inference.yml

name: Inference Tests

on:
  pull_request_review:
    types: [submitted]
  workflow_dispatch:

permissions:
  contents: read
  actions: read

concurrency:
  group: inference-${{ github.event.pull_request.number || github.ref }}
  cancel-in-progress: true

jobs:
  gate:
    runs-on: ubuntu-latest
    timeout-minutes: 5
    outputs:
      approved: ${{ steps.check.outputs.approved }}
    steps:
      - id: check
        run: |
          if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
            echo "approved=true" >> $GITHUB_OUTPUT
          elif [ "${{ github.event.review.state }}" = "approved" ]; then
            echo "approved=true" >> $GITHUB_OUTPUT
          else
            echo "approved=false" >> $GITHUB_OUTPUT
          fi

  runner-check:
    needs: gate
    if: needs.gate.outputs.approved == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 5
    outputs:
      available: ${{ steps.check.outputs.available }}
    steps:
      - id: check
        env:
          GH_TOKEN: ${{ github.token }}
          REPO: ${{ github.repository }}
        run: |
          set -euo pipefail
          status=$(curl -sS -L -w "%{http_code}" -o runners.json \
            -H "Authorization: Bearer ${GH_TOKEN}" \
            -H "Accept: application/vnd.github+json" \
            -H "X-GitHub-Api-Version: 2022-11-28" \
            "https://api.github.com/repos/${REPO}/actions/runners?per_page=100" || true)
          if [ "$status" != "200" ]; then
            echo "available=false" >> "$GITHUB_OUTPUT"
            echo "::notice::Could not inspect self-hosted runners (HTTP $status); skipping inference tests instead of queueing."
            exit 0
          fi
          count=$(jq '[.runners[] | select(.status == "online" and .busy == false) | select(([.labels[].name] | contains(["self-hosted", "linux", "arm64"])))] | length' runners.json)
          if [ "$count" -gt 0 ]; then
            echo "available=true" >> "$GITHUB_OUTPUT"
          else
            echo "available=false" >> "$GITHUB_OUTPUT"
            echo "::notice::No idle self-hosted linux arm64 runner is available; skipping inference tests."
          fi

  build:
    needs: [gate, runner-check]
    if: needs.gate.outputs.approved == 'true' && needs.runner-check.outputs.available == 'true'
    runs-on: [self-hosted, linux, arm64]
    timeout-minutes: 60
    steps:
      - uses: actions/checkout@v4
        with:
          clean: false

      - name: Setup Python environment
        run: |
          if [ ! -f venv/bin/activate ]; then
            source ./setup
          fi

      - name: Build libcactus + engine tests
        run: |
          bash cactus-engine/build.sh
          cd cactus-engine/tests
          rm -rf build && mkdir -p build && cd build
          cmake .. -DCMAKE_BUILD_TYPE=Release
          make -j$(nproc)

  model-tests:
    needs: build
    runs-on: [self-hosted, linux, arm64]
    timeout-minutes: 60
    env:
      CACTUS_NO_CLOUD_TELE: "1"
    strategy:
      fail-fast: false
      matrix:
        model:
          - LiquidAI/LFM2-VL-450M
          - openai/whisper-small
    steps:
      - uses: actions/checkout@v4
        with:
          clean: false

      - name: Run tests for ${{ matrix.model }}
        env:
          HF_TOKEN: ${{ secrets.HF_TOKEN }}
        run: |
          source venv/bin/activate
          MODEL="${{ github.event.inputs.model || matrix.model }}"
          python3 -m cactus test --component engine --model "$MODEL"