ragleap-observability v0.8.3 release notes

tests/live/slo-live-test.sh: a manual test (about 20-25 minutes, needs Docker, not run in CI) that runs the SLO rules on a real Prometheus with a stand-in target and removes what it creates.

The package description no longer says "live-tested against a real cluster"; it points to this changelog. README wording on what is verified is updated.

Ran tests/live/slo-live-test.sh on 2026-10-17 against Prometheus v3.13.4 (single container, no Kubernetes, 15 s scrape interval, rules from 0.8.2, which are unchanged): a healthy target fired nothing; a stopped target fired RagleapAvailabilityFastBurn (after 140 s) and PostgresExporterDown (191 s) and both cleared after recovery; a target removed from the scrape config fired RagleapTargetAbsent (342 s) while PostgresExporterDown stayed inactive, and it cleared when the target returned. 12 of 12 checks passed. This supersedes "not run on a live Prometheus" in the 0.8.0 and 0.8.2 entries below.
With defaults, helm template output is identical to 0.8.2.

Not exercised: RagleapAvailabilitySlowBurn firing (needs an outage of 15 minutes or more), delivery through AlertManager, and the Kubernetes wiring.
Timings come from a fresh Prometheus with little history, so the burn-rate windows averaged only the data that existed.

Verified on PyPI: the 0.8.3 wheel contains 17 files under helm/ and the RagleapTargetAbsent rule.

pip install ragleap-observability==0.8.3