Skip to main content
GET
Get platform handoff latency

Authorizations

Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Response

Platform latency retrieved

Platform handoff latency over a trailing window.

Measured from the intents API receiving a request to the broker client being invoked for it, across every organization. This describes the platform, not a tenant, so every caller sees the same figure.

sample_count says how many handoffs the window held; every figure below is null when that is zero.

measured_at
string<date-time>
required

When the window was computed. Responses are cached briefly, so this can trail the request by a few seconds.

sample_count
integer<int64>
required
Required range: x >= 0
window_seconds
integer<int64>
required

The span the figures actually cover: from the oldest sample read to now.

The trailing hour on a busy platform, and further back on a quiet one, where the read reaches past the hour to find enough handoffs to take a tail over. Not a constant: render it rather than assuming an hour, or a three-day-old figure will be presented as this afternoon's.

max_ms
number<double> | null

Fractional, like the percentiles beside it. The window is stored and computed in microseconds; reporting the largest span as a whole number of milliseconds would be the one figure here that discards that resolution.

p50_ms
number<double> | null

Typical handoff cost, reported over any non-empty window.

p95_ms
number<double> | null

The tail, null until the read holds enough samples to characterise a distribution. Over a handful these describe the worst of that handful rather than the shape of anything.

Reaching that floor is what window_seconds can stretch for: a platform that dispatched nothing this hour still has a real tail, just an older one. These stay null only when the whole lookback is that thin.

p99_ms
number<double> | null