feat: add customizer as a plugin - AALGO-215

.cursor/rules/nemo-platform.mdc

@@ -37,9 +37,8 @@ User-facing skills in `packages/nemo_platform_ext/src/nemo_platform_ext/skills/`
 - `nemo-try-agent`: test a deployed agent or chat with a model.
 - `nemo-status`: read-only health dashboard. Run this before assuming the platform is up.
 - `nemo-teardown`: guided shutdown with confirmation.
-- `nemo-fine-tune`: fine-tuning. Not yet available; the skill tells the user this honestly instead of letting you improvise.
 
-Plugin-owned skills under `plugins/*/src/*/skills/` handle their own routing for guardrails, evaluations, optimization, data designer, anonymizer, and auditor.
+Plugin-owned skills under `plugins/*/src/*/skills/` handle their own routing for customization, guardrails, evaluations, optimization, data designer, anonymizer, and auditor.
 
 ## Sandboxed environments
 

AGENTS.md

@@ -33,9 +33,8 @@ User-facing skills in `packages/nemo_platform_ext/src/nemo_platform_ext/skills/`
 - `nemo-try-agent`: test a deployed agent or chat with a model.
 - `nemo-status`: read-only health dashboard.
 - `nemo-teardown`: guided shutdown with confirmation.
-- `nemo-fine-tune`: fine-tuning. Not yet available; the skill tells the user it's not shipped instead of improvising with another training library.
 
-Plugin-owned skills under `plugins/*/src/*/skills/` handle guardrails, evaluations, optimization, data designer, anonymizer, and auditor.
+Plugin-owned skills under `plugins/*/src/*/skills/` handle their own routing for customization, guardrails, evaluations, optimization, data designer, anonymizer, and auditor.
 
 ### Working in a sandboxed environment
 

CLAUDE.md

@@ -33,9 +33,8 @@ User-facing skills in `packages/nemo_platform_ext/src/nemo_platform_ext/skills/`
 - `nemo-try-agent`: test a deployed agent or chat with a model.
 - `nemo-status`: read-only health dashboard. Run this before assuming the platform is up.
 - `nemo-teardown`: guided shutdown with confirmation.
-- `nemo-fine-tune`: fine-tuning. Not yet available; the skill tells the user it's not shipped instead of letting the agent improvise with another training library.
 
-Plugin-owned skills live under `plugins/*/src/*/skills/` and handle their own routing for guardrails, evaluations, optimization, data designer, anonymizer, and auditor.
+Plugin-owned skills live under `plugins/*/src/*/skills/` and handle their own routing for customization, guardrails, evaluations, optimization, data designer, anonymizer, and auditor.
 
 ### Working in a sandboxed coding-agent environment
 

docker-bake.automodel.hcl

@@ -0,0 +1,203 @@
+# nmp-automodel image bake - run from Platform repo root (context = ".").
+#
+# Inspect targets (no build; finishes in ~0s):
+#   docker buildx bake --print -f docker-bake.automodel.hcl nmp-automodel-gpu-wheels
+#
+# Build wheels (override registry/tag via env, not --set):
+#   export WHEELS_REGISTRY=nvcr.io/0921617854601259/nemo-platform-dev
+#   export WHEELS_TAG=$(git rev-parse --short HEAD)
+#   docker buildx bake -f docker-bake.automodel.hcl nmp-automodel-gpu-wheels --push
+#
+# Build automodel images:
+#   docker buildx bake -f docker-bake.automodel.hcl nmp-automodel-base-builder
+#
+# Published tags: nvcr.io/0921617854601259/nemo-platform-dev/nmp-automodel-{base,tasks,training}:<tag>
+# NVCR allows only one repo segment after the registry prefix (no nmp/automodel-base nesting).
+
+variable "IMAGE_REGISTRY" {
+  default = "nvcr.io/0921617854601259/nemo-platform-dev"
+}
+
+variable "BASE_REGISTRY" {
+  default = "nvcr.io/0921617854601259/nemo-platform-dev"
+}
+
+variable "WHEELS_REGISTRY" {
+  default = "nvcr.io/0921617854601259/nemo-platform-dev"
+}
+
+variable "BAKE_TAG" {
+  default = "local"
+}
+
+variable "BASE_TAG_AUTOMODEL" {
+  default = "local"
+}
+
+variable "WHEELS_TAG" {
+  default = "3fd6986ff173b598446ffac06d9be3f84b482495"
+}
+
+variable "CUDA_VERSION" {
+  default = "12.8.1"
+}
+
+variable "MAMBA_22_COMMIT" {
+  default = "6b32be06d026e170b3fdaf3ae6282c5a6ff57b06"
+}
+
+variable "MAMBA_23_COMMIT" {
+  default = "v2.3.0"
+}
+
+variable "CAUSAL_CONV1D_VERSION" {
+  default = "v1.5.3"
+}
+
+# For local builds: --set "*.platform=linux/amd64"
+variable "BUILD_PLATFORMS" {
+  default = ["linux/arm64"]
+}
+
+function "wheel_tags" {
+  params = [name]
+  result = ["${WHEELS_REGISTRY}/${name}:${WHEELS_TAG}"]
+}
+
+function "get_causal_conv1d_wheel_image" {
+  params = []
+  result = "${WHEELS_REGISTRY}/causal-conv1d-wheel:${WHEELS_TAG}"
+}
+
+function "get_mamba_ssm_wheel_image" {
+  params = []
+  result = "${WHEELS_REGISTRY}/mamba-ssm-wheel:${WHEELS_TAG}"
+}
+
+group "nmp-automodel-gpu-wheels" {
+  targets = [
+    "causal-conv1d-wheel",
+    "mamba-ssm-wheel",
+  ]
+}
+
+group "nmp-automodel" {
+  targets = [
+    "nmp-automodel-base-builder",
+    "nmp-automodel-tasks-docker",
+    "nmp-automodel-training-docker",
+    "nmp-automodel-tasks-smoke-test",
+    "nmp-automodel-training-smoke-test",
+  ]
+}
+
+# Pre-built mamba-ssm / causal-conv1d wheels (cp311, cp312, cu13.1.1). Pushed to WHEELS_REGISTRY.
+target "causal-conv1d-wheel" {
+  target     = "causal-conv1d-wheel"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.mamba-wheel"
+  tags       = wheel_tags("causal-conv1d-wheel")
+  args = {
+    CUDA_VERSION          = CUDA_VERSION
+    CAUSAL_CONV1D_VERSION = CAUSAL_CONV1D_VERSION
+  }
+  platforms = BUILD_PLATFORMS
+}
+
+target "mamba-ssm-wheel" {
+  target     = "mamba-ssm-wheel"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.mamba-wheel"
+  tags       = wheel_tags("mamba-ssm-wheel")
+  args = {
+    CUDA_VERSION    = CUDA_VERSION
+    MAMBA_22_COMMIT = MAMBA_22_COMMIT
+    MAMBA_23_COMMIT = MAMBA_23_COMMIT
+  }
+  platforms = BUILD_PLATFORMS
+}
+
+target "platform-workspace" {
+  target     = "platform-workspace"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.platform-workspace"
+}
+
+target "nmp-automodel-base-builder" {
+  target          = "nmp-automodel-base-builder"
+  context         = "."
+  dockerfile      = "services/automodel/docker/Dockerfile.nmp-automodel-base"
+  no-cache-filter = ["automodel-clone"]
+  tags            = ["${IMAGE_REGISTRY}/nmp-automodel-base:${BAKE_TAG}"]
+  args = {
+    CAUSAL_CONV1D_WHEEL_IMAGE = get_causal_conv1d_wheel_image()
+    MAMBA_SSM_WHEEL_IMAGE     = get_mamba_ssm_wheel_image()
+  }
+  platforms = BUILD_PLATFORMS
+}
+
+target "nmp-automodel-tasks-docker" {
+  target     = "runtime"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.nmp-automodel-tasks"
+  contexts = {
+    platform-workspace = "target:platform-workspace"
+    nmp-automodel-base = "target:nmp-automodel-base-builder"
+  }
+  tags = ["${IMAGE_REGISTRY}/nmp-automodel-tasks:${BAKE_TAG}"]
+  args = {
+    BASE_REGISTRY      = BASE_REGISTRY
+    BASE_TAG_AUTOMODEL = BASE_TAG_AUTOMODEL
+  }
+  platforms = BUILD_PLATFORMS
+}
+
+target "nmp-automodel-training-docker" {
+  target     = "runtime"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.nmp-automodel-training"
+  contexts = {
+    platform-workspace = "target:platform-workspace"
+    nmp-automodel-base = "target:nmp-automodel-base-builder"
+  }
+  tags = ["${IMAGE_REGISTRY}/nmp-automodel-training:${BAKE_TAG}"]
+  args = {
+    BASE_REGISTRY      = BASE_REGISTRY
+    BASE_TAG_AUTOMODEL = BASE_TAG_AUTOMODEL
+  }
+  platforms = BUILD_PLATFORMS
+}
+
+target "nmp-automodel-tasks-smoke-test" {
+  target     = "smoke-test"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.nmp-automodel-tasks"
+  contexts = {
+    platform-workspace = "target:platform-workspace"
+    nmp-automodel-base = "target:nmp-automodel-base-builder"
+  }
+  args = {
+    BASE_REGISTRY      = BASE_REGISTRY
+    BASE_TAG_AUTOMODEL = BASE_TAG_AUTOMODEL
+    SMOKE_MARKER       = "smoke_nmp_automodel_tasks"
+  }
+  output    = ["type=cacheonly"]
+  platforms = BUILD_PLATFORMS
+}
+
+target "nmp-automodel-training-smoke-test" {
+  target     = "smoke-test"
+  context    = "."
+  dockerfile = "services/automodel/docker/Dockerfile.nmp-automodel-training"
+  contexts = {
+    platform-workspace = "target:platform-workspace"
+    nmp-automodel-base = "target:nmp-automodel-base-builder"
+  }
+  args = {
+    BASE_REGISTRY      = BASE_REGISTRY
+    BASE_TAG_AUTOMODEL = BASE_TAG_AUTOMODEL
+    SMOKE_MARKER       = "smoke_nmp_automodel_training"
+  }
+  output    = ["type=cacheonly"]
+  platforms = BUILD_PLATFORMS
+}

docs/agents/plugins.md

@@ -68,7 +68,7 @@ The skills that drive the agent lifecycle are:
 | `agents-optimize` | Selects a deployed agent, establishes an evaluation baseline, and suggests Switchyard routing, model swaps, skill optimization, prompt tuning, and new-model evaluations. See [Optimize Agents](optimization.md). |
 | `agents-secure` | Selects a deployed agent, checks guardrail coverage, and scans recent telemetry for sensitive data. See [Secure Agents](security.md). |
 
-Plugin-owned skills cover guardrails, evaluations, optimization, data
+Plugin-owned skills cover customization, guardrails, evaluations, optimization, data
 designer, anonymizer, and auditor. They are installed with their plugin and
 appear in `nemo skills list` once the platform restarts.
 

docs/set-up/config-reference.md

@@ -409,6 +409,8 @@ jobs:
   reconcile_interval_seconds: 2
   # Interval in seconds for the job scheduler to run | default: 5
   schedule_interval_seconds: 5
+  # Register the subprocess/default execution profile. When unset, defaults to true for docker/none runtimes and false for kubernetes.
+  enable_subprocess_executor:
 ```
 
 ### `models`