Skip to main content
A span is one unit of work inside a request — a query, a cache call, an outbound request. These endpoints aggregate spans across requests, which is how you find the individual operation responsible for a slow endpoint. For request-level time, use the APM API. All span endpoints require a bearer token with metrics read access, and all times are ISO 8601 in UTC. Span types: db.query, redis, cache, http.out, queue, rpc, graphql, custom.

Operation Report

Summarize spans by fingerprint, type, endpoint, name, system, host, database, code location, bucket or environment, with per-group quantiles.

Path Parameters

string
required
Account identifier (format: acc_ followed by 27 alphanumeric characters)

Query Parameters

string
required
One of fingerprint, type, endpoint, name, system, host, db, call_site, bucket_sid, environment_sid
string
required
Start of the window (ISO 8601, UTC)
string
required
End of the window (ISO 8601, UTC)
string
Limit to one bucket
string
Limit to one environment
string
Limit to one normalized endpoint
string
Limit to one span type
string
Limit to one span fingerprint
number
Maximum groups to return (1–1000, default 50)

Response

string
The field the results are grouped by.
number
Number of groups matched.
array
Each entry carries key, runs, errorCount, avg_us, p50_us, p95_us, p99_us, total_us, requests (distinct requests the group ran in), runsPerRequest (high values indicate an N+1), lastSeen, and sample* fields describing a representative span (sampleType, sampleName, sampleSystem, sampleHost, sampleDb, sampleEndpoint, sampleStatement, plus sampleCallSite / sampleCallSiteFn / sampleCallStack, which are empty unless code location capture was on when the spans were recorded). sampleCallStack holds the caller chain innermost first, one file:line:col|function frame per line.
Durations are microseconds (_us) here, unlike the millisecond fields on the request-level APM endpoints — spans are frequently sub-millisecond.

Operation Trend

Chart one operation over time — runs, distinct requests, errors and latency quantiles. The bucket width is derived from the requested window, so the series stays bounded for any range.

Query Parameters

string
required
Start of the window (ISO 8601, UTC)
string
required
End of the window (ISO 8601, UTC)
string
Span fingerprint to chart. Strongly recommended — it is what keeps the read narrow.
string
Limit to one span type
string
Limit to one normalized endpoint
string
Limit to one bucket
string
Limit to one environment

Response

array
Each entry carries bucket, runs, requests, errorCount, avg_us, p95_us, p99_us and total_us.

Drill Back To Requests

The requests a span fingerprint ran in, so you can go from a slow query to a real request and inspect its whole waterfall.

Query Parameters

string
required
Span fingerprint to drill back from
string
required
Start of the window (ISO 8601, UTC)
string
required
End of the window (ISO 8601, UTC)
string
Limit to one normalized endpoint
string
Limit to one bucket
string
Limit to one environment
number
Maximum requests to return (1–1000, default 50)

Response

string
The fingerprint that was queried.
array
Each entry carries request_sid, endpoint, spanCount (times the operation ran in that request), total_us and lastSeen.