zorya-classifier 0.0.0-stage → 0.2.0

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 (369) hide show
  1. package/LICENSE +21 -0
  2. package/NOTICE.md +13 -0
  3. package/README.md +139 -2
  4. package/dist/app/cli-arguments.d.ts +40 -0
  5. package/dist/app/cli-arguments.js +67 -0
  6. package/dist/app/cli-route.d.ts +3 -0
  7. package/dist/app/cli-route.js +40 -0
  8. package/dist/app/create-classifier.d.ts +33 -0
  9. package/dist/app/create-classifier.js +312 -0
  10. package/dist/app/extraction-files.d.ts +9 -0
  11. package/dist/app/extraction-files.js +28 -0
  12. package/dist/cli.d.ts +2 -0
  13. package/dist/cli.js +77 -0
  14. package/dist/entities/boundary-view.d.ts +29 -0
  15. package/dist/entities/boundary-view.js +1 -0
  16. package/dist/entities/contracts.d.ts +471 -0
  17. package/dist/entities/contracts.js +97 -0
  18. package/dist/entities/cutout-raster.d.ts +7 -0
  19. package/dist/entities/cutout-raster.js +1 -0
  20. package/dist/entities/errors.d.ts +7 -0
  21. package/dist/entities/errors.js +16 -0
  22. package/dist/entities/object-proposals.d.ts +53 -0
  23. package/dist/entities/object-proposals.js +8 -0
  24. package/dist/entities/representation.d.ts +225 -0
  25. package/dist/entities/representation.js +33 -0
  26. package/dist/entities/segmentation-metadata.d.ts +11 -0
  27. package/dist/entities/segmentation-metadata.js +1 -0
  28. package/dist/entities/segmentation-verification.d.ts +30 -0
  29. package/dist/entities/segmentation-verification.js +1 -0
  30. package/dist/entities/segmentation.d.ts +159 -0
  31. package/dist/entities/segmentation.js +5 -0
  32. package/dist/entities/whole-object.d.ts +128 -0
  33. package/dist/entities/whole-object.js +7 -0
  34. package/dist/features/choice-beam-layout.d.ts +18 -0
  35. package/dist/features/choice-beam-layout.js +136 -0
  36. package/dist/features/choice.d.ts +4 -0
  37. package/dist/features/choice.js +83 -0
  38. package/dist/features/prompts.d.ts +8 -0
  39. package/dist/features/prompts.js +56 -0
  40. package/dist/features/segmentation/box-check.d.ts +10 -0
  41. package/dist/features/segmentation/box-check.js +299 -0
  42. package/dist/features/segmentation/mask.d.ts +6 -0
  43. package/dist/features/segmentation/mask.js +89 -0
  44. package/dist/features/segmentation/packed-mask.d.ts +5 -0
  45. package/dist/features/segmentation/packed-mask.js +53 -0
  46. package/dist/features/uniform-grid.d.ts +3 -0
  47. package/dist/features/uniform-grid.js +35 -0
  48. package/dist/index.d.ts +10 -0
  49. package/dist/index.js +2 -0
  50. package/dist/infrastructure/boundary-view.d.ts +2 -0
  51. package/dist/infrastructure/boundary-view.js +263 -0
  52. package/dist/infrastructure/openrouter-common.d.ts +130 -0
  53. package/dist/infrastructure/openrouter-common.js +317 -0
  54. package/dist/infrastructure/openrouter-vision.d.ts +11 -0
  55. package/dist/infrastructure/openrouter-vision.js +17 -0
  56. package/dist/infrastructure/openrouter.d.ts +19 -0
  57. package/dist/infrastructure/openrouter.js +207 -0
  58. package/dist/infrastructure/raster.d.ts +11 -0
  59. package/dist/infrastructure/raster.js +216 -0
  60. package/dist/infrastructure/segmentation/box-presentation.d.ts +2 -0
  61. package/dist/infrastructure/segmentation/box-presentation.js +136 -0
  62. package/dist/infrastructure/segmentation/cutout-raster.d.ts +3 -0
  63. package/dist/infrastructure/segmentation/cutout-raster.js +79 -0
  64. package/dist/infrastructure/segmentation/grabcut/index.d.ts +13 -0
  65. package/dist/infrastructure/segmentation/grabcut/index.js +118 -0
  66. package/dist/infrastructure/segmentation/grabcut/worker-inference.d.ts +47 -0
  67. package/dist/infrastructure/segmentation/grabcut/worker-inference.js +260 -0
  68. package/dist/infrastructure/segmentation/index.d.ts +2 -0
  69. package/dist/infrastructure/segmentation/index.js +2 -0
  70. package/dist/infrastructure/segmentation/mobile-sam/index.d.ts +58 -0
  71. package/dist/infrastructure/segmentation/mobile-sam/index.js +472 -0
  72. package/dist/infrastructure/segmentation/mobile-sam/worker-inference.d.ts +149 -0
  73. package/dist/infrastructure/segmentation/mobile-sam/worker-inference.js +845 -0
  74. package/dist/infrastructure/segmentation/native-child.d.ts +23 -0
  75. package/dist/infrastructure/segmentation/native-child.js +297 -0
  76. package/dist/infrastructure/segmentation/native-protocol.d.ts +923 -0
  77. package/dist/infrastructure/segmentation/native-protocol.js +242 -0
  78. package/dist/infrastructure/segmentation/native-worker.d.ts +1 -0
  79. package/dist/infrastructure/segmentation/native-worker.js +93 -0
  80. package/dist/infrastructure/segmentation/probe-backend.d.ts +1 -0
  81. package/dist/infrastructure/segmentation/probe-backend.js +31 -0
  82. package/dist/infrastructure/segmentation/proposal-presentation.d.ts +3 -0
  83. package/dist/infrastructure/segmentation/proposal-presentation.js +371 -0
  84. package/dist/infrastructure/segmentation/verify-asset.d.ts +4 -0
  85. package/dist/infrastructure/segmentation/verify-asset.js +32 -0
  86. package/dist/infrastructure/text-representation.d.ts +23 -0
  87. package/dist/infrastructure/text-representation.js +226 -0
  88. package/dist/infrastructure/vision-chat.d.ts +64 -0
  89. package/dist/infrastructure/vision-chat.js +318 -0
  90. package/dist/openai-compatible.d.ts +8 -0
  91. package/dist/openai-compatible.js +11 -0
  92. package/dist/openrouter.d.ts +4 -0
  93. package/dist/openrouter.js +2 -0
  94. package/dist/processes/boundary-state.d.ts +12 -0
  95. package/dist/processes/boundary-state.js +196 -0
  96. package/dist/processes/classify.d.ts +3 -0
  97. package/dist/processes/classify.js +39 -0
  98. package/dist/processes/extract-from-region.d.ts +19 -0
  99. package/dist/processes/extract-from-region.js +415 -0
  100. package/dist/processes/extract.d.ts +25 -0
  101. package/dist/processes/extract.js +223 -0
  102. package/dist/processes/locate.d.ts +3 -0
  103. package/dist/processes/locate.js +274 -0
  104. package/dist/processes/proposal-location.d.ts +19 -0
  105. package/dist/processes/proposal-location.js +396 -0
  106. package/dist/processes/whole-object.d.ts +7 -0
  107. package/dist/processes/whole-object.js +272 -0
  108. package/dist/segmentation.d.ts +10 -0
  109. package/dist/segmentation.js +4 -0
  110. package/dist/shared/choice-mass.d.ts +2 -0
  111. package/dist/shared/choice-mass.js +19 -0
  112. package/dist/shared/gate.d.ts +11 -0
  113. package/dist/shared/gate.js +52 -0
  114. package/dist/shared/input-budget.d.ts +7 -0
  115. package/dist/shared/input-budget.js +21 -0
  116. package/dist/shared/scope.d.ts +12 -0
  117. package/dist/shared/scope.js +29 -0
  118. package/docs/0.2.0-plan.md +190 -0
  119. package/docs/README.ru.md +258 -0
  120. package/docs/acceptance.md +144 -0
  121. package/docs/api.md +49 -0
  122. package/docs/architecture.md +45 -0
  123. package/docs/benchmarks.md +63 -0
  124. package/docs/categorical-presence-port.md +26 -0
  125. package/docs/evidence/current-linux-check.json +15 -0
  126. package/docs/evidence/decisions-completion-v1-manifest.json +96 -0
  127. package/docs/evidence/decisions-completion-v1-report.json +1205 -0
  128. package/docs/evidence/decisions-selection-v1-manifest.json +96 -0
  129. package/docs/evidence/decisions-selection-v1-report.json +1457 -0
  130. package/docs/evidence/jev-pricing-2026-10-18.json +69 -0
  131. package/docs/evidence/linux-check.json +20 -0
  132. package/docs/evidence/linux-check.txt +205 -0
  133. package/docs/evidence/live-release-smoke-manifest.json +82 -0
  134. package/docs/evidence/live-release-smoke-provenance.json +9 -0
  135. package/docs/evidence/live-release-smoke-report.json +2136 -0
  136. package/docs/evidence/live-v1-manifest.json +84 -0
  137. package/docs/evidence/live-v1-provenance.json +9 -0
  138. package/docs/evidence/live-v1-report.json +15555 -0
  139. package/docs/evidence/luna-decisions-fine-probe.json +19 -0
  140. package/docs/evidence/luna-decisions-locator-smoke-manifest.json +39 -0
  141. package/docs/evidence/luna-decisions-locator-smoke-report.json +149 -0
  142. package/docs/evidence/pack-check.json +26 -0
  143. package/docs/evidence/protocol-probes.json +61 -0
  144. package/docs/evidence/rejected-vision-selection-v1-qwen-photo-dev-01.json +42 -0
  145. package/docs/evidence/rejected-vision-selection-v1-qwen-photo-dev-17.json +42 -0
  146. package/docs/evidence/rejected-vision-selection-v1-qwen-photo-heldout-09.json +42 -0
  147. package/docs/evidence/rejected-vision-selection-v2-haiku-photo-dev-16.json +42 -0
  148. package/docs/evidence/rejected-vision-selection-v2-haiku-support-heldout-016.json +42 -0
  149. package/docs/evidence/v02/baseline.json +1043 -0
  150. package/docs/evidence/v02/boundary-variants.md +17 -0
  151. package/docs/evidence/v02/box-confirmation-dev40/README.md +15 -0
  152. package/docs/evidence/v02/box-confirmation-dev40/component-refinement.json +246 -0
  153. package/docs/evidence/v02/box-confirmation-dev40/evaluation.json +1663 -0
  154. package/docs/evidence/v02/box-confirmation-dev40/provenance.json +130 -0
  155. package/docs/evidence/v02/box-confirmation-dev40/report.json +127 -0
  156. package/docs/evidence/v02/box-confirmation-dev40-freeze-20261020-v1.json +230 -0
  157. package/docs/evidence/v02/box-confirmation-method-selection-20261020-v1.json +85 -0
  158. package/docs/evidence/v02/box-heldout-cancelled-before-dispatch-20261021-v1.json +32 -0
  159. package/docs/evidence/v02/box-heldout-combined-seal-20261021-v1.json +89 -0
  160. package/docs/evidence/v02/box-heldout-evaluator-truth-seal-20261021-v1.json +61 -0
  161. package/docs/evidence/v02/box-heldout-exclusion-audit-20261020-v1.json +81 -0
  162. package/docs/evidence/v02/box-heldout-map-seal-20261021-v1.json +48 -0
  163. package/docs/evidence/v02/box-heldout-photo-seal-20261020-v1.json +62 -0
  164. package/docs/evidence/v02/box-heldout-run-freeze-20261021-v1.json +61 -0
  165. package/docs/evidence/v02/box-heldout-source-review-20261020-v1.json +134 -0
  166. package/docs/evidence/v02/box-heldout-ui-seal-20261020-v1.json +44 -0
  167. package/docs/evidence/v02/calculator-proof-root-review-20261020-v1.json +63 -0
  168. package/docs/evidence/v02/categorical-dev40-freeze.json +94 -0
  169. package/docs/evidence/v02/categorical-dev40-preserve.json +460 -0
  170. package/docs/evidence/v02/categorical-dev40-target-only.json +460 -0
  171. package/docs/evidence/v02/categorical-heldout-admission-20261020-v3.json +215 -0
  172. package/docs/evidence/v02/categorical-heldout-admission-20261020-v4.json +270 -0
  173. package/docs/evidence/v02/categorical-heldout-dispatch-authorization-20261020-v4.json +79 -0
  174. package/docs/evidence/v02/categorical-heldout-predispatch-20261020-v3.json +57 -0
  175. package/docs/evidence/v02/categorical-heldout-producer-source-map-v4.json +262 -0
  176. package/docs/evidence/v02/categorical-heldout-producer-source-map.json +263 -0
  177. package/docs/evidence/v02/categorical-heldout-run-freeze-20261020-v3.json +35 -0
  178. package/docs/evidence/v02/categorical-heldout-run-freeze-20261020-v4.json +35 -0
  179. package/docs/evidence/v02/categorical-heldout-v4/README.md +35 -0
  180. package/docs/evidence/v02/categorical-heldout-v4/candidate-evaluation.json +2812 -0
  181. package/docs/evidence/v02/categorical-heldout-v4/candidate-physical-receipts.json +15261 -0
  182. package/docs/evidence/v02/categorical-heldout-v4/candidate-provenance.json +131 -0
  183. package/docs/evidence/v02/categorical-heldout-v4/control-evaluation.json +855 -0
  184. package/docs/evidence/v02/categorical-heldout-v4/control-physical-receipts.json +4402 -0
  185. package/docs/evidence/v02/categorical-heldout-v4/control-provenance.json +31 -0
  186. package/docs/evidence/v02/categorical-heldout-v4/mask-normalized-exploratory.json +682 -0
  187. package/docs/evidence/v02/categorical-heldout-v4/paired-qualification.json +21 -0
  188. package/docs/evidence/v02/categorical-heldout-v4/report.json +258 -0
  189. package/docs/evidence/v02/categorical-presence-dev6.json +144 -0
  190. package/docs/evidence/v02/categorical-presence-dev6.md +18 -0
  191. package/docs/evidence/v02/categorical-vision-method-selection.json +90 -0
  192. package/docs/evidence/v02/checked-region-component-v1/README.md +7 -0
  193. package/docs/evidence/v02/checked-region-component-v1/evidence.json +311 -0
  194. package/docs/evidence/v02/checked-region-resource-interrupted-2b0d09f.json +102 -0
  195. package/docs/evidence/v02/checked-region-segmentation-resource-4df544c-receipt.json +167 -0
  196. package/docs/evidence/v02/checked-region-segmentation-resource-4df544c-root-review.json +58 -0
  197. package/docs/evidence/v02/checked-region-segmentation-resource-987ac32-receipt.json +158 -0
  198. package/docs/evidence/v02/ci-012b935.json +17 -0
  199. package/docs/evidence/v02/ci-0c365cb.json +372 -0
  200. package/docs/evidence/v02/ci-185bde8.json +32 -0
  201. package/docs/evidence/v02/ci-24f849f.json +95 -0
  202. package/docs/evidence/v02/ci-25c19cd.json +375 -0
  203. package/docs/evidence/v02/ci-2922caf.json +375 -0
  204. package/docs/evidence/v02/ci-29357d7.json +29 -0
  205. package/docs/evidence/v02/ci-32aa90d.json +24 -0
  206. package/docs/evidence/v02/ci-35e3064.json +376 -0
  207. package/docs/evidence/v02/ci-4331b1b.json +384 -0
  208. package/docs/evidence/v02/ci-43b7426.json +376 -0
  209. package/docs/evidence/v02/ci-450439c.json +373 -0
  210. package/docs/evidence/v02/ci-46283fe.json +29 -0
  211. package/docs/evidence/v02/ci-4c9c09f.json +34 -0
  212. package/docs/evidence/v02/ci-5579646.json +27 -0
  213. package/docs/evidence/v02/ci-6f9cc15.json +127 -0
  214. package/docs/evidence/v02/ci-7c1051f.json +27 -0
  215. package/docs/evidence/v02/ci-9837888.json +34 -0
  216. package/docs/evidence/v02/ci-987ac32.json +29 -0
  217. package/docs/evidence/v02/ci-9d00fe5.json +25 -0
  218. package/docs/evidence/v02/ci-9f85acc.json +127 -0
  219. package/docs/evidence/v02/ci-be275a6.json +201 -0
  220. package/docs/evidence/v02/ci-c107ddf.json +372 -0
  221. package/docs/evidence/v02/ci-c35f67d.json +127 -0
  222. package/docs/evidence/v02/ci-c5eddd8.json +32 -0
  223. package/docs/evidence/v02/ci-d2042e0.json +372 -0
  224. package/docs/evidence/v02/ci-d684259.json +375 -0
  225. package/docs/evidence/v02/ci-e0c190b.json +127 -0
  226. package/docs/evidence/v02/ci-e551253.json +372 -0
  227. package/docs/evidence/v02/ci-f9d4633.json +29 -0
  228. package/docs/evidence/v02/dev1-proposal-evaluation.md +13 -0
  229. package/docs/evidence/v02/dev2-mobile-sam-proposal-oracle.md +21 -0
  230. package/docs/evidence/v02/dev2-proposal-evaluation.md +21 -0
  231. package/docs/evidence/v02/fresh-development.md +16 -0
  232. package/docs/evidence/v02/fresh-native-numeric/artifact-hashes.json +9 -0
  233. package/docs/evidence/v02/fresh-native-numeric/evaluation-root-v1.json +568 -0
  234. package/docs/evidence/v02/fresh-native-numeric/run-manifest.json +342 -0
  235. package/docs/evidence/v02/fresh-native-numeric/summary.json +5654 -0
  236. package/docs/evidence/v02/fresh-native-outline-axis/artifact-hashes.json +9 -0
  237. package/docs/evidence/v02/fresh-native-outline-axis/evaluation-root-v1.json +568 -0
  238. package/docs/evidence/v02/fresh-native-outline-axis/run-manifest.json +378 -0
  239. package/docs/evidence/v02/fresh-native-outline-axis/summary.json +5605 -0
  240. package/docs/evidence/v02/fresh-text/README.md +23 -0
  241. package/docs/evidence/v02/fresh-text/ascii-contours-v1/classify-followup-run-manifest.json +405 -0
  242. package/docs/evidence/v02/fresh-text/ascii-contours-v1/classify-followup-summary.json +1646 -0
  243. package/docs/evidence/v02/fresh-text/ascii-contours-v1/evaluation.json +568 -0
  244. package/docs/evidence/v02/fresh-text/ascii-contours-v1/original-run-manifest.json +404 -0
  245. package/docs/evidence/v02/fresh-text/ascii-contours-v1/original-summary.json +4416 -0
  246. package/docs/evidence/v02/fresh-text/ascii-luma-v1/classify-followup-run-manifest.json +405 -0
  247. package/docs/evidence/v02/fresh-text/ascii-luma-v1/classify-followup-summary.json +1646 -0
  248. package/docs/evidence/v02/fresh-text/ascii-luma-v1/evaluation.json +568 -0
  249. package/docs/evidence/v02/fresh-text/ascii-luma-v1/original-run-manifest.json +404 -0
  250. package/docs/evidence/v02/fresh-text/ascii-luma-v1/original-summary.json +4647 -0
  251. package/docs/evidence/v02/fresh-text/ascii-svg-layered-v1/classify-followup-run-manifest.json +405 -0
  252. package/docs/evidence/v02/fresh-text/ascii-svg-layered-v1/classify-followup-summary.json +846 -0
  253. package/docs/evidence/v02/fresh-text/ascii-svg-layered-v1/evaluation.json +568 -0
  254. package/docs/evidence/v02/fresh-text/ascii-svg-layered-v1/original-run-manifest.json +404 -0
  255. package/docs/evidence/v02/fresh-text/ascii-svg-layered-v1/original-summary.json +1789 -0
  256. package/docs/evidence/v02/fresh-text/no-image-control/evaluation.json +568 -0
  257. package/docs/evidence/v02/fresh-text/no-image-control/original-run-manifest.json +404 -0
  258. package/docs/evidence/v02/fresh-text/no-image-control/original-summary.json +3953 -0
  259. package/docs/evidence/v02/fresh-text/palette-grid-v1/classify-followup-run-manifest.json +405 -0
  260. package/docs/evidence/v02/fresh-text/palette-grid-v1/classify-followup-summary.json +1646 -0
  261. package/docs/evidence/v02/fresh-text/palette-grid-v1/evaluation.json +568 -0
  262. package/docs/evidence/v02/fresh-text/palette-grid-v1/original-run-manifest.json +404 -0
  263. package/docs/evidence/v02/fresh-text/palette-grid-v1/original-summary.json +3974 -0
  264. package/docs/evidence/v02/fresh-text/pixel-rle-v1/classify-followup-run-manifest.json +405 -0
  265. package/docs/evidence/v02/fresh-text/pixel-rle-v1/classify-followup-summary.json +1646 -0
  266. package/docs/evidence/v02/fresh-text/pixel-rle-v1/evaluation.json +568 -0
  267. package/docs/evidence/v02/fresh-text/pixel-rle-v1/original-run-manifest.json +404 -0
  268. package/docs/evidence/v02/fresh-text/pixel-rle-v1/original-summary.json +4945 -0
  269. package/docs/evidence/v02/fresh-text/report.json +284 -0
  270. package/docs/evidence/v02/fresh-text/svg-contours-v1/classify-followup-run-manifest.json +405 -0
  271. package/docs/evidence/v02/fresh-text/svg-contours-v1/classify-followup-summary.json +846 -0
  272. package/docs/evidence/v02/fresh-text/svg-contours-v1/evaluation.json +568 -0
  273. package/docs/evidence/v02/fresh-text/svg-contours-v1/original-run-manifest.json +404 -0
  274. package/docs/evidence/v02/fresh-text/svg-contours-v1/original-summary.json +2189 -0
  275. package/docs/evidence/v02/heldout-evaluation-20261020.md +37 -0
  276. package/docs/evidence/v02/heldout-input-derivation.md +18 -0
  277. package/docs/evidence/v02/integration-0c365cb.json +90 -0
  278. package/docs/evidence/v02/integration-25c19cd.json +52 -0
  279. package/docs/evidence/v02/integration-35e3064.json +41 -0
  280. package/docs/evidence/v02/integration-4331b1b.json +42 -0
  281. package/docs/evidence/v02/integration-4c9c09f.json +46 -0
  282. package/docs/evidence/v02/integration-8010aad.json +48 -0
  283. package/docs/evidence/v02/integration-c1aa640.json +41 -0
  284. package/docs/evidence/v02/integration-eea1790.json +46 -0
  285. package/docs/evidence/v02/integration-fa82e65.json +42 -0
  286. package/docs/evidence/v02/jpeg-transport-offline.json +268 -0
  287. package/docs/evidence/v02/jpeg-transport-offline.md +7 -0
  288. package/docs/evidence/v02/local-archive-cleanup.json +59 -0
  289. package/docs/evidence/v02/native-base64-after.json +30 -0
  290. package/docs/evidence/v02/native-base64-before.json +32 -0
  291. package/docs/evidence/v02/native-checked-region-windows-46283fe.json +248 -0
  292. package/docs/evidence/v02/native-checked-region-windows-987ac32.json +141 -0
  293. package/docs/evidence/v02/native-grabcut-windows-4c494a3-run2.json +148 -0
  294. package/docs/evidence/v02/native-grabcut-windows-4c494a3-run3.json +148 -0
  295. package/docs/evidence/v02/native-grabcut-windows-4c494a3.json +56 -0
  296. package/docs/evidence/v02/native-proposal-windows-0c365cb.json +148 -0
  297. package/docs/evidence/v02/native-proposal-windows-450439c.json +156 -0
  298. package/docs/evidence/v02/native-proposal-windows-4c9c09f-retry1.json +147 -0
  299. package/docs/evidence/v02/native-proposal-windows-4c9c09f.json +52 -0
  300. package/docs/evidence/v02/native-region-windows-efc9318.json +163 -0
  301. package/docs/evidence/v02/native-windows-eea1790-baseline.json +198 -0
  302. package/docs/evidence/v02/native-windows-eea1790.json +228 -0
  303. package/docs/evidence/v02/ordinary-luna-axis/run-manifest.json +99 -0
  304. package/docs/evidence/v02/ordinary-luna-axis/summary.json +811 -0
  305. package/docs/evidence/v02/owned-temp-cleanup-20261020-v1.json +28 -0
  306. package/docs/evidence/v02/presentation-paired-20261020.json +1438 -0
  307. package/docs/evidence/v02/proposal-resource-soak-261dc23.json +201 -0
  308. package/docs/evidence/v02/proposal-resource-soak-261dc23.md +15 -0
  309. package/docs/evidence/v02/proposal-resource-soak-450439c-receipt.md +13 -0
  310. package/docs/evidence/v02/reconstructed-c03-box.json +41 -0
  311. package/docs/evidence/v02/region-seeded-dev-v1/README.md +17 -0
  312. package/docs/evidence/v02/region-seeded-dev-v1/evaluation.json +260 -0
  313. package/docs/evidence/v02/region-seeded-dev-v2/README.md +15 -0
  314. package/docs/evidence/v02/region-seeded-dev-v2/evaluation.json +232 -0
  315. package/docs/evidence/v02/region-seeded-dev-v3-candidate-only/README.md +9 -0
  316. package/docs/evidence/v02/region-seeded-dev-v3-candidate-only/control-identity-failure.json +22 -0
  317. package/docs/evidence/v02/region-seeded-dev-v3-candidate-only/evaluation.json +160 -0
  318. package/docs/evidence/v02/region-seeded-dev-v3-control-r2/README.md +9 -0
  319. package/docs/evidence/v02/region-seeded-dev-v3-control-r2/attempt-accounting.json +55 -0
  320. package/docs/evidence/v02/region-seeded-dev-v3-control-r2/evaluation.json +136 -0
  321. package/docs/evidence/v02/region-seeded-development-control-r2-freeze-20261021.json +132 -0
  322. package/docs/evidence/v02/region-seeded-development-freeze-20261021-v1.json +167 -0
  323. package/docs/evidence/v02/region-seeded-development-freeze-20261021-v2.json +186 -0
  324. package/docs/evidence/v02/region-seeded-development-freeze-20261021-v3.json +205 -0
  325. package/docs/evidence/v02/region-seeded-heldout-cancelled-before-dispatch-20261021-v2.json +20 -0
  326. package/docs/evidence/v02/region-seeded-heldout-dispatch-authorization-20261021-v3.json +47 -0
  327. package/docs/evidence/v02/region-seeded-heldout-readiness-audit-20261021-v3.json +125 -0
  328. package/docs/evidence/v02/region-seeded-heldout-run-freeze-20261021-v2.json +234 -0
  329. package/docs/evidence/v02/region-seeded-heldout-run-freeze-20261021-v3.json +357 -0
  330. package/docs/evidence/v02/region-seeded-heldout-v3/README.md +19 -0
  331. package/docs/evidence/v02/region-seeded-heldout-v3/evaluation-summary.json +445 -0
  332. package/docs/evidence/v02/region-seeded-method-selection-20261021-v2.json +66 -0
  333. package/docs/evidence/v02/region-seeded-method-selection-20261021-v3.json +120 -0
  334. package/docs/evidence/v02/region-segmentation-resource-0ba4c87-receipt.json +156 -0
  335. package/docs/evidence/v02/resource-soak-linux-8010aad.json +125 -0
  336. package/docs/evidence/v02/screen-v2-terminal-evaluation.md +22 -0
  337. package/docs/evidence/v02/seed-outline/run-manifest.json +59 -0
  338. package/docs/evidence/v02/seed-outline/summary.json +487 -0
  339. package/docs/evidence/v02/seed-unbounded/run-manifest.json +80 -0
  340. package/docs/evidence/v02/seed-unbounded/summary.json +730 -0
  341. package/docs/evidence/v02/segmentation/README.md +21 -0
  342. package/docs/evidence/v02/segmentation/component-comparison-summary.json +240 -0
  343. package/docs/evidence/v02/segmentation/provenance/MOBILE-SAM-LICENSE.txt +201 -0
  344. package/docs/evidence/v02/segmentation/provenance/OPENCV-LICENSE.txt +202 -0
  345. package/docs/evidence/v02/segmentation/provenance/mobile-sam-download-receipt.json +29 -0
  346. package/docs/evidence/v02/segmentation/provenance/mobile-sam-graph-audit.json +214 -0
  347. package/docs/evidence/v02/segmentation/provenance/mobile-sam-model-metadata.json +166 -0
  348. package/docs/evidence/v02/split-choice-noul-dev6.json +72 -0
  349. package/docs/evidence/v02/split-choice-noul-dev6.md +11 -0
  350. package/docs/evidence/v02/supported-vision-screen-20261020.md +65 -0
  351. package/docs/evidence/v02/wesnoth-1.16.11-minimap-projection-source-audit.md +56 -0
  352. package/docs/evidence/v02/wesnoth-proof-root-review-20261020-v1.json +76 -0
  353. package/docs/evidence/vision-cost-reconciliation.json +37 -0
  354. package/docs/evidence/vision-model-summary.json +562 -0
  355. package/docs/evidence/vision-selection-v1-manifest.json +95 -0
  356. package/docs/evidence/vision-selection-v1-report.json +4598 -0
  357. package/docs/evidence/vision-selection-v2-manifest.json +95 -0
  358. package/docs/evidence/vision-selection-v2-report.json +1139 -0
  359. package/docs/evidence/windows-native.json +73 -0
  360. package/docs/evidence/windows-secrets.json +46 -0
  361. package/docs/operations.md +21 -0
  362. package/docs/plan.md +27 -0
  363. package/docs/provenance.json +32 -0
  364. package/docs/publishing.md +30 -0
  365. package/docs/research/text-transport-audit.md +66 -0
  366. package/docs/results.md +69 -0
  367. package/docs/v02-usage.md +162 -0
  368. package/docs/vision-results.md +78 -0
  369. package/package.json +114 -4
@@ -0,0 +1,1205 @@
1
+ {
2
+ "status": "complete",
3
+ "startedAt": "2026-10-17T22:59:39.495Z",
4
+ "finishedAt": "2026-10-17T23:00:11.001Z",
5
+ "policy": "Complete the missing20 development photos and20 heldout texts for the same frozen Luna Decisions classifier; earlier40 heldout photos and10 development texts are retained, never replayed or repaired. No inference/prompt changes. At most40 new classification calls under expanded1000 aggregate request limit and unchanged $0.50 cap. No retries. Preserve previous locator provider error.",
6
+ "finalists": ["luna-decisions"],
7
+ "selected": null,
8
+ "summaries": [
9
+ {
10
+ "route": "luna-decisions",
11
+ "model": "openai/gpt-6-luna-decisions",
12
+ "provider": "openai",
13
+ "photos": 20,
14
+ "photoCorrect": 16,
15
+ "heldout": 0,
16
+ "heldoutCorrect": 0,
17
+ "texts": 20,
18
+ "textCorrect": 20,
19
+ "errors": 0,
20
+ "knownCostUsd": 0.003345499999999999,
21
+ "knownCostCount": 40,
22
+ "priceIndex": 0.1
23
+ }
24
+ ],
25
+ "physicalCalls": 40,
26
+ "aggregatePhysicalCalls": 328,
27
+ "prior": {
28
+ "sha256": "25c463dc9b005c2702dd94e78b12cee861e5f27827c5093cda857771837b19c7",
29
+ "physicalCalls": 288,
30
+ "budget": {
31
+ "admittedProviderCalls": 288,
32
+ "maxCalls": 1000,
33
+ "maxCostUsd": 0.5,
34
+ "knownCostUsd": 0.029322660000000056,
35
+ "unknownCostCalls": 3,
36
+ "inFlightCalls": 0,
37
+ "reservationPerUnknownUsd": 0.05,
38
+ "accountedExposureUsd": 0.17932266000000008,
39
+ "exceededByReportedCost": false
40
+ }
41
+ },
42
+ "protocol": "native-choice-v1: exact criteria and raw distributions; no chat rank protocol",
43
+ "budget": {
44
+ "admittedProviderCalls": 328,
45
+ "maxCalls": 1000,
46
+ "maxCostUsd": 0.5,
47
+ "knownCostUsd": 0.03266816000000008,
48
+ "unknownCostCalls": 3,
49
+ "inFlightCalls": 0,
50
+ "reservationPerUnknownUsd": 0.05,
51
+ "accountedExposureUsd": 0.1826681600000001,
52
+ "exceededByReportedCost": false
53
+ },
54
+ "stopped": null,
55
+ "rows": [
56
+ {
57
+ "route": "luna-decisions",
58
+ "caseId": "photo-dev-01",
59
+ "split": "dev",
60
+ "kind": "photo",
61
+ "durationMs": 1441.0478480000002,
62
+ "result": {
63
+ "choice": "absent",
64
+ "nativeChoice": "absent",
65
+ "probabilities": {
66
+ "present": 0.34,
67
+ "absent": 0.66
68
+ },
69
+ "tied": ["absent"],
70
+ "confidence": 0.32,
71
+ "metadata": {
72
+ "model": "openai/gpt-6-luna-decisions-20261006",
73
+ "provider": "OpenAI",
74
+ "requestId": "gen-dec-1792277982-Qnm7RpAFI1HTFdeTKjM8",
75
+ "durationMs": 1281.445572,
76
+ "costUsd": 0.0001168,
77
+ "attempts": 1
78
+ }
79
+ },
80
+ "expected": "present",
81
+ "correct": false
82
+ },
83
+ {
84
+ "route": "luna-decisions",
85
+ "caseId": "photo-dev-02",
86
+ "split": "dev",
87
+ "kind": "photo",
88
+ "durationMs": 711.966296,
89
+ "result": {
90
+ "choice": "present",
91
+ "nativeChoice": "present",
92
+ "probabilities": {
93
+ "present": 1,
94
+ "absent": 0
95
+ },
96
+ "tied": ["present"],
97
+ "confidence": 1,
98
+ "metadata": {
99
+ "model": "openai/gpt-6-luna-decisions-20261006",
100
+ "provider": "OpenAI",
101
+ "requestId": "gen-dec-1792277983-d3OyMSXttp5rQLX7SdBo",
102
+ "durationMs": 618.9590979999998,
103
+ "costUsd": 0.0000944,
104
+ "attempts": 1
105
+ }
106
+ },
107
+ "expected": "present",
108
+ "correct": true
109
+ },
110
+ {
111
+ "route": "luna-decisions",
112
+ "caseId": "photo-dev-03",
113
+ "split": "dev",
114
+ "kind": "photo",
115
+ "durationMs": 435.90419999999995,
116
+ "result": {
117
+ "choice": "present",
118
+ "nativeChoice": "present",
119
+ "probabilities": {
120
+ "present": 1,
121
+ "absent": 0
122
+ },
123
+ "tied": ["present"],
124
+ "confidence": 1,
125
+ "metadata": {
126
+ "model": "openai/gpt-6-luna-decisions-20261006",
127
+ "provider": "OpenAI",
128
+ "requestId": "gen-dec-1792277984-QUZCsHthIPNOdeb8jPYx",
129
+ "durationMs": 382.97728000000006,
130
+ "costUsd": 0.0000816,
131
+ "attempts": 1
132
+ }
133
+ },
134
+ "expected": "present",
135
+ "correct": true
136
+ },
137
+ {
138
+ "route": "luna-decisions",
139
+ "caseId": "photo-dev-04",
140
+ "split": "dev",
141
+ "kind": "photo",
142
+ "durationMs": 829.5554400000001,
143
+ "result": {
144
+ "choice": "present",
145
+ "nativeChoice": "present",
146
+ "probabilities": {
147
+ "present": 1,
148
+ "absent": 0
149
+ },
150
+ "tied": ["present"],
151
+ "confidence": 1,
152
+ "metadata": {
153
+ "model": "openai/gpt-6-luna-decisions-20261006",
154
+ "provider": "OpenAI",
155
+ "requestId": "gen-dec-1792277984-HAE42nTS2M5ZFFUrqtVa",
156
+ "durationMs": 751.0542809999997,
157
+ "costUsd": 0.0000944,
158
+ "attempts": 1
159
+ }
160
+ },
161
+ "expected": "present",
162
+ "correct": true
163
+ },
164
+ {
165
+ "route": "luna-decisions",
166
+ "caseId": "photo-dev-05",
167
+ "split": "dev",
168
+ "kind": "photo",
169
+ "durationMs": 730.5974340000002,
170
+ "result": {
171
+ "choice": "present",
172
+ "nativeChoice": "present",
173
+ "probabilities": {
174
+ "present": 0.99,
175
+ "absent": 0.01
176
+ },
177
+ "tied": ["present"],
178
+ "confidence": 0.98,
179
+ "metadata": {
180
+ "model": "openai/gpt-6-luna-decisions-20261006",
181
+ "provider": "OpenAI",
182
+ "requestId": "gen-dec-1792277985-nno3rG444Zs4ZN6Ilr8b",
183
+ "durationMs": 637.6476969999999,
184
+ "costUsd": 0.0001008,
185
+ "attempts": 1
186
+ }
187
+ },
188
+ "expected": "absent",
189
+ "correct": false
190
+ },
191
+ {
192
+ "route": "luna-decisions",
193
+ "caseId": "photo-dev-06",
194
+ "split": "dev",
195
+ "kind": "photo",
196
+ "durationMs": 872.4963850000004,
197
+ "result": {
198
+ "choice": "present",
199
+ "nativeChoice": "present",
200
+ "probabilities": {
201
+ "present": 0.73,
202
+ "absent": 0.27
203
+ },
204
+ "tied": ["present"],
205
+ "confidence": 0.46,
206
+ "metadata": {
207
+ "model": "openai/gpt-6-luna-decisions-20261006",
208
+ "provider": "OpenAI",
209
+ "requestId": "gen-dec-1792277986-tKwOEZvbBbktubz05DCH",
210
+ "durationMs": 773.8749189999999,
211
+ "costUsd": 0.0001008,
212
+ "attempts": 1
213
+ }
214
+ },
215
+ "expected": "absent",
216
+ "correct": false
217
+ },
218
+ {
219
+ "route": "luna-decisions",
220
+ "caseId": "photo-dev-07",
221
+ "split": "dev",
222
+ "kind": "photo",
223
+ "durationMs": 857.2547080000004,
224
+ "result": {
225
+ "choice": "absent",
226
+ "nativeChoice": "absent",
227
+ "probabilities": {
228
+ "present": 0.09,
229
+ "absent": 0.91
230
+ },
231
+ "tied": ["absent"],
232
+ "confidence": 0.82,
233
+ "metadata": {
234
+ "model": "openai/gpt-6-luna-decisions-20261006",
235
+ "provider": "OpenAI",
236
+ "requestId": "gen-dec-1792277987-TcvXMgDdcLpF30lcpL1q",
237
+ "durationMs": 760.0918329999995,
238
+ "costUsd": 0.0001008,
239
+ "attempts": 1
240
+ }
241
+ },
242
+ "expected": "absent",
243
+ "correct": true
244
+ },
245
+ {
246
+ "route": "luna-decisions",
247
+ "caseId": "photo-dev-08",
248
+ "split": "dev",
249
+ "kind": "photo",
250
+ "durationMs": 451.9101609999998,
251
+ "result": {
252
+ "choice": "absent",
253
+ "nativeChoice": "absent",
254
+ "probabilities": {
255
+ "present": 0.02,
256
+ "absent": 0.98
257
+ },
258
+ "tied": ["absent"],
259
+ "confidence": 0.96,
260
+ "metadata": {
261
+ "model": "openai/gpt-6-luna-decisions-20261006",
262
+ "provider": "OpenAI",
263
+ "requestId": "gen-dec-1792277988-Y0iQJJuRhbmjhPO7EyTL",
264
+ "durationMs": 404.29795900000045,
265
+ "costUsd": 0.0000784,
266
+ "attempts": 1
267
+ }
268
+ },
269
+ "expected": "absent",
270
+ "correct": true
271
+ },
272
+ {
273
+ "route": "luna-decisions",
274
+ "caseId": "photo-dev-09",
275
+ "split": "dev",
276
+ "kind": "photo",
277
+ "durationMs": 931.6173310000004,
278
+ "result": {
279
+ "choice": "present",
280
+ "nativeChoice": "present",
281
+ "probabilities": {
282
+ "present": 1,
283
+ "absent": 0
284
+ },
285
+ "tied": ["present"],
286
+ "confidence": 1,
287
+ "metadata": {
288
+ "model": "openai/gpt-6-luna-decisions-20261006",
289
+ "provider": "OpenAI",
290
+ "requestId": "gen-dec-1792277989-91I9CEsi8n9ZF1WX5MTL",
291
+ "durationMs": 814.3404379999993,
292
+ "costUsd": 0.0000976,
293
+ "attempts": 1
294
+ }
295
+ },
296
+ "expected": "present",
297
+ "correct": true
298
+ },
299
+ {
300
+ "route": "luna-decisions",
301
+ "caseId": "photo-dev-10",
302
+ "split": "dev",
303
+ "kind": "photo",
304
+ "durationMs": 687.858107,
305
+ "result": {
306
+ "choice": "present",
307
+ "nativeChoice": "present",
308
+ "probabilities": {
309
+ "present": 0.97,
310
+ "absent": 0.03
311
+ },
312
+ "tied": ["present"],
313
+ "confidence": 0.94,
314
+ "metadata": {
315
+ "model": "openai/gpt-6-luna-decisions-20261006",
316
+ "provider": "OpenAI",
317
+ "requestId": "gen-dec-1792277990-uH4VeftKG6bZfAxeyBID",
318
+ "durationMs": 603.3556960000005,
319
+ "costUsd": 0.0000944,
320
+ "attempts": 1
321
+ }
322
+ },
323
+ "expected": "present",
324
+ "correct": true
325
+ },
326
+ {
327
+ "route": "luna-decisions",
328
+ "caseId": "photo-dev-11",
329
+ "split": "dev",
330
+ "kind": "photo",
331
+ "durationMs": 714.497784000001,
332
+ "result": {
333
+ "choice": "present",
334
+ "nativeChoice": "present",
335
+ "probabilities": {
336
+ "present": 1,
337
+ "absent": 0
338
+ },
339
+ "tied": ["present"],
340
+ "confidence": 1,
341
+ "metadata": {
342
+ "model": "openai/gpt-6-luna-decisions-20261006",
343
+ "provider": "OpenAI",
344
+ "requestId": "gen-dec-1792277990-oEG4bQ4dEWiuc1OYH7HO",
345
+ "durationMs": 625.7863370000014,
346
+ "costUsd": 0.0001008,
347
+ "attempts": 1
348
+ }
349
+ },
350
+ "expected": "present",
351
+ "correct": true
352
+ },
353
+ {
354
+ "route": "luna-decisions",
355
+ "caseId": "photo-dev-12",
356
+ "split": "dev",
357
+ "kind": "photo",
358
+ "durationMs": 758.023733,
359
+ "result": {
360
+ "choice": "present",
361
+ "nativeChoice": "present",
362
+ "probabilities": {
363
+ "present": 0.99,
364
+ "absent": 0.01
365
+ },
366
+ "tied": ["present"],
367
+ "confidence": 0.98,
368
+ "metadata": {
369
+ "model": "openai/gpt-6-luna-decisions-20261006",
370
+ "provider": "OpenAI",
371
+ "requestId": "gen-dec-1792277991-AvOLoFhKO36lTltlBp76",
372
+ "durationMs": 672.4401229999985,
373
+ "costUsd": 0.0000944,
374
+ "attempts": 1
375
+ }
376
+ },
377
+ "expected": "present",
378
+ "correct": true
379
+ },
380
+ {
381
+ "route": "luna-decisions",
382
+ "caseId": "photo-dev-13",
383
+ "split": "dev",
384
+ "kind": "photo",
385
+ "durationMs": 1040.0477920000012,
386
+ "result": {
387
+ "choice": "present",
388
+ "nativeChoice": "present",
389
+ "probabilities": {
390
+ "present": 1,
391
+ "absent": 0
392
+ },
393
+ "tied": ["present"],
394
+ "confidence": 1,
395
+ "metadata": {
396
+ "model": "openai/gpt-6-luna-decisions-20261006",
397
+ "provider": "OpenAI",
398
+ "requestId": "gen-dec-1792277993-lGFprmry4xK6q2NVJVgp",
399
+ "durationMs": 961.8451479999985,
400
+ "costUsd": 0.0000944,
401
+ "attempts": 1
402
+ }
403
+ },
404
+ "expected": "present",
405
+ "correct": true
406
+ },
407
+ {
408
+ "route": "luna-decisions",
409
+ "caseId": "photo-dev-14",
410
+ "split": "dev",
411
+ "kind": "photo",
412
+ "durationMs": 1303.43606,
413
+ "result": {
414
+ "choice": "present",
415
+ "nativeChoice": "present",
416
+ "probabilities": {
417
+ "present": 1,
418
+ "absent": 0
419
+ },
420
+ "tied": ["present"],
421
+ "confidence": 1,
422
+ "metadata": {
423
+ "model": "openai/gpt-6-luna-decisions-20261006",
424
+ "provider": "OpenAI",
425
+ "requestId": "gen-dec-1792277994-pgpOqWWsjllQkKRTxbS0",
426
+ "durationMs": 1203.900878999999,
427
+ "costUsd": 0.0001104,
428
+ "attempts": 1
429
+ }
430
+ },
431
+ "expected": "present",
432
+ "correct": true
433
+ },
434
+ {
435
+ "route": "luna-decisions",
436
+ "caseId": "photo-dev-15",
437
+ "split": "dev",
438
+ "kind": "photo",
439
+ "durationMs": 1968.3092730000008,
440
+ "result": {
441
+ "choice": "present",
442
+ "nativeChoice": "present",
443
+ "probabilities": {
444
+ "present": 1,
445
+ "absent": 0
446
+ },
447
+ "tied": ["present"],
448
+ "confidence": 1,
449
+ "metadata": {
450
+ "model": "openai/gpt-6-luna-decisions-20261006",
451
+ "provider": "OpenAI",
452
+ "requestId": "gen-dec-1792277996-f24uU0nzr5JUY73N7p7G",
453
+ "durationMs": 1825.6368390000007,
454
+ "costUsd": 0.0001267,
455
+ "attempts": 1
456
+ }
457
+ },
458
+ "expected": "present",
459
+ "correct": true
460
+ },
461
+ {
462
+ "route": "luna-decisions",
463
+ "caseId": "photo-dev-16",
464
+ "split": "dev",
465
+ "kind": "photo",
466
+ "durationMs": 1211.1612849999983,
467
+ "result": {
468
+ "choice": "present",
469
+ "nativeChoice": "present",
470
+ "probabilities": {
471
+ "present": 0.99,
472
+ "absent": 0.01
473
+ },
474
+ "tied": ["present"],
475
+ "confidence": 0.98,
476
+ "metadata": {
477
+ "model": "openai/gpt-6-luna-decisions-20261006",
478
+ "provider": "OpenAI",
479
+ "requestId": "gen-dec-1792277997-3wEJmzE58zbMS9OcF0d7",
480
+ "durationMs": 1119.4658999999992,
481
+ "costUsd": 0.0001008,
482
+ "attempts": 1
483
+ }
484
+ },
485
+ "expected": "present",
486
+ "correct": true
487
+ },
488
+ {
489
+ "route": "luna-decisions",
490
+ "caseId": "photo-dev-17",
491
+ "split": "dev",
492
+ "kind": "photo",
493
+ "durationMs": 2455.050903999998,
494
+ "result": {
495
+ "choice": "absent",
496
+ "nativeChoice": "absent",
497
+ "probabilities": {
498
+ "present": 0,
499
+ "absent": 1
500
+ },
501
+ "tied": ["absent"],
502
+ "confidence": 1,
503
+ "metadata": {
504
+ "model": "openai/gpt-6-luna-decisions-20261006",
505
+ "provider": "OpenAI",
506
+ "requestId": "gen-dec-1792278000-6BrS1wCG4l3wzLzTFXEV",
507
+ "durationMs": 2333.680232999999,
508
+ "costUsd": 0.0001008,
509
+ "attempts": 1
510
+ }
511
+ },
512
+ "expected": "present",
513
+ "correct": false
514
+ },
515
+ {
516
+ "route": "luna-decisions",
517
+ "caseId": "photo-dev-18",
518
+ "split": "dev",
519
+ "kind": "photo",
520
+ "durationMs": 2094.691584,
521
+ "result": {
522
+ "choice": "present",
523
+ "nativeChoice": "present",
524
+ "probabilities": {
525
+ "present": 1,
526
+ "absent": 0
527
+ },
528
+ "tied": ["present"],
529
+ "confidence": 1,
530
+ "metadata": {
531
+ "model": "openai/gpt-6-luna-decisions-20261006",
532
+ "provider": "OpenAI",
533
+ "requestId": "gen-dec-1792278002-XktG6l0k6ntzKhIj61uX",
534
+ "durationMs": 1989.4556279999997,
535
+ "costUsd": 0.0001008,
536
+ "attempts": 1
537
+ }
538
+ },
539
+ "expected": "present",
540
+ "correct": true
541
+ },
542
+ {
543
+ "route": "luna-decisions",
544
+ "caseId": "photo-dev-19",
545
+ "split": "dev",
546
+ "kind": "photo",
547
+ "durationMs": 1632.8159440000018,
548
+ "result": {
549
+ "choice": "present",
550
+ "nativeChoice": "present",
551
+ "probabilities": {
552
+ "present": 1,
553
+ "absent": 0
554
+ },
555
+ "tied": ["present"],
556
+ "confidence": 1,
557
+ "metadata": {
558
+ "model": "openai/gpt-6-luna-decisions-20261006",
559
+ "provider": "OpenAI",
560
+ "requestId": "gen-dec-1792278004-7sElRmPIt0bHi4Tmrawo",
561
+ "durationMs": 1552.7531120000021,
562
+ "costUsd": 0.0000944,
563
+ "attempts": 1
564
+ }
565
+ },
566
+ "expected": "present",
567
+ "correct": true
568
+ },
569
+ {
570
+ "route": "luna-decisions",
571
+ "caseId": "photo-dev-20",
572
+ "split": "dev",
573
+ "kind": "photo",
574
+ "durationMs": 2821.4310640000003,
575
+ "result": {
576
+ "choice": "present",
577
+ "nativeChoice": "present",
578
+ "probabilities": {
579
+ "present": 1,
580
+ "absent": 0
581
+ },
582
+ "tied": ["present"],
583
+ "confidence": 1,
584
+ "metadata": {
585
+ "model": "openai/gpt-6-luna-decisions-20261006",
586
+ "provider": "OpenAI",
587
+ "requestId": "gen-dec-1792278007-54VNatHvi84S5mtcJD1n",
588
+ "durationMs": 2705.8991290000013,
589
+ "costUsd": 0.0001008,
590
+ "attempts": 1
591
+ }
592
+ },
593
+ "expected": "present",
594
+ "correct": true
595
+ },
596
+ {
597
+ "route": "luna-decisions",
598
+ "caseId": "support-heldout-001",
599
+ "split": "heldout",
600
+ "kind": "text",
601
+ "durationMs": 246.913195000001,
602
+ "result": {
603
+ "choice": "order_status",
604
+ "nativeChoice": "order_status",
605
+ "probabilities": {
606
+ "order_status": 0.98,
607
+ "refund": 0.01,
608
+ "technical_support": 0,
609
+ "general_feedback": 0,
610
+ "unknown": 0.01
611
+ },
612
+ "tied": ["order_status"],
613
+ "confidence": 0.98,
614
+ "metadata": {
615
+ "model": "openai/gpt-6-luna-decisions-20261006",
616
+ "provider": "OpenAI",
617
+ "requestId": "gen-dec-1792278008-pWneBZUBJFTLvOeRhkGT",
618
+ "durationMs": 245.10594499999934,
619
+ "costUsd": 0.000068,
620
+ "attempts": 1
621
+ }
622
+ },
623
+ "expected": "order_status",
624
+ "correct": true
625
+ },
626
+ {
627
+ "route": "luna-decisions",
628
+ "caseId": "support-heldout-002",
629
+ "split": "heldout",
630
+ "kind": "text",
631
+ "durationMs": 303.3520260000005,
632
+ "result": {
633
+ "choice": "order_status",
634
+ "nativeChoice": "order_status",
635
+ "probabilities": {
636
+ "order_status": 0.89,
637
+ "refund": 0.02,
638
+ "technical_support": 0.02,
639
+ "general_feedback": 0.01,
640
+ "unknown": 0.06
641
+ },
642
+ "tied": ["order_status"],
643
+ "confidence": 0.86,
644
+ "metadata": {
645
+ "model": "openai/gpt-6-luna-decisions-20261006",
646
+ "provider": "OpenAI",
647
+ "requestId": "gen-dec-1792278008-XL4tT0ffc5ygmwnnvQlP",
648
+ "durationMs": 301.9031750000031,
649
+ "costUsd": 0.0000674,
650
+ "attempts": 1
651
+ }
652
+ },
653
+ "expected": "order_status",
654
+ "correct": true
655
+ },
656
+ {
657
+ "route": "luna-decisions",
658
+ "caseId": "support-heldout-003",
659
+ "split": "heldout",
660
+ "kind": "text",
661
+ "durationMs": 201.33628699999826,
662
+ "result": {
663
+ "choice": "refund",
664
+ "nativeChoice": "refund",
665
+ "probabilities": {
666
+ "order_status": 0.01,
667
+ "refund": 0.97,
668
+ "technical_support": 0.01,
669
+ "general_feedback": 0,
670
+ "unknown": 0.01
671
+ },
672
+ "tied": ["refund"],
673
+ "confidence": 0.96,
674
+ "metadata": {
675
+ "model": "openai/gpt-6-luna-decisions-20261006",
676
+ "provider": "OpenAI",
677
+ "requestId": "gen-dec-1792278009-swByykVuTJ0N1RlwGWZ9",
678
+ "durationMs": 199.69924299999911,
679
+ "costUsd": 0.0000686,
680
+ "attempts": 1
681
+ }
682
+ },
683
+ "expected": "refund",
684
+ "correct": true
685
+ },
686
+ {
687
+ "route": "luna-decisions",
688
+ "caseId": "support-heldout-004",
689
+ "split": "heldout",
690
+ "kind": "text",
691
+ "durationMs": 260.627246,
692
+ "result": {
693
+ "choice": "refund",
694
+ "nativeChoice": "refund",
695
+ "probabilities": {
696
+ "order_status": 0,
697
+ "refund": 1,
698
+ "technical_support": 0,
699
+ "general_feedback": 0,
700
+ "unknown": 0
701
+ },
702
+ "tied": ["refund"],
703
+ "confidence": 1,
704
+ "metadata": {
705
+ "model": "openai/gpt-6-luna-decisions-20261006",
706
+ "provider": "OpenAI",
707
+ "requestId": "gen-dec-1792278009-Ey1aVw3RDCu3timdNjAN",
708
+ "durationMs": 259.2615509999996,
709
+ "costUsd": 0.0000674,
710
+ "attempts": 1
711
+ }
712
+ },
713
+ "expected": "refund",
714
+ "correct": true
715
+ },
716
+ {
717
+ "route": "luna-decisions",
718
+ "caseId": "support-heldout-005",
719
+ "split": "heldout",
720
+ "kind": "text",
721
+ "durationMs": 195.35287799999787,
722
+ "result": {
723
+ "choice": "technical_support",
724
+ "nativeChoice": "technical_support",
725
+ "probabilities": {
726
+ "order_status": 0,
727
+ "refund": 0,
728
+ "technical_support": 1,
729
+ "general_feedback": 0,
730
+ "unknown": 0
731
+ },
732
+ "tied": ["technical_support"],
733
+ "confidence": 1,
734
+ "metadata": {
735
+ "model": "openai/gpt-6-luna-decisions-20261006",
736
+ "provider": "OpenAI",
737
+ "requestId": "gen-dec-1792278009-CwKZmy1GRKHiKB8Y75Rq",
738
+ "durationMs": 194.00128199999745,
739
+ "costUsd": 0.0000683,
740
+ "attempts": 1
741
+ }
742
+ },
743
+ "expected": "technical_support",
744
+ "correct": true
745
+ },
746
+ {
747
+ "route": "luna-decisions",
748
+ "caseId": "support-heldout-006",
749
+ "split": "heldout",
750
+ "kind": "text",
751
+ "durationMs": 216.72845600000073,
752
+ "result": {
753
+ "choice": "technical_support",
754
+ "nativeChoice": "technical_support",
755
+ "probabilities": {
756
+ "order_status": 0,
757
+ "refund": 0,
758
+ "technical_support": 1,
759
+ "general_feedback": 0,
760
+ "unknown": 0
761
+ },
762
+ "tied": ["technical_support"],
763
+ "confidence": 1,
764
+ "metadata": {
765
+ "model": "openai/gpt-6-luna-decisions-20261006",
766
+ "provider": "OpenAI",
767
+ "requestId": "gen-dec-1792278009-xhyunE3yIxzjxRqeYvN3",
768
+ "durationMs": 215.53119000000152,
769
+ "costUsd": 0.0000678,
770
+ "attempts": 1
771
+ }
772
+ },
773
+ "expected": "technical_support",
774
+ "correct": true
775
+ },
776
+ {
777
+ "route": "luna-decisions",
778
+ "caseId": "support-heldout-007",
779
+ "split": "heldout",
780
+ "kind": "text",
781
+ "durationMs": 193.41025900000022,
782
+ "result": {
783
+ "choice": "general_feedback",
784
+ "nativeChoice": "general_feedback",
785
+ "probabilities": {
786
+ "order_status": 0,
787
+ "refund": 0,
788
+ "technical_support": 0,
789
+ "general_feedback": 0.99,
790
+ "unknown": 0.01
791
+ },
792
+ "tied": ["general_feedback"],
793
+ "confidence": 0.99,
794
+ "metadata": {
795
+ "model": "openai/gpt-6-luna-decisions-20261006",
796
+ "provider": "OpenAI",
797
+ "requestId": "gen-dec-1792278010-eyZP0xQt425JFx905MCp",
798
+ "durationMs": 192.1062039999997,
799
+ "costUsd": 0.0000685,
800
+ "attempts": 1
801
+ }
802
+ },
803
+ "expected": "general_feedback",
804
+ "correct": true
805
+ },
806
+ {
807
+ "route": "luna-decisions",
808
+ "caseId": "support-heldout-008",
809
+ "split": "heldout",
810
+ "kind": "text",
811
+ "durationMs": 216.69686900000306,
812
+ "result": {
813
+ "choice": "general_feedback",
814
+ "nativeChoice": "general_feedback",
815
+ "probabilities": {
816
+ "order_status": 0,
817
+ "refund": 0,
818
+ "technical_support": 0,
819
+ "general_feedback": 0.99,
820
+ "unknown": 0.01
821
+ },
822
+ "tied": ["general_feedback"],
823
+ "confidence": 0.99,
824
+ "metadata": {
825
+ "model": "openai/gpt-6-luna-decisions-20261006",
826
+ "provider": "OpenAI",
827
+ "requestId": "gen-dec-1792278010-gYu4F6L5XCCC67LXbEF6",
828
+ "durationMs": 215.2029450000009,
829
+ "costUsd": 0.0000683,
830
+ "attempts": 1
831
+ }
832
+ },
833
+ "expected": "general_feedback",
834
+ "correct": true
835
+ },
836
+ {
837
+ "route": "luna-decisions",
838
+ "caseId": "support-heldout-009",
839
+ "split": "heldout",
840
+ "kind": "text",
841
+ "durationMs": 271.3751020000018,
842
+ "result": {
843
+ "choice": "unknown",
844
+ "nativeChoice": "unknown",
845
+ "probabilities": {
846
+ "order_status": 0.01,
847
+ "refund": 0.01,
848
+ "technical_support": 0.05,
849
+ "general_feedback": 0,
850
+ "unknown": 0.93
851
+ },
852
+ "tied": ["unknown"],
853
+ "confidence": 0.91,
854
+ "metadata": {
855
+ "model": "openai/gpt-6-luna-decisions-20261006",
856
+ "provider": "OpenAI",
857
+ "requestId": "gen-dec-1792278010-wvQTSJSx4AV4erKthEJB",
858
+ "durationMs": 269.99628999999914,
859
+ "costUsd": 0.0000684,
860
+ "attempts": 1
861
+ }
862
+ },
863
+ "expected": "unknown",
864
+ "correct": true
865
+ },
866
+ {
867
+ "route": "luna-decisions",
868
+ "caseId": "support-heldout-010",
869
+ "split": "heldout",
870
+ "kind": "text",
871
+ "durationMs": 183.1355659999972,
872
+ "result": {
873
+ "choice": "unknown",
874
+ "nativeChoice": "unknown",
875
+ "probabilities": {
876
+ "order_status": 0.01,
877
+ "refund": 0,
878
+ "technical_support": 0,
879
+ "general_feedback": 0,
880
+ "unknown": 0.99
881
+ },
882
+ "tied": ["unknown"],
883
+ "confidence": 0.99,
884
+ "metadata": {
885
+ "model": "openai/gpt-6-luna-decisions-20261006",
886
+ "provider": "OpenAI",
887
+ "requestId": "gen-dec-1792278010-V8iCooRexe8nb2qP1mV5",
888
+ "durationMs": 181.67535199999838,
889
+ "costUsd": 0.0000672,
890
+ "attempts": 1
891
+ }
892
+ },
893
+ "expected": "unknown",
894
+ "correct": true
895
+ },
896
+ {
897
+ "route": "luna-decisions",
898
+ "caseId": "support-heldout-011",
899
+ "split": "heldout",
900
+ "kind": "text",
901
+ "durationMs": 197.27163799999835,
902
+ "result": {
903
+ "choice": "order_status",
904
+ "nativeChoice": "order_status",
905
+ "probabilities": {
906
+ "order_status": 0.99,
907
+ "refund": 0.01,
908
+ "technical_support": 0,
909
+ "general_feedback": 0,
910
+ "unknown": 0
911
+ },
912
+ "tied": ["order_status"],
913
+ "confidence": 0.99,
914
+ "metadata": {
915
+ "model": "openai/gpt-6-luna-decisions-20261006",
916
+ "provider": "OpenAI",
917
+ "requestId": "gen-dec-1792278011-Jiz4CAzjCiXfnMXVeKCN",
918
+ "durationMs": 195.96227600000202,
919
+ "costUsd": 0.0000687,
920
+ "attempts": 1
921
+ }
922
+ },
923
+ "expected": "order_status",
924
+ "correct": true
925
+ },
926
+ {
927
+ "route": "luna-decisions",
928
+ "caseId": "support-heldout-012",
929
+ "split": "heldout",
930
+ "kind": "text",
931
+ "durationMs": 621.9313650000004,
932
+ "result": {
933
+ "choice": "order_status",
934
+ "nativeChoice": "order_status",
935
+ "probabilities": {
936
+ "order_status": 0.91,
937
+ "refund": 0,
938
+ "technical_support": 0.01,
939
+ "general_feedback": 0,
940
+ "unknown": 0.08
941
+ },
942
+ "tied": ["order_status"],
943
+ "confidence": 0.89,
944
+ "metadata": {
945
+ "model": "openai/gpt-6-luna-decisions-20261006",
946
+ "provider": "OpenAI",
947
+ "requestId": "gen-dec-1792278011-Evh5DGzlzEQB3hP4YzIS",
948
+ "durationMs": 620.5858430000008,
949
+ "costUsd": 0.000068,
950
+ "attempts": 1
951
+ }
952
+ },
953
+ "expected": "order_status",
954
+ "correct": true
955
+ },
956
+ {
957
+ "route": "luna-decisions",
958
+ "caseId": "support-heldout-013",
959
+ "split": "heldout",
960
+ "kind": "text",
961
+ "durationMs": 210.15029100000174,
962
+ "result": {
963
+ "choice": "refund",
964
+ "nativeChoice": "refund",
965
+ "probabilities": {
966
+ "order_status": 0.01,
967
+ "refund": 0.99,
968
+ "technical_support": 0,
969
+ "general_feedback": 0,
970
+ "unknown": 0
971
+ },
972
+ "tied": ["refund"],
973
+ "confidence": 0.99,
974
+ "metadata": {
975
+ "model": "openai/gpt-6-luna-decisions-20261006",
976
+ "provider": "OpenAI",
977
+ "requestId": "gen-dec-1792278011-P6XE4iOlfa9ofHQVcrKJ",
978
+ "durationMs": 208.32587299999796,
979
+ "costUsd": 0.0000685,
980
+ "attempts": 1
981
+ }
982
+ },
983
+ "expected": "refund",
984
+ "correct": true
985
+ },
986
+ {
987
+ "route": "luna-decisions",
988
+ "caseId": "support-heldout-014",
989
+ "split": "heldout",
990
+ "kind": "text",
991
+ "durationMs": 203.72848599999998,
992
+ "result": {
993
+ "choice": "refund",
994
+ "nativeChoice": "refund",
995
+ "probabilities": {
996
+ "order_status": 0.01,
997
+ "refund": 0.81,
998
+ "technical_support": 0.02,
999
+ "general_feedback": 0.01,
1000
+ "unknown": 0.15
1001
+ },
1002
+ "tied": ["refund"],
1003
+ "confidence": 0.76,
1004
+ "metadata": {
1005
+ "model": "openai/gpt-6-luna-decisions-20261006",
1006
+ "provider": "OpenAI",
1007
+ "requestId": "gen-dec-1792278012-xApArfE36pFv2QavEbLt",
1008
+ "durationMs": 201.4932039999985,
1009
+ "costUsd": 0.0000676,
1010
+ "attempts": 1
1011
+ }
1012
+ },
1013
+ "expected": "refund",
1014
+ "correct": true
1015
+ },
1016
+ {
1017
+ "route": "luna-decisions",
1018
+ "caseId": "support-heldout-015",
1019
+ "split": "heldout",
1020
+ "kind": "text",
1021
+ "durationMs": 214.86429500000304,
1022
+ "result": {
1023
+ "choice": "technical_support",
1024
+ "nativeChoice": "technical_support",
1025
+ "probabilities": {
1026
+ "order_status": 0,
1027
+ "refund": 0.01,
1028
+ "technical_support": 0.98,
1029
+ "general_feedback": 0,
1030
+ "unknown": 0.01
1031
+ },
1032
+ "tied": ["technical_support"],
1033
+ "confidence": 0.98,
1034
+ "metadata": {
1035
+ "model": "openai/gpt-6-luna-decisions-20261006",
1036
+ "provider": "OpenAI",
1037
+ "requestId": "gen-dec-1792278012-E8NLKHNdE9oMwN3KA3aD",
1038
+ "durationMs": 213.5112509999999,
1039
+ "costUsd": 0.0000686,
1040
+ "attempts": 1
1041
+ }
1042
+ },
1043
+ "expected": "technical_support",
1044
+ "correct": true
1045
+ },
1046
+ {
1047
+ "route": "luna-decisions",
1048
+ "caseId": "support-heldout-016",
1049
+ "split": "heldout",
1050
+ "kind": "text",
1051
+ "durationMs": 247.88723400000163,
1052
+ "result": {
1053
+ "choice": "technical_support",
1054
+ "nativeChoice": "technical_support",
1055
+ "probabilities": {
1056
+ "order_status": 0,
1057
+ "refund": 0,
1058
+ "technical_support": 0.99,
1059
+ "general_feedback": 0,
1060
+ "unknown": 0.01
1061
+ },
1062
+ "tied": ["technical_support"],
1063
+ "confidence": 0.99,
1064
+ "metadata": {
1065
+ "model": "openai/gpt-6-luna-decisions-20261006",
1066
+ "provider": "OpenAI",
1067
+ "requestId": "gen-dec-1792278012-zirENgOqVPT4LvqfCLxi",
1068
+ "durationMs": 246.5892070000009,
1069
+ "costUsd": 0.0000677,
1070
+ "attempts": 1
1071
+ }
1072
+ },
1073
+ "expected": "technical_support",
1074
+ "correct": true
1075
+ },
1076
+ {
1077
+ "route": "luna-decisions",
1078
+ "caseId": "support-heldout-017",
1079
+ "split": "heldout",
1080
+ "kind": "text",
1081
+ "durationMs": 268.06552499999816,
1082
+ "result": {
1083
+ "choice": "general_feedback",
1084
+ "nativeChoice": "general_feedback",
1085
+ "probabilities": {
1086
+ "order_status": 0,
1087
+ "refund": 0,
1088
+ "technical_support": 0,
1089
+ "general_feedback": 0.98,
1090
+ "unknown": 0.02
1091
+ },
1092
+ "tied": ["general_feedback"],
1093
+ "confidence": 0.98,
1094
+ "metadata": {
1095
+ "model": "openai/gpt-6-luna-decisions-20261006",
1096
+ "provider": "OpenAI",
1097
+ "requestId": "gen-dec-1792278012-YblJSp9yUJO17TCZlA17",
1098
+ "durationMs": 266.7219179999993,
1099
+ "costUsd": 0.0000682,
1100
+ "attempts": 1
1101
+ }
1102
+ },
1103
+ "expected": "general_feedback",
1104
+ "correct": true
1105
+ },
1106
+ {
1107
+ "route": "luna-decisions",
1108
+ "caseId": "support-heldout-018",
1109
+ "split": "heldout",
1110
+ "kind": "text",
1111
+ "durationMs": 279.66683099999864,
1112
+ "result": {
1113
+ "choice": "general_feedback",
1114
+ "nativeChoice": "general_feedback",
1115
+ "probabilities": {
1116
+ "order_status": 0,
1117
+ "refund": 0,
1118
+ "technical_support": 0,
1119
+ "general_feedback": 0.99,
1120
+ "unknown": 0.01
1121
+ },
1122
+ "tied": ["general_feedback"],
1123
+ "confidence": 0.99,
1124
+ "metadata": {
1125
+ "model": "openai/gpt-6-luna-decisions-20261006",
1126
+ "provider": "OpenAI",
1127
+ "requestId": "gen-dec-1792278013-HTWHUjOJRSqzW1j6cF4m",
1128
+ "durationMs": 278.2255129999976,
1129
+ "costUsd": 0.0000675,
1130
+ "attempts": 1
1131
+ }
1132
+ },
1133
+ "expected": "general_feedback",
1134
+ "correct": true
1135
+ },
1136
+ {
1137
+ "route": "luna-decisions",
1138
+ "caseId": "support-heldout-019",
1139
+ "split": "heldout",
1140
+ "kind": "text",
1141
+ "durationMs": 204.25994699999865,
1142
+ "result": {
1143
+ "choice": "unknown",
1144
+ "nativeChoice": "unknown",
1145
+ "probabilities": {
1146
+ "order_status": 0,
1147
+ "refund": 0.01,
1148
+ "technical_support": 0.01,
1149
+ "general_feedback": 0.01,
1150
+ "unknown": 0.97
1151
+ },
1152
+ "tied": ["unknown"],
1153
+ "confidence": 0.96,
1154
+ "metadata": {
1155
+ "model": "openai/gpt-6-luna-decisions-20261006",
1156
+ "provider": "OpenAI",
1157
+ "requestId": "gen-dec-1792278013-2SdsjRyMhqDAI3R41q9D",
1158
+ "durationMs": 202.92036200000075,
1159
+ "costUsd": 0.0000682,
1160
+ "attempts": 1
1161
+ }
1162
+ },
1163
+ "expected": "unknown",
1164
+ "correct": true
1165
+ },
1166
+ {
1167
+ "route": "luna-decisions",
1168
+ "caseId": "support-heldout-020",
1169
+ "split": "heldout",
1170
+ "kind": "text",
1171
+ "durationMs": 216.86293599999772,
1172
+ "result": {
1173
+ "choice": "unknown",
1174
+ "nativeChoice": "unknown",
1175
+ "probabilities": {
1176
+ "order_status": 0,
1177
+ "refund": 0,
1178
+ "technical_support": 0,
1179
+ "general_feedback": 0.01,
1180
+ "unknown": 0.99
1181
+ },
1182
+ "tied": ["unknown"],
1183
+ "confidence": 0.99,
1184
+ "metadata": {
1185
+ "model": "openai/gpt-6-luna-decisions-20261006",
1186
+ "provider": "OpenAI",
1187
+ "requestId": "gen-dec-1792278013-eCZ4wQsoEfpKi0IQ0jXQ",
1188
+ "durationMs": 215.5590000000011,
1189
+ "costUsd": 0.0000683,
1190
+ "attempts": 1
1191
+ }
1192
+ },
1193
+ "expected": "unknown",
1194
+ "correct": true
1195
+ }
1196
+ ],
1197
+ "limitations": [
1198
+ "Public images can overlap model training",
1199
+ "Authored text fixtures are not customer traffic",
1200
+ "No accuracy guarantees outside this corpus",
1201
+ "Initial run evaluates two finalists; extension evaluates one fixed route; protocol variants are not paired comparisons",
1202
+ "Locator cases are integration smoke, not full localization quality",
1203
+ "Local cost reservations are not an account-side spending cap"
1204
+ ]
1205
+ }