You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
API monitoring: every API, full analysis, dashboards and service health (M16) #63
API monitoring for the workspace's own APIs: any project, written in any framework, sends its request records, health and uptime to Yuva through a middleware or OpenTelemetry, and the workspace gets request volume, errors and latency per route, consumers, alerts and uptime, custom dashboards like Grafana, the health of every service and its dependencies, and public status pages on the workspace's own domain. Separate from visitor analytics (#42, M15), which counts visits and cannot compute percentiles from event props.
Web SDK server helpers this builds on: #60, #61, #62.
Decided in docs/architecture.md › API monitoring (M16): an API is an analytics property of kind api;
no row per request is stored (minute, hour and day rollups with mergeable sketches); sampling is off by
default and, when an SDK samples, shows as "estimated".
changed the title [-]API monitoring: request volume, errors and latency per route[/-][+]API monitoring: every API, full analysis, dashboards and service health (M16)[/+]on Oct 14, 2026
Activity
[-]API monitoring: request volume, errors and latency per route[/-][+]API monitoring: every API, full analysis, dashboards and service health (M16)[/+]docs: plan API monitoring (M16): request rollups, every API stack, da…
productdevbook commentedon Oct 14, 2026
Closed for now: the API monitoring work moved off main into #72 and is not on the roadmap until that PR is reviewed.