ci.yml

name: CI

on:
  push:
    branches: [main]
  pull_request:

jobs:
  lint:
    name: Lint
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4

      - uses: astral-sh/setup-uv@38f3f104447c67c051c4a08e39b64a148898af3a  # v4
        with:
          python-version: "3.11"

      # Plain `uv sync`, deliberately, on two counts.
      #
      # Resolving fresh is wanted on the pull requests where it actually
      # happens, which is not all of them. `uv sync` follows the committed lock
      # while that lock satisfies pyproject.toml, so an ordinary pull request
      # installs the locked versions. When a pull request moves dependency
      # metadata without regenerating the lock, uv resolves the new set and
      # these jobs exercise it. `--locked` would instead refuse to run and
      # report that the lock needs updating, which says nothing about whether
      # the new version works.
      #
      # The two flags tried here also break the release, and for one shared
      # reason: the release bumps the version in pyproject.toml and leaves
      # uv.lock naming the old one, so uv has to re-resolve.
      #
      #   --locked     fails on exactly that mismatch, reporting that the
      #                lockfile needs to be updated.
      #   --no-build   fails because re-resolving means building this
      #                workspace's own three packages, which are source rather
      #                than wheels: "Building source distributions for
      #                coordinode is disabled".
      #
      # Verified on the release branch with a cold cache: both fail there, and
      # both pass on that same branch once uv.lock carries the new version. So
      # restoring either one means first making the release regenerate the
      # lock, and even then it gives up the fresh resolution described above.
      - run: uv sync
      - run: uv run ruff check coordinode/ langchain-coordinode/ llama-index-coordinode/ tests/
      - run: uv run ruff format --check coordinode/ langchain-coordinode/ llama-index-coordinode/ tests/

  test:
    name: Test (Python ${{ matrix.python-version }})
    runs-on: ubuntu-latest
    strategy:
      fail-fast: false
      matrix:
        python-version: ["3.11", "3.12", "3.13", "3.14"]
    steps:
      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
        with:
          submodules: recursive

      - uses: astral-sh/setup-uv@38f3f104447c67c051c4a08e39b64a148898af3a  # v4
        with:
          python-version: ${{ matrix.python-version }}

      - name: Install dependencies
        run: uv sync --all-packages

      - name: Generate proto stubs
        run: uv run make proto

      - name: Unit tests
        run: uv run pytest tests/unit/ -v

  build-embedded:
    name: Build embedded (CI check)
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
        with:
          submodules: recursive

      - name: Build (maturin / Linux x86_64)
        uses: PyO3/maturin-action@32307a466a178317e8c2ae343b38e73896a047be  # v1.47.0
        with:
          command: build
          args: >-
            --manifest-path coordinode-embedded/Cargo.toml
            --out dist
          manylinux: manylinux_2_28
          before-script-linux: |
            dnf install -y protobuf-compiler

      - uses: astral-sh/setup-uv@38f3f104447c67c051c4a08e39b64a148898af3a  # v4
        with:
          python-version: "3.12"

      - name: Install wheel + run embedded tests
        # The main `test` job skips coordinode_embedded with importorskip
        # because that runner doesn't build the wheel.  This job has the
        # wheel, so install it and run the embedded tests here, then assert
        # nothing was skipped (a skip with the wheel installed means a
        # broken manylinux glibc / missing numpy / etc., not the expected
        # "no wheel" condition).
        # maturin-action runs inside a manylinux container; only paths
        # under $GITHUB_WORKSPACE are bind-mounted back to the host, so
        # the wheel goes to ./dist/ (workspace-relative), not /tmp/.
        run: |
          set -euo pipefail
          WHL=$(ls dist/coordinode_embedded-*.whl | head -1)
          test -n "$WHL"
          uv venv --python 3.12 /tmp/test-venv
          # Pinned to the versions uv.lock already resolves for the workspace,
          # so this throwaway venv tests against the same stack as the other
          # jobs instead of whatever is newest on PyPI that morning.
          uv pip install --python /tmp/test-venv/bin/python "$WHL" \
            numpy==2.4.6 pytest==9.1.1 pytest-timeout==2.4.0
          # Every test file that needs the extension belongs here: the main
          # `test` job skips them all for want of a wheel, so a file left out
          # of this list runs in no job at all.
          OUTPUT=$(/tmp/test-venv/bin/python -m pytest \
            tests/unit/test_hnsw.py tests/unit/test_embedded_values.py \
            -v --strict-markers -ra 2>&1)
          echo "$OUTPUT"
          # `-ra` prints a "SKIPPED [N]" short summary header when anything was
          # skipped; importorskip-based skip also surfaces this way.  Use the
          # full output instead of the exit code because pytest treats skips
          # as success.
          if echo "$OUTPUT" | grep -qE '^SKIPPED|=+ .* skipped'; then
            echo "::error::Tests were skipped in build-embedded; the wheel is installed so this is a real failure (broken glibc / missing numpy / etc.)"
            exit 1
          fi
          if ! echo "$OUTPUT" | grep -qE '=+ [0-9]+ passed'; then
            echo "::error::No tests passed; the module was likely not collected"
            exit 1
          fi

  test-integration:
    name: Integration tests
    runs-on: ubuntu-latest
    services:
      coordinode:
        # Pinned by digest, not by tag: the proto submodule pins a server
        # version, and a tag can be re-pushed, so `:0.5.7` alone does not name
        # one fixed server. The digest below is 0.5.7; bump both together with
        # the submodule.
        image: ghcr.io/structured-world/coordinode@sha256:75a6242beb4cea8ab6726c842898fafc03ca9f59c6f7e36e462934a4ca64874c
        ports:
          - 7080:7080
          - 7084:7084
    steps:
      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
        with:
          submodules: recursive

      - uses: astral-sh/setup-uv@38f3f104447c67c051c4a08e39b64a148898af3a  # v4
        with:
          python-version: "3.11"

      - name: Wait for coordinode
        run: |
          echo "Waiting for coordinode to be ready (HTTP :7084 + gRPC :7080)..."
          for i in $(seq 1 30); do
            if curl -sf http://localhost:7084/health >/dev/null 2>&1 && \
               (echo > /dev/tcp/localhost/7080) 2>/dev/null; then
              echo "coordinode is ready (attempt $i)"
              exit 0
            fi
            echo "Attempt $i/30, not ready yet, sleeping 5s..."
            sleep 5
          done
          echo "Error: coordinode did not become healthy after 150s" >&2
          exit 1

      - name: Install + generate proto
        run: |
          uv sync --all-packages
          uv run make proto

      - name: Integration tests
        env:
          COORDINODE_ADDR: "localhost:7080"
        run: uv run pytest tests/integration/ -v --timeout=30