feat(recipe): add L40S accelerator support

.claude/skills/aicr-analyzing-snapshots/SKILL.md

@@ -99,6 +99,7 @@ Key fields from `GPU.smi`:
 | `gh200` | unresolved — Grace Hopper Superchip, not the discrete H200 GPU (check before h200) |
 | `h200` | h200 (discrete H200 GPU) |
 | `a100` | a100 |
+| `l40s` | l40s |
 | `l40` | l40 |
 | `rtx pro 6000` | rtx-pro-6000 |
 
@@ -274,7 +275,7 @@ aicr recipe \
 | Criteria | Extracted From | Valid Values |
 |----------|---------------|--------------|
 | service | K8s.node.provider / K8s.server.version | eks, gke, aks, oke, kind, lke |
-| accelerator | GPU.smi.gpu.model | h100, h200, gb200, b200, a100, l40, rtx-pro-6000 |
-| os | OS.release.ID | ubuntu, rhel, cos, amazonlinux |
+| accelerator | GPU.smi.gpu.model | h100, h200, gb200, b200, a100, l40s, l40, rtx-pro-6000 |
+| os | OS.release.ID | ubuntu, rhel, cos, amazonlinux, talos, ol |
 | intent | User-specified | training, inference |
 | platform | User-specified | dynamo, kubeflow, nim, runai, slurm |

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -117,7 +117,7 @@ body:
         - Kubernetes version:
         - OS (ubuntu/cos/other) + version:
         - Kernel version:
-        - GPU type (h100/h200/gb200/b200/a100/l40/rtx-pro-6000/other):
+        - GPU type (h100/h200/gb200/b200/a100/l40/l40s/rtx-pro-6000/other):
         - Workload intent (training/inference):
       value: |
         - AICR version (CLI `aicr version`, API image tag, or commit SHA):
@@ -126,7 +126,7 @@ body:
         - Kubernetes version:
         - OS (ubuntu/cos/other) + version:
         - Kernel version:
-        - GPU type (h100/h200/gb200/b200/a100/l40/rtx-pro-6000/other):
+        - GPU type (h100/h200/gb200/b200/a100/l40/l40s/rtx-pro-6000/other):
         - Workload intent (training/inference):
     validations:
       required: true

.settings.yaml

@@ -61,6 +61,8 @@ testing_tools:
   tilt: '0.37.4'
   # renovate: datasource=github-releases depName=helm/helm depType=testing_tools
   helm: 'v4.2.2'
+  # renovate: datasource=github-releases depName=databus23/helm-diff depType=testing_tools
+  helm_diff: 'v3.15.10'
   # renovate: datasource=github-releases depName=helmfile/helmfile depType=testing_tools
   helmfile: 'v1.6.0'
   # helmfile_checksums are refreshed automatically by tools/update-helmfile-checksums

api/aicr/v1/server.yaml

@@ -120,15 +120,15 @@ paths:
           description: GPU/accelerator type. If omitted, treated as "any" (wildcard).
           schema:
             type: string
-            enum: [h100, h200, gb200, b200, a100, l40, rtx-pro-6000, any]
+            enum: [h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000, any]
             default: any
         - name: gpu
           in: query
           required: false
           description: Alias for accelerator parameter (backwards compatibility).
           schema:
             type: string
-            enum: [h100, h200, gb200, b200, a100, l40, rtx-pro-6000, any]
+            enum: [h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000, any]
             default: any
         - name: intent
           in: query
@@ -144,7 +144,7 @@ paths:
           description: GPU node operating system. If omitted, treated as "any" (wildcard).
           schema:
             type: string
-            enum: [ubuntu, rhel, cos, amazonlinux, talos, any]
+            enum: [ubuntu, rhel, cos, amazonlinux, ol, talos, any]
             default: any
         - name: platform
           in: query
@@ -487,15 +487,15 @@ paths:
           description: GPU/accelerator type. If omitted, treated as "any" (wildcard).
           schema:
             type: string
-            enum: [h100, h200, gb200, b200, a100, l40, rtx-pro-6000, any]
+            enum: [h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000, any]
             default: any
         - name: gpu
           in: query
           required: false
           description: Alias for accelerator parameter (backwards compatibility).
           schema:
             type: string
-            enum: [h100, h200, gb200, b200, a100, l40, rtx-pro-6000, any]
+            enum: [h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000, any]
             default: any
         - name: intent
           in: query
@@ -511,7 +511,7 @@ paths:
           description: GPU node operating system. If omitted, treated as "any" (wildcard).
           schema:
             type: string
-            enum: [ubuntu, rhel, cos, amazonlinux, talos, any]
+            enum: [ubuntu, rhel, cos, amazonlinux, ol, talos, any]
             default: any
         - name: platform
           in: query
@@ -1262,7 +1262,7 @@ components:
         accelerator:
           type: string
           description: GPU/accelerator type
-          enum: [h100, h200, gb200, b200, a100, l40, rtx-pro-6000, any]
+          enum: [h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000, any]
           example: h100
         intent:
           type: string
@@ -1272,7 +1272,7 @@ components:
         os:
           type: string
           description: Operating system family
-          enum: [ubuntu, rhel, cos, amazonlinux, talos, any]
+          enum: [ubuntu, rhel, cos, amazonlinux, ol, talos, any]
           example: ubuntu
         platform:
           type: string

docs/contributor/index.md

@@ -179,7 +179,7 @@ both entry points share it. Adding business logic to `pkg/cli` or
 | `pkg/client/v1` | `aicr.Client` facade — shared SDK used by CLI, server, and external Go callers |
 | **Recipe and data** | |
 | `pkg/recipe` | Recipe resolution, overlay/mixin composition, registry. [recipe.md](recipe.md) |
-| `pkg/recipe/oskind` | Single source of truth for OS criterion string values (`ubuntu`, `rhel`, `cos`, `amazonlinux`, `talos`). Imported by `pkg/recipe`, `pkg/collector`, `pkg/snapshotter`, and the CLI. |
+| `pkg/recipe/oskind` | Single source of truth for OS criterion string values (`ubuntu`, `rhel`, `cos`, `amazonlinux`, `ol`, `talos`). Imported by `pkg/recipe`, `pkg/collector`, `pkg/snapshotter`, and the CLI. |
 | `pkg/constraints` | Declarative constraint operators (`>=`, `<=`, tolerance) and evaluation |
 | `pkg/measurement` | Schema for collector output and validator input |
 | `pkg/serializer` | Deterministic YAML/JSON for evidence and bundles |

docs/contributor/recipe.md

@@ -150,9 +150,9 @@ Criteria fields (see `pkg/recipe/criteria.go` `type Criteria`):
 | Field | Type | Wildcard | Static OSS values |
 |---|---|---|---|
 | `service` | `CriteriaServiceType` | `any` or empty | `eks`, `gke`, `aks`, `oke`, `ocp`, `kind`, `lke`, `bcm` |
-| `accelerator` | `CriteriaAcceleratorType` | `any` or empty | `h100`, `h200`, `gb200`, `b200`, `a100`, `l40`, `rtx-pro-6000` |
+| `accelerator` | `CriteriaAcceleratorType` | `any` or empty | `h100`, `h200`, `gb200`, `b200`, `a100`, `l40`, `l40s`, `rtx-pro-6000` |
 | `intent` | `CriteriaIntentType` | `any` or empty | `training`, `inference` |
-| `os` | `CriteriaOSType` | `any` or empty | `ubuntu`, `rhel`, `cos`, `amazonlinux`, `talos` |
+| `os` | `CriteriaOSType` | `any` or empty | `ubuntu`, `rhel`, `cos`, `amazonlinux`, `ol`, `talos` |
 | `platform` | `CriteriaPlatformType` | `any` or empty | `dynamo`, `kubeflow`, `nim`, `runai`, `slurm` |
 | `nodes` | int | `0` | any positive int |
 

docs/user/api-reference.md

@@ -119,10 +119,10 @@ Generate an optimized configuration recipe based on environment parameters.
 | Parameter | Type | Default | Description |
 |-----------|------|---------|-------------|
 | `service` | string | any | K8s service: `eks`, `gke`, `aks`, `oke`, `ocp`, `kind`, `lke`, `bcm`, `any` |
-| `accelerator` | string | any | GPU type: `h100`, `h200`, `gb200`, `b200`, `a100`, `l40`, `rtx-pro-6000`, `any` |
+| `accelerator` | string | any | GPU type: `h100`, `h200`, `gb200`, `b200`, `a100`, `l40`, `l40s`, `rtx-pro-6000`, `any` |
 | `gpu` | string | any | Alias for `accelerator` |
 | `intent` | string | any | Workload: `training`, `inference`, `any` |
-| `os` | string | any | Node OS: `ubuntu`, `rhel`, `cos`, `amazonlinux`, `talos`, `any` |
+| `os` | string | any | Node OS: `ubuntu`, `rhel`, `cos`, `amazonlinux`, `ol`, `talos`, `any` |
 | `platform` | string | any | Platform/framework: `dynamo`, `kubeflow`, `nim`, `runai`, `slurm`, `any` |
 | `nodes` | integer | 0 | GPU node count (0 = any) |
 
@@ -823,7 +823,7 @@ openapi-generator-cli generate -i openapi.yaml -g typescript-fetch -o ./ts-clien
 
 **"Invalid accelerator type" error:**
 ```shell
-# Use valid values: h100, h200, gb200, b200, a100, l40, rtx-pro-6000, any
+# Use valid values: h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000, any
 curl "http://localhost:8080/v1/recipe?accelerator=h100"
 ```
 

docs/user/cli-reference.md

@@ -96,7 +96,7 @@ aicr snapshot [flags]
 | `--runtime-class` | | string | | Runtime class for GPU access without consuming a GPU allocation (e.g., `nvidia`). Mutually exclusive with `--require-gpu`. |
 | `--template` | | string | | Path to Go template file for custom output formatting (requires YAML format) |
 | `--max-nodes-per-entry` | | int | 0 | Maximum node names per taint/label entry in topology collection (0 = unlimited) |
-| `--os` | | string | | Node OS family (`ubuntu`, `rhel`, `cos`, `amazonlinux`, `talos`). Selects the per-OS pod configuration and in-pod service collector backend. `talos` skips the `/run/systemd` and `/etc/os-release` hostPath mounts and uses the Kubernetes-API service backend. Reads `AICR_OS` env when unset. |
+| `--os` | | string | | Node OS family (`ubuntu`, `rhel`, `cos`, `amazonlinux`, `ol`, `talos`). Selects the per-OS pod configuration and in-pod service collector backend. `talos` skips the `/run/systemd` and `/etc/os-release` hostPath mounts and uses the Kubernetes-API service backend. Reads `AICR_OS` env when unset. |
 | `--requests` | | string | | Override agent container resource requests as a comma-separated list of `name=quantity` pairs (e.g. `cpu=500m,memory=1Gi,ephemeral-storage=1Gi`). Unspecified resources keep the built-in privileged or restricted defaults. Reads `AICR_REQUESTS` env when unset. |
 | `--limits` | | string | | Override agent container resource limits as a comma-separated list of `name=quantity` pairs (e.g. `cpu=1,memory=2Gi,ephemeral-storage=2Gi`). Unspecified resources keep the built-in defaults. With `--require-gpu`, the default `nvidia.com/gpu=1` is applied only when `--limits` does not already contain that key — an explicit `--limits nvidia.com/gpu=N` wins. Reads `AICR_LIMITS` env when unset. |
 | `--cluster-config` | | string | | Path to a pre-existing k8s-launch-kit (l8k) `cluster-config.yaml`. Ingests the file's per-hardware-group network topology (PFs, capabilities, kernel modules, machine/GPU type, fabric type) into the snapshot as a `NetworkTopology` Measurement. **Local agent mode only for now** (`AICR_AGENT_MODE=true`) — Job-mode rejects this flag with an `INVALID_REQUEST` error until ConfigMap mounting is implemented. Mutually exclusive with `--discover-network` at the collector level — file path wins when both are set, so callers can default discovery from a flag without inadvertent cluster contact. Reads `AICR_CLUSTER_CONFIG_PATH` env when unset. |
@@ -205,7 +205,7 @@ spec:
         - nvidia.com/gpu=present:NoSchedule
       requireGpu: false
       runtimeClassName: ""         # mutually exclusive with requireGpu
-      os: ""                       # ubuntu | rhel | cos | amazonlinux | talos
+      os: ""                       # ubuntu | rhel | cos | amazonlinux | ol | talos
       requests: ""                 # "cpu=500m,memory=1Gi"
       limits: ""                   # "cpu=1,memory=2Gi"
     execution:
@@ -387,9 +387,9 @@ Generate recipes using direct system parameters:
 | Flag | Short | Type | Description |
 |------|-------|------|-------------|
 | `--service` | | string | K8s service: eks, gke, aks, oke, ocp, kind, lke, bcm, any |
-| `--accelerator` | `--gpu` | string | Accelerator/GPU type: h100, h200, gb200, b200, a100, l40, rtx-pro-6000 |
+| `--accelerator` | `--gpu` | string | Accelerator/GPU type: h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000 |
 | `--intent` | | string | Workload intent: training, inference |
-| `--os` | | string | OS family: ubuntu, rhel, cos, amazonlinux, talos |
+| `--os` | | string | OS family: ubuntu, rhel, cos, amazonlinux, ol, talos |
 | `--platform` | | string | Platform/framework type: dynamo, kubeflow, nim, runai, slurm |
 | `--nodes` | | int | Number of GPU nodes in the cluster |
 | `--output` | `-o` | string | Output file (default: stdout) |

pkg/cli/recipe.go

@@ -100,9 +100,9 @@ func recipeCmd() *cli.Command {
 		Usage:    "Create optimized recipe for given intent and environment parameters.",
 		Description: `Generate configuration recipe based on specified environment parameters including:
   - Kubernetes service type (e.g. eks, gke, aks, oke, kind, lke, bcm)
-  - Accelerator type (e.g. h100, h200, gb200, b200, a100, l40, rtx-pro-6000)
+  - Accelerator type (e.g. h100, h200, gb200, b200, a100, l40, l40s, rtx-pro-6000)
   - Workload intent (e.g. training, inference)
-  - GPU node operating system (e.g. ubuntu, rhel, cos, amazonlinux, talos)
+  - GPU node operating system (e.g. ubuntu, rhel, cos, amazonlinux, ol, talos)
   - Number of GPU nodes in the cluster
 
 The recipe returns a list of components with deployment order based on dependencies.

pkg/cli/snapshot.go

@@ -379,7 +379,7 @@ func snapshotCmdFlags() []cli.Flag {
 		},
 		&cli.StringFlag{
 			Name:     "os",
-			Usage:    "Node OS family (ubuntu, rhel, cos, amazonlinux, talos). Selects the per-OS pod configuration and service collector backend. Talos skips systemd hostPath mounts and uses the Kubernetes-API service backend.",
+			Usage:    "Node OS family (ubuntu, rhel, cos, amazonlinux, ol, talos). Selects the per-OS pod configuration and service collector backend. Talos skips systemd hostPath mounts and uses the Kubernetes-API service backend.",
 			Sources:  cli.EnvVars("AICR_OS"),
 			Category: catAgentDeployment,
 		},

pkg/client/v1/types.go

@@ -122,9 +122,9 @@ type AgentConfig struct {
 //
 // Field meanings match the pkg/recipe.Criteria documentation:
 //   - Service: Kubernetes service flavor (eks/gke/aks/oke/kind/lke/bcm).
-//   - Accelerator: GPU model identifier (h100/h200/b200/gb200/a100/l40/rtx-pro-6000).
+//   - Accelerator: GPU model identifier (h100/h200/b200/gb200/a100/l40/l40s/rtx-pro-6000).
 //   - Intent: workload intent (training/inference).
-//   - OS: worker-node OS (ubuntu/rhel/cos/amazonlinux/talos).
+//   - OS: worker-node OS (ubuntu/rhel/cos/amazonlinux/talos/ol).
 //   - Platform: framework overlay (dynamo/kubeflow/nim/runai/slurm).
 //   - Nodes: worker-node count hint (0 = unspecified).
 //
@@ -221,7 +221,7 @@ type RecipeRequest struct {
 	Intent string
 
 	// OS is the worker-node operating system. Mapped to CriteriaOS.
-	// Supported values: "ubuntu", "rhel", "cos", "amazonlinux".
+	// Supported values: "ubuntu", "rhel", "cos", "amazonlinux", "talos", "ol".
 	// Empty means "unspecified" — recipe resolution will not select
 	// OS-pinned leaf overlays (e.g., h100-eks-ubuntu-training,
 	// h100-gke-cos-training) and will fall back to the OS-agnostic

pkg/collector/k8s/node.go

@@ -92,7 +92,7 @@ func (k *Collector) collectNode(ctx context.Context) (map[string]measurement.Rea
 //   - aws:///us-west-2a/i-0123456789abcdef0 → "eks"
 //   - gce://my-project/us-central1-a/gke-cluster-node → "gke"
 //   - azure:///subscriptions/.../virtualMachines/... → "aks"
-//   - oci://... → "oke"
+//   - ocid1.instance.oc1... → "oke" (OKE emits a raw OCID, no scheme prefix)
 //
 // If the format is unrecognized, it returns the raw provider prefix.
 func parseProvider(providerID string) string {
@@ -101,6 +101,11 @@ func parseProvider(providerID string) string {
 		return ""
 	}
 
+	// OKE nodes set providerID to a raw Oracle OCID (no "://" scheme).
+	if strings.HasPrefix(providerID, "ocid1.") {
+		return "oke"
+	}
+
 	// Split by "://" to get the provider prefix
 	parts := strings.SplitN(providerID, "://", 2)
 

pkg/collector/k8s/node_test.go

@@ -209,15 +209,25 @@ func TestParseProvider(t *testing.T) {
 			want:       "aks",
 		},
 		{
-			name:       "OCI OKE",
-			providerID: "oci://ocid1.instance.oc1.phx.abcdef123456",
+			name:       "OKE raw OCID (actual format)",
+			providerID: "ocid1.instance.oc1.us-chicago-1.anxxeljsaqwjupqcb4pa5kzxy4hef5dtclbkqsnmu6kedbkrne3s2bz5nwzq",
+			want:       "oke",
+		},
+		{
+			name:       "OKE raw OCID short",
+			providerID: "ocid1.instance.oc1.phx.abcdef123456",
 			want:       "oke",
 		},
 		{
 			name:       "empty provider",
 			providerID: "",
 			want:       "",
 		},
+		{
+			name:       "OKE oci:// scheme (legacy format)",
+			providerID: "oci://ocid1.instance.oc1.us-chicago-1.example",
+			want:       "oke",
+		},
 		{
 			name:       "unknown format",
 			providerID: "custom-provider://some-id",

pkg/fingerprint/doc.go

@@ -18,8 +18,8 @@
 //
 // A Fingerprint records the cluster-identity dimensions used to bind
 // a snapshot to a recipe — service (eks/gke/aks/oke/kind/lke/bcm),
-// accelerator (h100/h200/gb200/b200/a100/l40/rtx-pro-6000), OS
-// (ubuntu/rhel/cos/amazonlinux/talos plus raw VERSION_ID), Kubernetes
+// accelerator (h100/h200/gb200/b200/a100/l40/l40s/rtx-pro-6000), OS
+// (ubuntu/rhel/cos/amazonlinux/talos/ol plus raw VERSION_ID), Kubernetes
 // server version, region, total node count, and GPU node count. Each
 // dimension records the resolved value plus an optional source string
 // identifying which collector signal produced it (e.g.,

pkg/fingerprint/from_measurements.go

@@ -27,6 +27,7 @@ import (
 // constants because the collector packages keep them unexported and we
 // don't want to import them just for the names.
 const (
+	serviceOKE              = "oke"
 	subtypeK8sServer        = "server"
 	subtypeK8sNode          = "node"
 	subtypeGPUHardware      = "hardware"
@@ -223,7 +224,7 @@ func populateFromK8s(fp *Fingerprint, m *measurement.Measurement) {
 	}
 	if st := m.GetSubtype(subtypeK8sNode); st != nil {
 		if v, err := st.GetString(keyK8sNodeProvider); err == nil && v != "" {
-			fp.Service = Dimension{Value: v, Source: sourceServiceProvider}
+			fp.Service = Dimension{Value: normalizeProviderID(v), Source: sourceServiceProvider}
 		}
 	}
 }
@@ -320,6 +321,28 @@ func extractRegion(m *measurement.Measurement) (region string, multi bool) {
 	return value, false
 }
 
+// normalizeProviderID maps a raw Kubernetes spec.providerID (or an already-
+// normalized name stored by the collector) to the service type string used in
+// recipe criteria. OKE nodes can appear in four forms:
+//   - "ocid1.instance.oc1..." — bare OCID (no scheme); what K8s stores natively
+//   - "oci://ocid1.instance.oc1..." — full scheme URL; the collector normalizes
+//     this to "oke" via parseProvider, but hand-crafted snapshots may carry it
+//   - "oci" — bare string stored by the collector before the "oci://" → "oke"
+//     mapping existed
+//   - "oke" — already normalized; pass through
+func normalizeProviderID(v string) string {
+	if strings.HasPrefix(v, "ocid1.") {
+		return serviceOKE
+	}
+	if strings.HasPrefix(v, "oci://") {
+		return serviceOKE
+	}
+	if v == "oci" {
+		return serviceOKE
+	}
+	return v
+}
+
 // normalizeOSID maps an /etc/os-release ID value to the
 // recipe.CriteriaOSType enum. Returns "" for IDs that do not match a
 // supported OS kind so callers treat them as "fingerprint did not
@@ -335,6 +358,8 @@ func normalizeOSID(id string) string {
 		return oskind.COS
 	case oskind.AmazonLinux, "amzn", "amazon", "al2", "al2023":
 		return oskind.AmazonLinux
+	case oskind.OracleLinux:
+		return oskind.OracleLinux
 	case oskind.Talos:
 		return oskind.Talos
 	default: