agent-hitch 0.2.6 → 0.2.8
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +30 -3
- package/README.zh-CN.md +14 -2
- package/dist/bin/hitch.js +25 -3
- package/dist/bin/hitch.js.map +1 -1
- package/dist/scripts/canary-benchmark.js +54 -0
- package/dist/scripts/canary-benchmark.js.map +1 -0
- package/dist/scripts/canary-buildkit-secrets.js +166 -0
- package/dist/scripts/canary-buildkit-secrets.js.map +1 -0
- package/dist/scripts/canary-collision-domains.js +77 -0
- package/dist/scripts/canary-collision-domains.js.map +1 -0
- package/dist/scripts/canary-docker-image-observation.js +181 -0
- package/dist/scripts/canary-docker-image-observation.js.map +1 -0
- package/dist/scripts/canary-eval-scheduler-throughput.js +47 -0
- package/dist/scripts/canary-eval-scheduler-throughput.js.map +1 -0
- package/dist/scripts/canary-gpu-hardware.js +95 -0
- package/dist/scripts/canary-gpu-hardware.js.map +1 -0
- package/dist/scripts/canary-harbor-load.js +365 -0
- package/dist/scripts/canary-harbor-load.js.map +1 -0
- package/dist/scripts/canary-packaged-harness.js +75 -0
- package/dist/scripts/canary-packaged-harness.js.map +1 -0
- package/dist/scripts/canary-resource-load.js +188 -0
- package/dist/scripts/canary-resource-load.js.map +1 -0
- package/dist/scripts/check-architecture.js +10 -5
- package/dist/scripts/check-architecture.js.map +1 -1
- package/dist/scripts/docker-reaper-crash-child.js +21 -0
- package/dist/scripts/docker-reaper-crash-child.js.map +1 -0
- package/dist/src/adapters/catalog.js +3 -1
- package/dist/src/adapters/catalog.js.map +1 -1
- package/dist/src/adapters/discovery.js +1 -1
- package/dist/src/adapters/discovery.js.map +1 -1
- package/dist/src/adapters/providers/claude.js +6 -0
- package/dist/src/adapters/providers/claude.js.map +1 -1
- package/dist/src/adapters/providers/codex-auth.js +29 -0
- package/dist/src/adapters/providers/codex-auth.js.map +1 -0
- package/dist/src/adapters/providers/codex.js +29 -6
- package/dist/src/adapters/providers/codex.js.map +1 -1
- package/dist/src/adapters/providers/deepseek.js +6 -0
- package/dist/src/adapters/providers/deepseek.js.map +1 -1
- package/dist/src/adapters/providers/model-call.js +26 -0
- package/dist/src/adapters/providers/model-call.js.map +1 -0
- package/dist/src/adapters/providers/opencode.js +6 -0
- package/dist/src/adapters/providers/opencode.js.map +1 -1
- package/dist/src/adapters/providers/pi.js +6 -0
- package/dist/src/adapters/providers/pi.js.map +1 -1
- package/dist/src/artifacts/index.js +1 -0
- package/dist/src/artifacts/index.js.map +1 -1
- package/dist/src/artifacts/preparer.js +7 -2
- package/dist/src/artifacts/preparer.js.map +1 -1
- package/dist/src/backends/harbor/agent-budget.js +31 -0
- package/dist/src/backends/harbor/agent-budget.js.map +1 -0
- package/dist/src/backends/harbor/backend.js +59 -107
- package/dist/src/backends/harbor/backend.js.map +1 -1
- package/dist/src/backends/harbor/environment-config.js +92 -0
- package/dist/src/backends/harbor/environment-config.js.map +1 -0
- package/dist/src/backends/harbor/index.js +6 -1
- package/dist/src/backends/harbor/index.js.map +1 -1
- package/dist/src/backends/harbor/model-proxy-config.js +23 -0
- package/dist/src/backends/harbor/model-proxy-config.js.map +1 -0
- package/dist/src/backends/harbor/process.js +177 -0
- package/dist/src/backends/harbor/process.js.map +1 -0
- package/dist/src/backends/harbor/regrade.js +77 -0
- package/dist/src/backends/harbor/regrade.js.map +1 -0
- package/dist/src/backends/harbor/runtime-toolchain.js +30 -0
- package/dist/src/backends/harbor/runtime-toolchain.js.map +1 -0
- package/dist/src/backends/harbor/supervisor.js +74 -0
- package/dist/src/backends/harbor/supervisor.js.map +1 -0
- package/dist/src/backends/harbor/task-resources.js +206 -0
- package/dist/src/backends/harbor/task-resources.js.map +1 -0
- package/dist/src/backends/harbor/tools.js +6 -29
- package/dist/src/backends/harbor/tools.js.map +1 -1
- package/dist/src/backends/index.js +5 -1
- package/dist/src/backends/index.js.map +1 -1
- package/dist/src/benchmarks/index.js +3 -0
- package/dist/src/benchmarks/index.js.map +1 -0
- package/dist/src/benchmarks/loader.js +271 -0
- package/dist/src/benchmarks/loader.js.map +1 -0
- package/dist/src/benchmarks/metrics.js +16 -0
- package/dist/src/benchmarks/metrics.js.map +1 -0
- package/dist/src/benchmarks/toml.js +13 -0
- package/dist/src/benchmarks/toml.js.map +1 -0
- package/dist/src/benchmarks/validation.js +242 -0
- package/dist/src/benchmarks/validation.js.map +1 -0
- package/dist/src/cli/arguments.js +6 -4
- package/dist/src/cli/arguments.js.map +1 -1
- package/dist/src/cli/commands/benchmark.js +22 -0
- package/dist/src/cli/commands/benchmark.js.map +1 -0
- package/dist/src/cli/commands/capabilities.js +18 -0
- package/dist/src/cli/commands/capabilities.js.map +1 -0
- package/dist/src/cli/commands/daemon.js +135 -8
- package/dist/src/cli/commands/daemon.js.map +1 -1
- package/dist/src/cli/commands/eval.js +228 -6
- package/dist/src/cli/commands/eval.js.map +1 -1
- package/dist/src/cli/commands/images.js +40 -0
- package/dist/src/cli/commands/images.js.map +1 -0
- package/dist/src/cli/commands/run.js +42 -17
- package/dist/src/cli/commands/run.js.map +1 -1
- package/dist/src/cli/commands/runs.js +25 -3
- package/dist/src/cli/commands/runs.js.map +1 -1
- package/dist/src/cli/commands/trajectory.js +96 -5
- package/dist/src/cli/commands/trajectory.js.map +1 -1
- package/dist/src/cli/commands/verifier.js +36 -0
- package/dist/src/cli/commands/verifier.js.map +1 -0
- package/dist/src/cli/commands/worker.js +79 -0
- package/dist/src/cli/commands/worker.js.map +1 -0
- package/dist/src/cli/main.js +10 -0
- package/dist/src/cli/main.js.map +1 -1
- package/dist/src/cli/output.js +84 -4
- package/dist/src/cli/output.js.map +1 -1
- package/dist/src/control-plane/build-admission.js +69 -0
- package/dist/src/control-plane/build-admission.js.map +1 -0
- package/dist/src/control-plane/build-records.js +6 -0
- package/dist/src/control-plane/build-records.js.map +1 -0
- package/dist/src/control-plane/collisions.js +40 -0
- package/dist/src/control-plane/collisions.js.map +1 -0
- package/dist/src/control-plane/eval-control-state.js +22 -0
- package/dist/src/control-plane/eval-control-state.js.map +1 -0
- package/dist/src/control-plane/eval-control-work.js +44 -0
- package/dist/src/control-plane/eval-control-work.js.map +1 -0
- package/dist/src/control-plane/eval-image-resolution.js +52 -0
- package/dist/src/control-plane/eval-image-resolution.js.map +1 -0
- package/dist/src/control-plane/eval-image-services.js +30 -0
- package/dist/src/control-plane/eval-image-services.js.map +1 -0
- package/dist/src/control-plane/eval-records.js +252 -0
- package/dist/src/control-plane/eval-records.js.map +1 -0
- package/dist/src/control-plane/eval-recovery.js +181 -0
- package/dist/src/control-plane/eval-recovery.js.map +1 -0
- package/dist/src/control-plane/eval-scheduler.js +450 -0
- package/dist/src/control-plane/eval-scheduler.js.map +1 -0
- package/dist/src/control-plane/index.js +20 -0
- package/dist/src/control-plane/index.js.map +1 -0
- package/dist/src/control-plane/local-worker.js +49 -0
- package/dist/src/control-plane/local-worker.js.map +1 -0
- package/dist/src/control-plane/model-capture-planning.js +12 -0
- package/dist/src/control-plane/model-capture-planning.js.map +1 -0
- package/dist/src/control-plane/remote-result-transport.js +175 -0
- package/dist/src/control-plane/remote-result-transport.js.map +1 -0
- package/dist/src/control-plane/remote-work-coordinator.js +333 -0
- package/dist/src/control-plane/remote-work-coordinator.js.map +1 -0
- package/dist/src/control-plane/remote-work-inputs.js +266 -0
- package/dist/src/control-plane/remote-work-inputs.js.map +1 -0
- package/dist/src/control-plane/remote-work-item.js +62 -0
- package/dist/src/control-plane/remote-work-item.js.map +1 -0
- package/dist/src/control-plane/remote-work-recovery.js +220 -0
- package/dist/src/control-plane/remote-work-recovery.js.map +1 -0
- package/dist/src/control-plane/remote-worker-artifacts.js +135 -0
- package/dist/src/control-plane/remote-worker-artifacts.js.map +1 -0
- package/dist/src/control-plane/remote-worker-client.js +202 -0
- package/dist/src/control-plane/remote-worker-client.js.map +1 -0
- package/dist/src/control-plane/remote-worker-configuration.js +34 -0
- package/dist/src/control-plane/remote-worker-configuration.js.map +1 -0
- package/dist/src/control-plane/remote-worker-credentials.js +58 -0
- package/dist/src/control-plane/remote-worker-credentials.js.map +1 -0
- package/dist/src/control-plane/remote-worker-protocol.js +489 -0
- package/dist/src/control-plane/remote-worker-protocol.js.map +1 -0
- package/dist/src/control-plane/remote-worker-runner.js +214 -0
- package/dist/src/control-plane/remote-worker-runner.js.map +1 -0
- package/dist/src/control-plane/remote-workers.js +377 -0
- package/dist/src/control-plane/remote-workers.js.map +1 -0
- package/dist/src/control-plane/rerun-scheduler.js +448 -0
- package/dist/src/control-plane/rerun-scheduler.js.map +1 -0
- package/dist/src/control-plane/rerun-submission.js +62 -0
- package/dist/src/control-plane/rerun-submission.js.map +1 -0
- package/dist/src/control-plane/resources.js +182 -0
- package/dist/src/control-plane/resources.js.map +1 -0
- package/dist/src/control-plane/scheduler-eval-entry.js +32 -0
- package/dist/src/control-plane/scheduler-eval-entry.js.map +1 -0
- package/dist/src/control-plane/synthetic-result.js +21 -0
- package/dist/src/control-plane/synthetic-result.js.map +1 -0
- package/dist/src/control-plane/work-admission.js +23 -0
- package/dist/src/control-plane/work-admission.js.map +1 -0
- package/dist/src/control-plane/work-dispatcher.js +237 -0
- package/dist/src/control-plane/work-dispatcher.js.map +1 -0
- package/dist/src/controller-runtime/hash.js +19 -3
- package/dist/src/controller-runtime/hash.js.map +1 -1
- package/dist/src/controller-runtime/index.js +1 -1
- package/dist/src/controller-runtime/index.js.map +1 -1
- package/dist/src/controller-runtime/store.js +15 -6
- package/dist/src/controller-runtime/store.js.map +1 -1
- package/dist/src/daemon/auth.js +5 -0
- package/dist/src/daemon/auth.js.map +1 -1
- package/dist/src/daemon/client.js.map +1 -1
- package/dist/src/daemon/event-stream.js +46 -0
- package/dist/src/daemon/event-stream.js.map +1 -0
- package/dist/src/daemon/health.js +66 -0
- package/dist/src/daemon/health.js.map +1 -0
- package/dist/src/daemon/launcher.js +28 -1
- package/dist/src/daemon/launcher.js.map +1 -1
- package/dist/src/daemon/logging.js +23 -0
- package/dist/src/daemon/logging.js.map +1 -0
- package/dist/src/daemon/scheduler.js +26 -5
- package/dist/src/daemon/scheduler.js.map +1 -1
- package/dist/src/daemon/server.js +255 -98
- package/dist/src/daemon/server.js.map +1 -1
- package/dist/src/daemon/telemetry.js +163 -0
- package/dist/src/daemon/telemetry.js.map +1 -0
- package/dist/src/daemon/worker-routes.js +219 -0
- package/dist/src/daemon/worker-routes.js.map +1 -0
- package/dist/src/domain/benchmarks.js +2 -0
- package/dist/src/domain/benchmarks.js.map +1 -0
- package/dist/src/domain/bundles.js +2 -0
- package/dist/src/domain/bundles.js.map +1 -0
- package/dist/src/domain/control-plane.js +2 -0
- package/dist/src/domain/control-plane.js.map +1 -0
- package/dist/src/domain/execution-evidence.js +2 -0
- package/dist/src/domain/execution-evidence.js.map +1 -0
- package/dist/src/domain/execution-plan.js +2 -0
- package/dist/src/domain/execution-plan.js.map +1 -0
- package/dist/src/domain/images.js +2 -0
- package/dist/src/domain/images.js.map +1 -0
- package/dist/src/domain/index.js +13 -0
- package/dist/src/domain/index.js.map +1 -1
- package/dist/src/domain/interactions.js +2 -0
- package/dist/src/domain/interactions.js.map +1 -0
- package/dist/src/domain/providers.js +2 -0
- package/dist/src/domain/providers.js.map +1 -0
- package/dist/src/domain/resources.js +2 -0
- package/dist/src/domain/resources.js.map +1 -0
- package/dist/src/domain/runs.js.map +1 -1
- package/dist/src/domain/validation.js +11 -4
- package/dist/src/domain/validation.js.map +1 -1
- package/dist/src/domain/verifier-evidence-validation.js +214 -0
- package/dist/src/domain/verifier-evidence-validation.js.map +1 -0
- package/dist/src/domain/verifier-evidence.js +3 -0
- package/dist/src/domain/verifier-evidence.js.map +1 -0
- package/dist/src/domain/verifier-score-contract.js +232 -0
- package/dist/src/domain/verifier-score-contract.js.map +1 -0
- package/dist/src/domain/workers.js +2 -0
- package/dist/src/domain/workers.js.map +1 -0
- package/dist/src/evals/benchmark-adapter-manifest.js +219 -0
- package/dist/src/evals/benchmark-adapter-manifest.js.map +1 -0
- package/dist/src/evals/benchmark-candidate.js +36 -0
- package/dist/src/evals/benchmark-candidate.js.map +1 -0
- package/dist/src/evals/benchmark-run.js +166 -0
- package/dist/src/evals/benchmark-run.js.map +1 -0
- package/dist/src/evals/collect-only-rerun.js +183 -0
- package/dist/src/evals/collect-only-rerun.js.map +1 -0
- package/dist/src/evals/directory.js +30 -0
- package/dist/src/evals/directory.js.map +1 -0
- package/dist/src/evals/docker-engine-stats.js +89 -0
- package/dist/src/evals/docker-engine-stats.js.map +1 -0
- package/dist/src/evals/docker-ownership.js +51 -0
- package/dist/src/evals/docker-ownership.js.map +1 -0
- package/dist/src/evals/docker-reaper.js +144 -0
- package/dist/src/evals/docker-reaper.js.map +1 -0
- package/dist/src/evals/docker-resource-observer.js +207 -0
- package/dist/src/evals/docker-resource-observer.js.map +1 -0
- package/dist/src/evals/duration-estimator.js +126 -0
- package/dist/src/evals/duration-estimator.js.map +1 -0
- package/dist/src/evals/environment-image-planning.js +164 -0
- package/dist/src/evals/environment-image-planning.js.map +1 -0
- package/dist/src/evals/eval-artifact-planning.js +56 -0
- package/dist/src/evals/eval-artifact-planning.js.map +1 -0
- package/dist/src/evals/eval-finalization.js +10 -0
- package/dist/src/evals/eval-finalization.js.map +1 -0
- package/dist/src/evals/eval-lifecycle-events.js +29 -0
- package/dist/src/evals/eval-lifecycle-events.js.map +1 -0
- package/dist/src/evals/eval-logical-plan.js +51 -0
- package/dist/src/evals/eval-logical-plan.js.map +1 -0
- package/dist/src/evals/eval-result-builder.js +81 -0
- package/dist/src/evals/eval-result-builder.js.map +1 -0
- package/dist/src/evals/events.js +53 -5
- package/dist/src/evals/events.js.map +1 -1
- package/dist/src/evals/evolution-baseline.js +191 -0
- package/dist/src/evals/evolution-baseline.js.map +1 -0
- package/dist/src/evals/execution-evidence.js +116 -0
- package/dist/src/evals/execution-evidence.js.map +1 -0
- package/dist/src/evals/execution-leases.js +342 -0
- package/dist/src/evals/execution-leases.js.map +1 -0
- package/dist/src/evals/execution-plan-artifacts.js +72 -0
- package/dist/src/evals/execution-plan-artifacts.js.map +1 -0
- package/dist/src/evals/execution-plan-images.js +69 -0
- package/dist/src/evals/execution-plan-images.js.map +1 -0
- package/dist/src/evals/execution-plan-resources.js +174 -0
- package/dist/src/evals/execution-plan-resources.js.map +1 -0
- package/dist/src/evals/execution-plan.js +451 -0
- package/dist/src/evals/execution-plan.js.map +1 -0
- package/dist/src/evals/failure-classifier.js +61 -0
- package/dist/src/evals/failure-classifier.js.map +1 -0
- package/dist/src/evals/harbor-artifact-builder-image.js +99 -0
- package/dist/src/evals/harbor-artifact-builder-image.js.map +1 -0
- package/dist/src/evals/harbor-artifact-builder.js +358 -0
- package/dist/src/evals/harbor-artifact-builder.js.map +1 -0
- package/dist/src/evals/harbor-bridge-error.js +58 -3
- package/dist/src/evals/harbor-bridge-error.js.map +1 -1
- package/dist/src/evals/harbor-node-runtime.js +146 -0
- package/dist/src/evals/harbor-node-runtime.js.map +1 -0
- package/dist/src/evals/harbor-phase-timings.js +60 -0
- package/dist/src/evals/harbor-phase-timings.js.map +1 -0
- package/dist/src/evals/index.js +40 -2
- package/dist/src/evals/index.js.map +1 -1
- package/dist/src/evals/infrastructure-retry.js +208 -84
- package/dist/src/evals/infrastructure-retry.js.map +1 -1
- package/dist/src/evals/interaction-capture-import.js +49 -0
- package/dist/src/evals/interaction-capture-import.js.map +1 -0
- package/dist/src/evals/local-docker-provider.js +372 -0
- package/dist/src/evals/local-docker-provider.js.map +1 -0
- package/dist/src/evals/local-eval-planning.js +80 -0
- package/dist/src/evals/local-eval-planning.js.map +1 -0
- package/dist/src/evals/model-capture-plan.js +22 -0
- package/dist/src/evals/model-capture-plan.js.map +1 -0
- package/dist/src/evals/model-capture-runtime.js +114 -0
- package/dist/src/evals/model-capture-runtime.js.map +1 -0
- package/dist/src/evals/model-proxy-runtime-state.js +53 -0
- package/dist/src/evals/model-proxy-runtime-state.js.map +1 -0
- package/dist/src/evals/native-phase-evidence.js +440 -0
- package/dist/src/evals/native-phase-evidence.js.map +1 -0
- package/dist/src/evals/physical-retry-work.js +31 -0
- package/dist/src/evals/physical-retry-work.js.map +1 -0
- package/dist/src/evals/planned-execution-support.js +66 -0
- package/dist/src/evals/planned-execution-support.js.map +1 -0
- package/dist/src/evals/planned-execution.js +469 -0
- package/dist/src/evals/planned-execution.js.map +1 -0
- package/dist/src/evals/planned-progress-publisher.js +62 -0
- package/dist/src/evals/planned-progress-publisher.js.map +1 -0
- package/dist/src/evals/planned-remote-execution.js +40 -0
- package/dist/src/evals/planned-remote-execution.js.map +1 -0
- package/dist/src/evals/planned-retry-execution.js +174 -0
- package/dist/src/evals/planned-retry-execution.js.map +1 -0
- package/dist/src/evals/planned-retry-lifecycle.js +188 -0
- package/dist/src/evals/planned-retry-lifecycle.js.map +1 -0
- package/dist/src/evals/preparation-rerun.js +225 -0
- package/dist/src/evals/preparation-rerun.js.map +1 -0
- package/dist/src/evals/prepared-harness.js +76 -0
- package/dist/src/evals/prepared-harness.js.map +1 -0
- package/dist/src/evals/progress.js +52 -12
- package/dist/src/evals/progress.js.map +1 -1
- package/dist/src/evals/records.js +20 -3
- package/dist/src/evals/records.js.map +1 -1
- package/dist/src/evals/recovery.js +242 -0
- package/dist/src/evals/recovery.js.map +1 -0
- package/dist/src/evals/regrade-evidence.js +77 -0
- package/dist/src/evals/regrade-evidence.js.map +1 -0
- package/dist/src/evals/remote-infrastructure-retry.js +165 -0
- package/dist/src/evals/remote-infrastructure-retry.js.map +1 -0
- package/dist/src/evals/request.js +21 -3
- package/dist/src/evals/request.js.map +1 -1
- package/dist/src/evals/rerun-inputs.js +108 -0
- package/dist/src/evals/rerun-inputs.js.map +1 -0
- package/dist/src/evals/rerun-slots.js +4 -3
- package/dist/src/evals/rerun-slots.js.map +1 -1
- package/dist/src/evals/rerun-types.js +45 -0
- package/dist/src/evals/rerun-types.js.map +1 -0
- package/dist/src/evals/rerun.js +133 -88
- package/dist/src/evals/rerun.js.map +1 -1
- package/dist/src/evals/result-helpers.js +78 -0
- package/dist/src/evals/result-helpers.js.map +1 -0
- package/dist/src/evals/resume-state.js +36 -0
- package/dist/src/evals/resume-state.js.map +1 -0
- package/dist/src/evals/retry-backoff.js +20 -0
- package/dist/src/evals/retry-backoff.js.map +1 -0
- package/dist/src/evals/retry-state.js +195 -0
- package/dist/src/evals/retry-state.js.map +1 -0
- package/dist/src/evals/scheduler-metrics.js +95 -0
- package/dist/src/evals/scheduler-metrics.js.map +1 -0
- package/dist/src/evals/scheduler-trace-replay.js +107 -0
- package/dist/src/evals/scheduler-trace-replay.js.map +1 -0
- package/dist/src/evals/service-types.js +2 -0
- package/dist/src/evals/service-types.js.map +1 -0
- package/dist/src/evals/service.js +331 -347
- package/dist/src/evals/service.js.map +1 -1
- package/dist/src/evals/task-resources.js +209 -0
- package/dist/src/evals/task-resources.js.map +1 -0
- package/dist/src/evals/trial-environment-evidence.js +87 -0
- package/dist/src/evals/trial-environment-evidence.js.map +1 -0
- package/dist/src/evals/trial-execution-evidence.js +13 -0
- package/dist/src/evals/trial-execution-evidence.js.map +1 -0
- package/dist/src/evals/trial-import-identity.js +31 -0
- package/dist/src/evals/trial-import-identity.js.map +1 -0
- package/dist/src/evals/trial-import.js +183 -188
- package/dist/src/evals/trial-import.js.map +1 -1
- package/dist/src/evals/trial-publication-recovery.js +144 -0
- package/dist/src/evals/trial-publication-recovery.js.map +1 -0
- package/dist/src/evals/trial-publication.js +25 -0
- package/dist/src/evals/trial-publication.js.map +1 -0
- package/dist/src/evals/trial-reference-validation.js +49 -0
- package/dist/src/evals/trial-reference-validation.js.map +1 -0
- package/dist/src/evals/verifier-artifacts.js +317 -0
- package/dist/src/evals/verifier-artifacts.js.map +1 -0
- package/dist/src/evals/verifier-eligibility.js +25 -0
- package/dist/src/evals/verifier-eligibility.js.map +1 -0
- package/dist/src/evals/verifier-only-rerun.js +210 -0
- package/dist/src/evals/verifier-only-rerun.js.map +1 -0
- package/dist/src/evals/verifier-runtime.js +61 -0
- package/dist/src/evals/verifier-runtime.js.map +1 -0
- package/dist/src/evals/verifier-score-artifacts.js +119 -0
- package/dist/src/evals/verifier-score-artifacts.js.map +1 -0
- package/dist/src/evals/work-item-artifacts.js +23 -0
- package/dist/src/evals/work-item-artifacts.js.map +1 -0
- package/dist/src/foundation/config.js +17 -0
- package/dist/src/foundation/config.js.map +1 -1
- package/dist/src/foundation/contained-file.js +85 -0
- package/dist/src/foundation/contained-file.js.map +1 -0
- package/dist/src/foundation/credential-redaction.js +162 -0
- package/dist/src/foundation/credential-redaction.js.map +1 -0
- package/dist/src/foundation/environment-image-references.js +59 -0
- package/dist/src/foundation/environment-image-references.js.map +1 -0
- package/dist/src/foundation/executable.js +45 -5
- package/dist/src/foundation/executable.js.map +1 -1
- package/dist/src/foundation/fs.js +2 -1
- package/dist/src/foundation/fs.js.map +1 -1
- package/dist/src/foundation/index.js +6 -2
- package/dist/src/foundation/index.js.map +1 -1
- package/dist/src/foundation/process-identity.js +69 -0
- package/dist/src/foundation/process-identity.js.map +1 -0
- package/dist/src/images/context.js +47 -0
- package/dist/src/images/context.js.map +1 -0
- package/dist/src/images/docker-buildkit.js +137 -0
- package/dist/src/images/docker-buildkit.js.map +1 -0
- package/dist/src/images/dockerfile.js +67 -0
- package/dist/src/images/dockerfile.js.map +1 -0
- package/dist/src/images/gc.js +261 -0
- package/dist/src/images/gc.js.map +1 -0
- package/dist/src/images/index.js +10 -0
- package/dist/src/images/index.js.map +1 -0
- package/dist/src/images/manifest.js +109 -0
- package/dist/src/images/manifest.js.map +1 -0
- package/dist/src/images/ownership.js +5 -0
- package/dist/src/images/ownership.js.map +1 -0
- package/dist/src/images/records.js +32 -0
- package/dist/src/images/records.js.map +1 -0
- package/dist/src/images/registry.js +154 -0
- package/dist/src/images/registry.js.map +1 -0
- package/dist/src/images/service.js +224 -0
- package/dist/src/images/service.js.map +1 -0
- package/dist/src/model-access/capture.js +255 -0
- package/dist/src/model-access/capture.js.map +1 -0
- package/dist/src/model-access/index.js +5 -0
- package/dist/src/model-access/index.js.map +1 -0
- package/dist/src/model-access/policy.js +49 -0
- package/dist/src/model-access/policy.js.map +1 -0
- package/dist/src/model-access/proxy.js +336 -0
- package/dist/src/model-access/proxy.js.map +1 -0
- package/dist/src/model-access/records.js +167 -0
- package/dist/src/model-access/records.js.map +1 -0
- package/dist/src/runs/adapter-process.js +12 -0
- package/dist/src/runs/adapter-process.js.map +1 -0
- package/dist/src/runs/bundle.js +450 -0
- package/dist/src/runs/bundle.js.map +1 -0
- package/dist/src/runs/events.js +12 -1
- package/dist/src/runs/events.js.map +1 -1
- package/dist/src/runs/executor.js +46 -45
- package/dist/src/runs/executor.js.map +1 -1
- package/dist/src/runs/finalizer.js +17 -0
- package/dist/src/runs/finalizer.js.map +1 -1
- package/dist/src/runs/index.js +8 -1
- package/dist/src/runs/index.js.map +1 -1
- package/dist/src/runs/manifest.js +5 -5
- package/dist/src/runs/manifest.js.map +1 -1
- package/dist/src/runs/phase-bundle.js +86 -0
- package/dist/src/runs/phase-bundle.js.map +1 -0
- package/dist/src/runs/phase-cancellation.js +107 -0
- package/dist/src/runs/phase-cancellation.js.map +1 -0
- package/dist/src/runs/phase-group.js +113 -0
- package/dist/src/runs/phase-group.js.map +1 -0
- package/dist/src/runs/query.js +1 -1
- package/dist/src/runs/query.js.map +1 -1
- package/dist/src/runs/records.js +3 -0
- package/dist/src/runs/records.js.map +1 -1
- package/dist/src/runs/request.js +10 -3
- package/dist/src/runs/request.js.map +1 -1
- package/dist/src/runs/training-candidate.js +161 -0
- package/dist/src/runs/training-candidate.js.map +1 -0
- package/dist/src/runs/verifier-evidence-redaction.js +61 -0
- package/dist/src/runs/verifier-evidence-redaction.js.map +1 -0
- package/dist/src/runs/verifier-evidence.js +459 -0
- package/dist/src/runs/verifier-evidence.js.map +1 -0
- package/dist/src/runs/verifier-structured-evidence.js +83 -0
- package/dist/src/runs/verifier-structured-evidence.js.map +1 -0
- package/dist/src/trajectories/analysis.js +288 -0
- package/dist/src/trajectories/analysis.js.map +1 -0
- package/dist/src/trajectories/chunk-projection.js +148 -0
- package/dist/src/trajectories/chunk-projection.js.map +1 -0
- package/dist/src/trajectories/content-projection.js +389 -0
- package/dist/src/trajectories/content-projection.js.map +1 -0
- package/dist/src/trajectories/dsh-chunk-contract.js +176 -0
- package/dist/src/trajectories/dsh-chunk-contract.js.map +1 -0
- package/dist/src/trajectories/dsh-contract.js +284 -0
- package/dist/src/trajectories/dsh-contract.js.map +1 -0
- package/dist/src/trajectories/events-chunk-drill.js +95 -0
- package/dist/src/trajectories/events-chunk-drill.js.map +1 -0
- package/dist/src/trajectories/events-page.js +367 -0
- package/dist/src/trajectories/events-page.js.map +1 -0
- package/dist/src/trajectories/index.js +3 -0
- package/dist/src/trajectories/index.js.map +1 -1
- package/dist/src/trajectories/provider-capture.js +15 -24
- package/dist/src/trajectories/provider-capture.js.map +1 -1
- package/dist/src/trajectories/providers/deepseek.js +3 -3
- package/dist/src/trajectories/providers/deepseek.js.map +1 -1
- package/dist/src/trajectories/request-attempt.js +47 -0
- package/dist/src/trajectories/request-attempt.js.map +1 -0
- package/dist/src/trajectories/stream-reader.js +311 -0
- package/dist/src/trajectories/stream-reader.js.map +1 -0
- package/dist/src/trajectories/surface-fold.js +126 -0
- package/dist/src/trajectories/surface-fold.js.map +1 -0
- package/dist/src/workers/index.js +3 -0
- package/dist/src/workers/index.js.map +1 -0
- package/dist/src/workers/remote-harbor-work-spec.js +97 -0
- package/dist/src/workers/remote-harbor-work-spec.js.map +1 -0
- package/dist/src/workers/remote-harbor-worker.js +223 -0
- package/dist/src/workers/remote-harbor-worker.js.map +1 -0
- package/docs/schemas/benchmark-hook-request-v1.schema.json +65 -0
- package/docs/schemas/benchmark-hook-response-v1.schema.json +75 -0
- package/docs/schemas/benchmark-hook-v1.schema.json +33 -0
- package/docs/schemas/benchmark-lock-v1.schema.json +351 -0
- package/docs/schemas/benchmark-metric-v1.schema.json +37 -0
- package/docs/schemas/benchmark-package-v1.schema.json +167 -0
- package/docs/schemas/benchmark-phase-group.schema.json +41 -0
- package/docs/schemas/benchmark-profile-v1.schema.json +141 -0
- package/docs/schemas/benchmark-task-v1.schema.json +384 -0
- package/docs/schemas/benchmark-tool-result-v1.schema.json +30 -0
- package/docs/schemas/environment-build-index.schema.json +13 -0
- package/docs/schemas/environment-build-record.schema.json +25 -0
- package/docs/schemas/environment-image-manifest.schema.json +62 -0
- package/docs/schemas/error.schema.json +22 -0
- package/docs/schemas/eval-control.schema.json +36 -0
- package/docs/schemas/eval-execution-plan.schema.json +216 -0
- package/docs/schemas/eval-progress.schema.json +1 -15
- package/docs/schemas/eval-rerun-result.schema.json +58 -0
- package/docs/schemas/eval-rerun-state.schema.json +61 -0
- package/docs/schemas/eval-rerun-submission.schema.json +60 -0
- package/docs/schemas/eval-result.schema.json +32 -13
- package/docs/schemas/eval-submission-input.schema.json +14 -0
- package/docs/schemas/eval-submission.schema.json +68 -0
- package/docs/schemas/eval-trial-publication.schema.json +15 -0
- package/docs/schemas/eval-trial-reference.schema.json +129 -0
- package/docs/schemas/execution-evidence.schema.json +96 -0
- package/docs/schemas/execution-lease.schema.json +50 -0
- package/docs/schemas/execution-provider-status.schema.json +65 -0
- package/docs/schemas/execution-worker.schema.json +53 -0
- package/docs/schemas/harbor-process-exit.schema.json +14 -0
- package/docs/schemas/interaction-capture-ref.schema.json +45 -0
- package/docs/schemas/local-provider-execution.schema.json +33 -0
- package/docs/schemas/model-interaction.schema.json +41 -0
- package/docs/schemas/regrade-assessment-reference.schema.json +11 -0
- package/docs/schemas/remote-result-envelope.schema.json +32 -0
- package/docs/schemas/remote-tree-envelope.schema.json +25 -0
- package/docs/schemas/remote-work-offer.schema.json +58 -0
- package/docs/schemas/remote-work-receipt.schema.json +50 -0
- package/docs/schemas/remote-worker-artifact.schema.json +17 -0
- package/docs/schemas/remote-worker-event.schema.json +18 -0
- package/docs/schemas/remote-worker-heartbeat.schema.json +37 -0
- package/docs/schemas/remote-worker-registration.schema.json +52 -0
- package/docs/schemas/result-bundle-index.schema.json +133 -0
- package/docs/schemas/run-context.schema.json +15 -0
- package/docs/schemas/run-environment-images.schema.json +39 -0
- package/docs/schemas/run-manifest.schema.json +4 -0
- package/docs/schemas/run-request.schema.json +4 -0
- package/docs/schemas/training-data-candidate.schema.json +46 -0
- package/docs/schemas/trajectory-analysis.schema.json +258 -0
- package/docs/schemas/trajectory-events-page.schema.json +40 -0
- package/docs/schemas/verifier-assessment.schema.json +77 -0
- package/docs/schemas/verifier-evidence.schema.json +276 -0
- package/integrations/harbor/hitch_benchmark.py +252 -0
- package/integrations/harbor/hitch_candidate_recycle.py +262 -0
- package/integrations/harbor/hitch_harbor_agent.py +698 -568
- package/integrations/harbor/hitch_harbor_environment.py +469 -0
- package/integrations/harbor/hitch_harbor_task_resources.py +275 -0
- package/integrations/harbor/hitch_harbor_verifier.py +83 -2
- package/integrations/harbor/hitch_phase_supervisor.py +452 -0
- package/integrations/harbor/hitch_tool_client.mjs +91 -0
- package/integrations/model-call/cli.js +57 -0
- package/package.json +30 -6
|
@@ -0,0 +1,174 @@
|
|
|
1
|
+
const RESOURCE_FIELDS = ["cpu_millis", "memory_bytes", "container_slots", "build_slots"];
|
|
2
|
+
const OPTIONAL_RESOURCE_FIELDS = ["gpu_count", "ephemeral_disk_bytes"];
|
|
3
|
+
const SOURCES = new Set([
|
|
4
|
+
"task", "compose", "submission-default", "operator-default", "provider-policy", "derived-components",
|
|
5
|
+
]);
|
|
6
|
+
export function parseTaskResourceRequirements(value, taskIds) {
|
|
7
|
+
if (value === undefined)
|
|
8
|
+
return undefined;
|
|
9
|
+
if (!Array.isArray(value))
|
|
10
|
+
throw new TypeError("execution plan task resources must be an array");
|
|
11
|
+
const requirements = value.map((entry, index) => parseRequirement(entry, index));
|
|
12
|
+
const actual = requirements.map((entry) => entry.task_id);
|
|
13
|
+
const expected = [...taskIds].sort(compareBytes);
|
|
14
|
+
if (new Set(actual).size !== actual.length || JSON.stringify(actual) !== JSON.stringify(expected)) {
|
|
15
|
+
throw new TypeError("execution plan task resources do not exactly match tasks");
|
|
16
|
+
}
|
|
17
|
+
return requirements;
|
|
18
|
+
}
|
|
19
|
+
export function reservationForTasks(taskIds, parallelism, defaults, requirements) {
|
|
20
|
+
if (!requirements)
|
|
21
|
+
return scaleResources(defaults, parallelism);
|
|
22
|
+
const byTask = new Map(requirements.map((entry) => [entry.task_id, entry.reservation]));
|
|
23
|
+
const selected = taskIds.map((taskId) => {
|
|
24
|
+
const resource = byTask.get(taskId);
|
|
25
|
+
if (!resource)
|
|
26
|
+
throw new TypeError(`execution plan has no resources for task: ${taskId}`);
|
|
27
|
+
return resource;
|
|
28
|
+
});
|
|
29
|
+
const result = Object.fromEntries(RESOURCE_FIELDS.map((name) => [name, selected
|
|
30
|
+
.map((entry) => entry[name])
|
|
31
|
+
.sort((left, right) => right - left)
|
|
32
|
+
.slice(0, parallelism)
|
|
33
|
+
.reduce((sum, entry) => sum + entry, 0)]));
|
|
34
|
+
for (const field of OPTIONAL_RESOURCE_FIELDS)
|
|
35
|
+
if (selected.some((entry) => entry[field] !== undefined))
|
|
36
|
+
result[field] = selected
|
|
37
|
+
.map((entry) => entry[field] ?? 0)
|
|
38
|
+
.sort((left, right) => right - left)
|
|
39
|
+
.slice(0, parallelism)
|
|
40
|
+
.reduce((sum, entry) => sum + entry, 0);
|
|
41
|
+
return result;
|
|
42
|
+
}
|
|
43
|
+
export function resourceRequirementForTask(plan, taskId) {
|
|
44
|
+
return plan.task_resources?.find((entry) => entry.task_id === taskId);
|
|
45
|
+
}
|
|
46
|
+
export function runtimeResourcesForTask(plan, taskId, fallback) {
|
|
47
|
+
const requirement = resourceRequirementForTask(plan, taskId);
|
|
48
|
+
const sidecarLimits = Object.fromEntries((requirement?.components ?? [])
|
|
49
|
+
.filter((entry) => entry.role === "task-sidecar" || entry.role === "provider-sidecar")
|
|
50
|
+
.map((entry) => [entry.name, {
|
|
51
|
+
cpu_millis: entry.fields.cpu_millis.value,
|
|
52
|
+
memory_bytes: entry.fields.memory_bytes.value,
|
|
53
|
+
...(entry.fields.gpu_count && entry.fields.gpu_count.value > 0 ? { gpu_count: entry.fields.gpu_count.value } : {}),
|
|
54
|
+
}]));
|
|
55
|
+
return { mainLimits: requirement?.main_limits ?? fallback, sidecarLimits };
|
|
56
|
+
}
|
|
57
|
+
function parseRequirement(value, index) {
|
|
58
|
+
const record = exactRecord(value, ["task_id", "reservation", "main_limits", "fields", "components", "diagnostics"], `task resources ${index}`);
|
|
59
|
+
if (typeof record.task_id !== "string" || !record.task_id)
|
|
60
|
+
throw new TypeError(`task resources ${index} identity is invalid`);
|
|
61
|
+
const reservation = parseVector(record.reservation, `task resources ${index} reservation`);
|
|
62
|
+
const mainLimits = parseVector(record.main_limits, `task resources ${index} main limits`);
|
|
63
|
+
if (mainLimits.container_slots !== 1 || mainLimits.build_slots !== 0)
|
|
64
|
+
throw new TypeError(`task resources ${index} main limits are invalid`);
|
|
65
|
+
const fieldsRecord = exactRecord(record.fields, RESOURCE_FIELDS, `task resources ${index} fields`, OPTIONAL_RESOURCE_FIELDS);
|
|
66
|
+
const fields = Object.fromEntries(RESOURCE_FIELDS.map((name) => [name, parseField(fieldsRecord[name], `task resources ${index} ${name}`)]));
|
|
67
|
+
if (fieldsRecord.gpu_count !== undefined)
|
|
68
|
+
fields.gpu_count = parseField(fieldsRecord.gpu_count, `task resources ${index} gpu_count`);
|
|
69
|
+
if (fieldsRecord.ephemeral_disk_bytes !== undefined)
|
|
70
|
+
fields.ephemeral_disk_bytes = parseField(fieldsRecord.ephemeral_disk_bytes, `task resources ${index} ephemeral_disk_bytes`);
|
|
71
|
+
if (RESOURCE_FIELDS.some((name) => fields[name].value !== reservation[name])
|
|
72
|
+
|| (fields.gpu_count?.value ?? 0) !== (reservation.gpu_count ?? 0)
|
|
73
|
+
|| (fields.ephemeral_disk_bytes?.value ?? 0) !== (reservation.ephemeral_disk_bytes ?? 0)
|
|
74
|
+
|| fields.cpu_millis.source !== "derived-components" || fields.memory_bytes.source !== "derived-components"
|
|
75
|
+
|| fields.container_slots.source !== "derived-components" || fields.build_slots.source !== "derived-components") {
|
|
76
|
+
throw new TypeError(`task resources ${index} field evidence does not match reservation`);
|
|
77
|
+
}
|
|
78
|
+
if (!Array.isArray(record.components) || record.components.length === 0)
|
|
79
|
+
throw new TypeError(`task resources ${index} components are invalid`);
|
|
80
|
+
const components = record.components.map((entry, componentIndex) => parseComponent(entry, index, componentIndex));
|
|
81
|
+
if (components.filter((entry) => entry.role === "main").length !== 1 || new Set(components.map((entry) => `${entry.role}:${entry.name}`)).size !== components.length) {
|
|
82
|
+
throw new TypeError(`task resources ${index} component identities are invalid`);
|
|
83
|
+
}
|
|
84
|
+
const sum = sumComponents(components);
|
|
85
|
+
if (RESOURCE_FIELDS.some((name) => sum[name] !== reservation[name])
|
|
86
|
+
|| OPTIONAL_RESOURCE_FIELDS.some((name) => (sum[name] ?? 0) !== (reservation[name] ?? 0)))
|
|
87
|
+
throw new TypeError(`task resources ${index} components do not match reservation`);
|
|
88
|
+
for (const component of components.filter((entry) => entry.role === "main" || entry.role === "verifier")) {
|
|
89
|
+
if (component.fields.cpu_millis.value > mainLimits.cpu_millis || component.fields.memory_bytes.value > mainLimits.memory_bytes) {
|
|
90
|
+
throw new TypeError(`task resources ${index} main limits do not cover runtime components`);
|
|
91
|
+
}
|
|
92
|
+
}
|
|
93
|
+
if (!Array.isArray(record.diagnostics) || record.diagnostics.some((entry) => typeof entry !== "string" || !entry)
|
|
94
|
+
|| new Set(record.diagnostics).size !== record.diagnostics.length)
|
|
95
|
+
throw new TypeError(`task resources ${index} diagnostics are invalid`);
|
|
96
|
+
return {
|
|
97
|
+
task_id: record.task_id,
|
|
98
|
+
reservation,
|
|
99
|
+
main_limits: mainLimits,
|
|
100
|
+
fields,
|
|
101
|
+
components,
|
|
102
|
+
diagnostics: [...record.diagnostics],
|
|
103
|
+
};
|
|
104
|
+
}
|
|
105
|
+
function parseComponent(value, requirementIndex, componentIndex) {
|
|
106
|
+
const label = `task resources ${requirementIndex} component ${componentIndex}`;
|
|
107
|
+
const record = exactRecord(value, ["name", "role", "replicas", "resources", "fields"], label);
|
|
108
|
+
if (typeof record.name !== "string" || !record.name
|
|
109
|
+
|| !["main", "task-sidecar", "verifier", "provider-sidecar"].includes(record.role)
|
|
110
|
+
|| !Number.isSafeInteger(record.replicas) || record.replicas < 1)
|
|
111
|
+
throw new TypeError(`${label} identity is invalid`);
|
|
112
|
+
const resources = parseVector(record.resources, `${label} resources`);
|
|
113
|
+
const fieldsRecord = exactRecord(record.fields, ["cpu_millis", "memory_bytes"], `${label} fields`, OPTIONAL_RESOURCE_FIELDS);
|
|
114
|
+
const fields = {
|
|
115
|
+
cpu_millis: parseField(fieldsRecord.cpu_millis, `${label} cpu_millis`),
|
|
116
|
+
memory_bytes: parseField(fieldsRecord.memory_bytes, `${label} memory_bytes`),
|
|
117
|
+
...(fieldsRecord.gpu_count === undefined ? {} : { gpu_count: parseField(fieldsRecord.gpu_count, `${label} gpu_count`) }),
|
|
118
|
+
...(fieldsRecord.ephemeral_disk_bytes === undefined ? {} : { ephemeral_disk_bytes: parseField(fieldsRecord.ephemeral_disk_bytes, `${label} ephemeral_disk_bytes`) }),
|
|
119
|
+
};
|
|
120
|
+
const replicas = record.replicas;
|
|
121
|
+
if (resources.cpu_millis !== fields.cpu_millis.value * replicas || resources.memory_bytes !== fields.memory_bytes.value * replicas
|
|
122
|
+
|| (resources.gpu_count ?? 0) !== (fields.gpu_count?.value ?? 0) * replicas
|
|
123
|
+
|| (resources.ephemeral_disk_bytes ?? 0) !== (fields.ephemeral_disk_bytes?.value ?? 0) * replicas
|
|
124
|
+
|| resources.container_slots !== replicas || resources.build_slots !== 0)
|
|
125
|
+
throw new TypeError(`${label} resource totals are invalid`);
|
|
126
|
+
return { name: record.name, role: record.role, replicas, resources, fields };
|
|
127
|
+
}
|
|
128
|
+
function parseField(value, label) {
|
|
129
|
+
const record = exactRecord(value, ["value", "source", "estimated"], label);
|
|
130
|
+
if (!Number.isSafeInteger(record.value) || record.value < 0 || !SOURCES.has(record.source)
|
|
131
|
+
|| typeof record.estimated !== "boolean")
|
|
132
|
+
throw new TypeError(`${label} evidence is invalid`);
|
|
133
|
+
return { value: record.value, source: record.source, estimated: record.estimated };
|
|
134
|
+
}
|
|
135
|
+
function parseVector(value, label) {
|
|
136
|
+
const record = exactRecord(value, RESOURCE_FIELDS, label, OPTIONAL_RESOURCE_FIELDS);
|
|
137
|
+
if (RESOURCE_FIELDS.some((name) => !Number.isSafeInteger(record[name]) || record[name] < 0))
|
|
138
|
+
throw new TypeError(`${label} is invalid`);
|
|
139
|
+
for (const field of OPTIONAL_RESOURCE_FIELDS)
|
|
140
|
+
if (record[field] !== undefined && (!Number.isSafeInteger(record[field]) || record[field] < 0))
|
|
141
|
+
throw new TypeError(`${label} is invalid`);
|
|
142
|
+
return {
|
|
143
|
+
...Object.fromEntries(RESOURCE_FIELDS.map((name) => [name, record[name]])),
|
|
144
|
+
...(record.gpu_count === undefined ? {} : { gpu_count: record.gpu_count }),
|
|
145
|
+
...(record.ephemeral_disk_bytes === undefined ? {} : { ephemeral_disk_bytes: record.ephemeral_disk_bytes }),
|
|
146
|
+
};
|
|
147
|
+
}
|
|
148
|
+
function sumComponents(components) {
|
|
149
|
+
const result = Object.fromEntries(RESOURCE_FIELDS.map((name) => [name, components.reduce((sum, entry) => sum + entry.resources[name], 0)]));
|
|
150
|
+
if (components.some((entry) => entry.resources.gpu_count !== undefined))
|
|
151
|
+
result.gpu_count = components.reduce((sum, entry) => sum + (entry.resources.gpu_count ?? 0), 0);
|
|
152
|
+
if (components.some((entry) => entry.resources.ephemeral_disk_bytes !== undefined))
|
|
153
|
+
result.ephemeral_disk_bytes = components.reduce((sum, entry) => sum + (entry.resources.ephemeral_disk_bytes ?? 0), 0);
|
|
154
|
+
return result;
|
|
155
|
+
}
|
|
156
|
+
function scaleResources(resources, count) {
|
|
157
|
+
return {
|
|
158
|
+
...Object.fromEntries(RESOURCE_FIELDS.map((name) => [name, resources[name] * count])),
|
|
159
|
+
...(resources.gpu_count === undefined ? {} : { gpu_count: resources.gpu_count * count }),
|
|
160
|
+
...(resources.ephemeral_disk_bytes === undefined ? {} : { ephemeral_disk_bytes: resources.ephemeral_disk_bytes * count }),
|
|
161
|
+
};
|
|
162
|
+
}
|
|
163
|
+
function exactRecord(value, keys, label, optional = []) {
|
|
164
|
+
if (!value || typeof value !== "object" || Array.isArray(value))
|
|
165
|
+
throw new TypeError(`${label} must be an object`);
|
|
166
|
+
const record = value;
|
|
167
|
+
if (keys.some((key) => !(key in record)) || Object.keys(record).some((key) => !keys.includes(key) && !optional.includes(key)))
|
|
168
|
+
throw new TypeError(`${label} fields are invalid`);
|
|
169
|
+
return record;
|
|
170
|
+
}
|
|
171
|
+
function compareBytes(left, right) {
|
|
172
|
+
return Buffer.from(left).compare(Buffer.from(right));
|
|
173
|
+
}
|
|
174
|
+
//# sourceMappingURL=execution-plan-resources.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"execution-plan-resources.js","sourceRoot":"","sources":["../../../src/evals/execution-plan-resources.ts"],"names":[],"mappings":"AAQA,MAAM,eAAe,GAAG,CAAC,YAAY,EAAE,cAAc,EAAE,iBAAiB,EAAE,aAAa,CAAU,CAAC;AAClG,MAAM,wBAAwB,GAAG,CAAC,WAAW,EAAE,sBAAsB,CAAU,CAAC;AAChF,MAAM,OAAO,GAAG,IAAI,GAAG,CAA8B;IACnD,MAAM,EAAE,SAAS,EAAE,oBAAoB,EAAE,kBAAkB,EAAE,iBAAiB,EAAE,oBAAoB;CACrG,CAAC,CAAC;AAEH,MAAM,UAAU,6BAA6B,CAAC,KAAc,EAAE,OAA0B;IACtF,IAAI,KAAK,KAAK,SAAS;QAAE,OAAO,SAAS,CAAC;IAC1C,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,gDAAgD,CAAC,CAAC;IACjG,MAAM,YAAY,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE,CAAC,gBAAgB,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC;IACjF,MAAM,MAAM,GAAG,YAAY,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAC1D,MAAM,QAAQ,GAAG,CAAC,GAAG,OAAO,CAAC,CAAC,IAAI,CAAC,YAAY,CAAC,CAAC;IACjD,IAAI,IAAI,GAAG,CAAC,MAAM,CAAC,CAAC,IAAI,KAAK,MAAM,CAAC,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,KAAK,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,EAAE,CAAC;QAClG,MAAM,IAAI,SAAS,CAAC,0DAA0D,CAAC,CAAC;IAClF,CAAC;IACD,OAAO,YAAY,CAAC;AACtB,CAAC;AAED,MAAM,UAAU,mBAAmB,CACjC,OAA0B,EAC1B,WAAmB,EACnB,QAA0B,EAC1B,YAAmD;IAEnD,IAAI,CAAC,YAAY;QAAE,OAAO,cAAc,CAAC,QAAQ,EAAE,WAAW,CAAC,CAAC;IAChE,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC;IACxF,MAAM,QAAQ,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,EAAE;QACtC,MAAM,QAAQ,GAAG,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;QACpC,IAAI,CAAC,QAAQ;YAAE,MAAM,IAAI,SAAS,CAAC,6CAA6C,MAAM,EAAE,CAAC,CAAC;QAC1F,OAAO,QAAQ,CAAC;IAClB,CAAC,CAAC,CAAC;IACH,MAAM,MAAM,GAAG,MAAM,CAAC,WAAW,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,QAAQ;aAC5E,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;aAC3B,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,KAAK,GAAG,IAAI,CAAC;aACnC,KAAK,CAAC,CAAC,EAAE,WAAW,CAAC;aACrB,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,CAAgC,CAAC;IAC5E,KAAK,MAAM,KAAK,IAAI,wBAAwB;QAAE,IAAI,QAAQ,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,KAAK,CAAC,KAAK,SAAS,CAAC;YAAE,MAAM,CAAC,KAAK,CAAC,GAAG,QAAQ;iBAC7H,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;iBACjC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,KAAK,GAAG,IAAI,CAAC;iBACnC,KAAK,CAAC,CAAC,EAAE,WAAW,CAAC;iBACrB,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,EAAE,CAAC,CAAC,CAAC;IAC1C,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,MAAM,UAAU,0BAA0B,CACxC,IAA+D,EAC/D,MAAc;IAEd,OAAO,IAAI,CAAC,cAAc,EAAE,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,OAAO,KAAK,MAAM,CAAC,CAAC;AACxE,CAAC;AAED,MAAM,UAAU,uBAAuB,CACrC,IAA+D,EAC/D,MAAc,EACd,QAA0B;IAE1B,MAAM,WAAW,GAAG,0BAA0B,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IAC7D,MAAM,aAAa,GAAG,MAAM,CAAC,WAAW,CAAC,CAAC,WAAW,EAAE,UAAU,IAAI,EAAE,CAAC;SACrE,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,cAAc,IAAI,KAAK,CAAC,IAAI,KAAK,kBAAkB,CAAC;SACrF,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,EAAE;YAC3B,UAAU,EAAE,KAAK,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK;YACzC,YAAY,EAAE,KAAK,CAAC,MAAM,CAAC,YAAY,CAAC,KAAK;YAC7C,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,SAAS,IAAI,KAAK,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACnH,CAAC,CAAC,CAAC,CAAC;IACP,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,WAAW,IAAI,QAAQ,EAAE,aAAa,EAAE,CAAC;AAC7E,CAAC;AAED,SAAS,gBAAgB,CAAC,KAAc,EAAE,KAAa;IACrD,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,EAAE,CAAC,SAAS,EAAE,aAAa,EAAE,aAAa,EAAE,QAAQ,EAAE,YAAY,EAAE,aAAa,CAAC,EAAE,kBAAkB,KAAK,EAAE,CAAC,CAAC;IAC/I,IAAI,OAAO,MAAM,CAAC,OAAO,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,OAAO;QAAE,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,sBAAsB,CAAC,CAAC;IAC9H,MAAM,WAAW,GAAG,WAAW,CAAC,MAAM,CAAC,WAAW,EAAE,kBAAkB,KAAK,cAAc,CAAC,CAAC;IAC3F,MAAM,UAAU,GAAG,WAAW,CAAC,MAAM,CAAC,WAAW,EAAE,kBAAkB,KAAK,cAAc,CAAC,CAAC;IAC1F,IAAI,UAAU,CAAC,eAAe,KAAK,CAAC,IAAI,UAAU,CAAC,WAAW,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,0BAA0B,CAAC,CAAC;IAC7I,MAAM,YAAY,GAAG,WAAW,CAAC,MAAM,CAAC,MAAM,EAAE,eAAe,EAAE,kBAAkB,KAAK,SAAS,EAAE,wBAAwB,CAAC,CAAC;IAC7H,MAAM,MAAM,GAAG,MAAM,CAAC,WAAW,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,UAAU,CAAC,YAAY,CAAC,IAAI,CAAC,EAAE,kBAAkB,KAAK,IAAI,IAAI,EAAE,CAAC,CAAC,CAAC,CAAmD,CAAC;IAC9L,IAAI,YAAY,CAAC,SAAS,KAAK,SAAS;QAAE,MAAM,CAAC,SAAS,GAAG,UAAU,CAAC,YAAY,CAAC,SAAS,EAAE,kBAAkB,KAAK,YAAY,CAAC,CAAC;IACrI,IAAI,YAAY,CAAC,oBAAoB,KAAK,SAAS;QAAE,MAAM,CAAC,oBAAoB,GAAG,UAAU,CAAC,YAAY,CAAC,oBAAoB,EAAE,kBAAkB,KAAK,uBAAuB,CAAC,CAAC;IACjL,IAAI,eAAe,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,KAAK,KAAK,WAAW,CAAC,IAAI,CAAC,CAAC;WACvE,CAAC,MAAM,CAAC,SAAS,EAAE,KAAK,IAAI,CAAC,CAAC,KAAK,CAAC,WAAW,CAAC,SAAS,IAAI,CAAC,CAAC;WAC/D,CAAC,MAAM,CAAC,oBAAoB,EAAE,KAAK,IAAI,CAAC,CAAC,KAAK,CAAC,WAAW,CAAC,oBAAoB,IAAI,CAAC,CAAC;WACrF,MAAM,CAAC,UAAU,CAAC,MAAM,KAAK,oBAAoB,IAAI,MAAM,CAAC,YAAY,CAAC,MAAM,KAAK,oBAAoB;WACxG,MAAM,CAAC,eAAe,CAAC,MAAM,KAAK,oBAAoB,IAAI,MAAM,CAAC,WAAW,CAAC,MAAM,KAAK,oBAAoB,EAAE,CAAC;QAClH,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,4CAA4C,CAAC,CAAC;IAC3F,CAAC;IACD,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,UAAU,CAAC,IAAI,MAAM,CAAC,UAAU,CAAC,MAAM,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,yBAAyB,CAAC,CAAC;IAC/I,MAAM,UAAU,GAAG,MAAM,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,cAAc,EAAE,EAAE,CAAC,cAAc,CAAC,KAAK,EAAE,KAAK,EAAE,cAAc,CAAC,CAAC,CAAC;IAClH,IAAI,UAAU,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,MAAM,CAAC,CAAC,MAAM,KAAK,CAAC,IAAI,IAAI,GAAG,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,GAAG,KAAK,CAAC,IAAI,IAAI,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,UAAU,CAAC,MAAM,EAAE,CAAC;QACrK,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,mCAAmC,CAAC,CAAC;IAClF,CAAC;IACD,MAAM,GAAG,GAAG,aAAa,CAAC,UAAU,CAAC,CAAC;IACtC,IAAI,eAAe,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,GAAG,CAAC,IAAI,CAAC,KAAK,WAAW,CAAC,IAAI,CAAC,CAAC;WAC9D,wBAAwB,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,KAAK,CAAC,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,sCAAsC,CAAC,CAAC;IAChL,KAAK,MAAM,SAAS,IAAI,UAAU,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,MAAM,IAAI,KAAK,CAAC,IAAI,KAAK,UAAU,CAAC,EAAE,CAAC;QACzG,IAAI,SAAS,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK,GAAG,UAAU,CAAC,UAAU,IAAI,SAAS,CAAC,MAAM,CAAC,YAAY,CAAC,KAAK,GAAG,UAAU,CAAC,YAAY,EAAE,CAAC;YAC/H,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,8CAA8C,CAAC,CAAC;QAC7F,CAAC;IACH,CAAC;IACD,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,WAAW,CAAC,IAAI,MAAM,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,KAAK,CAAC;WAC5G,IAAI,GAAG,CAAC,MAAM,CAAC,WAAW,CAAC,CAAC,IAAI,KAAK,MAAM,CAAC,WAAW,CAAC,MAAM;QAAE,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,0BAA0B,CAAC,CAAC;IAC5I,OAAO;QACL,OAAO,EAAE,MAAM,CAAC,OAAO;QACvB,WAAW;QACX,WAAW,EAAE,UAAU;QACvB,MAAM;QACN,UAAU;QACV,WAAW,EAAE,CAAC,GAAG,MAAM,CAAC,WAAW,CAAa;KACjD,CAAC;AACJ,CAAC;AAED,SAAS,cAAc,CAAC,KAAc,EAAE,gBAAwB,EAAE,cAAsB;IACtF,MAAM,KAAK,GAAG,kBAAkB,gBAAgB,cAAc,cAAc,EAAE,CAAC;IAC/E,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,WAAW,EAAE,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC;IAC9F,IAAI,OAAO,MAAM,CAAC,IAAI,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,IAAI;WAC9C,CAAC,CAAC,MAAM,EAAE,cAAc,EAAE,UAAU,EAAE,kBAAkB,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAc,CAAC;WACzF,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAK,MAAM,CAAC,QAAmB,GAAG,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,sBAAsB,CAAC,CAAC;IACpI,MAAM,SAAS,GAAG,WAAW,CAAC,MAAM,CAAC,SAAS,EAAE,GAAG,KAAK,YAAY,CAAC,CAAC;IACtE,MAAM,YAAY,GAAG,WAAW,CAAC,MAAM,CAAC,MAAM,EAAE,CAAC,YAAY,EAAE,cAAc,CAAC,EAAE,GAAG,KAAK,SAAS,EAAE,wBAAwB,CAAC,CAAC;IAC7H,MAAM,MAAM,GAAG;QACb,UAAU,EAAE,UAAU,CAAC,YAAY,CAAC,UAAU,EAAE,GAAG,KAAK,aAAa,CAAC;QACtE,YAAY,EAAE,UAAU,CAAC,YAAY,CAAC,YAAY,EAAE,GAAG,KAAK,eAAe,CAAC;QAC5E,GAAG,CAAC,YAAY,CAAC,SAAS,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,UAAU,CAAC,YAAY,CAAC,SAAS,EAAE,GAAG,KAAK,YAAY,CAAC,EAAE,CAAC;QACxH,GAAG,CAAC,YAAY,CAAC,oBAAoB,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,oBAAoB,EAAE,UAAU,CAAC,YAAY,CAAC,oBAAoB,EAAE,GAAG,KAAK,uBAAuB,CAAC,EAAE,CAAC;KACrK,CAAC;IACF,MAAM,QAAQ,GAAG,MAAM,CAAC,QAAkB,CAAC;IAC3C,IAAI,SAAS,CAAC,UAAU,KAAK,MAAM,CAAC,UAAU,CAAC,KAAK,GAAG,QAAQ,IAAI,SAAS,CAAC,YAAY,KAAK,MAAM,CAAC,YAAY,CAAC,KAAK,GAAG,QAAQ;WAC7H,CAAC,SAAS,CAAC,SAAS,IAAI,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,SAAS,EAAE,KAAK,IAAI,CAAC,CAAC,GAAG,QAAQ;WACxE,CAAC,SAAS,CAAC,oBAAoB,IAAI,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,oBAAoB,EAAE,KAAK,IAAI,CAAC,CAAC,GAAG,QAAQ;WAC9F,SAAS,CAAC,eAAe,KAAK,QAAQ,IAAI,SAAS,CAAC,WAAW,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,8BAA8B,CAAC,CAAC;IACxI,OAAO,EAAE,IAAI,EAAE,MAAM,CAAC,IAAI,EAAE,IAAI,EAAE,MAAM,CAAC,IAAuC,EAAE,QAAQ,EAAE,SAAS,EAAE,MAAM,EAAE,CAAC;AAClH,CAAC;AAED,SAAS,UAAU,CAAC,KAAc,EAAE,KAAa;IAC/C,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,EAAE,CAAC,OAAO,EAAE,QAAQ,EAAE,WAAW,CAAC,EAAE,KAAK,CAAC,CAAC;IAC3E,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,KAAK,CAAC,IAAK,MAAM,CAAC,KAAgB,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,MAAqC,CAAC;WAChI,OAAO,MAAM,CAAC,SAAS,KAAK,SAAS;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,sBAAsB,CAAC,CAAC;IAChG,OAAO,EAAE,KAAK,EAAE,MAAM,CAAC,KAAe,EAAE,MAAM,EAAE,MAAM,CAAC,MAAqC,EAAE,SAAS,EAAE,MAAM,CAAC,SAAS,EAAE,CAAC;AAC9H,CAAC;AAED,SAAS,WAAW,CAAC,KAAc,EAAE,KAAa;IAChD,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,EAAE,eAAe,EAAE,KAAK,EAAE,wBAAwB,CAAC,CAAC;IACpF,IAAI,eAAe,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,IAAK,MAAM,CAAC,IAAI,CAAY,GAAG,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,aAAa,CAAC,CAAC;IACpJ,KAAK,MAAM,KAAK,IAAI,wBAAwB;QAAE,IAAI,MAAM,CAAC,KAAK,CAAC,KAAK,SAAS,IAAI,CAAC,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,IAAK,MAAM,CAAC,KAAK,CAAY,GAAG,CAAC,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,aAAa,CAAC,CAAC;IACrM,OAAO;QACL,GAAG,MAAM,CAAC,WAAW,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QAC1E,GAAG,CAAC,MAAM,CAAC,SAAS,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,MAAM,CAAC,SAAS,EAAE,CAAC;QAC1E,GAAG,CAAC,MAAM,CAAC,oBAAoB,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,oBAAoB,EAAE,MAAM,CAAC,oBAAoB,EAAE,CAAC;KAC7E,CAAC;AACnC,CAAC;AAED,SAAS,aAAa,CAAC,UAAqC;IAC1D,MAAM,MAAM,GAAG,MAAM,CAAC,WAAW,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,UAAU,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,CAAC,SAAS,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAgC,CAAC;IAC3K,IAAI,UAAU,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,SAAS,CAAC,SAAS,KAAK,SAAS,CAAC;QAAE,MAAM,CAAC,SAAS,GAAG,UAAU,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,KAAK,CAAC,SAAS,CAAC,SAAS,IAAI,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;IACzK,IAAI,UAAU,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,SAAS,CAAC,oBAAoB,KAAK,SAAS,CAAC;QAAE,MAAM,CAAC,oBAAoB,GAAG,UAAU,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,KAAK,CAAC,SAAS,CAAC,oBAAoB,IAAI,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;IAC1M,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,SAAS,cAAc,CAAC,SAA2B,EAAE,KAAa;IAChE,OAAO;QACL,GAAG,MAAM,CAAC,WAAW,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,SAAS,CAAC,IAAI,CAAC,GAAG,KAAK,CAAC,CAAC,CAAC;QACrF,GAAG,CAAC,SAAS,CAAC,SAAS,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,SAAS,CAAC,SAAS,GAAG,KAAK,EAAE,CAAC;QACxF,GAAG,CAAC,SAAS,CAAC,oBAAoB,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,oBAAoB,EAAE,SAAS,CAAC,oBAAoB,GAAG,KAAK,EAAE,CAAC;KAC3F,CAAC;AACnC,CAAC;AAED,SAAS,WAAW,CAAC,KAAc,EAAE,IAAuB,EAAE,KAAa,EAAE,WAA8B,EAAE;IAC3G,IAAI,CAAC,KAAK,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,oBAAoB,CAAC,CAAC;IACnH,MAAM,MAAM,GAAG,KAAgC,CAAC;IAChD,IAAI,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC,CAAC,GAAG,IAAI,MAAM,CAAC,CAAC,IAAI,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,qBAAqB,CAAC,CAAC;IAClL,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,SAAS,YAAY,CAAC,IAAY,EAAE,KAAa;IAC/C,OAAO,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC;AACvD,CAAC"}
|
|
@@ -0,0 +1,451 @@
|
|
|
1
|
+
import { sha256JSON } from "../foundation/index.js";
|
|
2
|
+
import { artifactPinFields, opaqueWorkId, parseArtifactAssignments, parseRuntimeContract, workItemId } from "./execution-plan-artifacts.js";
|
|
3
|
+
import { imagesForTasks, parseEnvironmentImageFallbacks, parseEnvironmentImageUses } from "./execution-plan-images.js";
|
|
4
|
+
import { parseTaskResourceRequirements, reservationForTasks } from "./execution-plan-resources.js";
|
|
5
|
+
export const DEFAULT_EVAL_TRIAL_RESOURCES = {
|
|
6
|
+
cpu_millis: 1_000,
|
|
7
|
+
memory_bytes: 1024 * 1024 * 1024,
|
|
8
|
+
container_slots: 1,
|
|
9
|
+
build_slots: 0,
|
|
10
|
+
};
|
|
11
|
+
export function buildEvalExecutionPlan(options) {
|
|
12
|
+
const resources = parseResourceVector(options.trialResources || DEFAULT_EVAL_TRIAL_RESOURCES, "default trial resources");
|
|
13
|
+
if (!isSha256(options.candidate.revisionIdentity) || !isSha256(options.candidate.artifactId)) {
|
|
14
|
+
throw new TypeError("execution plan candidate identity is invalid");
|
|
15
|
+
}
|
|
16
|
+
if (!Number.isSafeInteger(options.maxParallelism) || options.maxParallelism < 1 || options.maxParallelism > options.request.max_concurrent) {
|
|
17
|
+
throw new TypeError("execution plan max parallelism is invalid");
|
|
18
|
+
}
|
|
19
|
+
const provider = options.provider || "local-docker";
|
|
20
|
+
if (!provider)
|
|
21
|
+
throw new TypeError("execution plan provider is invalid");
|
|
22
|
+
const modelCapture = options.modelCapture === null ? undefined : parseModelCapturePlan(options.modelCapture ?? {
|
|
23
|
+
requested_mode: "native",
|
|
24
|
+
effective_mode: "native",
|
|
25
|
+
required: false,
|
|
26
|
+
});
|
|
27
|
+
const tasks = options.tasks === null ? null : canonicalTasks(options.tasks);
|
|
28
|
+
if (options.taskScheduling && (tasks === null || options.workItemMode !== "task-slots")) {
|
|
29
|
+
throw new TypeError("task scheduling hints require task-slot work items with known membership");
|
|
30
|
+
}
|
|
31
|
+
const taskScheduling = tasks === null ? undefined : parseTaskScheduling(options.taskScheduling, tasks);
|
|
32
|
+
const artifactAssignments = parseArtifactAssignments(options.candidate.artifactAssignments, tasks, options.candidate.artifactId);
|
|
33
|
+
const taskResources = tasks === null ? undefined : parseTaskResourceRequirements(options.taskResources, tasks);
|
|
34
|
+
const environmentImages = tasks === null ? [] : parseEnvironmentImageUses(options.environmentImages ?? [], tasks, "execution plan environment images");
|
|
35
|
+
const imageFallbacks = tasks === null ? [] : parseEnvironmentImageFallbacks(options.environmentImageFallbacks ?? [], tasks);
|
|
36
|
+
const candidateIdentity = sha256JSON({
|
|
37
|
+
harness_revision_identity: options.candidate.revisionIdentity,
|
|
38
|
+
artifact_assignments: artifactAssignments.length > 0
|
|
39
|
+
? artifactAssignments.map((entry) => ({ task_ids: entry.taskIds, artifact_id: entry.artifactId, runtime_contract: entry.runtimeContract }))
|
|
40
|
+
: [{ task_ids: tasks ?? [], artifact_id: options.candidate.artifactId }],
|
|
41
|
+
requested_model: options.request.model,
|
|
42
|
+
agent_args_sha256: sha256JSON(options.request.agent_args),
|
|
43
|
+
protocol: {
|
|
44
|
+
timeout_ms: options.request.timeout_ms,
|
|
45
|
+
setup_timeout_ms: options.request.setup_timeout_ms,
|
|
46
|
+
},
|
|
47
|
+
});
|
|
48
|
+
const slots = tasks === null ? [] : buildSlots(options.evalId, tasks, options.request.attempts, candidateIdentity);
|
|
49
|
+
const workItems = tasks === null
|
|
50
|
+
? [opaqueWorkItem(options.evalId, options.maxParallelism, resources, provider, artifactAssignments[0])]
|
|
51
|
+
: options.workItemMode === "task-slots"
|
|
52
|
+
? buildTaskWorkItems(options.evalId, slots, resources, provider, taskResources, environmentImages, artifactAssignments, taskScheduling)
|
|
53
|
+
: buildAttemptWorkItems(options.evalId, tasks, slots, options.request.attempts, options.maxParallelism, resources, provider, taskResources, environmentImages, artifactAssignments);
|
|
54
|
+
return parseEvalExecutionPlan({
|
|
55
|
+
schema_version: "1",
|
|
56
|
+
planner: "hitch-local-v1",
|
|
57
|
+
eval_id: options.evalId,
|
|
58
|
+
membership: tasks === null ? "opaque" : "known",
|
|
59
|
+
candidate_identity: candidateIdentity,
|
|
60
|
+
benchmark: {
|
|
61
|
+
id: options.request.benchmark_id,
|
|
62
|
+
revision: options.request.benchmark_revision,
|
|
63
|
+
verifier_identity: sha256JSON({
|
|
64
|
+
backend: "harbor",
|
|
65
|
+
benchmark_id: options.request.benchmark_id,
|
|
66
|
+
benchmark_revision: options.request.benchmark_revision,
|
|
67
|
+
verifier: "dataset",
|
|
68
|
+
}),
|
|
69
|
+
},
|
|
70
|
+
provider,
|
|
71
|
+
...(modelCapture ? { model_capture: modelCapture } : {}),
|
|
72
|
+
max_parallelism: options.maxParallelism,
|
|
73
|
+
default_trial_resources: resources,
|
|
74
|
+
...(taskResources ? { task_resources: taskResources } : {}),
|
|
75
|
+
...(imageFallbacks.length > 0 ? { image_fallbacks: imageFallbacks } : {}),
|
|
76
|
+
slots,
|
|
77
|
+
work_items: workItems,
|
|
78
|
+
retry_policy: {
|
|
79
|
+
infrastructure_retries: options.request.infrastructure_retries,
|
|
80
|
+
infrastructure_retry_backoff_ms: options.request.infrastructure_retry_backoff_ms,
|
|
81
|
+
verifier_execution: "same-trial-verifier-only",
|
|
82
|
+
candidate_rerun_on_verifier_failure: false,
|
|
83
|
+
},
|
|
84
|
+
created_at: options.createdAt || new Date().toISOString(),
|
|
85
|
+
});
|
|
86
|
+
}
|
|
87
|
+
export function parseEvalExecutionPlan(value) {
|
|
88
|
+
if (!isRecord(value))
|
|
89
|
+
throw new TypeError("eval execution plan must be an object");
|
|
90
|
+
const plan = value;
|
|
91
|
+
assertOnlyKeys(plan, [
|
|
92
|
+
"schema_version", "planner", "eval_id", "membership", "candidate_identity", "benchmark", "provider", "model_capture",
|
|
93
|
+
"max_parallelism", "default_trial_resources", "task_resources", "image_fallbacks", "slots", "work_items", "retry_policy", "created_at",
|
|
94
|
+
], "eval execution plan");
|
|
95
|
+
if (plan.schema_version !== "1" || plan.planner !== "hitch-local-v1" || !isEvalId(plan.eval_id)
|
|
96
|
+
|| (plan.membership !== "known" && plan.membership !== "opaque")
|
|
97
|
+
|| !isSha256(plan.candidate_identity) || typeof plan.provider !== "string" || !plan.provider
|
|
98
|
+
|| !Number.isSafeInteger(plan.max_parallelism) || plan.max_parallelism < 1
|
|
99
|
+
|| typeof plan.created_at !== "string" || !Number.isFinite(Date.parse(plan.created_at))) {
|
|
100
|
+
throw new TypeError("eval execution plan identity is invalid");
|
|
101
|
+
}
|
|
102
|
+
const benchmark = parseBenchmark(plan.benchmark);
|
|
103
|
+
const modelCapture = plan.model_capture === undefined ? undefined : parseModelCapturePlan(plan.model_capture);
|
|
104
|
+
const resources = parseResourceVector(plan.default_trial_resources, "execution plan default trial resources");
|
|
105
|
+
if (!Array.isArray(plan.slots) || !Array.isArray(plan.work_items))
|
|
106
|
+
throw new TypeError("eval execution plan work graph is invalid");
|
|
107
|
+
const slots = plan.slots.map((slot, index) => parseSlot(slot, plan.eval_id, plan.candidate_identity, index));
|
|
108
|
+
const taskResources = parseTaskResourceRequirements(plan.task_resources, slots.map((slot) => slot.task_id).filter((task, index, all) => all.indexOf(task) === index));
|
|
109
|
+
const taskIds = slots.map((slot) => slot.task_id).filter((task, index, all) => all.indexOf(task) === index);
|
|
110
|
+
const imageFallbacks = parseEnvironmentImageFallbacks(plan.image_fallbacks ?? [], taskIds);
|
|
111
|
+
const workItems = plan.work_items.map((item, index) => parseWorkItem(item, plan.eval_id, plan.provider, plan.max_parallelism, index));
|
|
112
|
+
assertPlanGraph(plan.membership, slots, workItems, resources, taskResources);
|
|
113
|
+
const retry = parseRetryPolicy(plan.retry_policy);
|
|
114
|
+
return {
|
|
115
|
+
schema_version: "1",
|
|
116
|
+
planner: "hitch-local-v1",
|
|
117
|
+
eval_id: plan.eval_id,
|
|
118
|
+
membership: plan.membership,
|
|
119
|
+
candidate_identity: plan.candidate_identity,
|
|
120
|
+
benchmark,
|
|
121
|
+
provider: plan.provider,
|
|
122
|
+
...(modelCapture ? { model_capture: modelCapture } : {}),
|
|
123
|
+
max_parallelism: plan.max_parallelism,
|
|
124
|
+
default_trial_resources: resources,
|
|
125
|
+
...(taskResources ? { task_resources: taskResources } : {}),
|
|
126
|
+
...(imageFallbacks.length > 0 ? { image_fallbacks: imageFallbacks } : {}),
|
|
127
|
+
slots,
|
|
128
|
+
work_items: workItems,
|
|
129
|
+
retry_policy: retry,
|
|
130
|
+
created_at: plan.created_at,
|
|
131
|
+
};
|
|
132
|
+
}
|
|
133
|
+
function parseModelCapturePlan(value) {
|
|
134
|
+
if (!isRecord(value))
|
|
135
|
+
throw new TypeError("execution plan model capture must be an object");
|
|
136
|
+
assertOnlyKeys(value, ["requested_mode", "effective_mode", "required", "topology", "degraded_reason"], "execution plan model capture");
|
|
137
|
+
const modes = new Set(["off", "native", "proxy", "hybrid"]);
|
|
138
|
+
if (!modes.has(value.requested_mode) || !modes.has(value.effective_mode)
|
|
139
|
+
|| typeof value.required !== "boolean"
|
|
140
|
+
|| value.topology !== undefined && value.topology !== "host-side" && value.topology !== "in-sandbox"
|
|
141
|
+
|| value.degraded_reason !== undefined && (typeof value.degraded_reason !== "string" || !value.degraded_reason)) {
|
|
142
|
+
throw new TypeError("execution plan model capture is invalid");
|
|
143
|
+
}
|
|
144
|
+
if ((value.effective_mode === "proxy" || value.effective_mode === "hybrid") !== (value.topology !== undefined)) {
|
|
145
|
+
throw new TypeError("execution plan model capture topology is invalid");
|
|
146
|
+
}
|
|
147
|
+
if ((value.requested_mode !== value.effective_mode) !== (value.degraded_reason !== undefined)) {
|
|
148
|
+
throw new TypeError("execution plan model capture degradation evidence is invalid");
|
|
149
|
+
}
|
|
150
|
+
if (value.required === true && value.requested_mode !== value.effective_mode) {
|
|
151
|
+
throw new TypeError("required model capture cannot be degraded");
|
|
152
|
+
}
|
|
153
|
+
return {
|
|
154
|
+
requested_mode: value.requested_mode,
|
|
155
|
+
effective_mode: value.effective_mode,
|
|
156
|
+
required: value.required,
|
|
157
|
+
...(value.topology === undefined ? {} : { topology: value.topology }),
|
|
158
|
+
...(value.degraded_reason === undefined ? {} : { degraded_reason: value.degraded_reason }),
|
|
159
|
+
};
|
|
160
|
+
}
|
|
161
|
+
function buildSlots(evalId, tasks, attempts, candidateIdentity) {
|
|
162
|
+
const slots = [];
|
|
163
|
+
for (const taskId of tasks) {
|
|
164
|
+
for (let attempt = 1; attempt <= attempts; attempt += 1) {
|
|
165
|
+
const identity = { eval_id: evalId, task_id: taskId, attempt, candidate_identity: candidateIdentity };
|
|
166
|
+
slots.push({
|
|
167
|
+
schema_version: "1",
|
|
168
|
+
slot_id: `slot_${sha256JSON(identity).slice("sha256:".length, "sha256:".length + 32)}`,
|
|
169
|
+
eval_id: evalId,
|
|
170
|
+
task_id: taskId,
|
|
171
|
+
attempt,
|
|
172
|
+
candidate_identity: candidateIdentity,
|
|
173
|
+
state: "pending",
|
|
174
|
+
physical_execution: 1,
|
|
175
|
+
});
|
|
176
|
+
}
|
|
177
|
+
}
|
|
178
|
+
return slots;
|
|
179
|
+
}
|
|
180
|
+
function buildAttemptWorkItems(evalId, tasks, slots, attempts, maxParallelism, resources, provider, taskResources, environmentImages = [], artifactAssignments = []) {
|
|
181
|
+
const items = [];
|
|
182
|
+
for (let attempt = 1; attempt <= attempts; attempt += 1) {
|
|
183
|
+
const groups = artifactAssignments.length > 0
|
|
184
|
+
? artifactAssignments
|
|
185
|
+
: [{ taskIds: [...tasks] }];
|
|
186
|
+
for (const group of groups) {
|
|
187
|
+
const groupTasks = tasks.filter((task) => group.taskIds.includes(task));
|
|
188
|
+
if (groupTasks.length === 0)
|
|
189
|
+
continue;
|
|
190
|
+
const selected = slots.filter((slot) => slot.attempt === attempt && groupTasks.includes(slot.task_id)).map((slot) => slot.slot_id);
|
|
191
|
+
const imageRefs = imagesForTasks(environmentImages, groupTasks);
|
|
192
|
+
const requestedParallelism = Math.min(maxParallelism, selected.length);
|
|
193
|
+
const artifactPin = artifactPinFields(group);
|
|
194
|
+
items.push({
|
|
195
|
+
schema_version: "1",
|
|
196
|
+
work_id: workItemId(evalId, attempt, selected, imageRefs, artifactPin),
|
|
197
|
+
eval_id: evalId,
|
|
198
|
+
backend: "harbor",
|
|
199
|
+
logical_attempt: attempt,
|
|
200
|
+
task_ids: [...groupTasks],
|
|
201
|
+
slots: selected,
|
|
202
|
+
opaque_membership: false,
|
|
203
|
+
requested_parallelism: requestedParallelism,
|
|
204
|
+
reservation: reservationForTasks(groupTasks, requestedParallelism, resources, taskResources),
|
|
205
|
+
provider,
|
|
206
|
+
...(imageRefs.length > 0 ? { image_refs: imageRefs } : {}),
|
|
207
|
+
...artifactPin,
|
|
208
|
+
});
|
|
209
|
+
}
|
|
210
|
+
}
|
|
211
|
+
return items;
|
|
212
|
+
}
|
|
213
|
+
function buildTaskWorkItems(evalId, slots, resources, provider, taskResources, environmentImages = [], artifactAssignments = [], taskScheduling) {
|
|
214
|
+
return slots.map((slot) => {
|
|
215
|
+
const imageRefs = imagesForTasks(environmentImages, [slot.task_id]);
|
|
216
|
+
const assignment = artifactAssignments.find((entry) => entry.taskIds.includes(slot.task_id));
|
|
217
|
+
const artifactPin = artifactPinFields(assignment);
|
|
218
|
+
return {
|
|
219
|
+
schema_version: "1",
|
|
220
|
+
work_id: workItemId(evalId, slot.attempt, [slot.slot_id], imageRefs, artifactPin),
|
|
221
|
+
eval_id: evalId,
|
|
222
|
+
backend: "harbor",
|
|
223
|
+
logical_attempt: slot.attempt,
|
|
224
|
+
task_ids: [slot.task_id],
|
|
225
|
+
slots: [slot.slot_id],
|
|
226
|
+
opaque_membership: false,
|
|
227
|
+
requested_parallelism: 1,
|
|
228
|
+
reservation: reservationForTasks([slot.task_id], 1, resources, taskResources),
|
|
229
|
+
provider,
|
|
230
|
+
...(imageRefs.length > 0 ? { image_refs: imageRefs } : {}),
|
|
231
|
+
...(taskScheduling?.[slot.task_id] ? { scheduling: taskScheduling[slot.task_id] } : {}),
|
|
232
|
+
...artifactPin,
|
|
233
|
+
};
|
|
234
|
+
});
|
|
235
|
+
}
|
|
236
|
+
function opaqueWorkItem(evalId, maxParallelism, resources, provider, assignment) {
|
|
237
|
+
const artifactPin = artifactPinFields(assignment);
|
|
238
|
+
const identity = { eval_id: evalId, backend: "harbor", membership: "opaque", ...artifactPin };
|
|
239
|
+
return {
|
|
240
|
+
schema_version: "1",
|
|
241
|
+
work_id: `work_${sha256JSON(identity).slice("sha256:".length, "sha256:".length + 32)}`,
|
|
242
|
+
eval_id: evalId,
|
|
243
|
+
backend: "harbor",
|
|
244
|
+
logical_attempt: null,
|
|
245
|
+
task_ids: [],
|
|
246
|
+
slots: [],
|
|
247
|
+
opaque_membership: true,
|
|
248
|
+
requested_parallelism: maxParallelism,
|
|
249
|
+
reservation: scaleResources(resources, maxParallelism),
|
|
250
|
+
provider,
|
|
251
|
+
...artifactPin,
|
|
252
|
+
};
|
|
253
|
+
}
|
|
254
|
+
function parseSlot(value, evalId, candidateIdentity, index) {
|
|
255
|
+
if (!isRecord(value))
|
|
256
|
+
throw new TypeError(`eval execution plan slot ${index} is invalid`);
|
|
257
|
+
assertOnlyKeys(value, [
|
|
258
|
+
"schema_version", "slot_id", "eval_id", "task_id", "task_digest", "attempt", "candidate_identity",
|
|
259
|
+
"state", "physical_execution", "authoritative_run_id", "invalid_reason",
|
|
260
|
+
], `eval execution plan slot ${index}`);
|
|
261
|
+
if (value.schema_version !== "1" || typeof value.slot_id !== "string" || !/^slot_[a-f0-9]{32}$/.test(value.slot_id)
|
|
262
|
+
|| value.eval_id !== evalId || typeof value.task_id !== "string" || !value.task_id
|
|
263
|
+
|| !Number.isSafeInteger(value.attempt) || value.attempt < 1 || value.candidate_identity !== candidateIdentity
|
|
264
|
+
|| value.state !== "pending" || value.physical_execution !== 1) {
|
|
265
|
+
throw new TypeError(`eval execution plan slot ${index} is invalid`);
|
|
266
|
+
}
|
|
267
|
+
if (value.task_digest !== undefined && !isSha256(value.task_digest))
|
|
268
|
+
throw new TypeError(`eval execution plan slot ${index} task digest is invalid`);
|
|
269
|
+
if (value.authoritative_run_id !== undefined || value.invalid_reason !== undefined)
|
|
270
|
+
throw new TypeError(`pending execution plan slot ${index} is already settled`);
|
|
271
|
+
const identity = { eval_id: evalId, task_id: value.task_id, attempt: value.attempt, candidate_identity: candidateIdentity };
|
|
272
|
+
const expected = `slot_${sha256JSON(identity).slice("sha256:".length, "sha256:".length + 32)}`;
|
|
273
|
+
if (value.slot_id !== expected)
|
|
274
|
+
throw new TypeError(`eval execution plan slot ${index} identity does not match`);
|
|
275
|
+
return value;
|
|
276
|
+
}
|
|
277
|
+
function parseWorkItem(value, evalId, provider, maxParallelism, index) {
|
|
278
|
+
if (!isRecord(value))
|
|
279
|
+
throw new TypeError(`eval execution plan work item ${index} is invalid`);
|
|
280
|
+
assertOnlyKeys(value, [
|
|
281
|
+
"schema_version", "work_id", "eval_id", "backend", "logical_attempt", "task_ids", "slots",
|
|
282
|
+
"opaque_membership", "requested_parallelism", "reservation", "provider",
|
|
283
|
+
"image_refs", "artifact_id", "runtime_contract", "scheduling",
|
|
284
|
+
], `eval execution plan work item ${index}`);
|
|
285
|
+
if (value.schema_version !== "1" || typeof value.work_id !== "string" || !/^work_[a-f0-9]{32}$/.test(value.work_id)
|
|
286
|
+
|| value.eval_id !== evalId || value.backend !== "harbor"
|
|
287
|
+
|| (value.logical_attempt !== null && (!Number.isSafeInteger(value.logical_attempt) || value.logical_attempt < 1))
|
|
288
|
+
|| !Array.isArray(value.task_ids) || value.task_ids.some((task) => typeof task !== "string" || !task)
|
|
289
|
+
|| !Array.isArray(value.slots) || value.slots.some((slot) => typeof slot !== "string" || !/^slot_[a-f0-9]{32}$/.test(slot))
|
|
290
|
+
|| typeof value.opaque_membership !== "boolean" || !Number.isSafeInteger(value.requested_parallelism) || value.requested_parallelism < 1
|
|
291
|
+
|| value.requested_parallelism > maxParallelism || value.provider !== provider) {
|
|
292
|
+
throw new TypeError(`eval execution plan work item ${index} is invalid`);
|
|
293
|
+
}
|
|
294
|
+
if (new Set(value.task_ids).size !== value.task_ids.length || new Set(value.slots).size !== value.slots.length) {
|
|
295
|
+
throw new TypeError(`eval execution plan work item ${index} members are duplicated`);
|
|
296
|
+
}
|
|
297
|
+
const imageRefs = parseEnvironmentImageUses(value.image_refs ?? [], value.task_ids, `eval execution plan work item ${index} image refs`);
|
|
298
|
+
const artifactId = value.artifact_id === undefined ? undefined : isSha256(value.artifact_id) ? value.artifact_id : (() => { throw new TypeError(`eval execution plan work item ${index} artifact id is invalid`); })();
|
|
299
|
+
const runtimeContract = value.runtime_contract === undefined ? undefined : parseRuntimeContract(value.runtime_contract, `eval execution plan work item ${index}`);
|
|
300
|
+
const scheduling = value.scheduling === undefined ? undefined : parseSchedulingHint(value.scheduling, `eval execution plan work item ${index}`);
|
|
301
|
+
if ((artifactId === undefined) !== (runtimeContract === undefined))
|
|
302
|
+
throw new TypeError(`eval execution plan work item ${index} artifact contract is incomplete`);
|
|
303
|
+
return {
|
|
304
|
+
...value,
|
|
305
|
+
reservation: parseResourceVector(value.reservation, `eval execution plan work item ${index} reservation`),
|
|
306
|
+
...(imageRefs.length > 0 ? { image_refs: imageRefs } : {}),
|
|
307
|
+
...(artifactId ? { artifact_id: artifactId } : {}),
|
|
308
|
+
...(runtimeContract ? { runtime_contract: runtimeContract } : {}),
|
|
309
|
+
...(scheduling ? { scheduling } : {}),
|
|
310
|
+
};
|
|
311
|
+
}
|
|
312
|
+
function parseSchedulingHint(value, label) {
|
|
313
|
+
if (!isRecord(value))
|
|
314
|
+
throw new TypeError(`${label} scheduling hint is invalid`);
|
|
315
|
+
assertOnlyKeys(value, ["policy", "estimated_duration_ms", "remaining_path_ms", "estimate_source", "estimate_sample_count"], `${label} scheduling hint`);
|
|
316
|
+
const sources = new Set(["evolution-baseline", "history-p75", "task-budget", "default"]);
|
|
317
|
+
if (value.policy !== "critical-path-lpt-v1"
|
|
318
|
+
|| !Number.isSafeInteger(value.estimated_duration_ms) || value.estimated_duration_ms < 1
|
|
319
|
+
|| !Number.isSafeInteger(value.remaining_path_ms) || value.remaining_path_ms < value.estimated_duration_ms
|
|
320
|
+
|| !sources.has(value.estimate_source)
|
|
321
|
+
|| !Number.isSafeInteger(value.estimate_sample_count) || value.estimate_sample_count < 0) {
|
|
322
|
+
throw new TypeError(`${label} scheduling hint is invalid`);
|
|
323
|
+
}
|
|
324
|
+
if (value.estimate_source === "default" || value.estimate_source === "task-budget") {
|
|
325
|
+
if (value.estimate_sample_count !== 0)
|
|
326
|
+
throw new TypeError(`${label} scheduling hint sample count is invalid`);
|
|
327
|
+
}
|
|
328
|
+
else if (value.estimate_sample_count < 1)
|
|
329
|
+
throw new TypeError(`${label} scheduling hint sample count is invalid`);
|
|
330
|
+
return value;
|
|
331
|
+
}
|
|
332
|
+
function parseTaskScheduling(value, tasks) {
|
|
333
|
+
if (value === undefined)
|
|
334
|
+
return undefined;
|
|
335
|
+
const keys = Object.keys(value).sort(compareBytes);
|
|
336
|
+
if (keys.length !== tasks.length || keys.some((task, index) => task !== tasks[index])) {
|
|
337
|
+
throw new TypeError("task scheduling hints must cover every planned task exactly once");
|
|
338
|
+
}
|
|
339
|
+
return Object.fromEntries(keys.map((task) => [task, parseSchedulingHint(value[task], `task ${task}`)]));
|
|
340
|
+
}
|
|
341
|
+
function assertPlanGraph(membership, slots, workItems, resources, taskResources) {
|
|
342
|
+
if (new Set(slots.map((slot) => slot.slot_id)).size !== slots.length || new Set(workItems.map((item) => item.work_id)).size !== workItems.length) {
|
|
343
|
+
throw new TypeError("eval execution plan identities are duplicated");
|
|
344
|
+
}
|
|
345
|
+
if (membership === "opaque") {
|
|
346
|
+
const item = workItems[0];
|
|
347
|
+
if (slots.length !== 0 || workItems.length !== 1 || item?.opaque_membership !== true || item.logical_attempt !== null
|
|
348
|
+
|| item.slots.length !== 0 || item.task_ids.length !== 0 || item.work_id !== opaqueWorkId(item.eval_id, artifactPinFields(item.artifact_id && item.runtime_contract ? { taskIds: [], artifactId: item.artifact_id, runtimeContract: item.runtime_contract } : undefined))
|
|
349
|
+
|| taskResources !== undefined
|
|
350
|
+
|| JSON.stringify(item.reservation) !== JSON.stringify(scaleResources(resources, item.requested_parallelism))) {
|
|
351
|
+
throw new TypeError("opaque execution plan shape is invalid");
|
|
352
|
+
}
|
|
353
|
+
return;
|
|
354
|
+
}
|
|
355
|
+
if (slots.length === 0 || workItems.length === 0 || workItems.some((item) => item.opaque_membership))
|
|
356
|
+
throw new TypeError("known execution plan shape is invalid");
|
|
357
|
+
const planned = new Set(slots.map((slot) => slot.slot_id));
|
|
358
|
+
const assigned = workItems.flatMap((item) => item.slots);
|
|
359
|
+
if (new Set(assigned).size !== assigned.length || assigned.length !== planned.size || assigned.some((slot) => !planned.has(slot))) {
|
|
360
|
+
throw new TypeError("eval execution plan slots are not assigned exactly once");
|
|
361
|
+
}
|
|
362
|
+
const byId = new Map(slots.map((slot) => [slot.slot_id, slot]));
|
|
363
|
+
for (const item of workItems) {
|
|
364
|
+
const members = item.slots.map((slotId) => byId.get(slotId));
|
|
365
|
+
const taskIds = [...new Set(members.map((slot) => slot.task_id))].sort(compareBytes);
|
|
366
|
+
if (item.logical_attempt === null || members.some((slot) => slot.attempt !== item.logical_attempt)
|
|
367
|
+
|| JSON.stringify(item.task_ids) !== JSON.stringify(taskIds)
|
|
368
|
+
|| item.requested_parallelism > item.slots.length
|
|
369
|
+
|| JSON.stringify(item.reservation) !== JSON.stringify(reservationForTasks(item.task_ids, item.requested_parallelism, resources, taskResources))) {
|
|
370
|
+
throw new TypeError(`eval execution plan work item does not match its slots: ${item.work_id}`);
|
|
371
|
+
}
|
|
372
|
+
const expected = workItemId(item.eval_id, item.logical_attempt, item.slots, item.image_refs ?? [], {
|
|
373
|
+
...(item.artifact_id ? { artifact_id: item.artifact_id } : {}),
|
|
374
|
+
...(item.runtime_contract ? { runtime_contract: item.runtime_contract } : {}),
|
|
375
|
+
});
|
|
376
|
+
if (item.work_id !== expected)
|
|
377
|
+
throw new TypeError(`eval execution plan work item identity does not match: ${item.work_id}`);
|
|
378
|
+
}
|
|
379
|
+
}
|
|
380
|
+
function parseBenchmark(value) {
|
|
381
|
+
if (!isRecord(value))
|
|
382
|
+
throw new TypeError("eval execution plan benchmark is invalid");
|
|
383
|
+
assertOnlyKeys(value, ["id", "revision", "verifier_identity"], "eval execution plan benchmark");
|
|
384
|
+
if (typeof value.id !== "string" || !value.id || typeof value.revision !== "string" || !value.revision || !isSha256(value.verifier_identity)) {
|
|
385
|
+
throw new TypeError("eval execution plan benchmark is invalid");
|
|
386
|
+
}
|
|
387
|
+
return { id: value.id, revision: value.revision, verifier_identity: value.verifier_identity };
|
|
388
|
+
}
|
|
389
|
+
function parseRetryPolicy(value) {
|
|
390
|
+
if (!isRecord(value))
|
|
391
|
+
throw new TypeError("eval execution plan retry policy is invalid");
|
|
392
|
+
assertOnlyKeys(value, [
|
|
393
|
+
"infrastructure_retries", "infrastructure_retry_backoff_ms", "verifier_execution", "candidate_rerun_on_verifier_failure",
|
|
394
|
+
], "eval execution plan retry policy");
|
|
395
|
+
if (!Number.isSafeInteger(value.infrastructure_retries) || value.infrastructure_retries < 0
|
|
396
|
+
|| typeof value.infrastructure_retry_backoff_ms !== "number" || !Number.isFinite(value.infrastructure_retry_backoff_ms) || value.infrastructure_retry_backoff_ms < 0
|
|
397
|
+
|| value.verifier_execution !== "same-trial-verifier-only" || value.candidate_rerun_on_verifier_failure !== false) {
|
|
398
|
+
throw new TypeError("eval execution plan retry policy is invalid");
|
|
399
|
+
}
|
|
400
|
+
return value;
|
|
401
|
+
}
|
|
402
|
+
function canonicalTasks(tasks) {
|
|
403
|
+
if (tasks.length === 0 || tasks.some((task) => typeof task !== "string" || !task) || new Set(tasks).size !== tasks.length) {
|
|
404
|
+
throw new TypeError("execution plan tasks must be a non-empty unique list");
|
|
405
|
+
}
|
|
406
|
+
return [...tasks].sort(compareBytes);
|
|
407
|
+
}
|
|
408
|
+
function parseResourceVector(value, label) {
|
|
409
|
+
if (!isRecord(value))
|
|
410
|
+
throw new TypeError(`${label} must be an object`);
|
|
411
|
+
const required = ["cpu_millis", "memory_bytes", "container_slots", "build_slots"];
|
|
412
|
+
const allowed = [...required, "gpu_count", "ephemeral_disk_bytes"];
|
|
413
|
+
if (Object.keys(value).some((key) => !allowed.includes(key)))
|
|
414
|
+
throw new TypeError(`${label} has unknown fields`);
|
|
415
|
+
for (const field of required) {
|
|
416
|
+
if (!Number.isSafeInteger(value[field]) || value[field] < 0)
|
|
417
|
+
throw new TypeError(`${label} ${field} is invalid`);
|
|
418
|
+
}
|
|
419
|
+
if (value.gpu_count !== undefined && (!Number.isSafeInteger(value.gpu_count) || value.gpu_count < 0))
|
|
420
|
+
throw new TypeError(`${label} gpu_count is invalid`);
|
|
421
|
+
if (value.ephemeral_disk_bytes !== undefined && (!Number.isSafeInteger(value.ephemeral_disk_bytes) || value.ephemeral_disk_bytes < 0))
|
|
422
|
+
throw new TypeError(`${label} ephemeral_disk_bytes is invalid`);
|
|
423
|
+
return {
|
|
424
|
+
...Object.fromEntries(required.map((field) => [field, value[field]])),
|
|
425
|
+
...(value.gpu_count === undefined ? {} : { gpu_count: value.gpu_count }),
|
|
426
|
+
...(value.ephemeral_disk_bytes === undefined ? {} : { ephemeral_disk_bytes: value.ephemeral_disk_bytes }),
|
|
427
|
+
};
|
|
428
|
+
}
|
|
429
|
+
function scaleResources(resources, count) {
|
|
430
|
+
const scaled = Object.fromEntries(Object.entries(resources).map(([name, value]) => [name, value * count]));
|
|
431
|
+
return parseResourceVector(scaled, "scaled work item reservation");
|
|
432
|
+
}
|
|
433
|
+
function isEvalId(value) {
|
|
434
|
+
return typeof value === "string" && /^eval_[a-f0-9]{32}$/.test(value);
|
|
435
|
+
}
|
|
436
|
+
function isSha256(value) {
|
|
437
|
+
return typeof value === "string" && /^sha256:[a-f0-9]{64}$/.test(value);
|
|
438
|
+
}
|
|
439
|
+
function isRecord(value) {
|
|
440
|
+
return Boolean(value) && typeof value === "object" && !Array.isArray(value);
|
|
441
|
+
}
|
|
442
|
+
function compareBytes(left, right) {
|
|
443
|
+
return Buffer.from(left).compare(Buffer.from(right));
|
|
444
|
+
}
|
|
445
|
+
function assertOnlyKeys(record, allowed, label) {
|
|
446
|
+
const allowedSet = new Set(allowed);
|
|
447
|
+
const unexpected = Object.keys(record).find((key) => !allowedSet.has(key));
|
|
448
|
+
if (unexpected)
|
|
449
|
+
throw new TypeError(`${label} has unknown field: ${unexpected}`);
|
|
450
|
+
}
|
|
451
|
+
//# sourceMappingURL=execution-plan.js.map
|