June 2026 - V1.20.0

✨✨✨ This release introduces GKE support, adds AKS support, improves Workload Autoscaler coverage and safety, and strengthens provider-specific installation, migration, and autoscaling reliability. It broadens CloudPilot AI’s multi-cloud coverage while making day-to-day optimization workflows safer and easier to operate.

GCP and Azure support

🚀 Highlights

GKE Support

CloudPilot AI now supports Google Kubernetes Engine (GKE) as a first-class provider. Customers can install CloudPilot AI on GKE, register GCP cluster metadata, and run optimization workflows for GKE node pools with provider-specific node configuration.

This release also includes the GKE follow-up work needed for production use: phase2 install assets, source pool discovery, replacement-node provisioning, restore and native autoscaling freeze handling, GCP project and cluster metadata fallback logic, and clearer warnings when provider resources are not ready.

AKS Support

CloudPilot AI now supports Azure Kubernetes Service (AKS). Operators can install CloudPilot AI on AKS, register Azure cluster metadata, and run optimization workflows against Azure node pools with Azure-specific node configuration.

Azure support is integrated across the in-cluster agent, controller, webhook, spot handler, pricing path, and onboard manifests. This extends the same CloudPilot AI workflow across AWS/EKS, GCP/GKE, Azure/AKS, and Universal environments.

Workload Autoscaler Improvements

Workload Autoscaler now supports DaemonSet workloads and improves recommendation safety during initial onboarding. New workloads can wait for a trustworthy initial metrics window before recommendations are marked ready, reducing premature optimization actions when telemetry is still warming up.

⚙️ Enhancements

Add GKE provider support, including GKE installation assets, GCP cluster registration, GKE node pool optimization, and GCP-specific interruption prediction.
Add Azure/AKS provider support, including AKS installation assets, Azure node pool configuration, webhook mutation support, interruption prediction, and spot interruption detection.
Improve GKE install, restore, and migration reliability, including safer freeze snapshot handling, source pool alignment, replacement-node provisioning, and reusable progressive migration rounds.
Improve Workload Autoscaler with DaemonSet workload support, an initial optimization data-window gate, and safer Java recommendation readiness checks.
Improve Workload Autoscaler pod attribution so workloads with shared selectors are matched by controller ownership instead of selector overlap alone.
Add GCP provider readiness warnings and align the GCP autoscaler contract across the agent, onboard manifests, and console.
Add a Saving Report baseline picker, giving users a clearer way to compare saving report results against a selected baseline.

🛠️ Bug Fixes

Fix Workload Autoscaler pod and metrics attribution issues for workloads that share selectors.
Fix Java Workload Autoscaler initial gate coverage after metrics query changes.
Fix stale node warning noise when Pods reference nodes that have already disappeared.
Fix proxy tunnel reconnect behavior so agent communication recovers more reliably after dead connections.
Fix node drain safety for single-replica workloads by triggering rollout behavior when needed.

These updates broaden CloudPilot AI’s provider coverage while improving autoscaling safety, GKE and AKS operational readiness, and day-to-day troubleshooting. For questions or support, join our Slack community

Stay tuned for more updates! 🚀