ci(actions): add manual gpu branch e2e path

Author: pimlock

.github/workflows/branch-e2e.yml

@@ -3,6 +3,16 @@ name: Branch E2E Checks
 on:
   pull_request:
     types: [opened, synchronize, reopened, labeled]
+  workflow_dispatch:
+    inputs:
+      gpu-runner:
+        description: GPU runner label to target
+        required: true
+        type: choice
+        default: linux-arm64-gpu-l4-latest-1
+        options:
+          - linux-arm64-gpu-l4-latest-1
+          - linux-amd64-gpu-rtxpro6000-latest-1
 
 permissions:
   contents: read
@@ -26,8 +36,34 @@ jobs:
       runner: build-arm64
 
   e2e:
+    if: github.event_name == 'pull_request' && contains(github.event.pull_request.labels.*.name, 'test:e2e')
     needs: [build-gateway, build-cluster]
     uses: ./.github/workflows/e2e-test.yml
     with:
       image-tag: ${{ github.sha }}
       runner: build-arm64
+
+  build-gateway-gpu:
+    if: github.event_name == 'workflow_dispatch'
+    uses: ./.github/workflows/docker-build.yml
+    with:
+      component: gateway
+      platform: ${{ inputs.gpu-runner == 'linux-arm64-gpu-l4-latest-1' && 'linux/arm64' || 'linux/amd64' }}
+      runner: ${{ inputs.gpu-runner == 'linux-arm64-gpu-l4-latest-1' && 'build-arm64' || 'build-amd64' }}
+
+  build-cluster-gpu:
+    if: github.event_name == 'workflow_dispatch'
+    uses: ./.github/workflows/docker-build.yml
+    with:
+      component: cluster
+      platform: ${{ inputs.gpu-runner == 'linux-arm64-gpu-l4-latest-1' && 'linux/arm64' || 'linux/amd64' }}
+      runner: ${{ inputs.gpu-runner == 'linux-arm64-gpu-l4-latest-1' && 'build-arm64' || 'build-amd64' }}
+
+  e2e-gpu:
+    if: github.event_name == 'workflow_dispatch'
+    needs: [build-gateway-gpu, build-cluster-gpu]
+    uses: ./.github/workflows/e2e-test.yml
+    with:
+      image-tag: ${{ github.sha }}
+      runner: ${{ inputs.gpu-runner }}
+      suite: python-gpu

.github/workflows/e2e-test.yml

@@ -1,6 +1,29 @@
 name: E2E Test
 
 on:
+  workflow_dispatch:
+    inputs:
+      image-tag:
+        description: "Image tag to test (typically the commit SHA)"
+        required: false
+        type: string
+        default: ""
+      runner:
+        description: "GitHub Actions runner label"
+        required: false
+        type: string
+        default: "build-amd64"
+      suite:
+        description: "E2E suite to run"
+        required: false
+        type: choice
+        default: all
+        options:
+          - all
+          - python
+          - rust
+          - gateway-resume
+          - python-gpu
   workflow_call:
     inputs:
       image-tag:
@@ -12,6 +35,11 @@ on:
         required: false
         type: string
         default: "build-amd64"
+      suite:
+        description: "E2E suite to run"
+        required: false
+        type: string
+        default: "all"
 
 permissions:
   contents: read
@@ -20,6 +48,7 @@ permissions:
 jobs:
   e2e:
     name: "E2E (${{ matrix.suite }})"
+    if: ${{ inputs.suite == 'all' || inputs.suite == matrix.suite }}
     runs-on: ${{ inputs.runner }}
     timeout-minutes: 30
     strategy:
@@ -29,15 +58,23 @@ jobs:
           - suite: python
             cluster: e2e-python
             port: "8080"
+            cluster_gpu: "0"
             cmd: "mise run --no-prepare --skip-deps e2e:python"
           - suite: rust
             cluster: e2e-rust
             port: "8081"
+            cluster_gpu: "0"
             cmd: "mise run --no-prepare --skip-deps e2e:rust"
           - suite: gateway-resume
             cluster: e2e-resume
             port: "8082"
+            cluster_gpu: "0"
             cmd: "cargo test --manifest-path e2e/rust/Cargo.toml --features e2e --test gateway_resume"
+          - suite: python-gpu
+            cluster: e2e-gpu
+            port: "8083"
+            cluster_gpu: "1"
+            cmd: "mise run --no-prepare --skip-deps e2e:python:gpu"
     container:
       image: ghcr.io/nvidia/openshell/ci:latest
       credentials:
@@ -48,7 +85,7 @@ jobs:
         - /var/run/docker.sock:/var/run/docker.sock
     env:
       MISE_GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-      IMAGE_TAG: ${{ inputs.image-tag }}
+      IMAGE_TAG: ${{ inputs.image-tag || github.sha }}
       OPENSHELL_REGISTRY: ghcr.io/nvidia/openshell
       OPENSHELL_REGISTRY_HOST: ghcr.io
       OPENSHELL_REGISTRY_NAMESPACE: nvidia/openshell
@@ -62,28 +99,29 @@ jobs:
         run: echo "${{ secrets.GITHUB_TOKEN }}" | docker login ghcr.io -u "${{ github.actor }}" --password-stdin
 
       - name: Pull cluster image
-        run: docker pull ghcr.io/nvidia/openshell/cluster:${{ inputs.image-tag }}
+        run: docker pull ghcr.io/nvidia/openshell/cluster:${{ inputs.image-tag || github.sha }}
 
       - name: Install Python dependencies and generate protobuf stubs
-        if: matrix.suite == 'python'
+        if: matrix.suite == 'python' || matrix.suite == 'python-gpu'
         run: uv sync --frozen && mise run --no-prepare python:proto
 
       - name: Build Rust CLI
-        if: matrix.suite != 'python'
+        if: matrix.suite == 'rust' || matrix.suite == 'gateway-resume'
         run: cargo build -p openshell-cli --features openshell-core/dev-settings
 
       - name: Install SSH client
-        if: matrix.suite != 'python'
+        if: matrix.suite == 'rust' || matrix.suite == 'gateway-resume'
         run: apt-get update && apt-get install -y --no-install-recommends openssh-client && rm -rf /var/lib/apt/lists/*
 
       - name: Bootstrap cluster
         env:
           GATEWAY_HOST: host.docker.internal
           GATEWAY_PORT: ${{ matrix.port }}
           CLUSTER_NAME: ${{ matrix.cluster }}
+          CLUSTER_GPU: ${{ matrix.cluster_gpu }}
           SKIP_IMAGE_PUSH: "1"
           SKIP_CLUSTER_IMAGE_BUILD: "1"
-          OPENSHELL_CLUSTER_IMAGE: ghcr.io/nvidia/openshell/cluster:${{ inputs.image-tag }}
+          OPENSHELL_CLUSTER_IMAGE: ghcr.io/nvidia/openshell/cluster:${{ inputs.image-tag || github.sha }}
         run: mise run --no-prepare --skip-deps cluster
 
       - name: Run tests