Skip to main content
Version: v2026.0.0

Monitoring statistics

The statistics section of the admin panel displays usage metrics for the current tenant. Data is read from OpenSearch via the Stats API (/api/v1/admin/stats).

The statistics page with overview cards, activity charts, LLM distribution, top prompts, and feature adoption

Figure: the statistics page — overview cards, daily activity, LLM distribution, top prompts by time saved, and feature adoption.

Available statistics​

Global stats​

The dashboard shows aggregated totals:

  • Total conversations
  • Total requests
  • Total tokens (input + output)
  • Estimated time saved

Fetched from GET /api/v1/admin/stats/global.

Time interval selector​

A set of tabs at the top of the statistics page lets you select the time granularity. The selected interval applies to all charts on the page.

TabIntervalDescription
DayDAYDaily buckets (default)
WeekWEEKWeekly buckets
MonthMONTHMonthly buckets
YearYEARYearly buckets
AllnoneAll-time aggregation, no interval filter

Activity charts​

Two area/line charts show activity over the selected interval:

  • Requests: request volume over time.
  • Tokens: token usage (input + output) over time.

Since 2026.0.0, the token counts of a request include every completion of its tool-calling loop (up to the configured number of tool cycles), not only the last one, and a plan node whose agent fails still records the tokens the provider billed. Totals are therefore higher than in earlier versions for conversations and plans that use tools; historical data is not recomputed, so figures before and after the upgrade are not comparable. The interval bounds are computed in the server's time zone, so the most recent requests are no longer dropped east of UTC.

Fetched from GET /api/v1/admin/stats/timeseries?interval={interval}.

LLM distribution​

A chart showing which LLM models were used for requests, with counts per provider. The interval selector applies to this chart as well.

Fetched from GET /api/v1/admin/stats/llm-distribution?interval={interval}.

Top prompts by time saved​

Ranks the top 5 prompts by cumulative time saved (displayed in hours). This helps identify which prompts deliver the most value.

Fetched from GET /api/v1/admin/stats/top-prompts-time-saved?interval={interval}.

Feature adoption​

Shows adoption rates for advanced capabilities:

FeatureDescription
MultimodalPercentage of requests using image inputs
Function callingPercentage of requests using tool calling

Fetched from GET /api/v1/admin/stats/feature-adoption.

Figure: Statistics page layout and data flow.

PDF export​

Click the "Export PDF" button at the top of the statistics page to download the current dashboard view as a PDF file. The export includes all visible charts and summary cards.

Metrics export to OpenSearch​

Uxopian AI exports custom metrics to OpenSearch via the Micrometer Elastic exporter, configured in metrics.yml. The default index is micrometer-metrics. Standard JVM, HTTP, and system metrics are disabled by default to reduce noise.

To enable additional metrics, edit metrics.yml:

management:
metrics:
enable:
jvm: true # Enable JVM metrics
http: true # Enable HTTP request metrics

Actuator endpoints​

Three actuator endpoints are exposed over HTTP:

EndpointPathDescription
Health/actuator/healthApplication health (public)
Info/actuator/infoBuild and version info
Loggers/actuator/loggersView and change log levels at runtime

The health endpoint is public in the default gateway configuration. Other actuator endpoints should be protected in production.

REST API endpoints​

MethodEndpointDescription
GET/api/v1/admin/stats/globalGlobal aggregated statistics
GET/api/v1/admin/stats/timeseries?interval={interval}Activity time series
GET/api/v1/admin/stats/llm-distribution?interval={interval}LLM model usage distribution
GET/api/v1/admin/stats/top-prompts-time-saved?interval={interval}Top 5 prompts by time saved
GET/api/v1/admin/stats/feature-adoptionFeature adoption rates