KubeSqueeze Agent

KubeSqueeze is a runnable Kubernetes cost-optimization demo built around deterministic discovery, planning, human approval, execution, verification, and rollback.

Dashboard tour

Historical evidence. Kubernetes state is joined with seven days of Prometheus utilization.	Deterministic plans. Every recommendation carries evidence, policy context, savings, and an approval gate.
Policy guardrails. An optional model can draft typed policy; deterministic validation and human activation remain mandatory.	Verify and restore. Scoped execution, health verification, automatic rollback, and manual restore are durable state transitions.
Immutable audit trail. Human decisions and agent actions are attributable, timestamped, and streamed to the dashboard.

Run the complete stack

With Docker, Kind, kubectl, and curl installed:

make kind-up

Open http://127.0.0.1:8080. The command builds the application image and deploys the React dashboard, Go server, read-only collector, scoped executor, Postgres, Prometheus, seven days of seeded utilization history, and customer- like workload fixtures into a four-node Kind cluster.

Use make kind-down to remove the cluster.

Documentation

• Development, debugging, fixtures, and commands