Report real-user monitoring samples

POST/api/rum

ingestRum

Accepts a batch of browser-measured samples from the web app — Web Vitals (LCP, INP, CLS, TTFB, FCP), long animation frames, and the API calls behind a page with their cache outcome — and records each as one Prometheus histogram observation. Nothing is stored and no identity is read; anonymous callers are accepted. A request whose `Origin` is another site is answered `204` and records nothing. `route_id` and `api_route` are validated against closed sets (the app's route ids and this server's API path templates); a value outside them is recorded as `unknown`, never echoed into a label. A sample with a negative, non-finite or implausibly large value is dropped. The body is parsed as JSON whatever its `Content-Type`, so a `navigator.sendBeacon` with a plain-text blob works. At most 100 samples per batch (more is `400`); the body is capped at 16 KiB (`413`). Rate limited per IP.

Authentication optional — this operation also accepts anonymous requests. When authenticating, use a bearer token.

curl

curl -X POST 'https://neosource.dev/api/rum' \
  -H 'Content-Type: application/json' \
  -d '{"samples":[]}'

fetch

fetch("https://neosource.dev/api/rum", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
  },
  body: JSON.stringify({"samples":[]}),
});

Request bodyrequired

application/json

RumBatchRequest

object

Body of `POST /api/rum`: up to [`RUM_MAX_SAMPLES`] samples.

samplesrequired

array

items

RumSample

object

One measurement from one page view.

api_route: string | null
`api` only: the API path TEMPLATE the call was served by. A `String` for the same reason as `route_id`; a template the server does not route is recorded as `unknown`. Absent on a query-cache hit, which has no request to attribute.
example"/api/repos/{owner}/{repo}"
cache: one of
RumCacheOutcome
null
metricrequired: RumMetric
route_idrequired: string
The app's route id for the page (`/[owner=owner]/[repo]`), never a concrete path. Deliberately a `String`: the closed set is generated from the app's route tree, and an id the server does not know is recorded as `unknown` rather than refused.
example"/[owner=owner]/[repo]"
server_seconds: number | nulldouble
`api` only: the server's own `app` duration from the response's `Server-Timing`, in seconds, so `value − server_seconds` is the network and queueing share.
valuerequired: numberdouble
Seconds for every metric but `cls`, which is unitless. Negative or non-finite values are dropped.

Responses

204Batch recorded

413The body exceeded the route's 16 KiB cap.

Standard errors

Bodies documented once for the whole API — see standard errors.

400Bad Request — one of: invalid_input
429Rate limited — retry after the interval in the `Retry-After` header
500Internal server error
503Service temporarily unavailable / at capacity — retry after the `Retry-After` header
504Gateway timeout — the request exceeded the server's handling budget

Schemas

Referenced above. Listed here rather than expanded inline, so the same definition is not repeated at every level.

`RumCacheOutcome`

string

Where an API call's answer came from, as the page saw it.

"prefetch""query""network"

`RumMetric`

string

What one RUM sample measures.

"lcp""inp""cls""ttfb""fcp""loaf""api"