Skip to content

2.12. Statistics Dashboard

The Statistics page provides comprehensive analytics and visualizations of your AI usage, helping you understand usage patterns, performance metrics, and resource consumption.

Overview

The Statistics dashboard collects and displays data about your interactions with AI models, including:

  • Total API calls made
  • Token consumption (input and output)
  • Performance metrics (tokens per second, latency)
  • Model-specific usage breakdowns
  • Historical trends

Accessing Statistics

Click the Statistics icon in the sidebar (resembles a bar chart) to open the dashboard.

Dashboard Sections

Time Period Selector

At the top-right of the page, select the time period for your statistics:

  • Today: Current day only
  • 7 Days: Last week
  • 30 Days: Last month
  • All Time: Complete history

Overview Cards

The dashboard uses a visual hierarchy to emphasize the most important metrics:

Primary Metrics (larger cards with sparkline trends):

Metric Description
API Calls Total number of inference requests made, with 7-day trend sparkline
Total Tokens Combined input and output tokens, with 7-day trend sparkline

Secondary Metrics (standard cards):

Metric Description
Prompt Tokens Total input tokens sent to models
Completion Tokens Total output tokens generated by models
Avg Tokens/Sec Average generation speed across all requests
Conversations Number of unique chat conversations

Tertiary Metrics (compact cards):

  • Avg Latency: Average time to first token
  • Success Rate: Percentage of successful requests

Usage Over Time Chart

An interactive area chart showing daily usage trends with dual Y-axes for requests (left) and tokens (right).

Interactive Features:

  • Hover tooltips: View detailed values for each day
  • Legend toggle: Click on legend items to show/hide individual series
  • Brush zoom: For datasets with more than 10 days, use the brush at the bottom to zoom into specific date ranges

This visualization helps you:

  • Identify usage patterns
  • Track growth or decline in usage
  • Spot unusual spikes in activity

Token Distribution by Model

A pie or donut chart showing how token usage is distributed across different models. This helps you understand:

  • Which models you use most frequently
  • Token consumption patterns per model
  • Balance between different model types

Performance Trend

A multi-line chart tracking performance metrics over time:

  • Tokens per Second: Generation speed trend
  • Latency: Response time trend

Use this to monitor if your system's performance is consistent or degrading.

Model Usage Table

A detailed table breaking down usage by model:

Column Description
Model The model name/ID
Requests Number of API calls to this model
Prompt Tokens Total input tokens for this model
Completion Tokens Total output tokens from this model
Avg Tok/Sec Average generation speed

The table is sortable by clicking column headers.

Data Collection

Statistics are collected automatically when you:

  • Use the Chat interface
  • Make API calls through the Continuum Router
  • Use Cowork

All data is stored locally on your machine. No usage data is sent to external servers.

Storage Location

Statistics data is stored in a local SQLite database:

Platform Path
macOS ~/Library/Application Support/backend.ai-go/stats.db
Windows %APPDATA%\backend.ai-go\stats.db
Linux ~/.local/share/backend.ai-go/stats.db

Exporting Data

The Export button in the top-right corner allows you to export your statistics data:

  • Export as CSV: Downloads daily statistics in CSV format, suitable for spreadsheet applications
  • Export as JSON: Downloads complete statistics including overview, daily, and model data in JSON format
  • Copy to clipboard: Copies the JSON data to your clipboard for quick pasting

Security Note: CSV exports are sanitized to prevent formula injection attacks in spreadsheet applications.

Accessibility Features

Backend.AI GO charts are designed to be accessible to all users, including those with visual impairments or color vision deficiencies.

View Toggle (Chart/Table)

Each chart component includes a View Toggle button in the top-right corner that allows switching between:

  • Chart View: Visual representation with graphs and plots
  • Table View: Tabular data format that is:
    • Fully keyboard navigable
    • Screen reader compatible
    • Easier to read for users who prefer text-based data

Color-Blind Accessible Patterns

Charts use SVG patterns in addition to colors to differentiate data series:

  • Solid fills for the first series
  • Diagonal stripes, dots, horizontal lines, crosshatch, and vertical lines for subsequent series
  • Ensures data is distinguishable even without color perception

Screen Reader Support

All charts include comprehensive ARIA attributes:

  • role="img" with descriptive aria-label for chart containers
  • Meaningful labels describing the data being visualized
  • Table views provide proper semantic structure with <caption>, <th>, and scope attributes

High Contrast Mode

When your system's high contrast mode is enabled (prefers-contrast: high):

  • Chart borders become more prominent
  • Text contrast is increased
  • Focus indicators are more visible

Reduced Motion

For users who prefer reduced motion (prefers-reduced-motion: reduce):

  • Chart animations are disabled
  • Transitions are minimized
  • Data updates appear instantly without animation effects

Tips

  • Check statistics regularly to understand your usage patterns
  • Use the model breakdown to optimize which models you use
  • Monitor performance trends to detect potential issues
  • Compare different time periods to track improvements
  • Export data periodically for backup or external analysis
  • Use the Table View for screen readers or when you need precise values