ci.yaml

name: CI

# Runs on every push to main and on every PR targeting main.
#
# Scope: platform-neutral unit tests with 100% line coverage on the
# library modules we actually ship for this commit. We deliberately
# DO NOT run:
#
#   * tests/core/                  — needs HuggingFace weights
#   * tests/system/                — same, plus is slow
#   * tests/inference_engine/proposer/ — uses real Qwen3 sparse
#                                        proposer; HF-cache-bound
#   * tests/backends/mlx/test_{verifier,proposer,cache,torch_bridge}.py
#                                  — Apple-Silicon only
#
# Mac and CUDA contributors run the full suite locally via
# scripts/run_platform_tests.sh and push the platform-test reports to
# the PR branch as evidence; this CI workflow guards the platform-
# neutral surface so a regression there cannot land on main.

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

# Cancel superseded runs on the same branch — saves CI time on
# rapid-fire pushes.
concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

jobs:
  unit-tests:
    name: unit tests + 100% coverage
    runs-on: ubuntu-latest
    strategy:
      fail-fast: false
      matrix:
        python-version: ["3.12"]
    steps:
      - name: Check out
        uses: actions/checkout@v4

      - name: Set up Python ${{ matrix.python-version }}
        uses: actions/setup-python@v5
        with:
          python-version: ${{ matrix.python-version }}
          cache: pip

      - name: Install dependencies
        run: |
          python -m pip install --upgrade pip
          pip install -r requirements.txt

      - name: Show installed key versions
        run: |
          python -c "import torch, fastapi, pydantic, prometheus_client, transformers; \
            print('torch', torch.__version__); \
            print('fastapi', fastapi.__version__); \
            print('pydantic', pydantic.VERSION); \
            print('transformers', transformers.__version__); \
            print('prometheus_client', __import__('importlib.metadata', fromlist=['version']).version('prometheus_client'))"

      - name: Run platform-neutral test suite with 100% coverage
        env:
          PYTHONPATH: .
        run: |
          pytest \
            tests/inference_engine/server/ \
            tests/inference_engine/memory/ \
            tests/inference_engine/scheduler/ \
            tests/inference_engine/pipeline/ \
            tests/training/repr_align/ \
            tests/backends/mlx/test_env.py \
            --cov=inference_engine.server \
            --cov=inference_engine.memory \
            --cov=inference_engine.scheduler \
            --cov=inference_engine.pipeline \
            --cov=training.repr_align \
            --cov-report=term \
            --cov-report=xml:coverage.xml \
            --cov-fail-under=100 \
            --junitxml=junit.xml \
            -v

      - name: Upload coverage artifact
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: coverage-py${{ matrix.python-version }}
          path: |
            coverage.xml
            junit.xml
          if-no-files-found: warn
          retention-days: 14

  package-import-smoke:
    name: package import smoke
    runs-on: ubuntu-latest
    steps:
      - name: Check out
        uses: actions/checkout@v4
      - name: Set up Python 3.12
        uses: actions/setup-python@v5
        with:
          python-version: "3.12"
          cache: pip
      - name: Install dependencies
        run: |
          python -m pip install --upgrade pip
          pip install -r requirements.txt
      - name: Import every shipping subpackage
        env:
          PYTHONPATH: .
        run: |
          python -c "import inference_engine; \
            import inference_engine.server; \
            import inference_engine.server.app; \
            import inference_engine.server.config; \
            import inference_engine.server.engine; \
            import inference_engine.server.metrics; \
            import inference_engine.server.errors; \
            import inference_engine.server.auth; \
            import inference_engine.server.tokenizer; \
            import inference_engine.server.streaming; \
            import inference_engine.server.schemas; \
            import inference_engine.memory; \
            import inference_engine.memory.slab; \
            import inference_engine.memory.pool; \
            import inference_engine.scheduler; \
            import inference_engine.scheduler.config; \
            import inference_engine.scheduler.scheduler; \
            import inference_engine.scheduler.session; \
            import inference_engine.pipeline; \
            import inference_engine.pipeline.coordinator; \
            import inference_engine.proposer; \
            import inference_engine.proposer.sparse_logits; \
            import inference_engine.backends.mlx.env; \
            import training.repr_align; \
            print('all imports succeeded')"

  docker-build:
    name: docker build + import smoke
    runs-on: ubuntu-latest
    steps:
      - name: Check out
        uses: actions/checkout@v4

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v3

      - name: Build image
        uses: docker/build-push-action@v6
        with:
          context: .
          file: Dockerfile
          tags: kakeya:ci-${{ github.sha }}
          push: false
          load: true
          cache-from: type=gha
          cache-to: type=gha,mode=max

      - name: Smoke-test image (import only — weights would require HF cache)
        run: |
          # Confirm the package imports inside the image without
          # touching HF cache.
          docker run --rm \
            --entrypoint python \
            kakeya:ci-${{ github.sha }} \
            -c "import inference_engine.server.app; print('image imports clean')"
          # Confirm the server CLI parses (--help exits 0 before any
          # model load, so we don't need HF cache).
          docker run --rm \
            kakeya:ci-${{ github.sha }} \
            --help