Perform an action, predict the signal, then check the dashboard and explain what changed (DGX Spark to Azure Private Endpoint lab).

How the Cloudutsuk Platform fits on one DGX Spark: k3d clusters, Backstage, central Argo CD with Karmada, monitoring, OpenCost, LitmusChaos, KEDA, a local LLM with RAG, and the ADRs behind each choice.

A tour of the lab's Grafana dashboard: which panel answers which question about the private path, the queries behind them, and how long a change takes to show.

Bring up the Azure side, the local probe and the DGX's WireGuard config, check each piece, and fix the failures I hit along the way.

Seven phases for the Cloudutsuk Platform, each ending usable: foundation, GitOps and CI/CD, monitoring, portal and labs, cost, local LLM and RAG, alert agent.

Every az CLI command that builds the VNet, the private Foundry account, its private endpoint and DNS zone, and the WireGuard VM, plus the VM's cloud-init and the DGX's wg0.conf.

What the Cloudutsuk Platform must do (inventory, cost, labs, monitoring, chat, alert agent) and how well it must do it on a single DGX Spark.

The lab repo file by file, and the probe that checks DNS, TCP, TLS and HTTP for each Foundry name and reports it all over OpenTelemetry.

Why I'm building a self-hosted control plane on a DGX Spark for every Cloudutsuk lab, cloud resource and its cost, and how I'll know it worked.

Reach a private-only Microsoft Foundry account from a DGX Spark at home for about $14 a month: a WireGuard VM in the VNet, split DNS on the DGX, and a probe that shows every DNS answer in Grafana.