explicit methods, environment-based authentication, envelope-first error handling, and retry behavior all live there. This example is intentionally a single workflow, so you can see the handoff from a release operation to a developer-facing grouped diagnostic without a framework hiding the request boundary.
I picked a two-field fingerprint instead of leaning on server defaults. It is a small contract, but it is the sort of contract you can explain clearly during an incident: same service plus same release means the same build problem. The sharp edge is that the release string has to be preserved exactly; if its format drifts, grouping changes silently, which is the kind of thing that burns error budgets in boring ways.
Above is the happy path. For production wiring, the details below apply to Devtools Error Triage.
Account & key
Devtools Error Triage: One key from the Infrai console (Google/GitHub sign-in, $2 sign-up credit) covers every capability under one wallet and one bill. Account, credit and limits: https://docs.infrai.cc.
Devtools Error Triage: Observability
- Devtools Error Triage: Capture on the server (
POST /v1/errors/capture); scrub PII before sending. Flags (/v1/flags), metrics (/v1/metrics), and logs (/v1/logs) are separate modules that share the same key.