name: PR Test (sgl-router)

# Trigger contract — modeled on `pr-test.yml`:
#
#   * No `paths:` filter at the trigger level. The workflow ALWAYS
#     fires on every PR synchronize / push, so the run record always
#     appears as a check on the PR (no more "workflow silently didn't
#     fire" mystery debugging). Path-based skip decisions are made
#     inside the `sgl-router-gate` job below, where we can log the
#     reason. Same for the `run-ci` label gate — moved off the
#     workflow trigger and into the gate job so its outcome is
#     observable.
#
#   * `push` on `main` keeps firing so post-merge runs still record
#     against the default branch.
on:
  push:
    branches: [main]
  pull_request:
    branches: [main]
    types: [opened, synchronize, reopened, labeled]
  workflow_dispatch:

concurrency:
  group: sgl-router-${{ github.event_name }}-${{ github.head_ref || github.ref_name || 'default' }}
  cancel-in-progress: true

env:
  SGLANG_IS_IN_CI: true

jobs:
  # Stage 0 — decide whether to run the heavy tiers. Always runs
  # (cheap, ubuntu-latest), and emits a single `should_run` output
  # consumed by every tier below. Reasons it might evict downstream
  # work:
  #   * `pull_request` event without the `run-ci` label (budget gate)
  #   * No files matching the sgl-router paths changed in this PR
  # Either outcome is logged on the gate job's page, so operators can
  # see *why* the tiers were skipped instead of guessing.
  sgl-router-gate:
    name: gate
    runs-on: ubuntu-latest
    outputs:
      should_run: ${{ steps.decide.outputs.should_run }}
      paths_changed: ${{ steps.paths.outputs.sgl_router }}
      has_run_ci_label: ${{ steps.label.outputs.has_run_ci }}
    steps:
      - uses: actions/checkout@v4
      # `dorny/paths-filter` computes the diff against the PR base
      # (for pull_request) or the push's `before` SHA (for push) and
      # exposes `sgl_router=true|false` on whether any tracked path
      # changed. Replaces the old workflow-level `paths:` filter so
      # the gate is observable.
      - name: Detect sgl-router path changes
        id: paths
        uses: dorny/paths-filter@v3
        with:
          filters: |
            sgl_router:
              - 'experimental/sgl-router/**'
              - '.github/workflows/pr-test-sgl-router.yml'
              - 'scripts/ci/cuda/ci_install_dependency.sh'
      # `run-ci` label gate. Cheap workflow runs (e.g. docs-only PRs
      # that happen to touch the sgl-router dir) still need the
      # opt-in label before consuming the H100 / kind tiers below.
      # `push` events on main and `workflow_dispatch` skip this gate.
      - name: Check run-ci label
        id: label
        run: |
          if [[ "${{ github.event_name }}" != "pull_request" ]]; then
            echo "has_run_ci=true" >> "$GITHUB_OUTPUT"
            echo "Non-PR event (${{ github.event_name }}); label gate bypassed."
            exit 0
          fi
          if [[ "${{ contains(github.event.pull_request.labels.*.name, 'run-ci') }}" == "true" ]]; then
            echo "has_run_ci=true" >> "$GITHUB_OUTPUT"
            echo "PR has run-ci label; downstream tiers will run."
          else
            echo "has_run_ci=false" >> "$GITHUB_OUTPUT"
            echo "::warning::PR is missing the 'run-ci' label; skipping sgl-router tiers. Apply the label to opt in."
          fi
      - name: Decide
        id: decide
        run: |
          paths='${{ steps.paths.outputs.sgl_router }}'
          label='${{ steps.label.outputs.has_run_ci }}'
          if [[ "$paths" == "true" && "$label" == "true" ]]; then
            echo "should_run=true" >> "$GITHUB_OUTPUT"
            echo "Both path filter and run-ci label match; running tiers."
          else
            echo "should_run=false" >> "$GITHUB_OUTPUT"
            echo "Skipping tiers (paths_changed=$paths, has_run_ci=$label)."
          fi

  sgl-router-lint:
    name: tier-1 — lint
    needs: sgl-router-gate
    if: needs.sgl-router-gate.outputs.should_run == 'true'
    runs-on: ubuntu-latest
    env:
      # Scoped per-job (not workflow-wide) to avoid SMG's documented breakage
      # of self-hosted-runner pip installs that compile Rust extensions.
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
    steps:
      - uses: actions/checkout@v4

      - name: Install protoc
        run: sudo bash scripts/ci/utils/install_protoc.sh

      # sccache install is best-effort: the GitHub release CDN intermittently
      # returns 5xx and the action's built-in retry is shallow. If it fails,
      # we unset RUSTC_WRAPPER below and proceed without compile caching —
      # rust-cache still covers the bulk of warm-cache wins.
      - name: Configure sccache
        id: sccache
        continue-on-error: true
        uses: mozilla-actions/sccache-action@v0.0.9
        with:
          version: "v0.12.0"
          disable_annotations: true

      - name: Disable sccache if install failed
        if: steps.sccache.outcome != 'success'
        run: |
          echo "::warning::sccache install failed (${{ steps.sccache.outcome }}); proceeding without compile caching"
          echo "RUSTC_WRAPPER=" >> "$GITHUB_ENV"

      - name: Rust cache
        uses: Swatinem/rust-cache@v2
        with:
          workspaces: experimental/sgl-router
          shared-key: "sgl-router-cache"
          cache-all-crates: true
          cache-on-failure: true
          save-if: ${{ github.ref == 'refs/heads/main' }}

      - name: cargo check
        working-directory: experimental/sgl-router
        run: cargo check --locked --workspace --all-targets

      - name: cargo clippy
        working-directory: experimental/sgl-router
        run: cargo clippy --locked --workspace --all-targets -- -D warnings

      - name: cargo fmt
        working-directory: experimental/sgl-router
        run: |
          rustup toolchain install nightly --profile minimal --component rustfmt
          cargo +nightly fmt --all -- --check

      # cargo-deny 0.19.6+ required for CVSS 4.0 parsing.
      - name: Install cargo-deny
        run: cargo install --locked cargo-deny@0.19.6

      - name: cargo deny check
        working-directory: experimental/sgl-router
        run: cargo deny check licenses bans sources advisories

      - name: Show sccache stats
        if: always() && steps.sccache.outcome == 'success'
        run: sccache --show-stats

  sgl-router-build-and-test:
    name: tier-2 — build + test
    needs: [sgl-router-gate, sgl-router-lint]
    if: needs.sgl-router-gate.outputs.should_run == 'true'
    runs-on: ubuntu-latest
    env:
      RUSTC_WRAPPER: sccache
      SCCACHE_GHA_ENABLED: "true"
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@v4

      - name: Install protoc
        run: sudo bash scripts/ci/utils/install_protoc.sh

      - name: Install OS deps (libssl, pkg-config)
        run: |
          sudo apt-get update
          sudo apt-get install -y libssl-dev pkg-config

      # sccache install is best-effort: the GitHub release CDN intermittently
      # returns 5xx and the action's built-in retry is shallow. If it fails,
      # we unset RUSTC_WRAPPER below and proceed without compile caching —
      # rust-cache still covers the bulk of warm-cache wins.
      - name: Configure sccache
        id: sccache
        continue-on-error: true
        uses: mozilla-actions/sccache-action@v0.0.9
        with:
          version: "v0.12.0"
          disable_annotations: true

      - name: Disable sccache if install failed
        if: steps.sccache.outcome != 'success'
        run: |
          echo "::warning::sccache install failed (${{ steps.sccache.outcome }}); proceeding without compile caching"
          echo "RUSTC_WRAPPER=" >> "$GITHUB_ENV"

      - name: Rust cache
        uses: Swatinem/rust-cache@v2
        with:
          workspaces: experimental/sgl-router
          shared-key: "sgl-router-cache"
          cache-all-crates: true
          cache-on-failure: true
          save-if: ${{ github.ref == 'refs/heads/main' }}

      - name: cargo build (record wall time)
        working-directory: experimental/sgl-router
        run: |
          START=$(date +%s)
          cargo build --locked --release --all-targets
          END=$(date +%s)
          BUILD_SECONDS=$((END - START))
          echo "tier_2_build_duration_seconds=${BUILD_SECONDS}" >> "$GITHUB_ENV"
          echo "::notice title=Tier-2 build duration::${BUILD_SECONDS}s"

      - name: cargo test
        working-directory: experimental/sgl-router
        # Skip the tokenizer parity_matrix test here: ubuntu-latest has
        # no HuggingFace cache, so the matrix would hard-fail (see the
        # test docstring). The e2e job runs it after pytest populates
        # the Qwen3-0.6B tokenizer.json. Filter is matched against the
        # full test path `tokenizer::parity::parity_matrix`; substring
        # `parity_matrix` is unique to that test.
        run: cargo test --locked --release --workspace -- --skip parity_matrix

      # Regenerate the cross-impl block-hash parity fixture and fail if it
      # differs from the committed file.  The Python script replicates
      # SGLang's `radix_cache.hash_page` algorithm verbatim (no
      # `import sglang`); a diff here means EITHER the Rust port's algorithm
      # drifted from the script, OR the script drifted from SGLang's actual
      # code — in both cases the operator must investigate and re-commit
      # the fixture in the same PR as the algorithm change.
      #
      # `set -euo pipefail` makes a Python-side failure (syntax error,
      # missing import) cause this step to fail instead of falling
      # through to a no-diff "pass" when no fixture was actually
      # regenerated.
      - name: kv_events hash parity fixture (drift check)
        working-directory: experimental/sgl-router
        run: |
          set -euo pipefail
          python3 tests/scripts/generate_kv_events_hash_parity.py
          if ! git diff --exit-code -- tests/fixtures/kv_events_hash_parity.json; then
            echo "::error::kv_events hash parity fixture drifted." \
              "Re-run experimental/sgl-router/tests/scripts/generate_kv_events_hash_parity.py" \
              "and commit the updated tests/fixtures/kv_events_hash_parity.json."
            exit 1
          fi

      - name: Show sccache stats
        if: always() && steps.sccache.outcome == 'success'
        run: sccache --show-stats

      - name: Summary
        if: always()
        run: |
          echo "## Tier-2 results" >> "$GITHUB_STEP_SUMMARY"
          echo "- Build duration: \`${tier_2_build_duration_seconds:-(not recorded)}\` s" >> "$GITHUB_STEP_SUMMARY"

  sgl-router-docker-build-test:
    name: tier-3 — docker (placeholder)
    needs: [sgl-router-gate, sgl-router-build-and-test]
    if: needs.sgl-router-gate.outputs.should_run == 'true'
    runs-on: ubuntu-24.04
    timeout-minutes: 15
    steps:
      - name: Placeholder
        run: |
          echo "docker-build-test not implemented yet."
          exit 0

  # tier-3 k8s integration is decoupled from tier-3 e2e: each runs on a
  # different runner type (ubuntu kind cluster vs H100), each manages its
  # own test scope (k8s_integration/ vs everything-else-in-e2e), and
  # each is allowed to fail without blocking the other. They both still
  # gate on tier-2's build+test passing — a Rust compile failure blocks
  # the GPU/cluster runners from spinning up.
  sgl-router-k8s-integration:
    name: tier-3 — k8s integration
    needs: [sgl-router-gate, sgl-router-build-and-test]
    if: needs.sgl-router-gate.outputs.should_run == 'true'
    runs-on: ubuntu-22.04
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@v4
      - name: Setup Docker buildx
        uses: docker/setup-buildx-action@v3
      - name: Install kind
        run: |
          curl -Lo /tmp/kind https://kind.sigs.k8s.io/dl/v0.24.0/kind-linux-amd64
          chmod +x /tmp/kind && sudo mv /tmp/kind /usr/local/bin/kind
      - name: Install kubectl
        run: |
          curl -Lo /tmp/kubectl https://dl.k8s.io/release/v1.31.0/bin/linux/amd64/kubectl
          chmod +x /tmp/kubectl && sudo mv /tmp/kubectl /usr/local/bin/kubectl
      - name: Build router image
        run: docker build -t sgl-router:e2e -f experimental/sgl-router/tests/e2e/k8s_integration/Dockerfile.router .
      - name: Build fake_worker image
        run: |
          docker build -t sgl-router-fake-worker:e2e \
            -f experimental/sgl-router/tests/e2e/k8s_integration/Dockerfile.fake_worker \
            experimental/sgl-router/tests/e2e/k8s_integration/
      - name: Bootstrap kind + deploy
        run: bash experimental/sgl-router/tests/e2e/k8s_integration/setup.sh
      - name: Set up Python
        run: |
          python3 -m venv /tmp/e2e-venv
          /tmp/e2e-venv/bin/pip install -r experimental/sgl-router/tests/e2e/k8s_integration/requirements.txt
      - name: Run E2E
        run: /tmp/e2e-venv/bin/pytest experimental/sgl-router/tests/e2e/k8s_integration/ -v --tb=short
      - name: Dump cluster + router diagnostics on failure
        if: failure()
        run: |
          set +e
          kubectl -n sgl-router-test get pods -o wide
          kubectl -n sgl-router-test describe pod -l app=sgl-router
          kubectl -n sgl-router-test describe pod -l app=sglang
          kubectl -n sgl-router-test logs deploy/sgl-router --tail=300
          kubectl -n sgl-router-test logs deploy/sgl-router --tail=300 --previous
          kubectl -n sgl-router-test get events --sort-by=.lastTimestamp
          kubectl -n sgl-router-test get endpointslices -o wide
          kubectl -n sgl-router-test get svc
          true

  sgl-router-e2e:
    name: tier-3 — e2e
    needs: [sgl-router-gate, sgl-router-build-and-test]
    if: needs.sgl-router-gate.outputs.should_run == 'true'
    runs-on: 2-gpu-h100
    timeout-minutes: 45
    steps:
      - uses: actions/checkout@v4
      - name: Install Rust toolchain
        run: bash scripts/ci/utils/install_rust_protoc.sh
      - name: Rust cache
        uses: Swatinem/rust-cache@v2
        with:
          workspaces: experimental/sgl-router
          shared-key: "sgl-router-cache"
      - name: cargo build (release)
        working-directory: experimental/sgl-router
        run: |
          source "$HOME/.cargo/env"
          cargo build --locked --release --workspace

      # Install SGLang from the local checkout in editable mode (no PyPI
      # version pin) — mirrors `pr-test-rust.yml` so the router e2e runs
      # against whatever SGLang version is on the branch. Replaces the
      # previous `pip install sglang[srt]==X.Y.Z` pin, which forced the
      # e2e to test against a stale wheel and would silently miss any
      # new ServerArgs field (e.g. `disaggregation_mode` /
      # `disaggregation_bootstrap_port`) added in the branch.
      - name: Install SGLang dependencies
        run: |
          bash scripts/ci/cuda/ci_install_dependency.sh

      - name: Install e2e test dependencies
        run: |
          python3 -m pip install -r experimental/sgl-router/tests/e2e/requirements.txt

      # IMPORTANT: --ignore=tests/e2e/k8s_integration. The k8s integration
      # suite is owned by the `sgl-router-k8s-integration` job which has
      # kind + kubectl installed and a real cluster running. The e2e GPU
      # runner has neither, so pytest's recursive collection of
      # tests/e2e/ would ERROR every k8s test at setup. Both tiers must
      # be allowed to fail independently — k8s flakes (cluster boot,
      # image pull) must not surface as e2e failures on H100, and e2e
      # flakes (model load, HF auth) must not surface as k8s failures.
      - name: Run e2e
        working-directory: experimental/sgl-router
        env:
          HF_TOKEN: ${{ secrets.HF_TOKEN }}
        run: |
          python3 -m pytest tests/e2e/ -v -s --tb=short \
            --ignore=tests/e2e/k8s_integration

      - name: Tokenizer parity matrix (uses HF cache populated by e2e)
        working-directory: experimental/sgl-router
        env:
          HF_TOKEN: ${{ secrets.HF_TOKEN }}
        run: |
          source "$HOME/.cargo/env"
          cargo test --locked --release --test component tokenizer::parity

  sgl-router-finish:
    name: finish
    needs:
      - sgl-router-gate
      - sgl-router-lint
      - sgl-router-build-and-test
      - sgl-router-docker-build-test
      - sgl-router-k8s-integration
      - sgl-router-e2e
    # `always()` lets `finish` run after upstream `skipped` outcomes
    # (when the gate decided to skip). The downstream-skipped path is
    # an explicit success — the workflow was correctly bypassed for a
    # PR that didn't need full CI. The downstream-failure /
    # downstream-cancelled path is a real fail. We also accept the
    # gate's `skipped` (impossible today, kept for symmetry) and
    # require the gate itself to have succeeded.
    if: |
      always() &&
      needs.sgl-router-gate.result == 'success' &&
      !contains(needs.*.result, 'failure') &&
      !contains(needs.*.result, 'cancelled')
    runs-on: ubuntu-latest
    steps:
      - name: All required checks completed
        run: echo "OK"