M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

name: Build llama.cpp

on:
  workflow_dispatch:
    inputs:
      version:
        description: 'llama.cpp version (tag or branch)'
        required: false
        default: 'b4535'

jobs:
  build:
    strategy:
      fail-fast: false
      matrix:
        include:
          - os: ubuntu-latest
            artifact: libllama_linux_amd64.a
            platform: linux-amd64
          - os: macos-14
            artifact: libllama_darwin_arm64.a
            platform: darwin-arm64
          - os: macos-13
            artifact: libllama_darwin_amd64.a
            platform: darwin-amd64

    runs-on: ${{ matrix.os }}
    
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Install dependencies (Ubuntu)
        if: runner.os == 'Linux'
        run: |
          sudo apt-get update
          sudo apt-get install -y cmake build-essential

      - name: Build llama.cpp
        run: |
          chmod +x ./nornicdb/scripts/build-llama.sh
          cd nornicdb && ./scripts/build-llama.sh ${{ github.event.inputs.version }}

      - name: Upload artifact
        uses: actions/upload-artifact@v4
        with:
          name: ${{ matrix.platform }}
          path: |
            nornicdb/lib/llama/${{ matrix.artifact }}
            nornicdb/lib/llama/llama.h
            nornicdb/lib/llama/ggml.h
            nornicdb/lib/llama/VERSION
          retention-days: 30

  # CUDA build requires self-hosted runner with NVIDIA GPU
  build-cuda:
    runs-on: ubuntu-latest
    if: false  # Disabled by default - enable with self-hosted runner
    
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Setup CUDA
        uses: Jimver/cuda-toolkit@v0.2.14
        with:
          cuda: '12.2.0'

      - name: Build llama.cpp with CUDA
        run: |
          chmod +x ./nornicdb/scripts/build-llama.sh
          cd nornicdb && ./scripts/build-llama.sh ${{ github.event.inputs.version }}

      - name: Upload CUDA artifact
        uses: actions/upload-artifact@v4
        with:
          name: linux-amd64-cuda
          path: |
            nornicdb/lib/llama/libllama_linux_amd64_cuda.a
            nornicdb/lib/llama/llama.h
            nornicdb/lib/llama/ggml.h
            nornicdb/lib/llama/VERSION
          retention-days: 30

  release:
    needs: build
    runs-on: ubuntu-latest
    if: github.event_name == 'workflow_dispatch'
    
    steps:
      - name: Download all artifacts
        uses: actions/download-artifact@v4
        with:
          path: artifacts

      - name: Create combined archive
        run: |
          mkdir -p release/lib/llama
          cp artifacts/linux-amd64/* release/lib/llama/
          cp artifacts/darwin-arm64/*.a release/lib/llama/
          cp artifacts/darwin-amd64/*.a release/lib/llama/
          cd release && tar -czvf ../llama-libs-${{ github.event.inputs.version }}.tar.gz lib

      - name: Upload combined artifact
        uses: actions/upload-artifact@v4
        with:
          name: llama-libs-all-platforms
          path: llama-libs-${{ github.event.inputs.version }}.tar.gz
          retention-days: 90