andrewpaprotsky · Oct 1, 2026 · Sep 9, 2026 · Sep 9, 2026 · Sep 9, 2026 · Sep 9, 2026
@@ -425,6 +425,11 @@ areas:
   - deploy/utils/tests/
   - examples/deployments/
   - tests/deploy/
+- label: lpx
+  github_team: '@ai-dynamo/dynamo-lpx-codeowners'
+  path_globs:
+  - deploy/operator/internal/controller/lpx/
+  - deploy/operator/internal/dynamo/lpx/
 - label: gms
   github_team: '@ai-dynamo/dynamo-gms-codeowners'
   path_globs:
@@ -575,6 +580,7 @@ shared:
   - frontend
   - gms
   - kv-memory
+  - lpx
   - multimodal
   - observability
   - operator
@@ -980,6 +986,14 @@ shared:
   owners:
   - operator
   - observability
+- glob: deploy/operator/internal/controller/lpx/
+  owners:
+  - operator
+  - lpx
+- glob: deploy/operator/internal/dynamo/lpx/
+  owners:
+  - operator
+  - lpx
 - glob: deploy/operator/internal/controller/*checkpoint*
   owners:
   - operator

@@ -18,6 +18,7 @@
 #   frontend         @ai-dynamo/dynamo-frontend-codeowners
 #   gms              @ai-dynamo/dynamo-gms-codeowners
 #   kv-memory        @ai-dynamo/dynamo-kv-memory-codeowners
+#   lpx              @ai-dynamo/dynamo-lpx-codeowners
 #   multimodal       @ai-dynamo/dynamo-multimodal-codeowners
 #   observability    @ai-dynamo/dynamo-observability-codeowners
 #   operator         @ai-dynamo/dynamo-operator-codeowners
@@ -45,6 +46,7 @@
 #   @ai-dynamo/dynamo-frontend-codeowners
 #   @ai-dynamo/dynamo-gms-codeowners
 #   @ai-dynamo/dynamo-kv-memory-codeowners
+#   @ai-dynamo/dynamo-lpx-codeowners
 #   @ai-dynamo/dynamo-multimodal-codeowners
 #   @ai-dynamo/dynamo-observability-codeowners
 #   @ai-dynamo/dynamo-operator-codeowners
@@ -539,6 +541,7 @@ requirements.txt
 /lib/llm/src/block_manager/layout.rs                                                                    @ai-dynamo/dynamo-kv-memory-codeowners
 /tests/fault_tolerance/cancellation/                                                                    @ai-dynamo/dynamo-fault-tolerance-codeowners
 /deploy/operator/internal/checkpoint/                                                                   @ai-dynamo/dynamo-gms-codeowners
+/deploy/operator/internal/dynamo/lpx/                                                                   @ai-dynamo/dynamo-lpx-codeowners
 /dev/observability/prometheus-xpu.yml                                                                   @ai-dynamo/dynamo-xpu-codeowners
 /lib/llm/src/block_manager/connector/                                                                   @ai-dynamo/dynamo-kv-memory-codeowners
 /lib/llm/src/block_manager/offload.rs                                                                   @ai-dynamo/dynamo-kv-memory-codeowners
@@ -556,6 +559,7 @@ requirements.txt
 /lib/llm/src/block_manager/controller.rs                                                                @ai-dynamo/dynamo-kv-memory-codeowners
 /components/src/dynamo/common/multimodal/                                                               @ai-dynamo/dynamo-multimodal-codeowners
 /components/src/dynamo/trtllm/multimodal/                                                               @ai-dynamo/dynamo-multimodal-codeowners
+/deploy/operator/internal/controller/lpx/                                                               @ai-dynamo/dynamo-lpx-codeowners
 /lib/llm/src/block_manager/distributed.rs                                                               @ai-dynamo/dynamo-kv-memory-codeowners
 /lib/llm/src/block_manager/kv_consolidator/                                                             @ai-dynamo/dynamo-kv-memory-codeowners
 /tests/serve/multimodal_profiles/vllm_xpu.py                                                            @ai-dynamo/dynamo-xpu-codeowners
@@ -576,7 +580,7 @@ requirements.txt
 /.codex/                                                                                                @ai-dynamo/dynamo-process-codeowners @ai-dynamo/dynamo-agents-codeowners
 /agents/                                                                                                @ai-dynamo/dynamo-process-codeowners @ai-dynamo/dynamo-agents-codeowners
 /.agents/                                                                                               @ai-dynamo/dynamo-process-codeowners @ai-dynamo/dynamo-agents-codeowners
-/CODEOWNERS                                                                                             @ai-dynamo/dynamo-process-codeowners @ai-dynamo/dynamo-agents-codeowners @ai-dynamo/dynamo-backend-sglang-codeowners @ai-dynamo/dynamo-backend-trtllm-codeowners @ai-dynamo/dynamo-backend-vllm-codeowners @ai-dynamo/dynamo-diffusion-codeowners @ai-dynamo/dynamo-docs-codeowners @ai-dynamo/dynamo-epp-codeowners @ai-dynamo/dynamo-fault-tolerance-codeowners @ai-dynamo/dynamo-frontend-codeowners @ai-dynamo/dynamo-gms-codeowners @ai-dynamo/dynamo-kv-memory-codeowners @ai-dynamo/dynamo-multimodal-codeowners @ai-dynamo/dynamo-observability-codeowners @ai-dynamo/dynamo-operator-codeowners @ai-dynamo/dynamo-ops-codeowners @ai-dynamo/dynamo-performance-codeowners @ai-dynamo/dynamo-planner-codeowners @ai-dynamo/dynamo-rl-codeowners @ai-dynamo/dynamo-router-codeowners @ai-dynamo/dynamo-runtime-codeowners @ai-dynamo/dynamo-tokenspeed-codeowners @ai-dynamo/dynamo-xpu-codeowners
+/CODEOWNERS                                                                                             @ai-dynamo/dynamo-process-codeowners @ai-dynamo/dynamo-agents-codeowners @ai-dynamo/dynamo-backend-sglang-codeowners @ai-dynamo/dynamo-backend-trtllm-codeowners @ai-dynamo/dynamo-backend-vllm-codeowners @ai-dynamo/dynamo-diffusion-codeowners @ai-dynamo/dynamo-docs-codeowners @ai-dynamo/dynamo-epp-codeowners @ai-dynamo/dynamo-fault-tolerance-codeowners @ai-dynamo/dynamo-frontend-codeowners @ai-dynamo/dynamo-gms-codeowners @ai-dynamo/dynamo-kv-memory-codeowners @ai-dynamo/dynamo-lpx-codeowners @ai-dynamo/dynamo-multimodal-codeowners @ai-dynamo/dynamo-observability-codeowners @ai-dynamo/dynamo-operator-codeowners @ai-dynamo/dynamo-ops-codeowners @ai-dynamo/dynamo-performance-codeowners @ai-dynamo/dynamo-planner-codeowners @ai-dynamo/dynamo-rl-codeowners @ai-dynamo/dynamo-router-codeowners @ai-dynamo/dynamo-runtime-codeowners @ai-dynamo/dynamo-tokenspeed-codeowners @ai-dynamo/dynamo-xpu-codeowners
 /agent-docs/                                                                                            @ai-dynamo/dynamo-process-codeowners @ai-dynamo/dynamo-agents-codeowners
 /.lycheeignore                                                                                          @ai-dynamo/dynamo-ops-codeowners @ai-dynamo/dynamo-docs-codeowners
 /tests/deploy/                                                                                          @ai-dynamo/dynamo-ops-codeowners @ai-dynamo/dynamo-operator-codeowners
@@ -646,6 +650,7 @@ requirements.txt
 /lib/llm/src/http/service/metrics.rs                                                                    @ai-dynamo/dynamo-frontend-codeowners @ai-dynamo/dynamo-observability-codeowners
 /container/deps/requirements.vllm.txt                                                                   @ai-dynamo/dynamo-ops-codeowners @ai-dynamo/dynamo-backend-vllm-codeowners
 /deploy/operator/internal/checkpoint/                                                                   @ai-dynamo/dynamo-gms-codeowners @ai-dynamo/dynamo-operator-codeowners
+/deploy/operator/internal/dynamo/lpx/                                                                   @ai-dynamo/dynamo-operator-codeowners @ai-dynamo/dynamo-lpx-codeowners
 /lib/llm/src/preprocessor/mm_routing/                                                                   @ai-dynamo/dynamo-frontend-codeowners @ai-dynamo/dynamo-multimodal-codeowners @ai-dynamo/dynamo-router-codeowners
 /lib/llm/src/protocols/openai/audios/                                                                   @ai-dynamo/dynamo-frontend-codeowners @ai-dynamo/dynamo-multimodal-codeowners
 /lib/llm/src/protocols/openai/images/                                                                   @ai-dynamo/dynamo-frontend-codeowners @ai-dynamo/dynamo-diffusion-codeowners
@@ -680,6 +685,7 @@ requirements.txt
 /components/src/dynamo/trtllm/snapshot.py                                                               @ai-dynamo/dynamo-backend-trtllm-codeowners @ai-dynamo/dynamo-gms-codeowners
 /components/src/dynamo/vllm/cache_info.py                                                               @ai-dynamo/dynamo-backend-vllm-codeowners @ai-dynamo/dynamo-kv-memory-codeowners
 /container/deps/requirements.frontend.txt                                                               @ai-dynamo/dynamo-ops-codeowners @ai-dynamo/dynamo-frontend-codeowners
+/deploy/operator/internal/controller/lpx/                                                               @ai-dynamo/dynamo-operator-codeowners @ai-dynamo/dynamo-lpx-codeowners
 /lib/kv-router/src/services/shared_cache/                                                               @ai-dynamo/dynamo-router-codeowners @ai-dynamo/dynamo-kv-memory-codeowners
 /lib/kv-router/src/zmq_wire/extra_keys.rs                                                               @ai-dynamo/dynamo-router-codeowners @ai-dynamo/dynamo-multimodal-codeowners
 /lib/runtime/src/metrics/frontend_perf.rs                                                               @ai-dynamo/dynamo-observability-codeowners @ai-dynamo/dynamo-frontend-codeowners

@@ -84,7 +84,7 @@ verify-operator-chart-crds: ## Verify operator CRDs are managed from the image,
 
 .PHONY: test
 test: helm-unittest-install charts-prep verify-operator-chart-crds ## Run helm-unittest chart tests
-	@helm unittest .
+	@helm unittest . components/operator
 
 .PHONY: lint
 lint: charts-prep verify-operator-chart-crds ## Run helm lint

@@ -170,6 +170,8 @@ Kubernetes: `>=1.30.0-0`
 | dynamo-operator.natsAddr | string | `""` | NATS server address for operator communication. When empty, the operator uses bundled NATS only if global.nats.install=true; otherwise NATS is not configured. Format: `nats://hostname:4222` |
 | dynamo-operator.etcdAddr | string | `""` | etcd server address for an external etcd instance. Only needed when using external etcd without the bundled subchart. Format: `http://hostname:2379` or `https://hostname:2379` |
 | dynamo-operator.modelExpressURL | string | `""` | URL for the Model Express server if not deployed by this helm chart. This is ignored if Model Express server is installed by this helm chart (global.model-express.enabled is true). |
+| dynamo-operator.lpx.enabled | bool | `false` | Enable the Dynamo operator's LPX integration. Startup requires the externally installed scheduling.lpu.nvidia.com/v1alpha1 LpuPipelineRequest API. |
+| dynamo-operator.lpx.modelRegistryURL | string | `""` | Configure the LPU model registry used by the LPX integration. |
 | dynamo-operator.namespaceRestriction | object | `{"enabled":false,"lease":{"duration":"30s","renewInterval":"10s"},"targetNamespace":null}` | DEVELOPMENT AND TESTING ONLY: Namespace-restricted mode is not supported for production. Use cluster-wide mode for production deployments. |
 | dynamo-operator.namespaceRestriction.enabled | bool | `false` | DEVELOPMENT AND TESTING ONLY: Enable namespace-restricted reconciliation and admission. Not supported for production. |
 | dynamo-operator.namespaceRestriction.targetNamespace | string | `nil` | DEVELOPMENT AND TESTING ONLY: Target namespace. Defaults to the Helm release namespace. |

@@ -21,3 +21,4 @@
 .idea/
 *.tmproj
 .vscode/
+tests/
@@ -254,6 +254,7 @@ rules:
   - dynamographdeployments
   - dynamographdeploymentscalingadapters
   - dynamomodels
+  - lpxgraphdeployments
   verbs:
   - create
   - delete
@@ -269,6 +270,7 @@ rules:
   - dynamographdeploymentrequests/finalizers
   - dynamographdeployments/finalizers
   - dynamomodels/finalizers
+  - lpxgraphdeployments/finalizers
   verbs:
   - update
 - apiGroups:
@@ -279,6 +281,7 @@ rules:
   - dynamographdeployments/status
   - dynamographdeploymentscalingadapters/status
   - dynamomodels/status
+  - lpxgraphdeployments/status
   verbs:
   - get
   - patch
@@ -354,6 +357,16 @@ rules:
   - patch
   - update
   - watch
+- apiGroups:
+  - scheduling.lpu.nvidia.com
+  resources:
+  - lpupipelinerequests
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - watch
 - apiGroups:
   - scheduling.run.ai
   resources:

@@ -136,6 +136,9 @@ spec:
         - mountPath: /tmp/k8s-webhook-server/serving-certs
           name: cert
           readOnly: true
+        {{- with .Values.controllerManager.manager.volumeMounts }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
       securityContext:
         runAsNonRoot: true
       serviceAccountName: {{ include "dynamo-operator.fullname" . }}-controller-manager
@@ -151,3 +154,6 @@ spec:
           # Allow the pod to start before the secret exists so the
           # in-operator cert-controller can bootstrap it at runtime.
           optional: true
+      {{- with .Values.controllerManager.manager.volumes }}
+      {{- toYaml . | nindent 6 }}
+      {{- end }}
@@ -213,3 +213,8 @@ data:
     gpu:
       discoveryEnabled: {{ .Values.gpuDiscovery.enabled }}
     {{- end }}
+    lpx:
+      enabled: {{ .Values.lpx.enabled }}
+      {{- with .Values.lpx.modelRegistryURL }}
+      modelRegistryURL: {{ . | quote }}
+      {{- end }}
@@ -19,6 +19,25 @@
   Each check detects an old value path and fails with a migration message.
 */}}
 
+{{/*
+  LPX requires Grove integration. Only a parent chart can install Grove;
+  a standalone operator chart must explicitly enable an external Grove.
+*/}}
+
+{{- if and .Values.lpx.enabled (not (and
+  .Values.global
+  .Values.global.grove
+  (or
+    .Values.global.grove.enabled
+    (and (not .Chart.IsRoot) .Values.global.grove.install)
+  )
+)) }}
+  {{- fail `
+CONFIGURATION ERROR: LPX requires Grove integration to be enabled.
+Set global.grove.enabled=true to use an externally installed Grove operator.
+` }}
+{{- end }}
+
 {{/* ---- NATS toggle (moved to global.nats.install) ---- */}}
 
 {{- if and (hasKey .Values "nats") (hasKey .Values.nats "enabled") }}

@@ -0,0 +1,55 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+suite: standalone LPX and Grove configuration
+templates:
+  - templates/operator-config.yaml
+  - templates/validate-values.yaml
+tests:
+  - it: leaves LPX disabled when Grove is not configured
+    template: templates/operator-config.yaml
+    asserts:
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^lpx:\n  enabled: false$'
+
+  - it: rejects LPX when global values are absent
+    template: templates/validate-values.yaml
+    set:
+      lpx.enabled: true
+    asserts:
+      - failedTemplate:
+          errorPattern: "LPX requires Grove integration to be enabled"
+
+  - it: rejects LPX when Grove integration is explicitly disabled
+    template: templates/validate-values.yaml
+    set:
+      lpx.enabled: true
+      global.grove.enabled: false
+      global.grove.install: false
+    asserts:
+      - failedTemplate:
+          errorPattern: "LPX requires Grove integration to be enabled"
+
+  - it: enables LPX with externally installed Grove
+    template: templates/operator-config.yaml
+    set:
+      lpx.enabled: true
+      global.grove.enabled: true
+    asserts:
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^lpx:\n  enabled: true$'
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^  grove:\n    enabled: true$'
+
+  - it: rejects install-only Grove settings without a parent chart
+    template: templates/validate-values.yaml
+    set:
+      lpx.enabled: true
+      global.grove.enabled: false
+      global.grove.install: true
+    asserts:
+      - failedTemplate:
+          errorPattern: "LPX requires Grove integration to be enabled"
@@ -22,6 +22,14 @@
 # the cluster-wide operator.
 upgradeCRD: true
 
+# The LPX scheduler and its CRDs must be installed separately.
+lpx:
+  # -- Enable the Dynamo operator's LPX integration. Startup requires the
+  # externally installed scheduling.lpu.nvidia.com/v1alpha1 LpuPipelineRequest API.
+  enabled: false
+  # -- Configure the LPU model registry used by the LPX integration.
+  modelRegistryURL: ""
+
 # Environment variables to pass to operator Deployment.
 env: []
 
@@ -90,6 +98,8 @@ controllerManager:
       requests:
         cpu: 512m
         memory: 1Gi
+    volumeMounts: []
+    volumes: []
   replicas: 1
   serviceAccount:
     annotations: {}

@@ -0,0 +1,68 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+suite: LPX operator configuration and RBAC
+templates:
+  - charts/dynamo-operator/templates/operator-config.yaml
+  - charts/dynamo-operator/templates/manager-rbac.yaml
+  - charts/dynamo-operator/templates/validate-values.yaml
+tests:
+  - it: rejects LPX without Grove integration
+    template: charts/dynamo-operator/templates/validate-values.yaml
+    set:
+      dynamo-operator.lpx.enabled: true
+    asserts:
+      - failedTemplate:
+          errorPattern: "LPX requires Grove integration to be enabled"
+
+  - it: keeps LPX disabled by default
+    template: charts/dynamo-operator/templates/operator-config.yaml
+    asserts:
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^lpx:\n  enabled: false$'
+
+  - it: enables LPX and externally installed Grove together
+    template: charts/dynamo-operator/templates/operator-config.yaml
+    set:
+      global.grove.enabled: true
+      dynamo-operator.lpx.enabled: true
+    asserts:
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^lpx:\n  enabled: true$'
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^  grove:\n    enabled: true$'
+
+  - it: renders canonical LPX model registry configuration
+    template: charts/dynamo-operator/templates/operator-config.yaml
+    set:
+      global.grove.enabled: true
+      dynamo-operator.lpx.enabled: true
+      dynamo-operator.lpx.modelRegistryURL: gs://models/canonical
+    asserts:
+      - matchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^lpx:\n  enabled: true\n  modelRegistryURL: "gs://models/canonical"$'
+      - notMatchRegex:
+          path: data["config.yaml"]
+          pattern: '(?m)^lpu:'
+
+  - it: renders the exact inert LpuPipelineRequest manager grant
+    template: charts/dynamo-operator/templates/manager-rbac.yaml
+    asserts:
+      - contains:
+          path: rules
+          content:
+            apiGroups:
+              - scheduling.lpu.nvidia.com
+            resources:
+              - lpupipelinerequests
+            verbs:
+              - create
+              - delete
+              - get
+              - list
+              - watch
+        documentIndex: 0
@@ -96,7 +96,6 @@ tests:
       global.kai-scheduler.install: true
     asserts:
       - notFailedTemplate: {}
-
   # --- Default values should render cleanly ---
 
   - it: should render successfully with default values

@@ -87,6 +87,15 @@ dynamo-operator:
 
   # -- URL for the Model Express server if not deployed by this helm chart. This is ignored if Model Express server is installed by this helm chart (global.model-express.enabled is true).
   modelExpressURL: ""
+
+  # The LPX scheduler and its CRDs must be installed separately.
+  lpx:
+    # -- Enable the Dynamo operator's LPX integration. Startup requires the
+    # externally installed scheduling.lpu.nvidia.com/v1alpha1 LpuPipelineRequest API.
+    enabled: false
+    # -- Configure the LPU model registry used by the LPX integration.
+    modelRegistryURL: ""
+
   # -- DEVELOPMENT AND TESTING ONLY: Namespace-restricted mode is not supported for production. Use cluster-wide mode for production deployments.
   namespaceRestriction:
     # -- DEVELOPMENT AND TESTING ONLY: Enable namespace-restricted reconciliation and admission. Not supported for production.
-Original file line number
+Diff line change
@@ Expand Up / @@ -21,3 +21,4 @@ @@
     .idea/
     *.tmproj
     .vscode/
+    tests/