Prometheus
Scrape cost, cache and latency metrics from /metrics or a Push Gateway.
DeepintShield exposes Prometheus metrics for cost, caching and latency, labelled so you can break them down by team.
What works
- A
/metricsendpoint (the telemetry plugin is on by default) or push to a Push Gateway (port 9091) - Metrics such as
deepintshield_cost_total,deepintshield_cache_hits_total, and time-to-first-token and inter-token histograms - Labels for Virtual Key, team, customer, routing rule, selected key and fallback index
- Custom labels from
x-deepintshield-prom-*request headers
How to connect
Add the gateway (for example deepintshield-host:8080) as a scrape target, or configure the Push Gateway address for clusters.
Read the Prometheus docs. See Observability & Audit for the full picture.