Skip to content

AI Usage

The AI Usage group covers two dashboards that answer the question what did the agent spend in token and economic terms — one tab per perspective. Both pages read from the same trace stream documented in Observability Architecture, and both feed off gen_ai.* span attributes that Spring AI emits per OpenTelemetry GenAI semantic conventions.

flowchart LR
    Span["gen_ai.client.operation<br/>span"]
    Price["Per-model<br/>rates"]
    Curr["Display<br/>currency"]
    TC["Tokens & Cost"]
    AM["AI Models"]
    Span --> TC
    Span --> AM
    Price --> TC
    Curr --> TC

Per-model rates and Display currency are both configured exclusively through the Model Pricing Manager dialog opened from the Tokens & Cost dashboard — the dialog is the only supported edit surface.

The two tabs share the same span data — they slice it differently. Tokens & Cost rolls up by money and tokens; AI Models rolls up by model identity, provider, and latency characteristic (including streaming TTFT). Cost lives only on Tokens & Cost because pricing is a per-model lookup applied at read time, not a property of the trace itself.

Pages in this group

  • Tokens & Cost

    gen_ai.usage.* spans × the configured per-model rates × display currency. $ per model, paid call share, cost projection, per-model token volume.

  • AI Models

    gen_ai.response.* + Spring AI streaming observation timers. Provider mix, model mix, TTFT, per-chunk latency, finish reasons, embedding and image duration.

For the underlying pipeline that captures these spans, see Observability Architecture → The pipeline. For the Model Pricing Manager dialog and how cost is computed, see the Tokens & Cost page.

Cross-references

Was this page helpful? Ask on Discussions · Report a docs issue