reeval 0.24.0 → 0.26.0

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 (560) hide show
  1. package/CHANGELOG.md +64 -0
  2. package/README.md +16 -16
  3. package/THIRD_PARTY_NOTICES.md +0 -2068
  4. package/assets/package.json +1 -1
  5. package/assets/pnpm-lock.yaml +65 -50
  6. package/dist/{adapter-dHdepWKb.js → adapter-BHMbIeH9.js} +7 -7
  7. package/dist/{adapter-dHdepWKb.js.map → adapter-BHMbIeH9.js.map} +1 -1
  8. package/dist/{agent-CZpiwA_r.js → agent-D67VCd6Q.js} +8 -8
  9. package/dist/{agent-CZpiwA_r.js.map → agent-D67VCd6Q.js.map} +1 -1
  10. package/dist/{agents-CEapsHV-.js → agents-Dnv1MmgY.js} +40 -59
  11. package/dist/agents-Dnv1MmgY.js.map +1 -0
  12. package/dist/{analysis-BnG_1x_O.js → analysis-8vvMnUJB.js} +4 -5
  13. package/dist/analysis-8vvMnUJB.js.map +1 -0
  14. package/dist/api-ZeGnedB3.js +141 -0
  15. package/dist/api-ZeGnedB3.js.map +1 -0
  16. package/dist/assemble-BlA7FWRn.js +1 -0
  17. package/dist/{assemble-Dw1bF13I.js → assemble-CuvlsOaE.js} +117 -66
  18. package/dist/assemble-CuvlsOaE.js.map +1 -0
  19. package/dist/{attach-DvYiMu6w.js → attach-6av4DmO3.js} +24 -6
  20. package/dist/attach-6av4DmO3.js.map +1 -0
  21. package/dist/{attempt-DxR1agFo.js → attempt-D5WH0yFa.js} +3 -3
  22. package/dist/{attempt-DxR1agFo.js.map → attempt-D5WH0yFa.js.map} +1 -1
  23. package/dist/{attempt-host-DgeJFSZw.js → attempt-host-Be_-uDKJ.js} +37 -14
  24. package/dist/attempt-host-Be_-uDKJ.js.map +1 -0
  25. package/dist/boat-BCNNEJni.js +298 -0
  26. package/dist/boat-BCNNEJni.js.map +1 -0
  27. package/dist/boat-CwBW2sbp.js +1 -0
  28. package/dist/{boat-CUNgSc75.js → boat-CygRcaM0.js} +115 -196
  29. package/dist/boat-CygRcaM0.js.map +1 -0
  30. package/dist/{boat-images-BuFigiEb.js → boat-images-asGLQXD5.js} +19 -23
  31. package/dist/boat-images-asGLQXD5.js.map +1 -0
  32. package/dist/{bundle-yEgI3RsI.js → bundle-B7P-Awnm.js} +105 -17
  33. package/dist/bundle-B7P-Awnm.js.map +1 -0
  34. package/dist/{capture-DdEKVrwk.js → capture-BwLa-4s-.js} +20 -19
  35. package/dist/{capture-DdEKVrwk.js.map → capture-BwLa-4s-.js.map} +1 -1
  36. package/dist/capture-CG-J-1-j.js +1 -0
  37. package/dist/cases-Bsp-G-ZL.js +1 -0
  38. package/dist/{cases-B7QkvMGE.js → cases-C6vnV-gn.js} +60 -42
  39. package/dist/cases-C6vnV-gn.js.map +1 -0
  40. package/dist/classify-BIJQ2Xqo.js +163 -0
  41. package/dist/classify-BIJQ2Xqo.js.map +1 -0
  42. package/dist/{cli-Z7aw8k25.js → cli-BoCPYLUQ.js} +305 -124
  43. package/dist/cli-BoCPYLUQ.js.map +1 -0
  44. package/dist/{cli-spec-3zvLC9YW.js → cli-spec-DgEWHOt3.js} +164 -50
  45. package/dist/cli-spec-DgEWHOt3.js.map +1 -0
  46. package/dist/cli.js +2 -2
  47. package/dist/client-Dj65gXG3.js +105 -0
  48. package/dist/client-Dj65gXG3.js.map +1 -0
  49. package/dist/{cloud-owVBaSRt.js → cloud-TuNUlZ0L.js} +3 -3
  50. package/dist/{cloud-owVBaSRt.js.map → cloud-TuNUlZ0L.js.map} +1 -1
  51. package/dist/{cloud-files-BjinrTTe.js → cloud-files-CBovrfZA.js} +8 -8
  52. package/dist/{cloud-files-BjinrTTe.js.map → cloud-files-CBovrfZA.js.map} +1 -1
  53. package/dist/{cloud-DuHcNQVV.js → cloud-igGIS4Bm.js} +43 -20
  54. package/dist/cloud-igGIS4Bm.js.map +1 -0
  55. package/dist/cloud-project-Ch7jjlwh.js +110 -0
  56. package/dist/cloud-project-Ch7jjlwh.js.map +1 -0
  57. package/dist/cloud-project-DdUVB4EE.js +1 -0
  58. package/dist/cloud-push-DvnqOgRd.js +817 -0
  59. package/dist/cloud-push-DvnqOgRd.js.map +1 -0
  60. package/dist/cloud-push-MXKreNyq.js +1 -0
  61. package/dist/{cloud-run-DqMkSGYB.js → cloud-run-CbGF8Lwt.js} +285 -48
  62. package/dist/cloud-run-CbGF8Lwt.js.map +1 -0
  63. package/dist/cloud-run-CghE2dG6.js +1 -0
  64. package/dist/{cloud-secrets-Bak83tik.js → cloud-secrets-BWoV4ddM.js} +43 -35
  65. package/dist/cloud-secrets-BWoV4ddM.js.map +1 -0
  66. package/dist/cloud-secrets-DlPRfS2l.js +1 -0
  67. package/dist/command-Bps8kBML.js +141 -0
  68. package/dist/command-Bps8kBML.js.map +1 -0
  69. package/dist/{command-CtmSJzPC.js → command-ByUmSm2D.js} +29 -23
  70. package/dist/command-ByUmSm2D.js.map +1 -0
  71. package/dist/{command-CeygdoVe.js → command-CQk7FCI1.js} +92 -26
  72. package/dist/command-CQk7FCI1.js.map +1 -0
  73. package/dist/{command-CoHN1ibe.js → command-CxuGky7A.js} +44 -20
  74. package/dist/command-CxuGky7A.js.map +1 -0
  75. package/dist/{command-efTGfUYb.js → command-pxfK2bkV.js} +26 -16
  76. package/dist/command-pxfK2bkV.js.map +1 -0
  77. package/dist/commands-Cl3Y_UXg.js +1 -0
  78. package/dist/{commands-DR-NolV2.js → commands-DGWNNyRD.js} +54 -38
  79. package/dist/commands-DGWNNyRD.js.map +1 -0
  80. package/dist/{config-Bn2OhCN_.js → config-mU7wFvQq.js} +46 -32
  81. package/dist/config-mU7wFvQq.js.map +1 -0
  82. package/dist/core-BXYzOEyQ.js +2058 -0
  83. package/dist/core-BXYzOEyQ.js.map +1 -0
  84. package/dist/{slug-IHh3f770.js → cost-DQ0ayy6h.js} +37 -2
  85. package/dist/cost-DQ0ayy6h.js.map +1 -0
  86. package/dist/{cursor-catalog-BYIDSsLn.js → cursor-catalog-EOFft6LJ.js} +4 -4
  87. package/dist/{cursor-catalog-BYIDSsLn.js.map → cursor-catalog-EOFft6LJ.js.map} +1 -1
  88. package/dist/{daytona-CWs9j9mt.js → daytona-DtCjv8rH.js} +22 -17
  89. package/dist/daytona-DtCjv8rH.js.map +1 -0
  90. package/dist/daytona-image-Xy5o53Ji.js +1 -0
  91. package/dist/{daytona-image-DmGlFAyq.js → daytona-image-w4uTRZ0z.js} +6 -6
  92. package/dist/{daytona-image-DmGlFAyq.js.map → daytona-image-w4uTRZ0z.js.map} +1 -1
  93. package/dist/db-BcplCdzu.js +1 -0
  94. package/dist/{db-D6yQib7M.js → db-yigMuHM2.js} +31 -119
  95. package/dist/db-yigMuHM2.js.map +1 -0
  96. package/dist/debug-CGg7KHKq.js +1 -0
  97. package/dist/{debug-CIXzB4wK.js → debug-T29p6HOX.js} +2 -10
  98. package/dist/debug-T29p6HOX.js.map +1 -0
  99. package/dist/{default-effort-Ck6ViSaI.js → default-effort-B1b8t81b.js} +1 -1
  100. package/dist/{default-effort-Bpz5ULfC.js → default-effort-BDYXBzu5.js} +3 -3
  101. package/dist/{default-effort-Bpz5ULfC.js.map → default-effort-BDYXBzu5.js.map} +1 -1
  102. package/dist/{dist-CZtW7eI5.js → dist-DCG7vZGQ.js} +3 -3
  103. package/dist/{dist-CZtW7eI5.js.map → dist-DCG7vZGQ.js.map} +1 -1
  104. package/dist/{docker-Dgw4aaRk.js → docker-DcnX2WpM.js} +4 -4
  105. package/dist/{docker-Dgw4aaRk.js.map → docker-DcnX2WpM.js.map} +1 -1
  106. package/dist/{docker-Bbi4kmFj.js → docker-XCyDUd7_.js} +1 -1
  107. package/dist/{doctor-BfS6jswW.js → doctor-DTKa8rrq.js} +27 -27
  108. package/dist/{doctor-BfS6jswW.js.map → doctor-DTKa8rrq.js.map} +1 -1
  109. package/dist/duration-Bz2H7pXb.js +34 -0
  110. package/dist/duration-Bz2H7pXb.js.map +1 -0
  111. package/dist/{environment-BY8lVFFK.js → environment-BCBg6c0p.js} +38 -12
  112. package/dist/environment-BCBg6c0p.js.map +1 -0
  113. package/dist/errors-BjzEWESY.js +93 -0
  114. package/dist/errors-BjzEWESY.js.map +1 -0
  115. package/dist/{errors-P8oHhNzy.js → errors-Cda7UzRA.js} +2 -2
  116. package/dist/{errors-P8oHhNzy.js.map → errors-Cda7UzRA.js.map} +1 -1
  117. package/dist/{eval-source-BB8In3xo.js → eval-source-CSPyKuV6.js} +6 -6
  118. package/dist/{eval-source-BB8In3xo.js.map → eval-source-CSPyKuV6.js.map} +1 -1
  119. package/dist/{evaluate-B9_nukrc.d.ts → evaluate-Nnh0iPuo.d.ts} +1 -1
  120. package/dist/{evaluate-B8I9AKgw.js → evaluate-wgW1lGQj.js} +280 -60
  121. package/dist/evaluate-wgW1lGQj.js.map +1 -0
  122. package/dist/{events-D97ZzAeq.js → events-DG_Cq5Pn.js} +154 -33
  123. package/dist/events-DG_Cq5Pn.js.map +1 -0
  124. package/dist/events-RaK1pDmT.js +1 -0
  125. package/dist/{fake-bootstrap-4DFtzib0.js → fake-bootstrap-uCyHhnMO.js} +2 -2
  126. package/dist/{fake-bootstrap-4DFtzib0.js.map → fake-bootstrap-uCyHhnMO.js.map} +1 -1
  127. package/dist/files-BnsRehge.js +1 -0
  128. package/dist/{files-B0DbptiO.js → files-Bv6ipFjc.js} +4 -4
  129. package/dist/{files-B0DbptiO.js.map → files-Bv6ipFjc.js.map} +1 -1
  130. package/dist/{files-GBTUBAek.js → files-CcnSILye.js} +3 -3
  131. package/dist/{files-GBTUBAek.js.map → files-CcnSILye.js.map} +1 -1
  132. package/dist/{files-BesexM88.js → files-CjPm3yyj.js} +173 -41
  133. package/dist/files-CjPm3yyj.js.map +1 -0
  134. package/dist/files-CkmCV5rH.js +1 -0
  135. package/dist/{files-DE13BqBC.js → files-ClUbQPj9.js} +2 -2
  136. package/dist/{files-DE13BqBC.js.map → files-ClUbQPj9.js.map} +1 -1
  137. package/dist/{files-B4K1k__s.js → files-CqfVnRXw.js} +5 -5
  138. package/dist/{files-B4K1k__s.js.map → files-CqfVnRXw.js.map} +1 -1
  139. package/dist/files-D4Hb9HZS.js +1 -0
  140. package/dist/files-DCcRkJN4.js +1 -0
  141. package/dist/{files-TwSVeWwM.js → files-GAocj_Z1.js} +27 -41
  142. package/dist/files-GAocj_Z1.js.map +1 -0
  143. package/dist/gateway.d.ts +1 -1
  144. package/dist/{git-DatgUcnf.js → git-CRPr6I2T.js} +3 -3
  145. package/dist/{git-DatgUcnf.js.map → git-CRPr6I2T.js.map} +1 -1
  146. package/dist/{harness-FcTyyPco.js → harness-Dcd3ffkY.js} +3 -2
  147. package/dist/harness-Dcd3ffkY.js.map +1 -0
  148. package/dist/{harness-EZNoeUBf.js → harness-DvqNGwl6.js} +2 -2
  149. package/dist/{harness-EZNoeUBf.js.map → harness-DvqNGwl6.js.map} +1 -1
  150. package/dist/{harness-install-HgR73tpF.js → harness-install-B0Y2GjnO.js} +4 -4
  151. package/dist/{harness-install-HgR73tpF.js.map → harness-install-B0Y2GjnO.js.map} +1 -1
  152. package/dist/harness.d.ts +1 -1
  153. package/dist/harness.js +1 -1
  154. package/dist/http-DtlSzw4A.js +96 -0
  155. package/dist/http-DtlSzw4A.js.map +1 -0
  156. package/dist/{ids-DqEvrBFZ.js → ids-CtxRn13a.js} +4 -3
  157. package/dist/ids-CtxRn13a.js.map +1 -0
  158. package/dist/{images-SRE4f3SM.js → images-JspTw1vt.js} +13 -13
  159. package/dist/{images-SRE4f3SM.js.map → images-JspTw1vt.js.map} +1 -1
  160. package/dist/index.d.ts +7 -2
  161. package/dist/index.js +4 -5
  162. package/dist/init-BHIWfE2L.js +1 -0
  163. package/dist/{init-D5F9MYPC.js → init-wmMbauok.js} +6 -6
  164. package/dist/{init-D5F9MYPC.js.map → init-wmMbauok.js.map} +1 -1
  165. package/dist/{ink-list-CLq6YVB8.js → ink-list-B5Gul5iZ.js} +7 -6
  166. package/dist/ink-list-B5Gul5iZ.js.map +1 -0
  167. package/dist/{inputs-B_FsOD2N.js → inputs-BWTtz1-v.js} +6 -6
  168. package/dist/{inputs-B_FsOD2N.js.map → inputs-BWTtz1-v.js.map} +1 -1
  169. package/dist/inputs-load-C8HEtAej.js +1 -0
  170. package/dist/{inputs-load-B4MFkp0Q.js → inputs-load-DQRC3dUA.js} +3 -3
  171. package/dist/{inputs-load-B4MFkp0Q.js.map → inputs-load-DQRC3dUA.js.map} +1 -1
  172. package/dist/internal/vm-api/boat.d.ts +46 -0
  173. package/dist/internal/vm-api/boat.js +1 -0
  174. package/dist/internal/vm-api/daytona.d.ts +26 -0
  175. package/dist/internal/vm-api/daytona.js +220 -0
  176. package/dist/internal/vm-api/daytona.js.map +1 -0
  177. package/dist/internal/vm-api.d.ts +86 -0
  178. package/dist/internal/vm-api.js +3 -0
  179. package/dist/internal/vm.d.ts +1 -1
  180. package/dist/internal/vm.js +1 -1
  181. package/dist/{invocations-BxQqe5GU.js → invocations-CSA9OKFn.js} +5 -5
  182. package/dist/{invocations-BxQqe5GU.js.map → invocations-CSA9OKFn.js.map} +1 -1
  183. package/dist/issue-ipG9Yy1h.js +1 -0
  184. package/dist/{issue-iErT9KkK.js → issue-lKqtL3iX.js} +4 -4
  185. package/dist/{issue-iErT9KkK.js.map → issue-lKqtL3iX.js.map} +1 -1
  186. package/dist/{jiti-CtahaHaY.js → jiti-aNfS_lSj.js} +2 -2
  187. package/dist/{jiti-CtahaHaY.js.map → jiti-aNfS_lSj.js.map} +1 -1
  188. package/dist/{judge-MOKPsr6R.js → judge-etHC4K02.js} +5 -5
  189. package/dist/{judge-MOKPsr6R.js.map → judge-etHC4K02.js.map} +1 -1
  190. package/dist/judge.d.ts +1 -1
  191. package/dist/judge.js +1 -1
  192. package/dist/{judgments-Dur4D2pj.js → judgments-CVCUWT-g.js} +7 -4
  193. package/dist/{judgments-Dur4D2pj.js.map → judgments-CVCUWT-g.js.map} +1 -1
  194. package/dist/{kinds-Dohijzi-.js → kinds-BQX5ZxpK.js} +37 -11
  195. package/dist/{kinds-Dohijzi-.js.map → kinds-BQX5ZxpK.js.map} +1 -1
  196. package/dist/labels-Dtyf5-pd.js +251 -0
  197. package/dist/labels-Dtyf5-pd.js.map +1 -0
  198. package/dist/{layout-BhhJAEgf.js → layout-Co4tPHWI.js} +3 -3
  199. package/dist/{layout-BhhJAEgf.js.map → layout-Co4tPHWI.js.map} +1 -1
  200. package/dist/link-Cqs2DhxR.js +127 -0
  201. package/dist/link-Cqs2DhxR.js.map +1 -0
  202. package/dist/{list-ClDb8qiX.js → list-CjcNH-5e.js} +10 -7
  203. package/dist/list-CjcNH-5e.js.map +1 -0
  204. package/dist/{live-QFILMmql.js → live-BRF0ogR6.js} +3 -5
  205. package/dist/live-BRF0ogR6.js.map +1 -0
  206. package/dist/live-DTpM9b4D.js +56 -0
  207. package/dist/live-DTpM9b4D.js.map +1 -0
  208. package/dist/{machine-service-xFhakTs5.js → machine-service-BetvBWHz.js} +146 -13
  209. package/dist/machine-service-BetvBWHz.js.map +1 -0
  210. package/dist/{machine-service-DpJScNtA.js → machine-service-XFymGCHj.js} +1 -1
  211. package/dist/{machines-Cm-rU7jp.js → machines-CkDTbMZ6.js} +25 -25
  212. package/dist/{machines-Cm-rU7jp.js.map → machines-CkDTbMZ6.js.map} +1 -1
  213. package/dist/markdown-COts-jxf.js +653 -0
  214. package/dist/markdown-COts-jxf.js.map +1 -0
  215. package/dist/{model-Rmv-L9Jb.js → model-BYHXnYLe.js} +1 -1
  216. package/dist/model-CrsA2Evf.js +1 -0
  217. package/dist/{model-CcvSrrq9.js → model-DNC53cPL.js} +4 -4
  218. package/dist/{model-CcvSrrq9.js.map → model-DNC53cPL.js.map} +1 -1
  219. package/dist/{model-D9WMMEWD.js → model-F4UH_YG6.js} +4 -4
  220. package/dist/{model-D9WMMEWD.js.map → model-F4UH_YG6.js.map} +1 -1
  221. package/dist/{model-mTaaEkeQ.js → model-xemZQ7DK.js} +251 -65
  222. package/dist/model-xemZQ7DK.js.map +1 -0
  223. package/dist/{node-Bl2CtSjB.js → node-gPT36n_n.js} +19 -8
  224. package/dist/{node-Bl2CtSjB.js.map → node-gPT36n_n.js.map} +1 -1
  225. package/dist/options-CAZfxE0h.js +1 -0
  226. package/dist/otlp-BxPnO99e.js +1 -0
  227. package/dist/otlp-DbNz-occ.js +25 -0
  228. package/dist/otlp-DbNz-occ.js.map +1 -0
  229. package/dist/outcome-1LBk2e3w.js +38 -0
  230. package/dist/outcome-1LBk2e3w.js.map +1 -0
  231. package/dist/{patch-DhJDHRFs.js → patch-Dxzv8msl.js} +3 -3
  232. package/dist/{patch-DhJDHRFs.js.map → patch-Dxzv8msl.js.map} +1 -1
  233. package/dist/{patch-C7yBztZj.js → patch-uwo6t4Zy.js} +26 -2
  234. package/dist/patch-uwo6t4Zy.js.map +1 -0
  235. package/dist/{pi-providers-HygmL1y-.js → pi-providers-BjcQv2A1.js} +23 -9
  236. package/dist/{pi-providers-HygmL1y-.js.map → pi-providers-BjcQv2A1.js.map} +1 -1
  237. package/dist/{plain-Ctn8oUdp.js → plain-DXUNwqpO.js} +2 -2
  238. package/dist/{plain-Ctn8oUdp.js.map → plain-DXUNwqpO.js.map} +1 -1
  239. package/dist/{plan-CC0bKbm9.js → plan-DSJX7Oyq.js} +9 -9
  240. package/dist/{plan-CC0bKbm9.js.map → plan-DSJX7Oyq.js.map} +1 -1
  241. package/dist/{plugins-PwEtW8GH.js → plugins-DVEOMsYw.js} +7 -7
  242. package/dist/{plugins-PwEtW8GH.js.map → plugins-DVEOMsYw.js.map} +1 -1
  243. package/dist/{preflight-BHUEfA4m.js → preflight-B3oLZttl.js} +4 -4
  244. package/dist/{preflight-BHUEfA4m.js.map → preflight-B3oLZttl.js.map} +1 -1
  245. package/dist/preflight-CLYRuQ3r.js +1 -0
  246. package/dist/{preflight-DEV_vrdB.js → preflight-DXMD6jou.js} +1 -1
  247. package/dist/{preflight-BzvLYQ70.js → preflight-dUe0QiPK.js} +9 -9
  248. package/dist/{preflight-BzvLYQ70.js.map → preflight-dUe0QiPK.js.map} +1 -1
  249. package/dist/{prepare-D7hPrh7Q.js → prepare-BTaX7vcC.js} +8 -8
  250. package/dist/{prepare-D7hPrh7Q.js.map → prepare-BTaX7vcC.js.map} +1 -1
  251. package/dist/{prepare-DS42k409.js → prepare-QdiYPOwR.js} +2 -2
  252. package/dist/prices-CfnkJOth.js +1 -0
  253. package/dist/{prices-DXJsP2Ha.js → prices-CoLymMRd.js} +3 -3
  254. package/dist/{prices-DXJsP2Ha.js.map → prices-CoLymMRd.js.map} +1 -1
  255. package/dist/{process-RPlEhpHz.js → process-C_nh1H7X.js} +9 -3
  256. package/dist/process-C_nh1H7X.js.map +1 -0
  257. package/dist/provenance-Cylt_YxT.d.ts +222 -0
  258. package/dist/{provenance-DgQJ-dq7.js → provenance-Di42xQIS.js} +5 -5
  259. package/dist/{provenance-DgQJ-dq7.js.map → provenance-Di42xQIS.js.map} +1 -1
  260. package/dist/provider-B24hZX95.js +1 -0
  261. package/dist/{provider-ieWOn4YS.js → provider-Cn7lmECy.js} +2 -2
  262. package/dist/{provider-ieWOn4YS.js.map → provider-Cn7lmECy.js.map} +1 -1
  263. package/dist/{provider-stop-CxzWW7Ft.js → provider-stop-DjBMfZW5.js} +2 -2
  264. package/dist/{provider-stop-CxzWW7Ft.js.map → provider-stop-DjBMfZW5.js.map} +1 -1
  265. package/dist/{providers-N4HDMeeo.js → providers-Bm5NLaRh.js} +1 -1
  266. package/dist/{providers-B5Vr0hKH.js → providers-CTTfYbm0.js} +28 -28
  267. package/dist/{providers-B5Vr0hKH.js.map → providers-CTTfYbm0.js.map} +1 -1
  268. package/dist/query-CTMdWXbx.js +1 -0
  269. package/dist/{query-CmTjYbqh.js → query-TzNtTvwf.js} +2 -2
  270. package/dist/{query-CmTjYbqh.js.map → query-TzNtTvwf.js.map} +1 -1
  271. package/dist/{questions-DnqkJga0.js → questions-BqisU7A0.js} +19 -13
  272. package/dist/questions-BqisU7A0.js.map +1 -0
  273. package/dist/{recipe-B0NeT_Sn.js → recipe-DzjXne35.js} +4 -4
  274. package/dist/{recipe-B0NeT_Sn.js.map → recipe-DzjXne35.js.map} +1 -1
  275. package/dist/{record-2liAelrB.js → record-Cj3yWR0E.js} +43 -12
  276. package/dist/record-Cj3yWR0E.js.map +1 -0
  277. package/dist/{redact-DMueYS-v.js → redact-CfTHYWLv.js} +30 -1
  278. package/dist/redact-CfTHYWLv.js.map +1 -0
  279. package/dist/{remarks-D5qOZzcX.js → remarks-BPKGorBW.js} +4 -2
  280. package/dist/remarks-BPKGorBW.js.map +1 -0
  281. package/dist/{remarks-D-9f1y1B.js → remarks-CYXYiB_O.js} +21 -11
  282. package/dist/{remarks-D-9f1y1B.js.map → remarks-CYXYiB_O.js.map} +1 -1
  283. package/dist/{remote-B7BIRUTK.js → remote-D4UEa7Rs.js} +10 -8
  284. package/dist/{remote-B7BIRUTK.js.map → remote-D4UEa7Rs.js.map} +1 -1
  285. package/dist/renderer/ascii.d.ts +6 -0
  286. package/dist/renderer/ascii.js +7 -0
  287. package/dist/renderer/ascii.js.map +1 -0
  288. package/dist/renderer/markdown.d.ts +31 -0
  289. package/dist/renderer/markdown.js +1 -0
  290. package/dist/report-BSi-R8hC.js +941 -0
  291. package/dist/report-BSi-R8hC.js.map +1 -0
  292. package/dist/report-inputs-5XwZBIRK.js +113 -0
  293. package/dist/report-inputs-5XwZBIRK.js.map +1 -0
  294. package/dist/{resolve-B60w6DcB.js → resolve-BEltmYhD.js} +6 -5
  295. package/dist/resolve-BEltmYhD.js.map +1 -0
  296. package/dist/resolve-C7A5Q2Jj.js +1 -0
  297. package/dist/{resolve-t9sOkGGS.js → resolve-nRzGt4Ff.js} +30 -13
  298. package/dist/resolve-nRzGt4Ff.js.map +1 -0
  299. package/dist/{results-Df910ONm.js → results-DXBaxfIM.js} +11 -7
  300. package/dist/{results-Df910ONm.js.map → results-DXBaxfIM.js.map} +1 -1
  301. package/dist/{resume-4Q9q8AcE.js → resume-DZPjjToJ.js} +46 -21
  302. package/dist/resume-DZPjjToJ.js.map +1 -0
  303. package/dist/{retry-C15f5yyy.js → retry-rLdXCinf.js} +5 -5
  304. package/dist/{retry-C15f5yyy.js.map → retry-rLdXCinf.js.map} +1 -1
  305. package/dist/{root-Cm93-6ix.js → root-Co0jhOT5.js} +2 -2
  306. package/dist/{root-Cm93-6ix.js.map → root-Co0jhOT5.js.map} +1 -1
  307. package/dist/run-D_o_8Fkv.js +128 -0
  308. package/dist/run-D_o_8Fkv.js.map +1 -0
  309. package/dist/{client-kUwtq-vt.js → runtime-DsRjnydE.js} +525 -105
  310. package/dist/runtime-DsRjnydE.js.map +1 -0
  311. package/dist/sandbox/boat.d.ts +1 -1
  312. package/dist/sandbox/boat.js +2 -2
  313. package/dist/sandbox/daytona.d.ts +1 -1
  314. package/dist/sandbox/daytona.js +1 -1
  315. package/dist/{sandbox-BffNervX.js → sandbox-B8gDRs5E.js} +3 -3
  316. package/dist/{sandbox-BffNervX.js.map → sandbox-B8gDRs5E.js.map} +1 -1
  317. package/dist/sandbox.d.ts +1 -1
  318. package/dist/sandbox.js +1 -1
  319. package/dist/{secrets-file-Dtd3h3kY.js → secrets-file-B8rUJV2L.js} +4 -4
  320. package/dist/{secrets-file-Dtd3h3kY.js.map → secrets-file-B8rUJV2L.js.map} +1 -1
  321. package/dist/server-eUWbjZoi.js +1 -0
  322. package/dist/server-mJzPXBMM.js +204 -0
  323. package/dist/server-mJzPXBMM.js.map +1 -0
  324. package/dist/{services-BG8xuU82.js → services-CHh83UDC.js} +4 -2
  325. package/dist/services-CHh83UDC.js.map +1 -0
  326. package/dist/{shared-DtZwIAfy.js → shared-bXaqVxk-.js} +5 -34
  327. package/dist/shared-bXaqVxk-.js.map +1 -0
  328. package/dist/skill-dirs-B7o83G5I.js +63 -0
  329. package/dist/skill-dirs-B7o83G5I.js.map +1 -0
  330. package/dist/source-C6JgdJdJ.js +1 -0
  331. package/dist/{source-zoRWRMql.js → source-DxgCge_u.js} +33 -18
  332. package/dist/source-DxgCge_u.js.map +1 -0
  333. package/dist/spans-B4MwgEyy.js +1 -0
  334. package/dist/spans-Dq3fo6HM.js +6 -0
  335. package/dist/spans-Dq3fo6HM.js.map +1 -0
  336. package/dist/spans-tf3biLpT.js +72 -0
  337. package/dist/spans-tf3biLpT.js.map +1 -0
  338. package/dist/src-B2psni1B.js +1 -0
  339. package/dist/{src-DiS13D_q.js → src-BI35cTig.js} +2 -2
  340. package/dist/src-BLePHsjw.js +2 -0
  341. package/dist/{src-Bj3WJs3t.js → src-CftydE9Q.js} +2 -2
  342. package/dist/{src-CDoAEO4u.js → src-ClMhIgi3.js} +2 -2
  343. package/dist/{src-CDoAEO4u.js.map → src-ClMhIgi3.js.map} +1 -1
  344. package/dist/{src-CUrcaZ91.js → src-CnOj4_n4.js} +1 -1
  345. package/dist/src-Cs3nmbrr.js +2 -0
  346. package/dist/{src-TMKittwH.js → src-D0GezGkJ.js} +2 -2
  347. package/dist/{src-OC89dAjO.js → src-D1Q6ElaF.js} +182 -36
  348. package/dist/src-D1Q6ElaF.js.map +1 -0
  349. package/dist/{src-CFBg2ecq.js → src-D_9qqp-Y.js} +2 -2
  350. package/dist/{src-DwcTn4JY.js → src-DhZuOcgq.js} +5 -5
  351. package/dist/{src-DwcTn4JY.js.map → src-DhZuOcgq.js.map} +1 -1
  352. package/dist/{src-CGBwMZgM.js → src-Dm2YIeXn.js} +8 -8
  353. package/dist/{src-CGBwMZgM.js.map → src-Dm2YIeXn.js.map} +1 -1
  354. package/dist/{src-CcbV6VpV.js → src-Dpzfe-vl.js} +3 -3
  355. package/dist/{src-c3ZQu7Uu.js → src-NYxoDS3b.js} +5 -5
  356. package/dist/{src-c3ZQu7Uu.js.map → src-NYxoDS3b.js.map} +1 -1
  357. package/dist/{src-C3OhA5pw.js → src-Y4AiZxLt.js} +5 -5
  358. package/dist/{src-C3OhA5pw.js.map → src-Y4AiZxLt.js.map} +1 -1
  359. package/dist/{src-BQWAxhKD.js → src-YpmRtJmP.js} +3 -3
  360. package/dist/{src-BQWAxhKD.js.map → src-YpmRtJmP.js.map} +1 -1
  361. package/dist/{src-Bb3k9Kn_.js → src-nIPDQ297.js} +66 -11
  362. package/dist/{src-Bb3k9Kn_.js.map → src-nIPDQ297.js.map} +1 -1
  363. package/dist/ssh-pool-B9bnSQIj.js +1 -0
  364. package/dist/{ssh-pool-DwboLN9O.js → ssh-pool-Gv8gAsgZ.js} +19 -16
  365. package/dist/ssh-pool-Gv8gAsgZ.js.map +1 -0
  366. package/dist/{start-BwMyDneA.js → start-YOYgbaNU.js} +45 -13
  367. package/dist/start-YOYgbaNU.js.map +1 -0
  368. package/dist/state-Bvsi2WgZ.js +1 -0
  369. package/dist/{state-CXTaT0t8.js → state-CJlYQtdl.js} +21 -38
  370. package/dist/state-CJlYQtdl.js.map +1 -0
  371. package/dist/store/local.d.ts +1 -1
  372. package/dist/store/s3.d.ts +1 -1
  373. package/dist/studio-share-4T6GYO7l.js +209 -0
  374. package/dist/studio-share-4T6GYO7l.js.map +1 -0
  375. package/dist/suite-fO-u8jlS.js +2 -0
  376. package/dist/{suite-C4XSg6Ls.js → suite-oH7uiCRq.js} +107 -106
  377. package/dist/suite-oH7uiCRq.js.map +1 -0
  378. package/dist/summary-CTMWfnOF.js +75 -0
  379. package/dist/summary-CTMWfnOF.js.map +1 -0
  380. package/dist/{sync-bnI6WWaL.js → sync-BNduS38K.js} +6 -6
  381. package/dist/{sync-bnI6WWaL.js.map → sync-BNduS38K.js.map} +1 -1
  382. package/dist/{tart-CSP5Q0qV.js → tart-B6Uz4S3I.js} +43 -13
  383. package/dist/tart-B6Uz4S3I.js.map +1 -0
  384. package/dist/{tart-DmVuVsmX.js → tart-DNrumWZu.js} +1 -1
  385. package/dist/{tart-CzViW3WZ.js → tart-Duqp33Zh.js} +2 -2
  386. package/dist/{tart-CzViW3WZ.js.map → tart-Duqp33Zh.js.map} +1 -1
  387. package/dist/{tart-bake-d9UvIfe3.js → tart-bake-B4QyNb7f.js} +1 -1
  388. package/dist/{tart-bake-BCKRfyKm.js → tart-bake-s99wBzxO.js} +9 -9
  389. package/dist/{tart-bake-BCKRfyKm.js.map → tart-bake-s99wBzxO.js.map} +1 -1
  390. package/dist/{tart-cli-CqSpUYGV.js → tart-cli-B0xVYBti.js} +6 -4
  391. package/dist/tart-cli-B0xVYBti.js.map +1 -0
  392. package/dist/{tart-cli-C4d5FXG1.js → tart-cli-DfteJCN9.js} +1 -1
  393. package/dist/{tart-hosts-Du7SZM4w.js → tart-hosts-B3Ze8aMp.js} +1 -1
  394. package/dist/{tart-hosts-G0EUDt07.js → tart-hosts-DfVUcZqA.js} +6 -6
  395. package/dist/{tart-hosts-G0EUDt07.js.map → tart-hosts-DfVUcZqA.js.map} +1 -1
  396. package/dist/{tart-images-De0-ib7R.js → tart-images-yzIyb0bQ.js} +8 -3
  397. package/dist/tart-images-yzIyb0bQ.js.map +1 -0
  398. package/dist/telemetry-bBhK3DSs.js +399 -0
  399. package/dist/telemetry-bBhK3DSs.js.map +1 -0
  400. package/dist/testing.d.ts +11 -2
  401. package/dist/testing.js +4 -4
  402. package/dist/{tool-class-B3ZFTSOw.js → tool-class-CB-m4PiF.js} +2 -2
  403. package/dist/{tool-class-B3ZFTSOw.js.map → tool-class-CB-m4PiF.js.map} +1 -1
  404. package/dist/{trace-CfppUzD9.js → trace-CJQLlAbB.js} +2 -2
  405. package/dist/{trace-CfppUzD9.js.map → trace-CJQLlAbB.js.map} +1 -1
  406. package/dist/{traits-Bj1PM5F7.js → traits-DPIffsN8.js} +2 -2
  407. package/dist/{traits-Bj1PM5F7.js.map → traits-DPIffsN8.js.map} +1 -1
  408. package/dist/{transcript-CW02A4s9.js → transcript-TWiaAuoL.js} +2 -2
  409. package/dist/{transcript-CW02A4s9.js.map → transcript-TWiaAuoL.js.map} +1 -1
  410. package/dist/{resources-DqdwihhT.js → turn-error-Dmv_Muj7.js} +15 -2
  411. package/dist/turn-error-Dmv_Muj7.js.map +1 -0
  412. package/dist/{types-D4WEnzhY.d.ts → types-BZaItiui.d.ts} +159 -13
  413. package/dist/types-Buut5yGe.js +29 -0
  414. package/dist/{types-CbcnJxCy.js.map → types-Buut5yGe.js.map} +1 -1
  415. package/dist/types-C48zuZbT.d.ts +156 -0
  416. package/dist/types-CYz4dqAd.d.ts +1366 -0
  417. package/dist/{ui-DfvPNmPP.js → ui-BIYUGm8D.js} +21 -17
  418. package/dist/ui-BIYUGm8D.js.map +1 -0
  419. package/dist/{vertex-BLxYasfw.js → vertex-DQQGhUn7.js} +2 -2
  420. package/dist/{vertex-BLxYasfw.js.map → vertex-DQQGhUn7.js.map} +1 -1
  421. package/dist/vertex-DtWe0Xf6.js +1 -0
  422. package/dist/view/index.d.ts +142 -809
  423. package/dist/view/index.js +17 -12
  424. package/dist/view/report/index.d.ts +217 -0
  425. package/dist/view/report/index.js +4 -0
  426. package/dist/views-DOYgIWRs.js +2 -0
  427. package/dist/{views-ChoSQxfF.js → views-DPsEJBBP.js} +6 -6
  428. package/dist/{views-ChoSQxfF.js.map → views-DPsEJBBP.js.map} +1 -1
  429. package/dist/{vm-container-DOtIi_rd.js → vm-container-rHUmgBRf.js} +13 -11
  430. package/dist/vm-container-rHUmgBRf.js.map +1 -0
  431. package/dist/{vm-images-CHutQVml.js → vm-images-BC2ZjZl5.js} +19 -21
  432. package/dist/vm-images-BC2ZjZl5.js.map +1 -0
  433. package/dist/wait-CixM6W0G.js +65 -0
  434. package/dist/wait-CixM6W0G.js.map +1 -0
  435. package/dist/{worker-entry-CKBfEt0h.js → worker-entry-CN1eVLzz.js} +7 -7
  436. package/dist/{worker-entry-CKBfEt0h.js.map → worker-entry-CN1eVLzz.js.map} +1 -1
  437. package/dist/{workspace-Cip2dVwU.js → workspace-BS-FUp5I.js} +9 -9
  438. package/dist/{workspace-Cip2dVwU.js.map → workspace-BS-FUp5I.js.map} +1 -1
  439. package/dist/workspace-D8HqjfsQ.js +1 -0
  440. package/package.json +38 -1
  441. package/skills/reeval/references/cli.md +1 -1
  442. package/skills/reeval/references/config.md +1 -1
  443. package/skills/reeval/references/environments.md +1 -1
  444. package/dist/agents-CEapsHV-.js.map +0 -1
  445. package/dist/analysis-BnG_1x_O.js.map +0 -1
  446. package/dist/app-files-Dsr3U1Q7.js +0 -10
  447. package/dist/app-files-Dsr3U1Q7.js.map +0 -1
  448. package/dist/assemble-DUtbByvj.js +0 -1
  449. package/dist/assemble-Dw1bF13I.js.map +0 -1
  450. package/dist/attach-DvYiMu6w.js.map +0 -1
  451. package/dist/attempt-host-DgeJFSZw.js.map +0 -1
  452. package/dist/boat-BBPOc_bs.js +0 -1
  453. package/dist/boat-CUNgSc75.js.map +0 -1
  454. package/dist/boat-images-BuFigiEb.js.map +0 -1
  455. package/dist/bundle-yEgI3RsI.js.map +0 -1
  456. package/dist/capture-DxxB3d7p.js +0 -1
  457. package/dist/cases-B7QkvMGE.js.map +0 -1
  458. package/dist/cases-CA7gxxdr.js +0 -1
  459. package/dist/cli-Z7aw8k25.js.map +0 -1
  460. package/dist/cli-spec-3zvLC9YW.js.map +0 -1
  461. package/dist/client-kUwtq-vt.js.map +0 -1
  462. package/dist/cloud-DuHcNQVV.js.map +0 -1
  463. package/dist/cloud-run-C08P_Uap.js +0 -1
  464. package/dist/cloud-run-DqMkSGYB.js.map +0 -1
  465. package/dist/cloud-secrets-Bak83tik.js.map +0 -1
  466. package/dist/command-Bs6Knfm4.js +0 -298
  467. package/dist/command-Bs6Knfm4.js.map +0 -1
  468. package/dist/command-CeygdoVe.js.map +0 -1
  469. package/dist/command-CoHN1ibe.js.map +0 -1
  470. package/dist/command-CtmSJzPC.js.map +0 -1
  471. package/dist/command-efTGfUYb.js.map +0 -1
  472. package/dist/commands-B_cAs8nG.js +0 -1
  473. package/dist/commands-DR-NolV2.js.map +0 -1
  474. package/dist/config-Bn2OhCN_.js.map +0 -1
  475. package/dist/context-Ba7PNMMe.js +0 -120
  476. package/dist/context-Ba7PNMMe.js.map +0 -1
  477. package/dist/daytona-CWs9j9mt.js.map +0 -1
  478. package/dist/daytona-image-BjVpi3KS.js +0 -1
  479. package/dist/db-Bjq6qr6c.js +0 -2
  480. package/dist/db-D6yQib7M.js.map +0 -1
  481. package/dist/debug-CF4p4Lp_.js +0 -1
  482. package/dist/debug-CIXzB4wK.js.map +0 -1
  483. package/dist/environment-BY8lVFFK.js.map +0 -1
  484. package/dist/evaluate-B8I9AKgw.js.map +0 -1
  485. package/dist/events-D97ZzAeq.js.map +0 -1
  486. package/dist/files-5HsFIpxw.js +0 -1
  487. package/dist/files-BNg3wgw_.js +0 -1
  488. package/dist/files-BesexM88.js.map +0 -1
  489. package/dist/files-DNgQwgA0.js +0 -1
  490. package/dist/files-DvDUSrWL.js +0 -1
  491. package/dist/files-TwSVeWwM.js.map +0 -1
  492. package/dist/gate-CgYdaprG.js +0 -49
  493. package/dist/gate-CgYdaprG.js.map +0 -1
  494. package/dist/harness-FcTyyPco.js.map +0 -1
  495. package/dist/ids-DqEvrBFZ.js.map +0 -1
  496. package/dist/init-CS5kEaVg.js +0 -1
  497. package/dist/ink-list-CLq6YVB8.js.map +0 -1
  498. package/dist/inputs-load-D4IsEjq3.js +0 -1
  499. package/dist/issue-DmSorsTF.js +0 -1
  500. package/dist/list-ClDb8qiX.js.map +0 -1
  501. package/dist/live-C7iGaDCA.js +0 -120
  502. package/dist/live-C7iGaDCA.js.map +0 -1
  503. package/dist/live-C7kN5v2P.js +0 -1
  504. package/dist/live-QFILMmql.js.map +0 -1
  505. package/dist/machine-service-xFhakTs5.js.map +0 -1
  506. package/dist/model-Colu5mEp.js +0 -1
  507. package/dist/model-mTaaEkeQ.js.map +0 -1
  508. package/dist/options-CmRy82gl.js +0 -1
  509. package/dist/pairs-BGb134Vu.js +0 -230
  510. package/dist/pairs-BGb134Vu.js.map +0 -1
  511. package/dist/patch-C7yBztZj.js.map +0 -1
  512. package/dist/paths-Dwmi7L4X.js +0 -75
  513. package/dist/paths-Dwmi7L4X.js.map +0 -1
  514. package/dist/preflight-xa8_eGOW.js +0 -1
  515. package/dist/prices-C_mbZa5a.js +0 -1
  516. package/dist/process-RPlEhpHz.js.map +0 -1
  517. package/dist/provider-DHT20Gl8.js +0 -1
  518. package/dist/query-BXV6pRyI.js +0 -1
  519. package/dist/questions-DnqkJga0.js.map +0 -1
  520. package/dist/record-2liAelrB.js.map +0 -1
  521. package/dist/redact-DMueYS-v.js.map +0 -1
  522. package/dist/remarks-D5qOZzcX.js.map +0 -1
  523. package/dist/resolve-B60w6DcB.js.map +0 -1
  524. package/dist/resolve-CCLiJX4i.js +0 -1
  525. package/dist/resolve-t9sOkGGS.js.map +0 -1
  526. package/dist/resources-DqdwihhT.js.map +0 -1
  527. package/dist/resume-4Q9q8AcE.js.map +0 -1
  528. package/dist/runtime-CIsljVBy.js +0 -457
  529. package/dist/runtime-CIsljVBy.js.map +0 -1
  530. package/dist/server-CeYUmuuu.js +0 -1
  531. package/dist/server-fSNuT3Z4.js +0 -284
  532. package/dist/server-fSNuT3Z4.js.map +0 -1
  533. package/dist/services-BG8xuU82.js.map +0 -1
  534. package/dist/shared-DtZwIAfy.js.map +0 -1
  535. package/dist/slug-IHh3f770.js.map +0 -1
  536. package/dist/source-n7nUVm9p.js +0 -1
  537. package/dist/source-zoRWRMql.js.map +0 -1
  538. package/dist/src-9V3C6A7C.js +0 -2
  539. package/dist/src-CB5cZhaN.js +0 -2
  540. package/dist/src-DB0vml0O.js +0 -1
  541. package/dist/src-OC89dAjO.js.map +0 -1
  542. package/dist/ssh-pool-ChpT-fDU.js +0 -1
  543. package/dist/ssh-pool-DwboLN9O.js.map +0 -1
  544. package/dist/start-BwMyDneA.js.map +0 -1
  545. package/dist/state-CXTaT0t8.js.map +0 -1
  546. package/dist/state-CltRWlCu.js +0 -1
  547. package/dist/suite-C4XSg6Ls.js.map +0 -1
  548. package/dist/suite-CIsyPqPA.js +0 -2
  549. package/dist/summary-DeeQJAR5.js +0 -169
  550. package/dist/summary-DeeQJAR5.js.map +0 -1
  551. package/dist/tart-CSP5Q0qV.js.map +0 -1
  552. package/dist/tart-cli-CqSpUYGV.js.map +0 -1
  553. package/dist/tart-images-De0-ib7R.js.map +0 -1
  554. package/dist/types-CbcnJxCy.js +0 -15
  555. package/dist/ui-DfvPNmPP.js.map +0 -1
  556. package/dist/vertex-DZDUpp0W.js +0 -1
  557. package/dist/views-DU0PM0UW.js +0 -2
  558. package/dist/vm-container-DOtIi_rd.js.map +0 -1
  559. package/dist/vm-images-CHutQVml.js.map +0 -1
  560. package/dist/workspace-BKcNJAPb.js +0 -1
@@ -0,0 +1,1366 @@
1
+ import { Fn as CellRecord, Gn as HistoryEntry, Jt as Outcome, Kn as JudgmentRecord, L as CompareCut, Ln as EvalEntry, Mn as ArtifactLevel, Pn as AttemptRecord, Qn as StackSummary, Wn as HistoryCell, Xn as RunRecord, Zn as RunScope, bn as TokenUsage, er as CellEffort, ir as EffortContrast, qn as Lift, wn as TurnStatus, xn as TranscriptEvent, zn as FamilyLiftSummary } from "./types-BZaItiui.js";
2
+ /** Bumped when the system prompt or output schema changes: an older analysis is superseded. */
3
+ declare const ANALYSIS_PROMPT_VERSION = "analysis-v14";
4
+ /** Bumped when the digest's shape or selection rules change. */
5
+ declare const DIGEST_VERSION = 15;
6
+ type Confidence = 'high' | 'medium' | 'low';
7
+ type AnalysisClaimRole = 'finding' | 'caveat' | 'action';
8
+ type AnalysisClaim = {
9
+ text: string;
10
+ /** Fact (`F3`), evidence (`E2`, `E2:L39`), attempt (`r2/scaffold/reeval-good#1`) or check ids. */
11
+ cites: string[];
12
+ confidence: Confidence;
13
+ /** Exact spans from the cited evidence; a claim whose quote isn't there is dropped. */
14
+ quotes?: string[];
15
+ /** Numbers in the text that no cited fact or evidence contains. */
16
+ unverifiedNumbers?: string[];
17
+ /** Eval id this claim is about, when resolved from attempt cites; optional on older docs. */
18
+ evalId?: string;
19
+ /** Section role; optional on older docs (derived from which body array it sits in). */
20
+ role?: AnalysisClaimRole;
21
+ /** Attempt cites resolved to ids and paths; optional on older docs. */
22
+ attempts?: Array<{
23
+ attemptId: string;
24
+ path: string;
25
+ }>;
26
+ };
27
+ type AnomalyKind = 'reeval-bug' | 'eval-design' | 'judge' | 'noise' | 'other';
28
+ type AnalysisAnomaly = {
29
+ text: string;
30
+ kind: AnomalyKind;
31
+ cites: string[];
32
+ /** Numbers in the text that no cited fact or evidence contains. */
33
+ unverifiedNumbers?: string[];
34
+ evalId?: string;
35
+ role?: AnalysisClaimRole;
36
+ attempts?: Array<{
37
+ attemptId: string;
38
+ path: string;
39
+ }>;
40
+ };
41
+ type ActionKind = 'skill' | 'docs' | 'eval' | 'judge' | 'reeval' | 'harness';
42
+ type Level = 'high' | 'medium' | 'low';
43
+ type AnalysisAction = {
44
+ title: string;
45
+ target: string;
46
+ kind: ActionKind;
47
+ change: string;
48
+ expectedEffect: string;
49
+ /** An `reeval` command that shows the effect; null when the model's didn't parse as reeval's CLI. */
50
+ verify: string | null;
51
+ /** Why the model's verify command was rejected. */
52
+ verifyProblem?: string;
53
+ impact: Level;
54
+ effort: Level;
55
+ cites: string[];
56
+ evalId?: string;
57
+ role?: AnalysisClaimRole;
58
+ attempts?: Array<{
59
+ attemptId: string;
60
+ path: string;
61
+ }>;
62
+ };
63
+ type AnalysisBody = {
64
+ tldr: string;
65
+ whatHappened: AnalysisClaim[];
66
+ results: AnalysisClaim[];
67
+ why: AnalysisClaim[];
68
+ anomalies: AnalysisAnomaly[];
69
+ actions: AnalysisAction[];
70
+ confidence: {
71
+ overall: Confidence;
72
+ why: string;
73
+ };
74
+ };
75
+ type AnalysisValidation = {
76
+ /** Citations checked, and those that named nothing in the digest. */
77
+ citations: number;
78
+ invalid: string[];
79
+ /** Claims, anomalies and actions dropped for a bad citation, invented quote, wrong judge score, or rejected verify. */
80
+ dropped: number;
81
+ commands: number;
82
+ invalidCommands: string[];
83
+ unverifiedNumbers: number;
84
+ };
85
+ type AnalysisScope = {
86
+ kind: 'run';
87
+ head: string;
88
+ /** Resolved once, so a nearer ancestor recorded later doesn't change what this analyzed. */
89
+ base: string | null;
90
+ /** An eval filter (substring of the eval id), or null for every eval. */
91
+ evals: string | null;
92
+ };
93
+ type AnalysisDoc = {
94
+ kind: 'reeval-analysis';
95
+ schemaVersion: 1;
96
+ scope: AnalysisScope;
97
+ /** The run the analysis belongs to: the scope's head. */
98
+ anchor: string;
99
+ inputs: {
100
+ records: Array<{
101
+ runId: string;
102
+ sha256: string;
103
+ }>;
104
+ digestVersion: number;
105
+ digestHash: string;
106
+ };
107
+ promptVersion: string;
108
+ model: string;
109
+ resolvedModel: string | null;
110
+ route: 'direct' | 'gateway' | 'openrouter' | 'vertex' | 'fake';
111
+ usage: {
112
+ inputTokens: number | null;
113
+ outputTokens: number | null;
114
+ };
115
+ latencyMs: number;
116
+ createdAt: string;
117
+ author: string;
118
+ validation: AnalysisValidation;
119
+ analysis: AnalysisBody;
120
+ /** Short attempt ids used in citations, and where each attempt is. */
121
+ attempts: Record<string, {
122
+ runId: string;
123
+ evalSlug: string;
124
+ attempt: string;
125
+ }>;
126
+ /** Each cited fact or evidence id, summarized in one line. */
127
+ cited: Record<string, string>;
128
+ /**
129
+ * Why the analysis failed; its `analysis` is then empty. `reeval analyze` treats a cached one as
130
+ * a miss and calls the model again.
131
+ */
132
+ error?: string;
133
+ };
134
+ /**
135
+ * A note on one eval and harness of a run, which the studio shows:
136
+ * `.reeval/analysis/<runId>/notes/<eval slug>__<harness slug>.json`.
137
+ */
138
+ type AnalysisNote = {
139
+ kind: 'reeval-analysis-note';
140
+ schemaVersion: 1;
141
+ runId: string;
142
+ evalId: string;
143
+ /** The harness id, as in the cell key. */
144
+ harness: string;
145
+ model: string;
146
+ resolvedModel: string | null;
147
+ createdAt: string;
148
+ /** The attempt ids (`<cell key>#<n>`) the note covers. */
149
+ attempts: string[];
150
+ /** Empty when the call failed. */
151
+ text: string;
152
+ /** Why the call failed; the run's outcomes don't change. */
153
+ error?: string;
154
+ usage: {
155
+ inputTokens: number | null;
156
+ outputTokens: number | null;
157
+ };
158
+ };
159
+ /** One analysis file the studio can show: saved in the run's folder, or in the local cache. */
160
+ type AnalysisEntry = {
161
+ file: string;
162
+ source: 'saved' | 'cached';
163
+ doc: AnalysisDoc;
164
+ };
165
+ type AnalysisIndex = {
166
+ runId: string;
167
+ entries: AnalysisEntry[];
168
+ /** The run's notes (`.reeval/analysis/<run>/notes/`), by eval and harness; absent when there are none. */
169
+ notes?: AnalysisNote[];
170
+ };
171
+ /** The newest analysis that wrote a TL;DR, for the studio run header. */
172
+ declare function pinnedAnalysis(entries: readonly AnalysisEntry[]): AnalysisEntry | undefined;
173
+ /** Newest first, saved before cached at the same time. */
174
+ declare function sortAnalyses(entries: AnalysisEntry[]): AnalysisEntry[];
175
+ /** Superseded by a newer prompt or digest version: the studio says so. */
176
+ declare const isSuperseded: (doc: AnalysisDoc) => boolean;
177
+ type ToolClass = 'context' | 'dependency' | 'web' | 'verify' | 'workspace';
178
+ /**
179
+ * Where a used skill came from: `placed` by the environment's `skills` (or a plugin), in a
180
+ * `skill-dir` some agent reads (a setup step, the workspace or a package manager put it there),
181
+ * or a `SKILL.md` read `elsewhere`, such as a package cache.
182
+ */
183
+ type SkillSource = 'placed' | 'skill-dir' | 'elsewhere';
184
+ type CompareStatus = 'regressed' | 'fixed' | 'same' | 'gate' | 'drift' | 'added' | 'removed' | 'infra';
185
+ type CompareRow = {
186
+ id: string;
187
+ status: CompareStatus;
188
+ base: Outcome | null;
189
+ head: Outcome | null;
190
+ /**
191
+ * For `drift`: what changed, e.g. `body changed`, `revision 2 → 3` or `harness config changed`. On
192
+ * any other row, `gate changed` when the cells' gate did.
193
+ */
194
+ changes?: string[];
195
+ /**
196
+ * When the cells' gate changed but nothing else did: each side's gate, and base's outcome with
197
+ * its recorded attempts scored under head's gate. `status` compares that with head, so a flip
198
+ * the gate alone explains is `gate`, not `regressed` or `fixed`.
199
+ */
200
+ gate?: {
201
+ base: string;
202
+ head: string;
203
+ rescored: Outcome;
204
+ };
205
+ /**
206
+ * When the runs ran different cells of the eval: the cells only head or only base has. Then
207
+ * `status`, `base` and `head` are over the cells both have.
208
+ */
209
+ cells?: {
210
+ added: string[];
211
+ removed: string[];
212
+ };
213
+ /** Cells in both runs, with each side's lift (variants) and effort. */
214
+ perCell?: CompareCell[];
215
+ /**
216
+ * The sides whose `fail` is only attempts the run's interrupt canceled, which `reeval show`
217
+ * reads as `canceled`. Such an eval is never `regressed` or `fixed`: it's `drift`.
218
+ */
219
+ canceled?: Array<'base' | 'head'>;
220
+ /** An eval with `cases`: each case's row, its `id` the case key, in head's order then base's. */
221
+ cases?: CompareRow[];
222
+ /**
223
+ * An eval with `cases` when the runs ran different cases: the case keys only head or only base
224
+ * has. Then `base` and `head` are over the cases both have.
225
+ */
226
+ caseScope?: {
227
+ added: string[];
228
+ removed: string[];
229
+ };
230
+ /** An eval with `cases`: each variant's lift paired by case on each side, null where that run has none. */
231
+ lifts?: CompareFamilyLift[];
232
+ /** An eval with `cases`: each cell's rate over the cases that a `passRate` gates, on each side. */
233
+ rates?: CompareFamilyRate[];
234
+ /**
235
+ * `family` on a case row whose cell's rate over the cases gates it (`CaseRecord.verdict`):
236
+ * `base` and `head` are what it did, and the eval's row is the verdict, so
237
+ * `--fail-on-regression` leaves it to that row.
238
+ */
239
+ verdict?: 'family';
240
+ };
241
+ type CompareFamilyLift = {
242
+ key: string;
243
+ vs: string;
244
+ base: Omit<FamilyLiftSummary, 'variant' | 'vs'> | null;
245
+ head: Omit<FamilyLiftSummary, 'variant' | 'vs'> | null;
246
+ };
247
+ type CompareFamilyRate = {
248
+ key: string;
249
+ base: {
250
+ rate: number;
251
+ passRate: number;
252
+ outcome: 'pass' | 'fail';
253
+ } | null;
254
+ head: {
255
+ rate: number;
256
+ passRate: number;
257
+ outcome: 'pass' | 'fail';
258
+ } | null;
259
+ };
260
+ type CompareCell = {
261
+ key: string;
262
+ base: Pick<HistoryCell, 'lift' | 'effort' | 'pass' | 'fail'>;
263
+ head: Pick<HistoryCell, 'lift' | 'effort' | 'pass' | 'fail'>;
264
+ };
265
+ type StackChange = {
266
+ field: keyof StackSummary;
267
+ label: string;
268
+ base: string;
269
+ head: string;
270
+ /** Not covered by eval fingerprints, so outcome flips across it are drift, not regressions. */
271
+ outsideFingerprint: boolean;
272
+ };
273
+ type StackCell = Pick<HistoryCell, 'key' | 'harness' | 'environment' | 'environmentDigest' | 'fingerprint' | 'fingerprintParts'>;
274
+ /** A run's stack, and its cells when known: the per-cell fields then compare over shared cells. */
275
+ type StackSource = Pick<HistoryEntry, 'stack'> & {
276
+ evals?: ReadonlyArray<{
277
+ id: string;
278
+ cells: readonly StackCell[];
279
+ }>;
280
+ };
281
+ /**
282
+ * What changed in the stack between two runs. With both runs' cells, images, agent CLIs,
283
+ * harnesses and environments compare over the cells both have, so a run over fewer cells isn't
284
+ * drift. Reported served-judge lists only compare when both have one.
285
+ */
286
+ declare function stackChanges(base: StackSource, head: StackSource): StackChange[];
287
+ /**
288
+ * Whether any of these cells changed fingerprint since `seen` last had it, then records them.
289
+ * A cell `seen` doesn't have yet is new coverage, not a change.
290
+ */
291
+ declare function cellsChanged(seen: Map<string, string>, evalId: string, cells: ReadonlyArray<Pick<HistoryCell, 'key' | 'fingerprint'>>): boolean;
292
+ type HistoryEval = HistoryEntry['evals'][number];
293
+ type ComparableEval = Pick<HistoryEval, 'id' | 'fingerprint' | 'outcome' | 'sources' | 'source' | 'revision' | 'cells'> & Partial<Pick<HistoryEval, 'cases' | 'lifts' | 'rates'>> & {
294
+ verdict?: 'family';
295
+ };
296
+ type Comparable = Pick<HistoryEntry, 'stack'> & {
297
+ evals: ComparableEval[];
298
+ };
299
+ /** Which source files were added, removed or changed. */
300
+ declare function sourceChanges(before: ComparableEval['sources'], after: ComparableEval['sources']): string[];
301
+ /**
302
+ * Why two runs of an eval aren't comparable: what in its source differs (`body`, `options`,
303
+ * `case data`, `revision 2 → 3`; a record from before `source`, its files), then the fingerprint
304
+ * parts that differ. Each reason but the revision ends in ` changed`.
305
+ */
306
+ declare function whyChanged(base: ComparableEval, head: ComparableEval): string[];
307
+ /** An eval only head has, with each of its cases. */
308
+ declare const addedRow: (h: ComparableEval) => CompareRow;
309
+ /** The rows of a run with nothing to compare it with: every eval is new. */
310
+ declare const newRows: (head: Comparable) => CompareRow[];
311
+ /** Each eval's row, an eval with `cases` with a row for each of its cases. */
312
+ declare function compare(base: Comparable, head: Comparable): CompareRow[];
313
+ /** Short name for a judge question, e.g. `login` from `login?`. */
314
+ declare function judgeLabel(question: string): string;
315
+ /**
316
+ * One short label per key. Shared prefixes take the distinctive tail so
317
+ * `Does this change…` does not collapse several checks; a tail that still
318
+ * collides falls back to the key.
319
+ */
320
+ declare function uniqueJudgeLabels(items: readonly {
321
+ key: string;
322
+ question: string;
323
+ }[]): Map<string, string>;
324
+ type JudgeKeyRate = {
325
+ key: string;
326
+ /** The `expect` label when the judge had one, else a short name from its question. */
327
+ label: string;
328
+ /** Named by an `expect` label, so the cell's checks carry it too. */
329
+ labelled: boolean;
330
+ question: string;
331
+ pass: number;
332
+ counted: number;
333
+ /** Compact lift vs the paired control, when both sides counted an attempt. */
334
+ lift: Pick<Lift, 'points' | 'control' | 'variant' | 'withinNoise'> | null;
335
+ };
336
+ /** Per-key pass/counted for one cell's counted attempts, with lift on a variant when both sides counted an attempt. */
337
+ declare function cellJudgeRates(e: Pick<EvalEntry, 'attempts' | 'cells'>, cellKey: string): JudgeKeyRate[];
338
+ /**
339
+ * The judge rates a cell's summary line adds to its checks: a labeled judge whose label is one
340
+ * of the cell's checks against its control is already there.
341
+ */
342
+ declare function summaryJudgeRates(e: Pick<EvalEntry, 'attempts' | 'cells'>, cell: Pick<CellRecord, 'key' | 'checks'>): JudgeKeyRate[];
343
+ /** `login 4/5 · checkout 0/5`, with `(+80 pts)` on a key that has lift. */
344
+ declare function judgeRatesLabel(rates: readonly JudgeKeyRate[]): string;
345
+ type CollapsedJudge = {
346
+ verdict: 'pass' | 'fail';
347
+ score: number;
348
+ threshold: number;
349
+ label: string;
350
+ key: string;
351
+ };
352
+ /** Each key, or the lowest score plus how many judges there were. */
353
+ declare function collapsedJudges(judgments: readonly JudgmentRecord[]): {
354
+ shown: CollapsedJudge[];
355
+ extra: number;
356
+ };
357
+ /** `fail 0.78/0.80`, or `fail 0.33/0.80 · 3 judges` when there are many. */
358
+ declare function collapsedJudgeLabel(judgments: readonly JudgmentRecord[]): string | null;
359
+ /**
360
+ * Each judgment's `expect` label, by position: the attempt's `toPass` assertions (less those on a
361
+ * failed judge call, which recorded no judgment) in call order, when they are exactly as many as
362
+ * its judgments. `check` is that assertion's label, the key of its check; a judgment no label
363
+ * reaches is named by its question.
364
+ */
365
+ declare function labeledJudgments(a: Pick<AttemptRecord, 'judgements' | 'assertions'>): Array<{
366
+ judgment: JudgmentRecord;
367
+ label: string;
368
+ labelled: boolean;
369
+ check?: string;
370
+ }>;
371
+ /** Sum of counted attempts' usage; null when none of them reported any. */
372
+ declare function cellUsage(e: Pick<EvalEntry, 'attempts'>, cellKey: string): TokenUsage | null;
373
+ /**
374
+ * Each `c.exec` check's `expect` label, by position: the attempt's `toSucceed` assertions in call
375
+ * order, when they are exactly as many as its checks; undefined where no label reaches.
376
+ */
377
+ declare function checkLabels(a: Pick<AttemptRecord, 'checks' | 'assertions'>): Array<string | undefined>;
378
+ /**
379
+ * Each `toSucceed` label's `c.exec` check, by position, when the attempt's `toSucceed` assertions
380
+ * are exactly as many as its checks; empty otherwise.
381
+ */
382
+ declare function checksByLabel<C>(a: {
383
+ checks: readonly C[];
384
+ } & Pick<AttemptRecord, 'assertions'>): Map<string, C>;
385
+ /**
386
+ * How a `c.exec` given `retries` ended: `passed on try 2 of 3`, `failed all 3 tries`, or null
387
+ * when it had none or passed on its first try.
388
+ */
389
+ declare function triesNote(exitCode: number, tried: number, of: number): string | null;
390
+ /** {@link triesNote} for a recorded check. */
391
+ declare const checkTries: (c: {
392
+ exitCode: number;
393
+ retries?: number;
394
+ tries?: readonly unknown[];
395
+ }) => string | null;
396
+ /** A check's command on one line: the first non-empty line, clipped; its log has the rest. */
397
+ declare function commandLine(command: string, max?: number): string;
398
+ /** Splits `file › group › … › name` back into its parts. */
399
+ declare function splitEvalId(id: string): {
400
+ file: string;
401
+ path: string[];
402
+ };
403
+ declare function formatDuration(ms: number | null | undefined): string;
404
+ declare const formatPercent: (rate: number | null) => string;
405
+ /**
406
+ * A harness of the run's stack as `cursor (model composer-2.5, served composer-2.5[fast=true])`:
407
+ * the configured model and reasoning effort, then what the harness says it resolved.
408
+ */
409
+ declare function harnessLabel(h: RunRecord['stack']['harnesses'][number]): string;
410
+ /**
411
+ * The runtime reeval ran on, as `node 22.14.0`, or `bun 1.4.2 (node 26.3.0)` with the Node version
412
+ * it's compatible with. A run recorded before `stack.runtime` ran on Node.
413
+ */
414
+ declare function runtimeLabel(stack: Pick<RunRecord['stack'], 'node' | 'runtime'>): string;
415
+ /** The evals' language as `typescript evals` or `dart 3.13.5 evals`; null when not recorded. */
416
+ declare function languageLabel(stack: Pick<RunRecord['stack'], 'language'>): string | null;
417
+ /** Harness-reported tokens as `127k in (98k cached) · 776 out`; null when none were reported. */
418
+ declare function formatUsage(usage: TokenUsage | null | undefined): string | null;
419
+ declare const shortSha: (sha: string | null | undefined) => string;
420
+ /** Reasoning time, or `n/a` when the harness couldn't measure it: never `0s` for unknown. */
421
+ declare const reasoningLabel: (ms: number | null) => string;
422
+ /**
423
+ * An attempt's effort: tool calls, then the files the workspace diff changed, which counts a file
424
+ * the shell wrote, then the calls to the agent's edit tools, which doesn't.
425
+ */
426
+ declare function activityLabel(a: Pick<AttemptRecord['activity'], 'tools' | 'fileEdits'>, changed: number): string;
427
+ /**
428
+ * A run's outcome, which `reeval run`'s exit code is read from: an error outside any attempt, a
429
+ * failed gated cell or a budget stop fails the run, while failures in ungated cells (controls, by
430
+ * default) don't. An interrupted run is `skipped`, whatever ran; a resume its filters left
431
+ * `interrupted` exits on its outcome as if it were complete.
432
+ */
433
+ declare function runOutcome(run: {
434
+ status: RunRecord['status'];
435
+ evals: ReadonlyArray<Pick<EvalEntry, 'outcome'>>;
436
+ errors?: readonly string[];
437
+ }): Outcome;
438
+ /**
439
+ * The eval hasn't run yet: its run stopped (interrupted, or out of budget) or a filtered resume
440
+ * left it out, and it's `skipped` without being a todo or declared skipped. A resume runs it.
441
+ */
442
+ declare const notRunYet: (run: Pick<RunRecord, 'status'>, e: Pick<EvalEntry, 'outcome' | 'todo' | 'skip'>) => boolean;
443
+ /** Distinct judge labels that failed on a counted attempt, in first-seen order. */
444
+ declare function failedJudgeLabels(record: Pick<RunRecord, 'evals'>): string[];
445
+ /**
446
+ * One sentence for the studio run header: counted pass/fail, which judges failed, and infra.
447
+ * Works without `reeval analyze`.
448
+ */
449
+ declare function runLead(record: Pick<RunRecord, 'summary' | 'evals'>): string;
450
+ type RunRow = {
451
+ entry: HistoryEntry;
452
+ passed: number;
453
+ /** Attempts that reached a verdict or errored; skips excluded. */
454
+ total: number;
455
+ /** Passes and fails, the denominator of the run's pass@1. */
456
+ counted: number;
457
+ /** Stack changes in the cells it shares with earlier runs, and in the judge since the previous complete run. */
458
+ drift: StackChange[];
459
+ /** Evals with a cell whose fingerprint changed since the last run that had that cell. */
460
+ changedEvals: number;
461
+ };
462
+ /**
463
+ * Newest first. Drift is measured per cell, against the last run that had it: a run over fewer
464
+ * cells (a filtered run) is no drift, and neither is the full run after it. The judge's settings
465
+ * compare with the next older complete run.
466
+ */
467
+ declare function runRows(entries: HistoryEntry[]): RunRow[];
468
+ /** Attempt counts; `missed` is ungated cells' failures, which `fail` leaves out. */
469
+ type Tally = {
470
+ pass: number;
471
+ fail: number;
472
+ missed: number;
473
+ infra: number;
474
+ skipped: number;
475
+ };
476
+ type TreeNode = {
477
+ kind: 'group';
478
+ key: string;
479
+ name: string;
480
+ tally: Tally;
481
+ children: TreeNode[];
482
+ } | {
483
+ kind: 'eval';
484
+ key: string;
485
+ name: string;
486
+ tally: Tally;
487
+ eval: EvalEntry;
488
+ };
489
+ /** Files, then groups, then evals, each with attempt counts; failing branches sort first. */
490
+ declare function evalTree(record: Pick<RunRecord, 'evals'>): TreeNode[];
491
+ type CellRow = {
492
+ key: string;
493
+ harness: string;
494
+ environment: string;
495
+ sandbox: string;
496
+ role: CellRecord['role'];
497
+ /** Whether the cell's failures fail the eval: its gate isn't `'none'`. */
498
+ gated: boolean;
499
+ gate: CellRecord['gate'];
500
+ overrides: string[];
501
+ /** The reasoning effort the harness factory asked for; null when it left it unset. */
502
+ reasoningEffort: string | null;
503
+ runs: number;
504
+ pass: number;
505
+ fail: number;
506
+ infraError: number;
507
+ skipped: number;
508
+ /** pass@1: passes over passes and fails. */
509
+ passAt1: number | null;
510
+ outcome: Outcome;
511
+ /** A variant's lift over its control, in whole points (from the record), else null. */
512
+ lift: number | null;
513
+ /** Its 95% interval in whole points, and whether that spans 0. */
514
+ liftInterval: {
515
+ lo: number;
516
+ hi: number;
517
+ withinNoise: boolean;
518
+ } | null;
519
+ /** The interval's lower bound is below the compare's `minLift`. */
520
+ belowMinLift: boolean;
521
+ /** Median tool calls, agent time and tokens over the cell's counted attempts. */
522
+ effort: CellEffort | null;
523
+ /** A variant's effort against its control, with the rank test's verdict. */
524
+ effortVsControl: EffortContrast | null;
525
+ /** Per-key judge pass/counted over this cell's counted attempts. */
526
+ judgeRates: JudgeKeyRate[];
527
+ /** Sum of counted attempts' tokens; null when none reported. */
528
+ usage: TokenUsage | null;
529
+ attempts: AttemptRecord[];
530
+ };
531
+ /** One row per cell: controls and variants, with each variant's recorded lift. */
532
+ declare function cellRows(e: EvalEntry): CellRow[];
533
+ /** One side's wins over the decided votes it was in; `rate` is null when none decided. */
534
+ type AnswerRate = {
535
+ name: string;
536
+ wins: number;
537
+ n: number;
538
+ rate: number | null;
539
+ };
540
+ type AnswerRow = {
541
+ question: string;
542
+ /** The model that answered: as served, when the provider said, else as configured. */
543
+ model: string;
544
+ votes: number;
545
+ ties: number;
546
+ /** Arms against each other on one harness, then harnesses against each other on one arm. */
547
+ within: AnswerRate[];
548
+ across: AnswerRate[];
549
+ /** Why some or all of its votes are missing: the model call failed. */
550
+ error: string | null;
551
+ /** No votes and no error: no pair had both sides pass their checks. */
552
+ noPairs: boolean;
553
+ /** Distinct vote reasons, in vote order, at most three. */
554
+ reasons: string[];
555
+ /** `questions/<n>.json` under the run's directory: what the model was sent; null when not kept. */
556
+ input: string | null;
557
+ };
558
+ /** One row per pairwise question, with each cut's win rates in the record's order. */
559
+ declare function answerRows(e: Pick<EvalEntry, 'answers'>): AnswerRow[];
560
+ /** One attempt's answer to a question asked of it on its own. */
561
+ type AttemptAnswerItem = {
562
+ /** The attempt's id, `<cell key>#<n>`. */
563
+ attempt: string;
564
+ outcome: Outcome;
565
+ choice: string | null;
566
+ text: string;
567
+ error: string | null;
568
+ /** `questions/<n>.json` in the attempt's directory: what the model was sent; null when not kept. */
569
+ input: string | null;
570
+ };
571
+ type AttemptAnswerRow = {
572
+ question: string;
573
+ /** The model that answered: as served, when the provider said, else as configured. */
574
+ model: string;
575
+ /** How many attempts picked each choice, in the question's order; null on a free-text question. */
576
+ choices: Array<{
577
+ choice: string;
578
+ n: number;
579
+ }> | null;
580
+ answers: AttemptAnswerItem[];
581
+ };
582
+ /** One row per question asked of each attempt on its own, in first-asked order. */
583
+ declare function attemptAnswerRows(e: Pick<EvalEntry, 'attempts'>): AttemptAnswerRow[];
584
+ /** Whether an eval needs the cells table: several cells, or several runs of one. */
585
+ declare const isMatrix: (e: Pick<EvalEntry, 'cells'>) => boolean;
586
+ type EvalHistoryPoint = {
587
+ run: HistoryEntry;
588
+ outcome: Outcome;
589
+ /** pass@1 over the gated cells' attempts. */
590
+ passAt1: number | null;
591
+ cells: HistoryCell[];
592
+ /** A cell's fingerprint differs from the last run that had that cell. */
593
+ changed: boolean;
594
+ };
595
+ /** One eval across runs, oldest first. */
596
+ declare function evalHistory(entries: HistoryEntry[], slug: string): EvalHistoryPoint[];
597
+ type CompareViewRow = {
598
+ id: string;
599
+ slug: string | null;
600
+ status: CompareStatus;
601
+ base: Outcome | null;
602
+ head: Outcome | null;
603
+ basePassAt1: number | null;
604
+ headPassAt1: number | null;
605
+ /** See `CompareRow`: then pass@1 is over the cells both runs have too. */
606
+ cells?: CompareRow['cells'];
607
+ changes?: string[];
608
+ gate?: CompareRow['gate'];
609
+ /** Mean judge score per run; null without judgments or without the record. */
610
+ baseScore: number | null;
611
+ headScore: number | null;
612
+ /** Cells in both runs, with lift and effort on each side. */
613
+ perCell: CompareCell[];
614
+ /** See `CompareRow`: an eval with `cases`, each case's row, and its gates over the cases. */
615
+ cases?: CompareRow['cases'];
616
+ caseScope?: CompareRow['caseScope'];
617
+ lifts?: CompareRow['lifts'];
618
+ rates?: CompareRow['rates'];
619
+ };
620
+ declare const COMPARE_ORDER: CompareStatus[];
621
+ /**
622
+ * Whether some of an eval's cells aren't gated (`gate: 'none'`, e.g. controls under
623
+ * `compare.gate: 'variants'`). Its pass@1 then covers fewer attempts than its pass count, so
624
+ * the eval line labels it `gated`: `2/4 · pass@1 100% gated`.
625
+ */
626
+ declare const hasUngatedCells: (cells: readonly Pick<HistoryCell, 'gate'>[]) => boolean;
627
+ /** A study's pass: every cell `gate: 'none'`, so nothing it saw passed or failed it. */
628
+ declare const isNotGatedPass: (e: {
629
+ outcome: Outcome;
630
+ cells: readonly Pick<HistoryCell, 'gate'>[];
631
+ }) => boolean;
632
+ /**
633
+ * An expected miss: a failed attempt in an ungated cell (a control, by default). It's reported
634
+ * apart from failures, since it fails neither its eval nor the run.
635
+ */
636
+ declare const isUngatedMiss: (e: {
637
+ cells: readonly Pick<HistoryCell, 'key' | 'gate'>[];
638
+ }, a: Pick<AttemptRecord, 'cell' | 'outcome' | 'error'>) => boolean;
639
+ /**
640
+ * The outcome an eval's mark shows. A pass needs a counted (passed or failed) attempt and no infra
641
+ * error: a gate that lets those through, such as `gate: 'none'`, still shows `!`, not `✓`.
642
+ */
643
+ declare function evalMarkOutcome(e: Pick<EvalEntry, 'outcome' | 'attempts'>): Outcome;
644
+ /** `reeval compare`'s classification, plus pass@1 and judge-score deltas from the records. */
645
+ declare function compareView(base: HistoryEntry, head: HistoryEntry, records?: {
646
+ base?: RunRecord;
647
+ head?: RunRecord;
648
+ }): {
649
+ rows: CompareViewRow[];
650
+ counts: Record<CompareStatus, number>;
651
+ changes: StackChange[];
652
+ };
653
+ type TranscriptLine = ({
654
+ turn: number;
655
+ } & TranscriptEvent) | {
656
+ turn: number;
657
+ ts: string;
658
+ atMs: number;
659
+ type: 'prompt';
660
+ text: string;
661
+ } | {
662
+ turn: number;
663
+ ts: string;
664
+ atMs: number;
665
+ type: 'turn-end';
666
+ status: TurnStatus;
667
+ durationMs: number;
668
+ };
669
+ type TranscriptItem = {
670
+ kind: 'turn';
671
+ turn: number;
672
+ prompt: string;
673
+ ts: string;
674
+ } | {
675
+ kind: 'event';
676
+ turn: number;
677
+ event: TranscriptLine;
678
+ } | {
679
+ kind: 'turn-end';
680
+ turn: number;
681
+ status: TurnStatus;
682
+ durationMs: number;
683
+ };
684
+ /** Parses `transcript.jsonl`, skipping lines that aren't JSON (a truncated tail, say). */
685
+ declare function parseTranscript(text: string): TranscriptLine[];
686
+ /**
687
+ * `transcript.jsonl`'s lines in file order, each as `parseTranscriptLine` reads it, so line `n`
688
+ * (from 1) is at index `n - 1`.
689
+ */
690
+ declare function transcriptByLine(text: string): Array<TranscriptLine | null>;
691
+ /** One line of a transcript, or null when it's blank, isn't JSON or has no `type`. */
692
+ declare function parseTranscriptLine(line: string): TranscriptLine | null;
693
+ declare function transcriptItems(lines: TranscriptLine[]): TranscriptItem[];
694
+ declare const findEval: (record: Pick<RunRecord, 'evals'>, slug: string) => EvalEntry | undefined;
695
+ declare const findAttempt: (e: EvalEntry, attemptId: string) => AttemptRecord | undefined;
696
+ /** The attempt to open by default: the first failure, else the first attempt. */
697
+ declare const defaultAttempt: (e: EvalEntry) => AttemptRecord | undefined;
698
+ /** What one attempt's transcript shows it did, counted from its classified tool calls. */
699
+ type AttemptFacts = {
700
+ toolCalls: number;
701
+ byClass: Record<ToolClass, number>;
702
+ /** Each skill its calls used, by where it came from, in order of first use. */
703
+ skills: Array<{
704
+ skill: string;
705
+ source: SkillSource;
706
+ /** The calls that used it. */
707
+ reads: number;
708
+ /** The transcript line of its first use, from 1. */
709
+ firstLine: number;
710
+ }>;
711
+ /** Each placed context file its calls read, directly or through a glob, in order of first read. */
712
+ contextFiles: Array<{
713
+ path: string;
714
+ reads: number;
715
+ firstLine: number;
716
+ }>;
717
+ };
718
+ /**
719
+ * One attempt's facts from its transcript's lines (as `transcriptByLine` parses them), its
720
+ * calls classified as `reeval query`'s `tool_calls`, `context_reads` and `skill_uses` classify them,
721
+ * against the skills and MCP servers of the attempt's environment in `record`.
722
+ */
723
+ declare function attemptFacts(record: RunRecord, attempt: AttemptRecord, lines: readonly (TranscriptLine | null)[]): AttemptFacts;
724
+ /** `<history>/<run>/remarks.json`, beside the record. */
725
+ declare const REMARKS_FILE = "remarks.json";
726
+ /** A run's remarks, `<history>/<run>/remarks.json`. Written only when the run raised one. */
727
+ type RunRemarks = {
728
+ kind: 'reeval-remarks';
729
+ schemaVersion: 1;
730
+ runId: string;
731
+ createdAt: string;
732
+ /**
733
+ * Every remark this run raised, ranked, including those an earlier run printed first
734
+ * (`firstSeenRun`), which don't print again.
735
+ */
736
+ remarks: Remark[];
737
+ };
738
+ /** Each remark kind's severity. */
739
+ declare const REMARK_KINDS: Record<Remark['kind'], {
740
+ severity: 'concern' | 'suggestion';
741
+ }>;
742
+ /** A remark about the eval's quality, not about the agent. */
743
+ type Remark = {
744
+ /**
745
+ * sha256(kind, evalId, the eval's sources hash, and for `slow-mcp-start` the server): one remark
746
+ * per kind (and server) per eval version.
747
+ */
748
+ id: string;
749
+ evalId: string;
750
+ /**
751
+ * The eval's record key (`EvalEntry.key`), when known. Optional on older remarks; `evalId` keeps
752
+ * the encoded fact path used for `remarkId`.
753
+ */
754
+ evalKey?: string;
755
+ kind: 'too-few-tries' | 'cant-tell-apart' | 'check-never-fails' | 'flaky-check' | 'slow-mcp-start' | 'proxy-only-egress' | 'setup-process-secret' | 'setup-tunnel-closed' | 'snapshot-too-large' | 'observe-throws';
756
+ /** Fixed by kind (`REMARK_KINDS`). */
757
+ severity: 'concern' | 'suggestion';
758
+ /** One plain sentence: what's wrong. */
759
+ text: string;
760
+ /**
761
+ * The remark as one short line, with its fix and where (`1 run per side; set runs: 4 at
762
+ * evals/a.eval.ts:12 to confirm these differences.`), which `reeval run`'s list prints without
763
+ * `-v`; absent when the kind has none.
764
+ */
765
+ brief?: string;
766
+ /** Facts from this run (or history facts) showing it mattered now. */
767
+ evidence: string[];
768
+ /** An edit at a source location (`source:<file>#L<n>`) in the eval. */
769
+ fix: {
770
+ at: string;
771
+ edit: string;
772
+ };
773
+ /**
774
+ * The first run that printed this `id`, or this run when none has: it prints only in that run,
775
+ * except the kinds about the environment, which print in every run they happen.
776
+ */
777
+ firstSeenRun: string;
778
+ /**
779
+ * Whether this run's `reeval run` output printed it: false when an earlier run printed it, or
780
+ * when the cap or a run without `-v` left it out, so a later run prints it.
781
+ */
782
+ printed?: boolean;
783
+ };
784
+ /** A displayed number: `value` for charts, `text` as every surface prints it (`461k`, `1m 05s`, `not reported`). */
785
+ type Figure = {
786
+ value: number | null;
787
+ text: string;
788
+ };
789
+ /** `text`: `-42%`, `±0%`, `+50 pts`, `n/a`. */
790
+ type Delta = {
791
+ value: number | null;
792
+ unit: 'pct' | 'pts';
793
+ text: string;
794
+ };
795
+ /** A 95% interval in the parent's unit. `text`: `+66 pts to +100 pts`. */
796
+ type Interval = {
797
+ lo: number;
798
+ hi: number;
799
+ text: string;
800
+ };
801
+ type Spread = {
802
+ n: number;
803
+ median: Figure;
804
+ q1: Figure;
805
+ q3: Figure;
806
+ };
807
+ /** `text`: `12 of 12`. */
808
+ type Count = {
809
+ k: number;
810
+ n: number;
811
+ text: string;
812
+ };
813
+ /** `text`: `75%`. `interval` is Wilson's; null at n = 0. */
814
+ type Rate = {
815
+ k: number;
816
+ n: number;
817
+ rate: number | null;
818
+ text: string;
819
+ interval: Interval | null;
820
+ };
821
+ /** `href` is the citation target (Section 2.10). */
822
+ type AttemptRef = {
823
+ /** The display label, `A1..AN`, numbered over the evals the report covers. */
824
+ id: string;
825
+ /** The attempt's `AttemptRecord.key`, the same in every report of its run. */
826
+ key: string;
827
+ href: string;
828
+ };
829
+ type Fairness = 'clear' | 'within-noise' | 'too-few-runs';
830
+ type MeasureKey = 'tokens' | 'toolCalls' | 'turns' | 'costUsd' | 'agentMs' | 'filesChanged' | 'linesAdded' | 'linesRemoved';
831
+ type MeasureCells = Record<MeasureKey, Spread | null>;
832
+ type EvalShape = 'task' | 'ab' | 'discoverability' | 'regression';
833
+ type RunReport = {
834
+ reportVersion: 1;
835
+ runId: string;
836
+ /** `evals.length === 1`: the eval is the `#` title and every block moves up one level. */
837
+ single: boolean;
838
+ header: RunHeader;
839
+ /** Empty when clean; the block is then omitted. */
840
+ integrity: IntegrityIssue[];
841
+ rollup: ExecutiveRollup | null;
842
+ /** Empty unless `evals.length >= 2` and two or more harnesses ran. */
843
+ roster: RosterRow[];
844
+ evals: EvalSection[];
845
+ /** From the analysis; empty without one. */
846
+ actions: ReportClaim[];
847
+ provenance: ProvenanceState;
848
+ attempts: AttemptIndexRow[];
849
+ };
850
+ type RunTally = {
851
+ /** Pass plus fail over gated cells: the pass-rate denominator. */
852
+ counted: number;
853
+ passed: number;
854
+ failed: number;
855
+ /** Ungated cells (baselines by default), never in `counted`. */
856
+ ungated: {
857
+ passed: number;
858
+ failed: number;
859
+ };
860
+ infraError: number;
861
+ /** Of `infraError`: attempts an unhandled turn error moved out of pass or fail. */
862
+ reclassified: number;
863
+ skipped: number;
864
+ };
865
+ type RunHeader = {
866
+ title: string;
867
+ status: RunRecord['status'];
868
+ startedAt: string;
869
+ author: string;
870
+ reeval: string;
871
+ profile: string | null;
872
+ /** Which runs this one is like-with-like with (`runScope`); unset in reports built before it. */
873
+ scope?: RunScope;
874
+ /** Null when cells ran different counts. */
875
+ runsPerCell: number | null;
876
+ wall: Figure;
877
+ tally: RunTally;
878
+ /** `input + output + cacheRead + cacheWrite` (`totalTokens`); null when no attempt reported usage. */
879
+ tokens: {
880
+ total: Figure;
881
+ cacheRead: Figure;
882
+ } | null;
883
+ /** Null when no attempt or judge call has a cost. */
884
+ cost: {
885
+ agent: Figure | null;
886
+ judge: Figure | null;
887
+ /** Labeled total when agent and judge are both present; else null. */
888
+ total: Figure | null;
889
+ basis: 'reported' | 'estimated' | 'mixed';
890
+ } | null;
891
+ /** `repo` is the basename of `git.remote`. */
892
+ git: {
893
+ sha: string;
894
+ branch: string | null;
895
+ dirty: boolean;
896
+ repo: string | null;
897
+ } | null;
898
+ };
899
+ type IntegrityKind = 'run-incomplete' | 'turn-error' | 'pre-satisfied' | 'fairness-floor' | 'served-model-drift' | 'unreported-usage' | 'prompt-confounded' | 'duplicate-environment' | 'baseline-never-passes' | 'dirty-worktree';
900
+ type IntegrityIssue = {
901
+ kind: IntegrityKind;
902
+ severity: 'error' | 'warn';
903
+ /** One plain sentence that names the metric it affects. */
904
+ text: string;
905
+ evalId: string | null;
906
+ cites: AttemptRef[];
907
+ };
908
+ type ExecutiveRollup = {
909
+ /** `analysis.tldr`; null without an analysis. */
910
+ lead: ReportClaim | null;
911
+ rows: Array<{
912
+ evalId: string;
913
+ name: string;
914
+ anchor: string;
915
+ shapes: EvalShape[];
916
+ /** `3 arms × 4 coding harnesses`. */
917
+ design: string;
918
+ /** The eval's headline in one cell, in its native unit. */
919
+ primary: string;
920
+ medianTokens: Figure;
921
+ }>;
922
+ /** Run-level analysis claims (no `evalId`). */
923
+ findings: ReportClaim[];
924
+ };
925
+ type RosterRow = {
926
+ harness: string;
927
+ model: string | null;
928
+ effort: string | null;
929
+ servedModels: string[];
930
+ gateway: string | null;
931
+ attempts: number;
932
+ };
933
+ type EvalSection = {
934
+ /** `EvalEntry.key`. */
935
+ id: string;
936
+ /** GFM slug of the eval's heading. */
937
+ anchor: string;
938
+ /** Unique in the report; `<name> (<file>)` when two evals share a name. */
939
+ name: string;
940
+ file: string;
941
+ shapes: EvalShape[];
942
+ /** `3 prompts, 4 coding harnesses, 3 arms, 36 attempts`. */
943
+ design: string;
944
+ headline: EvalHeadline;
945
+ insights: ReportClaim[];
946
+ /** Null on a check-free eval, whose first choice, judge, or measure block leads. */
947
+ primary: TaskResults | VariantComparison | RegressionChanges | null;
948
+ choices: ChoiceBreakdown[];
949
+ discovery: DiscoveryRate[];
950
+ judgments: JudgmentBreakdown | null;
951
+ qualitative: QualitativeAnswers[];
952
+ captures: CaptureGallery | null;
953
+ failures: FailureClusters | null;
954
+ };
955
+ type ArmRef = {
956
+ id: string;
957
+ label: string;
958
+ };
959
+ type EvalHeadline = {
960
+ /** The eval's name: the heading above the headline. */
961
+ task: string;
962
+ /** What differed (`Revised skill against the JS SDK baseline`); null on a task eval. */
963
+ kicker: string | null;
964
+ fairness: Fairness;
965
+ primary: HeadlinePrimary;
966
+ /** Sample size, harness count, interval, and floor, as sentences. */
967
+ foot: string;
968
+ };
969
+ type HeadlinePrimary = {
970
+ kind: 'pass-rate';
971
+ rate: Rate;
972
+ assertions: number;
973
+ harnesses: number;
974
+ /** Product outcome noun: passed (default), found, or used. */
975
+ outcome: 'passed' | 'found' | 'used';
976
+ } | {
977
+ kind: 'lift';
978
+ baseline: ArmRef;
979
+ variant: ArmRef;
980
+ points: Delta;
981
+ interval: Interval;
982
+ baselineRate: Rate;
983
+ variantRate: Rate;
984
+ /** Every other variant, with its own interval; never dropped. */
985
+ others: Array<{
986
+ variant: ArmRef;
987
+ points: Delta;
988
+ interval: Interval;
989
+ rate: Rate;
990
+ }>;
991
+ } | {
992
+ kind: 'tied-ceiling';
993
+ baseline: ArmRef;
994
+ variant: ArmRef;
995
+ rate: Rate;
996
+ assertions: number;
997
+ lead: {
998
+ measure: MeasureKey;
999
+ baseline: Figure;
1000
+ variant: Figure;
1001
+ delta: Delta;
1002
+ };
1003
+ } | {
1004
+ kind: 'choice';
1005
+ question: string;
1006
+ options: Array<{
1007
+ label: string;
1008
+ count: number;
1009
+ }>;
1010
+ attempts: number;
1011
+ noPick: number;
1012
+ } | {
1013
+ kind: 'measure';
1014
+ label: string;
1015
+ unit: 'tokens' | 'ms' | 'usd' | 'count' | 'lines' | 'files';
1016
+ spread: Spread;
1017
+ } | {
1018
+ kind: 'judge';
1019
+ label: string;
1020
+ question: string;
1021
+ scale: [number, number];
1022
+ threshold: number | null;
1023
+ spread: Spread;
1024
+ atOrAbove: Count | null;
1025
+ } | {
1026
+ kind: 'text';
1027
+ question: string;
1028
+ answered: Count;
1029
+ lead: Quote | null;
1030
+ /** When many answers collapse to one quote shown in the headline. */
1031
+ leadRole: 'sole' | 'representative' | null;
1032
+ } | {
1033
+ kind: 'regression';
1034
+ worse: number;
1035
+ fixed: number;
1036
+ /** First gated worse/fixed change, else first change row. */
1037
+ lead: ChangeRow | null;
1038
+ };
1039
+ type CheckCell = {
1040
+ status: 'pass' | 'fail' | 'mixed' | 'na' | 'error' | 'not-run';
1041
+ passed: number;
1042
+ /** Pass plus fail; never counts `notApplicable`. */
1043
+ answered: number;
1044
+ notApplicable: number;
1045
+ preSatisfied: number;
1046
+ /** `12 of 12`, `0 of 4`, `n/a`, `12 of 12, 12 pre-satisfied`. */
1047
+ text: string;
1048
+ /** Why the check was n/a when known (`expect.skip` reason); null otherwise. */
1049
+ reason: string | null;
1050
+ };
1051
+ type AssertionRow = {
1052
+ label: string;
1053
+ matcher: string;
1054
+ at: string | null;
1055
+ /** One entry per column key; every column present. */
1056
+ cells: Record<string, CheckCell>;
1057
+ total: CheckCell;
1058
+ };
1059
+ type BreakdownRow = {
1060
+ /** One value per axis, repeated on every row. */
1061
+ keys: string[];
1062
+ rate: Rate | null;
1063
+ measures: MeasureCells;
1064
+ };
1065
+ type TaskResults = {
1066
+ kind: 'task';
1067
+ columns: string[];
1068
+ assertions: AssertionRow[];
1069
+ axes: string[];
1070
+ breakdown: BreakdownRow[];
1071
+ /** The cross-group median row; null with one group. */
1072
+ median: BreakdownRow | null;
1073
+ };
1074
+ type Arm = {
1075
+ id: string;
1076
+ label: string;
1077
+ environment: string;
1078
+ environmentDigest: string | null;
1079
+ role: 'baseline' | 'variant';
1080
+ change: string | null;
1081
+ /** Distinct first-turn prompt hashes over the arm's attempts. */
1082
+ prompts: string[];
1083
+ };
1084
+ type MeasureVsBaseline = {
1085
+ value: Spread | null;
1086
+ delta: Delta | null;
1087
+ /** `3 of 4` harnesses lower than the baseline; null on the baseline. */
1088
+ lower: Count | null;
1089
+ };
1090
+ type ArmRow = {
1091
+ arm: string;
1092
+ rate: Rate;
1093
+ lift: Delta | null;
1094
+ interval: Interval | null;
1095
+ /** Pass-rate certainty only; measure direction lives in measure cells. */
1096
+ verdict: 'baseline' | 'helped' | 'made-worse' | 'within-noise' | 'too-few-runs';
1097
+ /** Always `pass-rate` for this column today. */
1098
+ verdictScope: 'pass-rate';
1099
+ measures: Record<MeasureKey, MeasureVsBaseline>;
1100
+ };
1101
+ type EffortPivotRow = {
1102
+ /** A harness label, or `Median` for the cross-harness row. */
1103
+ harness: string;
1104
+ arm: string;
1105
+ rate: Rate;
1106
+ measures: Record<MeasureKey, MeasureVsBaseline>;
1107
+ };
1108
+ type BehaviorRow = {
1109
+ label: string;
1110
+ matcher: string;
1111
+ cells: Record<string, Count | null>;
1112
+ };
1113
+ type ContrastRow = {
1114
+ from: string;
1115
+ to: string;
1116
+ measure: 'pass-rate' | MeasureKey;
1117
+ delta: Delta;
1118
+ interval: Interval | null;
1119
+ verdict: 'helped' | 'made-worse' | 'within-noise' | 'too-few-runs' | 'fewer' | 'more';
1120
+ /** `pass-rate` when the contrast is on pass rate; else the measure's own direction. */
1121
+ verdictScope: 'pass-rate' | 'measure';
1122
+ };
1123
+ type VariantComparison = {
1124
+ kind: 'comparison';
1125
+ /** Baseline first, then variants in declaration order. */
1126
+ arms: Arm[];
1127
+ /** Set when every arm passed every applicable assertion; `assertionsByArm` is then empty. */
1128
+ uniformLead: string | null;
1129
+ summary: ArmRow[];
1130
+ assertionsByArm: AssertionRow[];
1131
+ behavior: BehaviorRow[];
1132
+ /** From `contrastsOf`; empty with one variant. */
1133
+ contrasts: ContrastRow[];
1134
+ effortByHarness: EffortPivotRow[];
1135
+ };
1136
+ type ChangeRow = {
1137
+ signal: string;
1138
+ kind: 'outcome' | 'assertion' | 'judgment' | 'choice' | 'measure' | 'observation';
1139
+ harness: string | null;
1140
+ prior: string;
1141
+ current: string;
1142
+ delta: Delta | null;
1143
+ status: CompareStatus;
1144
+ gate: boolean;
1145
+ };
1146
+ type RegressionChanges = {
1147
+ kind: 'regression';
1148
+ baseRunId: string;
1149
+ /** Prior-run identity for reproducible comparison. */
1150
+ prior: {
1151
+ runId: string;
1152
+ startedAt: string;
1153
+ gitSha: string | null;
1154
+ branch: string | null;
1155
+ author: string;
1156
+ };
1157
+ /** Experimental inputs that changed (`stackChanges` / config), not outcome rows. */
1158
+ whatChanged: string[];
1159
+ rows: ChangeRow[];
1160
+ unchanged: ChangeRow[];
1161
+ };
1162
+ type ChoiceBreakdown = {
1163
+ question: string;
1164
+ /** Sorted by count, then by first appearance. */
1165
+ options: Array<{
1166
+ label: string;
1167
+ count: number;
1168
+ }>;
1169
+ noPick: number;
1170
+ rows: Array<{
1171
+ arm: string | null;
1172
+ prompt: string | null;
1173
+ harness: string;
1174
+ counts: Record<string, number>;
1175
+ attempts: number;
1176
+ text: string;
1177
+ }>;
1178
+ passByPick: Array<{
1179
+ option: string;
1180
+ rate: Rate;
1181
+ }>;
1182
+ };
1183
+ type DiscoveryRate = {
1184
+ what: string;
1185
+ matcher: string;
1186
+ found: Count;
1187
+ rows: Array<{
1188
+ keys: string[];
1189
+ found: Count;
1190
+ }>;
1191
+ /**
1192
+ * For a skill, the attempts that found it by where the skill they checked came from, its first
1193
+ * use in the report's `facts`: one source an attempt, so the counts sum to at most `found`. An
1194
+ * attempt counts in none when it has no facts, when its passing `toHaveCalledSkill` had a label
1195
+ * of its own (which doesn't name the skill), when the check that found it was a `not.`, or when
1196
+ * the matcher passed on a call the facts leave out (a failed call, or a file under
1197
+ * `skills/<name>/` that isn't its `SKILL.md`). Empty without facts.
1198
+ */
1199
+ sources: Array<{
1200
+ source: 'placed' | 'skill-dir' | 'elsewhere' | 'mcp' | 'web';
1201
+ count: number;
1202
+ }>;
1203
+ };
1204
+ type JudgeRow = {
1205
+ label: string;
1206
+ question: string;
1207
+ /** The `expect(verdict).toPass()` label that gates on it; null for a diagnostic judgment. */
1208
+ assertion: string | null;
1209
+ threshold: number | null;
1210
+ keys: string[];
1211
+ spread: Spread;
1212
+ /** Null when `threshold` is null (diagnostic): omit from "At or above". */
1213
+ atOrAbove: Count | null;
1214
+ };
1215
+ type PairwiseRow = {
1216
+ question: string;
1217
+ cut: CompareCut;
1218
+ a: string;
1219
+ b: string;
1220
+ aWins: number;
1221
+ bWins: number;
1222
+ ties: number;
1223
+ pairs: number;
1224
+ reasons: Array<{
1225
+ text: string;
1226
+ cites: [AttemptRef, AttemptRef];
1227
+ input: string | null;
1228
+ }>;
1229
+ };
1230
+ type JudgmentBreakdown = {
1231
+ axes: string[];
1232
+ scores: JudgeRow[];
1233
+ pairwise: PairwiseRow[];
1234
+ };
1235
+ type Quote = {
1236
+ text: string;
1237
+ attempt: AttemptRef;
1238
+ harness: string;
1239
+ arm: string | null;
1240
+ case: string | null;
1241
+ input: string | null;
1242
+ };
1243
+ type QualitativeAnswers = {
1244
+ question: string;
1245
+ answered: Count;
1246
+ /** From the analysis; null without one. */
1247
+ themes: Array<{
1248
+ title: string;
1249
+ quotes: Quote[];
1250
+ }> | null;
1251
+ entries: Quote[];
1252
+ };
1253
+ type CaptureShot = {
1254
+ name: string;
1255
+ /** Null when not kept. */
1256
+ href: string | null;
1257
+ meta: string | null;
1258
+ attempt: AttemptRef;
1259
+ status: number | null;
1260
+ pageErrors: number | null;
1261
+ consoleErrors: number | null;
1262
+ };
1263
+ type CaptureGallery = {
1264
+ rows: Array<{
1265
+ harness: string;
1266
+ arm: string | null;
1267
+ case: string | null;
1268
+ shots: CaptureShot[];
1269
+ }>;
1270
+ };
1271
+ type FailureCluster = {
1272
+ kind: 'assertion' | 'turn-error' | 'infra-error' | 'capture-error' | 'error';
1273
+ label: string;
1274
+ at: string | null;
1275
+ /** Members of the eval's counted attempts. */
1276
+ count: Count;
1277
+ attempts: AttemptRef[];
1278
+ };
1279
+ type FailureClusters = {
1280
+ clusters: FailureCluster[];
1281
+ remarks: Array<{
1282
+ kind: Remark['kind'];
1283
+ severity: Remark['severity'];
1284
+ text: string;
1285
+ fix: Remark['fix'];
1286
+ }>;
1287
+ };
1288
+ type ReportClaim = {
1289
+ role: 'finding' | 'caveat' | 'action';
1290
+ text: string;
1291
+ confidence: Confidence;
1292
+ cites: AttemptRef[];
1293
+ action: {
1294
+ kind: ActionKind;
1295
+ target: string;
1296
+ change: string;
1297
+ verify: string | null;
1298
+ } | null;
1299
+ };
1300
+ type ProvenanceState = {
1301
+ /** `stateRows` without its `Context` row. */
1302
+ rows: Array<{
1303
+ field: string;
1304
+ value: string;
1305
+ }>;
1306
+ context: {
1307
+ count: number;
1308
+ files: Array<{
1309
+ path: string;
1310
+ sha256: string;
1311
+ }>;
1312
+ };
1313
+ artifacts: ArtifactLevel;
1314
+ /** `../<runId>.json`. */
1315
+ record: string;
1316
+ warnings: string[];
1317
+ };
1318
+ type AttemptIndexRow = {
1319
+ id: string;
1320
+ /** `AttemptRecord.key`. */
1321
+ key: string;
1322
+ evalId: string;
1323
+ harness: string;
1324
+ arm: string | null;
1325
+ case: string | null;
1326
+ outcome: Exclude<Outcome, 'skipped'> | 'canceled';
1327
+ reclassified: boolean;
1328
+ turnError: string | null;
1329
+ tokens: Figure;
1330
+ agentMs: Figure;
1331
+ toolCalls: Figure;
1332
+ /**
1333
+ * What its transcript shows it did, from the report's `facts`; absent when the report had none
1334
+ * for it, such as a passing attempt whose transcript `evidence` didn't keep.
1335
+ */
1336
+ facts?: AttemptFacts;
1337
+ /** Lines exclude lockfiles; null counts print `not recorded`. */
1338
+ diff: {
1339
+ files: number;
1340
+ added: number | null;
1341
+ removed: number | null;
1342
+ text: string;
1343
+ } | null;
1344
+ links: {
1345
+ transcript: string | null;
1346
+ diff: string | null;
1347
+ captures: Array<{
1348
+ name: string;
1349
+ href: string;
1350
+ }>;
1351
+ questions: string[];
1352
+ };
1353
+ /** The target every `AttemptRef` to this attempt uses. */
1354
+ href: string;
1355
+ };
1356
+ type BlockMarkdown = {
1357
+ runHeader(header: RunHeader, single: boolean): string;
1358
+ integrity(issues: IntegrityIssue[]): string;
1359
+ rollup(rollup: ExecutiveRollup): string;
1360
+ roster(rows: RosterRow[]): string;
1361
+ evalSection(section: EvalSection, single: boolean): string;
1362
+ actions(claims: ReportClaim[]): string;
1363
+ provenance(state: ProvenanceState): string;
1364
+ attemptIndex(rows: AttemptIndexRow[], single: boolean): string;
1365
+ };
1366
+ export { Remark as $, checkLabels as $t, IntegrityKind as A, AnalysisClaim as An, formatUsage as At, Quote as B, DIGEST_VERSION as Bn, reasoningLabel as Bt, ExecutiveRollup as C, stackChanges as Cn, evalMarkOutcome as Ct, Figure as D, AnalysisAction as Dn, findEval as Dt, Fairness as E, ActionKind as En, findAttempt as Et, MeasureKey as F, AnalysisNote as Fn, isUngatedMiss as Ft, RunHeader as G, shortSha as Gt, RegressionChanges as H, isSuperseded as Hn, runOutcome as Ht, MeasureVsBaseline as I, AnalysisScope as In, languageLabel as It, Spread as J, transcriptItems as Jt, RunReport as K, splitEvalId as Kt, PairwiseRow as L, AnalysisValidation as Ln, notRunYet as Lt, JudgeRow as M, AnalysisDoc as Mn, hasUngatedCells as Mt, JudgmentBreakdown as N, AnalysisEntry as Nn, isMatrix as Nt, HeadlinePrimary as O, AnalysisAnomaly as On, formatDuration as Ot, MeasureCells as P, AnalysisIndex as Pn, isNotGatedPass as Pt, REMARK_KINDS as Q, cellUsage as Qt, ProvenanceState as R, AnomalyKind as Rn, parseTranscript as Rt, EvalShape as S, sourceChanges as Sn, evalHistory as St, FailureClusters as T, ANALYSIS_PROMPT_VERSION as Tn, failedJudgeLabels as Tt, ReportClaim as U, pinnedAnalysis as Un, runRows as Ut, Rate as V, Level as Vn, runLead as Vt, RosterRow as W, sortAnalyses as Wn, runtimeLabel as Wt, VariantComparison as X, JudgeKeyRate as Xt, TaskResults as Y, CollapsedJudge as Yt, REMARKS_FILE as Z, cellJudgeRates as Zt, Delta as _, StackSource as _n, answerRows as _t, AttemptIndexRow as a, judgeLabel as an, AttemptAnswerItem as at, EvalHeadline as b, compare as bn, compareView as bt, BlockMarkdown as c, summaryJudgeRates as cn, CellRow as ct, CaptureShot as d, CompareCell as dn, RunRow as dt, checkTries as en, RunRemarks as et, ChangeRow as f, CompareFamilyLift as fn, Tally as ft, Count as g, StackChange as gn, activityLabel as gt, ContrastRow as h, CompareStatus as hn, TreeNode as ht, AssertionRow as i, commandLine as in, AnswerRow as it, Interval as j, AnalysisClaimRole as jn, harnessLabel as jt, IntegrityIssue as k, AnalysisBody as kn, formatPercent as kt, BreakdownRow as l, triesNote as ln, CompareViewRow as lt, ChoiceBreakdown as m, CompareRow as mn, TranscriptLine as mt, ArmRef as n, collapsedJudgeLabel as nn, attemptFacts as nt, AttemptRef as o, judgeRatesLabel as on, AttemptAnswerRow as ot, CheckCell as p, CompareFamilyRate as pn, TranscriptItem as pt, RunTally as q, transcriptByLine as qt, ArmRow as r, collapsedJudges as rn, AnswerRate as rt, BehaviorRow as s, labeledJudgments as sn, COMPARE_ORDER as st, Arm as t, checksByLabel as tn, AttemptFacts as tt, CaptureGallery as u, uniqueJudgeLabels as un, EvalHistoryPoint as ut, DiscoveryRate as v, addedRow as vn, attemptAnswerRows as vt, FailureCluster as w, whyChanged as wn, evalTree as wt, EvalSection as x, newRows as xn, defaultAttempt as xt, EffortPivotRow as y, cellsChanged as yn, cellRows as yt, QualitativeAnswers as z, Confidence as zn, parseTranscriptLine as zt };