EAI-6030: Add per-hardware-family AIM model source selection

docs/values_inheritance_pattern.md

@@ -146,6 +146,13 @@ When ArgoCD renders applications with multi-source:
 3. **External overrides** from `cluster-values/values.yaml`
 4. **Runtime parameters** (domain, targetRevision) injected by bootstrap
 
+   AIM hardware-family selection is injected separately: cluster-bloom writes the
+   selected families as a YAML list into
+   `apps.aim-cluster-model-source.valuesObject.hardwareFamilies` (see
+   `sources/aim-cluster-model-source`). The value travels as a structured list,
+   not a string, so no comma parsing is involved. The base `root/values.yaml`
+   default is an empty list, which selects the legacy (install-all) branch.
+
 ## Developer Workflow
 
 ### Local Configuration Management (Local Mode)

root/values.yaml

@@ -31,6 +31,11 @@ apps:
     namespace: kaiwo-system
     path: aim-cluster-model-source
     syncWave: -20
+    valuesObject:
+      # Hardware families to install (cpu,epyc,instinct,radeon).
+      # Empty = legacy (install all generic model sources). cluster-bloom
+      # injects the selected families as a YAML list at deploy time.
+      hardwareFamilies: []
   airm:
     repoURL: "{{ .Values.ociRegistry.dockerHub }}"
     repoVersion: "1.1.9"

sources/aim-cluster-model-source/Chart.yaml

@@ -0,0 +1,8 @@
+# Copyright © Advanced Micro Devices, Inc., or its affiliates.
+#
+# SPDX-License-Identifier: MIT
+
+apiVersion: v2
+name: aim-cluster-model-source
+version: 0.1.0
+description: AIMClusterModelSource catalog with GPU-family profile selection

sources/aim-cluster-model-source/README.md

@@ -0,0 +1,59 @@
+<!--
+Copyright © Advanced Micro Devices, Inc., or its affiliates.
+
+SPDX-License-Identifier: MIT
+-->
+
+# aim-cluster-model-source
+
+Helm chart that installs `AIMClusterModelSource` resources. It renders one of
+two mutually exclusive branches, selected by `hardwareFamilies`:
+
+- **Legacy (default):** when `hardwareFamilies` is empty, the chart installs the
+  full set of generic `amd-aim-release-*` model sources (versions 0.8.5, 0.9.0,
+  0.10.0, 0.11.0), unchanged from the pre-chart directory app.
+- **Per-hardware-family profiles:** when `hardwareFamilies` is non-empty, the
+  chart installs only the `AIMClusterModelSource` resources for the listed
+  families. The legacy generic sources are not installed.
+
+## `hardwareFamilies`
+
+A YAML list (the primary form) or a comma-separated string. Allowed values:
+`cpu`, `epyc`, `instinct`, `radeon`. Empty (the default) selects the legacy
+branch.
+
+```yaml
+hardwareFamilies:
+  - epyc
+  - instinct
+```
+
+| Family | Source name | Registry | Notes |
+|---|---|---|---|
+| `instinct` | `amd-aim-instinct-0.12.0` | docker.io | works today |
+| `epyc` | `amd-aim-epyc-0.11.0` | docker.io | works today |
+| `cpu` | `amd-aim-cpu-0.12.0-rc1` | ghcr.io | placeholder, image pull fails until a docker.io release exists |
+| `radeon` | `amd-aim-radeon-0.12.0-rc1` | ghcr.io | placeholder, image pull fails until a docker.io release exists |
+
+`instinct` and `radeon` are GPU families; `cpu` and `epyc` are CPU inference
+targets. `cpu` and `radeon` are only available as `-rc1` on `ghcr.io` and
+require the `ghcr-regcred` pull secret, which this cluster does not provision.
+They are pinned as placeholders, their pull will fail until the team publishes a
+docker.io version.
+
+## Installing
+
+This chart is normally driven by cluster-bloom via the `AIM_HARDWARE_FAMILY`
+install flag, which injects the selected families as a YAML list into
+`apps.aim-cluster-model-source.valuesObject.hardwareFamilies` (see the
+cluster-forge `root` chart). No comma parsing is involved on that path, the
+value travels as a structured list.
+
+For a manual `helm` install, prefer a values file or pass a JSON list. A
+comma-separated string also works because the chart splits it, but note that
+Helm's `--set` and `--set-string` both treat a comma as a list separator and
+will silently drop a multi-value string, so use `--set-json` for the list form:
+
+```bash
+helm install ... --set-json 'hardwareFamilies=["epyc","instinct"]'
+```

sources/aim-cluster-model-source/templates/_helpers.tpl

@@ -0,0 +1,22 @@
+{{/*
+Normalize .Values.hardwareFamilies into a clean list of family tokens.
+Accepts a native list (the primary path, injected by cluster-bloom) or a
+comma-separated string. Trims whitespace and drops empty tokens. Empty input
+yields an empty list, which triggers the legacy branch.
+*/}}
+{{- define "aim.hardwareFamilies" -}}
+{{- $raw := .Values.hardwareFamilies -}}
+{{- $out := list -}}
+{{- if kindIs "string" $raw -}}
+  {{- range (splitList "," $raw) -}}
+    {{- $t := trim . -}}
+    {{- if $t -}}{{- $out = append $out $t -}}{{- end -}}
+  {{- end -}}
+{{- else if kindIs "slice" $raw -}}
+  {{- range $raw -}}
+    {{- $t := trim (toString .) -}}
+    {{- if $t -}}{{- $out = append $out $t -}}{{- end -}}
+  {{- end -}}
+{{- end -}}
+{{- $out | toJson -}}
+{{- end -}}

sources/aim-cluster-model-source/templates/legacy.yaml

@@ -0,0 +1,101 @@
+{{- $families := include "aim.hardwareFamilies" . | fromJsonArray }}
+{{- if not $families }}
+# Copyright © Advanced Micro Devices, Inc., or its affiliates.
+#
+# SPDX-License-Identifier: MIT
+
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-release-0.8.5
+spec:
+  filters:
+    - image: amdenterpriseai/aim-meta-llama-llama-3-1-405b-instruct:0.8.5
+    - image: amdenterpriseai/aim-meta-llama-llama-3-1-8b-instruct:0.8.5
+    - image: amdenterpriseai/aim-meta-llama-llama-3-2-1b-instruct:0.8.5
+    - image: amdenterpriseai/aim-meta-llama-llama-3-2-3b-instruct:0.8.5
+    - image: amdenterpriseai/aim-meta-llama-llama-3-3-70b-instruct:0.8.5-preview
+    - image: amdenterpriseai/aim-mistralai-mistral-small-3-2-24b-instruct-2506:0.8.5
+    - image: amdenterpriseai/aim-mistralai-mixtral-8x22b-instruct-v0-1:0.8.5
+    - image: amdenterpriseai/aim-mistralai-mixtral-8x7b-instruct-v0-1:0.8.5
+    - image: amdenterpriseai/aim-qwen-qwen3-32b:0.8.5
+  maxModels: 100
+  registry: docker.io
+  syncInterval: 1h
+---
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-release-0.9.0
+spec:
+  filters:
+    - image: amdenterpriseai/aim-coherelabs-command-a-reasoning-08-2025:0.9.0
+    - image: amdenterpriseai/aim-openai-gpt-oss-20b:0.9.0
+    - image: amdenterpriseai/aim-openai-gpt-oss-120b:0.9.0
+    - image: amdenterpriseai/aim-qwen-qwen3-235b-a22b:0.9.0
+  maxModels: 100
+  registry: docker.io
+  syncInterval: 1h
+---
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-release-0.10.0
+spec:
+  filters:
+  - image: amdenterpriseai/aim-coherelabs-command-a-reasoning-08-2025:0.10.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-r1-0528:0.10.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-r1:0.10.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-v3-1-terminus:0.10.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-v3-1:0.10.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-1-405b-instruct:0.10.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-1-8b-instruct:0.10.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-2-1b-instruct:0.10.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-2-3b-instruct:0.10.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-3-70b-instruct:0.10.0
+  - image: amdenterpriseai/aim-mistralai-ministral-3-14b-instruct-2512:0.10.0
+  - image: amdenterpriseai/aim-mistralai-mistral-large-3-675b-instruct-2512:0.10.0
+  - image: amdenterpriseai/aim-mistralai-mistral-small-3-2-24b-instruct-2506:0.10.0
+  - image: amdenterpriseai/aim-mistralai-mixtral-8x22b-instruct-v0-1:0.10.0
+  - image: amdenterpriseai/aim-mistralai-mixtral-8x7b-instruct-v0-1:0.10.0
+  - image: amdenterpriseai/aim-openai-gpt-oss-120b:0.10.0
+  - image: amdenterpriseai/aim-openai-gpt-oss-20b:0.10.0
+  - image: amdenterpriseai/aim-qwen-qwen3-235b-a22b:0.10.0
+  - image: amdenterpriseai/aim-qwen-qwen3-32b:0.10.0
+  maxModels: 100
+  registry: docker.io
+  syncInterval: 1h
+---
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-release-0.11.0
+spec:
+  filters:
+  - image: amdenterpriseai/aim-coherelabs-command-a-reasoning-08-2025:0.11.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-r1:0.11.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-r1-0528:0.11.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-v3-1:0.11.0
+  - image: amdenterpriseai/aim-deepseek-ai-deepseek-v3-1-terminus:0.11.0
+  - image: amdenterpriseai/aim-google-gemma-3-27b-it:0.11.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-1-405b-instruct:0.11.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-1-8b-instruct:0.11.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-2-1b-instruct:0.11.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-2-3b-instruct:0.11.0
+  - image: amdenterpriseai/aim-meta-llama-llama-3-3-70b-instruct:0.11.0
+  - image: amdenterpriseai/aim-mistralai-ministral-3-14b-instruct-2512:0.11.0
+  - image: amdenterpriseai/aim-mistralai-ministral-3-14b-reasoning-2512:0.11.0
+  - image: amdenterpriseai/aim-mistralai-mistral-small-3-2-24b-instruct-2506:0.11.0
+  - image: amdenterpriseai/aim-mistralai-mixtral-8x22b-instruct-v0-1:0.11.0
+  - image: amdenterpriseai/aim-mistralai-mixtral-8x7b-instruct-v0-1:0.11.0
+  - image: amdenterpriseai/aim-openai-gpt-oss-120b:0.11.0
+  - image: amdenterpriseai/aim-openai-gpt-oss-20b:0.11.0
+  - image: amdenterpriseai/aim-qwen-qwen3-235b-a22b:0.11.0
+  - image: amdenterpriseai/aim-qwen-qwen3-32b:0.11.0
+  - image: amdenterpriseai/aim-qwen-qwen3-coder-next:0.11.0
+  - image: amdenterpriseai/aim-qwen-qwen3-vl-235b-a22b-instruct:0.11.0
+  - image: amdenterpriseai/aim-qwen-qwen3-vl-235b-a22b-thinking:0.11.0
+  maxModels: 100
+  registry: docker.io
+  syncInterval: 1h
+{{- end }}

sources/aim-cluster-model-source/templates/profiles.yaml

@@ -0,0 +1,68 @@
+{{- $families := include "aim.hardwareFamilies" . | fromJsonArray }}
+{{- if has "instinct" $families }}
+# Copyright © Advanced Micro Devices, Inc., or its affiliates.
+#
+# SPDX-License-Identifier: MIT
+
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-instinct-0.12.0
+spec:
+  filters:
+    - image: amdenterpriseai/aim-google-gemma-3-1b-it:0.12.0
+    - image: amdenterpriseai/aim-minimaxai-minimax-m2-5:0.12.0
+    - image: amdenterpriseai/aim-zai-org-glm-4-7:0.12.0
+  maxModels: 100
+  registry: docker.io
+  syncInterval: 1h
+{{- end }}
+{{- if has "epyc" $families }}
+---
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-epyc-0.11.0
+spec:
+  filters:
+    - image: amdenterpriseai/aim-epyc-meta-llama-llama-3-1-8b-instruct:0.11.0-preview
+    - image: amdenterpriseai/aim-epyc-meta-llama-llama-3-2-1b-instruct:0.11.0-preview
+    - image: amdenterpriseai/aim-epyc-meta-llama-llama-3-2-3b-instruct:0.11.0-preview
+  maxModels: 100
+  registry: docker.io
+  syncInterval: 1h
+{{- end }}
+{{- if has "cpu" $families }}
+---
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-cpu-0.12.0-rc1
+spec:
+  filters:
+    - image: silogen/aim-cpu-qwen-qwen3-0-6b:0.12.0-rc1
+  imagePullSecrets:
+    - name: ghcr-regcred
+  maxModels: 100
+  registry: ghcr.io
+  syncInterval: 1h
+{{- end }}
+{{- if has "radeon" $families }}
+---
+apiVersion: aim.eai.amd.com/v1alpha1
+kind: AIMClusterModelSource
+metadata:
+  name: amd-aim-radeon-0.12.0-rc1
+spec:
+  filters:
+    - image: silogen/aim-radeon-meta-llama-llama-3-1-8b-instruct:0.12.0-rc1
+    - image: silogen/aim-radeon-openai-gpt-oss-20b:0.12.0-rc1
+    - image: silogen/aim-radeon-qwen-qwen3-5-9b:0.12.0-rc1
+    - image: silogen/aim-radeon-qwen-qwen3-vl-8b-instruct:0.12.0-rc1
+    - image: silogen/aim-radeon-zai-org-glm-4-7-flash:0.12.0-rc1
+  imagePullSecrets:
+    - name: ghcr-regcred
+  maxModels: 100
+  registry: ghcr.io
+  syncInterval: 1h
+{{- end }}