fix: benchmark index schema allows nulls and dotted ids used by exist… #54
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: CI | |
| # SHA-pinning policy: actions are pinned to the tagged releases below; the | |
| # maintainers must replace the tags with full commit SHAs (with a version | |
| # comment) via Dependabot/Renovate PRs before any release. Never accept a | |
| # dependency-update PR that moves an action to an untagged ref. | |
| on: | |
| push: | |
| branches: [main] | |
| pull_request: | |
| workflow_dispatch: | |
| concurrency: | |
| group: ci-${{ github.workflow }}-${{ github.ref }} | |
| cancel-in-progress: true | |
| permissions: | |
| contents: read | |
| jobs: | |
| fast-tests-windows: | |
| runs-on: windows-latest | |
| timeout-minutes: 60 | |
| strategy: | |
| matrix: | |
| python-version: ['3.10', '3.11'] | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Install uv | |
| uses: astral-sh/setup-uv@v9.0.0 # v9.0.0 | |
| with: | |
| python-version: ${{ matrix.python-version }} | |
| enable-cache: true | |
| - name: Sync locked environment | |
| run: uv sync --locked --all-groups --python ${{ matrix.python-version }} | |
| - name: Run deterministic tests | |
| run: uv run --python ${{ matrix.python-version }} pytest -q | |
| - name: Parse PowerShell scripts | |
| shell: powershell | |
| run: | | |
| $parseErrors = @() | |
| Get-ChildItem -Recurse -Filter *.ps1 | ForEach-Object { | |
| $errors = $null | |
| [void][System.Management.Automation.Language.Parser]::ParseFile( | |
| $_.FullName, [ref]$null, [ref]$errors | |
| ) | |
| if ($errors) { $parseErrors += $errors } | |
| } | |
| if ($parseErrors.Count -gt 0) { | |
| $parseErrors | ForEach-Object { Write-Error $_.Message } | |
| exit 1 | |
| } | |
| - name: Benchmark table drift check | |
| run: | | |
| uv run --python ${{ matrix.python-version }} python scripts/render_benchmark_tables.py | |
| git diff --exit-code -- README.md README.zh-CN.md | |
| - name: Benchmark evidence + adapter manifest validation | |
| run: | | |
| uv run --python ${{ matrix.python-version }} python scripts/validate_benchmark_index.py | |
| uv run --python ${{ matrix.python-version }} python scripts/validate_adapter_manifest.py --adapter paddleocr-vl-1.6 --strict | |
| uv run --python ${{ matrix.python-version }} python scripts/validate_adapter_manifest.py --adapter mineru --strict | |
| fast-tests-ubuntu: | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 40 | |
| strategy: | |
| matrix: | |
| python-version: ['3.10', '3.11'] | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Install uv | |
| uses: astral-sh/setup-uv@v9.0.0 # v9.0.0 | |
| with: | |
| python-version: ${{ matrix.python-version }} | |
| enable-cache: true | |
| - name: Sync locked environment | |
| run: uv sync --locked --all-groups --python ${{ matrix.python-version }} | |
| - name: Run pure-Python tests (PowerShell tests auto-skip) | |
| run: uv run --python ${{ matrix.python-version }} pytest -q | |
| - name: "Ruff lint (safety-critical rules: E9 syntax errors, F63/F7 undefined names, F82 invalid literals)" | |
| run: | | |
| uv run --python ${{ matrix.python-version }} --with ruff ruff check --select E9,F63,F7,F82 scripts tests --exclude 'eval-infra/01-omnidocbench' | |
| - name: Pyright typecheck | |
| run: | | |
| uv run --python ${{ matrix.python-version }} --with pyright pyright scripts tests | |
| continue-on-error: true | |
| psscriptanalyzer: | |
| runs-on: windows-latest | |
| timeout-minutes: 30 | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Lint PowerShell | |
| shell: pwsh | |
| run: | | |
| Install-Module PSScriptAnalyzer -RequiredVersion 1.24.0 -Scope CurrentUser -Force -Repository PSGallery | |
| $errors = Invoke-ScriptAnalyzer -Path . -Recurse -Severity Error | |
| if ($errors) { | |
| $errors | Format-Table ScriptName, Line, RuleName, Message -AutoSize | Out-String | Write-Host | |
| exit 1 | |
| } | |
| Write-Host "PSScriptAnalyzer: 0 errors" | |
| bash-syntax: | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 10 | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Parse our Bash scripts (repo-owned .sh only; the cloned upstream OmniDocBench tree is excluded) | |
| run: | | |
| find eval-infra/02-cdm-environment eval-infra/03-scoring adapters -type f -name '*.sh' -print0 | xargs -0 -n1 bash -n | |
| - name: "ShellCheck (SC1090: dynamic source is a known false positive)" | |
| run: | | |
| sudo apt-get update -qq && sudo apt-get install -y -qq shellcheck | |
| find eval-infra/02-cdm-environment eval-infra/03-scoring adapters -type f -name '*.sh' -print0 | xargs -0 -n1 shellcheck -S warning -e SC1090 | |
| actionlint: | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 10 | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Run actionlint | |
| run: | | |
| go run github.com/rhysd/actionlint/cmd/actionlint@v1.7.7 -config-file actionlint.yaml .github/workflows/*.yml | |
| shell: bash | |
| markdown-links: | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 15 | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Install uv | |
| uses: astral-sh/setup-uv@v9.0.0 # v9.0.0 | |
| with: | |
| python-version: '3.11' | |
| enable-cache: true | |
| - name: Validate internal markdown links | |
| run: | | |
| uv sync --locked --all-groups --python 3.11 | |
| uv run --python 3.11 python -m pytest tests/test_markdown_links.py -q | |
| dependency-review: | |
| if: github.event_name == 'pull_request' | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 10 | |
| permissions: | |
| contents: read | |
| pull-requests: write | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| - name: Check Dependency Graph availability | |
| id: depgraph | |
| shell: bash | |
| run: | | |
| ENABLED=$(gh api repos/${{ github.repository }} --jq '.security_and_analysis.dependency_graph.enabled // false' 2>/dev/null || echo false) | |
| echo "enabled=$ENABLED" >> "$GITHUB_OUTPUT" | |
| if [ "$ENABLED" = "true" ]; then | |
| echo "Dependency Graph enabled - running Dependency Review" | |
| else | |
| echo "Dependency Graph not enabled on this repository - skipping Dependency Review (enable at Settings -> Code security and analysis)" | |
| fi | |
| - name: Dependency Review | |
| if: steps.depgraph.outputs.enabled == 'true' | |
| uses: actions/dependency-review-action@v5.0.0 # v5.0.0 - pin to full SHA before release | |
| with: | |
| fail-on-severity: high | |
| scorecard: | |
| if: github.event_name == 'push' | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 20 | |
| permissions: | |
| security-events: write | |
| id-token: write | |
| steps: | |
| - name: Checkout | |
| uses: actions/checkout@v7.0.1 # v7.0.1 - pin to full SHA before release | |
| with: | |
| persist-credentials: false | |
| - name: Run OpenSSF Scorecard | |
| uses: ossf/scorecard-action@v2.4.1 # v2.4.1 - pin to full SHA before release | |
| with: | |
| results_file: results.sarif | |
| results_format: sarif | |
| publish_results: true | |
| - name: Upload SARIF | |
| uses: actions/upload-artifact@v4.6.2 # v4.6.2 - pin to full SHA before release | |
| with: | |
| name: scorecard-sarif | |
| path: results.sarif | |
| # CI intentionally excludes WSL provisioning, AMD GPU inference, model/dataset | |
| # downloads, TeX/ImageMagick CDM, scoring, and benchmarks. Those require the | |
| # dated physical-machine evidence documented under docs/ (see | |
| # .github/workflows/amd-gpu.yml for the manual self-hosted HIP smoke path). |