@claude-flow/cli 3.10.36 → 3.10.37
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.
- package/.claude/agents/analysis/analyze-code-quality.md +178 -178
- package/.claude/agents/analysis/code-analyzer.md +209 -209
- package/.claude/agents/analysis/code-review/analyze-code-quality.md +178 -178
- package/.claude/agents/architecture/arch-system-design.md +156 -156
- package/.claude/agents/architecture/system-design/arch-system-design.md +154 -154
- package/.claude/agents/browser/browser-agent.yaml +182 -182
- package/.claude/agents/consensus/byzantine-coordinator.md +62 -62
- package/.claude/agents/consensus/crdt-synchronizer.md +996 -996
- package/.claude/agents/consensus/gossip-coordinator.md +62 -62
- package/.claude/agents/consensus/performance-benchmarker.md +850 -850
- package/.claude/agents/consensus/quorum-manager.md +822 -822
- package/.claude/agents/consensus/raft-manager.md +62 -62
- package/.claude/agents/consensus/security-manager.md +621 -621
- package/.claude/agents/core/planner.md +374 -374
- package/.claude/agents/custom/test-long-runner.md +44 -44
- package/.claude/agents/data/data-ml-model.md +444 -444
- package/.claude/agents/data/ml/data-ml-model.md +192 -192
- package/.claude/agents/development/backend/dev-backend-api.md +141 -141
- package/.claude/agents/development/dev-backend-api.md +344 -344
- package/.claude/agents/devops/ci-cd/ops-cicd-github.md +163 -163
- package/.claude/agents/devops/ops-cicd-github.md +164 -164
- package/.claude/agents/documentation/api-docs/docs-api-openapi.md +173 -173
- package/.claude/agents/documentation/docs-api-openapi.md +354 -354
- package/.claude/agents/flow-nexus/app-store.md +87 -87
- package/.claude/agents/flow-nexus/authentication.md +68 -68
- package/.claude/agents/flow-nexus/challenges.md +80 -80
- package/.claude/agents/flow-nexus/neural-network.md +87 -87
- package/.claude/agents/flow-nexus/payments.md +82 -82
- package/.claude/agents/flow-nexus/sandbox.md +75 -75
- package/.claude/agents/flow-nexus/swarm.md +75 -75
- package/.claude/agents/flow-nexus/user-tools.md +95 -95
- package/.claude/agents/flow-nexus/workflow.md +83 -83
- package/.claude/agents/github/code-review-swarm.md +377 -377
- package/.claude/agents/github/github-modes.md +172 -172
- package/.claude/agents/github/issue-tracker.md +575 -575
- package/.claude/agents/github/multi-repo-swarm.md +552 -552
- package/.claude/agents/github/pr-manager.md +437 -437
- package/.claude/agents/github/project-board-sync.md +508 -508
- package/.claude/agents/github/release-manager.md +604 -604
- package/.claude/agents/github/release-swarm.md +582 -582
- package/.claude/agents/github/repo-architect.md +397 -397
- package/.claude/agents/github/swarm-issue.md +572 -572
- package/.claude/agents/github/swarm-pr.md +427 -427
- package/.claude/agents/github/sync-coordinator.md +451 -451
- package/.claude/agents/github/workflow-automation.md +902 -902
- package/.claude/agents/goal/agent.md +815 -815
- package/.claude/agents/optimization/benchmark-suite.md +664 -664
- package/.claude/agents/optimization/load-balancer.md +430 -430
- package/.claude/agents/optimization/performance-monitor.md +671 -671
- package/.claude/agents/optimization/resource-allocator.md +673 -673
- package/.claude/agents/optimization/topology-optimizer.md +807 -807
- package/.claude/agents/payments/agentic-payments.md +126 -126
- package/.claude/agents/sona/sona-learning-optimizer.md +74 -74
- package/.claude/agents/sparc/architecture.md +698 -698
- package/.claude/agents/sparc/pseudocode.md +519 -519
- package/.claude/agents/sparc/refinement.md +801 -801
- package/.claude/agents/sparc/specification.md +477 -477
- package/.claude/agents/specialized/mobile/spec-mobile-react-native.md +224 -224
- package/.claude/agents/specialized/spec-mobile-react-native.md +226 -226
- package/.claude/agents/sublinear/consensus-coordinator.md +337 -337
- package/.claude/agents/sublinear/matrix-optimizer.md +184 -184
- package/.claude/agents/sublinear/pagerank-analyzer.md +298 -298
- package/.claude/agents/sublinear/performance-optimizer.md +367 -367
- package/.claude/agents/sublinear/trading-predictor.md +245 -245
- package/.claude/agents/swarm/adaptive-coordinator.md +1126 -1126
- package/.claude/agents/swarm/hierarchical-coordinator.md +709 -709
- package/.claude/agents/swarm/mesh-coordinator.md +962 -962
- package/.claude/agents/templates/automation-smart-agent.md +204 -204
- package/.claude/agents/templates/base-template-generator.md +289 -289
- package/.claude/agents/templates/coordinator-swarm-init.md +89 -89
- package/.claude/agents/templates/github-pr-manager.md +176 -176
- package/.claude/agents/templates/implementer-sparc-coder.md +258 -258
- package/.claude/agents/templates/memory-coordinator.md +186 -186
- package/.claude/agents/templates/orchestrator-task.md +138 -138
- package/.claude/agents/templates/performance-analyzer.md +198 -198
- package/.claude/agents/templates/sparc-coordinator.md +513 -513
- package/.claude/agents/testing/production-validator.md +394 -394
- package/.claude/agents/testing/tdd-london-swarm.md +243 -243
- package/.claude/agents/v3/aidefence-guardian.md +282 -282
- package/.claude/agents/v3/claims-authorizer.md +208 -208
- package/.claude/agents/v3/collective-intelligence-coordinator.md +993 -993
- package/.claude/agents/v3/ddd-domain-expert.md +220 -220
- package/.claude/agents/v3/injection-analyst.md +236 -236
- package/.claude/agents/v3/performance-engineer.md +1233 -1233
- package/.claude/agents/v3/pii-detector.md +151 -151
- package/.claude/agents/v3/reasoningbank-learner.md +213 -213
- package/.claude/agents/v3/security-architect-aidefence.md +410 -410
- package/.claude/agents/v3/security-architect.md +867 -867
- package/.claude/agents/v3/swarm-memory-manager.md +157 -157
- package/.claude/agents/v3/v3-integration-architect.md +205 -205
- package/.claude/commands/agents/README.md +50 -50
- package/.claude/commands/agents/agent-capabilities.md +140 -140
- package/.claude/commands/agents/agent-coordination.md +28 -28
- package/.claude/commands/agents/agent-spawning.md +28 -28
- package/.claude/commands/agents/agent-types.md +216 -216
- package/.claude/commands/agents/health.md +139 -139
- package/.claude/commands/agents/list.md +100 -100
- package/.claude/commands/agents/logs.md +130 -130
- package/.claude/commands/agents/metrics.md +122 -122
- package/.claude/commands/agents/pool.md +127 -127
- package/.claude/commands/agents/spawn.md +140 -140
- package/.claude/commands/agents/status.md +115 -115
- package/.claude/commands/agents/stop.md +102 -102
- package/.claude/commands/analysis/COMMAND_COMPLIANCE_REPORT.md +53 -53
- package/.claude/commands/analysis/README.md +9 -9
- package/.claude/commands/analysis/bottleneck-detect.md +162 -162
- package/.claude/commands/analysis/performance-bottlenecks.md +58 -58
- package/.claude/commands/analysis/performance-report.md +25 -25
- package/.claude/commands/analysis/token-efficiency.md +44 -44
- package/.claude/commands/analysis/token-usage.md +25 -25
- package/.claude/commands/automation/README.md +9 -9
- package/.claude/commands/automation/auto-agent.md +122 -122
- package/.claude/commands/automation/self-healing.md +105 -105
- package/.claude/commands/automation/session-memory.md +89 -89
- package/.claude/commands/automation/smart-agents.md +72 -72
- package/.claude/commands/automation/smart-spawn.md +25 -25
- package/.claude/commands/automation/workflow-select.md +25 -25
- package/.claude/commands/claude-flow-help.md +103 -103
- package/.claude/commands/claude-flow-memory.md +107 -107
- package/.claude/commands/claude-flow-swarm.md +205 -205
- package/.claude/commands/coordination/README.md +9 -9
- package/.claude/commands/coordination/agent-spawn.md +25 -25
- package/.claude/commands/coordination/init.md +44 -44
- package/.claude/commands/coordination/orchestrate.md +43 -43
- package/.claude/commands/coordination/spawn.md +45 -45
- package/.claude/commands/coordination/swarm-init.md +85 -85
- package/.claude/commands/coordination/task-orchestrate.md +25 -25
- package/.claude/commands/github/README.md +11 -11
- package/.claude/commands/github/code-review-swarm.md +513 -513
- package/.claude/commands/github/code-review.md +25 -25
- package/.claude/commands/github/github-modes.md +146 -146
- package/.claude/commands/github/github-swarm.md +121 -121
- package/.claude/commands/github/issue-tracker.md +291 -291
- package/.claude/commands/github/issue-triage.md +25 -25
- package/.claude/commands/github/multi-repo-swarm.md +518 -518
- package/.claude/commands/github/pr-enhance.md +26 -26
- package/.claude/commands/github/pr-manager.md +169 -169
- package/.claude/commands/github/project-board-sync.md +470 -470
- package/.claude/commands/github/release-manager.md +339 -339
- package/.claude/commands/github/release-swarm.md +543 -543
- package/.claude/commands/github/repo-analyze.md +25 -25
- package/.claude/commands/github/repo-architect.md +366 -366
- package/.claude/commands/github/swarm-issue.md +484 -484
- package/.claude/commands/github/swarm-pr.md +287 -287
- package/.claude/commands/github/sync-coordinator.md +302 -302
- package/.claude/commands/github/workflow-automation.md +441 -441
- package/.claude/commands/hive-mind/README.md +17 -17
- package/.claude/commands/hive-mind/hive-mind-consensus.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-init.md +18 -18
- package/.claude/commands/hive-mind/hive-mind-memory.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-metrics.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-resume.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-sessions.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-spawn.md +21 -21
- package/.claude/commands/hive-mind/hive-mind-status.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-stop.md +8 -8
- package/.claude/commands/hive-mind/hive-mind-wizard.md +8 -8
- package/.claude/commands/hive-mind/hive-mind.md +27 -27
- package/.claude/commands/hooks/README.md +11 -11
- package/.claude/commands/hooks/overview.md +57 -57
- package/.claude/commands/hooks/post-edit.md +117 -117
- package/.claude/commands/hooks/post-task.md +112 -112
- package/.claude/commands/hooks/pre-edit.md +113 -113
- package/.claude/commands/hooks/pre-task.md +111 -111
- package/.claude/commands/hooks/session-end.md +118 -118
- package/.claude/commands/hooks/setup.md +102 -102
- package/.claude/commands/memory/README.md +9 -9
- package/.claude/commands/memory/memory-persist.md +25 -25
- package/.claude/commands/memory/memory-search.md +25 -25
- package/.claude/commands/memory/memory-usage.md +25 -25
- package/.claude/commands/memory/neural.md +47 -47
- package/.claude/commands/monitoring/README.md +9 -9
- package/.claude/commands/monitoring/agent-metrics.md +25 -25
- package/.claude/commands/monitoring/agents.md +44 -44
- package/.claude/commands/monitoring/real-time-view.md +25 -25
- package/.claude/commands/monitoring/status.md +46 -46
- package/.claude/commands/monitoring/swarm-monitor.md +25 -25
- package/.claude/commands/optimization/README.md +9 -9
- package/.claude/commands/optimization/auto-topology.md +61 -61
- package/.claude/commands/optimization/cache-manage.md +25 -25
- package/.claude/commands/optimization/parallel-execute.md +25 -25
- package/.claude/commands/optimization/parallel-execution.md +49 -49
- package/.claude/commands/optimization/topology-optimize.md +25 -25
- package/.claude/commands/pair/README.md +260 -260
- package/.claude/commands/pair/commands.md +545 -545
- package/.claude/commands/pair/config.md +509 -509
- package/.claude/commands/pair/examples.md +511 -511
- package/.claude/commands/pair/modes.md +347 -347
- package/.claude/commands/pair/session.md +406 -406
- package/.claude/commands/pair/start.md +208 -208
- package/.claude/commands/sparc/analyzer.md +51 -51
- package/.claude/commands/sparc/architect.md +53 -53
- package/.claude/commands/sparc/ask.md +97 -97
- package/.claude/commands/sparc/batch-executor.md +54 -54
- package/.claude/commands/sparc/code.md +89 -89
- package/.claude/commands/sparc/coder.md +54 -54
- package/.claude/commands/sparc/debug.md +83 -83
- package/.claude/commands/sparc/debugger.md +54 -54
- package/.claude/commands/sparc/designer.md +53 -53
- package/.claude/commands/sparc/devops.md +109 -109
- package/.claude/commands/sparc/docs-writer.md +80 -80
- package/.claude/commands/sparc/documenter.md +54 -54
- package/.claude/commands/sparc/innovator.md +54 -54
- package/.claude/commands/sparc/integration.md +83 -83
- package/.claude/commands/sparc/mcp.md +117 -117
- package/.claude/commands/sparc/memory-manager.md +54 -54
- package/.claude/commands/sparc/optimizer.md +54 -54
- package/.claude/commands/sparc/orchestrator.md +131 -131
- package/.claude/commands/sparc/post-deployment-monitoring-mode.md +83 -83
- package/.claude/commands/sparc/refinement-optimization-mode.md +83 -83
- package/.claude/commands/sparc/researcher.md +54 -54
- package/.claude/commands/sparc/reviewer.md +54 -54
- package/.claude/commands/sparc/security-review.md +80 -80
- package/.claude/commands/sparc/sparc-modes.md +174 -174
- package/.claude/commands/sparc/sparc.md +111 -111
- package/.claude/commands/sparc/spec-pseudocode.md +80 -80
- package/.claude/commands/sparc/supabase-admin.md +348 -348
- package/.claude/commands/sparc/swarm-coordinator.md +54 -54
- package/.claude/commands/sparc/tdd.md +54 -54
- package/.claude/commands/sparc/tester.md +54 -54
- package/.claude/commands/sparc/tutorial.md +79 -79
- package/.claude/commands/sparc/workflow-manager.md +54 -54
- package/.claude/commands/sparc.md +166 -166
- package/.claude/commands/stream-chain/pipeline.md +120 -120
- package/.claude/commands/stream-chain/run.md +69 -69
- package/.claude/commands/swarm/README.md +15 -15
- package/.claude/commands/swarm/analysis.md +95 -95
- package/.claude/commands/swarm/development.md +96 -96
- package/.claude/commands/swarm/examples.md +168 -168
- package/.claude/commands/swarm/maintenance.md +102 -102
- package/.claude/commands/swarm/optimization.md +117 -117
- package/.claude/commands/swarm/research.md +136 -136
- package/.claude/commands/swarm/swarm-analysis.md +8 -8
- package/.claude/commands/swarm/swarm-background.md +8 -8
- package/.claude/commands/swarm/swarm-init.md +19 -19
- package/.claude/commands/swarm/swarm-modes.md +8 -8
- package/.claude/commands/swarm/swarm-monitor.md +8 -8
- package/.claude/commands/swarm/swarm-spawn.md +19 -19
- package/.claude/commands/swarm/swarm-status.md +8 -8
- package/.claude/commands/swarm/swarm-strategies.md +8 -8
- package/.claude/commands/swarm/swarm.md +87 -87
- package/.claude/commands/swarm/testing.md +131 -131
- package/.claude/commands/training/README.md +9 -9
- package/.claude/commands/training/model-update.md +25 -25
- package/.claude/commands/training/neural-patterns.md +107 -107
- package/.claude/commands/training/neural-train.md +75 -75
- package/.claude/commands/training/pattern-learn.md +25 -25
- package/.claude/commands/training/specialization.md +62 -62
- package/.claude/commands/truth/start.md +142 -142
- package/.claude/commands/verify/check.md +49 -49
- package/.claude/commands/verify/start.md +127 -127
- package/.claude/commands/workflows/README.md +9 -9
- package/.claude/commands/workflows/development.md +77 -77
- package/.claude/commands/workflows/research.md +62 -62
- package/.claude/commands/workflows/workflow-create.md +25 -25
- package/.claude/commands/workflows/workflow-execute.md +25 -25
- package/.claude/commands/workflows/workflow-export.md +25 -25
- package/.claude/helpers/README.md +96 -96
- package/.claude/helpers/adr-compliance.sh +186 -186
- package/.claude/helpers/auto-commit.sh +178 -178
- package/.claude/helpers/auto-memory-hook.mjs +368 -368
- package/.claude/helpers/checkpoint-manager.sh +251 -251
- package/.claude/helpers/daemon-manager.sh +252 -252
- package/.claude/helpers/ddd-tracker.sh +144 -144
- package/.claude/helpers/github-safe.js +156 -156
- package/.claude/helpers/github-setup.sh +45 -45
- package/.claude/helpers/guidance-hook.sh +13 -13
- package/.claude/helpers/guidance-hooks.sh +102 -102
- package/.claude/helpers/health-monitor.sh +108 -108
- package/.claude/helpers/hook-handler.cjs +286 -286
- package/.claude/helpers/intelligence.cjs +1031 -1031
- package/.claude/helpers/learning-hooks.sh +329 -329
- package/.claude/helpers/learning-optimizer.sh +127 -127
- package/.claude/helpers/learning-service.mjs +1144 -1144
- package/.claude/helpers/memory.js +83 -83
- package/.claude/helpers/metrics-db.mjs +488 -488
- package/.claude/helpers/pattern-consolidator.sh +86 -86
- package/.claude/helpers/perf-worker.sh +160 -160
- package/.claude/helpers/post-commit +16 -16
- package/.claude/helpers/pre-commit +26 -26
- package/.claude/helpers/quick-start.sh +19 -19
- package/.claude/helpers/router.js +105 -105
- package/.claude/helpers/security-scanner.sh +127 -127
- package/.claude/helpers/session.js +135 -135
- package/.claude/helpers/setup-mcp.sh +18 -18
- package/.claude/helpers/standard-checkpoint-hooks.sh +189 -189
- package/.claude/helpers/statusline-hook.sh +21 -21
- package/.claude/helpers/statusline.cjs +604 -604
- package/.claude/helpers/statusline.js +352 -352
- package/.claude/helpers/swarm-comms.sh +353 -353
- package/.claude/helpers/swarm-hooks.sh +761 -761
- package/.claude/helpers/swarm-monitor.sh +210 -210
- package/.claude/helpers/sync-v3-metrics.sh +245 -245
- package/.claude/helpers/update-v3-progress.sh +165 -165
- package/.claude/helpers/v3-quick-status.sh +57 -57
- package/.claude/helpers/v3.sh +110 -110
- package/.claude/helpers/validate-v3-config.sh +215 -215
- package/.claude/helpers/worker-manager.sh +170 -170
- package/.claude/settings.json +182 -182
- package/.claude/skills/agentdb-advanced/SKILL.md +550 -550
- package/.claude/skills/agentdb-learning/SKILL.md +545 -545
- package/.claude/skills/agentdb-memory-patterns/SKILL.md +339 -339
- package/.claude/skills/agentdb-optimization/SKILL.md +509 -509
- package/.claude/skills/agentdb-vector-search/SKILL.md +339 -339
- package/.claude/skills/browser/SKILL.md +204 -204
- package/.claude/skills/dual-mode/README.md +71 -71
- package/.claude/skills/dual-mode/dual-collect.md +103 -103
- package/.claude/skills/dual-mode/dual-coordinate.md +85 -85
- package/.claude/skills/dual-mode/dual-spawn.md +81 -81
- package/.claude/skills/flow-nexus-neural/SKILL.md +727 -727
- package/.claude/skills/flow-nexus-platform/SKILL.md +1154 -1154
- package/.claude/skills/flow-nexus-swarm/SKILL.md +604 -604
- package/.claude/skills/github-code-review/SKILL.md +1125 -1125
- package/.claude/skills/github-multi-repo/SKILL.md +862 -862
- package/.claude/skills/github-project-management/SKILL.md +1262 -1262
- package/.claude/skills/github-release-management/SKILL.md +1064 -1064
- package/.claude/skills/github-workflow-automation/SKILL.md +1047 -1047
- package/.claude/skills/hooks-automation/SKILL.md +1201 -1201
- package/.claude/skills/pair-programming/SKILL.md +1202 -1202
- package/.claude/skills/reasoningbank-agentdb/SKILL.md +446 -446
- package/.claude/skills/reasoningbank-intelligence/SKILL.md +201 -201
- package/.claude/skills/skill-builder/SKILL.md +910 -910
- package/.claude/skills/sparc-methodology/SKILL.md +1106 -1106
- package/.claude/skills/stream-chain/SKILL.md +560 -560
- package/.claude/skills/swarm-advanced/SKILL.md +970 -970
- package/.claude/skills/swarm-orchestration/SKILL.md +179 -179
- package/.claude/skills/v3-cli-modernization/SKILL.md +871 -871
- package/.claude/skills/v3-core-implementation/SKILL.md +796 -796
- package/.claude/skills/v3-ddd-architecture/SKILL.md +441 -441
- package/.claude/skills/v3-integration-deep/SKILL.md +240 -240
- package/.claude/skills/v3-mcp-optimization/SKILL.md +776 -776
- package/.claude/skills/v3-memory-unification/SKILL.md +173 -173
- package/.claude/skills/v3-performance-optimization/SKILL.md +389 -389
- package/.claude/skills/v3-security-overhaul/SKILL.md +81 -81
- package/.claude/skills/v3-swarm-coordination/SKILL.md +339 -339
- package/.claude/skills/verification-quality/SKILL.md +691 -691
- package/README.md +416 -416
- package/bin/cli.js +314 -314
- package/bin/mcp-server.js +224 -224
- package/bin/preinstall.cjs +2 -2
- package/dist/src/benchmarks/gaia-critic.js +24 -24
- package/dist/src/commands/completions.js +409 -409
- package/dist/src/commands/daemon.js +44 -44
- package/dist/src/commands/embeddings.js +29 -29
- package/dist/src/commands/embeddings.js.map +1 -1
- package/dist/src/commands/hive-mind.js +97 -97
- package/dist/src/commands/hooks.js +9 -9
- package/dist/src/commands/ruvector/backup.js +23 -23
- package/dist/src/commands/ruvector/benchmark.js +31 -31
- package/dist/src/commands/ruvector/import.js +14 -14
- package/dist/src/commands/ruvector/init.js +115 -115
- package/dist/src/commands/ruvector/migrate.js +99 -99
- package/dist/src/commands/ruvector/optimize.js +51 -51
- package/dist/src/commands/ruvector/setup.js +624 -624
- package/dist/src/commands/ruvector/status.js +38 -38
- package/dist/src/init/claudemd-generator.js +227 -227
- package/dist/src/init/executor.js +453 -453
- package/dist/src/init/helpers-generator.js +727 -727
- package/dist/src/init/statusline-generator.js +570 -570
- package/dist/src/mcp-tools/agentdb-tools.js +15 -15
- package/dist/src/memory/graph-edge-writer.js +18 -18
- package/dist/src/memory/memory-bridge.js +84 -84
- package/dist/src/memory/memory-initializer.js +393 -393
- package/dist/src/memory/rabitq-index.js +5 -5
- package/dist/src/runtime/headless.js +28 -28
- package/dist/src/ruvector/flash-attention.d.ts +195 -0
- package/dist/src/ruvector/flash-attention.d.ts.map +1 -0
- package/dist/src/ruvector/flash-attention.js +643 -0
- package/dist/src/ruvector/flash-attention.js.map +1 -0
- package/dist/src/ruvector/moe-router.d.ts +206 -0
- package/dist/src/ruvector/moe-router.d.ts.map +1 -0
- package/dist/src/ruvector/moe-router.js +626 -0
- package/dist/src/ruvector/moe-router.js.map +1 -0
- package/dist/src/services/event-stream.d.ts +25 -0
- package/dist/src/services/event-stream.js +27 -0
- package/dist/src/services/headless-worker-executor.js +84 -84
- package/dist/src/services/loop-worker-runner.d.ts +16 -0
- package/dist/src/services/loop-worker-runner.js +34 -0
- package/dist/src/services/runtime-capabilities.d.ts +22 -0
- package/dist/src/services/runtime-capabilities.js +45 -0
- package/dist/src/transfer/deploy-seraphine.js +23 -23
- package/dist/tsconfig.tsbuildinfo +1 -1
- package/package.json +134 -134
- package/scripts/beir-bootstrap-significance.mjs +162 -162
- package/scripts/bench-rvagent.mjs +312 -312
- package/scripts/benchmark-codemods.mjs +127 -127
- package/scripts/benchmark-cross-repo.mjs +206 -206
- package/scripts/benchmark-pretrained-retrieval.mjs +305 -305
- package/scripts/benchmark-self-learning.mjs +241 -241
- package/scripts/benchmark-trajectory-mrr.mjs +191 -191
- package/scripts/deploy-ipfs-node.sh +153 -153
- package/scripts/grid-search-retrieval.mjs +216 -216
- package/scripts/postinstall.cjs +153 -153
- package/scripts/pretrain-from-github.mjs +358 -358
- package/scripts/publish-registry.ts +345 -345
- package/scripts/publish.sh +57 -57
- package/scripts/run-beir-bge.mjs +324 -336
- package/scripts/run-beir-hybrid.mjs +261 -263
- package/scripts/run-beir-lucene-bm25.mjs +154 -154
- package/scripts/run-beir-nfcorpus.mjs +231 -231
- package/scripts/run-beir-rrf-ablation.mjs +221 -221
- package/scripts/setup-ipfs-registry.md +366 -366
- package/dist/src/benchmarks/capability-tasks.d.ts +0 -126
- package/dist/src/benchmarks/capability-tasks.d.ts.map +0 -1
- package/dist/src/benchmarks/capability-tasks.js +0 -147
- package/dist/src/benchmarks/capability-tasks.js.map +0 -1
- package/dist/src/benchmarks/gaia-agent-contrastive.d.ts +0 -53
- package/dist/src/benchmarks/gaia-agent-contrastive.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-agent-contrastive.js +0 -177
- package/dist/src/benchmarks/gaia-agent-contrastive.js.map +0 -1
- package/dist/src/benchmarks/gaia-agent-gemini.d.ts +0 -51
- package/dist/src/benchmarks/gaia-agent-gemini.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-agent-gemini.js +0 -352
- package/dist/src/benchmarks/gaia-agent-gemini.js.map +0 -1
- package/dist/src/benchmarks/gaia-claude-p.d.ts +0 -105
- package/dist/src/benchmarks/gaia-claude-p.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-claude-p.js +0 -269
- package/dist/src/benchmarks/gaia-claude-p.js.map +0 -1
- package/dist/src/benchmarks/gaia-claude-p.smoke.d.ts +0 -20
- package/dist/src/benchmarks/gaia-claude-p.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-claude-p.smoke.js +0 -168
- package/dist/src/benchmarks/gaia-claude-p.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-codeagent-runner.py +0 -556
- package/dist/src/benchmarks/gaia-codeagent.d.ts +0 -129
- package/dist/src/benchmarks/gaia-codeagent.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-codeagent.js +0 -590
- package/dist/src/benchmarks/gaia-codeagent.js.map +0 -1
- package/dist/src/benchmarks/gaia-codeagent.smoke.d.ts +0 -33
- package/dist/src/benchmarks/gaia-codeagent.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-codeagent.smoke.js +0 -241
- package/dist/src/benchmarks/gaia-codeagent.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-dag-pilot.d.ts +0 -22
- package/dist/src/benchmarks/gaia-dag-pilot.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-dag-pilot.js +0 -158
- package/dist/src/benchmarks/gaia-dag-pilot.js.map +0 -1
- package/dist/src/benchmarks/gaia-dag.d.ts +0 -110
- package/dist/src/benchmarks/gaia-dag.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-dag.js +0 -658
- package/dist/src/benchmarks/gaia-dag.js.map +0 -1
- package/dist/src/benchmarks/gaia-dag.smoke.d.ts +0 -21
- package/dist/src/benchmarks/gaia-dag.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-dag.smoke.js +0 -290
- package/dist/src/benchmarks/gaia-dag.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-ensemble-pilot.d.ts +0 -17
- package/dist/src/benchmarks/gaia-ensemble-pilot.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-ensemble-pilot.js +0 -106
- package/dist/src/benchmarks/gaia-ensemble-pilot.js.map +0 -1
- package/dist/src/benchmarks/gaia-ensemble.d.ts +0 -103
- package/dist/src/benchmarks/gaia-ensemble.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-ensemble.js +0 -462
- package/dist/src/benchmarks/gaia-ensemble.js.map +0 -1
- package/dist/src/benchmarks/gaia-gemini-pilot.d.ts +0 -17
- package/dist/src/benchmarks/gaia-gemini-pilot.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-gemini-pilot.js +0 -168
- package/dist/src/benchmarks/gaia-gemini-pilot.js.map +0 -1
- package/dist/src/benchmarks/gaia-mode-router.d.ts +0 -67
- package/dist/src/benchmarks/gaia-mode-router.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-mode-router.js +0 -126
- package/dist/src/benchmarks/gaia-mode-router.js.map +0 -1
- package/dist/src/benchmarks/gaia-mode-router.smoke.d.ts +0 -15
- package/dist/src/benchmarks/gaia-mode-router.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-mode-router.smoke.js +0 -131
- package/dist/src/benchmarks/gaia-mode-router.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.d.ts +0 -24
- package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.js +0 -276
- package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/image_describe.d.ts +0 -45
- package/dist/src/benchmarks/gaia-tools/image_describe.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/image_describe.js +0 -222
- package/dist/src/benchmarks/gaia-tools/image_describe.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/image_describe.smoke.d.ts +0 -24
- package/dist/src/benchmarks/gaia-tools/image_describe.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/image_describe.smoke.js +0 -169
- package/dist/src/benchmarks/gaia-tools/image_describe.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/pdf_read.d.ts +0 -27
- package/dist/src/benchmarks/gaia-tools/pdf_read.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/pdf_read.js +0 -226
- package/dist/src/benchmarks/gaia-tools/pdf_read.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/python_exec.d.ts +0 -31
- package/dist/src/benchmarks/gaia-tools/python_exec.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/python_exec.js +0 -94
- package/dist/src/benchmarks/gaia-tools/python_exec.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/python_exec.smoke.d.ts +0 -13
- package/dist/src/benchmarks/gaia-tools/python_exec.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/python_exec.smoke.js +0 -98
- package/dist/src/benchmarks/gaia-tools/python_exec.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/visit_webpage.d.ts +0 -52
- package/dist/src/benchmarks/gaia-tools/visit_webpage.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/visit_webpage.js +0 -304
- package/dist/src/benchmarks/gaia-tools/visit_webpage.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/web_browse.d.ts +0 -61
- package/dist/src/benchmarks/gaia-tools/web_browse.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/web_browse.js +0 -217
- package/dist/src/benchmarks/gaia-tools/web_browse.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/web_browse.smoke.d.ts +0 -26
- package/dist/src/benchmarks/gaia-tools/web_browse.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/web_browse.smoke.js +0 -152
- package/dist/src/benchmarks/gaia-tools/web_browse.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-tools/web_search.smoke.d.ts +0 -24
- package/dist/src/benchmarks/gaia-tools/web_search.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-tools/web_search.smoke.js +0 -311
- package/dist/src/benchmarks/gaia-tools/web_search.smoke.js.map +0 -1
- package/dist/src/benchmarks/gaia-visit-webpage.smoke.d.ts +0 -25
- package/dist/src/benchmarks/gaia-visit-webpage.smoke.d.ts.map +0 -1
- package/dist/src/benchmarks/gaia-visit-webpage.smoke.js +0 -304
- package/dist/src/benchmarks/gaia-visit-webpage.smoke.js.map +0 -1
- package/dist/src/commands/performance-capability.d.ts +0 -30
- package/dist/src/commands/performance-capability.d.ts.map +0 -1
- package/dist/src/commands/performance-capability.js +0 -349
- package/dist/src/commands/performance-capability.js.map +0 -1
|
@@ -1,269 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* GAIA Claude-p Wrapper — iter 54 (#2156)
|
|
3
|
-
*
|
|
4
|
-
* Delegates each GAIA question to `claude -p` (Claude Code headless mode),
|
|
5
|
-
* which gives us WebSearch, WebFetch, Read (multimodal incl. PDF/DOCX/images),
|
|
6
|
-
* and Bash (Python execution) for free — the same tools HAL uses.
|
|
7
|
-
*
|
|
8
|
-
* Why this approach over a native TS CodeAgent:
|
|
9
|
-
* - HAL gaps vs ruflo were: visit_webpage, file reading (PDF/DOCX/XLSX/images),
|
|
10
|
-
* Python execution. Claude Code's built-in tools solve ALL of these.
|
|
11
|
-
* - No wheel-reinvention: battle-tested tool infra, native multimodal, proper
|
|
12
|
-
* tool-budget management, Anthropic WebSearch API.
|
|
13
|
-
* - Baseline: 24/53 (45.3%). Target: ≥45/53 to surpass HAL's 82.07%.
|
|
14
|
-
*
|
|
15
|
-
* SECURITY NOTE on --dangerously-skip-permissions:
|
|
16
|
-
* This flag is ONLY used inside the GAIA harness context, which is a sandboxed
|
|
17
|
-
* benchmark evaluation environment. GAIA questions have no real-world
|
|
18
|
-
* side effects — they are read-only research questions. The flag lets Claude Code
|
|
19
|
-
* use its tools (WebSearch, WebFetch, Read, Bash) without per-tool permission
|
|
20
|
-
* prompts, which is required for unattended benchmark execution. It MUST NOT
|
|
21
|
-
* be used in production workflows where Claude Code could affect real systems.
|
|
22
|
-
*
|
|
23
|
-
* JSON output format from `claude -p --output-format json`:
|
|
24
|
-
* {
|
|
25
|
-
* type: "result",
|
|
26
|
-
* subtype: "success" | "error_max_budget_usd" | ...,
|
|
27
|
-
* is_error: boolean,
|
|
28
|
-
* result: string, // final assistant message text
|
|
29
|
-
* total_cost_usd: number,
|
|
30
|
-
* duration_ms: number,
|
|
31
|
-
* num_turns: number,
|
|
32
|
-
* ...
|
|
33
|
-
* }
|
|
34
|
-
*
|
|
35
|
-
* Refs: ADR-138 (reference, NOT implemented), iter 54, #2156
|
|
36
|
-
*/
|
|
37
|
-
import { spawn } from 'node:child_process';
|
|
38
|
-
import * as path from 'node:path';
|
|
39
|
-
// ---------------------------------------------------------------------------
|
|
40
|
-
// Constants
|
|
41
|
-
// ---------------------------------------------------------------------------
|
|
42
|
-
/** Default model for claude -p GAIA runs. Sonnet for quality parity with HAL. */
|
|
43
|
-
export const CLAUDE_P_DEFAULT_MODEL = 'claude-sonnet-4-6';
|
|
44
|
-
/** Per-question budget cap (USD). HAL uses Sonnet 4.5 so $0.30 headroom is safe. */
|
|
45
|
-
export const CLAUDE_P_PER_QUESTION_BUDGET_USD = 0.30;
|
|
46
|
-
/** Subprocess timeout: 5 minutes per question. */
|
|
47
|
-
export const CLAUDE_P_TIMEOUT_MS = 5 * 60 * 1000;
|
|
48
|
-
/** FINAL_ANSWER extraction pattern — same as gaia-agent.ts. */
|
|
49
|
-
const FINAL_ANSWER_RE = /FINAL_ANSWER:\s*(.+?)(?:\n|$)/i;
|
|
50
|
-
// ---------------------------------------------------------------------------
|
|
51
|
-
// Prompt builder
|
|
52
|
-
// ---------------------------------------------------------------------------
|
|
53
|
-
/**
|
|
54
|
-
* Build the prompt sent to claude -p for a GAIA question.
|
|
55
|
-
*
|
|
56
|
-
* Includes the question text, optional attachment path, and precise instructions
|
|
57
|
-
* for using available tools and producing FINAL_ANSWER: in the expected format.
|
|
58
|
-
*/
|
|
59
|
-
export function buildClaudePPrompt(question) {
|
|
60
|
-
const attachmentBlock = question.file_path
|
|
61
|
-
? [
|
|
62
|
-
'',
|
|
63
|
-
`Attachment file path: ${path.resolve(question.file_path)}`,
|
|
64
|
-
'You can use the Read tool to view this file directly.',
|
|
65
|
-
'For images, audio, PDF, spreadsheets — use Read (it handles multimodal formats natively).',
|
|
66
|
-
].join('\n')
|
|
67
|
-
: '';
|
|
68
|
-
return [
|
|
69
|
-
'You are answering a question from the GAIA benchmark.',
|
|
70
|
-
'',
|
|
71
|
-
`Question: ${question.question}`,
|
|
72
|
-
attachmentBlock,
|
|
73
|
-
'',
|
|
74
|
-
'Instructions:',
|
|
75
|
-
'1. Find the answer using your available tools:',
|
|
76
|
-
' - WebSearch: for current facts, people, events, statistics',
|
|
77
|
-
' - WebFetch: for full page content when a snippet is insufficient',
|
|
78
|
-
' - Read: for file attachments (PDF, DOCX, XLSX, images, audio transcripts)',
|
|
79
|
-
' - Bash: for computation, data processing, Python scripts',
|
|
80
|
-
'2. Be precise. If a number is asked, give just the number.',
|
|
81
|
-
' If a name is asked, give just the name.',
|
|
82
|
-
'3. Output your final answer on the last line as exactly:',
|
|
83
|
-
' FINAL_ANSWER: <your-answer>',
|
|
84
|
-
'4. Do NOT use commas in numbers (write 50000 not 50,000) unless explicitly required.',
|
|
85
|
-
'5. Do NOT include units (write "5" not "5 km") unless the question explicitly asks for them.',
|
|
86
|
-
'6. If you genuinely cannot find the answer after exhausting your tools, write:',
|
|
87
|
-
' FINAL_ANSWER: unknown',
|
|
88
|
-
'',
|
|
89
|
-
'Find the answer and commit it with FINAL_ANSWER: on the last line.',
|
|
90
|
-
].join('\n');
|
|
91
|
-
}
|
|
92
|
-
// ---------------------------------------------------------------------------
|
|
93
|
-
// Core runner
|
|
94
|
-
// ---------------------------------------------------------------------------
|
|
95
|
-
/**
|
|
96
|
-
* Run a single GAIA question via `claude -p` headless mode.
|
|
97
|
-
*
|
|
98
|
-
* Spawns a subprocess, captures JSON output, extracts the final answer.
|
|
99
|
-
*/
|
|
100
|
-
export async function runGaiaQuestionViaClaudeP(question, options = {}) {
|
|
101
|
-
const model = options.model ?? CLAUDE_P_DEFAULT_MODEL;
|
|
102
|
-
const budgetUsd = options.budgetUsd ?? CLAUDE_P_PER_QUESTION_BUDGET_USD;
|
|
103
|
-
const timeoutMs = options.timeoutMs ?? CLAUDE_P_TIMEOUT_MS;
|
|
104
|
-
const claudeBin = options.claudeBin ?? '/Users/cohen/.local/bin/claude';
|
|
105
|
-
const prompt = buildClaudePPrompt(question);
|
|
106
|
-
// Build claude -p arguments.
|
|
107
|
-
// --dangerously-skip-permissions: acceptable here because GAIA is a read-only
|
|
108
|
-
// sandboxed benchmark context with no real-world side effects. See module-level
|
|
109
|
-
// security note above.
|
|
110
|
-
const args = [
|
|
111
|
-
'-p',
|
|
112
|
-
prompt,
|
|
113
|
-
'--model', model,
|
|
114
|
-
'--max-budget-usd', String(budgetUsd),
|
|
115
|
-
'--output-format', 'json',
|
|
116
|
-
'--dangerously-skip-permissions',
|
|
117
|
-
];
|
|
118
|
-
const startMs = Date.now();
|
|
119
|
-
return new Promise((resolve) => {
|
|
120
|
-
let stdout = '';
|
|
121
|
-
let stderr = '';
|
|
122
|
-
let settled = false;
|
|
123
|
-
const child = spawn(claudeBin, args, {
|
|
124
|
-
env: { ...process.env },
|
|
125
|
-
stdio: ['ignore', 'pipe', 'pipe'],
|
|
126
|
-
});
|
|
127
|
-
child.stdout.on('data', (chunk) => {
|
|
128
|
-
stdout += chunk.toString('utf-8');
|
|
129
|
-
});
|
|
130
|
-
child.stderr.on('data', (chunk) => {
|
|
131
|
-
stderr += chunk.toString('utf-8');
|
|
132
|
-
});
|
|
133
|
-
const timer = setTimeout(() => {
|
|
134
|
-
if (!settled) {
|
|
135
|
-
settled = true;
|
|
136
|
-
child.kill('SIGTERM');
|
|
137
|
-
resolve({
|
|
138
|
-
finalAnswer: null,
|
|
139
|
-
rawResult: '',
|
|
140
|
-
isError: true,
|
|
141
|
-
errorMessage: `Timed out after ${timeoutMs}ms`,
|
|
142
|
-
costUsd: 0,
|
|
143
|
-
wallMs: Date.now() - startMs,
|
|
144
|
-
numTurns: 0,
|
|
145
|
-
stopReason: 'timeout',
|
|
146
|
-
});
|
|
147
|
-
}
|
|
148
|
-
}, timeoutMs);
|
|
149
|
-
child.on('close', (code) => {
|
|
150
|
-
if (settled)
|
|
151
|
-
return;
|
|
152
|
-
settled = true;
|
|
153
|
-
clearTimeout(timer);
|
|
154
|
-
const wallMs = Date.now() - startMs;
|
|
155
|
-
// Parse JSON output from claude -p
|
|
156
|
-
const parsed = parseClaudePOutput(stdout, stderr, code ?? 1);
|
|
157
|
-
resolve({ ...parsed, wallMs });
|
|
158
|
-
});
|
|
159
|
-
child.on('error', (err) => {
|
|
160
|
-
if (settled)
|
|
161
|
-
return;
|
|
162
|
-
settled = true;
|
|
163
|
-
clearTimeout(timer);
|
|
164
|
-
resolve({
|
|
165
|
-
finalAnswer: null,
|
|
166
|
-
rawResult: '',
|
|
167
|
-
isError: true,
|
|
168
|
-
errorMessage: `Failed to spawn claude: ${err.message}`,
|
|
169
|
-
costUsd: 0,
|
|
170
|
-
wallMs: Date.now() - startMs,
|
|
171
|
-
numTurns: 0,
|
|
172
|
-
});
|
|
173
|
-
});
|
|
174
|
-
});
|
|
175
|
-
}
|
|
176
|
-
function parseClaudePOutput(stdout, stderr, exitCode) {
|
|
177
|
-
const raw = stdout.trim();
|
|
178
|
-
// Attempt JSON parse
|
|
179
|
-
let parsed = null;
|
|
180
|
-
try {
|
|
181
|
-
parsed = JSON.parse(raw);
|
|
182
|
-
}
|
|
183
|
-
catch {
|
|
184
|
-
// JSON parse failed — treat as error
|
|
185
|
-
return {
|
|
186
|
-
finalAnswer: null,
|
|
187
|
-
rawResult: raw || stderr.trim(),
|
|
188
|
-
isError: true,
|
|
189
|
-
errorMessage: `JSON parse failed (exitCode=${exitCode}): ${(raw || stderr).slice(0, 200)}`,
|
|
190
|
-
costUsd: 0,
|
|
191
|
-
numTurns: 0,
|
|
192
|
-
};
|
|
193
|
-
}
|
|
194
|
-
const isError = parsed.is_error === true || exitCode !== 0;
|
|
195
|
-
const resultText = parsed.result ?? '';
|
|
196
|
-
const costUsd = parsed.total_cost_usd ?? 0;
|
|
197
|
-
const numTurns = parsed.num_turns ?? 0;
|
|
198
|
-
const stopReason = parsed.subtype ?? parsed.stop_reason ?? undefined;
|
|
199
|
-
if (isError) {
|
|
200
|
-
const errMsg = (parsed.errors ?? []).join('; ') || `subtype=${parsed.subtype}`;
|
|
201
|
-
return {
|
|
202
|
-
finalAnswer: null,
|
|
203
|
-
rawResult: resultText,
|
|
204
|
-
isError: true,
|
|
205
|
-
errorMessage: errMsg,
|
|
206
|
-
costUsd,
|
|
207
|
-
numTurns,
|
|
208
|
-
stopReason,
|
|
209
|
-
};
|
|
210
|
-
}
|
|
211
|
-
// Extract FINAL_ANSWER from result text
|
|
212
|
-
const finalAnswer = extractFinalAnswer(resultText);
|
|
213
|
-
return {
|
|
214
|
-
finalAnswer,
|
|
215
|
-
rawResult: resultText,
|
|
216
|
-
isError: false,
|
|
217
|
-
costUsd,
|
|
218
|
-
numTurns,
|
|
219
|
-
stopReason,
|
|
220
|
-
};
|
|
221
|
-
}
|
|
222
|
-
// ---------------------------------------------------------------------------
|
|
223
|
-
// Answer extraction
|
|
224
|
-
// ---------------------------------------------------------------------------
|
|
225
|
-
/**
|
|
226
|
-
* Extract the FINAL_ANSWER value from claude -p's result text.
|
|
227
|
-
*
|
|
228
|
-
* Primary: regex match on `FINAL_ANSWER: <value>`
|
|
229
|
-
* Fallback: last non-empty line if no FINAL_ANSWER marker found.
|
|
230
|
-
*/
|
|
231
|
-
export function extractFinalAnswer(text) {
|
|
232
|
-
if (!text || !text.trim())
|
|
233
|
-
return null;
|
|
234
|
-
// Primary: FINAL_ANSWER: pattern
|
|
235
|
-
const match = FINAL_ANSWER_RE.exec(text);
|
|
236
|
-
if (match && match[1]) {
|
|
237
|
-
return match[1].trim();
|
|
238
|
-
}
|
|
239
|
-
// Fallback: last meaningful line (heuristic)
|
|
240
|
-
const lines = text.split('\n').map((l) => l.trim()).filter(Boolean);
|
|
241
|
-
const lastLine = lines[lines.length - 1];
|
|
242
|
-
if (lastLine && lastLine.length < 200) {
|
|
243
|
-
return lastLine;
|
|
244
|
-
}
|
|
245
|
-
return null;
|
|
246
|
-
}
|
|
247
|
-
/**
|
|
248
|
-
* Run a batch of GAIA questions through the claude -p wrapper.
|
|
249
|
-
*
|
|
250
|
-
* Concurrency is limited (default 2) because each claude -p subprocess
|
|
251
|
-
* is heavyweight — it starts a full Claude Code session with LSP etc.
|
|
252
|
-
*/
|
|
253
|
-
export async function runGaiaQuestionsBatchViaClaudeP(questions, options = {}) {
|
|
254
|
-
const concurrency = options.concurrency ?? 2;
|
|
255
|
-
const results = new Array(questions.length);
|
|
256
|
-
for (let i = 0; i < questions.length; i += concurrency) {
|
|
257
|
-
const batch = questions.slice(i, Math.min(i + concurrency, questions.length));
|
|
258
|
-
const batchResults = await Promise.all(batch.map((q, batchIdx) => runGaiaQuestionViaClaudeP(q, options).then((r) => {
|
|
259
|
-
const globalIdx = i + batchIdx;
|
|
260
|
-
options.onProgress?.(globalIdx + 1, questions.length, q.task_id, r.finalAnswer, r.costUsd);
|
|
261
|
-
return r;
|
|
262
|
-
})));
|
|
263
|
-
for (let j = 0; j < batchResults.length; j++) {
|
|
264
|
-
results[i + j] = batchResults[j];
|
|
265
|
-
}
|
|
266
|
-
}
|
|
267
|
-
return results;
|
|
268
|
-
}
|
|
269
|
-
//# sourceMappingURL=gaia-claude-p.js.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"gaia-claude-p.js","sourceRoot":"","sources":["../../../src/benchmarks/gaia-claude-p.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAmCG;AAEH,OAAO,EAAE,KAAK,EAAE,MAAM,oBAAoB,CAAC;AAC3C,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAGlC,8EAA8E;AAC9E,YAAY;AACZ,8EAA8E;AAE9E,iFAAiF;AACjF,MAAM,CAAC,MAAM,sBAAsB,GAAG,mBAAmB,CAAC;AAE1D,oFAAoF;AACpF,MAAM,CAAC,MAAM,gCAAgC,GAAG,IAAI,CAAC;AAErD,kDAAkD;AAClD,MAAM,CAAC,MAAM,mBAAmB,GAAG,CAAC,GAAG,EAAE,GAAG,IAAI,CAAC;AAEjD,+DAA+D;AAC/D,MAAM,eAAe,GAAG,gCAAgC,CAAC;AAoCzD,8EAA8E;AAC9E,iBAAiB;AACjB,8EAA8E;AAE9E;;;;;GAKG;AACH,MAAM,UAAU,kBAAkB,CAAC,QAAsB;IACvD,MAAM,eAAe,GAAG,QAAQ,CAAC,SAAS;QACxC,CAAC,CAAC;YACE,EAAE;YACF,yBAAyB,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE;YAC3D,uDAAuD;YACvD,2FAA2F;SAC5F,CAAC,IAAI,CAAC,IAAI,CAAC;QACd,CAAC,CAAC,EAAE,CAAC;IAEP,OAAO;QACL,uDAAuD;QACvD,EAAE;QACF,aAAa,QAAQ,CAAC,QAAQ,EAAE;QAChC,eAAe;QACf,EAAE;QACF,eAAe;QACf,gDAAgD;QAChD,+DAA+D;QAC/D,qEAAqE;QACrE,8EAA8E;QAC9E,6DAA6D;QAC7D,4DAA4D;QAC5D,4CAA4C;QAC5C,0DAA0D;QAC1D,gCAAgC;QAChC,sFAAsF;QACtF,8FAA8F;QAC9F,gFAAgF;QAChF,0BAA0B;QAC1B,EAAE;QACF,oEAAoE;KACrE,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AACf,CAAC;AAED,8EAA8E;AAC9E,cAAc;AACd,8EAA8E;AAE9E;;;;GAIG;AACH,MAAM,CAAC,KAAK,UAAU,yBAAyB,CAC7C,QAAsB,EACtB,UAA0B,EAAE;IAE5B,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,IAAI,sBAAsB,CAAC;IACtD,MAAM,SAAS,GAAG,OAAO,CAAC,SAAS,IAAI,gCAAgC,CAAC;IACxE,MAAM,SAAS,GAAG,OAAO,CAAC,SAAS,IAAI,mBAAmB,CAAC;IAC3D,MAAM,SAAS,GAAG,OAAO,CAAC,SAAS,IAAI,gCAAgC,CAAC;IAExE,MAAM,MAAM,GAAG,kBAAkB,CAAC,QAAQ,CAAC,CAAC;IAE5C,6BAA6B;IAC7B,8EAA8E;IAC9E,iFAAiF;IACjF,uBAAuB;IACvB,MAAM,IAAI,GAAa;QACrB,IAAI;QACJ,MAAM;QACN,SAAS,EAAE,KAAK;QAChB,kBAAkB,EAAE,MAAM,CAAC,SAAS,CAAC;QACrC,iBAAiB,EAAE,MAAM;QACzB,gCAAgC;KACjC,CAAC;IAEF,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAE3B,OAAO,IAAI,OAAO,CAAgB,CAAC,OAAO,EAAE,EAAE;QAC5C,IAAI,MAAM,GAAG,EAAE,CAAC;QAChB,IAAI,MAAM,GAAG,EAAE,CAAC;QAChB,IAAI,OAAO,GAAG,KAAK,CAAC;QAEpB,MAAM,KAAK,GAAG,KAAK,CAAC,SAAS,EAAE,IAAI,EAAE;YACnC,GAAG,EAAE,EAAE,GAAG,OAAO,CAAC,GAAG,EAAE;YACvB,KAAK,EAAE,CAAC,QAAQ,EAAE,MAAM,EAAE,MAAM,CAAC;SAClC,CAAC,CAAC;QAEH,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAa,EAAE,EAAE;YACxC,MAAM,IAAI,KAAK,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;QACpC,CAAC,CAAC,CAAC;QAEH,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAa,EAAE,EAAE;YACxC,MAAM,IAAI,KAAK,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;QACpC,CAAC,CAAC,CAAC;QAEH,MAAM,KAAK,GAAG,UAAU,CAAC,GAAG,EAAE;YAC5B,IAAI,CAAC,OAAO,EAAE,CAAC;gBACb,OAAO,GAAG,IAAI,CAAC;gBACf,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;gBACtB,OAAO,CAAC;oBACN,WAAW,EAAE,IAAI;oBACjB,SAAS,EAAE,EAAE;oBACb,OAAO,EAAE,IAAI;oBACb,YAAY,EAAE,mBAAmB,SAAS,IAAI;oBAC9C,OAAO,EAAE,CAAC;oBACV,MAAM,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO;oBAC5B,QAAQ,EAAE,CAAC;oBACX,UAAU,EAAE,SAAS;iBACtB,CAAC,CAAC;YACL,CAAC;QACH,CAAC,EAAE,SAAS,CAAC,CAAC;QAEd,KAAK,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,IAAI,EAAE,EAAE;YACzB,IAAI,OAAO;gBAAE,OAAO;YACpB,OAAO,GAAG,IAAI,CAAC;YACf,YAAY,CAAC,KAAK,CAAC,CAAC;YAEpB,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO,CAAC;YAEpC,mCAAmC;YACnC,MAAM,MAAM,GAAG,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,IAAI,CAAC,CAAC,CAAC;YAE7D,OAAO,CAAC,EAAE,GAAG,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;QACjC,CAAC,CAAC,CAAC;QAEH,KAAK,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;YACxB,IAAI,OAAO;gBAAE,OAAO;YACpB,OAAO,GAAG,IAAI,CAAC;YACf,YAAY,CAAC,KAAK,CAAC,CAAC;YACpB,OAAO,CAAC;gBACN,WAAW,EAAE,IAAI;gBACjB,SAAS,EAAE,EAAE;gBACb,OAAO,EAAE,IAAI;gBACb,YAAY,EAAE,2BAA2B,GAAG,CAAC,OAAO,EAAE;gBACtD,OAAO,EAAE,CAAC;gBACV,MAAM,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO;gBAC5B,QAAQ,EAAE,CAAC;aACZ,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;AACL,CAAC;AAiBD,SAAS,kBAAkB,CACzB,MAAc,EACd,MAAc,EACd,QAAgB;IAEhB,MAAM,GAAG,GAAG,MAAM,CAAC,IAAI,EAAE,CAAC;IAE1B,qBAAqB;IACrB,IAAI,MAAM,GAA6B,IAAI,CAAC;IAC5C,IAAI,CAAC;QACH,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAAsB,CAAC;IAChD,CAAC;IAAC,MAAM,CAAC;QACP,qCAAqC;QACrC,OAAO;YACL,WAAW,EAAE,IAAI;YACjB,SAAS,EAAE,GAAG,IAAI,MAAM,CAAC,IAAI,EAAE;YAC/B,OAAO,EAAE,IAAI;YACb,YAAY,EAAE,+BAA+B,QAAQ,MAAM,CAAC,GAAG,IAAI,MAAM,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE;YAC1F,OAAO,EAAE,CAAC;YACV,QAAQ,EAAE,CAAC;SACZ,CAAC;IACJ,CAAC;IAED,MAAM,OAAO,GAAG,MAAM,CAAC,QAAQ,KAAK,IAAI,IAAI,QAAQ,KAAK,CAAC,CAAC;IAC3D,MAAM,UAAU,GAAG,MAAM,CAAC,MAAM,IAAI,EAAE,CAAC;IACvC,MAAM,OAAO,GAAG,MAAM,CAAC,cAAc,IAAI,CAAC,CAAC;IAC3C,MAAM,QAAQ,GAAG,MAAM,CAAC,SAAS,IAAI,CAAC,CAAC;IACvC,MAAM,UAAU,GAAG,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,WAAW,IAAI,SAAS,CAAC;IAErE,IAAI,OAAO,EAAE,CAAC;QACZ,MAAM,MAAM,GAAG,CAAC,MAAM,CAAC,MAAM,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,WAAW,MAAM,CAAC,OAAO,EAAE,CAAC;QAC/E,OAAO;YACL,WAAW,EAAE,IAAI;YACjB,SAAS,EAAE,UAAU;YACrB,OAAO,EAAE,IAAI;YACb,YAAY,EAAE,MAAM;YACpB,OAAO;YACP,QAAQ;YACR,UAAU;SACX,CAAC;IACJ,CAAC;IAED,wCAAwC;IACxC,MAAM,WAAW,GAAG,kBAAkB,CAAC,UAAU,CAAC,CAAC;IAEnD,OAAO;QACL,WAAW;QACX,SAAS,EAAE,UAAU;QACrB,OAAO,EAAE,KAAK;QACd,OAAO;QACP,QAAQ;QACR,UAAU;KACX,CAAC;AACJ,CAAC;AAED,8EAA8E;AAC9E,oBAAoB;AACpB,8EAA8E;AAE9E;;;;;GAKG;AACH,MAAM,UAAU,kBAAkB,CAAC,IAAY;IAC7C,IAAI,CAAC,IAAI,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE;QAAE,OAAO,IAAI,CAAC;IAEvC,iCAAiC;IACjC,MAAM,KAAK,GAAG,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACzC,IAAI,KAAK,IAAI,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC;QACtB,OAAO,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IACzB,CAAC;IAED,6CAA6C;IAC7C,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;IACpE,MAAM,QAAQ,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IACzC,IAAI,QAAQ,IAAI,QAAQ,CAAC,MAAM,GAAG,GAAG,EAAE,CAAC;QACtC,OAAO,QAAQ,CAAC;IAClB,CAAC;IAED,OAAO,IAAI,CAAC;AACd,CAAC;AAaD;;;;;GAKG;AACH,MAAM,CAAC,KAAK,UAAU,+BAA+B,CACnD,SAAyB,EACzB,UAA+B,EAAE;IAEjC,MAAM,WAAW,GAAG,OAAO,CAAC,WAAW,IAAI,CAAC,CAAC;IAC7C,MAAM,OAAO,GAAoB,IAAI,KAAK,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC;IAE7D,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,SAAS,CAAC,MAAM,EAAE,CAAC,IAAI,WAAW,EAAE,CAAC;QACvD,MAAM,KAAK,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,WAAW,EAAE,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC;QAC9E,MAAM,YAAY,GAAG,MAAM,OAAO,CAAC,GAAG,CACpC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,EAAE,CACxB,yBAAyB,CAAC,CAAC,EAAE,OAAO,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE;YAC/C,MAAM,SAAS,GAAG,CAAC,GAAG,QAAQ,CAAC;YAC/B,OAAO,CAAC,UAAU,EAAE,CAAC,SAAS,GAAG,CAAC,EAAE,SAAS,CAAC,MAAM,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC;YAC3F,OAAO,CAAC,CAAC;QACX,CAAC,CAAC,CACH,CACF,CAAC;QACF,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,YAAY,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YAC7C,OAAO,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,YAAY,CAAC,CAAC,CAAC,CAAC;QACnC,CAAC;IACH,CAAC;IAED,OAAO,OAAO,CAAC;AACjB,CAAC"}
|
|
@@ -1,20 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* Smoke tests for gaia-claude-p.ts — iter 54 (#2156)
|
|
3
|
-
*
|
|
4
|
-
* Three minimal cases to verify the claude -p wrapper end-to-end:
|
|
5
|
-
* 1. Simple arithmetic: "what is 2+2" → expect "4"
|
|
6
|
-
* 2. Current fact: population of Tokyo → expect a large number
|
|
7
|
-
* 3. (Optional) Attachment path test — only runs when HF cache has a file
|
|
8
|
-
*
|
|
9
|
-
* Cost cap: ~$0.25 × 3 = $0.75 worst case at Sonnet rates, but using Haiku
|
|
10
|
-
* for smoke to keep it under $0.30 total.
|
|
11
|
-
*
|
|
12
|
-
* Run:
|
|
13
|
-
* npx ts-node src/benchmarks/gaia-claude-p.smoke.ts
|
|
14
|
-
* # or after build:
|
|
15
|
-
* node dist/src/benchmarks/gaia-claude-p.smoke.js
|
|
16
|
-
*
|
|
17
|
-
* Refs: iter 54, #2156
|
|
18
|
-
*/
|
|
19
|
-
export {};
|
|
20
|
-
//# sourceMappingURL=gaia-claude-p.smoke.d.ts.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"gaia-claude-p.smoke.d.ts","sourceRoot":"","sources":["../../../src/benchmarks/gaia-claude-p.smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG"}
|
|
@@ -1,168 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* Smoke tests for gaia-claude-p.ts — iter 54 (#2156)
|
|
3
|
-
*
|
|
4
|
-
* Three minimal cases to verify the claude -p wrapper end-to-end:
|
|
5
|
-
* 1. Simple arithmetic: "what is 2+2" → expect "4"
|
|
6
|
-
* 2. Current fact: population of Tokyo → expect a large number
|
|
7
|
-
* 3. (Optional) Attachment path test — only runs when HF cache has a file
|
|
8
|
-
*
|
|
9
|
-
* Cost cap: ~$0.25 × 3 = $0.75 worst case at Sonnet rates, but using Haiku
|
|
10
|
-
* for smoke to keep it under $0.30 total.
|
|
11
|
-
*
|
|
12
|
-
* Run:
|
|
13
|
-
* npx ts-node src/benchmarks/gaia-claude-p.smoke.ts
|
|
14
|
-
* # or after build:
|
|
15
|
-
* node dist/src/benchmarks/gaia-claude-p.smoke.js
|
|
16
|
-
*
|
|
17
|
-
* Refs: iter 54, #2156
|
|
18
|
-
*/
|
|
19
|
-
import { runGaiaQuestionViaClaudeP, extractFinalAnswer, buildClaudePPrompt, CLAUDE_P_DEFAULT_MODEL, } from './gaia-claude-p.js';
|
|
20
|
-
// ---------------------------------------------------------------------------
|
|
21
|
-
// Test helpers
|
|
22
|
-
// ---------------------------------------------------------------------------
|
|
23
|
-
let passed = 0;
|
|
24
|
-
let failed = 0;
|
|
25
|
-
function assert(condition, label, detail) {
|
|
26
|
-
if (condition) {
|
|
27
|
-
console.log(` PASS ${label}`);
|
|
28
|
-
passed++;
|
|
29
|
-
}
|
|
30
|
-
else {
|
|
31
|
-
console.error(` FAIL ${label}${detail ? ': ' + detail : ''}`);
|
|
32
|
-
failed++;
|
|
33
|
-
}
|
|
34
|
-
}
|
|
35
|
-
// ---------------------------------------------------------------------------
|
|
36
|
-
// Unit tests (no claude -p invocation)
|
|
37
|
-
// ---------------------------------------------------------------------------
|
|
38
|
-
function testExtractFinalAnswer() {
|
|
39
|
-
console.log('\n-- Unit: extractFinalAnswer --');
|
|
40
|
-
assert(extractFinalAnswer('FINAL_ANSWER: 4') === '4', 'extracts plain number');
|
|
41
|
-
assert(extractFinalAnswer('Some reasoning...\nFINAL_ANSWER: Paris') === 'Paris', 'extracts after reasoning text');
|
|
42
|
-
assert(extractFinalAnswer('final_answer: 42') === '42', 'case-insensitive extraction');
|
|
43
|
-
assert(extractFinalAnswer('') === null, 'returns null for empty string');
|
|
44
|
-
assert(extractFinalAnswer('No answer marker here') !== null, 'fallback to last line when no marker');
|
|
45
|
-
}
|
|
46
|
-
function testBuildClaudePPrompt() {
|
|
47
|
-
console.log('\n-- Unit: buildClaudePPrompt --');
|
|
48
|
-
const q = {
|
|
49
|
-
task_id: 'test-001',
|
|
50
|
-
level: 1,
|
|
51
|
-
question: 'What is 2+2?',
|
|
52
|
-
final_answer: '4',
|
|
53
|
-
file_name: null,
|
|
54
|
-
file_path: null,
|
|
55
|
-
};
|
|
56
|
-
const prompt = buildClaudePPrompt(q);
|
|
57
|
-
assert(prompt.includes('What is 2+2?'), 'prompt contains question text');
|
|
58
|
-
assert(prompt.includes('FINAL_ANSWER:'), 'prompt instructs FINAL_ANSWER format');
|
|
59
|
-
assert(!prompt.includes('Attachment'), 'no attachment section for null file_path');
|
|
60
|
-
const qWithFile = { ...q, file_path: '/tmp/test.pdf' };
|
|
61
|
-
const promptWithFile = buildClaudePPrompt(qWithFile);
|
|
62
|
-
assert(promptWithFile.includes('/tmp/test.pdf'), 'attachment path included when present');
|
|
63
|
-
assert(promptWithFile.includes('Read tool'), 'mentions Read tool for attachments');
|
|
64
|
-
}
|
|
65
|
-
// ---------------------------------------------------------------------------
|
|
66
|
-
// Integration tests (invoke claude -p)
|
|
67
|
-
// ---------------------------------------------------------------------------
|
|
68
|
-
async function testArithmetic() {
|
|
69
|
-
console.log('\n-- Integration: arithmetic (2+2) --');
|
|
70
|
-
const q = {
|
|
71
|
-
task_id: 'smoke-arithmetic',
|
|
72
|
-
level: 1,
|
|
73
|
-
question: 'What is 2 + 2? Provide only the numeric answer.',
|
|
74
|
-
final_answer: '4',
|
|
75
|
-
file_name: null,
|
|
76
|
-
file_path: null,
|
|
77
|
-
};
|
|
78
|
-
console.log(' Spawning claude -p ...');
|
|
79
|
-
const result = await runGaiaQuestionViaClaudeP(q, {
|
|
80
|
-
model: 'claude-haiku-4-5',
|
|
81
|
-
budgetUsd: 0.25,
|
|
82
|
-
timeoutMs: 120_000,
|
|
83
|
-
});
|
|
84
|
-
console.log(` raw result: ${result.rawResult.slice(0, 100)}`);
|
|
85
|
-
console.log(` finalAnswer: ${result.finalAnswer}`);
|
|
86
|
-
console.log(` costUsd: $${result.costUsd.toFixed(4)}`);
|
|
87
|
-
console.log(` numTurns: ${result.numTurns}`);
|
|
88
|
-
console.log(` isError: ${result.isError}`);
|
|
89
|
-
assert(!result.isError, 'no error', result.errorMessage);
|
|
90
|
-
assert(result.finalAnswer !== null, 'finalAnswer is not null');
|
|
91
|
-
assert(result.finalAnswer === '4', `answer is "4"`, `got: "${result.finalAnswer}"`);
|
|
92
|
-
assert(result.costUsd > 0, 'cost is positive');
|
|
93
|
-
assert(result.wallMs > 0, 'wallMs is positive');
|
|
94
|
-
}
|
|
95
|
-
async function testCurrentFact() {
|
|
96
|
-
console.log('\n-- Integration: current fact (Tokyo population) --');
|
|
97
|
-
const q = {
|
|
98
|
-
task_id: 'smoke-tokyo-pop',
|
|
99
|
-
level: 1,
|
|
100
|
-
question: 'What is the approximate population of Tokyo (the city proper)? Give only a number in millions, rounded to the nearest million.',
|
|
101
|
-
final_answer: '14',
|
|
102
|
-
file_name: null,
|
|
103
|
-
file_path: null,
|
|
104
|
-
};
|
|
105
|
-
console.log(' Spawning claude -p ...');
|
|
106
|
-
const result = await runGaiaQuestionViaClaudeP(q, {
|
|
107
|
-
model: 'claude-haiku-4-5',
|
|
108
|
-
budgetUsd: 0.25,
|
|
109
|
-
timeoutMs: 120_000,
|
|
110
|
-
});
|
|
111
|
-
console.log(` raw result excerpt: ${result.rawResult.slice(0, 150)}`);
|
|
112
|
-
console.log(` finalAnswer: ${result.finalAnswer}`);
|
|
113
|
-
console.log(` costUsd: $${result.costUsd.toFixed(4)}`);
|
|
114
|
-
assert(!result.isError, 'no error', result.errorMessage);
|
|
115
|
-
assert(result.finalAnswer !== null, 'finalAnswer is not null');
|
|
116
|
-
// Tokyo metro area is ~37M, city proper is ~13-14M — any multi-digit number is reasonable
|
|
117
|
-
const numAnswer = result.finalAnswer ? parseFloat(result.finalAnswer.replace(/[^0-9.]/g, '')) : NaN;
|
|
118
|
-
assert(!isNaN(numAnswer) && numAnswer > 0, 'answer is a non-empty number', `got: "${result.finalAnswer}"`);
|
|
119
|
-
}
|
|
120
|
-
async function testAnswerFormat() {
|
|
121
|
-
console.log('\n-- Integration: answer format discipline --');
|
|
122
|
-
const q = {
|
|
123
|
-
task_id: 'smoke-format',
|
|
124
|
-
level: 1,
|
|
125
|
-
question: 'What is the capital of France? Give only the city name, nothing else.',
|
|
126
|
-
final_answer: 'Paris',
|
|
127
|
-
file_name: null,
|
|
128
|
-
file_path: null,
|
|
129
|
-
};
|
|
130
|
-
console.log(' Spawning claude -p ...');
|
|
131
|
-
const result = await runGaiaQuestionViaClaudeP(q, {
|
|
132
|
-
model: 'claude-haiku-4-5',
|
|
133
|
-
budgetUsd: 0.25,
|
|
134
|
-
timeoutMs: 120_000,
|
|
135
|
-
});
|
|
136
|
-
console.log(` raw result excerpt: ${result.rawResult.slice(0, 150)}`);
|
|
137
|
-
console.log(` finalAnswer: ${result.finalAnswer}`);
|
|
138
|
-
console.log(` costUsd: $${result.costUsd.toFixed(4)}`);
|
|
139
|
-
assert(!result.isError, 'no error', result.errorMessage);
|
|
140
|
-
assert(result.finalAnswer !== null, 'finalAnswer is not null');
|
|
141
|
-
assert(result.finalAnswer !== null && result.finalAnswer.toLowerCase().includes('paris'), 'answer contains "paris"', `got: "${result.finalAnswer}"`);
|
|
142
|
-
}
|
|
143
|
-
// ---------------------------------------------------------------------------
|
|
144
|
-
// Main
|
|
145
|
-
// ---------------------------------------------------------------------------
|
|
146
|
-
async function main() {
|
|
147
|
-
console.log('=== gaia-claude-p smoke tests ===');
|
|
148
|
-
console.log(`Model: claude-haiku-4-5 (smoke uses Haiku to minimize cost)`);
|
|
149
|
-
console.log(`Default production model: ${CLAUDE_P_DEFAULT_MODEL}`);
|
|
150
|
-
// Unit tests (no API calls)
|
|
151
|
-
testExtractFinalAnswer();
|
|
152
|
-
testBuildClaudePPrompt();
|
|
153
|
-
// Integration tests (invoke claude -p)
|
|
154
|
-
await testArithmetic();
|
|
155
|
-
await testCurrentFact();
|
|
156
|
-
await testAnswerFormat();
|
|
157
|
-
// Summary
|
|
158
|
-
const total = passed + failed;
|
|
159
|
-
console.log(`\n=== Results: ${passed}/${total} passed ===`);
|
|
160
|
-
if (failed > 0) {
|
|
161
|
-
process.exit(1);
|
|
162
|
-
}
|
|
163
|
-
}
|
|
164
|
-
main().catch((err) => {
|
|
165
|
-
console.error('Smoke test runner error:', err);
|
|
166
|
-
process.exit(1);
|
|
167
|
-
});
|
|
168
|
-
//# sourceMappingURL=gaia-claude-p.smoke.js.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"gaia-claude-p.smoke.js","sourceRoot":"","sources":["../../../src/benchmarks/gaia-claude-p.smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAEH,OAAO,EACL,yBAAyB,EACzB,kBAAkB,EAClB,kBAAkB,EAClB,sBAAsB,GACvB,MAAM,oBAAoB,CAAC;AAG5B,8EAA8E;AAC9E,eAAe;AACf,8EAA8E;AAE9E,IAAI,MAAM,GAAG,CAAC,CAAC;AACf,IAAI,MAAM,GAAG,CAAC,CAAC;AAEf,SAAS,MAAM,CAAC,SAAkB,EAAE,KAAa,EAAE,MAAe;IAChE,IAAI,SAAS,EAAE,CAAC;QACd,OAAO,CAAC,GAAG,CAAC,WAAW,KAAK,EAAE,CAAC,CAAC;QAChC,MAAM,EAAE,CAAC;IACX,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,KAAK,CAAC,WAAW,KAAK,GAAG,MAAM,CAAC,CAAC,CAAC,IAAI,GAAG,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;QAChE,MAAM,EAAE,CAAC;IACX,CAAC;AACH,CAAC;AAED,8EAA8E;AAC9E,uCAAuC;AACvC,8EAA8E;AAE9E,SAAS,sBAAsB;IAC7B,OAAO,CAAC,GAAG,CAAC,kCAAkC,CAAC,CAAC;IAEhD,MAAM,CACJ,kBAAkB,CAAC,iBAAiB,CAAC,KAAK,GAAG,EAC7C,uBAAuB,CACxB,CAAC;IACF,MAAM,CACJ,kBAAkB,CAAC,wCAAwC,CAAC,KAAK,OAAO,EACxE,+BAA+B,CAChC,CAAC;IACF,MAAM,CACJ,kBAAkB,CAAC,kBAAkB,CAAC,KAAK,IAAI,EAC/C,6BAA6B,CAC9B,CAAC;IACF,MAAM,CACJ,kBAAkB,CAAC,EAAE,CAAC,KAAK,IAAI,EAC/B,+BAA+B,CAChC,CAAC;IACF,MAAM,CACJ,kBAAkB,CAAC,uBAAuB,CAAC,KAAK,IAAI,EACpD,sCAAsC,CACvC,CAAC;AACJ,CAAC;AAED,SAAS,sBAAsB;IAC7B,OAAO,CAAC,GAAG,CAAC,kCAAkC,CAAC,CAAC;IAEhD,MAAM,CAAC,GAAiB;QACtB,OAAO,EAAE,UAAU;QACnB,KAAK,EAAE,CAAC;QACR,QAAQ,EAAE,cAAc;QACxB,YAAY,EAAE,GAAG;QACjB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,IAAI;KAChB,CAAC;IAEF,MAAM,MAAM,GAAG,kBAAkB,CAAC,CAAC,CAAC,CAAC;IACrC,MAAM,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,CAAC,EAAE,+BAA+B,CAAC,CAAC;IACzE,MAAM,CAAC,MAAM,CAAC,QAAQ,CAAC,eAAe,CAAC,EAAE,sCAAsC,CAAC,CAAC;IACjF,MAAM,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,YAAY,CAAC,EAAE,0CAA0C,CAAC,CAAC;IAEnF,MAAM,SAAS,GAAiB,EAAE,GAAG,CAAC,EAAE,SAAS,EAAE,eAAe,EAAE,CAAC;IACrE,MAAM,cAAc,GAAG,kBAAkB,CAAC,SAAS,CAAC,CAAC;IACrD,MAAM,CAAC,cAAc,CAAC,QAAQ,CAAC,eAAe,CAAC,EAAE,uCAAuC,CAAC,CAAC;IAC1F,MAAM,CAAC,cAAc,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,oCAAoC,CAAC,CAAC;AACrF,CAAC;AAED,8EAA8E;AAC9E,uCAAuC;AACvC,8EAA8E;AAE9E,KAAK,UAAU,cAAc;IAC3B,OAAO,CAAC,GAAG,CAAC,uCAAuC,CAAC,CAAC;IAErD,MAAM,CAAC,GAAiB;QACtB,OAAO,EAAE,kBAAkB;QAC3B,KAAK,EAAE,CAAC;QACR,QAAQ,EAAE,iDAAiD;QAC3D,YAAY,EAAE,GAAG;QACjB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,IAAI;KAChB,CAAC;IAEF,OAAO,CAAC,GAAG,CAAC,0BAA0B,CAAC,CAAC;IACxC,MAAM,MAAM,GAAG,MAAM,yBAAyB,CAAC,CAAC,EAAE;QAChD,KAAK,EAAE,kBAAkB;QACzB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,OAAO;KACnB,CAAC,CAAC;IAEH,OAAO,CAAC,GAAG,CAAC,iBAAiB,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IAC/D,OAAO,CAAC,GAAG,CAAC,kBAAkB,MAAM,CAAC,WAAW,EAAE,CAAC,CAAC;IACpD,OAAO,CAAC,GAAG,CAAC,eAAe,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IACxD,OAAO,CAAC,GAAG,CAAC,eAAe,MAAM,CAAC,QAAQ,EAAE,CAAC,CAAC;IAC9C,OAAO,CAAC,GAAG,CAAC,cAAc,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC;IAE5C,MAAM,CAAC,CAAC,MAAM,CAAC,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,YAAY,CAAC,CAAC;IACzD,MAAM,CAAC,MAAM,CAAC,WAAW,KAAK,IAAI,EAAE,yBAAyB,CAAC,CAAC;IAC/D,MAAM,CAAC,MAAM,CAAC,WAAW,KAAK,GAAG,EAAE,eAAe,EAAE,SAAS,MAAM,CAAC,WAAW,GAAG,CAAC,CAAC;IACpF,MAAM,CAAC,MAAM,CAAC,OAAO,GAAG,CAAC,EAAE,kBAAkB,CAAC,CAAC;IAC/C,MAAM,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,oBAAoB,CAAC,CAAC;AAClD,CAAC;AAED,KAAK,UAAU,eAAe;IAC5B,OAAO,CAAC,GAAG,CAAC,sDAAsD,CAAC,CAAC;IAEpE,MAAM,CAAC,GAAiB;QACtB,OAAO,EAAE,iBAAiB;QAC1B,KAAK,EAAE,CAAC;QACR,QAAQ,EAAE,gIAAgI;QAC1I,YAAY,EAAE,IAAI;QAClB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,IAAI;KAChB,CAAC;IAEF,OAAO,CAAC,GAAG,CAAC,0BAA0B,CAAC,CAAC;IACxC,MAAM,MAAM,GAAG,MAAM,yBAAyB,CAAC,CAAC,EAAE;QAChD,KAAK,EAAE,kBAAkB;QACzB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,OAAO;KACnB,CAAC,CAAC;IAEH,OAAO,CAAC,GAAG,CAAC,yBAAyB,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IACvE,OAAO,CAAC,GAAG,CAAC,kBAAkB,MAAM,CAAC,WAAW,EAAE,CAAC,CAAC;IACpD,OAAO,CAAC,GAAG,CAAC,eAAe,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAExD,MAAM,CAAC,CAAC,MAAM,CAAC,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,YAAY,CAAC,CAAC;IACzD,MAAM,CAAC,MAAM,CAAC,WAAW,KAAK,IAAI,EAAE,yBAAyB,CAAC,CAAC;IAC/D,0FAA0F;IAC1F,MAAM,SAAS,GAAG,MAAM,CAAC,WAAW,CAAC,CAAC,CAAC,UAAU,CAAC,MAAM,CAAC,WAAW,CAAC,OAAO,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC;IACpG,MAAM,CACJ,CAAC,KAAK,CAAC,SAAS,CAAC,IAAI,SAAS,GAAG,CAAC,EAClC,8BAA8B,EAC9B,SAAS,MAAM,CAAC,WAAW,GAAG,CAC/B,CAAC;AACJ,CAAC;AAED,KAAK,UAAU,gBAAgB;IAC7B,OAAO,CAAC,GAAG,CAAC,+CAA+C,CAAC,CAAC;IAE7D,MAAM,CAAC,GAAiB;QACtB,OAAO,EAAE,cAAc;QACvB,KAAK,EAAE,CAAC;QACR,QAAQ,EAAE,uEAAuE;QACjF,YAAY,EAAE,OAAO;QACrB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,IAAI;KAChB,CAAC;IAEF,OAAO,CAAC,GAAG,CAAC,0BAA0B,CAAC,CAAC;IACxC,MAAM,MAAM,GAAG,MAAM,yBAAyB,CAAC,CAAC,EAAE;QAChD,KAAK,EAAE,kBAAkB;QACzB,SAAS,EAAE,IAAI;QACf,SAAS,EAAE,OAAO;KACnB,CAAC,CAAC;IAEH,OAAO,CAAC,GAAG,CAAC,yBAAyB,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IACvE,OAAO,CAAC,GAAG,CAAC,kBAAkB,MAAM,CAAC,WAAW,EAAE,CAAC,CAAC;IACpD,OAAO,CAAC,GAAG,CAAC,eAAe,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAExD,MAAM,CAAC,CAAC,MAAM,CAAC,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,YAAY,CAAC,CAAC;IACzD,MAAM,CAAC,MAAM,CAAC,WAAW,KAAK,IAAI,EAAE,yBAAyB,CAAC,CAAC;IAC/D,MAAM,CACJ,MAAM,CAAC,WAAW,KAAK,IAAI,IAAI,MAAM,CAAC,WAAW,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,EACjF,yBAAyB,EACzB,SAAS,MAAM,CAAC,WAAW,GAAG,CAC/B,CAAC;AACJ,CAAC;AAED,8EAA8E;AAC9E,OAAO;AACP,8EAA8E;AAE9E,KAAK,UAAU,IAAI;IACjB,OAAO,CAAC,GAAG,CAAC,mCAAmC,CAAC,CAAC;IACjD,OAAO,CAAC,GAAG,CAAC,6DAA6D,CAAC,CAAC;IAC3E,OAAO,CAAC,GAAG,CAAC,6BAA6B,sBAAsB,EAAE,CAAC,CAAC;IAEnE,4BAA4B;IAC5B,sBAAsB,EAAE,CAAC;IACzB,sBAAsB,EAAE,CAAC;IAEzB,uCAAuC;IACvC,MAAM,cAAc,EAAE,CAAC;IACvB,MAAM,eAAe,EAAE,CAAC;IACxB,MAAM,gBAAgB,EAAE,CAAC;IAEzB,UAAU;IACV,MAAM,KAAK,GAAG,MAAM,GAAG,MAAM,CAAC;IAC9B,OAAO,CAAC,GAAG,CAAC,kBAAkB,MAAM,IAAI,KAAK,aAAa,CAAC,CAAC;IAE5D,IAAI,MAAM,GAAG,CAAC,EAAE,CAAC;QACf,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,0BAA0B,EAAE,GAAG,CAAC,CAAC;IAC/C,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC,CAAC,CAAC"}
|