ObservabilityTraffic Analytics
Usage
Tokens, cost trend, and the relationship between context size and latency.
Usage
The Usage tab is the what is it costing, and where do tokens go view.
Visualizations
| Chart | Answers |
|---|---|
| Tokens and Cost | Is the bill explainable? Daily token total and cost in USD on a dual axis. |
| Tokens vs Latency | Is the bill costing user time? Scatter where each dot is one AI request, with tokens on x and latency on y. |
Tokens and Cost moving together means volume is driving spend. Cost rising with flat tokens points at a model swap or a pricing change — see Model pricing.
Workflow
Open this tab when the Tokens and Cost KPI card on the header shows growth.
Confirm the two series rise together. If they diverge, switch the Usage Records breakdown to Model to find which model is responsible.
Look at Tokens vs Latency for a new high-token, high-latency cluster — that is usually a context-window change in an application.
Tips
The scatter only includes AI records. MCP usage is measured in bytes — open the MCP tab for that.