Skip to main content

Dashboard

The Admin Dashboard is the central monitoring hub for platform operators. It provides real-time visibility into API traffic, GPU resource consumption, and provider key health — all in a single unified view. Access requires the Dashboard permission.

Admin Dashboard — API Usage tab showing request count, token consumption, and active users with date range filter

Overview​

The Dashboard is organized into three primary tabs: API Usage, GPU Usage, and Provider Keys. Each tab surfaces the metrics most relevant to its domain, with configurable date ranges and comparison periods. A toolbar at the top allows you to set the observation window, compare against a previous period, and toggle test-account exclusion.

API usage tab​

The API Usage tab is the default view and presents the following key metrics:

Summary cards​

MetricDescription
Total RequestsAggregate API call count for the selected window, with period-over-period change percentage.
Total TokensCombined input and output token consumption (displayed in millions). Shows breakdown of input vs. output tokens.
Active UsersDaily Active Users (DAU) averaged over the selected window.
Success RatePercentage of non-error responses. Displays error count and timeout count separately.

Each card shows a directional indicator (↑ green for improvement, ↓ red for regression) comparing against the previous period.

Requests trend chart​

Below the summary cards, a Requests Trend line chart plots request volume over time. You can switch the bucket granularity (Day, Hour) using the dropdown in the top-right corner of the chart. The chart overlays current and previous period data for visual comparison.

Toolbar options​

  • Date Range Picker — Select start and end dates for the observation window.
  • Compare to — Choose a comparison baseline (e.g., Previous period, Same period last month).
  • Exclude test accounts — Toggle to filter out internal test traffic from all metrics.
  • Manage → — Link to manage test-account definitions.
  • Refresh — Manually reload metrics data.

GPU usage tab​

The GPU Usage tab displays resource utilization for self-deployed GPU clusters. This tab is available after you activate the Model Serving Platform.

Key metrics include:

  • GPU hours consumed per cluster and model.
  • Utilization percentage across provisioned capacity.
  • Cost attribution by deployment.

Use this tab to identify under-utilized clusters or models approaching capacity limits.

Provider keys tab​

The Provider Keys tab shows the status and usage of configured upstream provider API keys:

  • Key identifier and associated provider name.
  • Request count and token volume routed through each key.
  • Error rate and rate-limit proximity warnings.
  • Last-used timestamp for staleness detection.

Tips​

  • Dashboard data refreshes automatically every 60 seconds. Use the manual refresh button for immediate updates.
  • The "Exclude test accounts" toggle persists across sessions — remember to disable it when debugging test-user issues.
  • Export functionality is available on individual metric cards for reporting purposes.
  • All timestamps are displayed in the tenant's configured timezone.

Next Step​