zeph 0.22.4

name: CI

on:
  push:
    branches: [main]
  pull_request:

permissions:
  contents: read
  pull-requests: write
  security-events: write

env:
  CARGO_TERM_COLOR: always
  RUSTC_WRAPPER: ""
  CARGO_INCREMENTAL: "0"

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

jobs:
  detect-changes:
    name: Detect Changes
    runs-on: ubuntu-latest
    timeout-minutes: 2
    outputs:
      docs-only: ${{ steps.classify.outputs.docs-only }}
      specs-only: ${{ steps.classify.outputs.specs-only }}
      run-full-ci: ${{ steps.classify.outputs.run-full-ci }}
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - name: Filter changed paths
        uses: dorny/paths-filter@ceb8a2b8f2d89434be7ff52d3de7ec3738c5cc9d # v4.0.3
        id: filter
        with:
          filters: |
            docs:
              - 'book/src/**'
              - 'docs/**'
              - '**.md'
              - '.local/testing/**'
            specs:
              - 'specs/**'
            code:
              - 'crates/**'
              - 'src/**'
              - 'Cargo.toml'
              - 'Cargo.lock'
              - 'build.rs'
              - 'docker/**'
            workflows:
              - '.github/workflows/**'
              - '.github/nextest.toml'
              - '.github/testcontainers-images.txt'
              - '.cargo/**'
            shell:
              - 'install/**'
      - name: Classify changes
        id: classify
        run: |
          DOCS="${{ steps.filter.outputs.docs }}"
          SPECS="${{ steps.filter.outputs.specs }}"
          CODE="${{ steps.filter.outputs.code }}"
          WORKFLOWS="${{ steps.filter.outputs.workflows }}"
          SHELL_SCRIPTS="${{ steps.filter.outputs.shell }}"

          # docs-only: docs or specs changed, nothing that requires building
          if [[ "$CODE" == "false" && "$WORKFLOWS" == "false" && "$SHELL_SCRIPTS" == "false" && ("$DOCS" == "true" || "$SPECS" == "true") ]]; then
            echo "docs-only=true" >> "$GITHUB_OUTPUT"
          else
            echo "docs-only=false" >> "$GITHUB_OUTPUT"
          fi

          # specs-only: specs changed, no code or workflow changes
          if [[ "$SPECS" == "true" && "$CODE" == "false" && "$WORKFLOWS" == "false" && "$SHELL_SCRIPTS" == "false" && "$DOCS" == "false" ]]; then
            echo "specs-only=true" >> "$GITHUB_OUTPUT"
          else
            echo "specs-only=false" >> "$GITHUB_OUTPUT"
          fi

          # run-full-ci: any code, workflow, or shell change triggers full CI
          if [[ "$CODE" == "true" || "$WORKFLOWS" == "true" || "$SHELL_SCRIPTS" == "true" ]]; then
            echo "run-full-ci=true" >> "$GITHUB_OUTPUT"
          else
            echo "run-full-ci=false" >> "$GITHUB_OUTPUT"
          fi

  cla:
    name: CLA Check
    if: github.event_name == 'pull_request'
    runs-on: ubuntu-latest
    timeout-minutes: 5
    permissions:
      contents: write
      pull-requests: write
    steps:
      - uses: contributor-assistant/github-action@ca4a40a7d1004f18d9960b404b97e5f30a505a08 # v2.6.1
        with:
          path-to-signatures: cla-signatures.json
          path-to-document: https://github.com/bug-ops/zeph/blob/main/.github/CLA.md
          branch: cla-signatures
          allowlist: bug-ops,dependabot[bot],github-actions[bot],renovate[bot]
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  lint-shellcheck:
    name: Lint (shellcheck)
    needs: detect-changes
    # Run when shell scripts changed, or when doing full CI (workflows changed)
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - name: Run shellcheck
        uses: ludeeus/action-shellcheck@00cae500b08a931fb5698e11e79bfbd38e612a38 # 2.0.0
        with:
          scandir: install
      - name: Transcript-integrity audit gate (issue #6360)
        run: bash .github/scripts/check_integrity_audit.sh
      - name: ACP permission-gate test-isolation guard (issue #6515)
        run: bash .github/scripts/check_acp_permission_gate.sh

  lint-fmt:
    name: Lint (fmt)
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 5
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@7c8d7d138f5c09cef361f8214cf96882cd029cdb # nightly
        with:
          toolchain: nightly
          components: rustfmt
      - name: Check formatting
        run: cargo +nightly fmt --check

  lint-clippy:
    name: Lint (clippy, ${{ matrix.label }})
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    strategy:
      fail-fast: false
      matrix:
        include:
          - features: "desktop,ide,server,chat,pdf,scheduler,registry,testing,deep-link"
            label: desktop-ide-server-chat-pdf-scheduler
          - features: bench
            label: bench
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
          components: clippy
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          shared-key: "ci"
      - name: Clippy
        run: cargo clippy --profile ci --workspace --all-targets --features ${{ matrix.features }} -- -D warnings

  msrv:
    name: MSRV check (1.97, ${{ matrix.label }})
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    strategy:
      fail-fast: false
      matrix:
        include:
          - features: "desktop,ide,server,chat,pdf,scheduler,registry,deep-link"
            label: desktop-ide-server-chat-pdf-scheduler
          - features: bench
            label: bench
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: "1.97"
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          shared-key: "msrv-${{ matrix.label }}"
      - name: cargo check (MSRV, --all-targets)
        run: cargo check --workspace --all-targets --features ${{ matrix.features }} --locked

  build-tests:
    # Linux-only: macOS/Windows builds live in the on-demand ci-non-linux.yml
    # workflow. The artifact name keeps the -ubuntu-latest suffix for naming
    # parity with the per-OS archives produced there.
    # Nightly + -Z threads=4 (issue #6737), pinned to a dated toolchain so new
    # nightly lints don't fail this gate via build.warnings = "deny". RUSTFLAGS
    # set here, not .cargo/config.toml (af1be7ba/cf75b63f).
    name: Build Tests
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 45
    env:
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@7c8d7d138f5c09cef361f8214cf96882cd029cdb # nightly
        with:
          toolchain: nightly-2026-08-07
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          cache-targets: "false"
          shared-key: "ci-nightly-threads"
      - uses: mozilla-actions/sccache-action@fc920bf0ec8de6ee65d409111f7ec508035751ba # v0.0.11
      - uses: taiki-e/install-action@4f74fa8a3d358e568049b7b1a7fbb4fceabda454 # nextest
      - name: Build and archive tests
        env:
          RUSTFLAGS: "-Z threads=4"
        run: cargo nextest archive --config-file .github/nextest.toml --cargo-profile ci --workspace --features "desktop,ide,server,chat,pdf,scheduler,registry,deep-link" --lib --bins --tests --archive-file nextest-archive.tar.zst
      - name: Upload test archive
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: nextest-archive-ubuntu-latest
          path: nextest-archive.tar.zst
          retention-days: 1
      - name: Upload binary
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: zeph-binary
          path: target/ci/zeph
          retention-days: 1

  build-tests-postgres:
    # The four postgres archives build RUNNABLE test binaries (with
    # test-utils = testcontainers) consumed by the integration matrix. They do
    # not duplicate the build-postgres job, which is check-only: it type-checks
    # the whole workspace under the postgres backend and produces no artifacts.
    # Split out of build-tests so they compile in parallel with the (longer)
    # main workspace archive instead of serially after it — integration jobs
    # start as soon as the slower of the two build jobs finishes.
    name: Build Tests (postgres archives)
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    env:
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          cache-targets: "false"
          shared-key: "ci"
      - uses: mozilla-actions/sccache-action@fc920bf0ec8de6ee65d409111f7ec508035751ba # v0.0.11
      - uses: taiki-e/install-action@4f74fa8a3d358e568049b7b1a7fbb4fceabda454 # nextest
      - name: Build and archive zeph-db postgres integration tests
        run: cargo nextest archive --config-file .github/nextest.toml --cargo-profile ci -p zeph-db --no-default-features --features test-utils --tests --archive-file nextest-archive-zeph-db-postgres.tar.zst
      - name: Upload zeph-db postgres test archive
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: nextest-archive-zeph-db-postgres
          path: nextest-archive-zeph-db-postgres.tar.zst
          retention-days: 1
      - name: Build and archive zeph-memory postgres integration tests
        # sqlite and postgres are mutually exclusive (zeph-db's compile_error!
        # fires if both are active), so default features (which include "sqlite")
        # must be disabled here; "jsonschema" is re-added explicitly since it was
        # previously pulled in via zeph-memory's default feature set.
        # --lib --bins (not just --tests) is required so cfg!(feature = "postgres")
        # branches inside src/ (e.g. dialect-selected SQL literals) get archived too —
        # see the "Run zeph-memory postgres unit tests" step in the integration
        # job (#5540).
        run: cargo nextest archive --config-file .github/nextest.toml --cargo-profile ci -p zeph-memory --no-default-features --features test-utils,jsonschema --lib --bins --tests --archive-file nextest-archive-zeph-memory-postgres.tar.zst
      - name: Upload zeph-memory postgres test archive
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: nextest-archive-zeph-memory-postgres
          path: nextest-archive-zeph-memory-postgres.tar.zst
          retention-days: 1
      - name: Build and archive zeph-index postgres integration tests
        # sqlite and postgres are mutually exclusive; default features (which
        # include "sqlite") must be disabled so only postgres is active.
        run: cargo nextest archive --config-file .github/nextest.toml --cargo-profile ci -p zeph-index --no-default-features --features test-utils --tests --archive-file nextest-archive-zeph-index-postgres.tar.zst
      - name: Upload zeph-index postgres test archive
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: nextest-archive-zeph-index-postgres
          path: nextest-archive-zeph-index-postgres.tar.zst
          retention-days: 1
      - name: Build and archive zeph-orchestration postgres integration tests
        # sqlite and postgres are mutually exclusive; default features (which
        # include "sqlite") must be disabled so only postgres is active.
        # test-utils implies llm-planning: tests/postgres_integration.rs exercises
        # PlanCache, which lives behind that feature gate (#5839).
        run: cargo nextest archive --config-file .github/nextest.toml --cargo-profile ci -p zeph-orchestration --no-default-features --features test-utils --tests --archive-file nextest-archive-zeph-orchestration-postgres.tar.zst
      - name: Upload zeph-orchestration postgres test archive
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: nextest-archive-zeph-orchestration-postgres
          path: nextest-archive-zeph-orchestration-postgres.tar.zst
          retention-days: 1

  test:
    name: "Test (shard ${{ matrix.partition }})"
    needs: [detect-changes, lint-fmt, lint-clippy, build-tests]
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    strategy:
      fail-fast: false
      matrix:
        # Increased from 8 (issue #6737) — real timing data showed ~23%
        # avoidable tail from hash-partition skew.
        partition:
          [
            "1/12",
            "2/12",
            "3/12",
            "4/12",
            "5/12",
            "6/12",
            "7/12",
            "8/12",
            "9/12",
            "10/12",
            "11/12",
            "12/12",
          ]
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: taiki-e/install-action@4f74fa8a3d358e568049b7b1a7fbb4fceabda454 # nextest
      - name: Download test archive
        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
        with:
          name: nextest-archive-ubuntu-latest
      - name: "Run tests (shard ${{ matrix.partition }})"
        run: |
          cargo nextest run \
            --config-file .github/nextest.toml \
            --archive-file nextest-archive.tar.zst \
            --workspace-remap . \
            --profile ci-partition \
            --partition hash:${{ matrix.partition }}

  integration:
    # One job per container-backed suite so the suites run in parallel instead
    # of serially — the job was the tail of the critical path. Non-ignored
    # tests in *integration* binaries (e.g. zeph-acp's stdio tests) already run
    # in the sharded test job; only #[ignore]d container-backed tests run here.
    name: Integration Tests (${{ matrix.suite }})
    needs: [detect-changes, lint-fmt, lint-clippy, build-tests, build-tests-postgres]
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 15
    strategy:
      fail-fast: false
      matrix:
        include:
          - suite: qdrant
            # qdrant_integration and document_integration are both Qdrant-backed
            # (document_integration fakes embeddings in-test — no extra services);
            # without the second binary those #[ignore]d tests would run nowhere
            # in CI. Container concurrency is capped by the qdrant-containers
            # test-group in .github/nextest.toml.
            archive: nextest-archive-ubuntu-latest
            file: nextest-archive.tar.zst
            run_args: "-E 'binary(qdrant_integration) or binary(document_integration)'"
          - suite: zeph-db-postgres
            archive: nextest-archive-zeph-db-postgres
            file: nextest-archive-zeph-db-postgres.tar.zst
            run_args: ""
          - suite: zeph-memory-postgres
            # Scoped to the postgres_integration binary: zeph-memory's archive also
            # contains hela_spreading_activation.rs (SQLite `:memory:` ignored tests,
            # unrelated to postgres). This archive is built postgres-only (sqlite is
            # disabled, see the archive step's comment in build-tests-postgres) — the
            # vast majority of zeph-memory's tests use a `SqliteStore::new(":memory:")`
            # fixture unconditionally (not gated on `#[cfg(feature = "sqlite")]`), so
            # they'd try to parse ":memory:" as a Postgres URL and fail; running
            # crate-wide would hit those unrelated failures.
            archive: nextest-archive-zeph-memory-postgres
            file: nextest-archive-zeph-memory-postgres.tar.zst
            run_args: "-E 'binary(postgres_integration)'"
          - suite: zeph-index-postgres
            archive: nextest-archive-zeph-index-postgres
            file: nextest-archive-zeph-index-postgres.tar.zst
            run_args: "-E 'binary(postgres_integration)'"
          - suite: zeph-orchestration-postgres
            archive: nextest-archive-zeph-orchestration-postgres
            file: nextest-archive-zeph-orchestration-postgres.tar.zst
            run_args: "-E 'binary(postgres_integration)'"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: taiki-e/install-action@4f74fa8a3d358e568049b7b1a7fbb4fceabda454 # nextest
      - uses: docker/login-action@dbcb813823bdd20940b903addbd779551569679f # v4
        with:
          username: ${{ secrets.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}
      - name: Restore testcontainer image cache
        id: tc-image-cache
        uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
        with:
          path: /tmp/testcontainer-images
          key: testcontainer-images-${{ hashFiles('.github/testcontainers-images.txt') }}
      - name: Pull and save testcontainer images (cache miss)
        if: steps.tc-image-cache.outputs.cache-hit != 'true'
        run: |
          mkdir -p /tmp/testcontainer-images
          while IFS= read -r image; do
            [ -z "$image" ] && continue
            echo "Pulling $image"
            docker pull "$image"
            fname=$(echo "$image" | tr '/:' '__')
            docker save "$image" | zstd -T0 -3 -o /tmp/testcontainer-images/"$fname".tar.zst
          done < .github/testcontainers-images.txt
          ls -lh /tmp/testcontainer-images/
      - name: Load testcontainer images (cache hit)
        if: steps.tc-image-cache.outputs.cache-hit == 'true'
        run: |
          for f in /tmp/testcontainer-images/*.tar.zst; do
            echo "Loading $f"
            zstd -d -c "$f" | docker load
          done
          docker image ls
      - name: Download test archive
        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
        with:
          name: ${{ matrix.archive }}
      - name: Run integration tests (testcontainers)
        run: cargo nextest run --config-file .github/nextest.toml --archive-file ${{ matrix.file }} --workspace-remap . --profile ci --run-ignored ignored-only ${{ matrix.run_args }}
      - name: Run zeph-memory postgres unit tests (#5540)
        if: matrix.suite == 'zeph-memory-postgres'
        # Runs the --lib/--bins portion of the same archive (built postgres-only
        # via test-utils, see the archive step in build-tests-postgres), so
        # cfg!(feature = "postgres") branches inside src/ get unit-test coverage in
        # CI, not just integration-test coverage that depends on happening to exercise
        # the right code path. Scoped by name (not `kind(lib) + kind(bin)`) because the
        # vast majority of zeph-memory's ~1500 lib unit tests open a `SqliteStore::new(
        # ":memory:")` fixture unconditionally — under this postgres-only build, those
        # unrelated tests would try to parse ":memory:" as a Postgres URL and fail.
        # `_matches_active_dialect` is this codebase's established naming convention
        # (see store/overflow.rs) for pure dialect-literal pinning tests that take no
        # DB connection and are safe to run under either feature build.
        run: cargo nextest run --config-file .github/nextest.toml --archive-file nextest-archive-zeph-memory-postgres.tar.zst --workspace-remap . --profile ci -E 'kind(lib) and test(matches_active_dialect)'
      - name: Reap orphaned testcontainers
        # Defense-in-depth, not a fix: testcontainers-rs 0.27.3 (pinned) has no Ryuk
        # reaper and can leak containers on SIGKILL or a StartupTimeout cancel under
        # concurrent load (#5546, #5547). Runs even if a prior step failed/timed out.
        if: always()
        run: docker ps -aq --filter "label=org.testcontainers.managed-by=testcontainers" | xargs -r docker rm -f

  coverage:
    name: Coverage
    if: github.event_name == 'push' && github.ref == 'refs/heads/main' && needs.detect-changes.outputs.run-full-ci == 'true'
    needs: [detect-changes, lint-fmt, lint-clippy]
    runs-on: ubuntu-latest
    timeout-minutes: 20
    # coverage uses -C instrument-coverage which produces different artifacts than
    # normal builds; sccache still helps for unchanged crates between main pushes.
    # This job also builds with `--features full`, which includes crates (classifiers,
    # profiling, sandbox, gonka, cocoon) outside the lint-clippy matrix, so a warning
    # could appear here that lint-clippy never sees. CARGO_BUILD_WARNINGS overrides the
    # repo-wide `.cargo/config.toml` deny-by-default to prevent that from promoting
    # workspace lint warnings to errors during instrumented compilation — linting is
    # enforced by the dedicated lint-clippy job.
    env:
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
      CARGO_BUILD_WARNINGS: "allow"
      # -C instrument-coverage inflates per-frame size on top of already-large
      # Config-heavy async test frames (same root cause as the daemon-startup
      # stack overflow fixed in #5394); tests that build more than one Agent/
      # ServeAgentDeps in the same frame (e.g.
      # build_agent_factory_gates_trust_state_independently_per_session, which as of
      # #6699 spawns its own 32 MiB-stack thread and so no longer depends on this var
      # itself) can overflow the default per-test thread stack under instrumentation
      # even though they pass in the non-instrumented `test` job. Keep this set — it
      # protects any other, unnamed multi-agent test that doesn't manage its own stack.
      RUST_MIN_STACK: "33554432"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          cache-targets: "false"
          shared-key: "coverage"
      - uses: mozilla-actions/sccache-action@fc920bf0ec8de6ee65d409111f7ec508035751ba # v0.0.11
      - uses: taiki-e/install-action@eba66cc6f87204a1e73f96e528e759b6c1fcf573 # cargo-llvm-cov
      - uses: taiki-e/install-action@4f74fa8a3d358e568049b7b1a7fbb4fceabda454 # nextest
      - name: Generate coverage
        run: cargo llvm-cov nextest --config-file .github/nextest.toml --cargo-profile ci --workspace --features full --lib --bins --lcov --output-path lcov.info
      - name: Upload coverage
        uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f # v7
        with:
          token: ${{ secrets.CODECOV_TOKEN }}
          files: lcov.info
          fail_ci_if_error: false

  docker-build-and-scan:
    name: Docker Build and Security Scan
    needs: [detect-changes, lint-fmt, lint-clippy, build-tests]
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - name: Download binary
        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
        with:
          name: zeph-binary
          path: target/ci
      - name: Prepare binaries for Docker
        run: |
          mkdir -p binaries
          cp target/ci/zeph binaries/zeph-amd64
          cp target/ci/zeph binaries/zeph-arm64
      - uses: docker/setup-buildx-action@bb05f3f5519dd87d3ba754cc423b652a5edd6d2c # v4
      - name: Build Docker image
        uses: docker/build-push-action@53b7df96c91f9c12dcc8a07bcb9ccacbed38856a # v7
        with:
          context: .
          file: docker/Dockerfile
          load: true
          tags: zeph:local
          push: false
          cache-from: type=gha
          cache-to: type=gha,mode=max
      - name: Run Trivy vulnerability scanner
        uses: aquasecurity/trivy-action@ed142fd0673e97e23eac54620cfb913e5ce36c25 # v0.36.0
        # continue-on-error: trivy exits 1 when vulnerabilities are found; without
        # this the upload step would be skipped and findings lost from the dashboard.
        continue-on-error: true
        with:
          image-ref: zeph:local
          format: sarif
          output: trivy-results.sarif
          severity: CRITICAL,HIGH
          exit-code: '1'
          limit-severities-for-sarif: true
      - name: Upload Trivy results to GitHub Security tab
        uses: github/codeql-action/upload-sarif@d1ba80a13dd99fba24a470575428917156a28b43 # v4
        with:
          sarif_file: trivy-results.sarif

  bundle-check:
    name: Bundle Check (${{ matrix.bundle }})
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 15
    strategy:
      fail-fast: false
      matrix:
        bundle: [desktop, ide, server, chat, ml, bench, full]
        include:
          - bundle: ml
            allow_failure: true
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          shared-key: "ci"
      # --all-targets compiles test/bench/example targets too, so per-bundle
      # test-cfg-gate mismatches (e.g. #6607) are caught here, not just by the
      # broader combined-feature jobs above (lint-clippy, msrv, release-build).
      - name: Check bundle
        continue-on-error: ${{ matrix.allow_failure == true }}
        run: cargo check --all-targets --features ${{ matrix.bundle }}

  release-build:
    name: Release Build Check
    needs: [detect-changes, lint-fmt, lint-clippy]
    if: needs.detect-changes.outputs.run-full-ci == 'true' && github.event_name == 'push'
    runs-on: ubuntu-latest
    timeout-minutes: 45
    env:
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          cache-targets: "false"
          shared-key: "release-build"
      - uses: mozilla-actions/sccache-action@fc920bf0ec8de6ee65d409111f7ec508035751ba # v0.0.11
      - name: cargo build (release, --all-targets)
        # `cargo check --release` does not reproduce the release-profile query-depth overflow
        # class of bug fixed by #5407/#5408 (verified empirically: it passes even without the
        # `#![recursion_limit]` fix) because it skips codegen/LTO layout finalization. Only a
        # real `cargo build --release` with this workspace's `[profile.release]` (lto = true,
        # codegen-units = 1) exercises the same query depth CI is meant to gate on.
        run: cargo build --release --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,registry,testing,deep-link"

  release-build-full:
    name: Release Build Check (full)
    needs: [detect-changes, lint-fmt, lint-clippy]
    if: needs.detect-changes.outputs.run-full-ci == 'true' && github.event_name == 'push'
    runs-on: ubuntu-latest
    timeout-minutes: 45
    env:
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          cache-targets: "false"
          shared-key: "release-build-full"
      - uses: mozilla-actions/sccache-action@fc920bf0ec8de6ee65d409111f7ec508035751ba # v0.0.11
      - name: cargo build (release, --all-targets, full)
        # Best-effort candle/ml coverage under the release profile, mirroring the
        # bundle-check `ml` matrix leg (#5418): step-level continue-on-error keeps
        # this job's overall result `success` so it never blocks ci-status, while
        # still surfacing genuine full-feature release-build breakage in the logs.
        continue-on-error: true
        run: cargo build --release --workspace --all-targets --features full

  build-postgres:
    name: Build (postgres)
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          shared-key: "ci"
      - name: Verify sqlite+postgres compile_error guard
        # Intentionally attempts the forbidden both-enabled combination and asserts it
        # fails with the expected compile_error! message. Guards against a future
        # accidental deletion of the mutual-exclusivity guard in zeph-db/src/lib.rs —
        # this exact regression (guard silently dropped for #5377) is how #5571 happened.
        # `cargo check` here is EXPECTED to fail (exit 101, the compile_error firing).
        # GitHub Actions runs `run:` blocks under `bash -eo pipefail`, so piping that
        # exit code directly into `grep` would poison the pipeline status regardless of
        # whether grep matches — `|| true` swallows the expected failure so only the
        # grep result (checked separately against the redirected log) determines pass/fail.
        run: |
          cargo check -p zeph-db --features postgres > /tmp/guard.log 2>&1 || true
          if grep -q "mutually exclusive for \`zeph-db\`" /tmp/guard.log; then
            echo "guard fired as expected"
          else
            echo "::error::compile_error! guard did not fire — sqlite+postgres both-enabled regression"
            cat /tmp/guard.log
            exit 1
          fi
      - name: Check postgres build (workspace)
        # `full` is backend-agnostic (does not include sqlite or postgres); the default
        # sqlite feature must be disabled explicitly since Cargo features are
        # additive-only and cannot "override" an already-active default. Every
        # "sqlite"/"postgres"-gated crate's workspace-dependency entry in the root
        # Cargo.toml has `default-features = false` (with explicit forwarding through
        # each crate's own `sqlite`/`postgres` feature) precisely so this cascades
        # cleanly instead of silently pulling in a conflicting default via a
        # `foo.workspace = true` edge somewhere in the graph.
        run: cargo check --workspace --all-targets --no-default-features --features full,postgres
      - name: Check sqlite build (workspace, no-default-features parity)
        run: cargo check --workspace --all-targets --no-default-features --features full,sqlite

  rustdoc:
    name: Rustdoc
    needs: detect-changes
    if: needs.detect-changes.outputs.run-full-ci == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 10
    # `build.warnings` (repo-wide .cargo/config.toml) denies ALL local-package rustdoc
    # warnings, not just what RUSTDOCFLAGS explicitly denies below — overriding to "allow"
    # here keeps this job's enforcement scoped to exactly the three lints RUSTDOCFLAGS names,
    # unchanged from before this migration (#5873). `rustdoc::private_intra_doc_links` and
    # `rustdoc::redundant_explicit_links` are denied explicitly (not just left to
    # `build.warnings`, which this env var disables) so the 39-site cleanup from #5894 has a
    # forward regression guardrail in both CI and local dev, not just a one-time sweep.
    env:
      RUSTDOCFLAGS: "--deny rustdoc::broken_intra_doc_links --deny rustdoc::private_intra_doc_links --deny rustdoc::redundant_explicit_links"
      CARGO_BUILD_WARNINGS: "allow"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - uses: dtolnay/rust-toolchain@4360b52568e2003a75bf9bc1d59f33a8e3fc893c # stable
        with:
          toolchain: stable
      - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2
        with:
          shared-key: "ci"
      - name: Build docs (all features except candle)
        run: cargo doc --no-deps --workspace --features "desktop,ide,server,chat,pdf,scheduler,registry,deep-link"
      - name: Doc-tests
        run: cargo test --doc --workspace --features "desktop,ide,server,chat,pdf,scheduler,registry,deep-link"

  validate-specs:
    name: Validate Specs
    needs: detect-changes
    if: needs.detect-changes.outputs.specs-only == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 2
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      - name: Check spec structure
        run: |
          echo "Spec files found:"
          find specs -name "*.md" -type f | sort
          echo "Total: $(find specs -name "*.md" -type f | wc -l)"
          # Verify README index exists
          test -f specs/README.md || (echo "::error::specs/README.md index is missing" && exit 1)

  ci-status:
    name: CI Status
    if: always()
    needs: [detect-changes, cla, lint-shellcheck, lint-fmt, lint-clippy, msrv, build-tests, build-tests-postgres, test, integration, coverage, docker-build-and-scan, bundle-check, validate-specs, build-postgres, rustdoc, release-build, release-build-full]
    runs-on: ubuntu-latest
    timeout-minutes: 1
    steps:
      - name: Check all jobs
        run: |
          results=(
            "${{ needs.detect-changes.result }}"
            "${{ needs.cla.result }}"
            "${{ needs.lint-shellcheck.result }}"
            "${{ needs.lint-fmt.result }}"
            "${{ needs.lint-clippy.result }}"
            "${{ needs.msrv.result }}"
            "${{ needs.build-tests.result }}"
            "${{ needs.build-tests-postgres.result }}"
            "${{ needs.test.result }}"
            "${{ needs.integration.result }}"
            "${{ needs.coverage.result }}"
            "${{ needs.docker-build-and-scan.result }}"
            "${{ needs.bundle-check.result }}"
            "${{ needs.validate-specs.result }}"
            "${{ needs.build-postgres.result }}"
            "${{ needs.rustdoc.result }}"
            "${{ needs.release-build.result }}"
            "${{ needs.release-build-full.result }}"
          )
          for r in "${results[@]}"; do
            if [[ "$r" != "success" && "$r" != "skipped" ]]; then
              echo "::error::One or more jobs failed or were cancelled"
              exit 1
            fi
          done
          echo "All jobs passed"