Metrics

Overview of metrics and monitoring in IntelliAsk — the app's built-in Prometheus endpoint and the standalone Metrics Exporter. Points to the full exporter configuration and metrics catalog.

IntelliAsk exposes usage and operational data to Prometheus (or any OpenMetrics-compatible tool) so you can build Grafana dashboards for active users, token usage, models, ratings, and more.

Active users in IntelliAsk

There are two sources of metrics, covering different things:

  • App-side endpoint — the IntelliAsk API server's own /metrics endpoint, covering runtime/HTTP internals (documented below).
  • Metrics Exporter — a separate service (exporter) that reads the MongoDB database and exposes rich intelliask_* usage metrics (users, tokens, models, ratings, tools, files).

Metrics Exporter (usage metrics)

The Metrics Exporter is the recommended way to get usage analytics. It is a bundled, optional stack component — enable it and it scrapes MongoDB and serves intelliask_* metrics for Prometheus to scrape and Grafana to visualize.

Full reference

This page is just an overview. For the exporter's complete configuration (MongoDB connection, caching, per-group metric toggles, logging) and the full catalog of exposed intelliask_* metrics, see Services → Metrics Exporter.

Metrics Exporter (service reference)

What it is, how it works, and how it is wired into the stack.

Configuration

Every variable — MongoDB, caching, timezone, the per-group metric toggles, logging, and the optional health check.

Metrics catalog

The full list of exposed intelliask_* metrics — name, type, labels, and what each measures.

Deploy → Components → Exporter

Image tag, resource limits, env keys, and exposing it through the reverse proxy.

In the stack the exporter runs on container port 8000, published on host port 9100. Point your Prometheus scrape config at it and look for intelliask_* metrics (e.g. intelliask_registered_users).

App-side metrics endpoint

Separately, the IntelliAsk API server exposes its own built-in Prometheus metrics at /metrics when METRICS_SECRET is configured. Scrape it with an Authorization: Bearer <METRICS_SECRET> header.

Browser RUM proxy outcomes are reported as rum_proxy_requests_total{endpoint,result}, where endpoint is traces, logs, or unknown, and result can be success, auth_drop, auth_error, bad_request, not_configured, collector_4xx, collector_5xx, collector_error, or collector_timeout.

Last updated on