# SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0

name: PR

on:
  push:
    branches:
      - "pull-request/[0-9]+"

concurrency:
  # The group name is the ref_name, so that workflows on the same PR/branch have the same group name for cancelling.
  group: docker-build-test-${{ github.ref_name }}
  cancel-in-progress: true

env:
  BUILDER_NAME: b-${{ github.run_id }}-${{ github.run_attempt }}

jobs:
  # ============================================================================
  # SETUP & DETECTION JOBS
  # ============================================================================
  changed-files:
    runs-on: ubuntu-latest
    outputs:
      core: ${{ steps.changes.outputs.core }}
      planner: ${{ steps.changes.outputs.planner }}
      operator: ${{ steps.changes.outputs.operator }}
      deploy: ${{ steps.changes.outputs.deploy }}
      vllm: ${{ steps.changes.outputs.vllm }}
      sglang: ${{ steps.changes.outputs.sglang }}
      trtllm: ${{ steps.changes.outputs.trtllm }}
      builder_name: ${{ steps.export-builder-name.outputs.builder_name }}
    steps:
      - name: Checkout code
        uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955  # v4.3.0
        with:
          fetch-depth: 0
      - name: Check for changes
        id: changes
        uses: ./.github/actions/changed-files
        with:
          gh_token: ${{ github.token }}
      - name: Export builder name
        id: export-builder-name
        run: |
          echo "builder_name=${{ env.BUILDER_NAME }}" >> $GITHUB_OUTPUT

  backend-status-check:
    runs-on: ubuntu-latest
    needs:
      - changed-files
      - operator
      - vllm-build
      - vllm-dev-build
      - vllm-test
      - vllm-multi-gpu-test
      - vllm-copy-to-acr
      - sglang-build
      - sglang-dev-build
      - sglang-test
      - sglang-multi-gpu-test
      - sglang-copy-to-acr
      - trtllm-build
      - trtllm-dev-build
      - trtllm-test
      - trtllm-multi-gpu-test
      - trtllm-copy-to-acr
      - planner-build
      - planner-test
    if: always()
    steps:
      - name: Check all dependent jobs
        run: |
          echo '${{ toJson(needs) }}' | jq -e 'to_entries | map(.value.result) | all(. as $result | ["success", "skipped"] | any($result == .))'

  deploy-status-check:
    runs-on: ubuntu-latest
    needs: [deploy-operator, deploy-test-vllm, deploy-test-sglang, deploy-test-trtllm]
    if: always()
    steps:
      - name: Check all deploy test jobs
        run: |
          echo '${{ toJson(needs) }}' | jq -e 'to_entries | map(.value.result) | all(. as $result | ["success", "skipped", "cancelled"] | any($result == .))'

  operator:
    needs: changed-files
    if: needs.changed-files.outputs.operator == 'true'
    name: Operator
    runs-on: prod-default-v2
    env:
      IMAGE_REGISTRY: ai-dynamo
      IMAGE_REPOSITORY: dynamo
      ECR_HOSTNAME: ${{ secrets.AWS_ACCOUNT_ID }}.dkr.ecr.${{ secrets.AWS_DEFAULT_REGION }}.amazonaws.com
    outputs:
      operator_default_tag: ${{ steps.build-and-push-image.outputs.operator_default_tag }}
    steps:
      - name: Checkout code
        uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955  # v4.3.0
      - name: Initialize Dynamo Builder
        uses: ./.github/actions/init-dynamo-builder
        with:
          builder_name: ${{ needs.changed-files.outputs.builder_name }}
          flavor: general
          arch: 'linux/amd64,linux/arm64'
      - name: Docker Login
        uses: ./.github/actions/docker-login
        with:
          aws_default_region: ${{ secrets.AWS_DEFAULT_REGION }}
          aws_account_id: ${{ secrets.AWS_ACCOUNT_ID }}
          azure_acr_hostname: ${{ secrets.AZURE_ACR_HOSTNAME }}
          azure_acr_user: ${{ secrets.AZURE_ACR_USER }}
          azure_acr_password: ${{ secrets.AZURE_ACR_PASSWORD }}
      - name: Linter
        shell: bash
        working-directory: ./deploy/operator
        run: |
          docker buildx build --platform linux/arm64 --target linter --progress=plain --build-arg DOCKER_PROXY=${ECR_HOSTNAME}/dockerhub/ --build-context snapshot=../snapshot .
      - name: Tester
        shell: bash
        working-directory: ./deploy/operator
        run: |
          docker buildx build --platform linux/arm64 --target tester --progress=plain --build-arg DOCKER_PROXY=${ECR_HOSTNAME}/dockerhub/ --build-context snapshot=../snapshot .
      - name: Set up Go
        uses: actions/setup-go@44694675825211faa026b3c33043df3e48a5fa00 # v6.0.0
        with:
          go-version: '1.25'
      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: "3.11"
      - name: Install Python dependencies for operator codegen
        shell: bash
        working-directory: ./deploy/operator
        run: |
          python -m pip install --upgrade pip
          python -m pip install "pydantic>=2,<3" "black==23.1.0" "pyyaml>=6.0"
      - name: Check for uncommitted changes
        shell: bash
        working-directory: ./deploy/operator
        run: |
          make check
      - name: Build and push Container
        id: build-and-push-image
        shell: bash
        working-directory: ./deploy/operator
        env:
          NO_CACHE_FLAG: '' # placeholder for future logic to add no cache flag if needed
        run: |
          ECR_DEFAULT_IMAGE_BASE="${ECR_HOSTNAME}/${IMAGE_REGISTRY}/${IMAGE_REPOSITORY}"
          DEFAULT_TAG="${{ github.sha }}-operator"
          ACR_IMAGE_BASE="${{ secrets.AZURE_ACR_HOSTNAME }}/${IMAGE_REGISTRY}/${IMAGE_REPOSITORY}"
          IMAGE_URIS=(
            "${ECR_DEFAULT_IMAGE_BASE}:${DEFAULT_TAG}"
            "${ACR_IMAGE_BASE}:${DEFAULT_TAG}"
          )

          echo "operator_default_tag=${DEFAULT_TAG}" >> $GITHUB_OUTPUT
          TAGGING_FLAGS=$(printf -- '-t %s ' "${IMAGE_URIS[@]}")
          echo "flags for docker buildx: ${TAGGING_FLAGS}"

          if [[ "$NO_CACHE_FLAG" == "true" ]]; then
            NO_CACHE_FLAG="--no-cache"
          fi
          docker buildx build --push ${NO_CACHE_FLAG} \
              --platform linux/amd64,linux/arm64 \
              --build-arg DOCKER_PROXY=${ECR_HOSTNAME}/dockerhub/ \
              --build-context snapshot=../snapshot \
              ${TAGGING_FLAGS} -f Dockerfile .

          echo "### 🐳 Operator Container Images" >> $GITHUB_STEP_SUMMARY
          echo "" >> $GITHUB_STEP_SUMMARY
          echo "| Image URI |" >> $GITHUB_STEP_SUMMARY
          echo "|-----|" >> $GITHUB_STEP_SUMMARY
          for image_uri in "${IMAGE_URIS[@]}"; do
            echo "| \`${image_uri}\` |" >> $GITHUB_STEP_SUMMARY
          done

  # ============================================================================
  # BUILD PIPELINES
  # ============================================================================

  vllm-build:
    name: vllm-runtime # This name overlaps with other vllm jobs to group them in the UI
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: vllm
      target: runtime
      cuda_version: '["12.9", "13.0"]'
      platform: 'linux/amd64,linux/arm64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      build_timeout_minutes: 60
    secrets: inherit

  vllm-dev-build:
    name: vllm-dev
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: vllm
      target: dev
      cuda_version: '["12.9", "13.0"]'
      platform: 'linux/amd64,linux/arm64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      push_image: false
      build_timeout_minutes: 60
    secrets: inherit

  sglang-build:
    name: sglang-runtime # This name overlaps with other sglang jobs to group them in the UI
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.sglang == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: sglang
      target: runtime
      cuda_version: '["12.9", "13.0"]'
      platform: 'linux/amd64,linux/arm64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      build_timeout_minutes: 60
    secrets: inherit

  sglang-dev-build:
    name: sglang-dev
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.sglang == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: sglang
      target: dev
      cuda_version: '["12.9", "13.0"]'
      platform: 'linux/amd64,linux/arm64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      push_image: false
      build_timeout_minutes: 60
    secrets: inherit

  trtllm-build:
    name: trtllm-runtime # This name overlaps with other trtllm jobs to group them in the UI
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.trtllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: trtllm
      target: runtime
      cuda_version: '["13.1"]'
      platform: 'linux/amd64,linux/arm64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      build_timeout_minutes: 60
    secrets: inherit

  trtllm-dev-build:
    name: trtllm-dev
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.trtllm == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: trtllm
      target: dev
      cuda_version: '["13.1"]'
      platform: 'linux/amd64,linux/arm64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      push_image: false
      build_timeout_minutes: 60
    secrets: inherit

  planner-build:
    name: planner # This name overlaps with other planner jobs to group them in the UI
    needs: [changed-files]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.planner == 'true'
    uses: ./.github/workflows/shared-build-image.yml
    with:
      framework: dynamo
      target: planner
      cuda_version: '[""]'
      platform: 'linux/amd64'
      builder_name: ${{ needs.changed-files.outputs.builder_name }}
      build_timeout_minutes: 45
    secrets: inherit

  # ============================================================================
  # TEST PIPELINES
  # ============================================================================

  vllm-test:
    name: vllm-runtime # This name overlaps with other vllm jobs to group them in the UI
    needs: [changed-files, vllm-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: vllm
      test_type: Test
      amd_runner: prod-tester-amd-gpu-v1 # This runner is overridden for ARM platform
      target_tag_plain: ${{ needs.vllm-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      platform: '["amd64", "arm64"]' # arm64 for CPU tests, single GPU tests are skipped
      run_cpu_only_tests: true
      cpu_only_test_markers: pre_merge and vllm and gpu_0
      gpu_test_markers: pre_merge and vllm and gpu_1
      gpu_test_timeout_minutes: 35
    secrets: inherit

  vllm-multi-gpu-test:
    name: vllm-runtime # This name overlaps with other vllm jobs to group them in the UI
    needs: [changed-files, vllm-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: vllm
      test_type: Multi-GPU Test
      amd_runner: prod-tester-amd-gpu-4-v1
      target_tag_plain: ${{ needs.vllm-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      platform: '["amd64"]' # No ARM GPUs available
      run_sanity_check: false
      gpu_test_markers: pre_merge and vllm and (gpu_2 or gpu_4)
      gpu_test_timeout_minutes: 60
    secrets: inherit

  sglang-test:
    name: sglang-runtime # This name overlaps with other sglang jobs to group them in the UI
    needs: [changed-files, sglang-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.sglang == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: sglang
      test_type: Test
      amd_runner: prod-tester-amd-gpu-v1 # This runner is overridden for ARM platform
      target_tag_plain: ${{ needs.sglang-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      platform: '["amd64", "arm64"]' # arm64 for CPU tests, single GPU tests are skipped
      run_cpu_only_tests: true
      cpu_only_test_markers: pre_merge and sglang and gpu_0
      gpu_test_markers: pre_merge and sglang and gpu_1
    secrets: inherit

  sglang-multi-gpu-test:
    name: sglang-runtime # This name overlaps with other sglang jobs to group them in the UI
    needs: [changed-files, sglang-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.sglang == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: sglang
      test_type: Multi-GPU Test
      amd_runner: prod-tester-amd-gpu-4-v1
      target_tag_plain: ${{ needs.sglang-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      platform: '["amd64"]' # No ARM GPUs available
      run_sanity_check: false
      gpu_test_markers: pre_merge and sglang and (gpu_2 or gpu_4)
      gpu_test_timeout_minutes: 60
    secrets: inherit

  trtllm-test:
    name: trtllm-runtime # This name overlaps with other trtllm jobs to group them in the UI
    needs: [changed-files, trtllm-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.trtllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: trtllm
      test_type: Test
      amd_runner: prod-tester-amd-gpu-v1 # This runner is overridden for ARM platform
      target_tag_plain: ${{ needs.trtllm-build.outputs.target_tag_plain }}
      cuda_version: '["13.1"]'
      platform: '["amd64", "arm64"]' # arm64 for CPU tests, single GPU tests are skipped
      run_cpu_only_tests: true
      cpu_only_test_markers: pre_merge and trtllm and gpu_0
      gpu_test_markers: pre_merge and trtllm and gpu_1
    secrets: inherit

  trtllm-multi-gpu-test:
    name: trtllm-runtime # This name overlaps with other trtllm jobs to group them in the UI
    needs: [changed-files, trtllm-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.trtllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: trtllm
      test_type: Multi-GPU Test
      amd_runner: prod-tester-amd-gpu-4-v1
      target_tag_plain: ${{ needs.trtllm-build.outputs.target_tag_plain }}
      cuda_version: '["13.1"]'
      platform: '["amd64"]' # No ARM GPUs available
      run_sanity_check: false
      gpu_test_markers: pre_merge and trtllm and (gpu_2 or gpu_4)
      gpu_test_timeout_minutes: 60
    secrets: inherit

  planner-test:
    name: planner # This name overlaps with other planner jobs to group them in the UI
    needs: [changed-files, planner-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.planner == 'true'
    uses: ./.github/workflows/shared-test.yml
    with:
      test_suite_name: planner
      test_type: CPU Test
      amd_runner: prod-tester-amd-gpu-v1 # TODO: CPU only DinD runner for dynamo repo
      target_tag_plain: ${{ needs.planner-build.outputs.target_tag_plain }}
      cuda_version: '[""]'
      platform: '["amd64"]'
      run_sanity_check: false
      run_cpu_only_tests: true
      cpu_only_test_markers: 'pre_merge and planner and gpu_0'
      cpu_only_test_timeout_minutes: 30
      run_gpu_tests: false
    secrets: inherit

  # ============================================================================
  # IMAGE COMPLIANCE PIPELINES
  # ============================================================================

  vllm-compliance:
    name: vllm-runtime # This name overlaps with other vllm jobs to group them in the UI
    needs: [changed-files, vllm-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-compliance.yml
    with:
      framework: vllm
      target: runtime
      target_tag_plain: ${{ needs.vllm-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      platform: '["amd64"]'
    secrets: inherit

  sglang-compliance:
    name: sglang-runtime # This name overlaps with other sglang jobs to group them in the UI
    needs: [changed-files, sglang-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.sglang == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-compliance.yml
    with:
      framework: sglang
      target: runtime
      target_tag_plain: ${{ needs.sglang-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      platform: '["amd64"]'
    secrets: inherit

  trtllm-compliance:
    name: trtllm-runtime # This name overlaps with other trtllm jobs to group them in the UI
    needs: [changed-files, trtllm-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.trtllm == 'true' || needs.changed-files.outputs.deploy == 'true'
    uses: ./.github/workflows/shared-compliance.yml
    with:
      framework: trtllm
      target: runtime
      target_tag_plain: ${{ needs.trtllm-build.outputs.target_tag_plain }}
      cuda_version: '["13.1"]'
      platform: '["amd64"]'
    secrets: inherit

  planner-compliance:
    name: planner # This name overlaps with other planner jobs to group them in the UI
    needs: [changed-files, planner-build]
    if: needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.planner == 'true'
    uses: ./.github/workflows/shared-compliance.yml
    with:
      framework: dynamo
      target: planner
      target_tag_plain: ${{ needs.planner-build.outputs.target_tag_plain }}
      cuda_version: '[""]'
      platform: '["amd64"]'
    secrets: inherit

  # ============================================================================
  # IMAGE COPY PIPELINES
  # ============================================================================

  vllm-copy-to-acr:
    name: vllm-runtime # This name overlaps with other vllm jobs to group them in the UI
    needs: [changed-files, vllm-build, vllm-test]
    if: |
      always() && !cancelled() &&
      (needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true') &&
      needs.vllm-build.result == 'success' &&
      (needs.vllm-test.result == 'success' || needs.vllm-test.result == 'skipped')
    uses: ./.github/workflows/shared-copy.yml
    with:
      target_tag_plain: ${{ needs.vllm-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      override_arch: amd64 # We are using AMD64 images only on the rest of the clusters.
      copy_timeout_minutes: 10
    secrets: inherit

  sglang-copy-to-acr:
    name: sglang-runtime # This name overlaps with other sglang jobs to group them in the UI
    needs: [changed-files, sglang-build, sglang-test]
    if: |
      always() && !cancelled() &&
      (needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.sglang == 'true' || needs.changed-files.outputs.deploy == 'true') &&
      needs.sglang-build.result == 'success' &&
      (needs.sglang-test.result == 'success' || needs.sglang-test.result == 'skipped')
    uses: ./.github/workflows/shared-copy.yml
    with:
      target_tag_plain: ${{ needs.sglang-build.outputs.target_tag_plain }}
      cuda_version: '["12.9", "13.0"]'
      override_arch: amd64 # We are using AMD64 images only on the rest of the clusters.
      copy_timeout_minutes: 10
    secrets: inherit

  trtllm-copy-to-acr:
    name: trtllm-runtime # This name overlaps with other trtllm jobs to group them in the UI
    needs: [changed-files, trtllm-build, trtllm-test]
    if: |
      always() && !cancelled() &&
      (needs.changed-files.outputs.core == 'true' || needs.changed-files.outputs.trtllm == 'true' || needs.changed-files.outputs.deploy == 'true') &&
      needs.trtllm-build.result == 'success' &&
      (needs.trtllm-test.result == 'success' || needs.trtllm-test.result == 'skipped')
    uses: ./.github/workflows/shared-copy.yml
    with:
      target_tag_plain: ${{ needs.trtllm-build.outputs.target_tag_plain }}
      cuda_version: '["13.1"]'
      override_arch: amd64 # We are using AMD64 images only on the rest of the clusters.
      copy_timeout_minutes: 10
    secrets: inherit

  # ============================================================================
  # DEPLOY TEST PIPELINES
  # ============================================================================

  deploy-operator:
    if: |
      !cancelled() && !failure() &&
      (needs.changed-files.outputs.vllm == 'true' ||
       needs.changed-files.outputs.sglang == 'true' ||
       needs.changed-files.outputs.trtllm == 'true' ||
       needs.changed-files.outputs.deploy == 'true') &&
      (needs.operator.result == 'success' || needs.operator.result == 'skipped')
    needs: [changed-files, operator]
    runs-on: prod-default-small-v2
    outputs:
      namespace: ${{ steps.setup.outputs.namespace }}
      vcluster_name: ${{ steps.setup.outputs.vcluster_name }}
      operator_tag: ${{ steps.setup.outputs.operator_tag }}
    steps:
      - uses: actions/checkout@v4
      - name: Setup vCluster and operator
        id: setup
        uses: ./.github/actions/setup-dynamo-operator
        with:
          kubeconfig_base64: ${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}
          registry: ${{ secrets.AZURE_ACR_HOSTNAME }}
          operator_tag: ${{ needs.operator.result == 'success' && needs.operator.outputs.operator_default_tag || 'main-operator' }}
          hf_token: ${{ secrets.HF_TOKEN }}
          dockerhub_username: ${{ secrets.DOCKERHUB_LOGIN_USER }}
          dockerhub_password: ${{ secrets.DOCKERHUB_ACCESS_TOKEN }}

  deploy-test-vllm:
    name: vllm Deploy Test
    needs: [changed-files, deploy-operator, vllm-copy-to-acr, vllm-multi-gpu-test]
    if: |
      !cancelled() && !failure() &&
      (needs.changed-files.outputs.vllm == 'true' || needs.changed-files.outputs.deploy == 'true')
    uses: ./.github/workflows/shared-deploy-test.yml
    with:
      framework: vllm
      profiles: '["agg", "agg_router", "disagg", "disagg_router"]'
      image_suffix: vllm-runtime-cuda12
      namespace: ${{ needs.deploy-operator.outputs.namespace }}
      vcluster_name: ${{ needs.deploy-operator.outputs.vcluster_name }}
      operator_tag: ${{ needs.deploy-operator.outputs.operator_tag }}
    secrets: inherit

  deploy-test-sglang:
    name: sglang Deploy Test
    needs: [changed-files, deploy-operator, sglang-copy-to-acr, sglang-multi-gpu-test]
    if: |
      !cancelled() && !failure() &&
      (needs.changed-files.outputs.sglang == 'true' || needs.changed-files.outputs.deploy == 'true')
    uses: ./.github/workflows/shared-deploy-test.yml
    with:
      framework: sglang
      profiles: '["agg", "agg_router"]'
      image_suffix: sglang-runtime-cuda12
      namespace: ${{ needs.deploy-operator.outputs.namespace }}
      vcluster_name: ${{ needs.deploy-operator.outputs.vcluster_name }}
      operator_tag: ${{ needs.deploy-operator.outputs.operator_tag }}
    secrets: inherit

  deploy-test-trtllm:
    name: trtllm Deploy Test
    needs: [changed-files, deploy-operator, trtllm-copy-to-acr, trtllm-multi-gpu-test]
    if: |
      !cancelled() && !failure() &&
      (needs.changed-files.outputs.trtllm == 'true' || needs.changed-files.outputs.deploy == 'true')
    uses: ./.github/workflows/shared-deploy-test.yml
    with:
      framework: trtllm
      profiles: '["agg", "agg_router"]'
      image_suffix: trtllm-runtime-cuda13
      namespace: ${{ needs.deploy-operator.outputs.namespace }}
      vcluster_name: ${{ needs.deploy-operator.outputs.vcluster_name }}
      operator_tag: ${{ needs.deploy-operator.outputs.operator_tag }}
    secrets: inherit

  deploy-cleanup:
    if: always()
    needs: [deploy-operator, deploy-test-vllm, deploy-test-sglang, deploy-test-trtllm]
    runs-on: prod-default-small-v2
    steps:
      - uses: actions/checkout@v4
      - name: Teardown vCluster
        if: needs.deploy-operator.outputs.namespace != '' && needs.deploy-operator.outputs.vcluster_name != ''
        uses: ./.github/actions/teardown-dynamo-operator
        with:
          kubeconfig_base64: ${{ secrets.AZURE_AKS_CI_KUBECONFIG_B64 }}
          vcluster_name: ${{ needs.deploy-operator.outputs.vcluster_name }}
          vcluster_namespace: ${{ needs.deploy-operator.outputs.namespace }}

  clean-k8s-builder:
    name: Clean K8s builder if exists
    runs-on: prod-default-small-v2
    if: always()
    needs:
      - changed-files
      - operator
      - planner-test
      - vllm-copy-to-acr
      - vllm-multi-gpu-test
      - sglang-copy-to-acr
      - sglang-multi-gpu-test
      - trtllm-copy-to-acr
      - trtllm-multi-gpu-test
    steps:
      - name: Checkout repository
        uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955  # v4.3.0
      - name: Create K8s builders (skip bootstrap)
        uses: ./.github/actions/bootstrap-buildkit
        continue-on-error: true
        with:
          builder_name: ${{ needs.changed-files.outputs.builder_name }}
          buildkit_worker_addresses: ''
          skip_bootstrap: true
      - name: Builder Cleanup in case of k8s builder
        shell: bash
        run: |
          docker buildx rm ${{ needs.changed-files.outputs.builder_name }} || true

  # ============================================================================
  # ALLURE REPORT
  # Generate Allure test report and deploy to GitHub Pages
  # ============================================================================

  allure-report:
    needs:
      - changed-files
      - operator
      - vllm-build
      - vllm-test
      - vllm-multi-gpu-test
      - vllm-copy-to-acr
      - sglang-build
      - sglang-test
      - sglang-multi-gpu-test
      - sglang-copy-to-acr
      - trtllm-build
      - trtllm-test
      - trtllm-multi-gpu-test
      - trtllm-copy-to-acr
      - deploy-test-vllm
      - deploy-test-sglang
      - deploy-test-trtllm
    if: false
    # Disabled: gh-pages branch bloated to ~1GB after 72 commits of Allure reports
    # if: ${{ !cancelled() }}
    uses: ./.github/workflows/generate-allure-report.yml
    with:
      run_id: ${{ github.run_id }}
      subdir: pr
      workflow_label: PR
    permissions:
      contents: write
      actions: read