AI token usage and cost tracking

Generate safe AI activity, confirm token health, and distinguish healthy zero, no recent events, unavailable telemetry, and unconfigured routing.

Prerequisites

• AI provider configured
• At least one governed AI Chat or SDK request available

Validation path

1. Run one approved AI Chat or SDK request using non-sensitive test content.
2. Open Dashboard and confirm the token card shows tracking evidence, no recent routed events, disabled/unconfigured pipeline, or telemetry unavailable.
3. Use the token health endpoint evidence to confirm event count, last event time, source, coverage scope, cost basis, and ingestion lag.
4. If usage totals exist but no recent events appear, verify provider routing and analytics ingestion before escalating.

Validate

• Token health card loads
• Event count updates after routed AI activity
• Ingestion lag is visible when events exist
• No-data state is not treated as a failure
• Cost is labeled as estimated from observed routed traffic

Next steps

• Set token budgets
• Review grouped usage by workspace, user, session, provider, and model