avuru obs team

Until now the product opened on a screen that needed you to already have a hypothesis. The new Dashboard gives you one, and it is where the app now lands.

Deciding what Avuru Obs collects used to mean editing values.yaml, running helm upgrade, and holding the cluster permissions to do it — so in practice collection stayed whatever it was on install day. Now an admin changes it from the app, and the sensor follows within seconds.

GET /api/v1/green/summary, /green/budgets and /green/report were registered without the session middleware that every other signal route uses. On an install with authentication enabled, all three answered 200 to a caller with no session at all.

On a real cluster the Pods table is a hundred-plus rows in no particular order, and the only way through it was the browser's own find-in-page. Both tables on the Nodes screen now sort and filter.

Deciding that payments is your most critical service used to mean editing values.yaml, running helm upgrade, and holding the cluster permissions to do it. So in practice it never happened: the Service Health board showed one auto-discovered group per namespace and the tier lanes sat empty. Groups are now authored in the app.

Two questions an operator asks in the first week — where is my telemetry actually stored? and what can this role do? — now have a screen each.

v0.4.0 finishes the account lifecycle you started in v0.2: users can now be edited, reset and deleted from the UI, and anyone can change their own password. Reviewing that surface turned up three ways to get into an account that shouldn't have worked — all three are closed. Alongside them, an install that silently lost its schema migration now heals itself.

avuru obs v0.3.0 is tagged. v0.2 secured the read side — login, roles, per-project grants, SSO. v0.3 closes the write side and turns the project into something you actually administer.

v0.3.1 is a patch: it fixes the chart's image defaults, which never pointed at the images the release workflow actually publishes. Nothing else changes — no schema migration, no API or configuration surface.

The green module reported energy only where the hardware exposes RAPL — which excludes the overwhelming majority of public-cloud VMs. An opt-in TDP estimator now models CPU power from utilization on those nodes, so a fleet that runs entirely on cloud instances stops seeing an empty /green.