Contribution offer: Terraform (IaC) deploy example - Dynamo on Nebius managed K8s

Hi team, following a conversation with an NVIDIA contact who suggested opening an issue here, I'd like to offer a cloud + Terraform deployment example for Dynamo.

Repo: https://github.com/CloudThrill/nvidia-dynamo-stack-terraform (Nebius MK8s today, structured for per-cloud subfolders)

How it differs from the existing k8s examples. The current examples start from "you have a cluster with the operator installed" and walk through manifests. This one starts from an empty cloud project: terraform apply provisions the VPC, managed Kubernetes, GPU node group, ingress + cert-manager, kube-prometheus-stack, the dynamo-platform chart and the DGD, one command rather than a copy-paste sequence. As far as I can tell there's no IaC example in the repo today.

What's in it: all verified on a live 1.3.0 cluster:

• Multi-model behind one shared Frontend on a single GPU, including the per-worker --endpoint dyn://$DYN_NAMESPACE-$DYN_NAMESPACE_WORKER_SUFFIX.<name>.generate needed so the second model isn't rejected on the default backend component
• Operator-managed ingress + Let's Encrypt TLS (the operator references a TLS secret it doesn't create, the example ships the cert-manager Certificate that fills it)
• Prometheus/Grafana with the shipped Dynamo + operator dashboards, including the monitor selector relaxation needed for them to be scraped
• Optional KV-aware routing example (router-mode=kv + worker KV events): verified: 20/20 shared-prefix requests routed to the cache-warm replica
• Measured deployment timings and a troubleshooting section built from real failures
• Real HTTPS URL endpoints no IP no port forwarding during the demos

Pinned to Dynamo 1.3.0 / vllm-runtime:1.3.0.

Placement : your call entirely; you know the repo layout better than I do. My instinct would be a self-contained terraform/nebius/ entry alongside the existing k8s examples: same shape as what's already there, just with the cluster provisioning included, so it stays isolated from the rest of the tree.

If carrying IaC in-tree isn't something the project wants, a short docs page linking out works too, though it'd be easier to find sitting with the other examples.

Start by reviewing the linked CloudThrill/nvidia-dynamo-stack-terraform repository and the existing Kubernetes examples in this repository. Confirm whether the contribution should be an in-tree terraform/nebius/ example or a documentation link, then verify that the deployment, monitoring, TLS, and troubleshooting flow is reproducible against the pinned Dynamo 1.3.0 setup.