agent-hitch 0.2.7 → 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.
Files changed (287) hide show
  1. package/README.md +16 -2
  2. package/README.zh-CN.md +14 -2
  3. package/dist/bin/hitch.js +25 -3
  4. package/dist/bin/hitch.js.map +1 -1
  5. package/dist/scripts/canary-benchmark.js +54 -0
  6. package/dist/scripts/canary-benchmark.js.map +1 -0
  7. package/dist/scripts/canary-eval-scheduler-throughput.js +47 -0
  8. package/dist/scripts/canary-eval-scheduler-throughput.js.map +1 -0
  9. package/dist/scripts/check-architecture.js +3 -2
  10. package/dist/scripts/check-architecture.js.map +1 -1
  11. package/dist/src/adapters/catalog.js +2 -1
  12. package/dist/src/adapters/catalog.js.map +1 -1
  13. package/dist/src/adapters/providers/codex-auth.js +29 -0
  14. package/dist/src/adapters/providers/codex-auth.js.map +1 -0
  15. package/dist/src/adapters/providers/codex.js +23 -6
  16. package/dist/src/adapters/providers/codex.js.map +1 -1
  17. package/dist/src/adapters/providers/model-call.js +26 -0
  18. package/dist/src/adapters/providers/model-call.js.map +1 -0
  19. package/dist/src/artifacts/index.js +1 -1
  20. package/dist/src/artifacts/index.js.map +1 -1
  21. package/dist/src/artifacts/preparer.js +7 -2
  22. package/dist/src/artifacts/preparer.js.map +1 -1
  23. package/dist/src/backends/harbor/agent-budget.js +31 -0
  24. package/dist/src/backends/harbor/agent-budget.js.map +1 -0
  25. package/dist/src/backends/harbor/backend.js +6 -4
  26. package/dist/src/backends/harbor/backend.js.map +1 -1
  27. package/dist/src/backends/harbor/index.js +1 -0
  28. package/dist/src/backends/harbor/index.js.map +1 -1
  29. package/dist/src/backends/harbor/regrade.js +77 -0
  30. package/dist/src/backends/harbor/regrade.js.map +1 -0
  31. package/dist/src/backends/index.js +1 -0
  32. package/dist/src/backends/index.js.map +1 -1
  33. package/dist/src/benchmarks/index.js +3 -0
  34. package/dist/src/benchmarks/index.js.map +1 -0
  35. package/dist/src/benchmarks/loader.js +271 -0
  36. package/dist/src/benchmarks/loader.js.map +1 -0
  37. package/dist/src/benchmarks/metrics.js +16 -0
  38. package/dist/src/benchmarks/metrics.js.map +1 -0
  39. package/dist/src/benchmarks/toml.js +13 -0
  40. package/dist/src/benchmarks/toml.js.map +1 -0
  41. package/dist/src/benchmarks/validation.js +242 -0
  42. package/dist/src/benchmarks/validation.js.map +1 -0
  43. package/dist/src/cli/arguments.js +6 -4
  44. package/dist/src/cli/arguments.js.map +1 -1
  45. package/dist/src/cli/commands/benchmark.js +22 -0
  46. package/dist/src/cli/commands/benchmark.js.map +1 -0
  47. package/dist/src/cli/commands/capabilities.js +18 -0
  48. package/dist/src/cli/commands/capabilities.js.map +1 -0
  49. package/dist/src/cli/commands/eval.js +15 -3
  50. package/dist/src/cli/commands/eval.js.map +1 -1
  51. package/dist/src/cli/commands/run.js +37 -15
  52. package/dist/src/cli/commands/run.js.map +1 -1
  53. package/dist/src/cli/commands/trajectory.js +96 -5
  54. package/dist/src/cli/commands/trajectory.js.map +1 -1
  55. package/dist/src/cli/commands/verifier.js +36 -0
  56. package/dist/src/cli/commands/verifier.js.map +1 -0
  57. package/dist/src/cli/main.js +6 -0
  58. package/dist/src/cli/main.js.map +1 -1
  59. package/dist/src/cli/output.js +14 -2
  60. package/dist/src/cli/output.js.map +1 -1
  61. package/dist/src/control-plane/eval-control-work.js +5 -0
  62. package/dist/src/control-plane/eval-control-work.js.map +1 -1
  63. package/dist/src/control-plane/eval-scheduler.js +4 -1
  64. package/dist/src/control-plane/eval-scheduler.js.map +1 -1
  65. package/dist/src/control-plane/index.js +1 -1
  66. package/dist/src/control-plane/index.js.map +1 -1
  67. package/dist/src/control-plane/remote-result-transport.js +3 -0
  68. package/dist/src/control-plane/remote-result-transport.js.map +1 -1
  69. package/dist/src/control-plane/remote-work-coordinator.js +1 -0
  70. package/dist/src/control-plane/remote-work-coordinator.js.map +1 -1
  71. package/dist/src/control-plane/remote-work-item.js +62 -0
  72. package/dist/src/control-plane/remote-work-item.js.map +1 -0
  73. package/dist/src/control-plane/remote-work-recovery.js +44 -5
  74. package/dist/src/control-plane/remote-work-recovery.js.map +1 -1
  75. package/dist/src/control-plane/remote-worker-protocol.js +2 -23
  76. package/dist/src/control-plane/remote-worker-protocol.js.map +1 -1
  77. package/dist/src/control-plane/rerun-scheduler.js +51 -14
  78. package/dist/src/control-plane/rerun-scheduler.js.map +1 -1
  79. package/dist/src/control-plane/rerun-submission.js +10 -4
  80. package/dist/src/control-plane/rerun-submission.js.map +1 -1
  81. package/dist/src/control-plane/work-admission.js +2 -0
  82. package/dist/src/control-plane/work-admission.js.map +1 -1
  83. package/dist/src/control-plane/work-dispatcher.js +9 -1
  84. package/dist/src/control-plane/work-dispatcher.js.map +1 -1
  85. package/dist/src/controller-runtime/hash.js +7 -2
  86. package/dist/src/controller-runtime/hash.js.map +1 -1
  87. package/dist/src/daemon/auth.js +5 -0
  88. package/dist/src/daemon/auth.js.map +1 -1
  89. package/dist/src/domain/benchmarks.js +2 -0
  90. package/dist/src/domain/benchmarks.js.map +1 -0
  91. package/dist/src/domain/index.js +4 -0
  92. package/dist/src/domain/index.js.map +1 -1
  93. package/dist/src/domain/runs.js.map +1 -1
  94. package/dist/src/domain/validation.js +11 -4
  95. package/dist/src/domain/validation.js.map +1 -1
  96. package/dist/src/domain/verifier-evidence-validation.js +214 -0
  97. package/dist/src/domain/verifier-evidence-validation.js.map +1 -0
  98. package/dist/src/domain/verifier-evidence.js +3 -0
  99. package/dist/src/domain/verifier-evidence.js.map +1 -0
  100. package/dist/src/domain/verifier-score-contract.js +232 -0
  101. package/dist/src/domain/verifier-score-contract.js.map +1 -0
  102. package/dist/src/evals/benchmark-adapter-manifest.js +219 -0
  103. package/dist/src/evals/benchmark-adapter-manifest.js.map +1 -0
  104. package/dist/src/evals/benchmark-candidate.js +36 -0
  105. package/dist/src/evals/benchmark-candidate.js.map +1 -0
  106. package/dist/src/evals/benchmark-run.js +166 -0
  107. package/dist/src/evals/benchmark-run.js.map +1 -0
  108. package/dist/src/evals/collect-only-rerun.js +2 -1
  109. package/dist/src/evals/collect-only-rerun.js.map +1 -1
  110. package/dist/src/evals/directory.js +1 -1
  111. package/dist/src/evals/directory.js.map +1 -1
  112. package/dist/src/evals/duration-estimator.js +126 -0
  113. package/dist/src/evals/duration-estimator.js.map +1 -0
  114. package/dist/src/evals/eval-lifecycle-events.js +13 -0
  115. package/dist/src/evals/eval-lifecycle-events.js.map +1 -1
  116. package/dist/src/evals/eval-logical-plan.js +51 -0
  117. package/dist/src/evals/eval-logical-plan.js.map +1 -0
  118. package/dist/src/evals/eval-result-builder.js +81 -0
  119. package/dist/src/evals/eval-result-builder.js.map +1 -0
  120. package/dist/src/evals/evolution-baseline.js +191 -0
  121. package/dist/src/evals/evolution-baseline.js.map +1 -0
  122. package/dist/src/evals/execution-plan.js +39 -3
  123. package/dist/src/evals/execution-plan.js.map +1 -1
  124. package/dist/src/evals/failure-classifier.js +61 -0
  125. package/dist/src/evals/failure-classifier.js.map +1 -0
  126. package/dist/src/evals/harbor-artifact-builder.js +6 -2
  127. package/dist/src/evals/harbor-artifact-builder.js.map +1 -1
  128. package/dist/src/evals/harbor-bridge-error.js +41 -1
  129. package/dist/src/evals/harbor-bridge-error.js.map +1 -1
  130. package/dist/src/evals/harbor-node-runtime.js +146 -0
  131. package/dist/src/evals/harbor-node-runtime.js.map +1 -0
  132. package/dist/src/evals/index.js +15 -0
  133. package/dist/src/evals/index.js.map +1 -1
  134. package/dist/src/evals/infrastructure-retry.js +109 -22
  135. package/dist/src/evals/infrastructure-retry.js.map +1 -1
  136. package/dist/src/evals/native-phase-evidence.js +440 -0
  137. package/dist/src/evals/native-phase-evidence.js.map +1 -0
  138. package/dist/src/evals/physical-retry-work.js +31 -0
  139. package/dist/src/evals/physical-retry-work.js.map +1 -0
  140. package/dist/src/evals/planned-execution-support.js +22 -10
  141. package/dist/src/evals/planned-execution-support.js.map +1 -1
  142. package/dist/src/evals/planned-execution.js +120 -112
  143. package/dist/src/evals/planned-execution.js.map +1 -1
  144. package/dist/src/evals/planned-progress-publisher.js +62 -0
  145. package/dist/src/evals/planned-progress-publisher.js.map +1 -0
  146. package/dist/src/evals/planned-retry-execution.js +174 -0
  147. package/dist/src/evals/planned-retry-execution.js.map +1 -0
  148. package/dist/src/evals/planned-retry-lifecycle.js +4 -2
  149. package/dist/src/evals/planned-retry-lifecycle.js.map +1 -1
  150. package/dist/src/evals/preparation-rerun.js +225 -0
  151. package/dist/src/evals/preparation-rerun.js.map +1 -0
  152. package/dist/src/evals/progress.js +48 -8
  153. package/dist/src/evals/progress.js.map +1 -1
  154. package/dist/src/evals/recovery.js +54 -4
  155. package/dist/src/evals/recovery.js.map +1 -1
  156. package/dist/src/evals/regrade-evidence.js +77 -0
  157. package/dist/src/evals/regrade-evidence.js.map +1 -0
  158. package/dist/src/evals/remote-infrastructure-retry.js +84 -18
  159. package/dist/src/evals/remote-infrastructure-retry.js.map +1 -1
  160. package/dist/src/evals/request.js +21 -3
  161. package/dist/src/evals/request.js.map +1 -1
  162. package/dist/src/evals/rerun-inputs.js +19 -0
  163. package/dist/src/evals/rerun-inputs.js.map +1 -1
  164. package/dist/src/evals/rerun-types.js +2 -6
  165. package/dist/src/evals/rerun-types.js.map +1 -1
  166. package/dist/src/evals/rerun.js +27 -20
  167. package/dist/src/evals/rerun.js.map +1 -1
  168. package/dist/src/evals/result-helpers.js +13 -0
  169. package/dist/src/evals/result-helpers.js.map +1 -1
  170. package/dist/src/evals/retry-backoff.js +20 -0
  171. package/dist/src/evals/retry-backoff.js.map +1 -0
  172. package/dist/src/evals/retry-state.js +195 -0
  173. package/dist/src/evals/retry-state.js.map +1 -0
  174. package/dist/src/evals/scheduler-metrics.js +95 -0
  175. package/dist/src/evals/scheduler-metrics.js.map +1 -0
  176. package/dist/src/evals/scheduler-trace-replay.js +107 -0
  177. package/dist/src/evals/scheduler-trace-replay.js.map +1 -0
  178. package/dist/src/evals/service.js +87 -138
  179. package/dist/src/evals/service.js.map +1 -1
  180. package/dist/src/evals/trial-import.js +47 -50
  181. package/dist/src/evals/trial-import.js.map +1 -1
  182. package/dist/src/evals/trial-publication-recovery.js +45 -5
  183. package/dist/src/evals/trial-publication-recovery.js.map +1 -1
  184. package/dist/src/evals/trial-reference-validation.js +49 -0
  185. package/dist/src/evals/trial-reference-validation.js.map +1 -0
  186. package/dist/src/evals/verifier-artifacts.js +317 -0
  187. package/dist/src/evals/verifier-artifacts.js.map +1 -0
  188. package/dist/src/evals/verifier-eligibility.js +25 -0
  189. package/dist/src/evals/verifier-eligibility.js.map +1 -0
  190. package/dist/src/evals/verifier-only-rerun.js +210 -0
  191. package/dist/src/evals/verifier-only-rerun.js.map +1 -0
  192. package/dist/src/evals/verifier-runtime.js +61 -0
  193. package/dist/src/evals/verifier-runtime.js.map +1 -0
  194. package/dist/src/evals/verifier-score-artifacts.js +119 -0
  195. package/dist/src/evals/verifier-score-artifacts.js.map +1 -0
  196. package/dist/src/foundation/contained-file.js +85 -0
  197. package/dist/src/foundation/contained-file.js.map +1 -0
  198. package/dist/src/foundation/credential-redaction.js +26 -1
  199. package/dist/src/foundation/credential-redaction.js.map +1 -1
  200. package/dist/src/foundation/fs.js +2 -1
  201. package/dist/src/foundation/fs.js.map +1 -1
  202. package/dist/src/foundation/index.js +2 -1
  203. package/dist/src/foundation/index.js.map +1 -1
  204. package/dist/src/runs/adapter-process.js +12 -0
  205. package/dist/src/runs/adapter-process.js.map +1 -0
  206. package/dist/src/runs/executor.js +24 -30
  207. package/dist/src/runs/executor.js.map +1 -1
  208. package/dist/src/runs/finalizer.js +17 -0
  209. package/dist/src/runs/finalizer.js.map +1 -1
  210. package/dist/src/runs/index.js +5 -0
  211. package/dist/src/runs/index.js.map +1 -1
  212. package/dist/src/runs/phase-bundle.js +86 -0
  213. package/dist/src/runs/phase-bundle.js.map +1 -0
  214. package/dist/src/runs/phase-cancellation.js +107 -0
  215. package/dist/src/runs/phase-cancellation.js.map +1 -0
  216. package/dist/src/runs/phase-group.js +113 -0
  217. package/dist/src/runs/phase-group.js.map +1 -0
  218. package/dist/src/runs/query.js +1 -1
  219. package/dist/src/runs/query.js.map +1 -1
  220. package/dist/src/runs/records.js +3 -0
  221. package/dist/src/runs/records.js.map +1 -1
  222. package/dist/src/runs/request.js +4 -2
  223. package/dist/src/runs/request.js.map +1 -1
  224. package/dist/src/runs/verifier-evidence-redaction.js +61 -0
  225. package/dist/src/runs/verifier-evidence-redaction.js.map +1 -0
  226. package/dist/src/runs/verifier-evidence.js +459 -0
  227. package/dist/src/runs/verifier-evidence.js.map +1 -0
  228. package/dist/src/runs/verifier-structured-evidence.js +83 -0
  229. package/dist/src/runs/verifier-structured-evidence.js.map +1 -0
  230. package/dist/src/trajectories/analysis.js +288 -0
  231. package/dist/src/trajectories/analysis.js.map +1 -0
  232. package/dist/src/trajectories/chunk-projection.js +148 -0
  233. package/dist/src/trajectories/chunk-projection.js.map +1 -0
  234. package/dist/src/trajectories/content-projection.js +389 -0
  235. package/dist/src/trajectories/content-projection.js.map +1 -0
  236. package/dist/src/trajectories/dsh-chunk-contract.js +176 -0
  237. package/dist/src/trajectories/dsh-chunk-contract.js.map +1 -0
  238. package/dist/src/trajectories/dsh-contract.js +284 -0
  239. package/dist/src/trajectories/dsh-contract.js.map +1 -0
  240. package/dist/src/trajectories/events-chunk-drill.js +95 -0
  241. package/dist/src/trajectories/events-chunk-drill.js.map +1 -0
  242. package/dist/src/trajectories/events-page.js +367 -0
  243. package/dist/src/trajectories/events-page.js.map +1 -0
  244. package/dist/src/trajectories/index.js +3 -0
  245. package/dist/src/trajectories/index.js.map +1 -1
  246. package/dist/src/trajectories/request-attempt.js +47 -0
  247. package/dist/src/trajectories/request-attempt.js.map +1 -0
  248. package/dist/src/trajectories/stream-reader.js +311 -0
  249. package/dist/src/trajectories/stream-reader.js.map +1 -0
  250. package/dist/src/trajectories/surface-fold.js +126 -0
  251. package/dist/src/trajectories/surface-fold.js.map +1 -0
  252. package/dist/src/workers/remote-harbor-worker.js +3 -0
  253. package/dist/src/workers/remote-harbor-worker.js.map +1 -1
  254. package/docs/schemas/benchmark-hook-request-v1.schema.json +65 -0
  255. package/docs/schemas/benchmark-hook-response-v1.schema.json +75 -0
  256. package/docs/schemas/benchmark-hook-v1.schema.json +33 -0
  257. package/docs/schemas/benchmark-lock-v1.schema.json +351 -0
  258. package/docs/schemas/benchmark-metric-v1.schema.json +37 -0
  259. package/docs/schemas/benchmark-package-v1.schema.json +167 -0
  260. package/docs/schemas/benchmark-phase-group.schema.json +41 -0
  261. package/docs/schemas/benchmark-profile-v1.schema.json +141 -0
  262. package/docs/schemas/benchmark-task-v1.schema.json +384 -0
  263. package/docs/schemas/benchmark-tool-result-v1.schema.json +30 -0
  264. package/docs/schemas/error.schema.json +22 -0
  265. package/docs/schemas/eval-progress.schema.json +1 -15
  266. package/docs/schemas/eval-rerun-result.schema.json +6 -1
  267. package/docs/schemas/eval-rerun-submission.schema.json +3 -1
  268. package/docs/schemas/eval-result.schema.json +1 -15
  269. package/docs/schemas/eval-trial-publication.schema.json +1 -25
  270. package/docs/schemas/eval-trial-reference.schema.json +129 -0
  271. package/docs/schemas/regrade-assessment-reference.schema.json +11 -0
  272. package/docs/schemas/run-context.schema.json +15 -0
  273. package/docs/schemas/run-manifest.schema.json +4 -0
  274. package/docs/schemas/run-request.schema.json +4 -0
  275. package/docs/schemas/trajectory-analysis.schema.json +258 -0
  276. package/docs/schemas/trajectory-events-page.schema.json +40 -0
  277. package/docs/schemas/verifier-assessment.schema.json +77 -0
  278. package/docs/schemas/verifier-evidence.schema.json +276 -0
  279. package/integrations/harbor/hitch_benchmark.py +252 -0
  280. package/integrations/harbor/hitch_candidate_recycle.py +262 -0
  281. package/integrations/harbor/hitch_harbor_agent.py +541 -43
  282. package/integrations/harbor/hitch_harbor_environment.py +138 -0
  283. package/integrations/harbor/hitch_harbor_verifier.py +60 -2
  284. package/integrations/harbor/hitch_phase_supervisor.py +452 -0
  285. package/integrations/harbor/hitch_tool_client.mjs +91 -0
  286. package/integrations/model-call/cli.js +57 -0
  287. package/package.json +17 -6
@@ -0,0 +1 @@
1
+ {"version":3,"file":"retry-state.js","sourceRoot":"","sources":["../../../src/evals/retry-state.ts"],"names":[],"mappings":"AAAA,OAAO,IAAI,MAAM,WAAW,CAAC;AAE7B,OAAO,EAAE,eAAe,EAAE,SAAS,EAAE,QAAQ,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAExG,OAAO,EAAE,oBAAoB,EAAE,0BAA0B,EAAE,oBAAoB,EAAE,MAAM,yBAAyB,CAAC;AACjH,OAAO,EAAE,qBAAqB,EAAE,kCAAkC,EAAE,MAAM,0BAA0B,CAAC;AAiCrG,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,aAAqB,EAAE,MAAc;IAC5E,MAAM,KAAK,GAAG,MAAM,QAAQ,CAAiB,IAAI,CAAC,IAAI,CAAC,aAAa,EAAE,kBAAkB,CAAC,EAAE,IAAI,CAAC,CAAC;IACjG,OAAO,KAAK,KAAK,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,mBAAmB,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;AACpE,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAAC,KAQjD;IACC,MAAM,cAAc,GAAG,oBAAoB,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAC3D,IAAI,CAAC,cAAc,IAAI,CAAC,oBAAoB,CAAC,cAAc,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,6DAA6D,CAAC,CAAC;IACjJ,MAAM,SAAS,GAAG,qBAAqB,CAAC,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,UAAU,EAAE,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC;IACvF,OAAO,cAAc,CAAC;QACpB,GAAG,KAAK;QACR,cAAc;QACd,WAAW,EAAE,gBAAgB;QAC7B,WAAW,EAAE,SAAS,CAAC,OAAO;QAC9B,KAAK,EAAE,SAAS;KACjB,CAAC,CAAC;AACL,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAAC,KAQjD;IACC,MAAM,cAAc,GAAG,oBAAoB,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAC3D,IAAI,CAAC,cAAc;QAAE,OAAO,IAAI,CAAC;IACjC,MAAM,WAAW,GAAG,cAAc,CAAC,cAAc,CAAC,CAAC;IACnD,OAAO,cAAc,CAAC;QACpB,GAAG,KAAK;QACR,cAAc;QACd,WAAW;QACX,KAAK,EAAE,KAAK,CAAC,SAAS,IAAI,oBAAoB,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,SAAS;KACzF,CAAC,CAAC;AACL,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,uBAAuB,CAAC,KAM7C;IACC,OAAO,WAAW,CAAC,KAAK,CAAC,aAAa,EAAE,KAAK,CAAC,MAAM,EAAE,CAAC,OAAO,EAAE,EAAE;QAChE,MAAM,KAAK,GAAG,OAAO,CAAC,SAAS,CAAC,SAAS,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,WAAW,KAAK,KAAK,CAAC,UAAU,CAAC,CAAC;QACnG,IAAI,KAAK,GAAG,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,kCAAkC,KAAK,CAAC,UAAU,EAAE,CAAC,CAAC;QACzF,MAAM,QAAQ,GAAG,OAAO,CAAC,SAAS,CAAC,KAAK,CAAoB,CAAC;QAC7D,IAAI,QAAQ,CAAC,KAAK,KAAK,KAAK,CAAC,KAAK;YAAE,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC;QACxF,IAAI,CAAC,iBAAiB,CAAC,QAAQ,CAAC,KAAK,EAAE,KAAK,CAAC,KAAK,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,+CAA+C,QAAQ,CAAC,KAAK,OAAO,KAAK,CAAC,KAAK,EAAE,CAAC,CAAC;QAC5J,MAAM,OAAO,GAAG,EAAE,GAAG,QAAQ,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,UAAU,EAAE,cAAc,CAAC,KAAK,CAAC,GAAG,IAAI,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,EAAE,2BAA2B,CAAC,EAAE,CAAC;QACpJ,MAAM,SAAS,GAAG,CAAC,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC;QACzC,SAAS,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC;QAC3B,OAAO,EAAE,OAAO,EAAE,EAAE,GAAG,OAAO,EAAE,SAAS,EAAE,kBAAkB,CAAC,SAAS,CAAC,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC;IAC9G,CAAC,CAAC,CAAC;AACL,CAAC;AAED,MAAM,UAAU,gBAAgB,CAAC,IAAyB,EAAE,KAA8B,EAAE,MAAc;IACxG,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,MAAM,SAAS,GAAG,KAAK,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,aAAa,KAAK,MAAM,CAAC,CAAC;IAC1F,IAAI,SAAS,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IACxC,MAAM,KAAK,GAAG,SAAS,CAAC,CAAC,CAAoB,CAAC;IAC9C,IAAI,SAAS,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK,CAAC,OAAO,IAAI,QAAQ,CAAC,WAAW,KAAK,KAAK,CAAC,WAAW,CAAC,EAAE,CAAC;QACnH,MAAM,IAAI,SAAS,CAAC,kDAAkD,MAAM,EAAE,CAAC,CAAC;IAClF,CAAC;IACD,MAAM,MAAM,GAAG,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC;IAClF,IAAI,CAAC,MAAM;QAAE,MAAM,IAAI,SAAS,CAAC,sDAAsD,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;IACxG,MAAM,IAAI,GAAG,kCAAkC,CAAC,MAAM,EAAE,KAAK,CAAC,WAAW,EAAE,SAAS,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,gBAAgB,CAAC,CAAC,CAAC;IACnI,IAAI,IAAI,CAAC,OAAO,KAAK,MAAM;QAAE,MAAM,IAAI,SAAS,CAAC,4CAA4C,MAAM,EAAE,CAAC,CAAC;IACvG,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;AAC7B,CAAC;AAED,KAAK,UAAU,cAAc,CAAC,KAY7B;IACC,MAAM,MAAM,GAAG,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;IACnC,IAAI,CAAC,MAAM,IAAI,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,IAAI,KAAK,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,iDAAiD,CAAC,CAAC;IACzJ,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,KAAK,CAAC,UAAU,CAAC,IAAI,KAAK,CAAC,UAAU,GAAG,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,iCAAiC,CAAC,CAAC;IAC5H,MAAM,SAAS,GAAG,cAAc,CAAC,KAAK,CAAC,GAAG,IAAI,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,EAAE,2BAA2B,CAAC,CAAC;IACrG,MAAM,MAAM,GAAG,UAAU,CAAC,EAAE,OAAO,EAAE,KAAK,CAAC,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,gBAAgB,EAAE,KAAK,CAAC,OAAO,CAAC,QAAQ,EAAE,WAAW,EAAE,KAAK,CAAC,UAAU,EAAE,CAAC,CAAC;IAC/I,MAAM,QAAQ,GAAoB;QAChC,WAAW,EAAE,kBAAkB,MAAM,CAAC,KAAK,CAAC,SAAS,CAAC,MAAM,EAAE,SAAS,CAAC,MAAM,GAAG,EAAE,CAAC,EAAE;QACtF,OAAO,EAAE,MAAM;QACf,gBAAgB,EAAE,KAAK,CAAC,OAAO,CAAC,QAAQ;QACxC,GAAG,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,cAAc,EAAE,KAAK,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACzE,WAAW,EAAE,KAAK,CAAC,UAAU;QAC7B,cAAc,EAAE,KAAK,CAAC,cAAc;QACpC,WAAW,EAAE,KAAK,CAAC,WAAW;QAC9B,GAAG,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,aAAa,EAAE,KAAK,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAClE,GAAG,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,KAAK,CAAC,SAAS,EAAE,2BAA2B,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACxG,KAAK,EAAE,KAAK,CAAC,KAAK;QAClB,UAAU,EAAE,SAAS;QACrB,UAAU,EAAE,SAAS;KACtB,CAAC;IACF,OAAO,WAAW,CAAC,KAAK,CAAC,aAAa,EAAE,KAAK,CAAC,MAAM,EAAE,CAAC,OAAO,EAAE,EAAE;QAChE,MAAM,QAAQ,GAAG,OAAO,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,WAAW,KAAK,QAAQ,CAAC,WAAW,CAAC,CAAC;QAC/F,IAAI,QAAQ,EAAE,CAAC;YACb,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAAC,QAAQ,CAAC,CAAC,KAAK,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC;gBAChG,MAAM,IAAI,SAAS,CAAC,qCAAqC,QAAQ,CAAC,WAAW,EAAE,CAAC,CAAC;YACnF,CAAC;YACD,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC;QACtD,CAAC;QACD,OAAO,EAAE,OAAO,EAAE,EAAE,GAAG,OAAO,EAAE,SAAS,EAAE,kBAAkB,CAAC,CAAC,GAAG,OAAO,CAAC,SAAS,EAAE,QAAQ,CAAC,CAAC,EAAE,EAAE,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC;IACtI,CAAC,CAAC,CAAC;AACL,CAAC;AAED,KAAK,UAAU,WAAW,CACxB,aAAqB,EACrB,MAAc,EACd,SAAiG;IAEjG,MAAM,SAAS,CAAC,aAAa,CAAC,CAAC;IAC/B,OAAO,YAAY,CAAC,IAAI,CAAC,IAAI,CAAC,aAAa,EAAE,QAAQ,CAAC,EAAE,aAAa,EAAE,KAAK,IAAI,EAAE;QAChF,MAAM,QAAQ,GAAG,MAAM,kBAAkB,CAAC,aAAa,EAAE,MAAM,CAAC,CAAC;QACjE,MAAM,GAAG,GAAG,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;QACrC,MAAM,KAAK,GAAG,QAAQ,IAAI,EAAE,cAAc,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,EAAE,UAAU,EAAE,CAAC,EAAE,SAAS,EAAE,EAAE,EAAE,UAAU,EAAE,GAAG,EAAE,CAAC;QAClH,MAAM,MAAM,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC;QAChC,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC;YACnB,MAAM,IAAI,GAAG,mBAAmB,CAAC,EAAE,GAAG,MAAM,CAAC,OAAO,EAAE,UAAU,EAAE,KAAK,CAAC,UAAU,GAAG,CAAC,EAAE,UAAU,EAAE,GAAG,EAAE,EAAE,MAAM,CAAC,CAAC;YACnH,MAAM,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,aAAa,EAAE,kBAAkB,CAAC,EAAE,IAAI,CAAC,CAAC;QAC5E,CAAC;QACD,OAAO,MAAM,CAAC,KAAK,CAAC;IACtB,CAAC,EAAE,EAAE,WAAW,EAAE,yBAAyB,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC,CAAC;AACtE,CAAC;AAED,MAAM,UAAU,mBAAmB,CAAC,KAAc,EAAE,cAAuB;IACzE,IAAI,CAAC,KAAK,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,oCAAoC,CAAC,CAAC;IAC3H,MAAM,MAAM,GAAG,KAAgC,CAAC;IAChD,IAAI,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC,CAAC,gBAAgB,EAAE,SAAS,EAAE,YAAY,EAAE,WAAW,EAAE,YAAY,CAAC,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;WACvH,MAAM,CAAC,cAAc,KAAK,GAAG,IAAI,OAAO,MAAM,CAAC,OAAO,KAAK,QAAQ,IAAI,CAAC,qBAAqB,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC;WAClH,cAAc,KAAK,SAAS,IAAI,MAAM,CAAC,OAAO,KAAK,cAAc;WACjE,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,UAAU,CAAC,IAAK,MAAM,CAAC,UAAqB,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,SAAS,CAAC,EAAE,CAAC;QACvH,MAAM,IAAI,SAAS,CAAC,sCAAsC,CAAC,CAAC;IAC9D,CAAC;IACD,MAAM,SAAS,GAAG,MAAM,CAAC,SAAS,CAAC,GAAG,CAAC,aAAa,CAAC,CAAC;IACtD,IAAI,IAAI,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,WAAW,CAAC,CAAC,CAAC,IAAI,KAAK,SAAS,CAAC,MAAM;QAAE,MAAM,IAAI,SAAS,CAAC,0CAA0C,CAAC,CAAC;IAC1J,MAAM,MAAM,GAAG,kBAAkB,CAAC,SAAS,CAAC,CAAC;IAC7C,IAAI,MAAM,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,KAAK,EAAE,EAAE,CAAC,QAAQ,KAAK,SAAS,CAAC,KAAK,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,4CAA4C,CAAC,CAAC;IACvI,OAAO;QACL,cAAc,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,CAAC,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,UAAoB;QACrF,SAAS,EAAE,UAAU,EAAE,cAAc,CAAC,MAAM,CAAC,UAAU,EAAE,6BAA6B,CAAC;KACxF,CAAC;AACJ,CAAC;AAED,SAAS,aAAa,CAAC,KAAc;IACnC,IAAI,CAAC,KAAK,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,kCAAkC,CAAC,CAAC;IACzH,MAAM,MAAM,GAAG,KAAgC,CAAC;IAChD,MAAM,QAAQ,GAAG,CAAC,gBAAgB,EAAE,eAAe,EAAE,YAAY,CAAC,CAAC;IACnE,MAAM,QAAQ,GAAG,CAAC,aAAa,EAAE,SAAS,EAAE,kBAAkB,EAAE,aAAa,EAAE,gBAAgB,EAAE,aAAa,EAAE,OAAO,EAAE,YAAY,EAAE,YAAY,CAAC,CAAC;IACrJ,IAAI,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;WACpF,OAAO,MAAM,CAAC,WAAW,KAAK,QAAQ,IAAI,CAAC,+BAA+B,CAAC,IAAI,CAAC,MAAM,CAAC,WAAW,CAAC;WACnG,OAAO,MAAM,CAAC,OAAO,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,OAAO,MAAM,CAAC,gBAAgB,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,gBAAgB;WAChI,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,WAAW,CAAC,IAAK,MAAM,CAAC,WAAsB,GAAG,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,oCAAoC,CAAC,CAAC;IAChJ,MAAM,cAAc,GAAG,0BAA0B,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC;IACzE,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,CAAC,gBAAgB,EAAE,eAAe,EAAE,cAAc,EAAE,UAAU,EAAE,mBAAmB,CAAC,CAAC;WACjH,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC,SAAS,EAAE,SAAS,EAAE,UAAU,EAAE,SAAS,EAAE,SAAS,EAAE,WAAW,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,kCAAkC,CAAC,CAAC;IAC5J,KAAK,MAAM,KAAK,IAAI,QAAQ;QAAE,IAAI,MAAM,CAAC,KAAK,CAAC,KAAK,SAAS,IAAI,CAAC,OAAO,MAAM,CAAC,KAAK,CAAC,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,kBAAkB,KAAK,aAAa,CAAC,CAAC;IAClL,IAAI,MAAM,CAAC,UAAU,KAAK,SAAS;QAAE,cAAc,CAAC,MAAM,CAAC,UAAU,EAAE,2BAA2B,CAAC,CAAC;IACpG,IAAI,MAAM,CAAC,WAAW,KAAK,gBAAgB,IAAI,OAAO,MAAM,CAAC,aAAa,KAAK,QAAQ;QAAE,MAAM,IAAI,SAAS,CAAC,kDAAkD,CAAC,CAAC;IACjK,OAAO;QACL,WAAW,EAAE,MAAM,CAAC,WAAW,EAAE,OAAO,EAAE,MAAM,CAAC,OAAO,EAAE,gBAAgB,EAAE,MAAM,CAAC,gBAAgB;QACnG,GAAG,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,EAAE,cAAc,EAAE,MAAM,CAAC,cAAwB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,WAAW,EAAE,MAAM,CAAC,WAAqB;QAChI,cAAc,EAAE,WAAW,EAAE,MAAM,CAAC,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,aAAa,CAAC,CAAC,CAAC,EAAE,aAAa,EAAE,MAAM,CAAC,aAAuB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACnI,GAAG,CAAC,MAAM,CAAC,UAAU,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,MAAM,CAAC,UAAoB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,KAAK,EAAE,MAAM,CAAC,KAAK;QAC9F,UAAU,EAAE,cAAc,CAAC,MAAM,CAAC,UAAU,EAAE,2BAA2B,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,MAAM,CAAC,UAAU,EAAE,2BAA2B,CAAC;KACvJ,CAAC;AACJ,CAAC;AAED,SAAS,cAAc,CAAC,cAAuC;IAC7D,IAAI,cAAc,CAAC,YAAY,KAAK,eAAe;QAAE,OAAO,eAAe,CAAC;IAC5E,IAAI,cAAc,CAAC,YAAY,KAAK,cAAc;QAAE,OAAO,cAAc,CAAC;IAC1E,IAAI,cAAc,CAAC,YAAY,KAAK,mBAAmB;QAAE,OAAO,mBAAmB,CAAC;IACpF,OAAO,UAAU,CAAC;AACpB,CAAC;AAED,SAAS,iBAAiB,CAAC,QAAyB;IAClD,MAAM,EAAE,KAAK,EAAE,MAAM,EAAE,UAAU,EAAE,QAAQ,EAAE,UAAU,EAAE,QAAQ,EAAE,UAAU,EAAE,UAAU,EAAE,GAAG,QAAQ,EAAE,GAAG,QAAQ,CAAC;IACpH,OAAO,QAAQ,CAAC;AAClB,CAAC;AAED,SAAS,iBAAiB,CAAC,IAA0B,EAAE,EAAwB;IAC7E,OAAO,IAAI,KAAK,SAAS,IAAI,CAAC,EAAE,KAAK,SAAS,IAAI,EAAE,KAAK,SAAS,CAAC;WAC9D,IAAI,KAAK,SAAS,IAAI,CAAC,EAAE,KAAK,UAAU,IAAI,EAAE,KAAK,SAAS,IAAI,EAAE,KAAK,WAAW,CAAC,CAAC;AAC3F,CAAC;AAED,SAAS,kBAAkB,CAAC,SAA4B;IACtD,OAAO,CAAC,GAAG,SAAS,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;WAC5G,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC,WAAW,IAAI,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,WAAW,CAAC,EAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC;AAC9H,CAAC;AAED,SAAS,cAAc,CAAC,KAAc,EAAE,KAAa;IACnD,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,KAAK,aAAa,CAAC,CAAC;IACjH,OAAO,KAAK,CAAC;AACf,CAAC;AAED,SAAS,OAAO,CAAmB,KAAc,EAAE,OAAqB;IACtE,OAAO,OAAO,KAAK,KAAK,QAAQ,IAAI,OAAO,CAAC,QAAQ,CAAC,KAAU,CAAC,CAAC;AACnE,CAAC"}
@@ -0,0 +1,95 @@
1
+ import { performance } from "node:perf_hooks";
2
+ export class EvalSchedulerMetrics {
3
+ startedAt = performance.now();
4
+ lastTransitionAt = this.startedAt;
5
+ active = 0;
6
+ weightedActiveMs = 0;
7
+ maxActive = 0;
8
+ finalSingleTailStartedAt;
9
+ finalSingleTailMs = 0;
10
+ initialWorkMs = 0;
11
+ retryWorkMs = 0;
12
+ verifierWorkMs = 0;
13
+ resourceBlockedMs = 0;
14
+ collisionBlockedMs = 0;
15
+ backoffBlockedMs = 0;
16
+ running = new Map();
17
+ startWork(workId, kind) {
18
+ if (this.running.has(workId))
19
+ return;
20
+ const now = performance.now();
21
+ this.transition(now);
22
+ this.active += 1;
23
+ this.maxActive = Math.max(this.maxActive, this.active);
24
+ if (this.active === 1)
25
+ this.finalSingleTailStartedAt = now;
26
+ else
27
+ this.finalSingleTailStartedAt = undefined;
28
+ this.running.set(workId, { kind, startedAt: now });
29
+ }
30
+ finishWork(workId) {
31
+ const entry = this.running.get(workId);
32
+ if (!entry)
33
+ return 0;
34
+ const now = performance.now();
35
+ this.transition(now);
36
+ const duration = Math.max(0, now - entry.startedAt);
37
+ if (entry.kind === "initial")
38
+ this.initialWorkMs += duration;
39
+ else
40
+ this.retryWorkMs += duration;
41
+ this.running.delete(workId);
42
+ this.active = Math.max(0, this.active - 1);
43
+ if (this.active === 1)
44
+ this.finalSingleTailStartedAt = now;
45
+ else if (this.active === 0 && this.finalSingleTailStartedAt !== undefined) {
46
+ this.finalSingleTailMs = now - this.finalSingleTailStartedAt;
47
+ }
48
+ return Math.max(1, Math.round(duration));
49
+ }
50
+ addVerifier(durationMs) {
51
+ if (Number.isFinite(durationMs) && durationMs >= 0)
52
+ this.verifierWorkMs += durationMs;
53
+ }
54
+ addBlocked(kind, durationMs) {
55
+ if (!Number.isFinite(durationMs) || durationMs < 0)
56
+ return;
57
+ if (kind === "resource")
58
+ this.resourceBlockedMs += durationMs;
59
+ else if (kind === "collision")
60
+ this.collisionBlockedMs += durationMs;
61
+ else
62
+ this.backoffBlockedMs += durationMs;
63
+ }
64
+ summary(input) {
65
+ const now = performance.now();
66
+ this.transition(now);
67
+ const makespan = Math.max(1, now - this.startedAt);
68
+ const physical = this.initialWorkMs + this.retryWorkMs;
69
+ return {
70
+ policy: input.prioritized ? "critical-path-lpt-v1" : "fifo-compat",
71
+ makespan_ms: Math.round(makespan),
72
+ physical_work_ms: Math.round(physical),
73
+ initial_work_ms: Math.round(this.initialWorkMs),
74
+ retry_work_ms: Math.round(this.retryWorkMs),
75
+ verifier_work_ms: Math.round(this.verifierWorkMs),
76
+ max_active: this.maxActive,
77
+ effective_parallelism: rounded(this.weightedActiveMs / makespan),
78
+ slot_utilization: rounded(this.weightedActiveMs / (makespan * Math.max(1, input.maxParallelism))),
79
+ single_active_tail_ms: Math.round(this.active === 1 && this.finalSingleTailStartedAt !== undefined
80
+ ? now - this.finalSingleTailStartedAt : this.finalSingleTailMs),
81
+ resource_blocked_ms: Math.round(this.resourceBlockedMs),
82
+ collision_blocked_ms: Math.round(this.collisionBlockedMs),
83
+ backoff_blocked_ms: Math.round(this.backoffBlockedMs),
84
+ verifier_skipped: input.verifierSkipped,
85
+ };
86
+ }
87
+ transition(now) {
88
+ this.weightedActiveMs += Math.max(0, now - this.lastTransitionAt) * this.active;
89
+ this.lastTransitionAt = now;
90
+ }
91
+ }
92
+ function rounded(value) {
93
+ return Math.round(value * 1_000) / 1_000;
94
+ }
95
+ //# sourceMappingURL=scheduler-metrics.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"scheduler-metrics.js","sourceRoot":"","sources":["../../../src/evals/scheduler-metrics.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAG9C,MAAM,OAAO,oBAAoB;IACd,SAAS,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;IACvC,gBAAgB,GAAG,IAAI,CAAC,SAAS,CAAC;IAClC,MAAM,GAAG,CAAC,CAAC;IACX,gBAAgB,GAAG,CAAC,CAAC;IACrB,SAAS,GAAG,CAAC,CAAC;IACd,wBAAwB,CAAqB;IAC7C,iBAAiB,GAAG,CAAC,CAAC;IACtB,aAAa,GAAG,CAAC,CAAC;IAClB,WAAW,GAAG,CAAC,CAAC;IAChB,cAAc,GAAG,CAAC,CAAC;IACnB,iBAAiB,GAAG,CAAC,CAAC;IACtB,kBAAkB,GAAG,CAAC,CAAC;IACvB,gBAAgB,GAAG,CAAC,CAAC;IACZ,OAAO,GAAG,IAAI,GAAG,EAA4D,CAAC;IAE/F,SAAS,CAAC,MAAc,EAAE,IAAyB;QACjD,IAAI,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC;YAAE,OAAO;QACrC,MAAM,GAAG,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;QAC9B,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC;QACrB,IAAI,CAAC,MAAM,IAAI,CAAC,CAAC;QACjB,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;QACvD,IAAI,IAAI,CAAC,MAAM,KAAK,CAAC;YAAE,IAAI,CAAC,wBAAwB,GAAG,GAAG,CAAC;;YACtD,IAAI,CAAC,wBAAwB,GAAG,SAAS,CAAC;QAC/C,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,GAAG,EAAE,CAAC,CAAC;IACrD,CAAC;IAED,UAAU,CAAC,MAAc;QACvB,MAAM,KAAK,GAAG,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;QACvC,IAAI,CAAC,KAAK;YAAE,OAAO,CAAC,CAAC;QACrB,MAAM,GAAG,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;QAC9B,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC;QACrB,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,GAAG,KAAK,CAAC,SAAS,CAAC,CAAC;QACpD,IAAI,KAAK,CAAC,IAAI,KAAK,SAAS;YAAE,IAAI,CAAC,aAAa,IAAI,QAAQ,CAAC;;YACxD,IAAI,CAAC,WAAW,IAAI,QAAQ,CAAC;QAClC,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;QAC5B,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QAC3C,IAAI,IAAI,CAAC,MAAM,KAAK,CAAC;YAAE,IAAI,CAAC,wBAAwB,GAAG,GAAG,CAAC;aACtD,IAAI,IAAI,CAAC,MAAM,KAAK,CAAC,IAAI,IAAI,CAAC,wBAAwB,KAAK,SAAS,EAAE,CAAC;YAC1E,IAAI,CAAC,iBAAiB,GAAG,GAAG,GAAG,IAAI,CAAC,wBAAwB,CAAC;QAC/D,CAAC;QACD,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC,CAAC;IAC3C,CAAC;IAED,WAAW,CAAC,UAAkB;QAC5B,IAAI,MAAM,CAAC,QAAQ,CAAC,UAAU,CAAC,IAAI,UAAU,IAAI,CAAC;YAAE,IAAI,CAAC,cAAc,IAAI,UAAU,CAAC;IACxF,CAAC;IAED,UAAU,CAAC,IAA0C,EAAE,UAAkB;QACvE,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,UAAU,CAAC,IAAI,UAAU,GAAG,CAAC;YAAE,OAAO;QAC3D,IAAI,IAAI,KAAK,UAAU;YAAE,IAAI,CAAC,iBAAiB,IAAI,UAAU,CAAC;aACzD,IAAI,IAAI,KAAK,WAAW;YAAE,IAAI,CAAC,kBAAkB,IAAI,UAAU,CAAC;;YAChE,IAAI,CAAC,gBAAgB,IAAI,UAAU,CAAC;IAC3C,CAAC;IAED,OAAO,CAAC,KAAgF;QACtF,MAAM,GAAG,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;QAC9B,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC;QACrB,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,GAAG,IAAI,CAAC,SAAS,CAAC,CAAC;QACnD,MAAM,QAAQ,GAAG,IAAI,CAAC,aAAa,GAAG,IAAI,CAAC,WAAW,CAAC;QACvD,OAAO;YACL,MAAM,EAAE,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC,sBAAsB,CAAC,CAAC,CAAC,aAAa;YAClE,WAAW,EAAE,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC;YACjC,gBAAgB,EAAE,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC;YACtC,eAAe,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,aAAa,CAAC;YAC/C,aAAa,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,WAAW,CAAC;YAC3C,gBAAgB,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,cAAc,CAAC;YACjD,UAAU,EAAE,IAAI,CAAC,SAAS;YAC1B,qBAAqB,EAAE,OAAO,CAAC,IAAI,CAAC,gBAAgB,GAAG,QAAQ,CAAC;YAChE,gBAAgB,EAAE,OAAO,CAAC,IAAI,CAAC,gBAAgB,GAAG,CAAC,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,cAAc,CAAC,CAAC,CAAC;YACjG,qBAAqB,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,KAAK,CAAC,IAAI,IAAI,CAAC,wBAAwB,KAAK,SAAS;gBAChG,CAAC,CAAC,GAAG,GAAG,IAAI,CAAC,wBAAwB,CAAC,CAAC,CAAC,IAAI,CAAC,iBAAiB,CAAC;YACjE,mBAAmB,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,iBAAiB,CAAC;YACvD,oBAAoB,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,kBAAkB,CAAC;YACzD,kBAAkB,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,gBAAgB,CAAC;YACrD,gBAAgB,EAAE,KAAK,CAAC,eAAe;SACxC,CAAC;IACJ,CAAC;IAEO,UAAU,CAAC,GAAW;QAC5B,IAAI,CAAC,gBAAgB,IAAI,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,GAAG,IAAI,CAAC,gBAAgB,CAAC,GAAG,IAAI,CAAC,MAAM,CAAC;QAChF,IAAI,CAAC,gBAAgB,GAAG,GAAG,CAAC;IAC9B,CAAC;CACF;AAED,SAAS,OAAO,CAAC,KAAa;IAC5B,OAAO,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,GAAG,KAAK,CAAC;AAC3C,CAAC"}
@@ -0,0 +1,107 @@
1
+ export function replaySchedulerTrace(work, options) {
2
+ if (!Number.isSafeInteger(options.slots) || options.slots < 1)
3
+ throw new TypeError("trace replay slots are invalid");
4
+ const chains = traceChains(work);
5
+ const replay = options.retryScheduling === "batch-v1"
6
+ ? replayBatch(chains, options.slots, options.policy)
7
+ : replayImmediate(chains, options.slots, options.policy);
8
+ const physicalWorkMs = work.reduce((sum, item) => sum + item.duration_ms, 0);
9
+ return {
10
+ makespan_ms: replay.makespanMs,
11
+ physical_work_ms: physicalWorkMs,
12
+ max_active: replay.maxActive,
13
+ slot_utilization: rounded(physicalWorkMs / (Math.max(1, replay.makespanMs) * options.slots)),
14
+ runnable_idle_ms: 0,
15
+ };
16
+ }
17
+ function traceChains(work) {
18
+ const byTask = new Map();
19
+ const ids = new Set();
20
+ for (const item of work) {
21
+ validateTraceWork(item);
22
+ if (ids.has(item.work_id))
23
+ throw new TypeError(`trace work identity is duplicated: ${item.work_id}`);
24
+ ids.add(item.work_id);
25
+ const stages = byTask.get(item.task_id) ?? [];
26
+ stages.push(item);
27
+ byTask.set(item.task_id, stages);
28
+ }
29
+ const chains = [...byTask.entries()].map(([taskId, stages]) => ({
30
+ taskId,
31
+ stages: stages.sort((left, right) => left.retry_index - right.retry_index),
32
+ }));
33
+ for (const chain of chains) {
34
+ if (chain.stages[0]?.execution_kind !== "initial" || chain.stages[0]?.retry_index !== 0
35
+ || chain.stages.some((stage, index) => stage.retry_index !== index
36
+ || (index > 0 && stage.execution_kind !== "physical-infrastructure-retry"))) {
37
+ throw new TypeError(`trace work chain is invalid: ${chain.taskId}`);
38
+ }
39
+ }
40
+ return chains.sort((left, right) => compareBytes(left.taskId, right.taskId));
41
+ }
42
+ function replayBatch(chains, slots, policy) {
43
+ let now = 0;
44
+ let maxActive = 0;
45
+ const stages = Math.max(0, ...chains.map((chain) => chain.stages.length));
46
+ for (let index = 0; index < stages; index += 1) {
47
+ const ready = chains.flatMap((chain) => chain.stages[index]
48
+ ? [{ chain, index, remainingMs: remaining(chain, index) }]
49
+ : []);
50
+ maxActive = Math.max(maxActive, Math.min(slots, ready.length));
51
+ now = scheduleReady(ready, slots, policy, now);
52
+ }
53
+ return { makespanMs: now, maxActive };
54
+ }
55
+ function replayImmediate(chains, slots, policy) {
56
+ const ready = chains.map((chain) => ({ chain, index: 0, remainingMs: remaining(chain, 0) }));
57
+ const running = [];
58
+ let now = 0;
59
+ let maxActive = 0;
60
+ while (ready.length > 0 || running.length > 0) {
61
+ ready.sort((left, right) => compareReady(left, right, policy));
62
+ while (running.length < slots && ready.length > 0) {
63
+ const item = ready.shift();
64
+ running.push({ ...item, completedAt: now + (item.chain.stages[item.index]?.duration_ms ?? 0) });
65
+ }
66
+ maxActive = Math.max(maxActive, running.length);
67
+ running.sort((left, right) => left.completedAt - right.completedAt || compareBytes(left.chain.taskId, right.chain.taskId));
68
+ const completed = running.shift();
69
+ if (!completed)
70
+ break;
71
+ now = completed.completedAt;
72
+ const next = completed.index + 1;
73
+ if (completed.chain.stages[next])
74
+ ready.push({ chain: completed.chain, index: next, remainingMs: remaining(completed.chain, next) });
75
+ }
76
+ return { makespanMs: now, maxActive };
77
+ }
78
+ function scheduleReady(ready, slots, policy, start) {
79
+ ready.sort((left, right) => compareReady(left, right, policy));
80
+ const available = Array.from({ length: slots }, () => start);
81
+ for (const item of ready) {
82
+ const earliest = Math.min(...available);
83
+ const slot = available.indexOf(earliest);
84
+ available[slot] = earliest + (item.chain.stages[item.index]?.duration_ms ?? 0);
85
+ }
86
+ return Math.max(start, ...available);
87
+ }
88
+ function compareReady(left, right, policy) {
89
+ if (policy === "critical-path-v1" && left.remainingMs !== right.remainingMs)
90
+ return right.remainingMs - left.remainingMs;
91
+ const leftStage = left.chain.stages[left.index];
92
+ const rightStage = right.chain.stages[right.index];
93
+ return leftStage.queued_order - rightStage.queued_order || compareBytes(leftStage.work_id, rightStage.work_id);
94
+ }
95
+ function remaining(chain, index) {
96
+ return chain.stages.slice(index).reduce((sum, stage) => sum + stage.duration_ms, 0);
97
+ }
98
+ function validateTraceWork(item) {
99
+ if (!item.work_id || !item.task_id || !Number.isSafeInteger(item.duration_ms) || item.duration_ms < 1
100
+ || !Number.isSafeInteger(item.queued_order) || item.queued_order < 0
101
+ || !Number.isSafeInteger(item.retry_index) || item.retry_index < 0
102
+ || (item.execution_kind === "initial") !== (item.retry_index === 0))
103
+ throw new TypeError("scheduler trace work is invalid");
104
+ }
105
+ function rounded(value) { return Math.round(value * 1_000) / 1_000; }
106
+ function compareBytes(left, right) { return Buffer.compare(Buffer.from(left), Buffer.from(right)); }
107
+ //# sourceMappingURL=scheduler-trace-replay.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"scheduler-trace-replay.js","sourceRoot":"","sources":["../../../src/evals/scheduler-trace-replay.ts"],"names":[],"mappings":"AAiBA,MAAM,UAAU,oBAAoB,CAClC,IAAqC,EACrC,OAIC;IAED,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,OAAO,CAAC,KAAK,GAAG,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,gCAAgC,CAAC,CAAC;IACrH,MAAM,MAAM,GAAG,WAAW,CAAC,IAAI,CAAC,CAAC;IACjC,MAAM,MAAM,GAAG,OAAO,CAAC,eAAe,KAAK,UAAU;QACnD,CAAC,CAAC,WAAW,CAAC,MAAM,EAAE,OAAO,CAAC,KAAK,EAAE,OAAO,CAAC,MAAM,CAAC;QACpD,CAAC,CAAC,eAAe,CAAC,MAAM,EAAE,OAAO,CAAC,KAAK,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;IAC3D,MAAM,cAAc,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,IAAI,EAAE,EAAE,CAAC,GAAG,GAAG,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC;IAC7E,OAAO;QACL,WAAW,EAAE,MAAM,CAAC,UAAU;QAC9B,gBAAgB,EAAE,cAAc;QAChC,UAAU,EAAE,MAAM,CAAC,SAAS;QAC5B,gBAAgB,EAAE,OAAO,CAAC,cAAc,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,MAAM,CAAC,UAAU,CAAC,GAAG,OAAO,CAAC,KAAK,CAAC,CAAC;QAC5F,gBAAgB,EAAE,CAAC;KACpB,CAAC;AACJ,CAAC;AAkBD,SAAS,WAAW,CAAC,IAAqC;IACxD,MAAM,MAAM,GAAG,IAAI,GAAG,EAAkC,CAAC;IACzD,MAAM,GAAG,GAAG,IAAI,GAAG,EAAU,CAAC;IAC9B,KAAK,MAAM,IAAI,IAAI,IAAI,EAAE,CAAC;QACxB,iBAAiB,CAAC,IAAI,CAAC,CAAC;QACxB,IAAI,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,sCAAsC,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC;QACrG,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;QACtB,MAAM,MAAM,GAAG,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;QAC9C,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAClB,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;IACnC,CAAC;IACD,MAAM,MAAM,GAAG,CAAC,GAAG,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,EAAE,CAAC,CAAC;QAC9D,MAAM;QACN,MAAM,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC,WAAW,CAAC;KAC3E,CAAC,CAAC,CAAC;IACJ,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,IAAI,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,cAAc,KAAK,SAAS,IAAI,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,WAAW,KAAK,CAAC;eAClF,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,WAAW,KAAK,KAAK;mBAC7D,CAAC,KAAK,GAAG,CAAC,IAAI,KAAK,CAAC,cAAc,KAAK,+BAA+B,CAAC,CAAC,EAAE,CAAC;YAChF,MAAM,IAAI,SAAS,CAAC,gCAAgC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC;QACtE,CAAC;IACH,CAAC;IACD,OAAO,MAAM,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,YAAY,CAAC,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC;AAC/E,CAAC;AAED,SAAS,WAAW,CAAC,MAA6B,EAAE,KAAa,EAAE,MAAsC;IACvG,IAAI,GAAG,GAAG,CAAC,CAAC;IACZ,IAAI,SAAS,GAAG,CAAC,CAAC;IAClB,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC;IAC1E,KAAK,IAAI,KAAK,GAAG,CAAC,EAAE,KAAK,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC,EAAE,CAAC;QAC/C,MAAM,KAAK,GAAG,MAAM,CAAC,OAAO,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC;YACzD,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,WAAW,EAAE,SAAS,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,CAAC;YAC1D,CAAC,CAAC,EAAE,CAAC,CAAC;QACR,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC;QAC/D,GAAG,GAAG,aAAa,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,GAAG,CAAC,CAAC;IACjD,CAAC;IACD,OAAO,EAAE,UAAU,EAAE,GAAG,EAAE,SAAS,EAAE,CAAC;AACxC,CAAC;AAED,SAAS,eAAe,CAAC,MAA6B,EAAE,KAAa,EAAE,MAAsC;IAC3G,MAAM,KAAK,GAAiB,MAAM,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,EAAE,WAAW,EAAE,SAAS,CAAC,KAAK,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;IAC3G,MAAM,OAAO,GAAgD,EAAE,CAAC;IAChE,IAAI,GAAG,GAAG,CAAC,CAAC;IACZ,IAAI,SAAS,GAAG,CAAC,CAAC;IAClB,OAAO,KAAK,CAAC,MAAM,GAAG,CAAC,IAAI,OAAO,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC9C,KAAK,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,YAAY,CAAC,IAAI,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC,CAAC;QAC/D,OAAO,OAAO,CAAC,MAAM,GAAG,KAAK,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAClD,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,EAAgB,CAAC;YACzC,OAAO,CAAC,IAAI,CAAC,EAAE,GAAG,IAAI,EAAE,WAAW,EAAE,GAAG,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,WAAW,IAAI,CAAC,CAAC,EAAE,CAAC,CAAC;QAClG,CAAC;QACD,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;QAChD,OAAO,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC,WAAW,IAAI,YAAY,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,EAAE,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC;QAC3H,MAAM,SAAS,GAAG,OAAO,CAAC,KAAK,EAAE,CAAC;QAClC,IAAI,CAAC,SAAS;YAAE,MAAM;QACtB,GAAG,GAAG,SAAS,CAAC,WAAW,CAAC;QAC5B,MAAM,IAAI,GAAG,SAAS,CAAC,KAAK,GAAG,CAAC,CAAC;QACjC,IAAI,SAAS,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC;YAAE,KAAK,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,SAAS,CAAC,KAAK,EAAE,KAAK,EAAE,IAAI,EAAE,WAAW,EAAE,SAAS,CAAC,SAAS,CAAC,KAAK,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC;IACvI,CAAC;IACD,OAAO,EAAE,UAAU,EAAE,GAAG,EAAE,SAAS,EAAE,CAAC;AACxC,CAAC;AAED,SAAS,aAAa,CAAC,KAAmB,EAAE,KAAa,EAAE,MAAsC,EAAE,KAAa;IAC9G,KAAK,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,YAAY,CAAC,IAAI,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC,CAAC;IAC/D,MAAM,SAAS,GAAG,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE,EAAE,GAAG,EAAE,CAAC,KAAK,CAAC,CAAC;IAC7D,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;QACzB,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,SAAS,CAAC,CAAC;QACxC,MAAM,IAAI,GAAG,SAAS,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC;QACzC,SAAS,CAAC,IAAI,CAAC,GAAG,QAAQ,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,WAAW,IAAI,CAAC,CAAC,CAAC;IACjF,CAAC;IACD,OAAO,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,GAAG,SAAS,CAAC,CAAC;AACvC,CAAC;AAED,SAAS,YAAY,CAAC,IAAgB,EAAE,KAAiB,EAAE,MAAsC;IAC/F,IAAI,MAAM,KAAK,kBAAkB,IAAI,IAAI,CAAC,WAAW,KAAK,KAAK,CAAC,WAAW;QAAE,OAAO,KAAK,CAAC,WAAW,GAAG,IAAI,CAAC,WAAW,CAAC;IACzH,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAyB,CAAC;IACxE,MAAM,UAAU,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAyB,CAAC;IAC3E,OAAO,SAAS,CAAC,YAAY,GAAG,UAAU,CAAC,YAAY,IAAI,YAAY,CAAC,SAAS,CAAC,OAAO,EAAE,UAAU,CAAC,OAAO,CAAC,CAAC;AACjH,CAAC;AAED,SAAS,SAAS,CAAC,KAAiB,EAAE,KAAa;IACjD,OAAO,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC;AACtF,CAAC;AAED,SAAS,iBAAiB,CAAC,IAA0B;IACnD,IAAI,CAAC,IAAI,CAAC,OAAO,IAAI,CAAC,IAAI,CAAC,OAAO,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,WAAW,CAAC,IAAI,IAAI,CAAC,WAAW,GAAG,CAAC;WAChG,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,IAAI,CAAC,YAAY,GAAG,CAAC;WACjE,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,WAAW,CAAC,IAAI,IAAI,CAAC,WAAW,GAAG,CAAC;WAC/D,CAAC,IAAI,CAAC,cAAc,KAAK,SAAS,CAAC,KAAK,CAAC,IAAI,CAAC,WAAW,KAAK,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,iCAAiC,CAAC,CAAC;AAChI,CAAC;AAED,SAAS,OAAO,CAAC,KAAa,IAAY,OAAO,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,GAAG,KAAK,CAAC,CAAC,CAAC;AACrF,SAAS,YAAY,CAAC,IAAY,EAAE,KAAa,IAAY,OAAO,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC"}
@@ -7,14 +7,13 @@ import { ensureControllerRuntime, writeRuntimeReference } from "../controller-ru
7
7
  import { importEvalTrialRun, importEvalTrialRuns, TrialBundlePendingError, validateEvalTrialReferences } from "./trial-import.js";
8
8
  import { EvalEventSink } from "./events.js";
9
9
  import { createEvalProgress, mergeEvalProgressTrial, writeEvalProgress } from "./progress.js";
10
- import { infrastructureFailureTrials, runInfrastructureRetries } from "./infrastructure-retry.js";
10
+ import { runInfrastructureRetries } from "./infrastructure-retry.js";
11
11
  import { newEvalId, resolveLocalDatasetTaskIds, validateEvalId, validateEvalRequest } from "./request.js";
12
- import { invalidTrialSlots } from "./rerun-slots.js";
13
12
  import { prepareEvalDirectory } from "./directory.js";
14
13
  import { buildEvalExecutionPlan, DEFAULT_EVAL_TRIAL_RESOURCES } from "./execution-plan.js";
15
14
  import { planLocalEvalInputs } from "./local-eval-planning.js";
16
15
  import { resolvedImageMapping } from "./environment-image-planning.js";
17
- import { assertBackendTrialSet, attemptDirectoryName, localSourceBackendFailure, summarizeTrialRefs, transportSummary } from "./result-helpers.js";
16
+ import { assertBackendTrialSet, attemptDirectoryName, localSourceBackendFailure, transportSummary } from "./result-helpers.js";
18
17
  import { executePlannedHarborTasks } from "./planned-execution.js";
19
18
  import { assertEvalResumeState, executionPlanWorkState, loadEvalResumeState } from "./resume-state.js";
20
19
  import { modelCaptureDegradationEvent, resolveEvalModelCapturePlan } from "./model-capture-plan.js";
@@ -25,7 +24,11 @@ import { prepareEvalArtifactAssignments, preparedArtifactPlanFields } from "./ev
25
24
  import { startEvalModelCaptureRuntime } from "./model-capture-runtime.js";
26
25
  import { recoverPromotedEvalTrialPublications } from "./trial-publication-recovery.js";
27
26
  import { emitEvalPlanLifecycle } from "./eval-lifecycle-events.js";
28
- export async function runEval({ evalId = newEvalId(), request, root, env = process.env, harborExecutable, signal, onEvent, trialBundleGraceMs, precreated = false, normalizedRequest, maxConcurrentOverride, executionResources, executionResourceSource = "operator-default", executionStrategy = "legacy-attempt-shards", executionWorker, modelCapturePlan, workItemAdmission, remoteWorkExecutor, resumeExisting = false, onControlPhase, onWorkItemState, dockerResourceReaper, environmentBuildMode = "backend", environmentImageResolver, environmentImageBuilder, environmentImageManifestLoader, harborArtifactBuilder }) {
27
+ import { materializeEvalPlan, writeEvalPlanningCheckpoint } from "./eval-logical-plan.js";
28
+ import { planTaskSchedulingHints, schedulingHintsFromPlan } from "./duration-estimator.js";
29
+ import { buildCompletedEvalResult } from "./eval-result-builder.js";
30
+ import { loadBenchmarkAdapterManifest } from "./benchmark-adapter-manifest.js";
31
+ export async function runEval({ evalId = newEvalId(), request, root, env = process.env, harborExecutable, signal, onEvent, trialBundleGraceMs, precreated = false, replaceTerminal = false, normalizedRequest, maxConcurrentOverride, executionResources, executionResourceSource = "operator-default", executionStrategy = "legacy-attempt-shards", executionWorker, modelCapturePlan, workItemAdmission, remoteWorkExecutor, resumeExisting = false, onControlPhase, onWorkItemState, onWorkItemQueued, evolutionBaselineDurations, dockerResourceReaper, environmentBuildMode = "backend", environmentImageResolver, environmentImageBuilder, environmentImageManifestLoader, harborArtifactBuilder }) {
29
32
  if (!root)
30
33
  throw invalidInput("a Hitch state root is required for eval");
31
34
  evalId = validateEvalId(evalId);
@@ -35,14 +38,16 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
35
38
  }
36
39
  const normalized = maxConcurrentOverride === undefined ? persistedRequest : { ...persistedRequest, max_concurrent: maxConcurrentOverride };
37
40
  const evalsDirectory = await ensureDir(statePaths(root).evals);
38
- const evalDirectory = await prepareEvalDirectory({ evalsDirectory, evalId, request: persistedRequest, precreated });
41
+ const evalDirectory = await prepareEvalDirectory({ evalsDirectory, evalId, request: persistedRequest, precreated, replaceTerminal });
39
42
  let startedAt = new Date();
40
43
  const sink = new EvalEventSink(evalDirectory, evalId, onEvent);
41
44
  await sink.open();
42
45
  let result;
43
46
  let trialRefs = [];
44
47
  let progress = null;
48
+ let schedulerSummary;
45
49
  let captureRuntime;
50
+ let failureStage = "planning";
46
51
  try {
47
52
  const planningStartedAt = Date.now();
48
53
  sink.emit({ type: "eval.started", backend: normalized.backend, dataset: normalized.dataset });
@@ -50,6 +55,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
50
55
  await onControlPhase?.("planning");
51
56
  const requestedReference = parseHarnessReference(normalized.harness_ref);
52
57
  const resolvedRevision = await resolveHarness(normalized.harness_ref, { root, env });
58
+ failureStage = "preparing";
53
59
  await onControlPhase?.("preparing");
54
60
  let localTransport;
55
61
  if (resolvedRevision.source.type === "git" && resolvedRevision.source.registered !== true) {
@@ -84,16 +90,11 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
84
90
  reference: runtimeRefFile,
85
91
  });
86
92
  const localTaskIds = await resolveLocalDatasetTaskIds(normalized.dataset);
93
+ const standardDataset = localTaskIds !== null && await loadBenchmarkAdapterManifest(normalized.dataset) !== null;
87
94
  const resume = resumeExisting ? await loadEvalResumeState(evalDirectory) : null;
88
95
  await withEnvironmentImageReferenceLock(root, () => beginEvalEnvironmentImagePlanning(evalDirectory, evalId));
89
96
  const localPlanning = await planLocalEvalInputs({ root, dataset: normalized.dataset, taskIds: localTaskIds, defaultResources: executionResources ?? DEFAULT_EVAL_TRIAL_RESOURCES, defaultSource: executionResourceSource, benchmarkId: normalized.benchmark_id, benchmarkRevision: normalized.benchmark_revision, buildMode: environmentBuildMode, harborTaskResourceInspector: path.join(controllerRuntime.directory, "payload", "integrations", "harbor", "hitch_harbor_task_resources.py"), ...(environmentImageResolver ? { resolver: environmentImageResolver } : {}), ...(environmentImageBuilder ? { builder: environmentImageBuilder } : {}), ...(resume ? { resumePlan: resume.executionPlan } : {}), ...(harborExecutable ? { harborExecutable } : {}), env, ...(signal ? { signal } : {}) });
90
97
  await withEnvironmentImageReferenceLock(root, () => writeEvalEnvironmentImageReferences(evalDirectory, evalId, localPlanning.environmentImages));
91
- const selectedArtifactBuilder = harborArtifactBuilder ?? (env.NODE_ENV === "test" && env.HITCH_TEST_HOST_ARTIFACT_BUILDER === "1" ? prepareHostHarborArtifactForTest : prepareHarborArtifact);
92
- const prepared = await prepareEvalArtifactAssignments({ builder: selectedArtifactBuilder, root, resolvedRevision, requestedReference, controllerRuntime, ...(localTransport ? { localTransport } : {}), taskRuntimeContracts: localPlanning.taskRuntimeContracts, sink, env, ...(signal ? { signal } : {}) });
93
- const preparedAssignments = prepared.assignments;
94
- const preparedArtifact = prepared.primary;
95
- const preparedArtifacts = prepared.artifactsById;
96
- const multipleRuntimeContracts = preparedAssignments.length > 1;
97
98
  const taskResources = localPlanning.taskResources;
98
99
  const plannedTasks = localTaskIds?.length ?? null;
99
100
  const plannedTrials = plannedTasks === null ? null : plannedTasks * normalized.attempts;
@@ -102,26 +103,18 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
102
103
  if (resume)
103
104
  startedAt = new Date(resume.progress.started_at);
104
105
  const capture = resolveEvalModelCapturePlan({ requested: modelCapturePlan, resumed: resume?.executionPlan.model_capture, resuming: Boolean(resume) });
105
- const activeCaptureRuntime = await startEvalModelCaptureRuntime({ plan: capture.plan, evalId, evalDirectory, env });
106
- captureRuntime = activeCaptureRuntime;
107
- capture.plan = activeCaptureRuntime.plan;
108
- const plan = {
106
+ const multipleRuntimeContracts = localPlanning.taskRuntimeContracts.length > 1;
107
+ const logicalPlan = {
109
108
  schema_version: SCHEMA_VERSION,
109
+ kind: "eval-logical-plan",
110
110
  eval_id: evalId,
111
111
  backend: "harbor",
112
- candidate: {
113
- id: "candidate-1",
114
- requested_harness_ref: normalized.harness_ref,
115
- harness_ref: lockedHarnessRef(resolvedRevision),
116
- harness_id: resolvedRevision.harness_id,
117
- revision_identity: resolvedRevision.identity,
118
- model: normalized.model || null,
119
- },
112
+ candidate: { id: "candidate-1", requested_harness_ref: normalized.harness_ref, harness_ref: lockedHarnessRef(resolvedRevision), harness_id: resolvedRevision.harness_id, revision_identity: resolvedRevision.identity, model: normalized.model || null },
120
113
  dataset: normalized.dataset,
121
114
  benchmark_id: normalized.benchmark_id,
122
115
  benchmark_revision: normalized.benchmark_revision,
123
116
  attempts: normalized.attempts,
124
- attempt_execution: (executionStrategy === "local-task-slots-v1" || multipleRuntimeContracts) && localTaskIds !== null
117
+ attempt_execution: (executionStrategy === "local-task-slots-v1" || multipleRuntimeContracts || standardDataset) && localTaskIds !== null
125
118
  ? "harbor-task-slots-v1"
126
119
  : "harbor-attempt-shards-v1",
127
120
  max_concurrent: normalized.max_concurrent,
@@ -129,14 +122,49 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
129
122
  infrastructure_retries: normalized.infrastructure_retries,
130
123
  infrastructure_retry_backoff_ms: normalized.infrastructure_retry_backoff_ms,
131
124
  ...(localTaskIds === null ? {} : { tasks: localTaskIds }),
132
- controller_runtime: {
133
- runtime_id: controllerRuntime.runtime_id,
134
- manifest_digest: controllerRuntime.manifest_digest,
135
- },
136
- ...preparedArtifactPlanFields(preparedAssignments),
125
+ controller_runtime: { runtime_id: controllerRuntime.runtime_id, manifest_digest: controllerRuntime.manifest_digest },
137
126
  ...(localTransport ? { local_source_transport: transportSummary(localTransport) } : {}),
138
127
  created_at: typeof resume?.plan.created_at === "string" ? resume.plan.created_at : new Date().toISOString(),
139
128
  };
129
+ progress = resume?.progress ?? createEvalProgress({
130
+ evalId,
131
+ benchmarkId: normalized.benchmark_id,
132
+ benchmarkRevision: normalized.benchmark_revision,
133
+ plannedTasks,
134
+ plannedTrials,
135
+ startedAt: startedAt.toISOString(),
136
+ });
137
+ if (!resume)
138
+ await Promise.all([
139
+ atomicWriteJSON(path.join(evalDirectory, "resolution.json"), resolvedRevision),
140
+ writeEvalPlanningCheckpoint(evalDirectory, logicalPlan, progress),
141
+ ]);
142
+ const selectedArtifactBuilder = harborArtifactBuilder ?? (env.NODE_ENV === "test" && env.HITCH_TEST_HOST_ARTIFACT_BUILDER === "1" ? prepareHostHarborArtifactForTest : prepareHarborArtifact);
143
+ const prepared = await prepareEvalArtifactAssignments({ builder: selectedArtifactBuilder, root, resolvedRevision, requestedReference, controllerRuntime, ...(localTransport ? { localTransport } : {}), taskRuntimeContracts: localPlanning.taskRuntimeContracts, sink, env, ...(signal ? { signal } : {}) });
144
+ const preparedAssignments = prepared.assignments;
145
+ const preparedArtifact = prepared.primary;
146
+ const preparedArtifacts = prepared.artifactsById;
147
+ const activeCaptureRuntime = await startEvalModelCaptureRuntime({ plan: capture.plan, evalId, evalDirectory, env });
148
+ captureRuntime = activeCaptureRuntime;
149
+ capture.plan = activeCaptureRuntime.plan;
150
+ const plan = materializeEvalPlan(logicalPlan, capture.persist ? capture.plan : undefined, preparedArtifactPlanFields(preparedAssignments));
151
+ const taskSlotPlanning = (executionStrategy === "local-task-slots-v1" || multipleRuntimeContracts || standardDataset) && localTaskIds !== null;
152
+ const taskScheduling = taskSlotPlanning
153
+ ? resume
154
+ ? schedulingHintsFromPlan(resume.executionPlan)
155
+ : await planTaskSchedulingHints({
156
+ root,
157
+ dataset: normalized.dataset,
158
+ taskIds: localTaskIds,
159
+ benchmarkId: normalized.benchmark_id,
160
+ benchmarkRevision: normalized.benchmark_revision,
161
+ provider: executionWorker?.provider ?? "local-docker",
162
+ model: normalized.model,
163
+ requestTimeoutMs: normalized.timeout_ms,
164
+ infrastructureRetries: normalized.infrastructure_retries,
165
+ ...(evolutionBaselineDurations ? { evolutionBaselineDurations } : {}),
166
+ })
167
+ : undefined;
140
168
  const expectedExecutionPlan = buildEvalExecutionPlan({
141
169
  evalId,
142
170
  request: normalized,
@@ -161,14 +189,14 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
161
189
  ...(localPlanning.environmentImageFallbacks.length > 0 ? { environmentImageFallbacks: localPlanning.environmentImageFallbacks } : {}),
162
190
  ...(executionWorker ? { provider: executionWorker.provider } : {}),
163
191
  modelCapture: capture.persist ? capture.plan : null,
164
- ...((executionStrategy === "local-task-slots-v1" || multipleRuntimeContracts) && localTaskIds !== null ? { workItemMode: "task-slots" } : {}),
165
- createdAt: plan.created_at,
192
+ ...(taskSlotPlanning ? { workItemMode: "task-slots" } : {}),
193
+ ...(taskScheduling ? { taskScheduling } : {}),
194
+ createdAt: logicalPlan.created_at,
166
195
  });
167
196
  if (resume)
168
197
  assertEvalResumeState({ state: resume, expectedPlan: plan, expectedExecutionPlan, expectedResolutionIdentity: resolvedRevision.identity, plannedTasks, plannedTrials });
169
198
  else
170
199
  await Promise.all([
171
- atomicWriteJSON(path.join(evalDirectory, "resolution.json"), resolvedRevision),
172
200
  atomicWriteJSON(path.join(evalDirectory, "plan.json"), plan),
173
201
  atomicWriteJSON(path.join(evalDirectory, "execution-plan.json"), expectedExecutionPlan),
174
202
  ]);
@@ -176,17 +204,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
176
204
  const captureDegradation = modelCaptureDegradationEvent(executionPlan.model_capture);
177
205
  if (captureDegradation)
178
206
  sink.emit(captureDegradation);
179
- progress = resume?.progress ?? createEvalProgress({
180
- evalId,
181
- benchmarkId: normalized.benchmark_id,
182
- benchmarkRevision: normalized.benchmark_revision,
183
- plannedTasks,
184
- plannedTrials,
185
- startedAt: startedAt.toISOString(),
186
- });
187
- if (!resume)
188
- await writeEvalProgress(evalDirectory, progress);
189
- else
207
+ if (resume)
190
208
  progress = (await recoverPromotedEvalTrialPublications({ root, evalDirectory, plan: executionPlan, progress, sink })).progress;
191
209
  sink.emit({
192
210
  type: "eval.planned",
@@ -198,7 +216,18 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
198
216
  work_items: executionPlan.work_items.length,
199
217
  membership: executionPlan.membership,
200
218
  });
219
+ sink.emit({
220
+ type: "eval.execution-strategy.verified",
221
+ execution_strategy: taskSlotPlanning ? "local-task-slots-v1" : "legacy-attempt-shards",
222
+ attempt_execution: plan.attempt_execution,
223
+ controller_runtime_id: controllerRuntime.runtime_id,
224
+ harness_revision_identity: resolvedRevision.identity,
225
+ membership: executionPlan.membership,
226
+ planned_trials: plannedTrials,
227
+ work_items: executionPlan.work_items.length,
228
+ });
201
229
  emitEvalPlanLifecycle(sink, executionPlan, planningStartedAt);
230
+ failureStage = "running";
202
231
  await onControlPhase?.("running", executionPlanWorkState(executionPlan, progress));
203
232
  if (localTransport) {
204
233
  await verifyLocalGitTransport(localTransport, {
@@ -213,7 +242,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
213
242
  }
214
243
  const backendRuns = [];
215
244
  const infrastructureRetryRuns = [];
216
- const plannedTaskExecution = (executionStrategy === "local-task-slots-v1" || multipleRuntimeContracts) && localTaskIds !== null;
245
+ const plannedTaskExecution = (executionStrategy === "local-task-slots-v1" || multipleRuntimeContracts || standardDataset) && localTaskIds !== null;
217
246
  if (plannedTaskExecution) {
218
247
  const execution = await executePlannedHarborTasks({
219
248
  evalId,
@@ -239,6 +268,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
239
268
  },
240
269
  ...(workItemAdmission ? { admission: workItemAdmission } : {}),
241
270
  ...(onWorkItemState ? { onWorkItemState } : {}),
271
+ ...(onWorkItemQueued ? { onWorkItemQueued } : {}),
242
272
  ...(dockerResourceReaper ? { dockerResourceReaper } : {}),
243
273
  ...(environmentImageManifestLoader ? { environmentImageManifestLoader } : {}),
244
274
  ...(activeCaptureRuntime.exporter ? { interactionCaptureExporter: activeCaptureRuntime.exporter } : {}),
@@ -253,6 +283,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
253
283
  leaseId: entry.leaseId,
254
284
  })));
255
285
  infrastructureRetryRuns.push(...execution.infrastructureRetryRuns);
286
+ schedulerSummary = execution.schedulerSummary;
256
287
  }
257
288
  else
258
289
  for (let logicalAttempt = 1; logicalAttempt <= normalized.attempts; logicalAttempt += 1) {
@@ -326,7 +357,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
326
357
  benchmarkId: normalized.benchmark_id,
327
358
  benchmarkRevision: normalized.benchmark_revision,
328
359
  runtimeId: controllerRuntime.runtime_id,
329
- env,
360
+ env, ...(signal ? { signal } : {}),
330
361
  modelCapturePlan: activeCaptureRuntime.plan,
331
362
  ...(activeCaptureRuntime.exporter ? { interactionCaptureExporter: activeCaptureRuntime.exporter } : {}),
332
363
  requireCompleteMarker: true,
@@ -354,7 +385,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
354
385
  benchmarkId: normalized.benchmark_id,
355
386
  benchmarkRevision: normalized.benchmark_revision,
356
387
  runtimeId: controllerRuntime.runtime_id,
357
- env,
388
+ env, ...(signal ? { signal } : {}),
358
389
  modelCapturePlan: activeCaptureRuntime.plan,
359
390
  ...(activeCaptureRuntime.exporter ? { interactionCaptureExporter: activeCaptureRuntime.exporter } : {}),
360
391
  rawResult: backendRun.rawResult,
@@ -400,100 +431,17 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
400
431
  }
401
432
  if (progress === null)
402
433
  throw new Error("eval progress was not initialized");
434
+ failureStage = "finalizing";
403
435
  await onControlPhase?.("finalizing");
404
436
  trialRefs = progress.trials;
405
- const cancelled = signal?.aborted === true;
406
- const localSourceFailure = localTransport
407
- ? backendRuns.some(({ run }) => localSourceBackendFailure(run.rawResult))
408
- || infrastructureRetryRuns.some(({ run }) => localSourceBackendFailure(run.rawResult))
409
- : false;
410
- const expectedBackendRuns = plannedTaskExecution ? executionPlan.work_items.length : normalized.attempts;
411
- const backendsSucceeded = backendRuns.every(({ run }) => run.backend.process_exit_code === 0 && run.rawResult !== null)
412
- && (plannedTaskExecution ? progress.trials.length === plannedTrials : backendRuns.length === expectedBackendRuns);
413
- const invalidTrials = localTaskIds === null
414
- ? trialRefs.filter((trial) => trial.observation_status !== "valid").map((trial) => ({ task_id: trial.task_id, attempt: trial.attempt }))
415
- : invalidTrialSlots(localTaskIds, normalized.attempts, progress);
416
- const infrastructureFailures = infrastructureFailureTrials(trialRefs);
417
- const infrastructureFailureSlots = infrastructureFailures.map((trial) => ({ task_id: trial.task_id, attempt: trial.attempt }));
418
- const verifierRetriesExhausted = normalized.infrastructure_retries > 0 && infrastructureFailures.some((trial) => trial.invalid_reason === "verifier_infrastructure_failure");
419
- const infrastructureErrorCode = verifierRetriesExhausted || (normalized.infrastructure_retries > 0 && infrastructureRetryRuns.length > 0)
420
- ? "eval_infrastructure_retries_exhausted"
421
- : "eval_has_infrastructure_failures";
422
- const succeeded = !cancelled && !localSourceFailure && backendsSucceeded && invalidTrials.length === 0;
423
- const singleBackend = backendRuns.length === 1 ? backendRuns[0].run : undefined;
424
- result = {
425
- schema_version: SCHEMA_VERSION,
426
- eval_id: evalId,
427
- status: cancelled ? "cancelled" : succeeded ? "succeeded" : "failed",
428
- exit_code: cancelled ? 9 : succeeded ? 0 : 13,
429
- ...(singleBackend ? { backend: singleBackend.backend, backend_summary: singleBackend.summary } : {}),
430
- ...(plannedTaskExecution ? {
431
- backend_work_items: backendRuns.map(({ attempt, workId, tasks, leaseId, run }) => ({
432
- work_id: workId,
433
- lease_id: leaseId,
434
- attempt,
435
- tasks,
436
- backend: run.backend,
437
- backend_summary: run.summary,
438
- })),
439
- } : normalized.attempts > 1 ? {
440
- backend_runs: backendRuns.map(({ attempt, run }) => ({
441
- attempt,
442
- backend: run.backend,
443
- backend_summary: run.summary,
444
- })),
445
- } : {}),
446
- infrastructure_retry_policy: {
447
- max_retries: normalized.infrastructure_retries,
448
- backoff_ms: normalized.infrastructure_retry_backoff_ms,
449
- verifier_execution: "same_trial_verifier_only",
450
- candidate_rerun_on_verifier_failure: false,
451
- },
452
- ...(infrastructureRetryRuns.length > 0 ? {
453
- infrastructure_retry_runs: infrastructureRetryRuns.map(({ attempt, retry, tasks, triggers, refs, run, leaseId, workId }) => ({
454
- execution_kind: "physical-infrastructure-retry", ...(leaseId ? { lease_id: leaseId } : {}), ...(workId ? { work_id: workId } : {}), attempt,
455
- retry,
456
- tasks,
457
- trigger_trials: triggers,
458
- trials: refs,
459
- backend: run.backend,
460
- backend_summary: run.summary,
461
- })),
462
- } : {}),
463
- candidate: plan.candidate,
464
- dataset: normalized.dataset,
465
- benchmark_id: normalized.benchmark_id,
466
- benchmark_revision: normalized.benchmark_revision,
467
- generation: progress.generation,
468
- trials: trialRefs,
469
- summary: summarizeTrialRefs(trialRefs),
470
- ...preparedArtifactPlanFields(preparedAssignments),
471
- ...(localTransport ? { local_source_transport: transportSummary(localTransport) } : {}),
472
- ...(succeeded ? {} : {
473
- error: {
474
- code: cancelled
475
- ? "cancelled"
476
- : localSourceFailure
477
- ? "local_source_materialize_failed"
478
- : !backendsSucceeded
479
- ? "harbor_failed"
480
- : infrastructureFailures.length > 0
481
- ? infrastructureErrorCode
482
- : "eval_has_invalid_tasks",
483
- message: cancelled
484
- ? "eval was cancelled"
485
- : localSourceFailure
486
- ? "Harbor rejected the transported local Git source before candidate execution"
487
- : !backendsSucceeded
488
- ? `Harbor work items completed ${backendRuns.length}/${expectedBackendRuns}`
489
- : infrastructureFailures.length > 0
490
- ? `${infrastructureErrorCode === "eval_infrastructure_retries_exhausted" ? "infrastructure retries exhausted" : "verifier infrastructure failure"}: ${infrastructureFailureSlots.map(trial => `${trial.task_id}#${trial.attempt}`).join(", ")}`
491
- : `eval has invalid or missing trials: ${invalidTrials.map(trial => `${trial.task_id}#${trial.attempt}`).join(", ")}`,
492
- },
493
- }),
494
- started_at: startedAt.toISOString(),
495
- completed_at: new Date().toISOString(),
496
- };
437
+ result = buildCompletedEvalResult({
438
+ evalId, request: normalized, plannedTaskExecution, plannedTrials,
439
+ executionWorkItems: executionPlan.work_items.length, localTaskIds, backendRuns,
440
+ infrastructureRetryRuns, candidate: plan.candidate, progress,
441
+ ...(schedulerSummary ? { schedulerSummary } : {}), preparedAssignments,
442
+ ...(localTransport ? { localTransport } : {}), startedAt,
443
+ cancelled: signal?.aborted === true,
444
+ });
497
445
  }
498
446
  catch (error) {
499
447
  trialRefs = progress?.trials ?? trialRefs;
@@ -507,6 +455,7 @@ export async function runEval({ evalId = newEvalId(), request, root, env = proce
507
455
  code: signal?.aborted ? "cancelled" : typed ? error.code : "internal_error",
508
456
  message: safeDiagnosticMessage(error, credentialValuesFromEnv(normalized.pass_env, env)),
509
457
  },
458
+ failure_stage: failureStage,
510
459
  benchmark_id: normalized.benchmark_id,
511
460
  benchmark_revision: normalized.benchmark_revision,
512
461
  ...(progress === null ? {} : { generation: progress.generation }),