feat(examples): add Kubernetes GPU autoscaling recipe

THIRD-PARTY-NOTICES

@@ -31,6 +31,14 @@ Image: vllm/vllm-openai
 License: Apache-2.0 and included component licenses
 URL: https://github.com/vllm-project/vllm
 
+Image: node:22-bookworm-slim
+License: MIT and included Debian component licenses
+URL: https://hub.docker.com/_/node
+
+Image: ollama/ollama
+License: MIT and included component licenses
+URL: https://github.com/ollama/ollama
+
 ================================================================================
 Docker Compose Images
 ================================================================================
@@ -239,6 +247,46 @@ Project: NVIDIA/OpenShell installer and runtime
 License: Apache-2.0
 URL: https://github.com/NVIDIA/OpenShell
 
+Project: NVIDIA/NemoClaw source and sandbox base image
+License: Apache-2.0 and included component licenses
+URL: https://github.com/NVIDIA/NemoClaw
+
+Project: Kubernetes SIG Agent Sandbox controller and API
+License: Apache-2.0
+URL: https://github.com/kubernetes-sigs/agent-sandbox
+
+Project: Kubernetes and kubectl
+License: Apache-2.0
+URL: https://github.com/kubernetes/kubernetes
+
+Project: Helm
+License: Apache-2.0
+URL: https://github.com/helm/helm
+
+Project: Canonical MicroK8s distribution
+License: See the project and included component notices
+URL: https://github.com/canonical/microk8s
+
+Project: NVIDIA GPU Operator
+License: Apache-2.0
+URL: https://github.com/NVIDIA/gpu-operator
+
+Project: NVIDIA DCGM Exporter
+License: Apache-2.0
+URL: https://github.com/NVIDIA/dcgm-exporter
+
+Project: Kubernetes Metrics Server
+License: Apache-2.0
+URL: https://github.com/kubernetes-sigs/metrics-server
+
+Project: Prometheus Community Helm Charts (kube-prometheus-stack and prometheus-adapter)
+License: Apache-2.0 and included component licenses
+URL: https://github.com/prometheus-community/helm-charts
+
+Project: Kubernetes ingress-nginx
+License: Apache-2.0
+URL: https://github.com/kubernetes/ingress-nginx
+
 Project: Astral uv installer and package manager
 License: Apache-2.0 OR MIT
 URL: https://github.com/astral-sh/uv
@@ -263,6 +311,10 @@ Model: NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4
 License: NVIDIA Open Model License
 URL: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4
 
+Model: Llama 3.2 3B
+License: Llama 3.2 Community License Agreement and Acceptable Use Policy
+URL: https://ollama.com/library/llama3.2
+
 Service: GitHub API
 Terms: See provider terms
 URL: https://docs.github.com/

examples/README.md

@@ -11,6 +11,7 @@ again by contributor provenance.
 | Example | Description |
 | --- | --- |
 | [Developer Community Chief of Staff](recipes/nvidia/developer-community-chief-of-staff/README.md) | Synthesizes Slack, Outlook, GitHub, and mirrored community signals into operating briefs, gaps, priorities, and follow-up recommendations. |
+| [Kubernetes GPU Autoscaling](recipes/nvidia/kubernetes-gpu-autoscaling/README.md) | Runs a CPU-only NemoClaw/OpenClaw sandbox through OpenShell on Kubernetes and autoscales authenticated, GPU-backed Ollama inference pods from DCGM utilization. |
 | [NV Tech Assistant](recipes/nvidia/nv-tech-assistant/README.md) | Answers NVIDIA technical questions with cited evidence from allowlisted NVIDIA, GitHub, and arXiv sources. |
 | [Payment Operations Hermes Assistant](recipes/nvidia/payment-ops-hermes/README.md) | Demonstrates constrained payment screening, evidence preparation, and a platform-enforced human release boundary. |
 | [PR Review Advisor](recipes/nvidia/pr-review-advisor/README.md) | Reviews exact pull request heads with a constrained Hermes workflow, produces attested artifacts, and publishes only through a separate maintainer action. |

examples/recipes/nvidia/kubernetes-gpu-autoscaling/.helmignore

@@ -0,0 +1,5 @@
+.DS_Store
+*.swp
+*.bak
+*.tmp
+.git/

examples/recipes/nvidia/kubernetes-gpu-autoscaling/Chart.yaml

@@ -0,0 +1,14 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+apiVersion: v2
+name: nemoclaw-gpu
+description: GPU-backed Ollama inference pods on Kubernetes with GPU utilization HPA
+type: application
+version: 0.1.0
+appVersion: "2026.05.27"
+keywords:
+  - nemoclaw
+  - ollama
+  - gpu
+maintainers:
+  - name: maggiezha