@tangle-network/agent-eval 0.144.11 → 0.144.13

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 (378) hide show
  1. package/CHANGELOG.md +11 -0
  2. package/dist/agent-profile-B9_GGsG8.d.ts +84 -0
  3. package/dist/agent-profile-B9_GGsG8.d.ts.map +1 -0
  4. package/dist/{agent-profile-cell-BOP-iA9Q.d.ts → agent-profile-cell-BkcRDikH.d.ts} +2 -2
  5. package/dist/{agent-profile-cell-BOP-iA9Q.d.ts.map → agent-profile-cell-BkcRDikH.d.ts.map} +1 -1
  6. package/dist/analyst/index.d.ts +134 -16
  7. package/dist/analyst/index.d.ts.map +1 -1
  8. package/dist/analyst/index.js +364 -10
  9. package/dist/analyst/index.js.map +1 -1
  10. package/dist/backend-integrity-CsVin_Wb.d.ts +280 -0
  11. package/dist/backend-integrity-CsVin_Wb.d.ts.map +1 -0
  12. package/dist/{baseline-C-GocmIW.js → baseline-BhPRQBVn.js} +3 -2
  13. package/dist/{baseline-C-GocmIW.js.map → baseline-BhPRQBVn.js.map} +1 -1
  14. package/dist/{benchmark-CWeqGl7x.js → benchmark-BhT16ep9.js} +2 -2
  15. package/dist/{benchmark-CWeqGl7x.js.map → benchmark-BhT16ep9.js.map} +1 -1
  16. package/dist/{agent-profile-DPi7IZg7.d.ts → benchmark-Ceoan7vk.d.ts} +4 -91
  17. package/dist/benchmark-Ceoan7vk.d.ts.map +1 -0
  18. package/dist/{benchmark-command-BteMFN62.js → benchmark-command-CY6cbhOB.js} +36 -79
  19. package/dist/benchmark-command-CY6cbhOB.js.map +1 -0
  20. package/dist/benchmarks/index.d.ts +244 -2
  21. package/dist/benchmarks/index.d.ts.map +1 -0
  22. package/dist/benchmarks/index.js +733 -1
  23. package/dist/benchmarks/index.js.map +1 -0
  24. package/dist/builder-eval/index.d.ts +23 -2
  25. package/dist/builder-eval/index.d.ts.map +1 -1
  26. package/dist/builder-eval/index.js +227 -3
  27. package/dist/builder-eval/index.js.map +1 -1
  28. package/dist/campaign/index.d.ts +10 -8
  29. package/dist/campaign/index.js +9 -6
  30. package/dist/{campaign-C2TTzQII.js → campaign-BYjBAypg.js} +21 -634
  31. package/dist/campaign-BYjBAypg.js.map +1 -0
  32. package/dist/{canonical-D011XM8r.js → canonical-D-XsTQ6_.js} +2 -2
  33. package/dist/{canonical-D011XM8r.js.map → canonical-D-XsTQ6_.js.map} +1 -1
  34. package/dist/{index-CQTZ-4XN.d.ts → capture-fetch-DDvpjVRU.d.ts} +3 -3
  35. package/dist/capture-fetch-DDvpjVRU.d.ts.map +1 -0
  36. package/dist/{default-registry-BmktKy8r.js → chat-client-Bp4Ebuuc.js} +1276 -1276
  37. package/dist/chat-client-Bp4Ebuuc.js.map +1 -0
  38. package/dist/cli.js +2 -2
  39. package/dist/{client-C9gzZE59.d.ts → client-KUbGulm_.d.ts} +4 -4
  40. package/dist/{client-C9gzZE59.d.ts.map → client-KUbGulm_.d.ts.map} +1 -1
  41. package/dist/contract/index.d.ts +13 -390
  42. package/dist/contract/index.d.ts.map +1 -1
  43. package/dist/contract/index.js +18 -542
  44. package/dist/contract/index.js.map +1 -1
  45. package/dist/{cost-ledger-DMFxsLKr.js → cost-ledger-BSe92yAV.js} +3 -3
  46. package/dist/{cost-ledger-DMFxsLKr.js.map → cost-ledger-BSe92yAV.js.map} +1 -1
  47. package/dist/{cost-ledger-Bv_e8XHY.d.ts → cost-ledger-DbQdN3nO.d.ts} +2 -2
  48. package/dist/{cost-ledger-Bv_e8XHY.d.ts.map → cost-ledger-DbQdN3nO.d.ts.map} +1 -1
  49. package/dist/{counterfactual-CxmxAONP.d.ts → counterfactual--bpysZF0.d.ts} +2 -14
  50. package/dist/{counterfactual-CxmxAONP.d.ts.map → counterfactual--bpysZF0.d.ts.map} +1 -1
  51. package/dist/{counterfactual-CWPTrMH7.js → counterfactual-lDfCx0Uz.js} +3 -31
  52. package/dist/{counterfactual-CWPTrMH7.js.map → counterfactual-lDfCx0Uz.js.map} +1 -1
  53. package/dist/{dataset-C8xaLXdY.d.ts → dataset-CJjKqQfA.d.ts} +2 -8
  54. package/dist/dataset-CJjKqQfA.d.ts.map +1 -0
  55. package/dist/{default-registry-Bf8Woqmq.d.ts → default-registry-Di6HP6pG.d.ts} +7 -12
  56. package/dist/default-registry-Di6HP6pG.d.ts.map +1 -0
  57. package/dist/{analyze-runs-C30yljDJ.js → define-agent-eval-iqjT--ZZ.js} +542 -130
  58. package/dist/define-agent-eval-iqjT--ZZ.js.map +1 -0
  59. package/dist/define-agent-eval-sH24zBfM.d.ts +388 -0
  60. package/dist/define-agent-eval-sH24zBfM.d.ts.map +1 -0
  61. package/dist/descriptive-B5MwKfbf.js +144 -0
  62. package/dist/descriptive-B5MwKfbf.js.map +1 -0
  63. package/dist/{dspy-rlm-engine-DbTk4JdR.js → dspy-rlm-engine-B_qhSc21.js} +4 -4
  64. package/dist/{dspy-rlm-engine-DbTk4JdR.js.map → dspy-rlm-engine-B_qhSc21.js.map} +1 -1
  65. package/dist/effect-sizes-DiH8MGOH.js +82 -0
  66. package/dist/effect-sizes-DiH8MGOH.js.map +1 -0
  67. package/dist/{engine-3hL-XqwJ.d.ts → engine-otFpE2gF.d.ts} +10 -38
  68. package/dist/engine-otFpE2gF.d.ts.map +1 -0
  69. package/dist/{errors-CKPfb2aH.d.ts → errors-DEE6u6ot.d.ts} +2 -14
  70. package/dist/{errors-CKPfb2aH.d.ts.map → errors-DEE6u6ot.d.ts.map} +1 -1
  71. package/dist/{errors-D-LKuDhb.js → errors-Dngq5h35.js} +2 -8
  72. package/dist/{errors-D-LKuDhb.js.map → errors-Dngq5h35.js.map} +1 -1
  73. package/dist/{eval-campaign-B_7wcnav.js → eval-campaign-UB-usSQ2.js} +6 -6
  74. package/dist/{eval-campaign-B_7wcnav.js.map → eval-campaign-UB-usSQ2.js.map} +1 -1
  75. package/dist/{exact-types-DSFFpLLI.d.ts → exact-types-BH1twmAJ.d.ts} +2 -2
  76. package/dist/{exact-types-DSFFpLLI.d.ts.map → exact-types-BH1twmAJ.d.ts.map} +1 -1
  77. package/dist/experiment/index.d.ts +9 -6
  78. package/dist/experiment/index.d.ts.map +1 -1
  79. package/dist/experiment/index.js +11 -7
  80. package/dist/experiment/index.js.map +1 -1
  81. package/dist/experiment-tracker-C29gXM4B.js +269 -0
  82. package/dist/experiment-tracker-C29gXM4B.js.map +1 -0
  83. package/dist/{experiment-tracker-IMntXr6J.d.ts → experiment-tracker-DWHZBAYL.d.ts} +77 -121
  84. package/dist/experiment-tracker-DWHZBAYL.d.ts.map +1 -0
  85. package/dist/{external-optimizer-contracts-CZuJNcT5.d.ts → external-optimizer-contracts-lixrOZdX.d.ts} +3 -3
  86. package/dist/{external-optimizer-contracts-CZuJNcT5.d.ts.map → external-optimizer-contracts-lixrOZdX.d.ts.map} +1 -1
  87. package/dist/external-optimizer-process-BTiNB-RH.js +301 -0
  88. package/dist/external-optimizer-process-BTiNB-RH.js.map +1 -0
  89. package/dist/{single-run-lock-DFWHEB09.js → external-optimizer-subprocess-DrJ9hR8u.js} +144 -438
  90. package/dist/external-optimizer-subprocess-DrJ9hR8u.js.map +1 -0
  91. package/dist/extract-usage-BrQ8mCLX.js +155 -0
  92. package/dist/extract-usage-BrQ8mCLX.js.map +1 -0
  93. package/dist/{failure-cluster-CqcvCcdR.d.ts → failure-cluster-BLURuWG4.d.ts} +2 -3
  94. package/dist/failure-cluster-BLURuWG4.d.ts.map +1 -0
  95. package/dist/{feedback-trajectory-CacHpxsp.d.ts → feedback-trajectory-juOozjAc.d.ts} +5 -36
  96. package/dist/feedback-trajectory-juOozjAc.d.ts.map +1 -0
  97. package/dist/fuzz.d.ts +2 -2
  98. package/dist/fuzz.js +2 -2
  99. package/dist/hosted/index.d.ts +3 -3
  100. package/dist/{index-CvSN3IG1.d.ts → index-B8Ui1mr1.d.ts} +2 -2
  101. package/dist/{index-CvSN3IG1.d.ts.map → index-B8Ui1mr1.d.ts.map} +1 -1
  102. package/dist/{skill-usage-CO9OLRBx.d.ts → index-BWDrSVfw.d.ts} +11 -136
  103. package/dist/index-BWDrSVfw.d.ts.map +1 -0
  104. package/dist/index-Ba3YrbAL.d.ts +1 -0
  105. package/dist/{index-BnEuDAK2.d.ts → index-COQYtuRF.d.ts} +3 -3
  106. package/dist/{index-BnEuDAK2.d.ts.map → index-COQYtuRF.d.ts.map} +1 -1
  107. package/dist/{index-C3ssXVLv.d.ts → index-CvjYbU0D.d.ts} +2 -2
  108. package/dist/{index-C3ssXVLv.d.ts.map → index-CvjYbU0D.d.ts.map} +1 -1
  109. package/dist/{index-DPPGNJ_R.d.ts → index-DSmEylT9.d.ts} +17 -115
  110. package/dist/index-DSmEylT9.d.ts.map +1 -0
  111. package/dist/index.d.ts +2397 -5308
  112. package/dist/index.d.ts.map +1 -1
  113. package/dist/index.js +5914 -10496
  114. package/dist/index.js.map +1 -1
  115. package/dist/{insight-report-CgX_s0Ez.d.ts → insight-report-CRi-Ufrj.d.ts} +4 -4
  116. package/dist/{insight-report-CgX_s0Ez.d.ts.map → insight-report-CRi-Ufrj.d.ts.map} +1 -1
  117. package/dist/{integrity-MLzHOfV9.js → integrity-Cy9WHAtb.js} +2 -2
  118. package/dist/{integrity-MLzHOfV9.js.map → integrity-Cy9WHAtb.js.map} +1 -1
  119. package/dist/{integrity-DY6tIbl0.js → integrity-DysDBWDu.js} +2 -2
  120. package/dist/{integrity-DY6tIbl0.js.map → integrity-DysDBWDu.js.map} +1 -1
  121. package/dist/{integrity-BuqEKu-x.d.ts → integrity-OrcI9Nau.d.ts} +3 -3
  122. package/dist/{integrity-BuqEKu-x.d.ts.map → integrity-OrcI9Nau.d.ts.map} +1 -1
  123. package/dist/internal-BDHPCnjk.js +230 -0
  124. package/dist/internal-BDHPCnjk.js.map +1 -0
  125. package/dist/judge-calibration-C5CbMYce.d.ts +117 -0
  126. package/dist/judge-calibration-C5CbMYce.d.ts.map +1 -0
  127. package/dist/judge-calibration-DZkWrm5H.js +317 -0
  128. package/dist/judge-calibration-DZkWrm5H.js.map +1 -0
  129. package/dist/{kind-factory-B8-r8-y8.js → kind-factory-CPmSd58s.js} +208 -208
  130. package/dist/{kind-factory-B8-r8-y8.js.map → kind-factory-CPmSd58s.js.map} +1 -1
  131. package/dist/ledger-core/index.d.ts +1 -1
  132. package/dist/ledger-core/index.js +2 -2
  133. package/dist/{ledger-core-DXZIqu17.js → ledger-core-BmZt19oQ.js} +110 -110
  134. package/dist/{ledger-core-DXZIqu17.js.map → ledger-core-BmZt19oQ.js.map} +1 -1
  135. package/dist/{llm-client-DzvMUsS_.js → llm-client-d0-2TT1g.js} +4 -64
  136. package/dist/{llm-client-DzvMUsS_.js.map → llm-client-d0-2TT1g.js.map} +1 -1
  137. package/dist/{skillopt-optimization-method-CQdVeM8k.js → llm-judge-dZ8P6nGI.js} +3349 -5401
  138. package/dist/llm-judge-dZ8P6nGI.js.map +1 -0
  139. package/dist/matrix/index.d.ts +2 -2
  140. package/dist/meta-eval/index.d.ts +3 -3
  141. package/dist/meta-eval/index.js +3 -3
  142. package/dist/{metrics-C9YY1OcL.js → metrics-Cl0L1KUy.js} +2 -108
  143. package/dist/{metrics-C9YY1OcL.js.map → metrics-Cl0L1KUy.js.map} +1 -1
  144. package/dist/{mint-B2O60ACG.js → mint-Dj9Ww_3I.js} +4 -4
  145. package/dist/{mint-B2O60ACG.js.map → mint-Dj9Ww_3I.js.map} +1 -1
  146. package/dist/{multi-layer-verifier-DnAqwl0h.d.ts → multi-layer-verifier-DIguZc8Z.d.ts} +3 -3
  147. package/dist/{multi-layer-verifier-DnAqwl0h.d.ts.map → multi-layer-verifier-DIguZc8Z.d.ts.map} +1 -1
  148. package/dist/multiplicity-DIWHvysC.d.ts +43 -0
  149. package/dist/multiplicity-DIWHvysC.d.ts.map +1 -0
  150. package/dist/multishot/index.d.ts +3 -3
  151. package/dist/multishot/index.js +1 -1
  152. package/dist/openapi.json +1 -1
  153. package/dist/{opencode-sqlite-8r6WUfHc.js → opencode-sqlite-DJWAXLms.js} +2 -2
  154. package/dist/{opencode-sqlite-8r6WUfHc.js.map → opencode-sqlite-DJWAXLms.js.map} +1 -1
  155. package/dist/package-version-D7lQHt_-.js +34 -0
  156. package/dist/package-version-D7lQHt_-.js.map +1 -0
  157. package/dist/paired-arms-D-XRF_fy.js +1045 -0
  158. package/dist/paired-arms-D-XRF_fy.js.map +1 -0
  159. package/dist/paired-promotion-decision-CGzg0cI_.d.ts +251 -0
  160. package/dist/paired-promotion-decision-CGzg0cI_.d.ts.map +1 -0
  161. package/dist/paired-tests-BHIhYVdu.js +213 -0
  162. package/dist/paired-tests-BHIhYVdu.js.map +1 -0
  163. package/dist/pareto-BqNW3LJR.d.ts +117 -0
  164. package/dist/pareto-BqNW3LJR.d.ts.map +1 -0
  165. package/dist/pipelines/index.d.ts +3 -64
  166. package/dist/pipelines/index.d.ts.map +1 -1
  167. package/dist/pipelines/index.js +4 -284
  168. package/dist/pipelines/index.js.map +1 -1
  169. package/dist/power-and-mde-CHIrXJll.js +195 -0
  170. package/dist/power-and-mde-CHIrXJll.js.map +1 -0
  171. package/dist/{promotion-policy-CrLrmys8.js → power-preflight-DEw-uC7q.js} +4 -184
  172. package/dist/power-preflight-DEw-uC7q.js.map +1 -0
  173. package/dist/pre-registration-CZwSFQS4.d.ts +577 -0
  174. package/dist/pre-registration-CZwSFQS4.d.ts.map +1 -0
  175. package/dist/{prime-protocol-BfSalTfR.js → prime-protocol-6tZTVsWm.js} +72 -21
  176. package/dist/prime-protocol-6tZTVsWm.js.map +1 -0
  177. package/dist/produced-state-DU79a81m.js +586 -0
  178. package/dist/produced-state-DU79a81m.js.map +1 -0
  179. package/dist/profile-cell.d.ts +1 -1
  180. package/dist/profile-cell.js +1 -1
  181. package/dist/promotion-policy-D0nPhkSy.d.ts +134 -0
  182. package/dist/promotion-policy-D0nPhkSy.d.ts.map +1 -0
  183. package/dist/promotion-policy-xzA40Evo.js +186 -0
  184. package/dist/promotion-policy-xzA40Evo.js.map +1 -0
  185. package/dist/{skillopt-optimization-method-USDKhxSA.d.ts → provenance-CCdxgLDT.d.ts} +57 -478
  186. package/dist/provenance-CCdxgLDT.d.ts.map +1 -0
  187. package/dist/registry-oJeeI4-a.d.ts +178 -0
  188. package/dist/registry-oJeeI4-a.d.ts.map +1 -0
  189. package/dist/{release-report-DA2BCu5a.d.ts → release-confidence-BFRE5WSp.d.ts} +180 -112
  190. package/dist/release-confidence-BFRE5WSp.d.ts.map +1 -0
  191. package/dist/{release-report-BUYmoKo2.js → release-confidence-CxDuiAev.js} +125 -217
  192. package/dist/release-confidence-CxDuiAev.js.map +1 -0
  193. package/dist/reporting.d.ts +6 -5
  194. package/dist/reporting.js +7 -5
  195. package/dist/{researcher-BchpD55R.d.ts → researcher-DaN4GST-.d.ts} +7 -40
  196. package/dist/{researcher-BchpD55R.d.ts.map → researcher-DaN4GST-.d.ts.map} +1 -1
  197. package/dist/{reward-hacking-BDToousL.js → reward-hacking-DNgjilrV.js} +3 -3
  198. package/dist/reward-hacking-DNgjilrV.js.map +1 -0
  199. package/dist/{reward-hacking-BI0OMAlo.d.ts → reward-hacking-DSSTuI9r.d.ts} +5 -5
  200. package/dist/{reward-hacking-BI0OMAlo.d.ts.map → reward-hacking-DSSTuI9r.d.ts.map} +1 -1
  201. package/dist/rl.d.ts +7 -7
  202. package/dist/rl.js +11 -10
  203. package/dist/rl.js.map +1 -1
  204. package/dist/rollout/index.d.ts +2 -2
  205. package/dist/rollout/index.js +4 -4
  206. package/dist/{rollout-DRrksrcV.js → rollout-BWtw0I_6.js} +3 -3
  207. package/dist/{rollout-DRrksrcV.js.map → rollout-BWtw0I_6.js.map} +1 -1
  208. package/dist/{rubric-predictive-validity-DvkjPYCe.d.ts → rubric-predictive-validity-BgxtKe4G.d.ts} +2 -2
  209. package/dist/{rubric-predictive-validity-DvkjPYCe.d.ts.map → rubric-predictive-validity-BgxtKe4G.d.ts.map} +1 -1
  210. package/dist/{rubric-predictive-validity-BRR632r1.js → rubric-predictive-validity-Cwwyd7ah.js} +2 -2
  211. package/dist/{rubric-predictive-validity-BRR632r1.js.map → rubric-predictive-validity-Cwwyd7ah.js.map} +1 -1
  212. package/dist/{run-record-BmSPWXJR.js → run-record-BvHPVS-i.js} +3 -3
  213. package/dist/{run-record-BmSPWXJR.js.map → run-record-BvHPVS-i.js.map} +1 -1
  214. package/dist/{run-record-CF4Dwpxr.d.ts → run-record-CKiihE6f.d.ts} +4 -4
  215. package/dist/{run-record-CF4Dwpxr.d.ts.map → run-record-CKiihE6f.d.ts.map} +1 -1
  216. package/dist/{proposal-findings-2GIUo1et.js → run-score-lDzV0X8j.js} +31 -31
  217. package/dist/run-score-lDzV0X8j.js.map +1 -0
  218. package/dist/sandbox-harness-BlSOu4LX.d.ts +70 -0
  219. package/dist/sandbox-harness-BlSOu4LX.d.ts.map +1 -0
  220. package/dist/{schema-Cef2cFmb2.d.ts → schema-Cef2cFmb.d.ts} +1 -1
  221. package/dist/schema-Cef2cFmb.d.ts.map +1 -0
  222. package/dist/semantic-concept-judge-BiJxScqe.js +406 -0
  223. package/dist/semantic-concept-judge-BiJxScqe.js.map +1 -0
  224. package/dist/{sequential-D-BLJBKU.js → sequential-C458DXNf.js} +4 -3
  225. package/dist/{sequential-D-BLJBKU.js.map → sequential-C458DXNf.js.map} +1 -1
  226. package/dist/sequential-eprocess-CbUt2htw.js +83 -0
  227. package/dist/sequential-eprocess-CbUt2htw.js.map +1 -0
  228. package/dist/series-convergence-D1cL1f-4.d.ts +129 -0
  229. package/dist/series-convergence-D1cL1f-4.d.ts.map +1 -0
  230. package/dist/{server-iu0ede49.js → server-ulsOdrTI.js} +5 -21
  231. package/dist/server-ulsOdrTI.js.map +1 -0
  232. package/dist/skillopt-optimization-method-C5cotF4E.d.ts +427 -0
  233. package/dist/skillopt-optimization-method-C5cotF4E.d.ts.map +1 -0
  234. package/dist/skillopt-optimization-method-jjdnc3YK.js +1999 -0
  235. package/dist/skillopt-optimization-method-jjdnc3YK.js.map +1 -0
  236. package/dist/{statistical-heldout-TQ-4CYiN.d.ts → statistical-heldout-DoFF5KJX.d.ts} +5 -278
  237. package/dist/statistical-heldout-DoFF5KJX.d.ts.map +1 -0
  238. package/dist/{store-otlp-Dw8PPIlL.js → store-otlp-CsptLYpN.js} +3 -3
  239. package/dist/{store-otlp-Dw8PPIlL.js.map → store-otlp-CsptLYpN.js.map} +1 -1
  240. package/dist/store-tool-spans-Cq9mFd-q.js +667 -0
  241. package/dist/store-tool-spans-Cq9mFd-q.js.map +1 -0
  242. package/dist/store-tool-spans-DbDLOBOb.d.ts +342 -0
  243. package/dist/store-tool-spans-DbDLOBOb.d.ts.map +1 -0
  244. package/dist/student-t-CvBq2mve.js +38 -0
  245. package/dist/student-t-CvBq2mve.js.map +1 -0
  246. package/dist/{summary-report-Lf-5I7xh.js → summary-report-Blysd6Z2.js} +6 -3
  247. package/dist/{summary-report-Lf-5I7xh.js.map → summary-report-Blysd6Z2.js.map} +1 -1
  248. package/dist/{summary-report-BNR7DWTj.d.ts → summary-report-CaL-Hnxt.d.ts} +5 -5
  249. package/dist/{summary-report-BNR7DWTj.d.ts.map → summary-report-CaL-Hnxt.d.ts.map} +1 -1
  250. package/dist/supervisor-run/index.d.ts +391 -3
  251. package/dist/supervisor-run/index.d.ts.map +1 -0
  252. package/dist/supervisor-run/index.js +1689 -2
  253. package/dist/{supervisor-run-D_sokXcO.js.map → supervisor-run/index.js.map} +1 -1
  254. package/dist/{extract-usage-CdZdoj1s.js → task-failure-attributes-CpQ4y5RD.js} +5 -157
  255. package/dist/task-failure-attributes-CpQ4y5RD.js.map +1 -0
  256. package/dist/{tool-groups-CZPGGlHf.d.ts → tool-groups-Cteb03Ps.d.ts} +3 -3
  257. package/dist/tool-groups-Cteb03Ps.d.ts.map +1 -0
  258. package/dist/tool-waste-BDdBZG1F.js +803 -0
  259. package/dist/tool-waste-BDdBZG1F.js.map +1 -0
  260. package/dist/tool-waste-DjRDEsuI.d.ts +128 -0
  261. package/dist/tool-waste-DjRDEsuI.d.ts.map +1 -0
  262. package/dist/trace-repair/index.d.ts +14 -5
  263. package/dist/trace-repair/index.d.ts.map +1 -1
  264. package/dist/trace-repair/index.js +35 -7
  265. package/dist/trace-repair/index.js.map +1 -1
  266. package/dist/traces.d.ts +406 -7
  267. package/dist/traces.d.ts.map +1 -0
  268. package/dist/traces.js +1011 -10
  269. package/dist/traces.js.map +1 -0
  270. package/dist/trajectory-replay/index.d.ts +16 -3
  271. package/dist/trajectory-replay/index.d.ts.map +1 -1
  272. package/dist/trajectory-replay/index.js +52 -5
  273. package/dist/trajectory-replay/index.js.map +1 -1
  274. package/dist/types-BEPZc6eo.d.ts +93 -0
  275. package/dist/types-BEPZc6eo.d.ts.map +1 -0
  276. package/dist/{usage-receipt-t7vAzCRQ.js → types-BI4fT3HN.js} +67 -67
  277. package/dist/types-BI4fT3HN.js.map +1 -0
  278. package/dist/{types-BnjdJ70P.d.ts → types-BZ59Ahr8.d.ts} +5 -5
  279. package/dist/{types-BnjdJ70P.d.ts.map → types-BZ59Ahr8.d.ts.map} +1 -1
  280. package/dist/{types-D216SgwM.d.ts → types-Cx3YUh2r.d.ts} +4 -241
  281. package/dist/types-Cx3YUh2r.d.ts.map +1 -0
  282. package/dist/{types-BvDKaULh.d.ts → types-DLQx4mKU.d.ts} +4 -4
  283. package/dist/{types-BvDKaULh.d.ts.map → types-DLQx4mKU.d.ts.map} +1 -1
  284. package/dist/{types-D4mog56g.d.ts → types-yLK8gXE9.d.ts} +2 -2
  285. package/dist/{types-D4mog56g.d.ts.map → types-yLK8gXE9.d.ts.map} +1 -1
  286. package/dist/verdict-BndeTAh_.js +61 -0
  287. package/dist/verdict-BndeTAh_.js.map +1 -0
  288. package/dist/verdict-E4eRNf7-.d.ts +392 -0
  289. package/dist/verdict-E4eRNf7-.d.ts.map +1 -0
  290. package/dist/{verdict-cache-BCcOh0kF.js → verdict-cache-mZf5FEiY.js} +3 -55
  291. package/dist/{verdict-cache-BCcOh0kF.js.map → verdict-cache-mZf5FEiY.js.map} +1 -1
  292. package/dist/wire/index.d.ts +3 -3
  293. package/dist/wire/index.d.ts.map +1 -1
  294. package/dist/wire/index.js +1 -1
  295. package/docs/charter.md +3 -3
  296. package/docs/control-runtime.md +3 -42
  297. package/docs/experiment.md +0 -1
  298. package/docs/feature-guide.md +2 -2
  299. package/docs/trace-repair-grader.md +1 -0
  300. package/docs/trajectory-replay.md +1 -0
  301. package/docs/verdicts.md +43 -0
  302. package/docs/verification-strategies.md +3 -2
  303. package/package.json +6 -11
  304. package/dist/agent-profile-DPi7IZg7.d.ts.map +0 -1
  305. package/dist/analyze-runs-C30yljDJ.js.map +0 -1
  306. package/dist/baseline-CavEbRyH.d.ts +0 -136
  307. package/dist/baseline-CavEbRyH.d.ts.map +0 -1
  308. package/dist/benchmark-command-BteMFN62.js.map +0 -1
  309. package/dist/benchmarks-Dzs8CKb1.js +0 -755
  310. package/dist/benchmarks-Dzs8CKb1.js.map +0 -1
  311. package/dist/campaign-C2TTzQII.js.map +0 -1
  312. package/dist/completion-verifier-DJA5BhPb.d.ts +0 -414
  313. package/dist/completion-verifier-DJA5BhPb.d.ts.map +0 -1
  314. package/dist/control.d.ts +0 -3
  315. package/dist/control.js +0 -2
  316. package/dist/dataset-C8xaLXdY.d.ts.map +0 -1
  317. package/dist/default-registry-Bf8Woqmq.d.ts.map +0 -1
  318. package/dist/default-registry-BmktKy8r.js.map +0 -1
  319. package/dist/engine-3hL-XqwJ.d.ts.map +0 -1
  320. package/dist/experiment-tracker-CnRICnMl.js +0 -500
  321. package/dist/experiment-tracker-CnRICnMl.js.map +0 -1
  322. package/dist/experiment-tracker-IMntXr6J.d.ts.map +0 -1
  323. package/dist/extract-usage-CdZdoj1s.js.map +0 -1
  324. package/dist/failure-cluster-CqcvCcdR.d.ts.map +0 -1
  325. package/dist/feedback-trajectory-CacHpxsp.d.ts.map +0 -1
  326. package/dist/index-BZ3-y4YL.d.ts +0 -391
  327. package/dist/index-BZ3-y4YL.d.ts.map +0 -1
  328. package/dist/index-CQTZ-4XN.d.ts.map +0 -1
  329. package/dist/index-DPPGNJ_R.d.ts.map +0 -1
  330. package/dist/index-YE4KdKbO2.d.ts +0 -335
  331. package/dist/index-YE4KdKbO2.d.ts.map +0 -1
  332. package/dist/paired-arms-iZ08VFMN.js +0 -260
  333. package/dist/paired-arms-iZ08VFMN.js.map +0 -1
  334. package/dist/paired-promotion-decision-B6zJ3gYM.d.ts +0 -114
  335. package/dist/paired-promotion-decision-B6zJ3gYM.d.ts.map +0 -1
  336. package/dist/prime-protocol-BfSalTfR.js.map +0 -1
  337. package/dist/promotion-policy-Ckjhzg_4.d.ts +0 -289
  338. package/dist/promotion-policy-Ckjhzg_4.d.ts.map +0 -1
  339. package/dist/promotion-policy-CrLrmys8.js.map +0 -1
  340. package/dist/proposal-findings-2GIUo1et.js.map +0 -1
  341. package/dist/propose-review-control-dSNPjFUH.js +0 -1458
  342. package/dist/propose-review-control-dSNPjFUH.js.map +0 -1
  343. package/dist/release-report-BUYmoKo2.js.map +0 -1
  344. package/dist/release-report-DA2BCu5a.d.ts.map +0 -1
  345. package/dist/replay-CohS93nE.js +0 -1859
  346. package/dist/replay-CohS93nE.js.map +0 -1
  347. package/dist/replay-DbhZ4Ked.d.ts +0 -834
  348. package/dist/replay-DbhZ4Ked.d.ts.map +0 -1
  349. package/dist/reward-hacking-BDToousL.js.map +0 -1
  350. package/dist/run-evidence-Dhi3C81V.d.ts +0 -225
  351. package/dist/run-evidence-Dhi3C81V.d.ts.map +0 -1
  352. package/dist/schema-Cef2cFmb2.d.ts.map +0 -1
  353. package/dist/semantic-concept-judge-D1z-KepS.js +0 -767
  354. package/dist/semantic-concept-judge-D1z-KepS.js.map +0 -1
  355. package/dist/series-convergence-ofsqPWhs.d.ts +0 -35
  356. package/dist/series-convergence-ofsqPWhs.d.ts.map +0 -1
  357. package/dist/server-iu0ede49.js.map +0 -1
  358. package/dist/single-run-lock-DFWHEB09.js.map +0 -1
  359. package/dist/skill-usage-CO9OLRBx.d.ts.map +0 -1
  360. package/dist/skillopt-optimization-method-CQdVeM8k.js.map +0 -1
  361. package/dist/skillopt-optimization-method-USDKhxSA.d.ts.map +0 -1
  362. package/dist/statistical-heldout-TQ-4CYiN.d.ts.map +0 -1
  363. package/dist/statistics-ByxzSiOM.js +0 -2212
  364. package/dist/statistics-ByxzSiOM.js.map +0 -1
  365. package/dist/statistics-D6Uebe_4.d.ts +0 -968
  366. package/dist/statistics-D6Uebe_4.d.ts.map +0 -1
  367. package/dist/supervisor-run-D_sokXcO.js +0 -1690
  368. package/dist/test-graded-scenario-D1TaI2va.d.ts +0 -141
  369. package/dist/test-graded-scenario-D1TaI2va.d.ts.map +0 -1
  370. package/dist/test-graded-scenario-JHcKQNpq.js +0 -318
  371. package/dist/test-graded-scenario-JHcKQNpq.js.map +0 -1
  372. package/dist/tool-groups-CZPGGlHf.d.ts.map +0 -1
  373. package/dist/tool-use-metrics-DEGMKycK.js +0 -370
  374. package/dist/tool-use-metrics-DEGMKycK.js.map +0 -1
  375. package/dist/types-D216SgwM.d.ts.map +0 -1
  376. package/dist/usage-receipt-t7vAzCRQ.js.map +0 -1
  377. package/dist/verdict-DExhxfgR.d.ts +0 -201
  378. package/dist/verdict-DExhxfgR.d.ts.map +0 -1
@@ -0,0 +1,301 @@
1
+ import { n as contentHash } from "./verdict-cache-mZf5FEiY.js";
2
+ import { m as tryAcquireAtomicFileLock, p as probeAtomicFileLock } from "./ledger-core-BmZt19oQ.js";
3
+ import { _ as resolveExternalOptimizerCallbackLimits, a as closeServer, h as isRecord, m as isExternalTextCandidate, o as listenLocal, s as sendJson } from "./external-optimizer-subprocess-DrJ9hR8u.js";
4
+ import { createServer } from "node:http";
5
+ //#region src/campaign/single-run-lock.ts
6
+ /**
7
+ * Single-run lock for evaluations that share one mutable environment.
8
+ *
9
+ * Two concurrent runs against a shared stateful gym silently corrupt each
10
+ * other: each resets/mutates environment state mid-cell of the other, and
11
+ * every score from both becomes garbage that LOOKS like worker variance
12
+ * (agent-lab R357 burned hours on flip-flopping scores before tracing them
13
+ * to exactly this). The fix is a pid lockfile: refuse to start while a live
14
+ * holder exists, reclaim stale locks whose pid is gone, release only if the
15
+ * lock is still ours.
16
+ *
17
+ * `alsoCheck` exists because independent runners can guard the same shared
18
+ * resource with differently named lockfiles; a runner must respect all of
19
+ * them even though it writes only its own.
20
+ */
21
+ function assertAvailable(path) {
22
+ const unavailable = probeAtomicFileLock({ lockPath: path });
23
+ if (unavailable) throw unavailableError(path, unavailable);
24
+ }
25
+ function unavailableError(path, unavailable) {
26
+ if (unavailable.reason === "recovery") return /* @__PURE__ */ new Error(`single-run lock recovery is already in progress (${path}); refusing a concurrent run`);
27
+ return /* @__PURE__ */ new Error(`single-run lock held by live pid ${unavailable.holder.pid} (${path}); refusing a concurrent run on the shared resource`);
28
+ }
29
+ /**
30
+ * Acquire the lock or throw naming the live holder. A stale lock (holder pid
31
+ * no longer running) is reclaimed by one contender. An interrupted reclaim
32
+ * leaves a marker that fails closed instead of admitting overlapping runs.
33
+ */
34
+ function acquireSingleRunLock(opts) {
35
+ const pid = opts.pid ?? process.pid;
36
+ if (!Number.isSafeInteger(pid) || pid <= 0) throw new Error("single-run lock pid must be a positive integer");
37
+ for (const path of opts.alsoCheck ?? []) assertAvailable(path);
38
+ const acquisition = tryAcquireAtomicFileLock({
39
+ lockPath: opts.lockPath,
40
+ pid
41
+ });
42
+ if (!acquisition.acquired) throw unavailableError(opts.lockPath, acquisition);
43
+ const releaseOnExit = opts.releaseOnExit ?? true;
44
+ let released = false;
45
+ const release = () => {
46
+ if (released) return;
47
+ released = true;
48
+ if (releaseOnExit) process.off("exit", release);
49
+ try {
50
+ acquisition.lock.release();
51
+ } catch {}
52
+ };
53
+ try {
54
+ for (const path of opts.alsoCheck ?? []) assertAvailable(path);
55
+ } catch (error) {
56
+ release();
57
+ throw error;
58
+ }
59
+ if (releaseOnExit) process.on("exit", release);
60
+ return { release };
61
+ }
62
+ //#endregion
63
+ //#region src/campaign/external-optimizer-callback.ts
64
+ async function startExternalOptimizerCallback(args) {
65
+ assertCallbackConfig(args);
66
+ const limits = resolveExternalOptimizerCallbackLimits(args.limits);
67
+ args.signal?.throwIfAborted();
68
+ let evaluations = 0;
69
+ let accepting = true;
70
+ let closePromise;
71
+ const activeControllers = /* @__PURE__ */ new Set();
72
+ const activeHandlers = /* @__PURE__ */ new Set();
73
+ const proposedCandidateHashes = /* @__PURE__ */ new Set();
74
+ let observationSequence = 0;
75
+ const server = createServer((request, response) => {
76
+ if (!accepting) {
77
+ sendJsonIfOpen(response, 503, { error: "external optimizer callback is closing" });
78
+ return;
79
+ }
80
+ const controller = new AbortController();
81
+ const abortRequest = () => {
82
+ request.destroy();
83
+ response.destroy();
84
+ };
85
+ activeControllers.add(controller);
86
+ controller.signal.addEventListener("abort", abortRequest, { once: true });
87
+ let handler;
88
+ handler = handleCallback(request, response, controller.signal, args, limits, () => {
89
+ const accepted = args.acceptEvaluation ? args.acceptEvaluation() : evaluations + 1;
90
+ if (accepted === void 0) return void 0;
91
+ if (!Number.isSafeInteger(accepted) || accepted <= 0) throw new Error("external optimizer callback: invalid accepted evaluation count");
92
+ if (accepted > args.maxEvaluations) return void 0;
93
+ evaluations += 1;
94
+ return accepted;
95
+ }, (observation) => {
96
+ if (!args.observe) return;
97
+ args.observe({
98
+ ...observation,
99
+ sequence: ++observationSequence
100
+ });
101
+ }, proposedCandidateHashes).finally(() => {
102
+ controller.signal.removeEventListener("abort", abortRequest);
103
+ activeControllers.delete(controller);
104
+ activeHandlers.delete(handler);
105
+ });
106
+ activeHandlers.add(handler);
107
+ handler.catch(() => void 0);
108
+ });
109
+ const port = await listenLocal(server);
110
+ const close = () => {
111
+ closePromise ??= closeCallbackServer();
112
+ return closePromise;
113
+ };
114
+ const onAbort = () => {
115
+ close().catch(() => void 0);
116
+ };
117
+ args.signal?.addEventListener("abort", onAbort, { once: true });
118
+ if (args.signal?.aborted) onAbort();
119
+ return {
120
+ url: `http://127.0.0.1:${port}/evaluate`,
121
+ token: args.token,
122
+ evaluations: () => evaluations,
123
+ close
124
+ };
125
+ async function closeCallbackServer() {
126
+ args.signal?.removeEventListener("abort", onAbort);
127
+ accepting = false;
128
+ const closingServer = closeServer(server);
129
+ server.closeIdleConnections?.();
130
+ for (const controller of activeControllers) controller.abort();
131
+ const [serverResult] = await Promise.allSettled([closingServer, waitForActiveHandlers(activeHandlers)]);
132
+ if (activeControllers.size !== 0 || activeHandlers.size !== 0) throw new Error("external optimizer callback closed with active request work");
133
+ if (serverResult?.status === "rejected") throw serverResult.reason;
134
+ }
135
+ }
136
+ async function handleCallback(request, response, signal, args, limits, nextEvaluation, observe, proposedCandidateHashes) {
137
+ try {
138
+ if (request.method !== "POST" || request.url !== "/evaluate") {
139
+ sendJsonIfOpen(response, 404, { error: "not found" });
140
+ return;
141
+ }
142
+ if (request.headers.authorization !== `Bearer ${args.token}`) {
143
+ sendJsonIfOpen(response, 401, { error: "unauthorized" });
144
+ return;
145
+ }
146
+ let body;
147
+ try {
148
+ body = await readJson(request, limits.maxRequestBytes);
149
+ } catch {
150
+ observe({
151
+ kind: "refusal",
152
+ reason: "invalid-request"
153
+ });
154
+ sendJsonIfOpen(response, 400, { error: "request body must be bounded valid JSON" });
155
+ return;
156
+ }
157
+ if (!isRecord(body) || !isExternalTextCandidate(body.candidate) || typeof body.exampleId !== "string") {
158
+ observe({
159
+ kind: "refusal",
160
+ reason: "invalid-request"
161
+ });
162
+ sendJsonIfOpen(response, 400, { error: "candidate and exampleId are required strings" });
163
+ return;
164
+ }
165
+ const candidate = cloneCandidate(body.candidate);
166
+ const candidateHash = contentHash({
167
+ kind: "external-text-candidate",
168
+ candidate
169
+ });
170
+ if (!proposedCandidateHashes.has(candidateHash)) {
171
+ observe({
172
+ kind: "proposal",
173
+ candidate,
174
+ candidateHash
175
+ });
176
+ proposedCandidateHashes.add(candidateHash);
177
+ }
178
+ const count = nextEvaluation();
179
+ if (count === void 0) {
180
+ observe({
181
+ kind: "refusal",
182
+ reason: "evaluation-limit",
183
+ candidate,
184
+ candidateHash,
185
+ exampleId: body.exampleId
186
+ });
187
+ sendJsonIfOpen(response, 429, { error: "evaluation limit reached" });
188
+ return;
189
+ }
190
+ let result;
191
+ try {
192
+ result = await args.evaluate({
193
+ candidate,
194
+ exampleId: body.exampleId
195
+ }, signal);
196
+ } catch {
197
+ observe({
198
+ kind: "refusal",
199
+ reason: "evaluation-failed",
200
+ candidate,
201
+ candidateHash,
202
+ exampleId: body.exampleId
203
+ });
204
+ sendJsonIfOpen(response, 500, { error: "evaluation failed" });
205
+ return;
206
+ }
207
+ let encoded;
208
+ try {
209
+ encoded = encodeJsonResponse(result);
210
+ } catch {
211
+ observe({
212
+ kind: "refusal",
213
+ reason: "evaluation-failed",
214
+ candidate,
215
+ candidateHash,
216
+ exampleId: body.exampleId
217
+ });
218
+ sendJsonIfOpen(response, 500, { error: "evaluation response must be valid JSON" });
219
+ return;
220
+ }
221
+ if (Buffer.byteLength(encoded) > limits.maxResponseBytes) {
222
+ observe({
223
+ kind: "refusal",
224
+ reason: "evaluation-failed",
225
+ candidate,
226
+ candidateHash,
227
+ exampleId: body.exampleId
228
+ });
229
+ sendJsonIfOpen(response, 413, { error: "evaluation response too large" });
230
+ return;
231
+ }
232
+ observe({
233
+ kind: "evaluation",
234
+ candidate,
235
+ candidateHash,
236
+ exampleId: body.exampleId,
237
+ evaluationNumber: count,
238
+ response: structuredClone(result)
239
+ });
240
+ sendEncodedJsonIfOpen(response, 200, encoded);
241
+ } catch {
242
+ sendJsonIfOpen(response, 500, { error: "evaluation failed" });
243
+ }
244
+ }
245
+ async function waitForActiveHandlers(activeHandlers) {
246
+ while (activeHandlers.size > 0) await Promise.allSettled([...activeHandlers]);
247
+ }
248
+ function sendJsonIfOpen(response, status, body) {
249
+ if (response.destroyed || response.writableEnded) return;
250
+ sendJson(response, status, body);
251
+ }
252
+ function readJson(request, maxRequestBytes) {
253
+ return new Promise((resolvePromise, reject) => {
254
+ let size = 0;
255
+ const chunks = [];
256
+ request.on("data", (chunk) => {
257
+ size += chunk.length;
258
+ if (size > maxRequestBytes) {
259
+ reject(/* @__PURE__ */ new Error("callback body too large"));
260
+ request.destroy();
261
+ return;
262
+ }
263
+ chunks.push(chunk);
264
+ });
265
+ request.on("error", reject);
266
+ request.on("end", () => {
267
+ try {
268
+ resolvePromise(JSON.parse(Buffer.concat(chunks).toString("utf8")));
269
+ } catch (error) {
270
+ reject(error);
271
+ }
272
+ });
273
+ });
274
+ }
275
+ function encodeJsonResponse(value) {
276
+ const encoded = JSON.stringify(value);
277
+ if (encoded === void 0) throw new Error("evaluation response must be JSON-serializable");
278
+ return encoded;
279
+ }
280
+ function sendEncodedJsonIfOpen(response, status, encoded) {
281
+ if (response.destroyed || response.writableEnded) return;
282
+ response.writeHead(status, {
283
+ "content-type": "application/json; charset=utf-8",
284
+ "content-length": String(Buffer.byteLength(encoded))
285
+ });
286
+ response.end(encoded);
287
+ }
288
+ function assertCallbackConfig(args) {
289
+ if (typeof args.token !== "string" || !args.token.trim()) throw new Error("external optimizer callback: token must be non-empty");
290
+ if (!Number.isSafeInteger(args.maxEvaluations) || args.maxEvaluations <= 0) throw new Error("external optimizer callback: maxEvaluations must be a positive safe integer");
291
+ if (args.acceptEvaluation !== void 0 && typeof args.acceptEvaluation !== "function") throw new Error("external optimizer callback: acceptEvaluation must be a function");
292
+ if (typeof args.evaluate !== "function") throw new Error("external optimizer callback: evaluate must be a function");
293
+ if (args.observe !== void 0 && typeof args.observe !== "function") throw new Error("external optimizer callback: observe must be a function");
294
+ }
295
+ function cloneCandidate(candidate) {
296
+ return typeof candidate === "string" ? candidate : { ...candidate };
297
+ }
298
+ //#endregion
299
+ export { acquireSingleRunLock as n, startExternalOptimizerCallback as t };
300
+
301
+ //# sourceMappingURL=external-optimizer-process-BTiNB-RH.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"external-optimizer-process-BTiNB-RH.js","names":[],"sources":["../src/campaign/single-run-lock.ts","../src/campaign/external-optimizer-callback.ts"],"sourcesContent":["/**\n * Single-run lock for evaluations that share one mutable environment.\n *\n * Two concurrent runs against a shared stateful gym silently corrupt each\n * other: each resets/mutates environment state mid-cell of the other, and\n * every score from both becomes garbage that LOOKS like worker variance\n * (agent-lab R357 burned hours on flip-flopping scores before tracing them\n * to exactly this). The fix is a pid lockfile: refuse to start while a live\n * holder exists, reclaim stale locks whose pid is gone, release only if the\n * lock is still ours.\n *\n * `alsoCheck` exists because independent runners can guard the same shared\n * resource with differently named lockfiles; a runner must respect all of\n * them even though it writes only its own.\n */\n\nimport {\n type AtomicFileLockUnavailable,\n probeAtomicFileLock,\n tryAcquireAtomicFileLock,\n} from '../ledger-core/atomic-file-lock'\n\nexport interface SingleRunLockOptions {\n /** Lockfile this runner writes (and checks). */\n readonly lockPath: string\n /** Other runners' lockfiles guarding the same resource; checked, never written. */\n readonly alsoCheck?: readonly string[]\n /** Install a process 'exit' hook that releases the lock. Default true. */\n readonly releaseOnExit?: boolean\n /** Owner pid recorded in the lockfile metadata. Default process.pid. */\n readonly pid?: number\n}\n\nexport interface SingleRunLock {\n /** Remove the lockfile if this process still owns it. Idempotent. */\n release(): void\n}\n\nfunction assertAvailable(path: string): void {\n const unavailable = probeAtomicFileLock({ lockPath: path })\n if (unavailable) throw unavailableError(path, unavailable)\n}\n\nfunction unavailableError(path: string, unavailable: AtomicFileLockUnavailable): Error {\n if (unavailable.reason === 'recovery') {\n return new Error(\n `single-run lock recovery is already in progress (${path}); refusing a concurrent run`,\n )\n }\n return new Error(\n `single-run lock held by live pid ${unavailable.holder.pid} (${path}); refusing a concurrent run on the shared resource`,\n )\n}\n\n/**\n * Acquire the lock or throw naming the live holder. A stale lock (holder pid\n * no longer running) is reclaimed by one contender. An interrupted reclaim\n * leaves a marker that fails closed instead of admitting overlapping runs.\n */\nexport function acquireSingleRunLock(opts: SingleRunLockOptions): SingleRunLock {\n const pid = opts.pid ?? process.pid\n if (!Number.isSafeInteger(pid) || pid <= 0)\n throw new Error('single-run lock pid must be a positive integer')\n for (const path of opts.alsoCheck ?? []) assertAvailable(path)\n const acquisition = tryAcquireAtomicFileLock({\n lockPath: opts.lockPath,\n pid,\n })\n if (!acquisition.acquired) throw unavailableError(opts.lockPath, acquisition)\n const releaseOnExit = opts.releaseOnExit ?? true\n let released = false\n const release = (): void => {\n if (released) return\n released = true\n if (releaseOnExit) process.off('exit', release)\n try {\n acquisition.lock.release()\n } catch {\n // release is best-effort; a leftover stale lock is reclaimed on next acquire\n }\n }\n try {\n for (const path of opts.alsoCheck ?? []) assertAvailable(path)\n } catch (error) {\n release()\n throw error\n }\n if (releaseOnExit) process.on('exit', release)\n return { release }\n}\n","import { createServer, type IncomingMessage, type ServerResponse } from 'node:http'\nimport { contentHash } from '../verdict-cache'\nimport {\n type ExternalOptimizerCallback,\n type ExternalOptimizerCallbackLimits,\n type ExternalOptimizerEvaluationObservation,\n type ExternalTextCandidate,\n type ExternalTextEvaluationRequest,\n isExternalTextCandidate,\n isRecord,\n resolveExternalOptimizerCallbackLimits,\n} from './external-optimizer-contracts'\nimport { closeServer, listenLocal, sendJson } from './external-optimizer-http'\n\ntype UnsequencedObservation = ExternalOptimizerEvaluationObservation extends infer T\n ? T extends ExternalOptimizerEvaluationObservation\n ? Omit<T, 'sequence'>\n : never\n : never\n\nexport async function startExternalOptimizerCallback<TResponse>(args: {\n token: string\n maxEvaluations: number\n acceptEvaluation?: () => number | undefined\n evaluate: (request: ExternalTextEvaluationRequest, signal: AbortSignal) => Promise<TResponse>\n observe?: (observation: ExternalOptimizerEvaluationObservation) => void\n /** Loopback JSON byte limits. Omitted fields use finite defaults. */\n limits?: Partial<ExternalOptimizerCallbackLimits>\n signal?: AbortSignal\n}): Promise<ExternalOptimizerCallback> {\n assertCallbackConfig(args)\n const limits = resolveExternalOptimizerCallbackLimits(args.limits)\n args.signal?.throwIfAborted()\n let evaluations = 0\n let accepting = true\n let closePromise: Promise<void> | undefined\n const activeControllers = new Set<AbortController>()\n const activeHandlers = new Set<Promise<void>>()\n const proposedCandidateHashes = new Set<string>()\n let observationSequence = 0\n const server = createServer((request, response) => {\n if (!accepting) {\n sendJsonIfOpen(response, 503, { error: 'external optimizer callback is closing' })\n return\n }\n\n const controller = new AbortController()\n const abortRequest = (): void => {\n request.destroy()\n response.destroy()\n }\n activeControllers.add(controller)\n controller.signal.addEventListener('abort', abortRequest, { once: true })\n\n let handler!: Promise<void>\n handler = handleCallback(\n request,\n response,\n controller.signal,\n args,\n limits,\n () => {\n const accepted = args.acceptEvaluation ? args.acceptEvaluation() : evaluations + 1\n if (accepted === undefined) return undefined\n if (!Number.isSafeInteger(accepted) || accepted <= 0) {\n throw new Error('external optimizer callback: invalid accepted evaluation count')\n }\n if (accepted > args.maxEvaluations) return undefined\n evaluations += 1\n return accepted\n },\n (observation) => {\n if (!args.observe) return\n args.observe({ ...observation, sequence: ++observationSequence })\n },\n proposedCandidateHashes,\n ).finally(() => {\n controller.signal.removeEventListener('abort', abortRequest)\n activeControllers.delete(controller)\n activeHandlers.delete(handler)\n })\n activeHandlers.add(handler)\n void handler.catch(() => undefined)\n })\n const port = await listenLocal(server)\n const close = (): Promise<void> => {\n closePromise ??= closeCallbackServer()\n return closePromise\n }\n const onAbort = (): void => {\n void close().catch(() => undefined)\n }\n args.signal?.addEventListener('abort', onAbort, { once: true })\n if (args.signal?.aborted) onAbort()\n return {\n url: `http://127.0.0.1:${port}/evaluate`,\n token: args.token,\n evaluations: () => evaluations,\n close,\n }\n\n async function closeCallbackServer(): Promise<void> {\n args.signal?.removeEventListener('abort', onAbort)\n accepting = false\n const closingServer = closeServer(server)\n server.closeIdleConnections?.()\n for (const controller of activeControllers) controller.abort()\n const [serverResult] = await Promise.allSettled([\n closingServer,\n waitForActiveHandlers(activeHandlers),\n ])\n if (activeControllers.size !== 0 || activeHandlers.size !== 0) {\n throw new Error('external optimizer callback closed with active request work')\n }\n if (serverResult?.status === 'rejected') throw serverResult.reason\n }\n}\n\nasync function handleCallback<TResponse>(\n request: IncomingMessage,\n response: ServerResponse,\n signal: AbortSignal,\n args: {\n token: string\n maxEvaluations: number\n acceptEvaluation?: () => number | undefined\n evaluate: (request: ExternalTextEvaluationRequest, signal: AbortSignal) => Promise<TResponse>\n observe?: (observation: ExternalOptimizerEvaluationObservation) => void\n },\n limits: ExternalOptimizerCallbackLimits,\n nextEvaluation: () => number | undefined,\n observe: (observation: UnsequencedObservation) => void,\n proposedCandidateHashes: Set<string>,\n): Promise<void> {\n try {\n if (request.method !== 'POST' || request.url !== '/evaluate') {\n sendJsonIfOpen(response, 404, { error: 'not found' })\n return\n }\n if (request.headers.authorization !== `Bearer ${args.token}`) {\n sendJsonIfOpen(response, 401, { error: 'unauthorized' })\n return\n }\n let body: unknown\n try {\n body = await readJson(request, limits.maxRequestBytes)\n } catch {\n observe({ kind: 'refusal', reason: 'invalid-request' })\n sendJsonIfOpen(response, 400, { error: 'request body must be bounded valid JSON' })\n return\n }\n if (\n !isRecord(body) ||\n !isExternalTextCandidate(body.candidate) ||\n typeof body.exampleId !== 'string'\n ) {\n observe({ kind: 'refusal', reason: 'invalid-request' })\n sendJsonIfOpen(response, 400, { error: 'candidate and exampleId are required strings' })\n return\n }\n const candidate = cloneCandidate(body.candidate)\n const candidateHash = contentHash({ kind: 'external-text-candidate', candidate })\n if (!proposedCandidateHashes.has(candidateHash)) {\n observe({ kind: 'proposal', candidate, candidateHash })\n proposedCandidateHashes.add(candidateHash)\n }\n const count = nextEvaluation()\n if (count === undefined) {\n observe({\n kind: 'refusal',\n reason: 'evaluation-limit',\n candidate,\n candidateHash,\n exampleId: body.exampleId,\n })\n sendJsonIfOpen(response, 429, { error: 'evaluation limit reached' })\n return\n }\n let result: TResponse\n try {\n result = await args.evaluate({ candidate, exampleId: body.exampleId }, signal)\n } catch {\n observe({\n kind: 'refusal',\n reason: 'evaluation-failed',\n candidate,\n candidateHash,\n exampleId: body.exampleId,\n })\n sendJsonIfOpen(response, 500, { error: 'evaluation failed' })\n return\n }\n let encoded: string\n try {\n encoded = encodeJsonResponse(result)\n } catch {\n observe({\n kind: 'refusal',\n reason: 'evaluation-failed',\n candidate,\n candidateHash,\n exampleId: body.exampleId,\n })\n sendJsonIfOpen(response, 500, { error: 'evaluation response must be valid JSON' })\n return\n }\n if (Buffer.byteLength(encoded) > limits.maxResponseBytes) {\n observe({\n kind: 'refusal',\n reason: 'evaluation-failed',\n candidate,\n candidateHash,\n exampleId: body.exampleId,\n })\n sendJsonIfOpen(response, 413, { error: 'evaluation response too large' })\n return\n }\n observe({\n kind: 'evaluation',\n candidate,\n candidateHash,\n exampleId: body.exampleId,\n evaluationNumber: count,\n response: structuredClone(result),\n })\n sendEncodedJsonIfOpen(response, 200, encoded)\n } catch {\n sendJsonIfOpen(response, 500, { error: 'evaluation failed' })\n }\n}\n\nasync function waitForActiveHandlers(activeHandlers: Set<Promise<void>>): Promise<void> {\n while (activeHandlers.size > 0) {\n await Promise.allSettled([...activeHandlers])\n }\n}\n\nfunction sendJsonIfOpen(response: ServerResponse, status: number, body: unknown): void {\n if (response.destroyed || response.writableEnded) return\n sendJson(response, status, body)\n}\n\nfunction readJson(request: IncomingMessage, maxRequestBytes: number): Promise<unknown> {\n return new Promise((resolvePromise, reject) => {\n let size = 0\n const chunks: Buffer[] = []\n request.on('data', (chunk: Buffer) => {\n size += chunk.length\n if (size > maxRequestBytes) {\n reject(new Error('callback body too large'))\n request.destroy()\n return\n }\n chunks.push(chunk)\n })\n request.on('error', reject)\n request.on('end', () => {\n try {\n resolvePromise(JSON.parse(Buffer.concat(chunks).toString('utf8')))\n } catch (error) {\n reject(error)\n }\n })\n })\n}\n\nfunction encodeJsonResponse(value: unknown): string {\n const encoded = JSON.stringify(value)\n if (encoded === undefined) throw new Error('evaluation response must be JSON-serializable')\n return encoded\n}\n\nfunction sendEncodedJsonIfOpen(response: ServerResponse, status: number, encoded: string): void {\n if (response.destroyed || response.writableEnded) return\n response.writeHead(status, {\n 'content-type': 'application/json; charset=utf-8',\n 'content-length': String(Buffer.byteLength(encoded)),\n })\n response.end(encoded)\n}\n\nfunction assertCallbackConfig(args: {\n token: string\n maxEvaluations: number\n acceptEvaluation?: () => number | undefined\n evaluate: (request: ExternalTextEvaluationRequest, signal: AbortSignal) => Promise<unknown>\n observe?: (observation: ExternalOptimizerEvaluationObservation) => void\n limits?: Partial<ExternalOptimizerCallbackLimits>\n signal?: AbortSignal\n}): void {\n if (typeof args.token !== 'string' || !args.token.trim()) {\n throw new Error('external optimizer callback: token must be non-empty')\n }\n if (!Number.isSafeInteger(args.maxEvaluations) || args.maxEvaluations <= 0) {\n throw new Error('external optimizer callback: maxEvaluations must be a positive safe integer')\n }\n if (args.acceptEvaluation !== undefined && typeof args.acceptEvaluation !== 'function') {\n throw new Error('external optimizer callback: acceptEvaluation must be a function')\n }\n if (typeof args.evaluate !== 'function') {\n throw new Error('external optimizer callback: evaluate must be a function')\n }\n if (args.observe !== undefined && typeof args.observe !== 'function') {\n throw new Error('external optimizer callback: observe must be a function')\n }\n}\n\nfunction cloneCandidate(candidate: ExternalTextCandidate): ExternalTextCandidate {\n return typeof candidate === 'string' ? candidate : { ...candidate }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAsCA,SAAS,gBAAgB,MAAoB;CAC3C,MAAM,cAAc,oBAAoB,EAAE,UAAU,KAAK,CAAC;CAC1D,IAAI,aAAa,MAAM,iBAAiB,MAAM,WAAW;AAC3D;AAEA,SAAS,iBAAiB,MAAc,aAA+C;CACrF,IAAI,YAAY,WAAW,YACzB,uBAAO,IAAI,MACT,oDAAoD,KAAK,6BAC3D;CAEF,uBAAO,IAAI,MACT,oCAAoC,YAAY,OAAO,IAAI,IAAI,KAAK,oDACtE;AACF;;;;;;AAOA,SAAgB,qBAAqB,MAA2C;CAC9E,MAAM,MAAM,KAAK,OAAO,QAAQ;CAChC,IAAI,CAAC,OAAO,cAAc,GAAG,KAAK,OAAO,GACvC,MAAM,IAAI,MAAM,gDAAgD;CAClE,KAAK,MAAM,QAAQ,KAAK,aAAa,CAAC,GAAG,gBAAgB,IAAI;CAC7D,MAAM,cAAc,yBAAyB;EAC3C,UAAU,KAAK;EACf;CACF,CAAC;CACD,IAAI,CAAC,YAAY,UAAU,MAAM,iBAAiB,KAAK,UAAU,WAAW;CAC5E,MAAM,gBAAgB,KAAK,iBAAiB;CAC5C,IAAI,WAAW;CACf,MAAM,gBAAsB;EAC1B,IAAI,UAAU;EACd,WAAW;EACX,IAAI,eAAe,QAAQ,IAAI,QAAQ,OAAO;EAC9C,IAAI;GACF,YAAY,KAAK,QAAQ;EAC3B,QAAQ,CAER;CACF;CACA,IAAI;EACF,KAAK,MAAM,QAAQ,KAAK,aAAa,CAAC,GAAG,gBAAgB,IAAI;CAC/D,SAAS,OAAO;EACd,QAAQ;EACR,MAAM;CACR;CACA,IAAI,eAAe,QAAQ,GAAG,QAAQ,OAAO;CAC7C,OAAO,EAAE,QAAQ;AACnB;;;ACrEA,eAAsB,+BAA0C,MASzB;CACrC,qBAAqB,IAAI;CACzB,MAAM,SAAS,uCAAuC,KAAK,MAAM;CACjE,KAAK,QAAQ,eAAe;CAC5B,IAAI,cAAc;CAClB,IAAI,YAAY;CAChB,IAAI;CACJ,MAAM,oCAAoB,IAAI,IAAqB;CACnD,MAAM,iCAAiB,IAAI,IAAmB;CAC9C,MAAM,0CAA0B,IAAI,IAAY;CAChD,IAAI,sBAAsB;CAC1B,MAAM,SAAS,cAAc,SAAS,aAAa;EACjD,IAAI,CAAC,WAAW;GACd,eAAe,UAAU,KAAK,EAAE,OAAO,yCAAyC,CAAC;GACjF;EACF;EAEA,MAAM,aAAa,IAAI,gBAAgB;EACvC,MAAM,qBAA2B;GAC/B,QAAQ,QAAQ;GAChB,SAAS,QAAQ;EACnB;EACA,kBAAkB,IAAI,UAAU;EAChC,WAAW,OAAO,iBAAiB,SAAS,cAAc,EAAE,MAAM,KAAK,CAAC;EAExE,IAAI;EACJ,UAAU,eACR,SACA,UACA,WAAW,QACX,MACA,cACM;GACJ,MAAM,WAAW,KAAK,mBAAmB,KAAK,iBAAiB,IAAI,cAAc;GACjF,IAAI,aAAa,KAAA,GAAW,OAAO,KAAA;GACnC,IAAI,CAAC,OAAO,cAAc,QAAQ,KAAK,YAAY,GACjD,MAAM,IAAI,MAAM,gEAAgE;GAElF,IAAI,WAAW,KAAK,gBAAgB,OAAO,KAAA;GAC3C,eAAe;GACf,OAAO;EACT,IACC,gBAAgB;GACf,IAAI,CAAC,KAAK,SAAS;GACnB,KAAK,QAAQ;IAAE,GAAG;IAAa,UAAU,EAAE;GAAoB,CAAC;EAClE,GACA,uBACF,CAAC,CAAC,cAAc;GACd,WAAW,OAAO,oBAAoB,SAAS,YAAY;GAC3D,kBAAkB,OAAO,UAAU;GACnC,eAAe,OAAO,OAAO;EAC/B,CAAC;EACD,eAAe,IAAI,OAAO;EAC1B,QAAa,YAAY,KAAA,CAAS;CACpC,CAAC;CACD,MAAM,OAAO,MAAM,YAAY,MAAM;CACrC,MAAM,cAA6B;EACjC,iBAAiB,oBAAoB;EACrC,OAAO;CACT;CACA,MAAM,gBAAsB;EAC1B,MAAW,CAAC,CAAC,YAAY,KAAA,CAAS;CACpC;CACA,KAAK,QAAQ,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;CAC9D,IAAI,KAAK,QAAQ,SAAS,QAAQ;CAClC,OAAO;EACL,KAAK,oBAAoB,KAAK;EAC9B,OAAO,KAAK;EACZ,mBAAmB;EACnB;CACF;CAEA,eAAe,sBAAqC;EAClD,KAAK,QAAQ,oBAAoB,SAAS,OAAO;EACjD,YAAY;EACZ,MAAM,gBAAgB,YAAY,MAAM;EACxC,OAAO,uBAAuB;EAC9B,KAAK,MAAM,cAAc,mBAAmB,WAAW,MAAM;EAC7D,MAAM,CAAC,gBAAgB,MAAM,QAAQ,WAAW,CAC9C,eACA,sBAAsB,cAAc,CACtC,CAAC;EACD,IAAI,kBAAkB,SAAS,KAAK,eAAe,SAAS,GAC1D,MAAM,IAAI,MAAM,6DAA6D;EAE/E,IAAI,cAAc,WAAW,YAAY,MAAM,aAAa;CAC9D;AACF;AAEA,eAAe,eACb,SACA,UACA,QACA,MAOA,QACA,gBACA,SACA,yBACe;CACf,IAAI;EACF,IAAI,QAAQ,WAAW,UAAU,QAAQ,QAAQ,aAAa;GAC5D,eAAe,UAAU,KAAK,EAAE,OAAO,YAAY,CAAC;GACpD;EACF;EACA,IAAI,QAAQ,QAAQ,kBAAkB,UAAU,KAAK,SAAS;GAC5D,eAAe,UAAU,KAAK,EAAE,OAAO,eAAe,CAAC;GACvD;EACF;EACA,IAAI;EACJ,IAAI;GACF,OAAO,MAAM,SAAS,SAAS,OAAO,eAAe;EACvD,QAAQ;GACN,QAAQ;IAAE,MAAM;IAAW,QAAQ;GAAkB,CAAC;GACtD,eAAe,UAAU,KAAK,EAAE,OAAO,0CAA0C,CAAC;GAClF;EACF;EACA,IACE,CAAC,SAAS,IAAI,KACd,CAAC,wBAAwB,KAAK,SAAS,KACvC,OAAO,KAAK,cAAc,UAC1B;GACA,QAAQ;IAAE,MAAM;IAAW,QAAQ;GAAkB,CAAC;GACtD,eAAe,UAAU,KAAK,EAAE,OAAO,+CAA+C,CAAC;GACvF;EACF;EACA,MAAM,YAAY,eAAe,KAAK,SAAS;EAC/C,MAAM,gBAAgB,YAAY;GAAE,MAAM;GAA2B;EAAU,CAAC;EAChF,IAAI,CAAC,wBAAwB,IAAI,aAAa,GAAG;GAC/C,QAAQ;IAAE,MAAM;IAAY;IAAW;GAAc,CAAC;GACtD,wBAAwB,IAAI,aAAa;EAC3C;EACA,MAAM,QAAQ,eAAe;EAC7B,IAAI,UAAU,KAAA,GAAW;GACvB,QAAQ;IACN,MAAM;IACN,QAAQ;IACR;IACA;IACA,WAAW,KAAK;GAClB,CAAC;GACD,eAAe,UAAU,KAAK,EAAE,OAAO,2BAA2B,CAAC;GACnE;EACF;EACA,IAAI;EACJ,IAAI;GACF,SAAS,MAAM,KAAK,SAAS;IAAE;IAAW,WAAW,KAAK;GAAU,GAAG,MAAM;EAC/E,QAAQ;GACN,QAAQ;IACN,MAAM;IACN,QAAQ;IACR;IACA;IACA,WAAW,KAAK;GAClB,CAAC;GACD,eAAe,UAAU,KAAK,EAAE,OAAO,oBAAoB,CAAC;GAC5D;EACF;EACA,IAAI;EACJ,IAAI;GACF,UAAU,mBAAmB,MAAM;EACrC,QAAQ;GACN,QAAQ;IACN,MAAM;IACN,QAAQ;IACR;IACA;IACA,WAAW,KAAK;GAClB,CAAC;GACD,eAAe,UAAU,KAAK,EAAE,OAAO,yCAAyC,CAAC;GACjF;EACF;EACA,IAAI,OAAO,WAAW,OAAO,IAAI,OAAO,kBAAkB;GACxD,QAAQ;IACN,MAAM;IACN,QAAQ;IACR;IACA;IACA,WAAW,KAAK;GAClB,CAAC;GACD,eAAe,UAAU,KAAK,EAAE,OAAO,gCAAgC,CAAC;GACxE;EACF;EACA,QAAQ;GACN,MAAM;GACN;GACA;GACA,WAAW,KAAK;GAChB,kBAAkB;GAClB,UAAU,gBAAgB,MAAM;EAClC,CAAC;EACD,sBAAsB,UAAU,KAAK,OAAO;CAC9C,QAAQ;EACN,eAAe,UAAU,KAAK,EAAE,OAAO,oBAAoB,CAAC;CAC9D;AACF;AAEA,eAAe,sBAAsB,gBAAmD;CACtF,OAAO,eAAe,OAAO,GAC3B,MAAM,QAAQ,WAAW,CAAC,GAAG,cAAc,CAAC;AAEhD;AAEA,SAAS,eAAe,UAA0B,QAAgB,MAAqB;CACrF,IAAI,SAAS,aAAa,SAAS,eAAe;CAClD,SAAS,UAAU,QAAQ,IAAI;AACjC;AAEA,SAAS,SAAS,SAA0B,iBAA2C;CACrF,OAAO,IAAI,SAAS,gBAAgB,WAAW;EAC7C,IAAI,OAAO;EACX,MAAM,SAAmB,CAAC;EAC1B,QAAQ,GAAG,SAAS,UAAkB;GACpC,QAAQ,MAAM;GACd,IAAI,OAAO,iBAAiB;IAC1B,uBAAO,IAAI,MAAM,yBAAyB,CAAC;IAC3C,QAAQ,QAAQ;IAChB;GACF;GACA,OAAO,KAAK,KAAK;EACnB,CAAC;EACD,QAAQ,GAAG,SAAS,MAAM;EAC1B,QAAQ,GAAG,aAAa;GACtB,IAAI;IACF,eAAe,KAAK,MAAM,OAAO,OAAO,MAAM,CAAC,CAAC,SAAS,MAAM,CAAC,CAAC;GACnE,SAAS,OAAO;IACd,OAAO,KAAK;GACd;EACF,CAAC;CACH,CAAC;AACH;AAEA,SAAS,mBAAmB,OAAwB;CAClD,MAAM,UAAU,KAAK,UAAU,KAAK;CACpC,IAAI,YAAY,KAAA,GAAW,MAAM,IAAI,MAAM,+CAA+C;CAC1F,OAAO;AACT;AAEA,SAAS,sBAAsB,UAA0B,QAAgB,SAAuB;CAC9F,IAAI,SAAS,aAAa,SAAS,eAAe;CAClD,SAAS,UAAU,QAAQ;EACzB,gBAAgB;EAChB,kBAAkB,OAAO,OAAO,WAAW,OAAO,CAAC;CACrD,CAAC;CACD,SAAS,IAAI,OAAO;AACtB;AAEA,SAAS,qBAAqB,MAQrB;CACP,IAAI,OAAO,KAAK,UAAU,YAAY,CAAC,KAAK,MAAM,KAAK,GACrD,MAAM,IAAI,MAAM,sDAAsD;CAExE,IAAI,CAAC,OAAO,cAAc,KAAK,cAAc,KAAK,KAAK,kBAAkB,GACvE,MAAM,IAAI,MAAM,6EAA6E;CAE/F,IAAI,KAAK,qBAAqB,KAAA,KAAa,OAAO,KAAK,qBAAqB,YAC1E,MAAM,IAAI,MAAM,kEAAkE;CAEpF,IAAI,OAAO,KAAK,aAAa,YAC3B,MAAM,IAAI,MAAM,0DAA0D;CAE5E,IAAI,KAAK,YAAY,KAAA,KAAa,OAAO,KAAK,YAAY,YACxD,MAAM,IAAI,MAAM,yDAAyD;AAE7E;AAEA,SAAS,eAAe,WAAyD;CAC/E,OAAO,OAAO,cAAc,WAAW,YAAY,EAAE,GAAG,UAAU;AACpE"}