build_ai_layer.yml

name: Build AI Layer

on:
  push:
    paths:
      - 'model/**'
  workflow_dispatch:

permissions:
  contents: write
  packages: write
  id-token: write
  attestations: write

env:
  REGISTRY: ghcr.io
  IMAGE_NAME: ${{ github.repository }}/executorch-arm-container

jobs:
  check-docker-image:
    runs-on: ubuntu-latest
    outputs:
      image-exists: ${{ steps.check.outputs.exists }}
      repo-slug: ${{ steps.slug.outputs.repo_slug }}
    steps:
      - name: Derive lowercase repo slug
        id: slug
        run: echo "repo_slug=$(echo '${{ github.repository }}' | tr '[:upper:]' '[:lower:]')" >> $GITHUB_OUTPUT
      - name: Check if Docker image exists
        id: check
        run: |
          if docker manifest inspect ghcr.io/$(echo '${{ github.repository }}' | tr '[:upper:]' '[:lower:]')/executorch-arm-container:latest > /dev/null 2>&1; then
            echo "exists=true" >> $GITHUB_OUTPUT
          else
            echo "exists=false" >> $GITHUB_OUTPUT
          fi

  build-docker-if-needed:
    needs: check-docker-image
    if: needs.check-docker-image.outputs.image-exists == 'false'
    uses: ./.github/workflows/build_docker.yml
    permissions:
      contents: read      # must be >= permissions requested by called workflow
      packages: write
      id-token: write
      # attestations: write  # uncomment if attestation step in called workflow is enabled
    secrets: inherit

  build-ai-layer:
    needs: [check-docker-image, build-docker-if-needed]
    if: always() && (needs.check-docker-image.outputs.image-exists == 'true' || needs.build-docker-if-needed.result == 'success')
    runs-on: ubuntu-latest
    container:
      image: ghcr.io/${{ needs.check-docker-image.outputs.repo-slug }}/executorch-arm-container:latest
      credentials:
        username: ${{ github.actor }}
        password: ${{ secrets.GITHUB_TOKEN }}
    
    env:
      GIT_AUTHOR_EMAIL: "action@github.com"
      GIT_AUTHOR_NAME: "GitHub Action"
      GIT_COMMITTER_EMAIL: "action@github.com"
      GIT_COMMITTER_NAME: "GitHub Action"

    steps:
    - name: Export lowercase repo slug env
      run: echo "REPO_SLUG=${{ needs.check-docker-image.outputs.repo-slug }}" >> $GITHUB_ENV
    - name: Checkout repository
      uses: actions/checkout@v7
      with:
        fetch-depth: 0

    - name: Setup git config and logging
      run: |
        git config --global --add safe.directory /github/workspace
        # Setup logging directory
        mkdir -p $GITHUB_WORKSPACE/ai_layer/logs
        export BUILD_TIMESTAMP=$(date '+%Y%m%d_%H%M%S')
        echo "BUILD_TIMESTAMP=$BUILD_TIMESTAMP" >> $GITHUB_ENV
        echo "LOG_DIR=$GITHUB_WORKSPACE/ai_layer/logs" >> $GITHUB_ENV
        echo "MAIN_LOG=$GITHUB_WORKSPACE/ai_layer/logs/build.log" >> $GITHUB_ENV
        
        # Log build start
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] ExecuTorch AI Layer Build Started (GitHub Actions)" | tee $GITHUB_WORKSPACE/ai_layer/logs/build.log
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Timestamp: $BUILD_TIMESTAMP" | tee -a $GITHUB_WORKSPACE/ai_layer/logs/build.log
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Workflow: ${{ github.workflow }}" | tee -a $GITHUB_WORKSPACE/ai_layer/logs/build.log
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Run ID: ${{ github.run_id }}" | tee -a $GITHUB_WORKSPACE/ai_layer/logs/build.log

    - name: Convert and build model
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Starting: Model conversion" | tee -a $MAIN_LOG
        cd $GITHUB_WORKSPACE/model && python3 aot_model.py 2>&1 | tee $LOG_DIR/model_conversion.log $MAIN_LOG
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Completed: Model conversion" | tee -a $MAIN_LOG 

    - name: Build ExecuTorch Core Libraries
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Starting: Stage1 build" | tee -a $MAIN_LOG
        $GITHUB_WORKSPACE/scripts/build_stage1.sh /workspace/executorch \
                               $GITHUB_WORKSPACE/out/stage1  \
                               $GITHUB_WORKSPACE/model/arm-none-eabi-gcc.cmake 2>&1 | tee $LOG_DIR/stage1_build.log $MAIN_LOG
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Completed: Stage1 build" | tee -a $MAIN_LOG

    - name: Build ExecuTorch Selective Kernel Libraries
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Starting: Stage2 build" | tee -a $MAIN_LOG
        $GITHUB_WORKSPACE/scripts/build_stage2_selective.sh \
                               /workspace/executorch \
                               "" \
                               $GITHUB_WORKSPACE/out/stage2 \
                               $GITHUB_WORKSPACE/model/arm-none-eabi-gcc.cmake \
                               $GITHUB_WORKSPACE/model/operators_minimal.txt 2>&1 | tee $LOG_DIR/stage2_build.log $MAIN_LOG
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Completed: Stage2 build" | tee -a $MAIN_LOG

    - name: Collect artifacts
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Starting: Package artifacts" | tee -a $MAIN_LOG
        cd "$GITHUB_WORKSPACE" && ./scripts/package_sdk.sh \
          "$GITHUB_WORKSPACE/out/stage1/assets" \
          "$GITHUB_WORKSPACE/out/stage2/assets" \
          "$GITHUB_WORKSPACE/model/ethos_u_minimal_example.pte" \
          "$GITHUB_WORKSPACE/ai_layer/engine" 2>&1 | tee $LOG_DIR/package_artifacts.log $MAIN_LOG
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Completed: Package artifacts" | tee -a $MAIN_LOG

    - name: Convert the model to a header file
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Starting: Model to header conversion" | tee -a $MAIN_LOG
        cd "$GITHUB_WORKSPACE" && python3 scripts/pte_to_header.py -p model/ethos_u_minimal_example.pte -d ai_layer/model -o model_pte.h 2>&1 | tee $LOG_DIR/model_to_header.log $MAIN_LOG
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Completed: Model to header conversion" | tee -a $MAIN_LOG

    - name: Generate AI layer report
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Starting: Generate report" | tee -a $MAIN_LOG
        python3 scripts/generate_ai_layer_report.py 2>&1 | tee $LOG_DIR/generate_report.log $MAIN_LOG || { echo "$(date '+%Y-%m-%d %H:%M:%S') [ERROR] Report generation failed" | tee -a $MAIN_LOG; exit 1; }
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Completed: Generate report" | tee -a $MAIN_LOG

    - name: Show report summary and finalize logs
      run: |
        echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] Build summary:" | tee -a $MAIN_LOG
        head -n 50 ai_layer/REPORT.md | tee -a $MAIN_LOG || echo "Report could not be displayed" | tee -a $MAIN_LOG
        
        # Create build summary
        cat > $LOG_DIR/build_summary.txt << EOF
        ExecuTorch AI Layer Build Summary (GitHub Actions)
        ==================================================
        Build Timestamp: $BUILD_TIMESTAMP
        Workflow: ${{ github.workflow }}
        Run ID: ${{ github.run_id }}
        Commit: ${{ github.sha }}
        Branch: ${{ github.ref }}
        Build Status: SUCCESS
        
        Log Files Generated:
        - Main build log: ai_layer/logs/build.log
        - Model conversion: ai_layer/logs/model_conversion.log
        - Stage1 build: ai_layer/logs/stage1_build.log  
        - Stage2 build: ai_layer/logs/stage2_build.log
        - Package artifacts: ai_layer/logs/package_artifacts.log
        - Model to header: ai_layer/logs/model_to_header.log
        - Generate report: ai_layer/logs/generate_report.log
        
        Outputs Generated:
        - AI Layer libraries: ai_layer/engine/lib/
        - AI Layer headers: ai_layer/engine/include/
        - Model header: ai_layer/model/model_pte.h
        - Build report: ai_layer/REPORT.md
        EOF
        
        echo "$(date '+%Y-%m-%d %H:%M:%S') [SUCCESS] Build completed successfully!" | tee -a $MAIN_LOG
        echo "📋 Build logs saved to: ai_layer/logs/" | tee -a $MAIN_LOG

    - name: Commit changes including logs
      run: |
        git add ai_layer/
        git diff --staged --quiet || git commit -m "Update AI layer with new model build - Build $BUILD_TIMESTAMP

        Build artifacts and logs included:
        - Libraries: ai_layer/engine/lib/
        - Headers: ai_layer/engine/include/  
        - Model: ai_layer/model/
        - Report: ai_layer/REPORT.md
        - Build logs: ai_layer/logs/

        Workflow run: ${{ github.run_id }}
        Commit: ${{ github.sha }}"
        git push