Skip to main content

AI cost and usage tracking

Auditable token, request, estimated-cost, pricing, freshness, and attribution reporting for AI traffic observed through governed routes.

Prerequisites​

  • Observed routed usage
  • A pricing entry for priced models

What it is​

Auditable token, request, estimated-cost, pricing, freshness, and attribution reporting for AI traffic observed through governed routes.

When to use it​

  1. Review routed usage and estimated spend
  2. Find unpriced or unattributed traffic

When not to use it​

  1. Do not interpret missing pricing or stale telemetry as zero cost

How to use it​

  1. Choose Cost
  2. Check status and freshness
  3. Review provider and consumer breakdowns
  4. Resolve unpriced or unattributed records

Example​

A priced request shows input/output tokens, a pricing snapshot, and a small nonzero estimated cost.

Permissions​

  1. Dashboard read access

Common failures​

  1. Usage event is stale
  2. Model pricing is unknown
  3. Principal attribution is missing

Validate​

  • Unpriced is distinct from zero-priced
  • Totals reconcile to source events

Next steps​

  • Review provider and consumer breakdowns
  • Resolve unpriced or unattributed records
Last updated on