ragleap-observability

version — loading live from PyPI…

Prometheus, Grafana, Loki and AlertManager observability stack for RagLeap Core. The changelog states exactly what each release verified, and what it did not.

pip install ragleap-observability

Highlights

• Prometheus + postgres_exporter — live-verified, real non-zero metrics
• Grafana dashboards-as-code, datasources confirmed live
• Loki + Promtail log shipping — 21+ real log streams confirmed
• AlertManager with optional Slack and email receivers (off by default; the rendered config is checked with amtool)
• Opt-in availability SLO alerts since 0.8.0 (fast and slow burn on the Prometheus up metric, plus a RagleapTargetAbsent alert since 0.8.2); rule logic tested with promtool and, since 0.8.3, run on a real Prometheus in a single container without Kubernetes (a stopped target fired the fast-burn alert); slow-burn firing and delivery through AlertManager were not exercised. SLI dashboards are not built yet

Tech stack

Core dependencies: (no Python runtime dependencies — Helm charts, not an importable library)

Testing

1 test function across 1 test file, verified directly from the repo.

This package is infrastructure manifests, not a unit-tested library — its "tests" are live cluster verification, documented in the CHANGELOG, not a pytest suite. 1 automated test file exists; real confidence comes from the documented live-verification runs above.

Quickstart

helm install ragleap-observability ./helm/ragleap-observability \
  --namespace monitoring --create-namespace