Preserve unavailable benchmark metrics as null #7
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: ContextBench Five Lane Score | |
| on: | |
| push: | |
| branches: [master] | |
| paths: | |
| - .github/workflows/contextbench-five-lane-score.yml | |
| - scripts/contextbench-score-five-lane-selections.mjs | |
| - scripts/contextbench-score-five-lane-artifact-selections.mjs | |
| - scripts/contextbench-build-publishable-report.mjs | |
| - scripts/contextbench-print-publishable-report.mjs | |
| - scripts/contextbench-five-lane-selections.json | |
| - tests/fixtures/contextbench-benchmark-protocol.json | |
| - tests/fixtures/contextbench-lanes.json | |
| - tests/fixtures/contextbench-task-manifest.json | |
| workflow_dispatch: | |
| permissions: | |
| contents: read | |
| actions: read | |
| jobs: | |
| five-lane-score: | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 45 | |
| env: | |
| ROOT: /tmp/contextbench-five-lane-score | |
| TASK_PAYLOADS: /tmp/contextbench-five-lane-score/task-payloads.json | |
| CHECKOUT_ROOT: /tmp/contextbench-five-lane-score-checkouts | |
| OFFICIAL_CONTEXTBENCH: /tmp/contextbench-five-lane-score/ContextBench-official | |
| TARGET_TASK_ID: SWE-Bench-Pro__go__maintenance__bugfix__4df06349 | |
| SOURCE_SELECTIONS_PATH: scripts/contextbench-five-lane-selections.json | |
| EXTERNAL_READINESS_ROOT: /tmp/contextbench-five-lane-score/external-readiness | |
| REQUIRED_LANES: raw-native,codebase-context,codebase-memory-mcp,grepai,ripgrep-lexical | |
| steps: | |
| - uses: actions/checkout@v4 | |
| - uses: pnpm/action-setup@v2 | |
| with: | |
| version: 10 | |
| - uses: actions/setup-node@v4 | |
| with: | |
| node-version: '24' | |
| cache: pnpm | |
| - uses: actions/setup-python@v5 | |
| with: | |
| python-version: '3.11' | |
| - name: Install and materialize selected Go task | |
| shell: bash | |
| run: | | |
| set -euo pipefail | |
| mkdir -p "$ROOT" "$CHECKOUT_ROOT" "$ROOT/logs" "$EXTERNAL_READINESS_ROOT" | |
| pnpm install --frozen-lockfile > "$ROOT/logs/pnpm-install.log" 2>&1 | |
| python -m pip install "tree-sitter==0.20.4" "tree-sitter-languages==1.10.2" datasets pyarrow > "$ROOT/logs/pip-install.log" 2>&1 | |
| git clone --depth 1 https://github.com/EuniAI/ContextBench.git "$OFFICIAL_CONTEXTBENCH" > "$ROOT/logs/contextbench-clone.log" 2>&1 | |
| node scripts/contextbench-runner.mjs --validate-fixtures > "$ROOT/logs/validate-fixtures.log" 2>&1 | |
| for attempt in 1 2 3; do | |
| node scripts/contextbench-select-slice.mjs --write-task-payloads --out "$TASK_PAYLOADS.all" --checkout-root "$CHECKOUT_ROOT" > "$ROOT/logs/write-payloads-$attempt.log" 2>&1 && break | |
| if [ "$attempt" = 3 ]; then exit 1; fi | |
| sleep 5 | |
| done | |
| node - <<'NODE' | |
| const fs = require('node:fs'); | |
| const payloadPath = process.env.TASK_PAYLOADS; | |
| const target = process.env.TARGET_TASK_ID; | |
| const payload = JSON.parse(fs.readFileSync(`${payloadPath}.all`, 'utf8')); | |
| const task = payload.tasks.find((candidate) => candidate.instance_id === target); | |
| if (!task) throw new Error(`target task ${target} not found`); | |
| fs.writeFileSync(payloadPath, `${JSON.stringify({ ...payload, task_count: 1, tasks: [task] }, null, 2)}\n`); | |
| NODE | |
| for attempt in 1 2 3; do | |
| node scripts/contextbench-select-slice.mjs --materialize-checkouts --payloads "$TASK_PAYLOADS" --max-tasks 1 > "$ROOT/logs/materialize-$attempt.log" 2>&1 && break | |
| if [ "$attempt" = 3 ]; then exit 1; fi | |
| sleep 5 | |
| done | |
| - name: Download GrepAI readiness artifact | |
| uses: actions/download-artifact@v4 | |
| with: | |
| name: contextbench-grepai-readiness | |
| run-id: 25643757046 | |
| github-token: ${{ github.token }} | |
| path: ${{ env.EXTERNAL_READINESS_ROOT }}/grepai | |
| - name: Download ripgrep readiness artifact | |
| uses: actions/download-artifact@v4 | |
| with: | |
| name: contextbench-ripgrep-readiness | |
| run-id: 25644197513 | |
| github-token: ${{ github.token }} | |
| path: ${{ env.EXTERNAL_READINESS_ROOT }}/ripgrep | |
| - name: Score five ready lane selections | |
| shell: bash | |
| run: node scripts/contextbench-score-five-lane-artifact-selections.mjs | |
| - name: Build publishable pilot report | |
| shell: bash | |
| run: | | |
| node scripts/contextbench-build-publishable-report.mjs \ | |
| --summary "$ROOT/summary.json" \ | |
| --protocol tests/fixtures/contextbench-benchmark-protocol.json \ | |
| --lanes tests/fixtures/contextbench-lanes.json \ | |
| --task-manifest tests/fixtures/contextbench-task-manifest.json \ | |
| --out "$ROOT/publishable-summary.json" \ | |
| --validation-out "$ROOT/publishable-validation.json" \ | |
| --humanized-out "$ROOT/humanized-summary.md" | |
| - name: Print compact publishable pilot report | |
| shell: bash | |
| run: node scripts/contextbench-print-publishable-report.mjs "$ROOT/publishable-summary.json" | |
| - name: Upload five-lane score artifacts | |
| if: always() | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| name: contextbench-five-lane-score | |
| path: /tmp/contextbench-five-lane-score | |
| retention-days: 14 |