Dashboard

Call log

The call log is a real-time feed of every tracked LLM call, ordered most recent first. It's the ground truth for everything the dashboard aggregates.

Columns

  • Time — timestamp of the call
  • Feature — the feature tag passed in the SDK call
  • Model — the LLM model used
  • User — the user ID passed in the SDK call, if provided
  • Input tokens — tokens in the prompt and context
  • Output tokens — tokens in the model's response
  • Cost — calculated cost in USD for this call
  • Latency — end-to-end response time in milliseconds

Limits

The call log displays the most recent 100 events. For aggregate analysis across all events use the By feature and By user views.

Debugging with the call log

The call log is useful for catching issues immediately after a deploy. If your avg cost / call jumps after a release, find the first call after the deploy timestamp and compare its token counts against calls from before — the difference will tell you what changed.

Tip: pass a promptVersion in your SDK calls to make before/after comparisons easier to filter.

await trackedCall{feature: 'search', promptVersion: 'v2.1', ...})

Questions? hello@llmcosttracker.com