Skip to main content
Usage & Logs is the per-request record of everything that went through your AIsa keys in the last 30 days. Where the Dashboard answers how much, this page answers which request — endpoint, model, latency, tokens, cache hits, cost, and status, one row at a time. Open it at Console → Usage & Logs. Every figure on the page reflects the filters currently applied, not just the rows visible on screen.

​
Summary Cards

Requests is the number of calls matching the current filters. Tokens is input plus output tokens across those calls. Cache Tokens is cache read plus cache write hits. On models that support prompt caching, a high read count against a low token count is what a working cache looks like. Usage Value is what the matching traffic was worth at list rates, with the amount actually charged to your balance below it. The two differ when a discount, promotion, or refund applied.

​
Usage Value Trend and Model Split

Usage Value Trend plots usage value in USD per day across the window. Hover — or focus the chart and use the arrow keys — to read any single point; the number in the top-right marks the peak. Model Split ranks models by usage value, up to five entries. Note that it is computed from the rows on the current page, so it follows your filters and paging rather than summarizing all 30 days — for the full-window breakdown, use the model split on the Dashboard.

​
Search and Filters

The search field matches against request ID, key, endpoint, and model — pasting a request ID from an error response or from the Playground jumps straight to that call. Three controls narrow the table further:
  • Rows — 10, 20, or 50 entries per page
  • Stream — All, Streaming, or Non-Streaming
  • Status — All, Success, Failed, or Refunded
Changing any filter resets you to the first page.

​
Log Table Columns

Each row is one API request. The footer reports which slice of the result set you are viewing, with previous and next controls. If nothing matches, the table says so rather than showing an empty grid — widen the date-independent filters or clear the search box.

​
Practical Uses

  • Debug a failed call — filter by Failed, find the request ID, and match it against your application logs
  • Explain a cost spike — sort out which model and key drove usage value on the peak day
  • Verify caching works — compare cache read counts against total tokens for repeated prompts
  • Check streaming behavior — filter to Streaming and compare time to first token against total latency
  • Audit key activity — search a key name to see everything it has done in the window

Dashboard Overview

High-level spend and usage metrics.

Pricing

Per-token rates driving the costs shown in logs.

Rate Limits

How request and token rates affect your traffic.