TraceCore Nightly Distributed Acceptance #34
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: TraceCore Nightly Distributed Acceptance | |
| on: | |
| schedule: | |
| - cron: "30 3 * * *" | |
| workflow_dispatch: | |
| permissions: | |
| contents: read | |
| jobs: | |
| distributed-acceptance: | |
| name: Distributed-style batch acceptance | |
| runs-on: ubuntu-latest | |
| env: | |
| AGENT_BENCH_CONFIG: agent-bench.toml | |
| steps: | |
| - name: Checkout code | |
| uses: actions/checkout@v4 | |
| - name: Set up Python | |
| uses: actions/setup-python@v5 | |
| with: | |
| python-version: "3.12" | |
| cache: pip | |
| - name: Install dependencies | |
| run: | | |
| python -m pip install --upgrade pip | |
| pip install -e .[dev] | |
| - name: Write distributed acceptance batch file | |
| run: | | |
| python - <<'PY' | |
| import json | |
| jobs = [ | |
| {"agent": "agents/toy_agent.py", "task_ref": "filesystem_hidden_config@1", "seed": 42}, | |
| {"agent": "agents/toy_agent.py", "task_ref": "filesystem_hidden_config@1", "seed": 7}, | |
| {"agent": "agents/rate_limit_agent.py", "task_ref": "rate_limited_api@1", "seed": 11}, | |
| {"agent": "agents/rate_limit_agent.py", "task_ref": "rate_limited_api@1", "seed": 17}, | |
| {"agent": "agents/chain_agent.py", "task_ref": "rate_limited_chain@1", "seed": 7}, | |
| {"agent": "agents/chain_agent.py", "task_ref": "rate_limited_chain@1", "seed": 13}, | |
| {"agent": "agents/ops_triage_agent.py", "task_ref": "log_alert_triage@1", "seed": 21}, | |
| {"agent": "agents/ops_triage_agent.py", "task_ref": "config_drift_remediation@1", "seed": 33}, | |
| {"agent": "agents/ops_triage_agent.py", "task_ref": "incident_recovery_chain@1", "seed": 17}, | |
| {"agent": "agents/log_stream_monitor_agent.py", "task_ref": "log_stream_monitor@1", "seed": 55}, | |
| ] | |
| with open("nightly-distributed-batch.json", "w", encoding="utf-8") as fh: | |
| json.dump(jobs, fh, indent=2) | |
| PY | |
| - name: Run batch acceptance suite under strict-spec | |
| run: | | |
| python -m agent_bench.cli run batch --batch-file nightly-distributed-batch.json --workers 10 --timeout 180 --strict-spec | |
| - name: Export distributed acceptance metrics | |
| if: always() | |
| run: | | |
| python -m agent_bench.cli runs metrics --limit 40 --format json > nightly-distributed-metrics.json | |
| cat nightly-distributed-metrics.json | |
| - name: Summarize recent batch acceptance runs | |
| if: always() | |
| run: | | |
| python - <<'PY' | |
| import json | |
| from agent_bench.runner.runlog import list_runs | |
| with open("nightly-distributed-runs.json", "w", encoding="utf-8") as fh: | |
| json.dump(list_runs(limit=40), fh, indent=2) | |
| PY | |
| cat nightly-distributed-runs.json | |
| - name: Upload distributed acceptance artifacts | |
| if: always() | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| name: nightly-distributed-acceptance-${{ github.run_id }} | |
| path: | | |
| nightly-distributed-batch.json | |
| nightly-distributed-metrics.json | |
| nightly-distributed-runs.json | |
| .agent_bench/runs/ | |
| retention-days: 30 |