Metrics Reference

Total Spend

Sum of cost_usd across all events in the selected period. If cost_usd is not provided, Cloudshrink auto-calculates it using published token prices for the given provider and model.

Event Count

Number of LLM API calls tracked. Each POST to /api/ingest can contain multiple events; each event increments the count by one.

Token Types

• input_tokens — Prompt tokens sent to the model
• output_tokens — Completion tokens generated by the model
• cached_tokens — Prompt cache hits (reduces effective input cost)

Delta %

Period-over-period change: ((current - previous) / previous) x 100. Computed for total spend and for each project, actor, and model individually.

Spike Threshold

A >20% increase in cost for any project, actor, or model triggers a spike flag. Spikes appear on the dashboard and can trigger Slack or webhook notifications on Pro and Enterprise plans.

Unattributed %

Percentage of total spend from events missing project or actor metadata. A high unattributed percentage indicates instrumentation gaps that should be fixed.