RivenGet started

Search docs

Find a page across the Riven documentation

Riven Telemetry

Live

Metering and observability for the platform: tokens, latency, cost, and health, recorded per request and attributable to a key.

Telemetry is the platform's memory of what happened. It is the source for the usage you see in Console, the health that Status publishes, and the pool health that Ultra Router routes on.

What is recorded#

  • Tokens — prompt, completion, and total, per request, with the tier multiplier applied to give allowance spend.
  • Attribution — which key, plan, surface, alias, and resolved pool served the request.
  • Latency — time to first token and total duration, held as percentiles rather than averages.
  • Outcomes — success, failover, and error class, which is what makes a degraded upstream visible before users report it.
  • Cost — for cloud-served requests, the upstream vendor cost, so on-prem savings are measurable.

Reading your own usage#

Two paths, same underlying data. Every completion carries live quota state in response headers, and the quota endpoint returns the current window on demand.

response headers
x-riven-quota-plan: pro
x-riven-quota-used: 412000
x-riven-quota-limit: 5000000
x-riven-quota-reset: 2026-08-01T00:00:00.000Z

Spend aggregation is asynchronous. A request made a moment ago may not yet appear in GET /billing/quota, though its response headers were accurate at the time. Do not build a hard spending gate on the endpoint alone — read the headers.

Privacy#

Telemetry records the shape of traffic, not its content: counts, timings, and identifiers rather than prompts and completions. Retention and access commitments are published on Riven Trust.