feat: add nvidia-platform components for NVIDIA Dynamo LLM inference (#102)

* add k8s platform setting for debuging

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* gpu operator installation on EKS

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* Install dynamo platform

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* Install dynamo platform

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix bug for installing dynamo-platform

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* vllm aggregated serving mode

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix blocking issue after deployment

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix to local pvc -> nfs

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* remove stale rock file automatically

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* dynamo kv router and kv cache offloading

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix deployment name bug, add EP setting

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* add readme

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* udate dynamo version and fix monitoring error

* add dra mode

* fix bug

* update dynamo version and fix bugs

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* add monitoring filter option: model name

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix v0.9.0 dynamo crd bug: missing component

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* add aiperf benchmark and benchmark monitoring dashboard

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix dashboard bug

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* change discovery backend to etcd

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* aiconfigurator setup

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix benchmark bug

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix benchmark dashboard bug

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* fix dashboard bug

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* minor change

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* minor change of benchmark

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* update readme and dashboard lang

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

* delete ingress for vllm

Signed-off-by: jihyeonRyu <jryu@nvidia.com>

---------

Signed-off-by: jihyeonRyu <jryu@nvidia.com>
Co-authored-by: jihyeonRyu <jryu@nvidia.com>

Author: jihyeonRyu

.env

@@ -1,3 +1,7 @@
+# Platform type: eks | k8s
+PLATFORM=k8s
+
+# EKS Settings (required when PLATFORM=eks)
 REGION=us-west-2
 EKS_CLUSTER_NAME=genai-on-eks
 EKS_MODE=auto

cli

@@ -35,7 +35,10 @@ const configPath = path.join(DIR, "config.json");
 let config = JSON.parse(fs.readFileSync(configPath, "utf8"));
 mergeConfig("config.workshop.json");
 mergeConfig("config.local.json");
-process.env.KUBECONFIG = path.join(DIR, ".kubeconfig");
+// Only set custom kubeconfig for EKS mode; K8s mode uses existing kubeconfig
+if (process.env.PLATFORM !== "k8s") {
+  process.env.KUBECONFIG = path.join(DIR, ".kubeconfig");
+}
 const TERRAFORM_DIR = path.join(DIR, "terraform");
 const COMPONENTS_DIR = path.join(DIR, "components");
 const EXAMPLES_DIR = path.join(DIR, "examples");

cli-menu.json

@@ -1,5 +1,17 @@
 {
   "componentCategories": [
+    {
+      "dir": "nvidia-platform",
+      "name": "NVIDIA Platform",
+      "components": [
+        { "dir": "gpu-operator", "name": "GPU Operator" },
+        { "dir": "monitoring", "name": "Monitoring (Prometheus + Grafana)" },
+        { "dir": "dynamo-platform", "name": "Dynamo Platform" },
+        { "dir": "dynamo-vllm", "name": "Dynamo vLLM Serving" },
+        { "dir": "benchmark", "name": "AIPerf Benchmark" },
+        { "dir": "aiconfigurator", "name": "AIConfigurator (Auto Config + SLA Deploy)" }
+      ]
+    },
     {
       "dir": "ai-gateway",
       "name": "AI Gateway",