headroom/.github/workflows/ci.yml

name: CI

# Intelligent + parallel pipeline (cutover from the old 4-version matrix):
#   changes      — paths-filter; skips heavy work for docs-only changes
#   build-wheel  — compile the Rust ext ONCE (fast `ci` cargo profile), share via artifact
#   lint         — ruff + mypy, once
#   prefetch-model — download the embedding model ONCE (authenticated), warm shared cache
#   test         — 4 parallel shards (pytest-split), each a fresh runner VM; run offline
#   test-extras / test-agno / build / commitlint / workflow-validation / *-e2e — preserved
#
# Notes: CPU-only torch everywhere (no CUDA stack); test shards run HF_HUB_OFFLINE.
# Multi-version (3.10/3.11/3.13) coverage on main is a planned follow-up.

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]
  workflow_dispatch:

permissions:
  contents: read

concurrency:
  group: ci-${{ github.workflow }}-${{ github.ref }}
  # Cancel superseded runs on PRs/branches, but never cancel a main build.
  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}

env:
  PY_VERSION: "3.12"
  # CPU-only torch — runners have no GPU; the default CUDA wheels pull ~2.5 GB.
  PIP_EXTRA_INDEX_URL: https://download.pytorch.org/whl/cpu

jobs:
  changes:
    runs-on: ubuntu-latest
    timeout-minutes: 5
    outputs:
      code: ${{ steps.filter.outputs.code }}
      e2e: ${{ steps.filter.outputs.e2e }}
      workflows: ${{ steps.filter.outputs.workflows }}
    steps:
      - uses: actions/checkout@v6
      - uses: dorny/paths-filter@v4
        id: filter
        with:
          filters: |
            code:
              - 'headroom/**'
              - 'crates/**'
              - '**/*.rs'
              - 'pyproject.toml'
              - 'Cargo.toml'
              - 'Cargo.lock'
              - 'tests/**'
              - 'scripts/**'
              - '.github/workflows/ci.yml'
            e2e:
              - 'headroom/**'
              - 'crates/**'
              - 'docker/**'
              - 'Dockerfile'
              - 'e2e/**'
              - 'scripts/install*'
              - 'pyproject.toml'
            workflows:
              - '.github/workflows/**'

  lint:
    needs: changes
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}
      - name: Cache pip
        uses: actions/cache@v5
        with:
          path: ~/.cache/pip
          key: ${{ runner.os }}-pip-lint-${{ hashFiles('pyproject.toml') }}
          restore-keys: ${{ runner.os }}-pip-lint-
      - run: python -m pip install --upgrade pip ruff mypy
      - name: ruff check
        run: ruff check .
      - name: ruff format --check
        run: ruff format --check .
      - name: mypy
        run: mypy headroom --ignore-missing-imports

  build-wheel:
    needs: changes
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}
      - uses: dtolnay/rust-toolchain@1.96.0
      - uses: Swatinem/rust-cache@v2
        with:
          workspaces: ". -> target"
      - name: Build wheel once (fast CI cargo profile)
        run: |
          python -m pip install --upgrade pip maturin
          maturin build --profile ci --out dist --interpreter "python${PY_VERSION}"
      - uses: actions/upload-artifact@v7
        with:
          name: headroom-wheel
          path: dist/*.whl
          retention-days: 1

  prefetch-model:
    needs: changes
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    steps:
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}
      - name: Cache HuggingFace model
        id: hfcache
        uses: actions/cache@v5
        with:
          path: ~/.cache/huggingface
          key: ${{ runner.os }}-models-allMiniLM-v2
      - name: Fetch all-MiniLM-L6-v2 once (authenticated, resilient)
        if: steps.hfcache.outputs.cache-hit != 'true'
        env:
          HF_TOKEN: ${{ secrets.HF_TOKEN }}
          HF_HUB_DISABLE_TELEMETRY: "1"
        run: |
          python -m pip install --upgrade pip huggingface_hub
          for i in 1 2 3 4 5 6; do
            if python -c "from huggingface_hub import snapshot_download; snapshot_download('sentence-transformers/all-MiniLM-L6-v2')"; then exit 0; fi
            echo "::warning::model fetch attempt $i failed; backing off"; sleep $((i * 30))
          done
          echo "::error::could not fetch all-MiniLM-L6-v2 from HuggingFace"; exit 1

  test:
    needs: [changes, build-wheel, prefetch-model]
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 30
    strategy:
      fail-fast: false
      matrix:
        shard: [1, 2, 3, 4]
    env:
      HF_HUB_OFFLINE: "1"
      TRANSFORMERS_OFFLINE: "1"
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}

      - name: Cache pip
        uses: actions/cache@v5
        with:
          path: ~/.cache/pip
          key: ${{ runner.os }}-pip-${{ env.PY_VERSION }}-${{ hashFiles('pyproject.toml') }}
          restore-keys: ${{ runner.os }}-pip-${{ env.PY_VERSION }}-

      - name: Restore HuggingFace model cache (warmed by prefetch-model)
        uses: actions/cache@v5
        with:
          path: ~/.cache/huggingface
          key: ${{ runner.os }}-models-allMiniLM-v2

      - name: Download prebuilt wheel
        uses: actions/download-artifact@v8
        with:
          name: headroom-wheel
          path: dist

      - name: Install (CPU torch + prebuilt wheel + dev deps, no cargo rebuild)
        run: |
          python -m pip install --upgrade pip
          pip install torch --index-url https://download.pytorch.org/whl/cpu
          WHEEL="$(ls dist/*.whl)"
          pip install "${WHEEL}[dev]" pytest-split
          # cwd's ./headroom source tree shadows the installed wheel; copy the
          # compiled extension in so tests import it (no second cargo build).
          SITE="$(python -c 'import sysconfig; print(sysconfig.get_path("platlib"))')"
          cp "${SITE}/headroom/"_core*.so headroom/
          python -c "from headroom._core import DiffCompressor; print('headroom._core OK')"

      # Coverage upload: without this, codecov only receives reports from
      # the two native-e2e workflows (3 CLI test files total), so head
      # coverage reads ~6% and codecov/patch fails for ANY diff not
      # exercised by those files — a false negative on every PR. The main
      # suite runs here; its coverage must be what codecov sees.
      - name: Run test shard ${{ matrix.shard }}/4
        run: |
          pytest tests scripts/tests \
            --splits 4 --group ${{ matrix.shard }} \
            --cov=headroom --cov-branch \
            --cov-report=xml:coverage-${{ matrix.shard }}.xml \
            --cov-report= \
            --tb=short -q

      - name: Upload coverage shard ${{ matrix.shard }} to Codecov
        uses: codecov/codecov-action@v5
        with:
          files: coverage-${{ matrix.shard }}.xml
          flags: python
          name: python-shard-${{ matrix.shard }}
          # Token is sent so uploads authenticate once the repo is activated on
          # Codecov. Until then Codecov may 404 ("Repository not found"); either
          # way, coverage upload is reporting-only and must never fail a build
          # whose tests pass — so this stays non-blocking.
          token: ${{ secrets.CODECOV_TOKEN }}
          fail_ci_if_error: false

  test-extras:
    needs: [changes, build-wheel]
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 30
    env:
      FASTEMBED_CACHE_PATH: ${{ github.workspace }}/.fastembed-cache
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}
      - name: Cache pip
        uses: actions/cache@v5
        with:
          path: ~/.cache/pip
          key: ${{ runner.os }}-pip-extras-${{ hashFiles('pyproject.toml') }}
          restore-keys: ${{ runner.os }}-pip-extras-
      - name: Cache fastembed model
        uses: actions/cache@v5
        with:
          path: ${{ github.workspace }}/.fastembed-cache
          key: ${{ runner.os }}-fastembed-bge-small-v1
      - name: Download prebuilt wheel
        uses: actions/download-artifact@v8
        with:
          name: headroom-wheel
          path: dist
      - name: Install (CPU torch + wheel[dev,relevance])
        run: |
          python -m pip install --upgrade pip
          pip install torch --index-url https://download.pytorch.org/whl/cpu
          WHEEL="$(ls dist/*.whl)"
          pip install "${WHEEL}[dev,relevance]"
          SITE="$(python -c 'import sysconfig; print(sysconfig.get_path("platlib"))')"
          cp "${SITE}/headroom/"_core*.so headroom/
          python -c "from headroom._core import SmartCrusher; print('headroom._core OK')"
      - name: Pre-fetch fastembed model (authenticated, resilient)
        env:
          HF_TOKEN: ${{ secrets.HF_TOKEN }}
          HF_HUB_DISABLE_TELEMETRY: "1"
        run: |
          for i in 1 2 3 4 5; do
            if python -c "from fastembed import TextEmbedding; TextEmbedding('BAAI/bge-small-en-v1.5')"; then exit 0; fi
            echo "::warning::fastembed fetch attempt $i failed; backing off"; sleep $((i * 20))
          done
          echo "::error::could not fetch fastembed model from HuggingFace"; exit 1
      - name: Run relevance tests
        # Offline so fastembed reads the cache the prefetch step just warmed,
        # without an unauthenticated cache-validation HEAD that could 429.
        env:
          HF_HUB_OFFLINE: "1"
          TRANSFORMERS_OFFLINE: "1"
        run: pytest tests/test_relevance.py -v

  test-agno:
    needs: [changes, build-wheel]
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}
      - name: Download prebuilt wheel
        uses: actions/download-artifact@v8
        with:
          name: headroom-wheel
          path: dist
      - name: Install (CPU torch + wheel[dev,agno])
        run: |
          python -m pip install --upgrade pip
          pip install torch --index-url https://download.pytorch.org/whl/cpu
          WHEEL="$(ls dist/*.whl)"
          pip install "${WHEEL}[dev,agno]"
          SITE="$(python -c 'import sysconfig; print(sysconfig.get_path("platlib"))')"
          cp "${SITE}/headroom/"_core*.so headroom/
      - name: Run agno tests
        run: pytest tests/test_integrations/agno/ -v

  test-dashboard-ui:
    needs: [changes, build-wheel]
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: ${{ env.PY_VERSION }}
      - name: Download prebuilt wheel
        uses: actions/download-artifact@v8
        with:
          name: headroom-wheel
          path: dist
      - name: Install (CPU torch + wheel[dev] + playwright)
        run: |
          python -m pip install --upgrade pip
          pip install torch --index-url https://download.pytorch.org/whl/cpu
          WHEEL="$(ls dist/*.whl)"
          pip install "${WHEEL}[dev]" playwright
          SITE="$(python -c 'import sysconfig; print(sysconfig.get_path("platlib"))')"
          cp "${SITE}/headroom/"_core*.so headroom/
      - name: Install chromium
        run: playwright install --with-deps chromium
      - name: Run dashboard playwright tests
        # Stub-based dashboard tests only (routes fully mocked, no network).
        # tests/test_dashboard/test_live_feed.py needs a live proxy on
        # localhost:8787 and stays excluded; the main shards keep skipping
        # these via importorskip since playwright is not installed there.
        env:
          HEADROOM_PLAYWRIGHT_ARTIFACT_DIR: ${{ runner.temp }}/playwright-artifacts
        run: pytest tests/test_dashboard_*_playwright.py -v
      - name: Upload dashboard screenshots
        if: always()
        uses: actions/upload-artifact@v7
        with:
          name: dashboard-playwright-artifacts
          path: ${{ runner.temp }}/playwright-artifacts
          if-no-files-found: ignore
          retention-days: 7

  commitlint:
    if: github.event_name == 'pull_request'
    runs-on: ubuntu-latest
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@v6
        with:
          fetch-depth: 0
      - uses: wagoid/commitlint-github-action@v6
        with:
          configFile: .commitlintrc.json

  build:
    needs: changes
    if: needs.changes.outputs.code == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: "3.11"
      - name: Cache pip
        uses: actions/cache@v5
        with:
          path: ~/.cache/pip
          key: ${{ runner.os }}-pip-build-${{ hashFiles('pyproject.toml') }}
          restore-keys: ${{ runner.os }}-pip-build-
      - uses: dtolnay/rust-toolchain@1.96.0
      - uses: Swatinem/rust-cache@v2
        with:
          workspaces: ". -> target"
      # Smoke check that the SHIPPED build (release profile) + sdist are wired
      # right; release.yml's matrix is what actually publishes to PyPI.
      - name: Install build tools
        run: |
          python -m pip install --upgrade pip
          pip install 'maturin>=1.5,<2.0' twine
      - name: Build wheel + sdist
        run: |
          maturin sdist --out dist
          maturin build --release --out dist
      - name: Check package
        run: twine check dist/*

  workflow-validation:
    needs: changes
    if: needs.changes.outputs.workflows == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@v6
      - name: Cache actionlint + act
        id: tools-cache
        uses: actions/cache@v5
        with:
          path: |
            /usr/local/bin/actionlint
            /usr/local/bin/act
          # Key off the workflow file itself: when someone updates the
          # download URLs to a newer tool version, the hash changes and
          # the cache busts automatically.
          key: ${{ runner.os }}-ci-tools-${{ hashFiles('.github/workflows/ci.yml') }}

      - name: Install actionlint
        if: steps.tools-cache.outputs.cache-hit != 'true'
        run: |
          curl -fsSL https://raw.githubusercontent.com/rhysd/actionlint/main/scripts/download-actionlint.bash | bash
          sudo mv ./actionlint /usr/local/bin/actionlint

      - name: Install act
        if: steps.tools-cache.outputs.cache-hit != 'true'
        run: |
          curl -fsSL https://raw.githubusercontent.com/nektos/act/master/install.sh | sudo bash
          sudo install ./bin/act /usr/local/bin/act
      - name: Validate workflow files
        run: bash scripts/validate-workflows.sh

  docker-native-e2e:
    needs: changes
    if: needs.changes.outputs.e2e == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 45
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: "3.11"
      - name: Build local Headroom image
        run: docker build -t headroom-native-e2e:latest .
      - name: Run Docker-native installer e2e
        env:
          HEADROOM_DOCKER_IMAGE: headroom-native-e2e:latest
        run: bash e2e/docker-native-install.sh
      - name: Run Docker-native compose smoke test
        env:
          HEADROOM_IMAGE: headroom-native-e2e:latest
          HEADROOM_HOST_HOME: ${{ github.workspace }}
          HEADROOM_WORKSPACE: ${{ github.workspace }}
        run: |
          mkdir -p .headroom .claude .codex .gemini
          trap 'docker compose -f docker/docker-compose.native.yml down -v' EXIT
          docker compose -f docker/docker-compose.native.yml up -d proxy
          for attempt in $(seq 1 30); do
            if curl --fail --silent http://127.0.0.1:8787/readyz >/dev/null; then
              break
            fi
            if [ "$attempt" -eq 30 ]; then
              docker compose -f docker/docker-compose.native.yml logs proxy
              exit 1
            fi
            sleep 1
          done
      - name: Run Docker-native wrap e2e
        run: |
          docker build -f e2e/wrap/Dockerfile -t headroom-wrap-e2e .
          docker run --rm headroom-wrap-e2e
      - name: Run Docker-native init e2e
        run: |
          docker build -f e2e/init/Dockerfile -t headroom-init-e2e .
          docker run --rm headroom-init-e2e

  windows-native-wrapper:
    needs: changes
    if: needs.changes.outputs.e2e == 'true'
    runs-on: windows-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: "3.12"
      - name: Install test dependencies
        run: |
          python -m pip install --upgrade pip
          pip install pytest
      - name: Run native installer wrapper tests
        run: pytest tests/test_install/test_native_installers.py -q

  macos-native-wrapper:
    needs: changes
    if: needs.changes.outputs.e2e == 'true'
    runs-on: macos-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v6
      - uses: actions/setup-python@v6
        with:
          python-version: "3.11"
      - name: Install bash and test dependencies
        run: |
          brew install bash
          python -m pip install --upgrade pip
          python -m pip install --retries 10 --timeout 60 pytest
      - name: Run native installer wrapper tests
        run: |
          BASH_PREFIX="$(brew --prefix bash)"
          export PATH="$BASH_PREFIX/bin:$PATH"
          pytest tests/test_install/test_native_installers.py -q