vllm-omni.pipeline.yml

name: "Pipeline - vLLM-Omni"

on:
  workflow_call:
    inputs:
      config-file:
        description: "Path to image config YAML"
        required: true
        type: string
      tag-suffix:
        description: "CI tag suffix (run_id or pr-NUMBER)"
        required: true
        type: string
      build:
        description: "Whether to build the image (false = skip build, test against prod)"
        required: false
        type: boolean
        default: true
      run-sanity-test:
        description: "Run sanity tests"
        required: false
        type: boolean
        default: true
      run-security-test:
        description: "Run security tests"
        required: false
        type: boolean
        default: false
      run-telemetry-test:
        description: "Run telemetry tests"
        required: false
        type: boolean
        default: true
      run-model-test:
        description: "Run model tests"
        required: false
        type: boolean
        default: true
      run-sagemaker-test:
        description: "Run SageMaker endpoint tests"
        required: false
        type: boolean
        default: true
      release:
        description: "Run release job after tests pass"
        required: false
        type: boolean
        default: false
    outputs:
      image-uri:
        description: "Built image URI (empty if build was skipped)"
        value: ${{ jobs.build.outputs.image-uri }}

jobs:
  ci-config:
    runs-on: ubuntu-latest
    outputs:
      customer-type: ${{ steps.parse.outputs.customer-type }}
    steps:
      - uses: actions/checkout@v6
      - id: parse
        run: |
          echo "customer-type=$(yq '.metadata.customer_type' '${{ inputs.config-file }}')" >> $GITHUB_OUTPUT

  build:
    if: ${{ inputs.build }}
    concurrency:
      group: ${{ github.workflow }}-build-${{ inputs.config-file }}-${{ github.ref }}
      cancel-in-progress: true
    runs-on:
      - codebuild-runner-${{ github.run_id }}-${{ github.run_attempt }}
        fleet:x86-vllm-build-runner
        buildspec-override:true
    timeout-minutes: 720
    env:
      WHEELS_BUCKET: ${{ vars.WHEELS_BUCKET }}
    outputs:
      image-uri: ${{ steps.build.outputs.image-uri }}
    steps:
      - uses: actions/checkout@v6
      - id: build
        uses: ./.github/actions/build-image
        with:
          config-file: ${{ inputs.config-file }}
          aws-account-id: ${{ vars.CI_AWS_ACCOUNT_ID }}
          aws-region: ${{ vars.AWS_REGION }}
          tag-suffix: ${{ inputs.tag-suffix }}

  sanity-test:
    if: ${{ always() && !failure() && !cancelled() && inputs.run-sanity-test }}
    needs: [build]
    concurrency:
      group: ${{ github.workflow }}-sanity-test-${{ inputs.config-file }}-${{ github.ref }}
      cancel-in-progress: true
    uses: ./.github/workflows/_reusable.sanity-tests.yml
    with:
      config-file: ${{ inputs.config-file }}
      image-uri: ${{ needs.build.outputs.image-uri || '' }}

  security-test:
    if: ${{ always() && !failure() && !cancelled() && inputs.run-security-test }}
    needs: [build]
    concurrency:
      group: ${{ github.workflow }}-security-test-${{ inputs.config-file }}-${{ github.ref }}
      cancel-in-progress: true
    uses: ./.github/workflows/_reusable.security-tests.yml
    with:
      config-file: ${{ inputs.config-file }}
      image-uri: ${{ needs.build.outputs.image-uri || '' }}

  telemetry-test:
    if: ${{ always() && !failure() && !cancelled() && inputs.run-telemetry-test }}
    needs: [build]
    concurrency:
      group: ${{ github.workflow }}-telemetry-test-${{ inputs.config-file }}-${{ github.ref }}
      cancel-in-progress: false
    uses: ./.github/workflows/_reusable.telemetry-tests.yml
    with:
      config-file: ${{ inputs.config-file }}
      image-uri: ${{ needs.build.outputs.image-uri || '' }}

  model-tests:
    if: ${{ always() && !failure() && !cancelled() && inputs.run-model-test }}
    needs: [build]
    concurrency:
      group: ${{ github.workflow }}-model-tests-${{ inputs.config-file }}-${{ github.ref }}
      cancel-in-progress: true
    uses: ./.github/workflows/vllm-omni.tests-model.yml
    with:
      config-file: ${{ inputs.config-file }}
      image-uri: ${{ needs.build.outputs.image-uri || '' }}
    secrets: inherit

  sagemaker-tests:
    if: ${{ always() && !failure() && !cancelled() && inputs.run-sagemaker-test && needs.ci-config.outputs.customer-type == 'sagemaker' }}
    needs: [build, ci-config]
    concurrency:
      group: ${{ github.workflow }}-sagemaker-tests-${{ inputs.config-file }}-${{ github.ref }}
      cancel-in-progress: false
    uses: ./.github/workflows/vllm-omni.tests-sagemaker.yml
    with:
      config-file: ${{ inputs.config-file }}
      image-uri: ${{ needs.build.outputs.image-uri || '' }}

  release-gate:
    if: >-
      github.ref == 'refs/heads/main' &&
      (contains(github.workflow_ref, '.autorelease') || contains(github.workflow_ref, '.dispatch-release')) &&
      !contains(github.workflow_ref, '.pr') &&
      inputs.release &&
      !failure() && !cancelled() &&
      needs.build.result == 'success'
    needs: [build, sanity-test, security-test, telemetry-test, model-tests, sagemaker-tests]
    runs-on: ubuntu-latest
    outputs:
      should-release: ${{ steps.check.outputs.should-release }}
      release-spec: ${{ steps.spec.outputs.release-spec }}
      environment: ${{ steps.check.outputs.environment }}
    steps:
      - uses: actions/checkout@v6
      - id: check
        run: |
          SHOULD_RELEASE=$(yq '.release.release' "${{ inputs.config-file }}")
          ENVIRONMENT=$(yq '.release.environment' "${{ inputs.config-file }}")
          echo "should-release=${SHOULD_RELEASE}" >> $GITHUB_OUTPUT
          echo "environment=${ENVIRONMENT}" >> $GITHUB_OUTPUT
      - id: spec
        if: steps.check.outputs.should-release == 'true'
        uses: ./.github/actions/generate-release-spec
        with:
          config-file: ${{ inputs.config-file }}

  release:
    if: ${{ !failure() && !cancelled() && needs.release-gate.outputs.should-release == 'true' }}
    needs: [build, release-gate]
    uses: ./.github/workflows/_reusable.release-image.yml
    with:
      source-image-uri: ${{ needs.build.outputs.image-uri }}
      release-spec: ${{ needs.release-gate.outputs.release-spec }}
      environment: ${{ needs.release-gate.outputs.environment }}
      aws-region: ${{ vars.AWS_REGION }}
    secrets: inherit

  notify-failure:
    # Notify Slack immediately when an auto-release/dispatch-release run fails.
    # Guarded on release context so PR runs never notify.
    if: >-
      failure() &&
      github.ref == 'refs/heads/main' &&
      (contains(github.workflow_ref, '.autorelease') || contains(github.workflow_ref, '.dispatch-release')) &&
      !contains(github.workflow_ref, '.pr') &&
      inputs.release
    needs: [ci-config, build, sanity-test, security-test, telemetry-test, model-tests, sagemaker-tests, release-gate, release]
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v6
      - uses: ./.github/actions/notify-slack-failure
        with:
          webhook-url: ${{ secrets.SLACK_RELEASE_FAILURE_WEBHOOK_URL }}
          workflow-name: ${{ github.workflow }}
          config-file: ${{ inputs.config-file }}
          run-url: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}