merge(main): incorporate current workflow prerequisites

Signed-off-by: Apurv Kumaria <akumaria@nvidia.com>

Author: apurvvkumaria

.github/workflows/podman-cpu-proof.yaml

@@ -80,6 +80,9 @@ jobs:
       - name: Build shared sandbox-name contract
         run: npm run build:policy-boundary
 
+      - name: Compile managed inference catalog
+        run: npm run catalog:compile
+
       - name: Prepare system and app slice CPU settings without service delegation
         shell: bash
         run: node --experimental-strip-types scripts/checks/run-portable-cpu-delegation-proof.mts prepare

.github/workflows/portable-profile-e2e.yaml

@@ -54,6 +54,9 @@ jobs:
       - name: Install root dependencies
         run: npm ci --ignore-scripts
 
+      - name: Compile managed inference catalogue
+        run: npm run catalog:compile
+
       - name: Build shared policy boundary
         run: npm run build:policy-boundary
 

.github/workflows/pr-self-hosted.yaml

@@ -28,17 +28,7 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  get-pr-info:
-    runs-on: ubuntu-latest
-    timeout-minutes: 5
-    outputs:
-      pr-info: ${{ steps.get-pr-info.outputs.pr-info }}
-    steps:
-      - id: get-pr-info
-        uses: nv-gha-runners/get-pr-info@090577647b8ddc4e06e809e264f7881650ecdccf # main
-
   select-llama-cpp-generic-gpu:
-    needs: get-pr-info
     runs-on: ubuntu-latest
     timeout-minutes: 5
     outputs:
@@ -48,12 +38,16 @@ jobs:
         name: Select llama.cpp generic GPU E2E from PR files
         env:
           GH_TOKEN: ${{ github.token }}
-          PR_INFO: ${{ needs.get-pr-info.outputs.pr-info }}
         shell: bash
         run: |
           set -euo pipefail
-          pr_number="$(jq -er '.number | select(type == "number" and . > 0)' <<<"$PR_INFO")"
-          head_sha="$(jq -er '.head.sha | select(test("^[a-f0-9]{40}$"))' <<<"$PR_INFO")"
+          [[ "$GITHUB_REF_NAME" =~ ^pull-request/([0-9]+)$ ]] || {
+            echo "::error::Copied branch name does not identify a PR" >&2
+            exit 1
+          }
+          pr_number="${BASH_REMATCH[1]}"
+          pr_json="$(gh api "repos/$GITHUB_REPOSITORY/pulls/$pr_number")"
+          head_sha="$(jq -er '.head.sha | select(test("^[a-f0-9]{40}$"))' <<<"$pr_json")"
           [[ "$head_sha" == "$GITHUB_SHA" ]] || {
             echo "::error::Copied PR branch SHA does not match the current PR head" >&2
             exit 1
@@ -89,9 +83,7 @@ jobs:
 
   llama-cpp-generic-gpu:
     name: llama.cpp on generic NVIDIA GPU
-    needs:
-      - get-pr-info
-      - select-llama-cpp-generic-gpu
+    needs: select-llama-cpp-generic-gpu
     if: ${{ needs.select-llama-cpp-generic-gpu.outputs.selected == 'true' }}
     runs-on: linux-amd64-gpu-rtxpro6000-latest-1
     timeout-minutes: 120
@@ -101,9 +93,9 @@ jobs:
       E2E_TARGET_ID: llama-cpp-generic-gpu
       NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
       NEMOCLAW_CLI_BIN: ${{ github.workspace }}/bin/nemoclaw.js
-      NEMOCLAW_E2E_EXPECTED_SHA: ${{ fromJSON(needs.get-pr-info.outputs.pr-info).head.sha }}
+      NEMOCLAW_E2E_EXPECTED_SHA: ${{ github.sha }}
       NEMOCLAW_E2E_SHARD: default
-      NEMOCLAW_LLAMA_CPP_QUALIFICATION_HEAD_SHA: ${{ fromJSON(needs.get-pr-info.outputs.pr-info).head.sha }}
+      NEMOCLAW_LLAMA_CPP_QUALIFICATION_HEAD_SHA: ${{ github.sha }}
       NEMOCLAW_LLAMACPP_RECIPE: llama-cpp.nemotron-3-nano-30b-a3b.spark-single.v1
       NEMOCLAW_NON_INTERACTIVE: "1"
       NEMOCLAW_PROVIDER: install-llama-cpp
@@ -115,7 +107,7 @@ jobs:
         uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
         with:
           persist-credentials: false
-          ref: ${{ fromJSON(needs.get-pr-info.outputs.pr-info).head.sha }}
+          ref: ${{ github.sha }}
 
       - name: Prepare E2E workspace
         uses: NVIDIA/NemoClaw/.github/actions/prepare-e2e@f6304bc25fc35bfaa441c8c2fbfee38f72805a75

ci/source-shape-test-budget.json

@@ -61,6 +61,11 @@
       "test": "routes only the measured heavy lanes on trusted main (#7145)",
       "category": "security"
     },
+    {
+      "file": "test/e2e/support/portable-profile-rootless-runtime-workflow.test.ts",
+      "test": "compiles the managed inference catalogue after dependency installation and before the live E2E test (#9680)",
+      "category": "compatibility"
+    },
     {
       "file": "test/growth-guardrails-workflow-boundary.test.ts",
       "test": "runs the trusted Vitest guardrails against pull request data",
@@ -76,6 +81,11 @@
       "test": "restores exact locked posture after root-separated repair and later failure (#7033)",
       "category": "security"
     },
+    {
+      "file": "test/muse-glimmer-vllm-image-provenance.test.ts",
+      "test": "binds the checked-in provenance to the selected runtime",
+      "category": "security"
+    },
     {
       "file": "test/muse-glimmer-vllm-image-provenance.test.ts",
       "test": "rejects %s",

docs/inference/set-up-vllm.mdx

@@ -158,8 +158,9 @@ Managed profiles and model-specific recipes use immutable image digests:
 
 - DGX Spark and DGX Station models without a model-specific runtime use the `linux/arm64` digest `sha256:9204569b17ee4c0eff75194b8e6e458479c8aee18953b5ab9cf359fcdac659e2` with a compressed layer size of `9.60 GB` under `nvcr.io/nvidia/vllm:26.05.post1-py3`.
   N1x uses the same image.
-- The DGX Spark Muse Glimmer recipe uses the `linux/arm64` post-merge vLLM nightly digest `sha256:677afd5bf3b4bb9881f91e107af7098f8410726b4c05b25cb4a815900b398204` with a compressed layer size of `9.70 GB` under `vllm/vllm-openai`.
-  The image was built from vLLM commit `ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9`, which contains merged Muse Glimmer support commit `6adad08767583f52eb4d2122111af0bf638ed5e6`.
+- The DGX Spark Muse Glimmer recipe uses the `linux/arm64` post-fix vLLM nightly digest `sha256:b0e84e5f2b00a7268e4fdda332790ebd4bfb166b64757e166914753afaeee965` with a compressed layer size of `9.71 GB` under `vllm/vllm-openai`.
+  The image was built from vLLM commit `5a4c8d99242e9e069b604d0e9b969e77f7dd501d`, which contains merged Muse Glimmer support commit `6adad08767583f52eb4d2122111af0bf638ed5e6` and the vLLM dependency fix from commit `90984ddbed27a09409506d6d6c0eea87f54b04b5`.
+  The image includes `huggingface_hub` `1.28.0`, which preserves the pinned model revision when vLLM transfers configuration to the engine process.
   It pins Hugging Face revision `d35cb79050f419c457611b1cee5c5d15b176f285` for the approximately `25.45 GB` model download.
 - The DGX Spark Nemotron 3.5 Lightning recipe uses the `linux/arm64` digest `sha256:3af90144a0926e5c5fe46ee16e5201e763dd854538b9d7ce433755f11dadaf78` with a compressed layer size of approximately `12.69 GB` under `vllm/vllm-openai`.
   It pins Hugging Face revision `0dcd680e5585c791728c83342b311d0a0026dbeb` for the approximately `21.56 GB` model download.

docs/manage-sandboxes/uninstall-nemoclaw.mdx

@@ -155,6 +155,9 @@ The legacy `default` namespace is not isolated across gateways, so it cannot sat
 Scoped uninstall stops before it deletes a sandbox, registry row, or gateway registration and preserves the selected gateway's runtime evidence and local state.
 Because the supported OpenShell Podman schema does not expose `sandbox_namespace`, scoped Podman uninstall fails closed before signaling and preserves the gateway runtime evidence and local state.
 A full uninstall for the portable experimental profile detects ownership only from durable lifecycle state.
+When no Portable lifecycle receipt or retirement record exists, uninstall runs ordinary cleanup and removes any abandoned Portable configuration directory.
+This path covers an incomplete or absent onboarding session and an absent state directory.
+Uninstall still stops before cleanup for an unknown Portable uninstall artifact or a completed Portable onboarding session whose lifecycle receipt is missing.
 It records the initial lifecycle receipts and sandbox registry state for comparison.
 NemoClaw first acquires the process-bound portable host lock at `~/.nemoclaw-portable-host.lock`.
 This host-wide fence coordinates portable onboarding, rebuild, and uninstall.

internal/security-reviews/muse-glimmer-vllm-image-provenance-v1.json

@@ -11,51 +11,64 @@
     "repository": "vllm/vllm-openai"
   },
   "image": {
-    "reference": "vllm/vllm-openai@sha256:677afd5bf3b4bb9881f91e107af7098f8410726b4c05b25cb4a815900b398204",
-    "manifestDigest": "sha256:677afd5bf3b4bb9881f91e107af7098f8410726b4c05b25cb4a815900b398204",
+    "reference": "vllm/vllm-openai@sha256:b0e84e5f2b00a7268e4fdda332790ebd4bfb166b64757e166914753afaeee965",
+    "manifestDigest": "sha256:b0e84e5f2b00a7268e4fdda332790ebd4bfb166b64757e166914753afaeee965",
     "manifestMediaType": "application/vnd.docker.distribution.manifest.v2+json",
-    "manifestUrl": "https://registry-1.docker.io/v2/vllm/vllm-openai/manifests/sha256:677afd5bf3b4bb9881f91e107af7098f8410726b4c05b25cb4a815900b398204",
-    "configDigest": "sha256:c3f199e54a26d2d7a9a41115cd07ce9d90a6488c5a4e75b17129e1006ce533fd",
+    "manifestUrl": "https://registry-1.docker.io/v2/vllm/vllm-openai/manifests/sha256:b0e84e5f2b00a7268e4fdda332790ebd4bfb166b64757e166914753afaeee965",
+    "configDigest": "sha256:49d2eb65dc2a8dea24e43c27b226f650481ac97d4ba9c567b6e1ca08bc472303",
     "configMediaType": "application/vnd.docker.container.image.v1+json",
-    "configSizeBytes": 34762,
-    "configUrl": "https://registry-1.docker.io/v2/vllm/vllm-openai/blobs/sha256:c3f199e54a26d2d7a9a41115cd07ce9d90a6488c5a4e75b17129e1006ce533fd",
+    "configSizeBytes": 34738,
+    "configUrl": "https://registry-1.docker.io/v2/vllm/vllm-openai/blobs/sha256:49d2eb65dc2a8dea24e43c27b226f650481ac97d4ba9c567b6e1ca08bc472303",
     "layerCount": 32,
-    "compressedLayerSizeBytes": 9699710136,
-    "createdAt": "2026-08-14T05:33:50.528328374Z",
+    "compressedLayerSizeBytes": 9706339423,
+    "createdAt": "2026-08-19T05:34:00.562836767Z",
     "platform": {
       "os": "linux",
       "architecture": "arm64"
     }
   },
   "build": {
-    "sourceRepository": "https://github.com/vllm-project/vllm",
-    "sourceRevision": "ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
-    "sourceRevisionUrl": "https://github.com/vllm-project/vllm/commit/ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
-    "imageTag": "vllm/vllm-openai:nightly-ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
-    "pipelineId": "019d130e-464e-4ff7-b84b-492992c0c06b",
-    "pipelineUrl": "https://buildkite.com/vllm/release-v2/builds/5174"
+    "sourceRevision": "5a4c8d99242e9e069b604d0e9b969e77f7dd501d",
+    "imageTag": "vllm/vllm-openai:nightly-5a4c8d99242e9e069b604d0e9b969e77f7dd501d",
+    "pipelineId": "019d130e-464e-4ff7-b84b-492992c0c06b"
   },
   "upstreamSupport": {
     "museMergeCommit": "6adad08767583f52eb4d2122111af0bf638ed5e6",
-    "museMergeCommitUrl": "https://github.com/vllm-project/vllm/commit/6adad08767583f52eb4d2122111af0bf638ed5e6",
-    "comparisonUrl": "https://github.com/vllm-project/vllm/compare/6adad08767583f52eb4d2122111af0bf638ed5e6...ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
-    "relationship": "direct-descendant",
-    "aheadBy": 1
+    "relationship": "descendant",
+    "aheadBy": 173,
+    "revisionFixCommit": "90984ddbed27a09409506d6d6c0eea87f54b04b5",
+    "revisionFixRelationship": "descendant",
+    "revisionFixAheadBy": 33
   },
   "reportedLabels": {
-    "ai.vllm.build.commit": "ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
+    "ai.vllm.build.commit": "5a4c8d99242e9e069b604d0e9b969e77f7dd501d",
     "ai.vllm.build.pipeline": "019d130e-464e-4ff7-b84b-492992c0c06b",
-    "ai.vllm.build.url": "https://buildkite.com/vllm/release-v2/builds/5174",
-    "ai.vllm.image.tag": "vllm/vllm-openai:nightly-ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
-    "org.opencontainers.image.revision": "ac7509e2b1db40fec2f03dde1ed4e9dfdc2338c9",
-    "org.opencontainers.image.source": "https://github.com/vllm-project/vllm"
+    "ai.vllm.image.tag": "vllm/vllm-openai:nightly-5a4c8d99242e9e069b604d0e9b969e77f7dd501d",
+    "org.opencontainers.image.revision": "5a4c8d99242e9e069b604d0e9b969e77f7dd501d"
+  },
+  "runtimeDependencies": {
+    "vllmVersion": "0.26.1rc1.dev942+g5a4c8d992",
+    "huggingfaceHubVersion": "1.28.0",
+    "sentencepieceVersion": "0.2.2",
+    "tiktokenVersion": "0.14.0"
+  },
+  "revisionSerialization": {
+    "model": "Inferact/Muse-Glimmer-30B-NVFP4-W4A4",
+    "requestedRevision": "d35cb79050f419c457611b1cee5c5d15b176f285",
+    "resolvedRevisionBeforePickle": "d35cb79050f419c457611b1cee5c5d15b176f285",
+    "resolvedRevisionAfterPickle": "d35cb79050f419c457611b1cee5c5d15b176f285",
+    "preserved": true
   },
   "verification": {
-    "observedAt": "2026-08-14T16:33:50Z",
+    "observedAt": "2026-08-19T22:53:46Z",
     "methods": [
       "docker-buildx-imagetools-inspect-raw",
       "docker-image-inspect",
-      "github.qkg1.toppare-api"
+      "github.qkg1.toppare-api",
+      "container-package-imports",
+      "resolved-revision-pickle-round-trip",
+      "dgx-spark-cold-cache-startup",
+      "vllm-openai-api-validation"
     ],
     "signedProvenanceAttestation": "not-available"
   }

managed-inference/presets/vllm.linux-amd64-nvidia.single.muse-glimmer-30b-nvfp4-w4a4.yaml

@@ -0,0 +1,101 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+apiVersion: nemoclaw.nvidia.com/managed-inference/v1
+kind: ServingPreset
+
+metadata:
+  id: vllm.linux-amd64-nvidia.single.muse-glimmer-30b-nvfp4-w4a4
+  displayName: Muse Glimmer 30B NVFP4 W4A4 on one Linux x86_64 NVIDIA GPU
+  supportState: experimental
+  validation:
+    level: hardware
+    evidence: brev-rtx-pro-server-6000-2026-08-18
+
+spec:
+  selection: explicit-only
+  priority: 310
+
+  requirements:
+    all:
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.platform.supported
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.docker.available
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.docker.daemon_reachable
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.docker.runtime_supported
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.docker.storage_compatible
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.gpu.nvidia_available
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.gpu.container_toolkit_available
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: capability
+          id: host.gpu.cdi_healthy
+          state: present
+      - readiness:
+          scope: everyNode
+          kind: observation
+          id: host.os.platform
+          comparison:
+            operator: equals
+            value: linux
+      - readiness:
+          scope: everyNode
+          kind: observation
+          id: host.os.architecture
+          comparison:
+            operator: equals
+            value: x64
+      - readiness:
+          scope: everyNode
+          kind: observation
+          id: host.docker.runtime
+          comparison:
+            operator: equals
+            value: docker
+      - readiness:
+          scope: everyNode
+          kind: observation
+          id: host.gpu.count
+          comparison:
+            operator: at-least
+            value: 1
+      - readiness:
+          scope: everyNode
+          kind: observation
+          id: host.gpu.driver_version
+          comparison:
+            operator: version-at-least
+            value: 580.65.06
+
+  plan:
+    backend: vllm
+    platform: linux
+    interactive: false
+    recipeRef: vllm.muse-glimmer-30b-nvfp4-w4a4.linux-amd64-single.v1