@claude-flow/cli 3.10.46 → 3.11.0

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 (551) hide show
  1. package/.claude/agents/analysis/analyze-code-quality.md +178 -178
  2. package/.claude/agents/analysis/code-analyzer.md +209 -209
  3. package/.claude/agents/analysis/code-review/analyze-code-quality.md +178 -178
  4. package/.claude/agents/architecture/arch-system-design.md +156 -156
  5. package/.claude/agents/architecture/system-design/arch-system-design.md +154 -154
  6. package/.claude/agents/browser/browser-agent.yaml +182 -182
  7. package/.claude/agents/consensus/byzantine-coordinator.md +62 -62
  8. package/.claude/agents/consensus/crdt-synchronizer.md +996 -996
  9. package/.claude/agents/consensus/gossip-coordinator.md +62 -62
  10. package/.claude/agents/consensus/performance-benchmarker.md +850 -850
  11. package/.claude/agents/consensus/quorum-manager.md +822 -822
  12. package/.claude/agents/consensus/raft-manager.md +62 -62
  13. package/.claude/agents/consensus/security-manager.md +621 -621
  14. package/.claude/agents/core/planner.md +374 -374
  15. package/.claude/agents/custom/test-long-runner.md +44 -44
  16. package/.claude/agents/data/data-ml-model.md +444 -444
  17. package/.claude/agents/data/ml/data-ml-model.md +192 -192
  18. package/.claude/agents/development/backend/dev-backend-api.md +141 -141
  19. package/.claude/agents/development/dev-backend-api.md +344 -344
  20. package/.claude/agents/devops/ci-cd/ops-cicd-github.md +163 -163
  21. package/.claude/agents/devops/ops-cicd-github.md +164 -164
  22. package/.claude/agents/documentation/api-docs/docs-api-openapi.md +173 -173
  23. package/.claude/agents/documentation/docs-api-openapi.md +354 -354
  24. package/.claude/agents/flow-nexus/app-store.md +87 -87
  25. package/.claude/agents/flow-nexus/authentication.md +68 -68
  26. package/.claude/agents/flow-nexus/challenges.md +80 -80
  27. package/.claude/agents/flow-nexus/neural-network.md +87 -87
  28. package/.claude/agents/flow-nexus/payments.md +82 -82
  29. package/.claude/agents/flow-nexus/sandbox.md +75 -75
  30. package/.claude/agents/flow-nexus/swarm.md +75 -75
  31. package/.claude/agents/flow-nexus/user-tools.md +95 -95
  32. package/.claude/agents/flow-nexus/workflow.md +83 -83
  33. package/.claude/agents/github/code-review-swarm.md +377 -377
  34. package/.claude/agents/github/github-modes.md +172 -172
  35. package/.claude/agents/github/issue-tracker.md +575 -575
  36. package/.claude/agents/github/multi-repo-swarm.md +552 -552
  37. package/.claude/agents/github/pr-manager.md +437 -437
  38. package/.claude/agents/github/project-board-sync.md +508 -508
  39. package/.claude/agents/github/release-manager.md +604 -604
  40. package/.claude/agents/github/release-swarm.md +582 -582
  41. package/.claude/agents/github/repo-architect.md +397 -397
  42. package/.claude/agents/github/swarm-issue.md +572 -572
  43. package/.claude/agents/github/swarm-pr.md +427 -427
  44. package/.claude/agents/github/sync-coordinator.md +451 -451
  45. package/.claude/agents/github/workflow-automation.md +902 -902
  46. package/.claude/agents/goal/agent.md +815 -815
  47. package/.claude/agents/optimization/benchmark-suite.md +664 -664
  48. package/.claude/agents/optimization/load-balancer.md +430 -430
  49. package/.claude/agents/optimization/performance-monitor.md +671 -671
  50. package/.claude/agents/optimization/resource-allocator.md +673 -673
  51. package/.claude/agents/optimization/topology-optimizer.md +807 -807
  52. package/.claude/agents/payments/agentic-payments.md +126 -126
  53. package/.claude/agents/sona/sona-learning-optimizer.md +74 -74
  54. package/.claude/agents/sparc/architecture.md +698 -698
  55. package/.claude/agents/sparc/pseudocode.md +519 -519
  56. package/.claude/agents/sparc/refinement.md +801 -801
  57. package/.claude/agents/sparc/specification.md +477 -477
  58. package/.claude/agents/specialized/mobile/spec-mobile-react-native.md +224 -224
  59. package/.claude/agents/specialized/spec-mobile-react-native.md +226 -226
  60. package/.claude/agents/sublinear/consensus-coordinator.md +337 -337
  61. package/.claude/agents/sublinear/matrix-optimizer.md +184 -184
  62. package/.claude/agents/sublinear/pagerank-analyzer.md +298 -298
  63. package/.claude/agents/sublinear/performance-optimizer.md +367 -367
  64. package/.claude/agents/sublinear/trading-predictor.md +245 -245
  65. package/.claude/agents/swarm/adaptive-coordinator.md +1126 -1126
  66. package/.claude/agents/swarm/hierarchical-coordinator.md +709 -709
  67. package/.claude/agents/swarm/mesh-coordinator.md +962 -962
  68. package/.claude/agents/templates/automation-smart-agent.md +204 -204
  69. package/.claude/agents/templates/base-template-generator.md +289 -289
  70. package/.claude/agents/templates/coordinator-swarm-init.md +89 -89
  71. package/.claude/agents/templates/github-pr-manager.md +176 -176
  72. package/.claude/agents/templates/implementer-sparc-coder.md +258 -258
  73. package/.claude/agents/templates/memory-coordinator.md +186 -186
  74. package/.claude/agents/templates/orchestrator-task.md +138 -138
  75. package/.claude/agents/templates/performance-analyzer.md +198 -198
  76. package/.claude/agents/templates/sparc-coordinator.md +513 -513
  77. package/.claude/agents/testing/production-validator.md +394 -394
  78. package/.claude/agents/testing/tdd-london-swarm.md +243 -243
  79. package/.claude/agents/v3/aidefence-guardian.md +282 -282
  80. package/.claude/agents/v3/claims-authorizer.md +208 -208
  81. package/.claude/agents/v3/collective-intelligence-coordinator.md +993 -993
  82. package/.claude/agents/v3/ddd-domain-expert.md +220 -220
  83. package/.claude/agents/v3/injection-analyst.md +236 -236
  84. package/.claude/agents/v3/performance-engineer.md +1233 -1233
  85. package/.claude/agents/v3/pii-detector.md +151 -151
  86. package/.claude/agents/v3/reasoningbank-learner.md +213 -213
  87. package/.claude/agents/v3/security-architect-aidefence.md +410 -410
  88. package/.claude/agents/v3/security-architect.md +867 -867
  89. package/.claude/agents/v3/swarm-memory-manager.md +157 -157
  90. package/.claude/agents/v3/v3-integration-architect.md +205 -205
  91. package/.claude/commands/agents/README.md +50 -50
  92. package/.claude/commands/agents/agent-capabilities.md +140 -140
  93. package/.claude/commands/agents/agent-coordination.md +28 -28
  94. package/.claude/commands/agents/agent-spawning.md +28 -28
  95. package/.claude/commands/agents/agent-types.md +216 -216
  96. package/.claude/commands/agents/health.md +139 -139
  97. package/.claude/commands/agents/list.md +100 -100
  98. package/.claude/commands/agents/logs.md +130 -130
  99. package/.claude/commands/agents/metrics.md +122 -122
  100. package/.claude/commands/agents/pool.md +127 -127
  101. package/.claude/commands/agents/spawn.md +140 -140
  102. package/.claude/commands/agents/status.md +115 -115
  103. package/.claude/commands/agents/stop.md +102 -102
  104. package/.claude/commands/analysis/COMMAND_COMPLIANCE_REPORT.md +53 -53
  105. package/.claude/commands/analysis/README.md +9 -9
  106. package/.claude/commands/analysis/bottleneck-detect.md +162 -162
  107. package/.claude/commands/analysis/performance-bottlenecks.md +58 -58
  108. package/.claude/commands/analysis/performance-report.md +25 -25
  109. package/.claude/commands/analysis/token-efficiency.md +44 -44
  110. package/.claude/commands/analysis/token-usage.md +25 -25
  111. package/.claude/commands/automation/README.md +9 -9
  112. package/.claude/commands/automation/auto-agent.md +122 -122
  113. package/.claude/commands/automation/self-healing.md +105 -105
  114. package/.claude/commands/automation/session-memory.md +89 -89
  115. package/.claude/commands/automation/smart-agents.md +72 -72
  116. package/.claude/commands/automation/smart-spawn.md +25 -25
  117. package/.claude/commands/automation/workflow-select.md +25 -25
  118. package/.claude/commands/claude-flow-help.md +103 -103
  119. package/.claude/commands/claude-flow-memory.md +107 -107
  120. package/.claude/commands/claude-flow-swarm.md +205 -205
  121. package/.claude/commands/coordination/README.md +9 -9
  122. package/.claude/commands/coordination/agent-spawn.md +25 -25
  123. package/.claude/commands/coordination/init.md +44 -44
  124. package/.claude/commands/coordination/orchestrate.md +43 -43
  125. package/.claude/commands/coordination/spawn.md +45 -45
  126. package/.claude/commands/coordination/swarm-init.md +85 -85
  127. package/.claude/commands/coordination/task-orchestrate.md +25 -25
  128. package/.claude/commands/github/README.md +11 -11
  129. package/.claude/commands/github/code-review-swarm.md +513 -513
  130. package/.claude/commands/github/code-review.md +25 -25
  131. package/.claude/commands/github/github-modes.md +146 -146
  132. package/.claude/commands/github/github-swarm.md +121 -121
  133. package/.claude/commands/github/issue-tracker.md +291 -291
  134. package/.claude/commands/github/issue-triage.md +25 -25
  135. package/.claude/commands/github/multi-repo-swarm.md +518 -518
  136. package/.claude/commands/github/pr-enhance.md +26 -26
  137. package/.claude/commands/github/pr-manager.md +169 -169
  138. package/.claude/commands/github/project-board-sync.md +470 -470
  139. package/.claude/commands/github/release-manager.md +339 -339
  140. package/.claude/commands/github/release-swarm.md +543 -543
  141. package/.claude/commands/github/repo-analyze.md +25 -25
  142. package/.claude/commands/github/repo-architect.md +366 -366
  143. package/.claude/commands/github/swarm-issue.md +484 -484
  144. package/.claude/commands/github/swarm-pr.md +287 -287
  145. package/.claude/commands/github/sync-coordinator.md +302 -302
  146. package/.claude/commands/github/workflow-automation.md +441 -441
  147. package/.claude/commands/hive-mind/README.md +17 -17
  148. package/.claude/commands/hive-mind/hive-mind-consensus.md +8 -8
  149. package/.claude/commands/hive-mind/hive-mind-init.md +18 -18
  150. package/.claude/commands/hive-mind/hive-mind-memory.md +8 -8
  151. package/.claude/commands/hive-mind/hive-mind-metrics.md +8 -8
  152. package/.claude/commands/hive-mind/hive-mind-resume.md +8 -8
  153. package/.claude/commands/hive-mind/hive-mind-sessions.md +8 -8
  154. package/.claude/commands/hive-mind/hive-mind-spawn.md +21 -21
  155. package/.claude/commands/hive-mind/hive-mind-status.md +8 -8
  156. package/.claude/commands/hive-mind/hive-mind-stop.md +8 -8
  157. package/.claude/commands/hive-mind/hive-mind-wizard.md +8 -8
  158. package/.claude/commands/hive-mind/hive-mind.md +27 -27
  159. package/.claude/commands/hooks/README.md +11 -11
  160. package/.claude/commands/hooks/overview.md +57 -57
  161. package/.claude/commands/hooks/post-edit.md +117 -117
  162. package/.claude/commands/hooks/post-task.md +112 -112
  163. package/.claude/commands/hooks/pre-edit.md +113 -113
  164. package/.claude/commands/hooks/pre-task.md +111 -111
  165. package/.claude/commands/hooks/session-end.md +118 -118
  166. package/.claude/commands/hooks/setup.md +102 -102
  167. package/.claude/commands/memory/README.md +9 -9
  168. package/.claude/commands/memory/memory-persist.md +25 -25
  169. package/.claude/commands/memory/memory-search.md +25 -25
  170. package/.claude/commands/memory/memory-usage.md +25 -25
  171. package/.claude/commands/memory/neural.md +47 -47
  172. package/.claude/commands/monitoring/README.md +9 -9
  173. package/.claude/commands/monitoring/agent-metrics.md +25 -25
  174. package/.claude/commands/monitoring/agents.md +44 -44
  175. package/.claude/commands/monitoring/real-time-view.md +25 -25
  176. package/.claude/commands/monitoring/status.md +46 -46
  177. package/.claude/commands/monitoring/swarm-monitor.md +25 -25
  178. package/.claude/commands/optimization/README.md +9 -9
  179. package/.claude/commands/optimization/auto-topology.md +61 -61
  180. package/.claude/commands/optimization/cache-manage.md +25 -25
  181. package/.claude/commands/optimization/parallel-execute.md +25 -25
  182. package/.claude/commands/optimization/parallel-execution.md +49 -49
  183. package/.claude/commands/optimization/topology-optimize.md +25 -25
  184. package/.claude/commands/pair/README.md +260 -260
  185. package/.claude/commands/pair/commands.md +545 -545
  186. package/.claude/commands/pair/config.md +509 -509
  187. package/.claude/commands/pair/examples.md +511 -511
  188. package/.claude/commands/pair/modes.md +347 -347
  189. package/.claude/commands/pair/session.md +406 -406
  190. package/.claude/commands/pair/start.md +208 -208
  191. package/.claude/commands/sparc/analyzer.md +51 -51
  192. package/.claude/commands/sparc/architect.md +53 -53
  193. package/.claude/commands/sparc/ask.md +97 -97
  194. package/.claude/commands/sparc/batch-executor.md +54 -54
  195. package/.claude/commands/sparc/code.md +89 -89
  196. package/.claude/commands/sparc/coder.md +54 -54
  197. package/.claude/commands/sparc/debug.md +83 -83
  198. package/.claude/commands/sparc/debugger.md +54 -54
  199. package/.claude/commands/sparc/designer.md +53 -53
  200. package/.claude/commands/sparc/devops.md +109 -109
  201. package/.claude/commands/sparc/docs-writer.md +80 -80
  202. package/.claude/commands/sparc/documenter.md +54 -54
  203. package/.claude/commands/sparc/innovator.md +54 -54
  204. package/.claude/commands/sparc/integration.md +83 -83
  205. package/.claude/commands/sparc/mcp.md +117 -117
  206. package/.claude/commands/sparc/memory-manager.md +54 -54
  207. package/.claude/commands/sparc/optimizer.md +54 -54
  208. package/.claude/commands/sparc/orchestrator.md +131 -131
  209. package/.claude/commands/sparc/post-deployment-monitoring-mode.md +83 -83
  210. package/.claude/commands/sparc/refinement-optimization-mode.md +83 -83
  211. package/.claude/commands/sparc/researcher.md +54 -54
  212. package/.claude/commands/sparc/reviewer.md +54 -54
  213. package/.claude/commands/sparc/security-review.md +80 -80
  214. package/.claude/commands/sparc/sparc-modes.md +174 -174
  215. package/.claude/commands/sparc/sparc.md +111 -111
  216. package/.claude/commands/sparc/spec-pseudocode.md +80 -80
  217. package/.claude/commands/sparc/supabase-admin.md +348 -348
  218. package/.claude/commands/sparc/swarm-coordinator.md +54 -54
  219. package/.claude/commands/sparc/tdd.md +54 -54
  220. package/.claude/commands/sparc/tester.md +54 -54
  221. package/.claude/commands/sparc/tutorial.md +79 -79
  222. package/.claude/commands/sparc/workflow-manager.md +54 -54
  223. package/.claude/commands/sparc.md +166 -166
  224. package/.claude/commands/stream-chain/pipeline.md +120 -120
  225. package/.claude/commands/stream-chain/run.md +69 -69
  226. package/.claude/commands/swarm/README.md +15 -15
  227. package/.claude/commands/swarm/analysis.md +95 -95
  228. package/.claude/commands/swarm/development.md +96 -96
  229. package/.claude/commands/swarm/examples.md +168 -168
  230. package/.claude/commands/swarm/maintenance.md +102 -102
  231. package/.claude/commands/swarm/optimization.md +117 -117
  232. package/.claude/commands/swarm/research.md +136 -136
  233. package/.claude/commands/swarm/swarm-analysis.md +8 -8
  234. package/.claude/commands/swarm/swarm-background.md +8 -8
  235. package/.claude/commands/swarm/swarm-init.md +19 -19
  236. package/.claude/commands/swarm/swarm-modes.md +8 -8
  237. package/.claude/commands/swarm/swarm-monitor.md +8 -8
  238. package/.claude/commands/swarm/swarm-spawn.md +19 -19
  239. package/.claude/commands/swarm/swarm-status.md +8 -8
  240. package/.claude/commands/swarm/swarm-strategies.md +8 -8
  241. package/.claude/commands/swarm/swarm.md +87 -87
  242. package/.claude/commands/swarm/testing.md +131 -131
  243. package/.claude/commands/training/README.md +9 -9
  244. package/.claude/commands/training/model-update.md +25 -25
  245. package/.claude/commands/training/neural-patterns.md +107 -107
  246. package/.claude/commands/training/neural-train.md +75 -75
  247. package/.claude/commands/training/pattern-learn.md +25 -25
  248. package/.claude/commands/training/specialization.md +62 -62
  249. package/.claude/commands/truth/start.md +142 -142
  250. package/.claude/commands/verify/check.md +49 -49
  251. package/.claude/commands/verify/start.md +127 -127
  252. package/.claude/commands/workflows/README.md +9 -9
  253. package/.claude/commands/workflows/development.md +77 -77
  254. package/.claude/commands/workflows/research.md +62 -62
  255. package/.claude/commands/workflows/workflow-create.md +25 -25
  256. package/.claude/commands/workflows/workflow-execute.md +25 -25
  257. package/.claude/commands/workflows/workflow-export.md +25 -25
  258. package/.claude/helpers/README.md +96 -96
  259. package/.claude/helpers/adr-compliance.sh +186 -186
  260. package/.claude/helpers/auto-commit.sh +178 -178
  261. package/.claude/helpers/auto-memory-hook.mjs +368 -368
  262. package/.claude/helpers/checkpoint-manager.sh +251 -251
  263. package/.claude/helpers/daemon-manager.sh +252 -252
  264. package/.claude/helpers/ddd-tracker.sh +144 -144
  265. package/.claude/helpers/github-safe.js +156 -156
  266. package/.claude/helpers/github-setup.sh +45 -45
  267. package/.claude/helpers/guidance-hook.sh +13 -13
  268. package/.claude/helpers/guidance-hooks.sh +102 -102
  269. package/.claude/helpers/health-monitor.sh +108 -108
  270. package/.claude/helpers/hook-handler.cjs +286 -286
  271. package/.claude/helpers/intelligence.cjs +1031 -1031
  272. package/.claude/helpers/learning-hooks.sh +329 -329
  273. package/.claude/helpers/learning-optimizer.sh +127 -127
  274. package/.claude/helpers/learning-service.mjs +1144 -1144
  275. package/.claude/helpers/memory.js +83 -83
  276. package/.claude/helpers/metrics-db.mjs +488 -488
  277. package/.claude/helpers/pattern-consolidator.sh +86 -86
  278. package/.claude/helpers/perf-worker.sh +160 -160
  279. package/.claude/helpers/post-commit +16 -16
  280. package/.claude/helpers/pre-commit +26 -26
  281. package/.claude/helpers/quick-start.sh +19 -19
  282. package/.claude/helpers/router.js +105 -105
  283. package/.claude/helpers/security-scanner.sh +127 -127
  284. package/.claude/helpers/session.js +157 -157
  285. package/.claude/helpers/setup-mcp.sh +18 -18
  286. package/.claude/helpers/standard-checkpoint-hooks.sh +189 -189
  287. package/.claude/helpers/statusline-hook.sh +21 -21
  288. package/.claude/helpers/statusline.cjs +604 -604
  289. package/.claude/helpers/statusline.js +352 -352
  290. package/.claude/helpers/swarm-comms.sh +353 -353
  291. package/.claude/helpers/swarm-hooks.sh +761 -761
  292. package/.claude/helpers/swarm-monitor.sh +210 -210
  293. package/.claude/helpers/sync-v3-metrics.sh +245 -245
  294. package/.claude/helpers/update-v3-progress.sh +165 -165
  295. package/.claude/helpers/v3-quick-status.sh +57 -57
  296. package/.claude/helpers/v3.sh +110 -110
  297. package/.claude/helpers/validate-v3-config.sh +215 -215
  298. package/.claude/helpers/worker-manager.sh +170 -170
  299. package/.claude/settings.json +182 -182
  300. package/.claude/skills/agentdb-advanced/SKILL.md +550 -550
  301. package/.claude/skills/agentdb-learning/SKILL.md +545 -545
  302. package/.claude/skills/agentdb-memory-patterns/SKILL.md +339 -339
  303. package/.claude/skills/agentdb-optimization/SKILL.md +509 -509
  304. package/.claude/skills/agentdb-vector-search/SKILL.md +339 -339
  305. package/.claude/skills/browser/SKILL.md +204 -204
  306. package/.claude/skills/dual-mode/README.md +71 -71
  307. package/.claude/skills/dual-mode/dual-collect.md +103 -103
  308. package/.claude/skills/dual-mode/dual-coordinate.md +85 -85
  309. package/.claude/skills/dual-mode/dual-spawn.md +81 -81
  310. package/.claude/skills/flow-nexus-neural/SKILL.md +727 -727
  311. package/.claude/skills/flow-nexus-platform/SKILL.md +1154 -1154
  312. package/.claude/skills/flow-nexus-swarm/SKILL.md +604 -604
  313. package/.claude/skills/github-code-review/SKILL.md +1125 -1125
  314. package/.claude/skills/github-multi-repo/SKILL.md +862 -862
  315. package/.claude/skills/github-project-management/SKILL.md +1262 -1262
  316. package/.claude/skills/github-release-management/SKILL.md +1064 -1064
  317. package/.claude/skills/github-workflow-automation/SKILL.md +1047 -1047
  318. package/.claude/skills/hooks-automation/SKILL.md +1201 -1201
  319. package/.claude/skills/pair-programming/SKILL.md +1202 -1202
  320. package/.claude/skills/reasoningbank-agentdb/SKILL.md +446 -446
  321. package/.claude/skills/reasoningbank-intelligence/SKILL.md +201 -201
  322. package/.claude/skills/skill-builder/SKILL.md +910 -910
  323. package/.claude/skills/sparc-methodology/SKILL.md +1106 -1106
  324. package/.claude/skills/stream-chain/SKILL.md +560 -560
  325. package/.claude/skills/swarm-advanced/SKILL.md +970 -970
  326. package/.claude/skills/swarm-orchestration/SKILL.md +179 -179
  327. package/.claude/skills/v3-cli-modernization/SKILL.md +871 -871
  328. package/.claude/skills/v3-core-implementation/SKILL.md +796 -796
  329. package/.claude/skills/v3-ddd-architecture/SKILL.md +441 -441
  330. package/.claude/skills/v3-integration-deep/SKILL.md +240 -240
  331. package/.claude/skills/v3-mcp-optimization/SKILL.md +776 -776
  332. package/.claude/skills/v3-memory-unification/SKILL.md +173 -173
  333. package/.claude/skills/v3-performance-optimization/SKILL.md +389 -389
  334. package/.claude/skills/v3-security-overhaul/SKILL.md +81 -81
  335. package/.claude/skills/v3-swarm-coordination/SKILL.md +339 -339
  336. package/.claude/skills/verification-quality/SKILL.md +691 -691
  337. package/README.md +412 -412
  338. package/bin/cli.js +314 -314
  339. package/bin/mcp-server.js +224 -224
  340. package/bin/preinstall.cjs +2 -2
  341. package/dist/src/benchmarks/capability-tasks.d.ts +126 -0
  342. package/dist/src/benchmarks/capability-tasks.d.ts.map +1 -0
  343. package/dist/src/benchmarks/capability-tasks.js +147 -0
  344. package/dist/src/benchmarks/capability-tasks.js.map +1 -0
  345. package/dist/src/benchmarks/gaia-agent-contrastive.d.ts +53 -0
  346. package/dist/src/benchmarks/gaia-agent-contrastive.d.ts.map +1 -0
  347. package/dist/src/benchmarks/gaia-agent-contrastive.js +177 -0
  348. package/dist/src/benchmarks/gaia-agent-contrastive.js.map +1 -0
  349. package/dist/src/benchmarks/gaia-agent-gemini.d.ts +51 -0
  350. package/dist/src/benchmarks/gaia-agent-gemini.d.ts.map +1 -0
  351. package/dist/src/benchmarks/gaia-agent-gemini.js +352 -0
  352. package/dist/src/benchmarks/gaia-agent-gemini.js.map +1 -0
  353. package/dist/src/benchmarks/gaia-claude-p.d.ts +105 -0
  354. package/dist/src/benchmarks/gaia-claude-p.d.ts.map +1 -0
  355. package/dist/src/benchmarks/gaia-claude-p.js +269 -0
  356. package/dist/src/benchmarks/gaia-claude-p.js.map +1 -0
  357. package/dist/src/benchmarks/gaia-claude-p.smoke.d.ts +20 -0
  358. package/dist/src/benchmarks/gaia-claude-p.smoke.d.ts.map +1 -0
  359. package/dist/src/benchmarks/gaia-claude-p.smoke.js +168 -0
  360. package/dist/src/benchmarks/gaia-claude-p.smoke.js.map +1 -0
  361. package/dist/src/benchmarks/gaia-codeagent-runner.py +556 -0
  362. package/dist/src/benchmarks/gaia-codeagent.d.ts +129 -0
  363. package/dist/src/benchmarks/gaia-codeagent.d.ts.map +1 -0
  364. package/dist/src/benchmarks/gaia-codeagent.js +590 -0
  365. package/dist/src/benchmarks/gaia-codeagent.js.map +1 -0
  366. package/dist/src/benchmarks/gaia-codeagent.smoke.d.ts +33 -0
  367. package/dist/src/benchmarks/gaia-codeagent.smoke.d.ts.map +1 -0
  368. package/dist/src/benchmarks/gaia-codeagent.smoke.js +241 -0
  369. package/dist/src/benchmarks/gaia-codeagent.smoke.js.map +1 -0
  370. package/dist/src/benchmarks/gaia-critic.js +24 -24
  371. package/dist/src/benchmarks/gaia-dag-pilot.d.ts +22 -0
  372. package/dist/src/benchmarks/gaia-dag-pilot.d.ts.map +1 -0
  373. package/dist/src/benchmarks/gaia-dag-pilot.js +158 -0
  374. package/dist/src/benchmarks/gaia-dag-pilot.js.map +1 -0
  375. package/dist/src/benchmarks/gaia-dag.d.ts +110 -0
  376. package/dist/src/benchmarks/gaia-dag.d.ts.map +1 -0
  377. package/dist/src/benchmarks/gaia-dag.js +658 -0
  378. package/dist/src/benchmarks/gaia-dag.js.map +1 -0
  379. package/dist/src/benchmarks/gaia-dag.smoke.d.ts +21 -0
  380. package/dist/src/benchmarks/gaia-dag.smoke.d.ts.map +1 -0
  381. package/dist/src/benchmarks/gaia-dag.smoke.js +290 -0
  382. package/dist/src/benchmarks/gaia-dag.smoke.js.map +1 -0
  383. package/dist/src/benchmarks/gaia-ensemble-pilot.d.ts +17 -0
  384. package/dist/src/benchmarks/gaia-ensemble-pilot.d.ts.map +1 -0
  385. package/dist/src/benchmarks/gaia-ensemble-pilot.js +106 -0
  386. package/dist/src/benchmarks/gaia-ensemble-pilot.js.map +1 -0
  387. package/dist/src/benchmarks/gaia-ensemble.d.ts +103 -0
  388. package/dist/src/benchmarks/gaia-ensemble.d.ts.map +1 -0
  389. package/dist/src/benchmarks/gaia-ensemble.js +462 -0
  390. package/dist/src/benchmarks/gaia-ensemble.js.map +1 -0
  391. package/dist/src/benchmarks/gaia-gemini-pilot.d.ts +17 -0
  392. package/dist/src/benchmarks/gaia-gemini-pilot.d.ts.map +1 -0
  393. package/dist/src/benchmarks/gaia-gemini-pilot.js +168 -0
  394. package/dist/src/benchmarks/gaia-gemini-pilot.js.map +1 -0
  395. package/dist/src/benchmarks/gaia-mode-router.d.ts +67 -0
  396. package/dist/src/benchmarks/gaia-mode-router.d.ts.map +1 -0
  397. package/dist/src/benchmarks/gaia-mode-router.js +126 -0
  398. package/dist/src/benchmarks/gaia-mode-router.js.map +1 -0
  399. package/dist/src/benchmarks/gaia-mode-router.smoke.d.ts +15 -0
  400. package/dist/src/benchmarks/gaia-mode-router.smoke.d.ts.map +1 -0
  401. package/dist/src/benchmarks/gaia-mode-router.smoke.js +131 -0
  402. package/dist/src/benchmarks/gaia-mode-router.smoke.js.map +1 -0
  403. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.d.ts +24 -0
  404. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.d.ts.map +1 -0
  405. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.js +276 -0
  406. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.js.map +1 -0
  407. package/dist/src/benchmarks/gaia-tools/image_describe.d.ts +45 -0
  408. package/dist/src/benchmarks/gaia-tools/image_describe.d.ts.map +1 -0
  409. package/dist/src/benchmarks/gaia-tools/image_describe.js +222 -0
  410. package/dist/src/benchmarks/gaia-tools/image_describe.js.map +1 -0
  411. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.d.ts +24 -0
  412. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.d.ts.map +1 -0
  413. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.js +169 -0
  414. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.js.map +1 -0
  415. package/dist/src/benchmarks/gaia-tools/pdf_read.d.ts +27 -0
  416. package/dist/src/benchmarks/gaia-tools/pdf_read.d.ts.map +1 -0
  417. package/dist/src/benchmarks/gaia-tools/pdf_read.js +226 -0
  418. package/dist/src/benchmarks/gaia-tools/pdf_read.js.map +1 -0
  419. package/dist/src/benchmarks/gaia-tools/python_exec.d.ts +31 -0
  420. package/dist/src/benchmarks/gaia-tools/python_exec.d.ts.map +1 -0
  421. package/dist/src/benchmarks/gaia-tools/python_exec.js +94 -0
  422. package/dist/src/benchmarks/gaia-tools/python_exec.js.map +1 -0
  423. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.d.ts +13 -0
  424. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.d.ts.map +1 -0
  425. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.js +98 -0
  426. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.js.map +1 -0
  427. package/dist/src/benchmarks/gaia-tools/visit_webpage.d.ts +52 -0
  428. package/dist/src/benchmarks/gaia-tools/visit_webpage.d.ts.map +1 -0
  429. package/dist/src/benchmarks/gaia-tools/visit_webpage.js +304 -0
  430. package/dist/src/benchmarks/gaia-tools/visit_webpage.js.map +1 -0
  431. package/dist/src/benchmarks/gaia-tools/web_browse.d.ts +61 -0
  432. package/dist/src/benchmarks/gaia-tools/web_browse.d.ts.map +1 -0
  433. package/dist/src/benchmarks/gaia-tools/web_browse.js +217 -0
  434. package/dist/src/benchmarks/gaia-tools/web_browse.js.map +1 -0
  435. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.d.ts +26 -0
  436. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.d.ts.map +1 -0
  437. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.js +152 -0
  438. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.js.map +1 -0
  439. package/dist/src/benchmarks/gaia-tools/web_search.smoke.d.ts +24 -0
  440. package/dist/src/benchmarks/gaia-tools/web_search.smoke.d.ts.map +1 -0
  441. package/dist/src/benchmarks/gaia-tools/web_search.smoke.js +311 -0
  442. package/dist/src/benchmarks/gaia-tools/web_search.smoke.js.map +1 -0
  443. package/dist/src/benchmarks/gaia-visit-webpage.smoke.d.ts +25 -0
  444. package/dist/src/benchmarks/gaia-visit-webpage.smoke.d.ts.map +1 -0
  445. package/dist/src/benchmarks/gaia-visit-webpage.smoke.js +304 -0
  446. package/dist/src/benchmarks/gaia-visit-webpage.smoke.js.map +1 -0
  447. package/dist/src/commands/completions.js +409 -409
  448. package/dist/src/commands/daemon.js +44 -44
  449. package/dist/src/commands/embeddings.js +26 -26
  450. package/dist/src/commands/hive-mind.js +97 -97
  451. package/dist/src/commands/hooks.js +9 -9
  452. package/dist/src/commands/neural.d.ts.map +1 -1
  453. package/dist/src/commands/neural.js +2351 -1
  454. package/dist/src/commands/neural.js.map +1 -1
  455. package/dist/src/commands/performance-capability.d.ts +30 -0
  456. package/dist/src/commands/performance-capability.d.ts.map +1 -0
  457. package/dist/src/commands/performance-capability.js +349 -0
  458. package/dist/src/commands/performance-capability.js.map +1 -0
  459. package/dist/src/commands/ruvector/backup.js +23 -23
  460. package/dist/src/commands/ruvector/benchmark.js +31 -31
  461. package/dist/src/commands/ruvector/import.js +14 -14
  462. package/dist/src/commands/ruvector/init.js +115 -115
  463. package/dist/src/commands/ruvector/migrate.js +99 -99
  464. package/dist/src/commands/ruvector/optimize.js +51 -51
  465. package/dist/src/commands/ruvector/setup.js +624 -624
  466. package/dist/src/commands/ruvector/status.js +38 -38
  467. package/dist/src/init/claudemd-generator.js +231 -231
  468. package/dist/src/init/executor.js +453 -453
  469. package/dist/src/init/helpers-generator.js +727 -727
  470. package/dist/src/init/statusline-generator.js +717 -717
  471. package/dist/src/mcp-tools/agent-execute-core.d.ts +23 -1
  472. package/dist/src/mcp-tools/agent-execute-core.d.ts.map +1 -1
  473. package/dist/src/mcp-tools/agent-execute-core.js +136 -9
  474. package/dist/src/mcp-tools/agent-execute-core.js.map +1 -1
  475. package/dist/src/mcp-tools/agent-tools.d.ts.map +1 -1
  476. package/dist/src/mcp-tools/agent-tools.js +48 -4
  477. package/dist/src/mcp-tools/agent-tools.js.map +1 -1
  478. package/dist/src/mcp-tools/agentdb-tools.js +15 -15
  479. package/dist/src/mcp-tools/hooks-tools.d.ts.map +1 -1
  480. package/dist/src/mcp-tools/hooks-tools.js +196 -0
  481. package/dist/src/mcp-tools/hooks-tools.js.map +1 -1
  482. package/dist/src/memory/graph-edge-writer.js +18 -18
  483. package/dist/src/memory/memory-bridge.js +84 -84
  484. package/dist/src/memory/memory-initializer.js +393 -393
  485. package/dist/src/memory/rabitq-index.js +5 -5
  486. package/dist/src/runtime/headless.js +28 -28
  487. package/dist/src/ruvector/enhanced-model-router.d.ts +5 -0
  488. package/dist/src/ruvector/enhanced-model-router.d.ts.map +1 -1
  489. package/dist/src/ruvector/enhanced-model-router.js +18 -1
  490. package/dist/src/ruvector/enhanced-model-router.js.map +1 -1
  491. package/dist/src/ruvector/model-prices.d.ts +50 -0
  492. package/dist/src/ruvector/model-prices.d.ts.map +1 -0
  493. package/dist/src/ruvector/model-prices.js +72 -0
  494. package/dist/src/ruvector/model-prices.js.map +1 -0
  495. package/dist/src/ruvector/model-router.d.ts +147 -8
  496. package/dist/src/ruvector/model-router.d.ts.map +1 -1
  497. package/dist/src/ruvector/model-router.js +418 -18
  498. package/dist/src/ruvector/model-router.js.map +1 -1
  499. package/dist/src/ruvector/neural-router.d.ts +172 -39
  500. package/dist/src/ruvector/neural-router.d.ts.map +1 -1
  501. package/dist/src/ruvector/neural-router.js +780 -91
  502. package/dist/src/ruvector/neural-router.js.map +1 -1
  503. package/dist/src/ruvector/router-calibrator.d.ts +65 -0
  504. package/dist/src/ruvector/router-calibrator.d.ts.map +1 -0
  505. package/dist/src/ruvector/router-calibrator.js +120 -0
  506. package/dist/src/ruvector/router-calibrator.js.map +1 -0
  507. package/dist/src/ruvector/router-trajectory.d.ts +137 -45
  508. package/dist/src/ruvector/router-trajectory.d.ts.map +1 -1
  509. package/dist/src/ruvector/router-trajectory.js +226 -63
  510. package/dist/src/ruvector/router-trajectory.js.map +1 -1
  511. package/dist/src/ruvector/task-embedder.d.ts +56 -0
  512. package/dist/src/ruvector/task-embedder.d.ts.map +1 -0
  513. package/dist/src/ruvector/task-embedder.js +237 -0
  514. package/dist/src/ruvector/task-embedder.js.map +1 -0
  515. package/dist/src/services/headless-worker-executor.js +84 -84
  516. package/dist/src/transfer/deploy-seraphine.js +23 -23
  517. package/dist/tsconfig.tsbuildinfo +1 -1
  518. package/package.json +136 -134
  519. package/scripts/beir-bootstrap-significance.mjs +162 -162
  520. package/scripts/bench-rvagent.mjs +312 -312
  521. package/scripts/benchmark-codemods.mjs +127 -127
  522. package/scripts/benchmark-cross-repo.mjs +206 -206
  523. package/scripts/benchmark-pretrained-retrieval.mjs +305 -305
  524. package/scripts/benchmark-self-learning.mjs +241 -241
  525. package/scripts/benchmark-trajectory-mrr.mjs +191 -191
  526. package/scripts/deploy-ipfs-node.sh +153 -153
  527. package/scripts/grid-search-retrieval.mjs +216 -216
  528. package/scripts/postinstall.cjs +153 -153
  529. package/scripts/pretrain-from-github.mjs +358 -358
  530. package/scripts/publish-registry.ts +345 -345
  531. package/scripts/publish.sh +57 -57
  532. package/scripts/run-beir-bge.mjs +324 -324
  533. package/scripts/run-beir-hybrid.mjs +261 -261
  534. package/scripts/run-beir-lucene-bm25.mjs +154 -154
  535. package/scripts/run-beir-nfcorpus.mjs +231 -231
  536. package/scripts/run-beir-rrf-ablation.mjs +221 -221
  537. package/scripts/setup-ipfs-registry.md +366 -366
  538. package/dist/src/ruvector/flash-attention.d.ts +0 -195
  539. package/dist/src/ruvector/flash-attention.d.ts.map +0 -1
  540. package/dist/src/ruvector/flash-attention.js +0 -643
  541. package/dist/src/ruvector/flash-attention.js.map +0 -1
  542. package/dist/src/ruvector/moe-router.d.ts +0 -206
  543. package/dist/src/ruvector/moe-router.d.ts.map +0 -1
  544. package/dist/src/ruvector/moe-router.js +0 -626
  545. package/dist/src/ruvector/moe-router.js.map +0 -1
  546. package/dist/src/services/event-stream.d.ts +0 -25
  547. package/dist/src/services/event-stream.js +0 -27
  548. package/dist/src/services/loop-worker-runner.d.ts +0 -16
  549. package/dist/src/services/loop-worker-runner.js +0 -34
  550. package/dist/src/services/runtime-capabilities.d.ts +0 -22
  551. package/dist/src/services/runtime-capabilities.js +0 -45
@@ -0,0 +1,241 @@
1
+ /**
2
+ * GAIA CodeAgent Smoke Tests — ADR-138 iter 54 (FINAL design)
3
+ *
4
+ * 5 tests validating the CodeAgent harness end-to-end.
5
+ * Tests 1-4 are unit-level (no API calls). Test 5 uses 1 API call.
6
+ *
7
+ * Tests:
8
+ * T1: Code block extraction — parser correctness (no subprocess)
9
+ * T2: Python runner — simple math (subprocess, no API)
10
+ * T3: Python runner — file read via ATTACHMENT_PATH
11
+ * T4: Python runner — code error recovery (traceback returned as obs)
12
+ * T5: End-to-end — simple factual question (1 API call, ~$0.01)
13
+ *
14
+ * Pass criteria: 5/5. Gate before iter 55 5Q run.
15
+ *
16
+ * Run: npx tsx gaia-codeagent.smoke.ts
17
+ * or: node dist/src/benchmarks/gaia-codeagent.smoke.js
18
+ *
19
+ * Refs: ADR-138, #2156, iter 54
20
+ */
21
+ // ---------------------------------------------------------------------------
22
+ // Imports
23
+ // ---------------------------------------------------------------------------
24
+ import * as fs from 'node:fs';
25
+ import * as os from 'node:os';
26
+ import * as path from 'node:path';
27
+ import { extractCodeBlock, runCodeAgentStep, runGaiaCodeAgent, } from './gaia-codeagent.js';
28
+ import { resolveAnthropicApiKey } from './gaia-agent.js';
29
+ function makeQ(overrides = {}) {
30
+ return {
31
+ task_id: 'smoke-test',
32
+ level: 1,
33
+ question: 'What is 2 + 2?',
34
+ final_answer: '4',
35
+ file_name: null,
36
+ file_path: null,
37
+ ...overrides,
38
+ };
39
+ }
40
+ // ---------------------------------------------------------------------------
41
+ // T1: Code block extraction (no subprocess, no API)
42
+ // ---------------------------------------------------------------------------
43
+ function testCodeBlockExtraction() {
44
+ const name = 'T1: Code block extraction';
45
+ const start = Date.now();
46
+ const cases = [
47
+ { input: '```python\nprint(2+2)\n```', expected: 'print(2+2)' },
48
+ { input: '```py\nresult = 42\n```', expected: 'result = 42' },
49
+ { input: '```\nfinal_answer("test")\n```', expected: 'final_answer("test")' },
50
+ { input: 'No code block here', expected: null },
51
+ {
52
+ input: 'Reasoning...\n```python\nweb_search("q")\nprint(r)\n```\nMore',
53
+ expected: 'web_search("q")\nprint(r)',
54
+ },
55
+ ];
56
+ for (const { input, expected } of cases) {
57
+ const result = extractCodeBlock(input);
58
+ if (result !== expected) {
59
+ return {
60
+ name,
61
+ passed: false,
62
+ message: `Case "${input.slice(0, 40)}": got "${result}", expected "${expected}"`,
63
+ durationMs: Date.now() - start,
64
+ };
65
+ }
66
+ }
67
+ return { name, passed: true, message: `${cases.length}/5 cases passed`, durationMs: Date.now() - start };
68
+ }
69
+ // ---------------------------------------------------------------------------
70
+ // T2: Python runner — simple math (subprocess, no API)
71
+ // ---------------------------------------------------------------------------
72
+ function testPythonRunnerMath() {
73
+ const name = 'T2: Python runner — simple math';
74
+ const start = Date.now();
75
+ const apiKey = process.env.ANTHROPIC_API_KEY ?? 'test-key';
76
+ const code = `result = 2 + 2\nprint(result)\nfinal_answer(str(result))`;
77
+ const stepResult = runCodeAgentStep(code, null, 15_000, apiKey);
78
+ if (stepResult.finalAnswer !== '4') {
79
+ return {
80
+ name,
81
+ passed: false,
82
+ message: `Expected finalAnswer "4", got "${stepResult.finalAnswer ?? 'null'}". Obs: ${stepResult.observation.slice(0, 200)}`,
83
+ durationMs: Date.now() - start,
84
+ };
85
+ }
86
+ return { name, passed: true, message: `finalAnswer="4" as expected`, durationMs: Date.now() - start };
87
+ }
88
+ // ---------------------------------------------------------------------------
89
+ // T3: Python runner — file read via ATTACHMENT_PATH
90
+ // ---------------------------------------------------------------------------
91
+ function testPythonRunnerFileRead() {
92
+ const name = 'T3: Python runner — file read';
93
+ const start = Date.now();
94
+ const tmpFile = path.join(os.tmpdir(), `gaia-smoke-${Date.now()}.txt`);
95
+ fs.writeFileSync(tmpFile, 'The magic number is forty-two (42).', 'utf-8');
96
+ try {
97
+ const apiKey = process.env.ANTHROPIC_API_KEY ?? 'test-key';
98
+ const code = `content = read_file(ATTACHMENT_PATH)\nprint("Read:", content[:80])\nfinal_answer("42")`;
99
+ const stepResult = runCodeAgentStep(code, tmpFile, 15_000, apiKey);
100
+ if (stepResult.finalAnswer !== '42') {
101
+ return {
102
+ name, passed: false,
103
+ message: `Expected finalAnswer "42", got "${stepResult.finalAnswer ?? 'null'}". Obs: ${stepResult.observation.slice(0, 200)}`,
104
+ durationMs: Date.now() - start,
105
+ };
106
+ }
107
+ if (!stepResult.observation.includes('forty-two') && !stepResult.observation.includes('42')) {
108
+ return {
109
+ name, passed: false,
110
+ message: `File content not in observation. Obs: ${stepResult.observation.slice(0, 200)}`,
111
+ durationMs: Date.now() - start,
112
+ };
113
+ }
114
+ return { name, passed: true, message: `File read OK, finalAnswer="42"`, durationMs: Date.now() - start };
115
+ }
116
+ finally {
117
+ try {
118
+ fs.unlinkSync(tmpFile);
119
+ }
120
+ catch { /* ignore */ }
121
+ }
122
+ }
123
+ // ---------------------------------------------------------------------------
124
+ // T4: Python runner — code error recovery
125
+ // ---------------------------------------------------------------------------
126
+ function testPythonRunnerErrorRecovery() {
127
+ const name = 'T4: Python runner — error recovery';
128
+ const start = Date.now();
129
+ const apiKey = process.env.ANTHROPIC_API_KEY ?? 'test-key';
130
+ const badCode = `result = undefined_variable_xyz + 1\nprint(result)`;
131
+ const stepResult = runCodeAgentStep(badCode, null, 15_000, apiKey);
132
+ if (stepResult.finalAnswer !== null) {
133
+ return {
134
+ name, passed: false,
135
+ message: `Expected no finalAnswer from errored code, got "${stepResult.finalAnswer}"`,
136
+ durationMs: Date.now() - start,
137
+ };
138
+ }
139
+ const obs = stepResult.observation.toLowerCase();
140
+ if (!obs.includes('error') && !obs.includes('nameerror')) {
141
+ return {
142
+ name, passed: false,
143
+ message: `Observation should contain error info, got: ${stepResult.observation.slice(0, 200)}`,
144
+ durationMs: Date.now() - start,
145
+ };
146
+ }
147
+ return {
148
+ name, passed: true,
149
+ message: `Error returned as observation: "${stepResult.observation.slice(0, 60)}..."`,
150
+ durationMs: Date.now() - start,
151
+ };
152
+ }
153
+ // ---------------------------------------------------------------------------
154
+ // T5: End-to-end — simple factual question (1 API call)
155
+ // ---------------------------------------------------------------------------
156
+ async function testEndToEndSimple() {
157
+ const name = 'T5: End-to-end — simple factual (1 API call)';
158
+ const start = Date.now();
159
+ let apiKey;
160
+ try {
161
+ apiKey = resolveAnthropicApiKey();
162
+ }
163
+ catch (err) {
164
+ return {
165
+ name, passed: false,
166
+ message: `No API key: ${String(err)}`,
167
+ durationMs: Date.now() - start,
168
+ };
169
+ }
170
+ const q = makeQ({
171
+ task_id: 'smoke-e2e-01',
172
+ question: 'What is 6 multiplied by 7? Answer as a number.',
173
+ final_answer: '42',
174
+ });
175
+ try {
176
+ const result = await runGaiaCodeAgent(q, {
177
+ model: 'claude-sonnet-4-6',
178
+ maxTurns: 5,
179
+ apiKey,
180
+ });
181
+ if (result.error && !result.finalAnswer) {
182
+ return { name, passed: false, message: `Agent error: ${result.error}`, durationMs: Date.now() - start };
183
+ }
184
+ if (result.finalAnswer === null) {
185
+ return { name, passed: false, message: `No final answer (timedOut=${result.timedOut})`, durationMs: Date.now() - start };
186
+ }
187
+ const num = parseFloat(result.finalAnswer.replace(/,/g, ''));
188
+ if (Math.abs(num - 42) > 0.01) {
189
+ return {
190
+ name, passed: false,
191
+ message: `Expected "42", got "${result.finalAnswer}" (turns=${result.turns})`,
192
+ durationMs: Date.now() - start,
193
+ };
194
+ }
195
+ const estCost = (result.totalInputTokens / 1e6) * 3.0 + (result.totalOutputTokens / 1e6) * 15.0;
196
+ return {
197
+ name, passed: true,
198
+ message: `finalAnswer="${result.finalAnswer}" turns=${result.turns} cost~=$${estCost.toFixed(4)}`,
199
+ durationMs: Date.now() - start,
200
+ };
201
+ }
202
+ catch (err) {
203
+ return { name, passed: false, message: `Exception: ${String(err)}`, durationMs: Date.now() - start };
204
+ }
205
+ }
206
+ // ---------------------------------------------------------------------------
207
+ // Main
208
+ // ---------------------------------------------------------------------------
209
+ async function main() {
210
+ console.log('\n=== GAIA CodeAgent Smoke Tests (ADR-138 iter 54) ===\n');
211
+ const results = [
212
+ testCodeBlockExtraction(),
213
+ testPythonRunnerMath(),
214
+ testPythonRunnerFileRead(),
215
+ testPythonRunnerErrorRecovery(),
216
+ await testEndToEndSimple(),
217
+ ];
218
+ let passed = 0;
219
+ for (const r of results) {
220
+ const status = r.passed ? 'PASS' : 'FAIL';
221
+ console.log(`[${status}] ${r.name} (${r.durationMs}ms)`);
222
+ console.log(` ${r.message}`);
223
+ if (r.passed)
224
+ passed++;
225
+ }
226
+ console.log(`\n=== ${passed}/${results.length} passed ===`);
227
+ if (passed === results.length) {
228
+ console.log('All smoke tests passed. CodeAgent harness ready for 1Q sanity check.\n');
229
+ process.exit(0);
230
+ }
231
+ else {
232
+ console.log(`SMOKE FAILED — ${results.length - passed} test(s) failed.\n`);
233
+ process.exit(1);
234
+ }
235
+ }
236
+ main().catch((err) => {
237
+ console.error('Smoke test crashed:', err);
238
+ process.exit(2);
239
+ });
240
+ export { testCodeBlockExtraction, testPythonRunnerMath, testPythonRunnerFileRead, testPythonRunnerErrorRecovery, testEndToEndSimple, };
241
+ //# sourceMappingURL=gaia-codeagent.smoke.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"gaia-codeagent.smoke.js","sourceRoot":"","sources":["../../../src/benchmarks/gaia-codeagent.smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAEH,8EAA8E;AAC9E,UAAU;AACV,8EAA8E;AAE9E,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,EACL,gBAAgB,EAChB,gBAAgB,EAChB,gBAAgB,GAEjB,MAAM,qBAAqB,CAAC;AAE7B,OAAO,EAAE,sBAAsB,EAAE,MAAM,iBAAiB,CAAC;AAazD,SAAS,KAAK,CAAC,YAAmC,EAAE;IAClD,OAAO;QACL,OAAO,EAAE,YAAY;QACrB,KAAK,EAAE,CAAU;QACjB,QAAQ,EAAE,gBAAgB;QAC1B,YAAY,EAAE,GAAG;QACjB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,IAAI;QACf,GAAG,SAAS;KACb,CAAC;AACJ,CAAC;AAED,8EAA8E;AAC9E,oDAAoD;AACpD,8EAA8E;AAE9E,SAAS,uBAAuB;IAC9B,MAAM,IAAI,GAAG,2BAA2B,CAAC;IACzC,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAEzB,MAAM,KAAK,GAAsD;QAC/D,EAAE,KAAK,EAAE,4BAA4B,EAAE,QAAQ,EAAE,YAAY,EAAE;QAC/D,EAAE,KAAK,EAAE,yBAAyB,EAAE,QAAQ,EAAE,aAAa,EAAE;QAC7D,EAAE,KAAK,EAAE,gCAAgC,EAAE,QAAQ,EAAE,sBAAsB,EAAE;QAC7E,EAAE,KAAK,EAAE,oBAAoB,EAAE,QAAQ,EAAE,IAAI,EAAE;QAC/C;YACE,KAAK,EAAE,+DAA+D;YACtE,QAAQ,EAAE,2BAA2B;SACtC;KACF,CAAC;IAEF,KAAK,MAAM,EAAE,KAAK,EAAE,QAAQ,EAAE,IAAI,KAAK,EAAE,CAAC;QACxC,MAAM,MAAM,GAAG,gBAAgB,CAAC,KAAK,CAAC,CAAC;QACvC,IAAI,MAAM,KAAK,QAAQ,EAAE,CAAC;YACxB,OAAO;gBACL,IAAI;gBACJ,MAAM,EAAE,KAAK;gBACb,OAAO,EAAE,SAAS,KAAK,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,WAAW,MAAM,gBAAgB,QAAQ,GAAG;gBAChF,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;aAC/B,CAAC;QACJ,CAAC;IACH,CAAC;IACD,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,GAAG,KAAK,CAAC,MAAM,iBAAiB,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,CAAC;AAC3G,CAAC;AAED,8EAA8E;AAC9E,uDAAuD;AACvD,8EAA8E;AAE9E,SAAS,oBAAoB;IAC3B,MAAM,IAAI,GAAG,iCAAiC,CAAC;IAC/C,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAEzB,MAAM,MAAM,GAAG,OAAO,CAAC,GAAG,CAAC,iBAAiB,IAAI,UAAU,CAAC;IAC3D,MAAM,IAAI,GAAG,0DAA0D,CAAC;IACxE,MAAM,UAAU,GAAG,gBAAgB,CAAC,IAAI,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;IAEhE,IAAI,UAAU,CAAC,WAAW,KAAK,GAAG,EAAE,CAAC;QACnC,OAAO;YACL,IAAI;YACJ,MAAM,EAAE,KAAK;YACb,OAAO,EAAE,kCAAkC,UAAU,CAAC,WAAW,IAAI,MAAM,WAAW,UAAU,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE;YAC5H,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;SAC/B,CAAC;IACJ,CAAC;IACD,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,6BAA6B,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,CAAC;AACxG,CAAC;AAED,8EAA8E;AAC9E,oDAAoD;AACpD,8EAA8E;AAE9E,SAAS,wBAAwB;IAC/B,MAAM,IAAI,GAAG,+BAA+B,CAAC;IAC7C,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IACzB,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,MAAM,EAAE,EAAE,cAAc,IAAI,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;IACvE,EAAE,CAAC,aAAa,CAAC,OAAO,EAAE,qCAAqC,EAAE,OAAO,CAAC,CAAC;IAE1E,IAAI,CAAC;QACH,MAAM,MAAM,GAAG,OAAO,CAAC,GAAG,CAAC,iBAAiB,IAAI,UAAU,CAAC;QAC3D,MAAM,IAAI,GAAG,wFAAwF,CAAC;QACtG,MAAM,UAAU,GAAG,gBAAgB,CAAC,IAAI,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;QAEnE,IAAI,UAAU,CAAC,WAAW,KAAK,IAAI,EAAE,CAAC;YACpC,OAAO;gBACL,IAAI,EAAE,MAAM,EAAE,KAAK;gBACnB,OAAO,EAAE,mCAAmC,UAAU,CAAC,WAAW,IAAI,MAAM,WAAW,UAAU,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE;gBAC7H,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;aAC/B,CAAC;QACJ,CAAC;QACD,IAAI,CAAC,UAAU,CAAC,WAAW,CAAC,QAAQ,CAAC,WAAW,CAAC,IAAI,CAAC,UAAU,CAAC,WAAW,CAAC,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC;YAC5F,OAAO;gBACL,IAAI,EAAE,MAAM,EAAE,KAAK;gBACnB,OAAO,EAAE,yCAAyC,UAAU,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE;gBACxF,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;aAC/B,CAAC;QACJ,CAAC;QACD,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,gCAAgC,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,CAAC;IAC3G,CAAC;YAAS,CAAC;QACT,IAAI,CAAC;YAAC,EAAE,CAAC,UAAU,CAAC,OAAO,CAAC,CAAC;QAAC,CAAC;QAAC,MAAM,CAAC,CAAC,YAAY,CAAC,CAAC;IACxD,CAAC;AACH,CAAC;AAED,8EAA8E;AAC9E,0CAA0C;AAC1C,8EAA8E;AAE9E,SAAS,6BAA6B;IACpC,MAAM,IAAI,GAAG,oCAAoC,CAAC;IAClD,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAEzB,MAAM,MAAM,GAAG,OAAO,CAAC,GAAG,CAAC,iBAAiB,IAAI,UAAU,CAAC;IAC3D,MAAM,OAAO,GAAG,oDAAoD,CAAC;IACrE,MAAM,UAAU,GAAG,gBAAgB,CAAC,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;IAEnE,IAAI,UAAU,CAAC,WAAW,KAAK,IAAI,EAAE,CAAC;QACpC,OAAO;YACL,IAAI,EAAE,MAAM,EAAE,KAAK;YACnB,OAAO,EAAE,mDAAmD,UAAU,CAAC,WAAW,GAAG;YACrF,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;SAC/B,CAAC;IACJ,CAAC;IACD,MAAM,GAAG,GAAG,UAAU,CAAC,WAAW,CAAC,WAAW,EAAE,CAAC;IACjD,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;QACzD,OAAO;YACL,IAAI,EAAE,MAAM,EAAE,KAAK;YACnB,OAAO,EAAE,+CAA+C,UAAU,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE;YAC9F,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;SAC/B,CAAC;IACJ,CAAC;IACD,OAAO;QACL,IAAI,EAAE,MAAM,EAAE,IAAI;QAClB,OAAO,EAAE,mCAAmC,UAAU,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,MAAM;QACrF,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;KAC/B,CAAC;AACJ,CAAC;AAED,8EAA8E;AAC9E,wDAAwD;AACxD,8EAA8E;AAE9E,KAAK,UAAU,kBAAkB;IAC/B,MAAM,IAAI,GAAG,8CAA8C,CAAC;IAC5D,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAEzB,IAAI,MAAc,CAAC;IACnB,IAAI,CAAC;QACH,MAAM,GAAG,sBAAsB,EAAE,CAAC;IACpC,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,OAAO;YACL,IAAI,EAAE,MAAM,EAAE,KAAK;YACnB,OAAO,EAAE,eAAe,MAAM,CAAC,GAAG,CAAC,EAAE;YACrC,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;SAC/B,CAAC;IACJ,CAAC;IAED,MAAM,CAAC,GAAG,KAAK,CAAC;QACd,OAAO,EAAE,cAAc;QACvB,QAAQ,EAAE,gDAAgD;QAC1D,YAAY,EAAE,IAAI;KACnB,CAAC,CAAC;IAEH,IAAI,CAAC;QACH,MAAM,MAAM,GAAG,MAAM,gBAAgB,CAAC,CAAC,EAAE;YACvC,KAAK,EAAE,mBAAmB;YAC1B,QAAQ,EAAE,CAAC;YACX,MAAM;SACP,CAAC,CAAC;QAEH,IAAI,MAAM,CAAC,KAAK,IAAI,CAAC,MAAM,CAAC,WAAW,EAAE,CAAC;YACxC,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,OAAO,EAAE,gBAAgB,MAAM,CAAC,KAAK,EAAE,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,CAAC;QAC1G,CAAC;QACD,IAAI,MAAM,CAAC,WAAW,KAAK,IAAI,EAAE,CAAC;YAChC,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,OAAO,EAAE,6BAA6B,MAAM,CAAC,QAAQ,GAAG,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,CAAC;QAC3H,CAAC;QACD,MAAM,GAAG,GAAG,UAAU,CAAC,MAAM,CAAC,WAAW,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,CAAC;QAC7D,IAAI,IAAI,CAAC,GAAG,CAAC,GAAG,GAAG,EAAE,CAAC,GAAG,IAAI,EAAE,CAAC;YAC9B,OAAO;gBACL,IAAI,EAAE,MAAM,EAAE,KAAK;gBACnB,OAAO,EAAE,uBAAuB,MAAM,CAAC,WAAW,YAAY,MAAM,CAAC,KAAK,GAAG;gBAC7E,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;aAC/B,CAAC;QACJ,CAAC;QACD,MAAM,OAAO,GAAG,CAAC,MAAM,CAAC,gBAAgB,GAAG,GAAG,CAAC,GAAG,GAAG,GAAG,CAAC,MAAM,CAAC,iBAAiB,GAAG,GAAG,CAAC,GAAG,IAAI,CAAC;QAChG,OAAO;YACL,IAAI,EAAE,MAAM,EAAE,IAAI;YAClB,OAAO,EAAE,gBAAgB,MAAM,CAAC,WAAW,WAAW,MAAM,CAAC,KAAK,WAAW,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE;YACjG,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;SAC/B,CAAC;IACJ,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,OAAO,EAAE,cAAc,MAAM,CAAC,GAAG,CAAC,EAAE,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,EAAE,CAAC;IACvG,CAAC;AACH,CAAC;AAED,8EAA8E;AAC9E,OAAO;AACP,8EAA8E;AAE9E,KAAK,UAAU,IAAI;IACjB,OAAO,CAAC,GAAG,CAAC,0DAA0D,CAAC,CAAC;IAExE,MAAM,OAAO,GAAiB;QAC5B,uBAAuB,EAAE;QACzB,oBAAoB,EAAE;QACtB,wBAAwB,EAAE;QAC1B,6BAA6B,EAAE;QAC/B,MAAM,kBAAkB,EAAE;KAC3B,CAAC;IAEF,IAAI,MAAM,GAAG,CAAC,CAAC;IACf,KAAK,MAAM,CAAC,IAAI,OAAO,EAAE,CAAC;QACxB,MAAM,MAAM,GAAG,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC;QAC1C,OAAO,CAAC,GAAG,CAAC,IAAI,MAAM,KAAK,CAAC,CAAC,IAAI,KAAK,CAAC,CAAC,UAAU,KAAK,CAAC,CAAC;QACzD,OAAO,CAAC,GAAG,CAAC,UAAU,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC;QACnC,IAAI,CAAC,CAAC,MAAM;YAAE,MAAM,EAAE,CAAC;IACzB,CAAC;IAED,OAAO,CAAC,GAAG,CAAC,SAAS,MAAM,IAAI,OAAO,CAAC,MAAM,aAAa,CAAC,CAAC;IAC5D,IAAI,MAAM,KAAK,OAAO,CAAC,MAAM,EAAE,CAAC;QAC9B,OAAO,CAAC,GAAG,CAAC,wEAAwE,CAAC,CAAC;QACtF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,GAAG,CAAC,kBAAkB,OAAO,CAAC,MAAM,GAAG,MAAM,oBAAoB,CAAC,CAAC;QAC3E,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;AACH,CAAC;AAED,IAAI,EAAE,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,EAAE;IACnB,OAAO,CAAC,KAAK,CAAC,qBAAqB,EAAE,GAAG,CAAC,CAAC;IAC1C,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC,CAAC,CAAC;AAEH,OAAO,EACL,uBAAuB,EACvB,oBAAoB,EACpB,wBAAwB,EACxB,6BAA6B,EAC7B,kBAAkB,GACnB,CAAC"}
@@ -70,25 +70,25 @@ function resolveApiKey(override) {
70
70
  * Instructs Sonnet to act as an adversarial reviewer and respond in JSON.
71
71
  */
72
72
  function buildCriticSystemPrompt() {
73
- return `You are an adversarial reviewer of agent answers on the GAIA benchmark. \
74
- Your job is to find flaws in candidate answers before they are submitted.
75
-
76
- Given a question, a candidate answer, and a summary of the agent's reasoning \
77
- trajectory, you must decide whether the candidate answer is correct.
78
-
79
- Evaluation criteria:
80
- 1. Does the answer directly address what the question asks?
81
- 2. Is there evidence in the trajectory that supports the answer?
82
- 3. Are there obvious flaws (wrong unit, wrong format, missed constraint, \
83
- fabricated source, off-by-one error, truncation)?
84
- 4. For numeric answers: is the magnitude and unit plausible?
85
- 5. For list answers: are all required items present and correctly ordered?
86
-
87
- Respond ONLY with a JSON object — no markdown, no prose outside the JSON:
88
- {"verdict":"pass"|"fail"|"uncertain","reasoning":"<1-3 sentences>","suggestedRevision":"<corrected answer or empty string>"}
89
-
90
- Use "uncertain" only when you genuinely cannot determine correctness from the \
91
- available information. Never use "uncertain" to avoid making a decision when \
73
+ return `You are an adversarial reviewer of agent answers on the GAIA benchmark. \
74
+ Your job is to find flaws in candidate answers before they are submitted.
75
+
76
+ Given a question, a candidate answer, and a summary of the agent's reasoning \
77
+ trajectory, you must decide whether the candidate answer is correct.
78
+
79
+ Evaluation criteria:
80
+ 1. Does the answer directly address what the question asks?
81
+ 2. Is there evidence in the trajectory that supports the answer?
82
+ 3. Are there obvious flaws (wrong unit, wrong format, missed constraint, \
83
+ fabricated source, off-by-one error, truncation)?
84
+ 4. For numeric answers: is the magnitude and unit plausible?
85
+ 5. For list answers: are all required items present and correctly ordered?
86
+
87
+ Respond ONLY with a JSON object — no markdown, no prose outside the JSON:
88
+ {"verdict":"pass"|"fail"|"uncertain","reasoning":"<1-3 sentences>","suggestedRevision":"<corrected answer or empty string>"}
89
+
90
+ Use "uncertain" only when you genuinely cannot determine correctness from the \
91
+ available information. Never use "uncertain" to avoid making a decision when \
92
92
  evidence is available.`;
93
93
  }
94
94
  /**
@@ -108,11 +108,11 @@ function buildCriticUserMessage(question, candidateAnswer, trajectory) {
108
108
  const result = s.result ? ` → ${s.result.slice(0, 200)}` : '';
109
109
  return `${label}${result}`;
110
110
  }).join('\n');
111
- return `QUESTION: ${question.question}
112
-
113
- CANDIDATE ANSWER: ${candidateAnswer}
114
-
115
- AGENT TRAJECTORY (${trajectory.turns} turns, summarised):
111
+ return `QUESTION: ${question.question}
112
+
113
+ CANDIDATE ANSWER: ${candidateAnswer}
114
+
115
+ AGENT TRAJECTORY (${trajectory.turns} turns, summarised):
116
116
  ${trajectoryText}`;
117
117
  }
118
118
  /**
@@ -0,0 +1,22 @@
1
+ /**
2
+ * GAIA DAG 5-Question Pilot Runner — iter 64
3
+ *
4
+ * Runs 5 specific GAIA L1 questions through the Co-Sight DAG harness
5
+ * and compares results against single-Sonnet baseline (from iter 63b).
6
+ *
7
+ * Pilot question mix (selected from iter 63b failures):
8
+ * 1. 5d0080cb — calculation (fish bag volume from academic paper, multi-hop)
9
+ * 2. cffe0e32 — reasoning puzzle (Secret Santa assignment chain)
10
+ * 3. ec09fa32 — riddle (game show, requires careful reasoning)
11
+ * 4. 46719c30 — retrieval (paper authors -> their other publications)
12
+ * 5. b816bfce — retrieval+reasoning (journal name from Norse mythology)
13
+ *
14
+ * Cost cap: $2.00
15
+ *
16
+ * Usage:
17
+ * node dist/src/benchmarks/gaia-dag-pilot.js
18
+ *
19
+ * Refs: ADR-139, iter 64, #2156
20
+ */
21
+ export {};
22
+ //# sourceMappingURL=gaia-dag-pilot.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"gaia-dag-pilot.d.ts","sourceRoot":"","sources":["../../../src/benchmarks/gaia-dag-pilot.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG"}
@@ -0,0 +1,158 @@
1
+ /**
2
+ * GAIA DAG 5-Question Pilot Runner — iter 64
3
+ *
4
+ * Runs 5 specific GAIA L1 questions through the Co-Sight DAG harness
5
+ * and compares results against single-Sonnet baseline (from iter 63b).
6
+ *
7
+ * Pilot question mix (selected from iter 63b failures):
8
+ * 1. 5d0080cb — calculation (fish bag volume from academic paper, multi-hop)
9
+ * 2. cffe0e32 — reasoning puzzle (Secret Santa assignment chain)
10
+ * 3. ec09fa32 — riddle (game show, requires careful reasoning)
11
+ * 4. 46719c30 — retrieval (paper authors -> their other publications)
12
+ * 5. b816bfce — retrieval+reasoning (journal name from Norse mythology)
13
+ *
14
+ * Cost cap: $2.00
15
+ *
16
+ * Usage:
17
+ * node dist/src/benchmarks/gaia-dag-pilot.js
18
+ *
19
+ * Refs: ADR-139, iter 64, #2156
20
+ */
21
+ import { loadGaia } from './gaia-loader.js';
22
+ import { runDagPilot } from './gaia-dag.js';
23
+ // ---------------------------------------------------------------------------
24
+ // Pilot question IDs (5 questions from iter 63b failures)
25
+ // ---------------------------------------------------------------------------
26
+ const PILOT_TASK_IDS = [
27
+ '5d0080cb-90d7-4712-bc33-848150e917d3', // calculation: fish bag volume
28
+ 'cffe0e32-c9a6-4c52-9877-78ceb4aaa9fb', // reasoning: Secret Santa
29
+ 'ec09fa32-d03f-4bf8-84b0-1f16922c3ae4', // riddle: game show bees
30
+ '46719c30-f4c3-4cad-be07-d5cb21eee6bb', // retrieval: paper authors + their work
31
+ 'b816bfce-3d80-4913-a07d-69b752ce6377', // retrieval: journal from Norse mythology
32
+ ];
33
+ // Baseline: single-Sonnet results from iter 63b
34
+ const BASELINE = {
35
+ '5d0080cb-90d7-4712-bc33-848150e917d3': { correct: false, answer: null },
36
+ 'cffe0e32-c9a6-4c52-9877-78ceb4aaa9fb': { correct: false, answer: null },
37
+ 'ec09fa32-d03f-4bf8-84b0-1f16922c3ae4': { correct: false, answer: null },
38
+ '46719c30-f4c3-4cad-be07-d5cb21eee6bb': { correct: false, answer: null },
39
+ 'b816bfce-3d80-4913-a07d-69b752ce6377': { correct: false, answer: 'cuddly' }, // close but wrong
40
+ };
41
+ const COST_CAP_USD = 2.00;
42
+ async function main() {
43
+ console.log('\n=== GAIA DAG 5-Question Pilot (iter 64) ===\n');
44
+ console.log('Questions: 5 (from iter 63b failures)');
45
+ console.log(`Planner: ${process.env['PLAN_MODEL'] ?? 'claude-sonnet-4-6'}`);
46
+ console.log(`Actor: ${process.env['ACT_MODEL'] ?? 'gemini-2.5-pro'}`);
47
+ console.log(`Cost cap: $${COST_CAP_USD}\n`);
48
+ // Load full GAIA L1 dataset and filter to pilot questions
49
+ console.log('Loading GAIA L1 dataset...');
50
+ let allQuestions;
51
+ try {
52
+ allQuestions = await loadGaia({ level: 1 });
53
+ }
54
+ catch (err) {
55
+ console.error('Failed to load GAIA dataset:', err instanceof Error ? err.message : String(err));
56
+ process.exit(1);
57
+ }
58
+ const pilotQuestions = PILOT_TASK_IDS.map((id) => {
59
+ const q = allQuestions.find((q) => q.task_id === id);
60
+ if (!q) {
61
+ console.warn(`WARNING: Question ${id} not found in dataset`);
62
+ }
63
+ return q;
64
+ }).filter(Boolean);
65
+ if (pilotQuestions.length === 0) {
66
+ console.error('No pilot questions found. Check HF token and dataset availability.');
67
+ process.exit(1);
68
+ }
69
+ console.log(`Loaded ${pilotQuestions.length} pilot questions\n`);
70
+ for (const q of pilotQuestions) {
71
+ console.log(` ${q.task_id.slice(0, 8)}: ${q.question.slice(0, 70)}...`);
72
+ console.log(` expected: "${q.final_answer}"`);
73
+ if (q.file_path)
74
+ console.log(` file: ${q.file_path}`);
75
+ }
76
+ console.log('');
77
+ // Run DAG pilot
78
+ const result = await runDagPilot(pilotQuestions, {});
79
+ // Check cost cap
80
+ if (result.totalCostUsd > COST_CAP_USD) {
81
+ console.warn(`\nWARNING: Cost $${result.totalCostUsd.toFixed(4)} exceeded cap $${COST_CAP_USD}\n`);
82
+ }
83
+ // Report results
84
+ console.log('\n=== DAG Pilot Results ===\n');
85
+ console.log(`Score: ${result.correct}/${result.total} (${(result.accuracy * 100).toFixed(1)}%)`);
86
+ console.log(`Avg steps/question: ${result.avgStepsPerQuestion.toFixed(1)}`);
87
+ console.log(`Total cost: $${result.totalCostUsd.toFixed(4)}`);
88
+ console.log(`Projected 53Q cost: $${result.projectedCost53Q.toFixed(2)}`);
89
+ console.log(`Mean wall time: ${(result.meanWallMs / 1000).toFixed(1)}s\n`);
90
+ console.log('Per-question breakdown:');
91
+ const baselineCorrect = Object.values(BASELINE).filter((b) => b.correct).length;
92
+ let dagBetter = 0;
93
+ let dagWorse = 0;
94
+ for (const q of result.perQuestion) {
95
+ const baseline = BASELINE[q.taskId];
96
+ const baselineResult = baseline?.correct ? 'PASS' : 'FAIL';
97
+ const dagResult = q.correct ? 'PASS' : 'FAIL';
98
+ const change = !baseline?.correct && q.correct ? ' (+RECOVERED)' :
99
+ baseline?.correct && !q.correct ? ' (-REGRESSED)' : '';
100
+ if (!baseline?.correct && q.correct)
101
+ dagBetter++;
102
+ if (baseline?.correct && !q.correct)
103
+ dagWorse++;
104
+ console.log(` ${dagResult} [was ${baselineResult}]${change} ${q.taskId.slice(0, 8)}: got="${q.got?.slice(0, 40) ?? 'null'}" expected="${q.expected.slice(0, 40)}"`);
105
+ console.log(` steps=${q.steps} (completed=${q.completedSteps}, blocked=${q.blockedSteps}) plannerCycles=${q.plannerCycles} cost=$${q.costUsd.toFixed(4)} wall=${(q.wallMs / 1000).toFixed(1)}s`);
106
+ }
107
+ console.log(`\nComparison vs single-Sonnet (iter 63b):`);
108
+ console.log(` Baseline: ${baselineCorrect}/5 (${(baselineCorrect * 20).toFixed(0)}%)`);
109
+ console.log(` DAG: ${result.correct}/5 (${(result.accuracy * 100).toFixed(1)}%)`);
110
+ console.log(` Recovered: ${dagBetter} questions (single-Sonnet failed, DAG solved)`);
111
+ console.log(` Regressed: ${dagWorse} questions (single-Sonnet passed, DAG failed)`);
112
+ console.log('\n=== Gate Verdict ===');
113
+ const multiStepExecuted = result.perQuestion.some((q) => q.steps > 1);
114
+ const avgSteps = result.avgStepsPerQuestion;
115
+ if (result.correct >= 4 && multiStepExecuted) {
116
+ console.log(`PROCEED to iter 65 full 53Q run.`);
117
+ console.log(` Reason: DAG ${result.correct}/5, avg ${avgSteps.toFixed(1)} steps/Q, multi-step plans executing.`);
118
+ }
119
+ else if (result.correct >= 3 && avgSteps >= 2) {
120
+ console.log(`PROCEED with CAUTION to iter 65 — diagnose blocking first.`);
121
+ console.log(` Reason: DAG ${result.correct}/5, avg ${avgSteps.toFixed(1)} steps/Q.`);
122
+ }
123
+ else {
124
+ console.log(`DIAGNOSE before iter 65.`);
125
+ console.log(` Reason: DAG ${result.correct}/5, avg ${avgSteps.toFixed(1)} steps/Q.`);
126
+ if (!multiStepExecuted) {
127
+ console.log(` Issue: Plans collapsed to single step — check planner prompt.`);
128
+ }
129
+ }
130
+ // Write JSON result
131
+ const jsonOut = {
132
+ run_type: 'dag-5q-pilot',
133
+ iter: 64,
134
+ timestamp: new Date().toISOString(),
135
+ planModel: process.env['PLAN_MODEL'] ?? 'claude-sonnet-4-6',
136
+ actModel: process.env['ACT_MODEL'] ?? 'gemini-2.5-pro',
137
+ summary: {
138
+ total: result.total,
139
+ correct: result.correct,
140
+ accuracy: result.accuracy,
141
+ avgStepsPerQuestion: result.avgStepsPerQuestion,
142
+ totalCostUsd: result.totalCostUsd,
143
+ projectedCost53Q: result.projectedCost53Q,
144
+ meanWallMs: result.meanWallMs,
145
+ },
146
+ baseline: { correct: baselineCorrect, total: 5, source: 'iter-63b' },
147
+ perQuestion: result.perQuestion,
148
+ };
149
+ const outPath = `/Users/cohen/Projects/ruflo/docs/benchmarks/runs/gaia-l1-iter64-dag-5q-pilot.json`;
150
+ const fs = await import('node:fs');
151
+ fs.writeFileSync(outPath, JSON.stringify(jsonOut, null, 2));
152
+ console.log(`\nResults written to: ${outPath}\n`);
153
+ }
154
+ main().catch((err) => {
155
+ console.error('Pilot failed:', err);
156
+ process.exit(1);
157
+ });
158
+ //# sourceMappingURL=gaia-dag-pilot.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"gaia-dag-pilot.js","sourceRoot":"","sources":["../../../src/benchmarks/gaia-dag-pilot.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAEH,OAAO,EAAE,QAAQ,EAAE,MAAM,kBAAkB,CAAC;AAC5C,OAAO,EAAE,WAAW,EAAkB,MAAM,eAAe,CAAC;AAG5D,8EAA8E;AAC9E,0DAA0D;AAC1D,8EAA8E;AAE9E,MAAM,cAAc,GAAG;IACrB,sCAAsC,EAAE,+BAA+B;IACvE,sCAAsC,EAAE,0BAA0B;IAClE,sCAAsC,EAAE,yBAAyB;IACjE,sCAAsC,EAAE,wCAAwC;IAChF,sCAAsC,EAAE,0CAA0C;CACnF,CAAC;AAEF,gDAAgD;AAChD,MAAM,QAAQ,GAAgE;IAC5E,sCAAsC,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,IAAI,EAAE;IACxE,sCAAsC,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,IAAI,EAAE;IACxE,sCAAsC,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,IAAI,EAAE;IACxE,sCAAsC,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,IAAI,EAAE;IACxE,sCAAsC,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,QAAQ,EAAE,EAAE,kBAAkB;CACjG,CAAC;AAEF,MAAM,YAAY,GAAG,IAAI,CAAC;AAE1B,KAAK,UAAU,IAAI;IACjB,OAAO,CAAC,GAAG,CAAC,iDAAiD,CAAC,CAAC;IAC/D,OAAO,CAAC,GAAG,CAAC,uCAAuC,CAAC,CAAC;IACrD,OAAO,CAAC,GAAG,CAAC,YAAY,OAAO,CAAC,GAAG,CAAC,YAAY,CAAC,IAAI,mBAAmB,EAAE,CAAC,CAAC;IAC5E,OAAO,CAAC,GAAG,CAAC,YAAY,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC,IAAI,gBAAgB,EAAE,CAAC,CAAC;IACxE,OAAO,CAAC,GAAG,CAAC,cAAc,YAAY,IAAI,CAAC,CAAC;IAE5C,0DAA0D;IAC1D,OAAO,CAAC,GAAG,CAAC,4BAA4B,CAAC,CAAC;IAC1C,IAAI,YAAY,CAAC;IACjB,IAAI,CAAC;QACH,YAAY,GAAG,MAAM,QAAQ,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC;IAC9C,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,OAAO,CAAC,KAAK,CAAC,8BAA8B,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;QAChG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,MAAM,cAAc,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,EAAE;QAC/C,MAAM,CAAC,GAAG,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,KAAK,EAAE,CAAC,CAAC;QACrD,IAAI,CAAC,CAAC,EAAE,CAAC;YACP,OAAO,CAAC,IAAI,CAAC,qBAAqB,EAAE,uBAAuB,CAAC,CAAC;QAC/D,CAAC;QACD,OAAO,CAAC,CAAC;IACX,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO,CAAwB,CAAC;IAE1C,IAAI,cAAc,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAChC,OAAO,CAAC,KAAK,CAAC,oEAAoE,CAAC,CAAC;QACpF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,OAAO,CAAC,GAAG,CAAC,UAAU,cAAc,CAAC,MAAM,oBAAoB,CAAC,CAAC;IACjE,KAAK,MAAM,CAAC,IAAI,cAAc,EAAE,CAAC;QAC/B,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,CAAC;QACzE,OAAO,CAAC,GAAG,CAAC,yBAAyB,CAAC,CAAC,YAAY,GAAG,CAAC,CAAC;QACxD,IAAI,CAAC,CAAC,SAAS;YAAE,OAAO,CAAC,GAAG,CAAC,oBAAoB,CAAC,CAAC,SAAS,EAAE,CAAC,CAAC;IAClE,CAAC;IACD,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;IAEhB,gBAAgB;IAChB,MAAM,MAAM,GAAmB,MAAM,WAAW,CAAC,cAAc,EAAE,EAAE,CAAC,CAAC;IAErE,iBAAiB;IACjB,IAAI,MAAM,CAAC,YAAY,GAAG,YAAY,EAAE,CAAC;QACvC,OAAO,CAAC,IAAI,CAAC,oBAAoB,MAAM,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,kBAAkB,YAAY,IAAI,CAAC,CAAC;IACrG,CAAC;IAED,iBAAiB;IACjB,OAAO,CAAC,GAAG,CAAC,+BAA+B,CAAC,CAAC;IAC7C,OAAO,CAAC,GAAG,CAAC,UAAU,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,KAAK,CAAC,MAAM,CAAC,QAAQ,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IACjG,OAAO,CAAC,GAAG,CAAC,uBAAuB,MAAM,CAAC,mBAAmB,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAC5E,OAAO,CAAC,GAAG,CAAC,gBAAgB,MAAM,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAC9D,OAAO,CAAC,GAAG,CAAC,wBAAwB,MAAM,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAC1E,OAAO,CAAC,GAAG,CAAC,mBAAmB,CAAC,MAAM,CAAC,UAAU,GAAG,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC;IAE3E,OAAO,CAAC,GAAG,CAAC,yBAAyB,CAAC,CAAC;IACvC,MAAM,eAAe,GAAG,MAAM,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,MAAM,CAAC;IAChF,IAAI,SAAS,GAAG,CAAC,CAAC;IAClB,IAAI,QAAQ,GAAG,CAAC,CAAC;IAEjB,KAAK,MAAM,CAAC,IAAI,MAAM,CAAC,WAAW,EAAE,CAAC;QACnC,MAAM,QAAQ,GAAG,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;QACpC,MAAM,cAAc,GAAG,QAAQ,EAAE,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC;QAC3D,MAAM,SAAS,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC;QAC9C,MAAM,MAAM,GAAG,CAAC,QAAQ,EAAE,OAAO,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,eAAe,CAAC,CAAC;YACnD,QAAQ,EAAE,OAAO,IAAI,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,eAAe,CAAC,CAAC,CAAC,EAAE,CAAC;QACtE,IAAI,CAAC,QAAQ,EAAE,OAAO,IAAI,CAAC,CAAC,OAAO;YAAE,SAAS,EAAE,CAAC;QACjD,IAAI,QAAQ,EAAE,OAAO,IAAI,CAAC,CAAC,CAAC,OAAO;YAAE,QAAQ,EAAE,CAAC;QAEhD,OAAO,CAAC,GAAG,CAAC,KAAK,SAAS,SAAS,cAAc,IAAI,MAAM,IAAI,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,UAAU,CAAC,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,IAAI,MAAM,eAAe,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,GAAG,CAAC,CAAC;QACrK,OAAO,CAAC,GAAG,CAAC,kBAAkB,CAAC,CAAC,KAAK,eAAe,CAAC,CAAC,cAAc,aAAa,CAAC,CAAC,YAAY,mBAAmB,CAAC,CAAC,aAAa,UAAU,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,MAAM,GAAG,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;IAC3M,CAAC;IAED,OAAO,CAAC,GAAG,CAAC,2CAA2C,CAAC,CAAC;IACzD,OAAO,CAAC,GAAG,CAAC,eAAe,eAAe,OAAO,CAAC,eAAe,GAAG,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IACxF,OAAO,CAAC,GAAG,CAAC,eAAe,MAAM,CAAC,OAAO,OAAO,CAAC,MAAM,CAAC,QAAQ,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IACxF,OAAO,CAAC,GAAG,CAAC,gBAAgB,SAAS,+CAA+C,CAAC,CAAC;IACtF,OAAO,CAAC,GAAG,CAAC,gBAAgB,QAAQ,+CAA+C,CAAC,CAAC;IAErF,OAAO,CAAC,GAAG,CAAC,wBAAwB,CAAC,CAAC;IACtC,MAAM,iBAAiB,GAAG,MAAM,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,GAAG,CAAC,CAAC,CAAC;IACtE,MAAM,QAAQ,GAAG,MAAM,CAAC,mBAAmB,CAAC;IAE5C,IAAI,MAAM,CAAC,OAAO,IAAI,CAAC,IAAI,iBAAiB,EAAE,CAAC;QAC7C,OAAO,CAAC,GAAG,CAAC,kCAAkC,CAAC,CAAC;QAChD,OAAO,CAAC,GAAG,CAAC,iBAAiB,MAAM,CAAC,OAAO,WAAW,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,uCAAuC,CAAC,CAAC;IACpH,CAAC;SAAM,IAAI,MAAM,CAAC,OAAO,IAAI,CAAC,IAAI,QAAQ,IAAI,CAAC,EAAE,CAAC;QAChD,OAAO,CAAC,GAAG,CAAC,4DAA4D,CAAC,CAAC;QAC1E,OAAO,CAAC,GAAG,CAAC,iBAAiB,MAAM,CAAC,OAAO,WAAW,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;IACxF,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,GAAG,CAAC,0BAA0B,CAAC,CAAC;QACxC,OAAO,CAAC,GAAG,CAAC,iBAAiB,MAAM,CAAC,OAAO,WAAW,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;QACtF,IAAI,CAAC,iBAAiB,EAAE,CAAC;YACvB,OAAO,CAAC,GAAG,CAAC,iEAAiE,CAAC,CAAC;QACjF,CAAC;IACH,CAAC;IAED,oBAAoB;IACpB,MAAM,OAAO,GAAG;QACd,QAAQ,EAAE,cAAc;QACxB,IAAI,EAAE,EAAE;QACR,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;QACnC,SAAS,EAAE,OAAO,CAAC,GAAG,CAAC,YAAY,CAAC,IAAI,mBAAmB;QAC3D,QAAQ,EAAE,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC,IAAI,gBAAgB;QACtD,OAAO,EAAE;YACP,KAAK,EAAE,MAAM,CAAC,KAAK;YACnB,OAAO,EAAE,MAAM,CAAC,OAAO;YACvB,QAAQ,EAAE,MAAM,CAAC,QAAQ;YACzB,mBAAmB,EAAE,MAAM,CAAC,mBAAmB;YAC/C,YAAY,EAAE,MAAM,CAAC,YAAY;YACjC,gBAAgB,EAAE,MAAM,CAAC,gBAAgB;YACzC,UAAU,EAAE,MAAM,CAAC,UAAU;SAC9B;QACD,QAAQ,EAAE,EAAE,OAAO,EAAE,eAAe,EAAE,KAAK,EAAE,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE;QACpE,WAAW,EAAE,MAAM,CAAC,WAAW;KAChC,CAAC;IAEF,MAAM,OAAO,GAAG,mFAAmF,CAAC;IACpG,MAAM,EAAE,GAAG,MAAM,MAAM,CAAC,SAAS,CAAC,CAAC;IACnC,EAAE,CAAC,aAAa,CAAC,OAAO,EAAE,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC;IAC5D,OAAO,CAAC,GAAG,CAAC,yBAAyB,OAAO,IAAI,CAAC,CAAC;AACpD,CAAC;AAED,IAAI,EAAE,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,EAAE;IACnB,OAAO,CAAC,KAAK,CAAC,eAAe,EAAE,GAAG,CAAC,CAAC;IACpC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC,CAAC,CAAC"}
@@ -0,0 +1,110 @@
1
+ /**
2
+ * GAIA DAG Harness — Co-Sight Architecture Port (ADR-139 Addendum)
3
+ *
4
+ * Ports the ZTE-AICloud/Co-Sight DAG orchestration pattern (Apache 2.0,
5
+ * arXiv 2510.21557) into the ruflo GAIA harness.
6
+ *
7
+ * Architecture:
8
+ * 1. PLAN — Claude Sonnet 4.6 reads the question and emits a DAG of 3-7
9
+ * steps: {id, description, depends_on: [], suggested_tool}.
10
+ * Claude-aware prompt: "3-5 steps max, direct answer when clear."
11
+ * 2. EXECUTE — Loop while ready steps exist (deps satisfied).
12
+ * Run all ready steps in PARALLEL (Promise.all, cap ≤5).
13
+ * Each step = a Gemini 2.5 Pro actor with the full tool suite.
14
+ * Actor marks step completed/blocked + writes step_notes.
15
+ * Blocked steps trigger planner re_plan before next cycle.
16
+ * 3. FINALIZE — Planner reads all step_notes → produces final answer
17
+ * using T2 extraction cascade from gaia-agent.ts.
18
+ * 4. CAMV — Async credibility labeling per step (stubbed, iter 65).
19
+ *
20
+ * Role assignment (env-configurable):
21
+ * PLAN_MODEL = claude-sonnet-4-6 (default)
22
+ * ACT_MODEL = gemini-2.5-pro (default)
23
+ * VISION_MODEL = gemini-2.5-pro (default, same as act)
24
+ *
25
+ * CLI:
26
+ * gaia-bench run --mode=dag --model claude-sonnet-4-6
27
+ *
28
+ * Cost: planner ~$0.02/Q + actors ~$0.03/Q = ~$0.05/Q (vs single-Sonnet ~$0.075)
29
+ *
30
+ * Refs: ADR-139, github.com/ZTE-AICloud/Co-Sight, arXiv 2510.21557, #2156
31
+ */
32
+ import { GaiaQuestion } from './gaia-loader.js';
33
+ import { GaiaToolCatalogue } from './gaia-tools/index.js';
34
+ export interface DagStep {
35
+ id: number;
36
+ description: string;
37
+ depends_on: number[];
38
+ suggested_tool?: string;
39
+ status: 'not_started' | 'in_progress' | 'completed' | 'blocked';
40
+ step_notes: string;
41
+ }
42
+ export interface DagPlan {
43
+ title: string;
44
+ question: string;
45
+ steps: DagStep[];
46
+ }
47
+ /**
48
+ * Get all steps whose dependencies are fully satisfied (completed).
49
+ * Mirrors Co-Sight's Plan.get_ready_steps().
50
+ */
51
+ export declare function getReadySteps(plan: DagPlan): DagStep[];
52
+ export interface DagResult {
53
+ questionId: string;
54
+ finalAnswer: string | null;
55
+ normalisedAnswer: string;
56
+ plan: DagPlan;
57
+ totalSteps: number;
58
+ completedSteps: number;
59
+ blockedSteps: number;
60
+ plannerCycles: number;
61
+ totalInputTokens: number;
62
+ totalOutputTokens: number;
63
+ estimatedCostUsd: number;
64
+ wallMs: number;
65
+ timedOut?: boolean;
66
+ error?: string;
67
+ }
68
+ export interface DagOptions {
69
+ planModel?: string;
70
+ actModel?: string;
71
+ anthropicApiKey?: string;
72
+ geminiApiKey?: string;
73
+ catalogue?: GaiaToolCatalogue;
74
+ /** Disable CAMV stub (default: false — stub is always disabled in this iter). */
75
+ enableCamv?: boolean;
76
+ }
77
+ /**
78
+ * Run a GAIA question through the Co-Sight DAG harness.
79
+ *
80
+ * Steps:
81
+ * 1. Planner (Claude Sonnet) creates a DAG plan.
82
+ * 2. Execute loop: parallel actors (Gemini 2.5 Pro) run ready steps.
83
+ * 3. Blocked steps trigger replan (up to MAX_REPLAN_CYCLES).
84
+ * 4. Finalizer (Claude Sonnet) reads all step notes → final answer.
85
+ */
86
+ export declare function runGaiaDAG(question: GaiaQuestion, options?: DagOptions): Promise<DagResult>;
87
+ export interface DagPilotResult {
88
+ correct: number;
89
+ total: number;
90
+ accuracy: number;
91
+ avgStepsPerQuestion: number;
92
+ perQuestion: Array<{
93
+ taskId: string;
94
+ question: string;
95
+ expected: string;
96
+ got: string | null;
97
+ correct: boolean;
98
+ steps: number;
99
+ completedSteps: number;
100
+ blockedSteps: number;
101
+ plannerCycles: number;
102
+ costUsd: number;
103
+ wallMs: number;
104
+ }>;
105
+ totalCostUsd: number;
106
+ projectedCost53Q: number;
107
+ meanWallMs: number;
108
+ }
109
+ export declare function runDagPilot(questions: GaiaQuestion[], options?: DagOptions): Promise<DagPilotResult>;
110
+ //# sourceMappingURL=gaia-dag.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"gaia-dag.d.ts","sourceRoot":"","sources":["../../../src/benchmarks/gaia-dag.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AAKH,OAAO,EAAE,YAAY,EAAE,MAAM,kBAAkB,CAAC;AAChD,OAAO,EAEL,iBAAiB,EAKlB,MAAM,uBAAuB,CAAC;AAiD/B,MAAM,WAAW,OAAO;IACtB,EAAE,EAAE,MAAM,CAAC;IACX,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,MAAM,EAAE,CAAC;IACrB,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,MAAM,EAAE,aAAa,GAAG,aAAa,GAAG,WAAW,GAAG,SAAS,CAAC;IAChE,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,OAAO;IACtB,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,OAAO,EAAE,CAAC;CAClB;AAED;;;GAGG;AACH,wBAAgB,aAAa,CAAC,IAAI,EAAE,OAAO,GAAG,OAAO,EAAE,CAQtD;AAygBD,MAAM,WAAW,SAAS;IACxB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,GAAG,IAAI,CAAC;IAC3B,gBAAgB,EAAE,MAAM,CAAC;IACzB,IAAI,EAAE,OAAO,CAAC;IACd,UAAU,EAAE,MAAM,CAAC;IACnB,cAAc,EAAE,MAAM,CAAC;IACvB,YAAY,EAAE,MAAM,CAAC;IACrB,aAAa,EAAE,MAAM,CAAC;IACtB,gBAAgB,EAAE,MAAM,CAAC;IACzB,iBAAiB,EAAE,MAAM,CAAC;IAC1B,gBAAgB,EAAE,MAAM,CAAC;IACzB,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,UAAU;IACzB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,SAAS,CAAC,EAAE,iBAAiB,CAAC;IAC9B,iFAAiF;IACjF,UAAU,CAAC,EAAE,OAAO,CAAC;CACtB;AA6BD;;;;;;;;GAQG;AACH,wBAAsB,UAAU,CAC9B,QAAQ,EAAE,YAAY,EACtB,OAAO,GAAE,UAAe,GACvB,OAAO,CAAC,SAAS,CAAC,CA6IpB;AAMD,MAAM,WAAW,cAAc;IAC7B,OAAO,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,EAAE,MAAM,CAAC;IACjB,mBAAmB,EAAE,MAAM,CAAC;IAC5B,WAAW,EAAE,KAAK,CAAC;QACjB,MAAM,EAAE,MAAM,CAAC;QACf,QAAQ,EAAE,MAAM,CAAC;QACjB,QAAQ,EAAE,MAAM,CAAC;QACjB,GAAG,EAAE,MAAM,GAAG,IAAI,CAAC;QACnB,OAAO,EAAE,OAAO,CAAC;QACjB,KAAK,EAAE,MAAM,CAAC;QACd,cAAc,EAAE,MAAM,CAAC;QACvB,YAAY,EAAE,MAAM,CAAC;QACrB,aAAa,EAAE,MAAM,CAAC;QACtB,OAAO,EAAE,MAAM,CAAC;QAChB,MAAM,EAAE,MAAM,CAAC;KAChB,CAAC,CAAC;IACH,YAAY,EAAE,MAAM,CAAC;IACrB,gBAAgB,EAAE,MAAM,CAAC;IACzB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,wBAAsB,WAAW,CAC/B,SAAS,EAAE,YAAY,EAAE,EACzB,OAAO,GAAE,UAAe,GACvB,OAAO,CAAC,cAAc,CAAC,CA6CzB"}