Skip to main content
KeyRing Docs
Documentation/Metrics & Analytics

Metrics & Analytics

Use the Metrics module to review local KPI cards, filtered request logs, session records, event streams, and exportable analytics data.

Read time: 8 min read

What Metrics Does

Metrics is KeyRing's local analytics workspace. It reads from the desktop app's local metrics data and turns that data into KPI cards, filterable tables, and exports.

Local dashboard

This is a local desktop intelligence module, not a cloud analytics dashboard.
KeyRing AI Metrics dashboard
Metrics turns local usage records into KPI cards, filterable tables, and exportable analytics views.

Layout

  • Top action area
  • Filter row
  • KPI cards
  • Five main data tables

Top actions currently include Reset History and Export Data.

KPI Cards

The KPI row reflects the current filter state, so the numbers update based on timeframe, provider, model, cost state, event type, and search query.

Requests
Total Tokens
Total Cost
Input Cost
Output Cost
Latency
Success Rate
Prompt Length
Session Duration
Cost / 1k Tokens

Filter Row

The current filter controls are Timeframe, Provider, Model, Cost State, Event Type, Search, and Refresh.

  • Timeframe options: Last 24 Hours, Last 7 Days, Last 30 Days, All Time
  • Search is slightly debounced before the workspace reloads
  • Refresh forces an immediate reload
1

Choose a timeframe first

Start broad, then narrow.

2

Filter by provider or model

Use provider/model filters when investigating one stack.

3

Use cost state or event type if needed

These help separate reconciled, estimated, unpriced, or event-specific records.

4

Use Search for final narrowing

The text search is best used after timeframe and provider/model filtering.

The Five Main Tables

Request / Response Log: row-level request details such as timestamp, request ID, provider, model, tokens, latency, cost status, and error code.

Sessions: grouped activity by session, including artifact chips such as JSON, Transcript, or Chatroom.

Events: event stream view for correlating event type, source, related session, and related request.

Model Stats: model-level usage, cost, latency, and success characteristics.

Provider Cost Efficiency: higher-level provider cost and latency comparison.

Artifact chips

Session chips currently copy the local artifact path to the clipboard. They do not directly open the file.

Exporting Data

The Export Data menu follows the current filter state.

  • Filtered CSV for spreadsheet-friendly export
  • Filtered XLSX for richer spreadsheet workflows
  • Raw JSON for the most direct programmatic representation

Reset History

Reset History is the destructive action in this module. Use it only when you intentionally want to clear the stored metrics history from your local analytics record.

Recommended Workflow

  1. Start with All Time if you are trying to understand overall usage.
  2. Narrow to a timeframe if the data is too broad.
  3. Filter by provider or model when investigating one stack.
  4. Use Request / Response Log for exact row-level review.
  5. Use Sessions for grouped run history.
  6. Export the filtered view if you need an outside record.

Troubleshooting

Common issues

  • Numbers look lower than expected: Check the active filters first.
  • Clicked an artifact chip and nothing opened: Current behavior is path copy, not automatic open.
  • Need the rawest export: Use Raw JSON.