🦥 Sloth V0.16.0: K8s Transformer Plugins & Web UI

[slok]

Hey everyone! 👋

This release brings two major features: a new plugin type that gives you full control over the Kubernetes objects Sloth produces, and a purpose-built web UI to visualize your SLOs in ways that weren't possible before.

 🎯 What's New

✨ Web UI

Warning

Experimental feature.

A new server command gives you a purpose-built web dashboard for visualizing your SLOs, solving problems that were impractical with Grafana alone.

Multi-dimensional SLO discovery: If you use SLO label grouping to create multiple SLO dimensions from a single spec (e.g. one SLO per endpoint, region, or customer), Grafana has no way to discover those dynamically generated SLO instances. The new UI understands Sloth's grouping model natively and can expand, list, and navigate all dimensions automatically.

Proper budget burn aggregation: The UI supports selecting static time ranges (monthly, weekly, quarterly or yearly) for burn rate and remaining budget visualization. In Grafana this required a hacky graph hardcoded to 30 days with expensive queries that didn't adapt to your actual SLO period. The new UI resolves this correctly with purpose-built charts.

On top of that:

• Service listing with search and sorting by name or alert status.
• SLO listing with search, filtering (alert firing, burning over budget, budget consumed) and sorting.
• SLO details with alert state, SLI error ratio chart and budget burn chart.

More info in the docs. Start by trying it locally with no dependencies:

sloth server --fake-prometheus

Note: Grafana will continue to be a valid UI for Sloth, this is just an alternative that can make sense on specific contexts or for SLO debugging/discovery

🧩 K8s Transformer Plugins

Warning

Experimental feature.

Sloth is no longer locked to producing Prometheus Operator PrometheusRule CRs. A new K8s transformer plugin system lets you control exactly what Kubernetes objects are generated from your SLOs. You can now:

• Output any Kubernetes resource type (ConfigMaps, custom CRDs like Google managed prometheus, victoria metrics specific CRs, etc.).
• Generate multiple objects per SLO group instead of a single PrometheusRule.
• Write your own transformer plugins using Sloth's standard plugin system.

The existing Prometheus Operator behavior is preserved as the default transformer plugin (sloth.dev/k8stransform/prom-operator-prometheus-rule/v1), so this is a non-breaking change for existing users.

The Go library also supports this via the new WriteResultAsK8sObjects method.

Check out the docs to get started.

🔍 SLO Duplicate Validation

The validate command now detects duplicate SLO IDs across your specs, catching copy-paste mistakes before they hit production. Disable with --ignore-slo-duplicates if you have intentional duplicates. (Thanks to @r3code for the contribution).

⚠️ Breaking Changes

Plugin loader ignores .. directories: The plugin filesystem loader now skips directories starting with ... This prevents accidental loading of plugins from parent paths. If you relied on this, adjust your plugin directory structure.

🔧 Under the Hood

• Updated to Kubernetes v1.35.
• Internally, the Prometheus Operator output path now uses the same dynamic unstructured plugin system as custom transformers, dogfooding the new plugin architecture.
• Contrib plugin changes.
• Dependency updates.