@tangle-network/agent-eval 0.191.0 → 0.192.0

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 (95) hide show
  1. package/CHANGELOG.md +38 -0
  2. package/dist/adapters/http.d.ts +2 -2
  3. package/dist/analyst/index.d.ts +6 -6
  4. package/dist/analyst/index.js +3 -3
  5. package/dist/{benchmark-command-3tZJnD82.js → benchmark-command-C458yEFK.js} +7 -7
  6. package/dist/{benchmark-command-3tZJnD82.js.map → benchmark-command-C458yEFK.js.map} +1 -1
  7. package/dist/benchmarks/index.d.ts +3 -3
  8. package/dist/benchmarks/index.js +3 -3
  9. package/dist/campaign/index.d.ts +4 -4
  10. package/dist/campaign/index.js +8 -9
  11. package/dist/{campaign-C8iWEF0M.js → campaign-BVQVBEfH.js} +12 -326
  12. package/dist/campaign-BVQVBEfH.js.map +1 -0
  13. package/dist/{campaign-evidence-BqjlTM3P.js → campaign-evidence-DcQiYhJI.js} +10 -45
  14. package/dist/campaign-evidence-DcQiYhJI.js.map +1 -0
  15. package/dist/cli.js +2 -2
  16. package/dist/contract/index.d.ts +6 -6
  17. package/dist/contract/index.js +6 -6
  18. package/dist/{default-registry-DK_9aVJf.d.ts → default-registry-DdqA_-_o.d.ts} +3 -3
  19. package/dist/{default-registry-DK_9aVJf.d.ts.map → default-registry-DdqA_-_o.d.ts.map} +1 -1
  20. package/dist/{define-agent-eval-BE43YTm4.js → define-agent-eval-DQuos3Aw.js} +8 -9
  21. package/dist/define-agent-eval-DQuos3Aw.js.map +1 -0
  22. package/dist/{define-agent-eval-P28y7mI1.d.ts → define-agent-eval-DcludNJS.d.ts} +13 -22
  23. package/dist/{define-agent-eval-P28y7mI1.d.ts.map → define-agent-eval-DcludNJS.d.ts.map} +1 -1
  24. package/dist/{dspy-rlm-engine-MfpwoEaj.js → dspy-rlm-engine-BX34vPVY.js} +2 -2
  25. package/dist/{dspy-rlm-engine-MfpwoEaj.js.map → dspy-rlm-engine-BX34vPVY.js.map} +1 -1
  26. package/dist/{engine-vNiM02Ub.d.ts → engine-C1DIAEpn.d.ts} +2 -2
  27. package/dist/{engine-vNiM02Ub.d.ts.map → engine-C1DIAEpn.d.ts.map} +1 -1
  28. package/dist/evaluation.d.ts +1 -1
  29. package/dist/experiment/index.d.ts +1 -1
  30. package/dist/experiment/index.js +1 -1
  31. package/dist/{external-optimizer-process-xFTliuiK.js → external-optimizer-process-D-Sab4v3.js} +3 -62
  32. package/dist/external-optimizer-process-D-Sab4v3.js.map +1 -0
  33. package/dist/{external-optimizer-subprocess-B234V9V0.js → external-optimizer-subprocess-kbqC-iWf.js} +64 -3
  34. package/dist/external-optimizer-subprocess-kbqC-iWf.js.map +1 -0
  35. package/dist/hosted/index.js +1 -1
  36. package/dist/{index-I4UJGRno.d.ts → index-Bj96W0jZ.d.ts} +28 -2
  37. package/dist/{index-I4UJGRno.d.ts.map → index-Bj96W0jZ.d.ts.map} +1 -1
  38. package/dist/{index-D-TN7ZRM.d.ts → index-DBoXFaz0.d.ts} +3 -3
  39. package/dist/{index-D-TN7ZRM.d.ts.map → index-DBoXFaz0.d.ts.map} +1 -1
  40. package/dist/{index-DH-MWH5f.d.ts → index-DNVzu9Vv.d.ts} +6 -6
  41. package/dist/{index-DH-MWH5f.d.ts.map → index-DNVzu9Vv.d.ts.map} +1 -1
  42. package/dist/{index-CzfnASj_.d.ts → index-DwQxdWve.d.ts} +338 -129
  43. package/dist/index-DwQxdWve.d.ts.map +1 -0
  44. package/dist/index.d.ts +9 -9
  45. package/dist/index.js +6 -6
  46. package/dist/jev.d.ts +1 -1
  47. package/dist/{journal-c3qY0SDP.js → journal-DvKCvjty.js} +94 -2
  48. package/dist/{journal-c3qY0SDP.js.map → journal-DvKCvjty.js.map} +1 -1
  49. package/dist/ledger-core/index.d.ts +2 -2
  50. package/dist/ledger-core/index.js +2 -2
  51. package/dist/{llm-judge-NJOHM0iA.js → llm-judge-Z6MkTkd0.js} +1819 -609
  52. package/dist/llm-judge-Z6MkTkd0.js.map +1 -0
  53. package/dist/{matrix-CgqdFm5N.d.ts → matrix-CQMMrEQ5.d.ts} +2 -2
  54. package/dist/{matrix-CgqdFm5N.d.ts.map → matrix-CQMMrEQ5.d.ts.map} +1 -1
  55. package/dist/multishot/golden/index.d.ts +1 -1
  56. package/dist/multishot/index.d.ts +2 -2
  57. package/dist/openapi.json +1 -1
  58. package/dist/{produced-state-CY-L7CqS.js → produced-state-CVmCPQDJ.js} +3 -3
  59. package/dist/{produced-state-CY-L7CqS.js.map → produced-state-CVmCPQDJ.js.map} +1 -1
  60. package/dist/rl.d.ts +1 -1
  61. package/dist/{search-ledger-CQ8Uy4FK.js → search-ledger-Ck7tok0h.js} +78 -12
  62. package/dist/search-ledger-Ck7tok0h.js.map +1 -0
  63. package/dist/{search-ledger-file-EvdDZRhv.js → search-ledger-file-DY0AXX3W.js} +2 -2
  64. package/dist/{search-ledger-file-EvdDZRhv.js.map → search-ledger-file-DY0AXX3W.js.map} +1 -1
  65. package/dist/{search-shipper-C55V7HE3.js → search-shipper-BG182p-U.js} +3 -3
  66. package/dist/{search-shipper-C55V7HE3.js.map → search-shipper-BG182p-U.js.map} +1 -1
  67. package/dist/{semantic-concept-judge-Bt7bXAAo.js → semantic-concept-judge-Bg-560H0.js} +3 -3
  68. package/dist/{semantic-concept-judge-Bt7bXAAo.js.map → semantic-concept-judge-Bg-560H0.js.map} +1 -1
  69. package/dist/{skillopt-optimization-method-DwO3TOAV.js → skillopt-optimization-method-DoayuM6D.js} +6 -6
  70. package/dist/{skillopt-optimization-method-DwO3TOAV.js.map → skillopt-optimization-method-DoayuM6D.js.map} +1 -1
  71. package/dist/{statistical-heldout-BED3wBtX.d.ts → statistical-heldout-C1Ej5Ce8.d.ts} +2 -2
  72. package/dist/{statistical-heldout-BED3wBtX.d.ts.map → statistical-heldout-C1Ej5Ce8.d.ts.map} +1 -1
  73. package/dist/{store-tool-spans-Cfh9XUCm.d.ts → store-tool-spans-ByLdU26j.d.ts} +3 -3
  74. package/dist/{store-tool-spans-Cfh9XUCm.d.ts.map → store-tool-spans-ByLdU26j.d.ts.map} +1 -1
  75. package/dist/{tool-groups-D-mkeN0N.d.ts → tool-groups-DiBi16ns.d.ts} +2 -2
  76. package/dist/tool-groups-DiBi16ns.d.ts.map +1 -0
  77. package/dist/trace-repair/index.d.ts +1 -1
  78. package/dist/traces.d.ts +2 -2
  79. package/dist/{types-ClK-sHs8.d.ts → types-ByZslIlr.d.ts} +2 -2
  80. package/dist/{types-ClK-sHs8.d.ts.map → types-ByZslIlr.d.ts.map} +1 -1
  81. package/dist/{types-BmQRs661.d.ts → types-C1CAD70_.d.ts} +14 -15
  82. package/dist/types-C1CAD70_.d.ts.map +1 -0
  83. package/docs/campaign-proposers.md +14 -16
  84. package/docs/search-ledger.md +62 -1
  85. package/package.json +1 -1
  86. package/dist/campaign-C8iWEF0M.js.map +0 -1
  87. package/dist/campaign-evidence-BqjlTM3P.js.map +0 -1
  88. package/dist/define-agent-eval-BE43YTm4.js.map +0 -1
  89. package/dist/external-optimizer-process-xFTliuiK.js.map +0 -1
  90. package/dist/external-optimizer-subprocess-B234V9V0.js.map +0 -1
  91. package/dist/index-CzfnASj_.d.ts.map +0 -1
  92. package/dist/llm-judge-NJOHM0iA.js.map +0 -1
  93. package/dist/search-ledger-CQ8Uy4FK.js.map +0 -1
  94. package/dist/tool-groups-D-mkeN0N.d.ts.map +0 -1
  95. package/dist/types-BmQRs661.d.ts.map +0 -1
@@ -1,6 +1,6 @@
1
1
  import { a as RunRecord } from "./run-record-DF-fuu31.js";
2
2
  import { a as PairedPromotionDecision, d as PairedBootstrapResult, i as PairedMcNemarEvidence, r as PairedDecisionStatistic, t as PairedDecisionMethod, u as PairedBootstrapOptions } from "./paired-promotion-decision-DPKz8ZnF.js";
3
- import { R as Scenario, a as CampaignResult, b as GenerationRecord, h as GateContext, j as MutableSurface, p as Gate, v as GateResult, w as JudgeScore } from "./types-BmQRs661.js";
3
+ import { R as Scenario, a as CampaignResult, b as GenerationRecord, h as GateContext, j as MutableSurface, p as Gate, v as GateResult, w as JudgeScore } from "./types-C1CAD70_.js";
4
4
  import { r as LedgerHash } from "./canonical-CFpojCN5.js";
5
5
  import { z } from "zod";
6
6
  //#region src/statistics/paired-binary.d.ts
@@ -1306,4 +1306,4 @@ declare function dimensionRegressions(candidate: Map<string, Record<string, Judg
1306
1306
  }): DimensionRegression[];
1307
1307
  //#endregion
1308
1308
  export { paretoSignificanceGate as $, PairArmsOptions as A, FinalEvidenceReservation as At, PowerPreflight as B, RiskDifferenceResult as Bt, hashJson as C, verifyAttestation as Ct, ComparePairedArmsOptions as D, FinalEvidenceMeasurement as Dt, verifyManifest as E, FinalEvidenceLedger as Et, PairedCorrectness as F, EProcessStep as Ft, BuildEvidenceVectorOptions as G, pairedRiskDifference as Gt, powerPreflight as H, isBinaryOutcomeVector as Ht, PairedMetricDelta as I, eProcess as It, ParetoSignificanceGateOptions as J, passAtK as Jt, EvidenceVector as K, pairedRiskDifferenceExact as Kt, comparePairedArms as L, ExactRiskDifferenceResult as Lt, PairRunRecordsResult as M, EProcess as Mt, PairedArmRow as N, EProcessOptions as Nt, MatchedPair as O, FinalEvidenceOutcome as Ot, PairedArmsComparison as P, EProcessState as Pt, paretoPolicy as Q, pairArms as R, McNemarResult as Rt, evaluateHypothesis as S, attest as St, signManifest as T, FinalEvidenceError as Tt, AxisEvidence as U, mcnemar as Ut, PowerPreflightOptions as V, ScoreRiskDifferenceResult as Vt, AxisVerdict as W, pairedBinaryScale as Wt, PromotionPolicy as X, PromotionObjective as Y, wilson as Yt, buildEvidenceVector as Z, sequentialPairedGate as _, verifyEvidenceReceipt as _t, detectScale as a, createCampaignEvidenceReceipt as at, SignedManifest as b, AttestationVerification as bt, pairHoldout as c, EVIDENCE_RECEIPT_VERSION as ct, SequentialDecision as d, EvidenceBinding as dt, Objective as et, SequentialObservation as f, EvidenceReceipt as ft, sequentialDecide as g, isIndependentEvidence as gt, SequentialStreamState as h, createEvidenceReceipt as ht, PairedHoldout as i, CampaignEvidenceContext as it, PairArmsResult as j, openFinalEvidenceLedger as jt, MatchedRunRecordPair as k, FinalEvidenceRecord as kt, SequentialDecideFn as l, EvidenceAuthority as lt, SequentialPairedGateOptions as m, INDEPENDENT_EVIDENCE_AUTHORITY_KINDS as mt, HeldoutSignificance as n, dominates as nt, dimensionRegressions as o, CreateEvidenceReceiptInput as ot, SequentialPairedGate as p, EvidenceReceiptVerification as pt, ObjectiveSource as q, pairedRiskDifferenceScore as qt, HeldoutSignificanceOptions as r, paretoFrontier as rt, heldoutSignificance as s, EVIDENCE_AUTHORITY_KINDS as st, DimensionRegression as t, ParetoResult as tt, SequentialDecideOptions as u, EvidenceAuthorityKind as ut, HypothesisManifest as v, ATTESTATION_ALGORITHM as vt, manifestContentDigest as w, FinalEvidenceConflictError as wt, SignedManifestAlgo as x, AttestedReport as xt, HypothesisResult as y, AttestationProvenance as yt, pairRunRecords as z, ProportionInterval as zt };
1309
- //# sourceMappingURL=statistical-heldout-BED3wBtX.d.ts.map
1309
+ //# sourceMappingURL=statistical-heldout-C1Ej5Ce8.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"statistical-heldout-BED3wBtX.d.ts","names":[],"sources":["../src/statistics/paired-binary.ts","../src/statistics/sequential-eprocess.ts","../src/experiment/final-evidence.ts","../src/attestation.ts","../src/experiment/evidence-receipt.ts","../src/experiment/campaign-evidence.ts","../src/pareto.ts","../src/campaign/gates/promotion-policy.ts","../src/campaign/gates/power-preflight.ts","../src/paired-arms.ts","../src/pre-registration.ts","../src/campaign/gates/sequential.ts","../src/campaign/gates/statistical-heldout.ts"],"mappings":";;;;;;;UAgBiB;;EAEf;;EAEA;;EAEA;;;;;;;;;iBAUc,OAAO,mBAAmB,WAAW,sBAAoB;;;;;;;;;;;;;;;;;;;;;;;;;;iBA2CzD,sBAAsB,QAAQ;;UAU7B;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;iBAec,QACd,SAAS,6BACT,WAAW,8BACV;;UAmBc;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;;;iBAkBc,qBACd,SAAS,6BACT,WAAW,6BACX,sBACC;;;;UAkCc;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA+Bc,0BACd,SAAS,6BACT,WAAW,6BACX,sBACC;;;UAyEc;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA8Dc,0BACd,SAAS,6BACT,WAAW,6BACX,sBACC;;;;;;;;;;;;;;;;;;;;iBAiFa,kBACd,QAAQ,mBACR,OAAO;;;;;;;;;;;iBA0BO,QAAQ,WAAW,WAAW;;;UCrgB7B;;;EAGf;;;;EAIA;;;;EAIA;;;;;;;;EAQA,SAAS;;UAGM;;EAEf;;EAEA;;EAEA;;;;;;;;;;UAWe,sBAAsB;EACrC;EACA;EACA;;EAEA;;EAEA;;EAEA;;;EAGA;;UAGe;;;EAGf,OAAO,YAAY;EACnB,SAAS;;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA8BK,SAAS,OAAM,kBAAuB;;;cCnFhD,mBAAiB,EAAA;;;;;;GAQZ,EAAA,KAAA;;KAGC,2BAA2B,EAAE,aAAa;cAEhD,mBAAiB,EAAA;;;GAKZ,EAAA,KAAA;KAEC,2BAA2B,EAAE,aAAa;UAgCrC;EACf,aAAa;EACb,iBAAiB;EACjB;IAAY,aAAa;IAA0B,WAAW;;;KAGpD,qBAAqB;EAC3B;EAAiB,OAAO;;EACxB;EAAkB;IAAS;IAA8C;;;UAE9D;EACf,QACE,OAAO,2BACN,QAAQ;IAAuB,QAAQ;IAAqB;;EAC/D,OACE,mBACA,aAAa,2BACZ,QAAQ;IAAuB,QAAQ;IAAqB;;EAC/D,QAAQ,QAAQ,qBAAqB;;;cAI1B,2BAA2B;WAE3B;EADX,YACW,8CACT;;;cAQS,mCAAmC;EAC9C,YAAY;;;iBA+HE,wBAAwB;EAAW;IAAiB;;;;;;;;;;;;;;;;;;;;;;;cCvMvD;UAEI;;EAEf,eAAe;;EAEf;;;EAGA;;EAEA;;EAEA;;;EAGA;;UAGe;;EAEf;EACA,YAAY;EACZ,kBAAkB;;EAElB;;UAGe;EACf;;EAEA;;;;;;;;iBAiBc,OAAO,iBAAiB,YAAY,wBAAwB;;;;;;;;;;iBAoB5D,kBACd,iBACA,UAAU,iBACT;;;cC1EU;;cAGA;KAQD,gCAAgC;cAE/B;UAOI;WACN,MAAM;;WAEN;;UAGM;WACN,sBAAsB;;WAEtB;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;WACA,WAAW;;WAEX;;WAEA;;UAGM;WACN,SAAS;WACT,aAAa;;UAGP;WACN;WACA;;UAGM,mCAAmC,KAAK;;;;;;;iBAQzC,sBACd,OAAO,4BACP,YAAY,wBACX;;;;;iBA4Ba,sBAAsB,SAAS,kBAAkB;;;;;;iBAkCjD,sBAAsB,SAAS;;;;KCjJnC,0BAA0B,KACpC;EAOI,YAAY,KAAK;;;iBAGP,8BAA8B,UAAU,UAAU,GAAG;EACnE,UAAU,eAAe,GAAG;EAC5B,SAAS;EACT,SAAS;IAAA;;;;;;;;;;;;;;;;;;KCPC;UAEK,UAAU;;EAEzB;EACA,WAAW;EACX,QAAQ,WAAW;;UAGJ,aAAa;EAC5B,UAAU;EACV,WAAW;;EAEX,cAAc;IAAQ,WAAW;IAAG,WAAW;;;;iBAIjC,UAAU,GAAG,GAAG,GAAG,GAAG,GAAG,YAAY,UAAU;;;;;;iBAmB/C,eAAe,GAAG,YAAY,KAAK,YAAY,UAAU,OAAO,aAAa;;;;;KCxBjF;EAAoB;;EAAwB;EAAmB;;UAE1D;;EAEf;EACA,QAAQ;;;EAGR,WAAW;;;;EAIX;;;;EAIA;;;;;EAKA;;;;KAKU;UAEK;EACf;EACA,QAAQ;EACR,WAAW;;;;;;;;EAQX,WAAW;;;;;EAKX;;;EAGA;IAAM;IAAa;;;EAEnB,mBAAmB;;EAEnB,SAAS;;;EAGT;;EAEA;EACA;EACA,gBAAgB;EAChB;EACA;EACA,SAAS;;UAGM;;EAEf,MAAM;;;EAGN;;;EAGA;IAAQ;IAAmB;;;;;;KAMjB,mBAAmB,IAAI,mBAAmB;UAErC;;;;EAIf;;EAEA;;EAEA;;EAEA;;;EAGA;;;;;;;;iBASc,oBAAoB,WAAW,kBAAkB,UAC/D,KAAK,YAAY,WAAW,YAC5B,YAAY,sBACZ,OAAM,6BACL;;;;;;cAkIU,cAAc;UA+EV,sCAAsC;;EAErD,YAAY;;;EAGZ,SAAS;;EAET;;;;;;;iBAQc,uBAAuB,qBAAqB,kBAAkB,WAAW,UACvF,SAAS,gCACR,KAAK,WAAW;;;;;;;;;;;;;;;;;UCzVF;;EAEf;;;EAGA;;EAEA;;EAEA;;;;EAIA;;UAGe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;EAGA;EACA;EACA;;EAEA;;EAEA;;;iBAYc,eAAe,MAAM,wBAAwB;;;;;UCzB5C;;;EAGf;;;;;EAKA;;EAEA;;EAEA;;EAEA,UAAU;;UAGK;;EAEf;;EAEA;;;UAIe;EACf;;;;EAIA;EACA,UAAU;EACV,WAAW;;UAGI;;EAEf,OAAO;;;EAGP,kBAAkB;;EAElB,mBAAmB;;;;;;;;;;;;;;;;;;;;iBAqBL,SAAS,eAAe,gBAAgB,MAAM,kBAAkB;;UA8G/D;;EAEf;;EAEA;;EAEA,SAAS;;EAET,gBAAgB;;;UAID;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;;;EAIA,aAAa;;EAEb;IAAY;IAAW;;;UAGR,iCAAiC;;;;EAIhD;;EAEA,YAAY;;UAGG;EACf;EACA;EACA;;;EAGA,aAAa;EACb,cAAc;;;;;;;;;;;;;;;iBAgBA,kBACd,eAAe,gBACf,MAAM,2BACL;UAmEc;EACf;EACA;EACA,UAAU;EACV,WAAW;;UAGI;EACf,OAAO;EACP,kBAAkB;EAClB,mBAAmB;;;;;;;;;iBAeL,eACd,uBAAuB,aACvB,wBAAwB,cACvB;;;;;;;;;;;;;;;;UCrWc;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;EACA;;;KAIU;UAEK,uBAAuB;;EAEtC;;EAEA,MAAM;;UAGS;EACf,UAAU;EACV;EACA;EACA;;;EAGA;;EAEA,kBAAkB;EAGlB;;;;;;;;;;;;;;;;;iBAkBoB,SAAS,GAAG,KAAK,IAAI;;;;;iBAQ3B,sBAAsB,UAAU;;;;;;iBAc1B,aAAa,GAAG,qBAAqB,QAAQ;;;;;iBAS7C,eAAe,GAAG,iBAAiB;;;;;iBAYnC,mBACpB,UAAU,gBACV;EAAY;EAAW;EAAgB;IACtC,QAAQ;;;KCrFC;UAEK;EACf,UAAU;;EAEV;;EAEA;;;EAGA;;UAGe;;;EAGf;;;;EAIA;;;EAGA;;EAEA;;;;EAIA;;;EAGA;;;;EAIA,8BAA8B;;;EAG9B,kBAAkB;;EAElB;;;;;;;EAOA,SAAS;;;;;;KAOC,wBAAwB;EAAkB,UAAU;;UAE/C,qBAAqB,qBAAqB,kBAAkB,WAAW,kBAC9E,KAAK,WAAW;;;;;;;;;;EAUxB,QAAQ,gBAAgB;;;EAGxB,SAAS;;;;;;;;;;;;;;;;;iBAuPK,qBAAqB,qBAAqB,kBAAkB,WAAW,UACrF,SAAS,8BACR,qBAAqB,WAAW;UA6GlB;;EAEf;;EAEA;;EAEA;;EAEA;;UAGe;GACd;IAAQ,SAAS;;IAAyB;IAAe;;;EAE1D,SAAS;;;;;;;;;;;;;;;;;;;;;;;;;iBA0BK,iBAAiB,UAAS,0BAA+B;;;UC9dxD;;EAEf;;EAEA;;EAEA;;;;;;;;;;;;;iBA4Ec,YACd,WAAW,YAAY,eAAe,cACtC,UAAU,YAAY,eAAe,cACrC,aAAa,aACb,SAAS,GAAG,oCACX;UAkEc;EACf,QAAQ;;;;;;;;;;;;EAYR,WAAW;;;;EAIX,iBAAiB;;;;;;;EAOjB,UAAU;;EAEV,mBAAmB;;EAEnB,SAAS;;;;EAIT;;EAEA;;EAEA;EACA;;EAEA;;EAEA;;EAEA,gBAAgB;;EAEhB;;;;EAIA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;;EAEA;EACA;;EAEA,8BAA8B;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA6BhB,oBACd,QAAQ,eACR,OAAM,6BACL;UA2Ec;EACf;;;EAGA,WAAW;;;;EAIX;;EAEA;IAAM;IAAa;;;EAEnB,mBAAmB;;EAEnB,SAAS;;EAET;;;;EAIA;EACA;EACA;EACA;EACA;;EAEA;EACA;;EAEA;;EAEA;;;;;iBAMc,YAAY;;;;;;;;;iBAYZ,qBACd,WAAW,YAAY,eAAe,cACtC,UAAU,YAAY,eAAe,cACrC,aAAa,aACb,8BACA;EACE;EACA;EACA;EACA;;;EAGA;EACA,8BAA8B;EAC9B;IAED"}
1
+ {"version":3,"file":"statistical-heldout-C1Ej5Ce8.d.ts","names":[],"sources":["../src/statistics/paired-binary.ts","../src/statistics/sequential-eprocess.ts","../src/experiment/final-evidence.ts","../src/attestation.ts","../src/experiment/evidence-receipt.ts","../src/experiment/campaign-evidence.ts","../src/pareto.ts","../src/campaign/gates/promotion-policy.ts","../src/campaign/gates/power-preflight.ts","../src/paired-arms.ts","../src/pre-registration.ts","../src/campaign/gates/sequential.ts","../src/campaign/gates/statistical-heldout.ts"],"mappings":";;;;;;;UAgBiB;;EAEf;;EAEA;;EAEA;;;;;;;;;iBAUc,OAAO,mBAAmB,WAAW,sBAAoB;;;;;;;;;;;;;;;;;;;;;;;;;;iBA2CzD,sBAAsB,QAAQ;;UAU7B;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;iBAec,QACd,SAAS,6BACT,WAAW,8BACV;;UAmBc;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;;;iBAkBc,qBACd,SAAS,6BACT,WAAW,6BACX,sBACC;;;;UAkCc;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA+Bc,0BACd,SAAS,6BACT,WAAW,6BACX,sBACC;;;UAyEc;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA8Dc,0BACd,SAAS,6BACT,WAAW,6BACX,sBACC;;;;;;;;;;;;;;;;;;;;iBAiFa,kBACd,QAAQ,mBACR,OAAO;;;;;;;;;;;iBA0BO,QAAQ,WAAW,WAAW;;;UCrgB7B;;;EAGf;;;;EAIA;;;;EAIA;;;;;;;;EAQA,SAAS;;UAGM;;EAEf;;EAEA;;EAEA;;;;;;;;;;UAWe,sBAAsB;EACrC;EACA;EACA;;EAEA;;EAEA;;EAEA;;;EAGA;;UAGe;;;EAGf,OAAO,YAAY;EACnB,SAAS;;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA8BK,SAAS,OAAM,kBAAuB;;;cCnFhD,mBAAiB,EAAA;;;;;;GAQZ,EAAA,KAAA;;KAGC,2BAA2B,EAAE,aAAa;cAEhD,mBAAiB,EAAA;;;GAKZ,EAAA,KAAA;KAEC,2BAA2B,EAAE,aAAa;UAgCrC;EACf,aAAa;EACb,iBAAiB;EACjB;IAAY,aAAa;IAA0B,WAAW;;;KAGpD,qBAAqB;EAC3B;EAAiB,OAAO;;EACxB;EAAkB;IAAS;IAA8C;;;UAE9D;EACf,QACE,OAAO,2BACN,QAAQ;IAAuB,QAAQ;IAAqB;;EAC/D,OACE,mBACA,aAAa,2BACZ,QAAQ;IAAuB,QAAQ;IAAqB;;EAC/D,QAAQ,QAAQ,qBAAqB;;;cAI1B,2BAA2B;WAE3B;EADX,YACW,8CACT;;;cAQS,mCAAmC;EAC9C,YAAY;;;iBA+HE,wBAAwB;EAAW;IAAiB;;;;;;;;;;;;;;;;;;;;;;;cCvMvD;UAEI;;EAEf,eAAe;;EAEf;;;EAGA;;EAEA;;EAEA;;;EAGA;;UAGe;;EAEf;EACA,YAAY;EACZ,kBAAkB;;EAElB;;UAGe;EACf;;EAEA;;;;;;;;iBAiBc,OAAO,iBAAiB,YAAY,wBAAwB;;;;;;;;;;iBAoB5D,kBACd,iBACA,UAAU,iBACT;;;cC1EU;;cAGA;KAQD,gCAAgC;cAE/B;UAOI;WACN,MAAM;;WAEN;;UAGM;WACN,sBAAsB;;WAEtB;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;WACA,WAAW;;WAEX;;WAEA;;UAGM;WACN,SAAS;WACT,aAAa;;UAGP;WACN;WACA;;UAGM,mCAAmC,KAAK;;;;;;;iBAQzC,sBACd,OAAO,4BACP,YAAY,wBACX;;;;;iBA4Ba,sBAAsB,SAAS,kBAAkB;;;;;;iBAkCjD,sBAAsB,SAAS;;;;KCjJnC,0BAA0B,KACpC;EAOI,YAAY,KAAK;;;iBAGP,8BAA8B,UAAU,UAAU,GAAG;EACnE,UAAU,eAAe,GAAG;EAC5B,SAAS;EACT,SAAS;IAAA;;;;;;;;;;;;;;;;;;KCPC;UAEK,UAAU;;EAEzB;EACA,WAAW;EACX,QAAQ,WAAW;;UAGJ,aAAa;EAC5B,UAAU;EACV,WAAW;;EAEX,cAAc;IAAQ,WAAW;IAAG,WAAW;;;;iBAIjC,UAAU,GAAG,GAAG,GAAG,GAAG,GAAG,YAAY,UAAU;;;;;;iBAmB/C,eAAe,GAAG,YAAY,KAAK,YAAY,UAAU,OAAO,aAAa;;;;;KCxBjF;EAAoB;;EAAwB;EAAmB;;UAE1D;;EAEf;EACA,QAAQ;;;EAGR,WAAW;;;;EAIX;;;;EAIA;;;;;EAKA;;;;KAKU;UAEK;EACf;EACA,QAAQ;EACR,WAAW;;;;;;;;EAQX,WAAW;;;;;EAKX;;;EAGA;IAAM;IAAa;;;EAEnB,mBAAmB;;EAEnB,SAAS;;;EAGT;;EAEA;EACA;EACA,gBAAgB;EAChB;EACA;EACA,SAAS;;UAGM;;EAEf,MAAM;;;EAGN;;;EAGA;IAAQ;IAAmB;;;;;;KAMjB,mBAAmB,IAAI,mBAAmB;UAErC;;;;EAIf;;EAEA;;EAEA;;EAEA;;;EAGA;;;;;;;;iBASc,oBAAoB,WAAW,kBAAkB,UAC/D,KAAK,YAAY,WAAW,YAC5B,YAAY,sBACZ,OAAM,6BACL;;;;;;cAkIU,cAAc;UA+EV,sCAAsC;;EAErD,YAAY;;;EAGZ,SAAS;;EAET;;;;;;;iBAQc,uBAAuB,qBAAqB,kBAAkB,WAAW,UACvF,SAAS,gCACR,KAAK,WAAW;;;;;;;;;;;;;;;;;UCzVF;;EAEf;;;EAGA;;EAEA;;EAEA;;;;EAIA;;UAGe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;EAGA;EACA;EACA;;EAEA;;EAEA;;;iBAYc,eAAe,MAAM,wBAAwB;;;;;UCzB5C;;;EAGf;;;;;EAKA;;EAEA;;EAEA;;EAEA,UAAU;;UAGK;;EAEf;;EAEA;;;UAIe;EACf;;;;EAIA;EACA,UAAU;EACV,WAAW;;UAGI;;EAEf,OAAO;;;EAGP,kBAAkB;;EAElB,mBAAmB;;;;;;;;;;;;;;;;;;;;iBAqBL,SAAS,eAAe,gBAAgB,MAAM,kBAAkB;;UA8G/D;;EAEf;;EAEA;;EAEA,SAAS;;EAET,gBAAgB;;;UAID;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;;;EAIA,aAAa;;EAEb;IAAY;IAAW;;;UAGR,iCAAiC;;;;EAIhD;;EAEA,YAAY;;UAGG;EACf;EACA;EACA;;;EAGA,aAAa;EACb,cAAc;;;;;;;;;;;;;;;iBAgBA,kBACd,eAAe,gBACf,MAAM,2BACL;UAmEc;EACf;EACA;EACA,UAAU;EACV,WAAW;;UAGI;EACf,OAAO;EACP,kBAAkB;EAClB,mBAAmB;;;;;;;;;iBAeL,eACd,uBAAuB,aACvB,wBAAwB,cACvB;;;;;;;;;;;;;;;;UCrWc;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;EACA;;;KAIU;UAEK,uBAAuB;;EAEtC;;EAEA,MAAM;;UAGS;EACf,UAAU;EACV;EACA;EACA;;;EAGA;;EAEA,kBAAkB;EAGlB;;;;;;;;;;;;;;;;;iBAkBoB,SAAS,GAAG,KAAK,IAAI;;;;;iBAQ3B,sBAAsB,UAAU;;;;;;iBAc1B,aAAa,GAAG,qBAAqB,QAAQ;;;;;iBAS7C,eAAe,GAAG,iBAAiB;;;;;iBAYnC,mBACpB,UAAU,gBACV;EAAY;EAAW;EAAgB;IACtC,QAAQ;;;KCrFC;UAEK;EACf,UAAU;;EAEV;;EAEA;;;EAGA;;UAGe;;;EAGf;;;;EAIA;;;EAGA;;EAEA;;;;EAIA;;;EAGA;;;;EAIA,8BAA8B;;;EAG9B,kBAAkB;;EAElB;;;;;;;EAOA,SAAS;;;;;;KAOC,wBAAwB;EAAkB,UAAU;;UAE/C,qBAAqB,qBAAqB,kBAAkB,WAAW,kBAC9E,KAAK,WAAW;;;;;;;;;;EAUxB,QAAQ,gBAAgB;;;EAGxB,SAAS;;;;;;;;;;;;;;;;;iBAuPK,qBAAqB,qBAAqB,kBAAkB,WAAW,UACrF,SAAS,8BACR,qBAAqB,WAAW;UA6GlB;;EAEf;;EAEA;;EAEA;;EAEA;;UAGe;GACd;IAAQ,SAAS;;IAAyB;IAAe;;;EAE1D,SAAS;;;;;;;;;;;;;;;;;;;;;;;;;iBA0BK,iBAAiB,UAAS,0BAA+B;;;UC9dxD;;EAEf;;EAEA;;EAEA;;;;;;;;;;;;;iBA4Ec,YACd,WAAW,YAAY,eAAe,cACtC,UAAU,YAAY,eAAe,cACrC,aAAa,aACb,SAAS,GAAG,oCACX;UAkEc;EACf,QAAQ;;;;;;;;;;;;EAYR,WAAW;;;;EAIX,iBAAiB;;;;;;;EAOjB,UAAU;;EAEV,mBAAmB;;EAEnB,SAAS;;;;EAIT;;EAEA;;EAEA;EACA;;EAEA;;EAEA;;EAEA,gBAAgB;;EAEhB;;;;EAIA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;;EAEA;EACA;;EAEA,8BAA8B;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA6BhB,oBACd,QAAQ,eACR,OAAM,6BACL;UA2Ec;EACf;;;EAGA,WAAW;;;;EAIX;;EAEA;IAAM;IAAa;;;EAEnB,mBAAmB;;EAEnB,SAAS;;EAET;;;;EAIA;EACA;EACA;EACA;EACA;;EAEA;EACA;;EAEA;;EAEA;;;;;iBAMc,YAAY;;;;;;;;;iBAYZ,qBACd,WAAW,YAAY,eAAe,cACtC,UAAU,YAAY,eAAe,cACrC,aAAa,aACb,8BACA;EACE;EACA;EACA;EACA;;;EAGA;EACA,8BAA8B;EAC9B;IAED"}
@@ -6,8 +6,8 @@ import { E as SpanSourceReader, F as SearchTraceResult, K as ViewTraceResult, N
6
6
  import { et as ProviderRedactor, rt as RawProviderSink } from "./types-C0bWoSKC.js";
7
7
  import { i as SearchArtifactRef } from "./search-ledger-types-C9anU2DX.js";
8
8
  import { s as TraceStore } from "./store-C4avFjIi.js";
9
- import { a as TraceAnalystLimits, i as TraceAnalysisEngineResult, n as TraceAnalysisEngine } from "./engine-vNiM02Ub.js";
10
- import { t as TraceToolGroupName } from "./tool-groups-D-mkeN0N.js";
9
+ import { a as TraceAnalystLimits, i as TraceAnalysisEngineResult, n as TraceAnalysisEngine } from "./engine-C1DIAEpn.js";
10
+ import { t as TraceToolGroupName } from "./tool-groups-DiBi16ns.js";
11
11
  //#region src/trace/extract-usage.d.ts
12
12
  type ExtractedUsage = RunTokenUsage;
13
13
  type SseUsageMode = 'cumulative' | 'delta';
@@ -407,4 +407,4 @@ declare class ToolTraceMissingError extends CaptureIntegrityError {
407
407
  declare function toolSpansToTraceAnalysisStore(spans: readonly ToolSpan[] | null | undefined, options?: ToolSpansToTraceAnalysisStoreOptions): TraceAnalysisStore;
408
408
  //#endregion
409
409
  export { AnalyzeTracesInput as A, CaptureFetchContext as B, inferDomainKeywords as C, planTraceInsightQuestions as D, isTraceInsightEvidence as E, OTEL_AGENT_EVAL_SCOPE as F, SseUsageMode as G, captureFetchToRawSink as H, OtlpExport as I, extractUsageFromSse as J, extractUsage as K, OtlpResourceSpans as L, AnalyzeTracesResult as M, analyzeTraces as N, scoreTraceInsightReadiness as O, OtlpFlatLine as P, OtlpSpan as R, domainEvidencePattern as S, isTraceInsightAttribution as T, ExtractUsageFromSseOptions as U, CaptureFetchOptions as V, ExtractedUsage as W, TraceInsightTask as _, ToolSpansToTraceAnalysisStoreOptions as a, defaultTraceInsightPanel as b, TraceInsightContext as c, TraceInsightPanelRole as d, TraceInsightPromptInput as f, TraceInsightSuite as g, TraceInsightReadiness as h, OtlpFileTraceStoreOptions as i, AnalyzeTracesOptions as j, tokenizeDomainWords as k, TraceInsightEvidence as l, TraceInsightQuestion as m, toolSpansToTraceAnalysisStore as n, otlpTextToTraceAnalysisStore as o, TraceInsightQualityGate as p, extractUsageFromResponse as q, OtlpFileTraceStore as r, TraceInsightAttribution as s, ToolTraceMissingError as t, TraceInsightFinding as u, buildTraceInsightContext as v, isTraceInsightArtifact as w, describeTraceInsightScope as x, buildTraceInsightPrompt as y, exportRunAsOtlp as z };
410
- //# sourceMappingURL=store-tool-spans-Cfh9XUCm.d.ts.map
410
+ //# sourceMappingURL=store-tool-spans-ByLdU26j.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"store-tool-spans-Cfh9XUCm.d.ts","names":[],"sources":["../src/trace/extract-usage.ts","../src/trace/capture-fetch.ts","../src/trace/otel.ts","../src/trace/otlp-flat.ts","../src/trace-analyst/analyst.ts","../src/trace-analyst/insights.ts","../src/trace-analyst/store-otlp.ts","../src/trace-analyst/store-tool-spans.ts"],"mappings":";;;;;;;;;;;KAgBY,iBAAiB;KACjB;UAEK;;EAEf,OAAO;;;;;;;iBAuDO,aAAa,gBAAgB;;;;;;iBA6C7B,oBACd,cACA,UAAS,6BACR;;;;;;;iBA4DmB,yBACpB,UAAU,UACV,aAAa,6BACZ,QAAQ;;;UCrKM;;;EAGf;;EAEA;;;EAGA;;EAEA;;EAEA;;UAGe;;EAEf,WAAW;;;EAGX;;;EAGA;;;;;;;;EAQA,WAAW,OAAO,gBAAgB,KAAK;;EAEvC,eAAe;;iBAkDD,sBACd,cAAc,WAAW,OACzB,MAAM,iBACN,KAAK,qBACL,OAAM,6BACE,WAAW;;;cCxFR;EAA0B;EAAoC;;UAE1D;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA,YAAY;IACV;IACA;MAAS;MAAsB;MAAmB;MAAsB;;;EAE1E,SAAS;IAAQ;IAAsB;IAAc,aAAa;;EAClE;IAAW;IAAc;;;UAGV;EACf;IAAY,YAAY;;EACxB,YAAY;IAAQ,cAAc;IAAuB,OAAO;;;UAGjD;EACf,eAAe;;;iBAIK,gBACpB,OAAO,YACP,eACA,gBAAe,4CACd,QAAQ;;;KClDC;UAEK;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;IACE,MAAM;IACN;;EAEF;IAAY,YAAY;;EACxB,YAAY;EACZ,SAAS;IAAQ;IAAc;IAAuB,aAAa;;;;;UCNpD;EACf;EACA;EACA;EACA;;KAGU,sBAAsB;UAEjB;;EAEf,iBAAiB;;EAEjB,QAAQ;;EAER;;EAEA,YAAY;EACZ,SAAS,QAAQ;EACjB;EACA;EACA,aAAa;EACb;EACA,yBAAyB;EACzB,4BAA4B;EAC5B,eAAe,SAAS;EACxB,OAAO;EACP,OAAO,iBAAiB,SAAS;EACjC,SAAS;;;;;;;;iBASW,cACpB,OAAO,oBACP,SAAS,uBACR,QAAQ;;;UCnDM;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,OAAO;;UAGQ;EACf;EACA;EACA;EACA;EACA;;EAEA,cAAc;;;UAIC;EACf;IACE;IACA;IACA;IACA;IACA;IACA;IACA;;EAEF,WAAW,gGAET;EAEF,OAAO;IAAQ;IAAiB;IAAgB,UAAU;;;KAGhD,0BAA0B;EAE9B;EAAoB;EAA0B;;EAE9C;EACA;EACA,OAAO;EACP,WAAW;;EAGX;EACA;EACA;EACA,OAAO;EACP,WAAW;;EAGX;EACA;EACA,OAAO;EACP,MAAM;EACN;EACA,mBAAmB;;;iBAaX,uBAAuB,iBAAiB,SAAS;iBAoBjD,uBAAuB,iBAAiB,SAAS;iBAyCjD,0BAA0B,iBAAiB,SAAS;UA8CnD;EACf;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf,OAAO;EACP,WAAW;EACX,QAAQ;EACR,SAAS;EACT;;UAGe;EACf,OAAO;EACP;EACA;EACA,WAAW;EACX,OAAO;EACP,UAAU;EACV,OAAO;EACP,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,OAAO;;iBA0BO,oBAAoB;iBAMpB,oBAAoB,OAAO;iBAuB3B,sBAAsB,qBAAqB;iBAS3C,0BAA0B,OAAO;iBAsBjC,0BAA0B,OAAO,0BAA0B;iBAsD3D,yBAAyB,OAAO,0BAA0B;iBAa1D,2BAA2B,SAAS,sBAAsB;iBA+E1D,4BAA4B;iBA6B5B,wBAAwB,OAAO;;;UCjW9B;;EAEf,eAAe;;EAEf;;EAEA;;EAEA;;EAEA;;KAGG,gCAAgC;UAEpB,kCAAkC;;EAEjD;;;;;;;;EAQA;;uBAMa,kCAAkC;WACtC,iBAAiB;mBACT;mBACA;mBACA;mBACA;UACT;UACA;UACA;UACA;EAER,YAAY,MAAM;EA+BZ,SAAS,kBAAkB,UAAU,4BAA4B;EAOjE,SACJ;IAAQ;IAAkB;KAC1B,UAAU,4BACT;EAeG,YACJ,UAAU,qBACV,UAAU,4BACT,QAAQ;EAoDL,YACJ;IAAQ,UAAU;IAAqB;IAAe;KACtD,UAAU,4BACT,QAAQ;EAmBL,YACJ,UAAU,qBACV,UAAU,4BACT;EAMG,UACJ;IAAQ;IAAkB;KAC1B,UAAU,4BACT,QAAQ;EAuCL,UACJ;IAAQ;IAAkB;IAA6B;KACvD,UAAU,4BACT,QAAQ;EAsFL,YACJ;IAAQ;IAAkB;IAAuB;KACjD,UAAU,4BACT,QAAQ;EA4CL,WACJ;IAAQ;IAAkB;IAAiB;IAAuB;KAClE,UAAU,4BACT,QAAQ;;;EA6CL,cAAc,UAAU,4BAA4B;UAI5C;qBAmBK,WAAW,SAAS,cAAc,QAAQ;qBAE1C;UAEL;UAmBA;UAsHA;UA+DN;UAkBA;UAwDA;UAmBM;;cA+CH,2BAA2B;mBACrB;mBACA;EAEjB,YAAY,MAAM;;;YAYF,WAAW,SAAS,cAAc,QAAQ;YAkBhD;;;iBAwBI,6BACd,cACA,OAAM,uCACL;;;;cC18BU,8BAA8B;EACzC;;;;;;iBASc,8BACd,gBAAgB,+BAChB,UAAS,uCACR"}
1
+ {"version":3,"file":"store-tool-spans-ByLdU26j.d.ts","names":[],"sources":["../src/trace/extract-usage.ts","../src/trace/capture-fetch.ts","../src/trace/otel.ts","../src/trace/otlp-flat.ts","../src/trace-analyst/analyst.ts","../src/trace-analyst/insights.ts","../src/trace-analyst/store-otlp.ts","../src/trace-analyst/store-tool-spans.ts"],"mappings":";;;;;;;;;;;KAgBY,iBAAiB;KACjB;UAEK;;EAEf,OAAO;;;;;;;iBAuDO,aAAa,gBAAgB;;;;;;iBA6C7B,oBACd,cACA,UAAS,6BACR;;;;;;;iBA4DmB,yBACpB,UAAU,UACV,aAAa,6BACZ,QAAQ;;;UCrKM;;;EAGf;;EAEA;;;EAGA;;EAEA;;EAEA;;UAGe;;EAEf,WAAW;;;EAGX;;;EAGA;;;;;;;;EAQA,WAAW,OAAO,gBAAgB,KAAK;;EAEvC,eAAe;;iBAkDD,sBACd,cAAc,WAAW,OACzB,MAAM,iBACN,KAAK,qBACL,OAAM,6BACE,WAAW;;;cCxFR;EAA0B;EAAoC;;UAE1D;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA,YAAY;IACV;IACA;MAAS;MAAsB;MAAmB;MAAsB;;;EAE1E,SAAS;IAAQ;IAAsB;IAAc,aAAa;;EAClE;IAAW;IAAc;;;UAGV;EACf;IAAY,YAAY;;EACxB,YAAY;IAAQ,cAAc;IAAuB,OAAO;;;UAGjD;EACf,eAAe;;;iBAIK,gBACpB,OAAO,YACP,eACA,gBAAe,4CACd,QAAQ;;;KClDC;UAEK;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;IACE,MAAM;IACN;;EAEF;IAAY,YAAY;;EACxB,YAAY;EACZ,SAAS;IAAQ;IAAc;IAAuB,aAAa;;;;;UCNpD;EACf;EACA;EACA;EACA;;KAGU,sBAAsB;UAEjB;;EAEf,iBAAiB;;EAEjB,QAAQ;;EAER;;EAEA,YAAY;EACZ,SAAS,QAAQ;EACjB;EACA;EACA,aAAa;EACb;EACA,yBAAyB;EACzB,4BAA4B;EAC5B,eAAe,SAAS;EACxB,OAAO;EACP,OAAO,iBAAiB,SAAS;EACjC,SAAS;;;;;;;;iBASW,cACpB,OAAO,oBACP,SAAS,uBACR,QAAQ;;;UCnDM;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,OAAO;;UAGQ;EACf;EACA;EACA;EACA;EACA;;EAEA,cAAc;;;UAIC;EACf;IACE;IACA;IACA;IACA;IACA;IACA;IACA;;EAEF,WAAW,gGAET;EAEF,OAAO;IAAQ;IAAiB;IAAgB,UAAU;;;KAGhD,0BAA0B;EAE9B;EAAoB;EAA0B;;EAE9C;EACA;EACA,OAAO;EACP,WAAW;;EAGX;EACA;EACA;EACA,OAAO;EACP,WAAW;;EAGX;EACA;EACA,OAAO;EACP,MAAM;EACN;EACA,mBAAmB;;;iBAaX,uBAAuB,iBAAiB,SAAS;iBAoBjD,uBAAuB,iBAAiB,SAAS;iBAyCjD,0BAA0B,iBAAiB,SAAS;UA8CnD;EACf;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf,OAAO;EACP,WAAW;EACX,QAAQ;EACR,SAAS;EACT;;UAGe;EACf,OAAO;EACP;EACA;EACA,WAAW;EACX,OAAO;EACP,UAAU;EACV,OAAO;EACP,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,OAAO;;iBA0BO,oBAAoB;iBAMpB,oBAAoB,OAAO;iBAuB3B,sBAAsB,qBAAqB;iBAS3C,0BAA0B,OAAO;iBAsBjC,0BAA0B,OAAO,0BAA0B;iBAsD3D,yBAAyB,OAAO,0BAA0B;iBAa1D,2BAA2B,SAAS,sBAAsB;iBA+E1D,4BAA4B;iBA6B5B,wBAAwB,OAAO;;;UCjW9B;;EAEf,eAAe;;EAEf;;EAEA;;EAEA;;EAEA;;KAGG,gCAAgC;UAEpB,kCAAkC;;EAEjD;;;;;;;;EAQA;;uBAMa,kCAAkC;WACtC,iBAAiB;mBACT;mBACA;mBACA;mBACA;UACT;UACA;UACA;UACA;EAER,YAAY,MAAM;EA+BZ,SAAS,kBAAkB,UAAU,4BAA4B;EAOjE,SACJ;IAAQ;IAAkB;KAC1B,UAAU,4BACT;EAeG,YACJ,UAAU,qBACV,UAAU,4BACT,QAAQ;EAoDL,YACJ;IAAQ,UAAU;IAAqB;IAAe;KACtD,UAAU,4BACT,QAAQ;EAmBL,YACJ,UAAU,qBACV,UAAU,4BACT;EAMG,UACJ;IAAQ;IAAkB;KAC1B,UAAU,4BACT,QAAQ;EAuCL,UACJ;IAAQ;IAAkB;IAA6B;KACvD,UAAU,4BACT,QAAQ;EAsFL,YACJ;IAAQ;IAAkB;IAAuB;KACjD,UAAU,4BACT,QAAQ;EA4CL,WACJ;IAAQ;IAAkB;IAAiB;IAAuB;KAClE,UAAU,4BACT,QAAQ;;;EA6CL,cAAc,UAAU,4BAA4B;UAI5C;qBAmBK,WAAW,SAAS,cAAc,QAAQ;qBAE1C;UAEL;UAmBA;UAsHA;UA+DN;UAkBA;UAwDA;UAmBM;;cA+CH,2BAA2B;mBACrB;mBACA;EAEjB,YAAY,MAAM;;;YAYF,WAAW,SAAS,cAAc,QAAQ;YAkBhD;;;iBAwBI,6BACd,cACA,OAAM,uCACL;;;;cC18BU,8BAA8B;EACzC;;;;;;iBASc,8BACd,gBAAgB,+BAChB,UAAS,uCACR"}
@@ -1,5 +1,5 @@
1
1
  import { O as TraceAnalysisStore } from "./types-CeDfCRXp.js";
2
- import { h as TraceAnalysisToolDescriptor } from "./engine-vNiM02Ub.js";
2
+ import { h as TraceAnalysisToolDescriptor } from "./engine-C1DIAEpn.js";
3
3
  //#region src/analyst/tool-groups.d.ts
4
4
  /** Named tool sets. Kinds pass `tools: TRACE_TOOL_GROUPS.failureForensics` etc. */
5
5
  type TraceToolGroupName =
@@ -25,4 +25,4 @@ type TraceToolGroupName =
25
25
  declare function buildTraceToolsForGroup(group: TraceToolGroupName, store: TraceAnalysisStore): TraceAnalysisToolDescriptor[];
26
26
  //#endregion
27
27
  export { buildTraceToolsForGroup as n, TraceToolGroupName as t };
28
- //# sourceMappingURL=tool-groups-D-mkeN0N.d.ts.map
28
+ //# sourceMappingURL=tool-groups-DiBi16ns.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"tool-groups-DiBi16ns.d.ts","names":[],"sources":["../src/analyst/tool-groups.ts"],"mappings":";;;;KAmBY;;;;;;;;;;;;;;;;;;;;iBAyDI,wBACd,OAAO,oBACP,OAAO,qBACN"}
@@ -2,7 +2,7 @@ import { r as CaptureIntegrityError } from "../errors-DEE6u6ot.js";
2
2
  import { b as CustomTokenPricing, c as CostLedgerHandle, p as CostProvenance } from "../cost-ledger-DbQdN3nO.js";
3
3
  import { d as RunTokenUsage } from "../run-record-DF-fuu31.js";
4
4
  import { t as DefaultVerdict } from "../verdict-E4eRNf7-.js";
5
- import { a as TraceAnalystLimits, n as TraceAnalysisEngine } from "../engine-vNiM02Ub.js";
5
+ import { a as TraceAnalystLimits, n as TraceAnalysisEngine } from "../engine-C1DIAEpn.js";
6
6
  import { t as PrimeBridgeTransport } from "../prime-bridge-transport-6feEglLf.js";
7
7
  import { a as RecordedTrajectoryStep, h as isRecordedTimeout } from "../steps-Bg5-KVx1.js";
8
8
  //#region src/trace-repair/mini-swe-scaffold.d.ts
package/dist/traces.d.ts CHANGED
@@ -4,11 +4,11 @@ import { a as RunRecord, c as RunSplitTag, d as RunTokenUsage, u as RunTerminalO
4
4
  import { A as DEFAULT_TRACE_ANALYST_BUDGETS, B as TraceAnalystSpan, C as BoundedTraceAnalysisStoreOptions, D as TRACE_ANALYSIS_LIMITS, E as SpanSourceReader, F as SearchTraceResult, G as ViewTraceOversized, H as TraceAnalystSpanStatus, I as SpanMatchRecord, K as ViewTraceResult, L as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, M as ErrorCluster, N as QueryTracesPage, O as TraceAnalysisStore, P as SearchSpanResult, R as TraceAnalystByteBudgets, S as createBoundedTraceAnalysisStore, T as ReadSpanSourceResult, U as TraceAnalystTraceSummary, V as TraceAnalystSpanKind, W as ViewSpansResult, j as DatasetOverview, k as TraceAnalysisStoreContext, w as ReadSpanSourceInput, z as TraceAnalystFilters } from "./types-CeDfCRXp.js";
5
5
  import { $ as NoopRawProviderSink, Q as InMemoryRawProviderSinkOptions, X as FileSystemRawProviderSinkOptions, Y as FileSystemRawProviderSink, Z as InMemoryRawProviderSink, at as defaultProviderRedactor, et as ProviderRedactor, it as RawProviderSinkFilter, nt as RawProviderEvent, ot as providerFromBaseUrl, rt as RawProviderSink, tt as RawProviderDirection } from "./types-C0bWoSKC.js";
6
6
  import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, r as FileSystemTraceStoreOptions, s as TraceStore, t as EventFilter } from "./store-C4avFjIi.js";
7
- import { _ as buildTraceAnalysisToolDescriptors, g as UNTRUSTED_TRACE_TEXT, h as TraceAnalysisToolDescriptor, m as TRACE_ANALYST_TOOL_NAMESPACE, p as BuildTraceAnalysisToolsOptions, v as traceAnalystFunctionGroup } from "./engine-vNiM02Ub.js";
7
+ import { _ as buildTraceAnalysisToolDescriptors, g as UNTRUSTED_TRACE_TEXT, h as TraceAnalysisToolDescriptor, m as TRACE_ANALYST_TOOL_NAMESPACE, p as BuildTraceAnalysisToolsOptions, v as traceAnalystFunctionGroup } from "./engine-C1DIAEpn.js";
8
8
  import { a as SpanHandle, i as RunCompleteHookContext, n as CaptureWrite, o as TraceEmitter, r as RunCompleteHook, s as TraceEmitterOptions, t as CaptureErrorHandler } from "./emitter-HcCmFQl_.js";
9
9
  import { S as shareAllowed, _ as emptyRedactionReport, a as RedactionFinding, b as redactForShare, c as SafetyCategory, d as ShareSafetyVerdict, f as assessShareSafety, g as detectPersonalData, h as detectCredential, i as RedactOptions, l as ShareSafetyFinding, m as combineVerdicts, n as REDACTION_PROFILES, o as RedactionProfile, p as classifyKey, r as REDACTION_VERSION, s as RedactionReport, t as AssessOptions, u as ShareSafetyStatus, v as normalizeKey, x as redactText, y as redact } from "./redact-CopuMxll.js";
10
10
  import { C as TOOL_LATENCY_MS, D as asNumber, E as applyLlmSpanOtlpAttributes, O as contextInputTokens, S as TOOL_ARGS_CAPTURED, T as TOOL_NAME_ATTR_KEYS, _ as LlmSpanOtlpInput, a as LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, b as RUN_COST_ATTR_KEYS, c as LLM_COST_USD, d as LLM_MODEL_ATTR_KEYS, f as LLM_MODEL_NAME, g as LLM_REASONING_TOKEN_ATTR_KEYS, h as LLM_REASONING_TOKENS, i as LLM_CACHE_WRITE_TOKENS, k as firstNumberAttr, l as LLM_INPUT_TOKENS, m as LLM_OUTPUT_TOKEN_ATTR_KEYS, n as LLM_CACHED_TOKENS, o as LLM_CONTEXT_TOKENS, p as LLM_OUTPUT_TOKENS, r as LLM_CACHED_TOKEN_ATTR_KEYS, s as LLM_COST_ATTR_KEYS, t as INPUT_VALUE, u as LLM_INPUT_TOKEN_ATTR_KEYS, v as OPENINFERENCE_SPAN_KIND, w as TOOL_NAME, x as SPAN_KIND_ATTR_KEYS, y as OUTPUT_VALUE } from "./attribute-vocabulary-CcXqHODT.js";
11
- import { A as AnalyzeTracesInput, B as CaptureFetchContext, C as inferDomainKeywords, D as planTraceInsightQuestions, E as isTraceInsightEvidence, F as OTEL_AGENT_EVAL_SCOPE, G as SseUsageMode, H as captureFetchToRawSink, I as OtlpExport, J as extractUsageFromSse, K as extractUsage, L as OtlpResourceSpans, M as AnalyzeTracesResult, N as analyzeTraces, O as scoreTraceInsightReadiness, P as OtlpFlatLine, R as OtlpSpan, S as domainEvidencePattern, T as isTraceInsightAttribution, U as ExtractUsageFromSseOptions, V as CaptureFetchOptions, W as ExtractedUsage, _ as TraceInsightTask, a as ToolSpansToTraceAnalysisStoreOptions, b as defaultTraceInsightPanel, c as TraceInsightContext, d as TraceInsightPanelRole, f as TraceInsightPromptInput, g as TraceInsightSuite, h as TraceInsightReadiness, i as OtlpFileTraceStoreOptions, j as AnalyzeTracesOptions, k as tokenizeDomainWords, l as TraceInsightEvidence, m as TraceInsightQuestion, n as toolSpansToTraceAnalysisStore, o as otlpTextToTraceAnalysisStore, p as TraceInsightQualityGate, q as extractUsageFromResponse, r as OtlpFileTraceStore, s as TraceInsightAttribution, t as ToolTraceMissingError, u as TraceInsightFinding, v as buildTraceInsightContext, w as isTraceInsightArtifact, x as describeTraceInsightScope, y as buildTraceInsightPrompt, z as exportRunAsOtlp } from "./store-tool-spans-Cfh9XUCm.js";
11
+ import { A as AnalyzeTracesInput, B as CaptureFetchContext, C as inferDomainKeywords, D as planTraceInsightQuestions, E as isTraceInsightEvidence, F as OTEL_AGENT_EVAL_SCOPE, G as SseUsageMode, H as captureFetchToRawSink, I as OtlpExport, J as extractUsageFromSse, K as extractUsage, L as OtlpResourceSpans, M as AnalyzeTracesResult, N as analyzeTraces, O as scoreTraceInsightReadiness, P as OtlpFlatLine, R as OtlpSpan, S as domainEvidencePattern, T as isTraceInsightAttribution, U as ExtractUsageFromSseOptions, V as CaptureFetchOptions, W as ExtractedUsage, _ as TraceInsightTask, a as ToolSpansToTraceAnalysisStoreOptions, b as defaultTraceInsightPanel, c as TraceInsightContext, d as TraceInsightPanelRole, f as TraceInsightPromptInput, g as TraceInsightSuite, h as TraceInsightReadiness, i as OtlpFileTraceStoreOptions, j as AnalyzeTracesOptions, k as tokenizeDomainWords, l as TraceInsightEvidence, m as TraceInsightQuestion, n as toolSpansToTraceAnalysisStore, o as otlpTextToTraceAnalysisStore, p as TraceInsightQualityGate, q as extractUsageFromResponse, r as OtlpFileTraceStore, s as TraceInsightAttribution, t as ToolTraceMissingError, u as TraceInsightFinding, v as buildTraceInsightContext, w as isTraceInsightArtifact, x as describeTraceInsightScope, y as buildTraceInsightPrompt, z as exportRunAsOtlp } from "./store-tool-spans-ByLdU26j.js";
12
12
  import { a as RunIntegrityReport, i as RunIntegrityIssueCode, n as RunIntegrityExpectations, o as assertRunCaptured, r as RunIntegrityIssue, s as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-CaeNeusS.js";
13
13
  import { a as groupBy, c as judgeSpans, d as runMetricExtractor, f as runsForScenario, i as argHash, l as llmSpans, n as RunMetric, o as hasCapturedToolArgs, p as toolSpans, r as aggregateLlm, s as isRunMetric, t as RUN_METRICS, u as runFailureClass } from "./query-DBUQhehH.js";
14
14
  import { SpanKind as SpanKind$1 } from "@tangle-network/agent-trace-contract";
@@ -1,5 +1,5 @@
1
1
  import { c as RunSplitTag } from "./run-record-DF-fuu31.js";
2
- import { R as Scenario, d as DispatchContext } from "./types-BmQRs661.js";
2
+ import { R as Scenario, d as DispatchContext } from "./types-C1CAD70_.js";
3
3
  //#region src/benchmarks/types.d.ts
4
4
  type BenchmarkTaskKind = 'retrieval' | 'rag-answer' | 'hallucination' | 'kb-improvement' | 'routing' | 'custom';
5
5
  type BenchmarkFamily = 'beir' | 'mteb-retrieval' | 'msmarco' | 'trec-dl' | 'miracl' | 'lotte' | 'bright' | 'crag' | 'hotpotqa' | 'kilt' | 'ragtruth' | 'faithbench' | 'first-party' | 'custom';
@@ -99,4 +99,4 @@ declare const BENCHMARK_SPLIT_SEED = "agent-eval-v1";
99
99
  declare function deterministicSplit(itemId: string, seed?: string): RunSplitTag;
100
100
  //#endregion
101
101
  export { BenchmarkFamily as a, BenchmarkSource as c, BenchmarkEvaluation as i, BenchmarkTaskKind as l, BenchmarkAdapter as n, BenchmarkResponder as o, BenchmarkDatasetItem as r, BenchmarkScenario as s, BENCHMARK_SPLIT_SEED as t, deterministicSplit as u };
102
- //# sourceMappingURL=types-ClK-sHs8.d.ts.map
102
+ //# sourceMappingURL=types-ByZslIlr.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"types-ClK-sHs8.d.ts","names":[],"sources":["../src/benchmarks/types.ts"],"mappings":";;;KASY;KAQA;UAgBK,qBAAqB;;;EAGpC;;EAEA,SAAS;;EAET,QAAQ;;EAER,SAAS;;EAET,WAAW;;EAEX;;EAEA,SAAS;EACT,WAAW;;UAGI;;;;EAIf;;EAEA;;EAEA,aAAa;;;EAGb,KAAK;EACL;;UAGe;EACf;EACA;EACA;;EAEA;EACA;EACA;;EAEA;;;;;EAKA;;;UAQe,iBAAiB,kBAAkB,oBAAoB;;EAEtE;EACA,SAAS;EACT,WAAW;EACX;EACA,SAAS;EACT;;;;;EAKA,YAAY,OAAO,cAAc,QAAQ,qBAAqB;;EAE9D,SAAS,MAAM,qBAAqB,WAAW,UAAU,YAAY,QAAQ;;;;EAI7E,YAAY,iBAAiB;;UAGd,kBAAkB,4BAA4B;EAC7D;EACA;EACA,QAAQ;EACR,UAAU;EACV,UAAU;EACV,MAAM,qBAAqB;;KAGjB,mBAAmB,oBAAoB,uBAAuB;EACxE,UAAU,kBAAkB;EAC5B,MAAM,qBAAqB;EAC3B,SAAS;MACL,QAAQ,aAAa;;;cAoBd;;;;;;;;;iBAUG,mBACd,gBACA,gBACC"}
1
+ {"version":3,"file":"types-ByZslIlr.d.ts","names":[],"sources":["../src/benchmarks/types.ts"],"mappings":";;;KASY;KAQA;UAgBK,qBAAqB;;;EAGpC;;EAEA,SAAS;;EAET,QAAQ;;EAER,SAAS;;EAET,WAAW;;EAEX;;EAEA,SAAS;EACT,WAAW;;UAGI;;;;EAIf;;EAEA;;EAEA,aAAa;;;EAGb,KAAK;EACL;;UAGe;EACf;EACA;EACA;;EAEA;EACA;EACA;;EAEA;;;;;EAKA;;;UAQe,iBAAiB,kBAAkB,oBAAoB;;EAEtE;EACA,SAAS;EACT,WAAW;EACX;EACA,SAAS;EACT;;;;;EAKA,YAAY,OAAO,cAAc,QAAQ,qBAAqB;;EAE9D,SAAS,MAAM,qBAAqB,WAAW,UAAU,YAAY,QAAQ;;;;EAI7E,YAAY,iBAAiB;;UAGd,kBAAkB,4BAA4B;EAC7D;EACA;EACA,QAAQ;EACR,UAAU;EACV,UAAU;EACV,MAAM,qBAAqB;;KAGjB,mBAAmB,oBAAoB,uBAAuB;EACxE,UAAU,kBAAkB;EAC5B,MAAM,qBAAqB;EAC3B,SAAS;MACL,QAAQ,aAAa;;;cAoBd;;;;;;;;;iBAUG,mBACd,gBACA,gBACC"}
@@ -266,9 +266,15 @@ interface ScoredSurfaceOutcome {
266
266
  /** Search state supplied to one candidate-generation call.
267
267
  * Final evaluation data is not represented in this contract. */
268
268
  interface ProposeContext<TFindings = ProposalFinding> {
269
- /** The parent surface this proposal mutates: the global incumbent by
270
- * default, or the frontier parent a `selectParent` policy drew. */
269
+ /** The parent surface this proposal mutates: the node the search policy
270
+ * chose (the incumbent by default). */
271
271
  readonly currentSurface: MutableSurface;
272
+ /** How the child should derive from the parent: `improve` for a hill-climb
273
+ * step; other policies ask for `draft`, `debug` or `merge`. Absent outside
274
+ * a search. */
275
+ readonly operator?: 'draft' | 'improve' | 'debug' | 'merge';
276
+ /** One record per earlier proposal: its candidates' measurements and the
277
+ * candidate that took the lead, if any. */
272
278
  readonly history: ReadonlyArray<GenerationRecord>;
273
279
  readonly findings: ReadonlyArray<TFindings>;
274
280
  /** BREADTH: how many candidate surfaces to return this generation. */
@@ -284,8 +290,8 @@ interface ProposeContext<TFindings = ProposalFinding> {
284
290
  * candidate must beat to promote. `runOptimization` always supplies it. */
285
291
  readonly incumbentOutcome?: ScoredSurfaceOutcome;
286
292
  /** Measured result for `currentSurface`, the parent this proposal mutates.
287
- * Equal to `incumbentOutcome` unless the run draws parents from the Pareto
288
- * frontier through `selectParent`. `runOptimization` always supplies it. */
293
+ * Equal to `incumbentOutcome` unless the policy draws parents from the Pareto
294
+ * frontier. `runOptimization` always supplies it. */
289
295
  readonly parentOutcome?: ScoredSurfaceOutcome;
290
296
  /** DEPTH: max iterations the agentic generator may take per candidate.
291
297
  * 1 = single-shot; >1 = it may iterate on its own change before handing it
@@ -569,17 +575,10 @@ interface GenerationRecord {
569
575
  * handled — the evidence a blind `Mutator` cannot see. */
570
576
  interface GenerationCandidate {
571
577
  surfaceHash: string;
572
- /** Mean over complete task-quality scores, or null when none were produced. */
578
+ /** Mean over complete task-quality scores, or null when none were produced.
579
+ * Lineage and paired contrasts live in the search ledger: read the node's
580
+ * edges from its `SearchStateView` and its contrast from `estimateNode`. */
573
581
  composite: number | null;
574
- /** Estimated interval for `composite`, or null when uncertainty was not estimated. */
575
- ci95: [number, number] | null;
576
- /** Exact surface this candidate mutated. */
577
- parentSurfaceHash?: string;
578
- /** Measured search-split composite of the exact parent surface. */
579
- parentComposite?: number;
580
- /** Candidate composite minus its parent's composite. Present only when the
581
- * candidate completed the designed denominator. */
582
- observedDeltaFromParent?: number;
583
582
  /** Whether this candidate had a scorable result for every designed campaign
584
583
  * cell and was therefore eligible for ranking, promotion, and Pareto
585
584
  * selection. */
@@ -670,4 +669,4 @@ interface CampaignResult<TArtifact = unknown, TScenario extends Scenario = Scena
670
669
  }
671
670
  //#endregion
672
671
  export { LabeledScenarioWrite as A, ScoredSurfaceOutcome as B, JudgeDimension as C, LabeledScenarioSampleArgs as D, LabeledScenarioRecord as E, ProposeContext as F, labelTrustRank as G, SurfaceProposer as H, ProposedCandidate as I, RedactionStatus as L, OptimizerConfig as M, ParetoParent as N, LabeledScenarioSource as O, ProposalTrackContext as P, Scenario as R, JudgeConfig as S, LabelTrust as T, TraceSpan as U, SessionScript as V, isProposedCandidate as W, GateDecision as _, CampaignResult as a, GenerationRecord as b, CampaignTraceWriter as c, DispatchContext as d, DispatchFn as f, GateContribution as g, GateContext as h, CampaignCostMeter as i, MutableSurface as j, LabeledScenarioStore as k, CodeSurface as l, GateCheckStatus as m, CampaignArtifactWriter as n, CampaignScenarioIdentity as o, Gate as p, CampaignCellResult as r, CampaignTokenUsage as s, CampaignAggregates as t, ComponentSurface as u, GateResult as v, JudgeScore as w, JudgeAggregate as x, GenerationCandidate as y, ScenarioAggregate as z };
673
- //# sourceMappingURL=types-BmQRs661.d.ts.map
672
+ //# sourceMappingURL=types-C1CAD70_.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"types-C1CAD70_.d.ts","names":[],"sources":["../src/campaign/types.ts"],"mappings":";;;;;;;;UAiCiB;EACf;EACA;EACA;;;;;EAKA;;;UAIe,iCAAiC,KAAK;EACrD;;;;;UAMe;EACf;;EAEA;EACA;EACA;EACA;EACA,QAAQ;EACR,OAAO;EACP,WAAW;EACX,MAAM;;EAEN;;EAEA;;;;;;;;EAQA;;;;KAKU,WAAW,kBAAkB,UAAU,cACjD,UAAU,WACV,KAAK,oBACF,QAAQ;;;;UAOI,cAAc,WAAW;EACxC;EACA;EACA;;EAEA;;;EAGA,sBAAsB,UAAU,WAAW,sBAAsB,UAAU,cAAc;;UAK1E;;EAEf;;EAEA;;;;;;;;;UAUe,YAAY,WAAW,kBAAkB,WAAW;EACnE;EACA,YAAY;;;;EAIZ;;;EAGA,MAAM;IACJ,UAAU;IACV,UAAU;IACV,QAAQ;;IAER,aAAa;IACb;IACA,WAAW;MACT,aAAa,QAAQ;EACzB,aAAa,UAAU;;;;;;;;;;;UAYR;EACf,YAAY;EACZ;EACA;;EAEA,UAAU;;;;;;;EAOV;;;;EAIA,eAAe,eAAe;IAAQ;IAAe;;;;;EAIrD;;;EAGA;;EAEA;;EAEA,WAAW,eAAe;;;;;;;UAUX;WACN;;;WAGA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;;WAGA;aACE;aACA;aACA;;;WAGF;;;UAIM;WACN;WACA,YAAY,SAAS;;;;;;;;;;;KAYpB,0BAA0B,mBAAmB;;;;;;;UAQxC;EACf,SAAS;;EAET;;;;EAIA;;;;;;EAMA,cAAc,SAAS;;;;iBAKT,oBACd,OAAO,iBAAiB,oBACvB,SAAS;;;;;;;;;UAkBK;EACf,SAAS;EACT;;;EAGA,YAAY;;;EAGZ;;EAEA;EACA;EACA;;;UAIe;EACf;EACA;EACA;EACA;EACA;;;;;UAMe;;;EAGf;;EAEA;EACA;EACA;EACA,YAAY;;;;;EAKZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;EAC1E;IACE;IACA;;;;;UAMa,eAAe,YAAY;;;WAGjC,gBAAgB;;;;WAIhB;;;WAGA,SAAS,cAAc;WACvB,UAAU,cAAc;;WAExB;WACA;WACA,QAAQ;;WAER,QAAQ;;;WAGR,kBAAkB;;;WAGlB,mBAAmB;;;;WAInB,gBAAgB;;;;WAIhB;;;;;;;WAOA,gBAAgB,cAAc;;WAE9B,aAAa;WACb;;;;;;;;;;;;;;UAeM,gBAAgB,YAAY;EAC3C;;;;;EAKA,QAAQ,KAAK,eAAe,aAAa,QAAQ,MAAM,iBAAiB;;;EAGxE,QAAQ;IAAQ,SAAS,cAAc;;IAAwB;IAAe;;;UAG/D;EACf;EACA;EACA,mBAAmB,qBAAqB;;UAGzB,wBAAwB;EACvC,UAAU;;;KAMA;;KAGA;UAEK;EACf;EACA,QAAQ;EACR;;UAGe,YAAY,WAAW,kBAAkB;EACxD,oBAAoB,YAAY;EAChC,oBAAoB,YAAY;;EAEhC,aAAa,YAAY,eAAe;;;;;EAKxC,sBAAsB,YAAY,eAAe;;;;;;;EAOjD,yBAAyB,YAAY,eAAe;;;EAGpD,uBAAuB,YAAY;EACnC,WAAW;EACX;IAAQ;IAAmB;;;EAE3B,aAAa;EACb;EACA,QAAQ;;UAGO;EACf,UAAU;EACV;EACA,mBAAmB;EACnB;;;UAIe,KAAK,qBAAqB,kBAAkB,WAAW;EACtE;EACA,OAAO,KAAK,YAAY,WAAW,aAAa,QAAQ;;;;UAOzC;EACf,KAAK,cAAc,aAAa,0BAA0B;EAC1D,SAAS;;UAGM;EACf,IAAI,aAAa;EACjB,aAAa,aAAa;;;;UAKX;EACf,MAAM,cAAc,kBAAkB,aAAa;EACnD,UAAU,cAAc,iBAAiB;;;;;;KAO/B,qBAAqB;;;;;UAMhB;;EAEf,YAAY,GACV,OAAO,KAAK,iBAAiB;IAC3B,UAAU;MAEX,QAAQ,eAAe;;;;;KAQhB;KAOA;;;;;;;;;;;;;;;KAgBA;;iBASI,eAAe,OAAO;;;;UAOrB,qBAAqB,kBAAkB,WAAW,UAAU;EAC3E,UAAU;EACV,UAAU;EACV,aAAa,eAAe;EAC5B,QAAQ;EACR;EACA;EACA,iBAAiB;;;;;EAKjB,aAAa;;EAEb;;UAGe,sBAAsB,kBAAkB,WAAW,UAAU,6BACpE,qBAAqB,WAAW;;EAExC;;;EAGA;;UAGe;EACf;;EAEA;;;;EAIA;EACA;IACE;IACA,SAAS,wBAAwB;IACjC;IACA;;;;;IAKA,WAAW;;;UAIE;EACf,QAAQ,OAAO,uBAAuB;EACtC,OAAO,MAAM,4BAA4B,QAAQ;EACjD,QAAQ;IACN;IACA;IACA,UAAU;;;IAGV,SAAS,OAAO;;;UAMH,mBAAmB;;;EAGlC;EACA;EACA;EACA;EACA;EACA,UAAU;EACV,aAAa,eAAe;;;EAG5B;;EAEA,gBAAgB;;EAEhB;;;EAGA,YAAY;;;EAGZ;;;EAGA;EACA;EACA;EACA;;;;EAIA;;EAEA;;EAEA;EACA;;UAGe;EACf;EACA;EACA;EACA;;;;;;EAMA,cAAc;;UAGC;EACf;EACA;EACA;;;EAGA,cAAc;;UAGC;EACf;EACA,YAAY;EACZ;;;;;;UAOe;EACf;;;;EAIA;;;;EAIA;;;;EAIA;IACE;IACA;IACA,iBAAiB;MAAQ;MAAgB;;;;;EAI3C,YAAY;;;;;;;;;;EAUZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;;;EAG1E;;;;EAIA;;;;EAIA,cAAc,SAAS;;UAGR;EACf,SAAS,eAAe;EACxB,YAAY,eAAe;;EAE3B,MAAM;;EAEN;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;UAGe,eAAe,qBAAqB,kBAAkB,WAAW;;EAEhF;;EAEA;EACA;;EAEA;EACA;EACA;EACA;EACA,OAAO,MAAM,mBAAmB;EAChC,YAAY;EACZ;IACE,aAAa;IACb;;EAEF,OAAO;EACP;EACA;EACA,iBAAiB;;;EAGjB,WAAW,MAAM,2BAA2B,KAAK"}
@@ -100,7 +100,7 @@ Underreported spending and incomplete receipts remain explicit; `raw.method.cost
100
100
  Inspect `cost.accountingComplete` and `cost.incompleteReasons` before treating the known subtotal as complete spending.
101
101
  The shared dollar limit controls calls admitted through the cost ledger; arbitrary off-ledger callbacks must enforce their own spending limits.
102
102
 
103
- Native generation records report `ci95: null` because search does not estimate candidate uncertainty.
103
+ Native generation records carry no interval; a candidate's paired contrast with its parent or the baseline is `estimateNode` over the run's search ledger.
104
104
  Final comparisons retain their independently computed statistics.
105
105
  Every final case and replica must have complete execution and judge results before comparison.
106
106
 
@@ -596,38 +596,36 @@ const proposer: SurfaceProposer = {
596
596
  Return a label and rationale when they will help later analysis.
597
597
  Candidate creation must not read final test results.
598
598
 
599
- A proposer may attach `attribution`: an opaque JSON-safe record retained on `GenerationCandidate.attribution` and in loop provenance.
599
+ A proposer may attach `attribution`: an opaque JSON-safe record stored with the proposal in the search ledger and retained on `GenerationCandidate.attribution` and in loop provenance.
600
600
  The loop never interprets it.
601
601
  Tag it with a schema field and validate it on readback.
602
602
  `makePolicyEditCandidateRecord` from `/analyst` records an edit forecast that can later be compared with the measured change.
603
603
 
604
- `runOptimization()` rejects a candidate whose `surfaceHash` was already admitted.
605
- This includes the baseline, an earlier generation, and another candidate in the same proposal.
606
- The complete proposal is checked before candidate dispatch, so duplicates cannot consume candidate cells.
604
+ A candidate is a content-addressed node in the run's search ledger.
605
+ A candidate identical to a surface the search already holds, including the baseline, is a re-proposal: the ledger records a second edge into the existing node, and the surface is not measured again.
607
606
  Use `reps` when one surface needs repeated measurements.
607
+ Lineage lives in the ledger: read a node's parents from its edges and its paired contrast from `estimateNode`, not from the generation record.
608
608
 
609
609
  ### Choose The Parent
610
610
 
611
- `runOptimization()` is incumbent-anchored by default.
612
- Every generation mutates the single best complete surface seen so far, and only a candidate that beats the incumbent becomes the next incumbent.
613
- The recorded `parentSurfaceHash` lineage is then a chain.
614
- Pass `selectParent` to draw the parent from the Pareto frontier instead.
615
- The selector receives the frontier so far, the measured incumbent, the generation history, and the generation index, and returns one frontier parent.
616
- The loop hands that parent to the proposer as `ctx.currentSurface` and `ctx.parentOutcome`, and records it as every candidate's `parentSurfaceHash`.
617
- `ctx.incumbentOutcome` stays the global promotion bar, and a candidate still has to beat the incumbent to promote.
618
- `crowdedFrontierParent({ seed })` is the provided policy: a seeded NSGA-II crowded tournament that prefers isolated frontier parents.
619
- The loop refuses a parent it has not measured to completion.
620
- `selfImprove({ selectParent })` forwards the same policy in proposer mode.
611
+ `runOptimization()` runs on the search kernel with a `SearchPolicy` ([search ledger](./search-ledger.md#run-a-search-the-kernel)).
612
+ The default, `incumbent()`, is the hill climb: each proposal extends the leader once every earlier candidate is measured, and a candidate that scored every scenario leads when it beats the leader on the scenarios they share.
613
+ Pass `policy: crowdedFrontierParent({ seed })` to draw the parent from the Pareto frontier instead: a seeded NSGA-II crowded tournament that prefers isolated frontier members.
614
+ The proposer receives that parent as `ctx.currentSurface` and `ctx.parentOutcome`; `ctx.incumbentOutcome` stays the leader, and only a candidate that beats the leader leads.
615
+ `selfImprove({ policy })` forwards the same policy in proposer mode.
621
616
 
622
617
  ```ts
623
618
  import { crowdedFrontierParent, runOptimization } from '@tangle-network/agent-eval/campaign'
624
619
 
625
620
  const result = await runOptimization({
626
621
  // ...scenarios, dispatchWithSurface, judges, proposer, populationSize, maxGenerations, runDir
627
- selectParent: crowdedFrontierParent({ seed: 42 }),
622
+ policy: crowdedFrontierParent({ seed: 42 }),
628
623
  })
629
624
  ```
630
625
 
626
+ `runOptimization` always writes its search ledger, at `<runDir>/search/ledger.jsonl` unless `searchLedger` puts it elsewhere, and returns the receipt on `searchHistory`.
627
+ Running it again on the same run directory continues an interrupted search and reruns nothing that settled; after a finished search it starts the next one beside it.
628
+
631
629
  ## Data And Cost Rules
632
630
 
633
631
  - Train and selection cases are visible to complete optimization methods.
@@ -134,13 +134,70 @@ const cellId = await recorder.allocateCell({ nodeId: root.nodeId, taskId, split:
134
134
  await recorder.settleCell({ cellId, outcome, accounting, identity, runRecord })
135
135
  ```
136
136
 
137
- `runOptimization({ searchLedger: { ledger, identity } })` and `selfImprove` record through the recorder and return the receipt on `searchHistory`.
137
+ `runOptimization` and `selfImprove` run on the kernel below and return the receipt on `searchHistory`.
138
138
  The loop's scenarios are its proposer's feedback, so they are the train split; its promotions are budget decisions and it makes no claim.
139
139
 
140
140
  `gepaOptimizationMethod({ searchLedger: { identity } })` records GEPA's search when it finishes.
141
141
  `importGepaPopulation` turns the population into nodes and `correlated` edges and reports collapsed duplicates.
142
142
  `importExternalEvaluations` turns every callback evaluation into an `external` cell; a candidate GEPA evaluated but kept out of its population gets an `unknown` edge and is decided `pruned`.
143
143
 
144
+ ## Run a search: the kernel
145
+
146
+ `runSearch` is the one loop every optimizer runs on.
147
+ It separates three decisions: where to expand (a `SearchPolicy`), where to spend rollouts (a `SearchAllocator`), and what to claim (the claim step on the sealed test split).
148
+ There is no generation barrier.
149
+ A lane that frees up takes the next allocated cell: claim cells first, then rung and root cells, then screens, then train cells.
150
+ The policy proposes when no cell waits, fewer than twice the lanes' capacity run, and the cap admits one proposal and its expected screens.
151
+
152
+ ```ts
153
+ import { incumbent, runSearch, SearchRecorder, uniform } from '@tangle-network/agent-eval/campaign'
154
+
155
+ const policy = incumbent({ patience: 3 })
156
+ const allocation = uniform({ reps: 1 })
157
+ const recorder = await SearchRecorder.open({ ledger }, {
158
+ ...opening,
159
+ policy: { expansion: policy.name, allocation: allocation.name, seed },
160
+ })
161
+ const { state, leader, reason } = await runSearch({
162
+ recorder, root, codec, policy, allocation, proposer, executor, maxExpansions: 10,
163
+ })
164
+ ```
165
+
166
+ The ports:
167
+
168
+ - **Executor:** `lanes()` declares pools of slots, each `hard` (it enforces `cellUsd` per cell) or `estimate` (it cannot); `place(cell)` picks a cell's lane; `run(work)` runs one attempt and returns its outcome, accounting and identity; `adopt(work)` returns an attempt an earlier process finished, or null.
169
+ An environment fault is an `errored` outcome; a rejection stops the search.
170
+ - **Proposer:** `propose({ parents, operator, leader })` returns children with a label, a rationale and optional typed `attribution`, plus the operation's accounting.
171
+ Its output is stored as a `proposal` blob on `operation-recorded` before any child is registered.
172
+ - **Codec:** `node(recorder, artifact)` content-addresses an artifact, `diff` stores the parent-to-child diff, and `load` reads a node's artifact back.
173
+ - **Policy:** `expand(view)` returns parents and an operator, or null to wait; `leader(view)` names the node the search keeps.
174
+ The view holds the policy split (selection, or train when a search has none) and no test cell.
175
+ `incumbent({ patience })` is the hill climb: it expands the leader once every earlier child is screened, and a node that scored every unit of its screen leads when it beats the leader on the units they share.
176
+ `crowdedFrontierParent({ seed })` draws the parent from the Pareto frontier by a seeded crowded tournament and keeps the incumbent's leader rule.
177
+ - **Allocator:** `plan(state, nodeId)` lists the cells a node needs now; the kernel allocates the ones the ledger lacks.
178
+ `uniform({ reps })` gives the root every train and selection task as `root` cells, and every other node every train task (`train`) and selection task (`screen`).
179
+
180
+ **Budget.**
181
+ Every reservation passes the ledger's admission rule: committed spend plus open reservations plus the unspent claim reserve plus the new hold stays within `maxUsd`.
182
+ The kernel checks `state.budget.headroomUsd` first and prices an expansion as one proposal plus `childrenPerProposal` screens.
183
+ A hard lane holds its maximum; an estimate lane holds 1.5 times the p99 of its settled cells once 20 settled, else its prior.
184
+ Spend above a hold is recorded as overspend, never refused.
185
+ Expansion stops at `maxNodes`, `maxExpansions`, the deadline, after `patience` expansions without a new leader, when the cap cannot admit one more expansion, or when the proposer stops.
186
+ Allocated cells still run; at the deadline the ones not yet started are cancelled.
187
+ At close the leader is decided `selected` when it has a scored cell, every other undecided node `rejected` with its estimate against the leader, and the search closes with no claim.
188
+
189
+ **Resume.**
190
+ The ledger is the only checkpoint.
191
+ Running the kernel again on an open ledger replays it and continues.
192
+ An operation that started without a result is recorded `failed` with an unknown cost and a floor of 0, and the next proposal runs under a new operation id.
193
+ A recorded proposal whose children were not all registered is finished from its stored output.
194
+ A cell the ledger shows unsettled is offered to `executor.adopt` before its first dispatch, so an attempt that finished before the restart is recorded once and not run again.
195
+ Aborting the `signal` pauses the search: in-flight cells are aborted, a scored result that still arrives is recorded, an attempt that ends in an error while the search stops stays unsettled (the error may be the interruption), and the ledger stays open.
196
+ One kernel runs a ledger file at a time on a host: a second one is refused by a pid lock beside the ledger, and a killed holder's lock is reclaimed.
197
+ A closed ledger returns its result; more work on a closed search is a new search.
198
+
199
+ `scripts/search-sim.ts` runs the real kernel, ledger and policies over a seeded synthetic objective, proposer and executor, and `kill-resume` SIGKILLs it at random ledger positions and compares the resumed search with an uninterrupted one.
200
+
144
201
  ## Ship a search
145
202
 
146
203
  A hosted store (Intelligence, or the reference receiver in `examples/hosted-ingest-server/`) receives a search through the [hosted ingest wire](./hosted-ingest-spec.md).
@@ -186,6 +243,10 @@ Those claims need the sealed test split, the claim's power check, and held-out e
186
243
  - `src/campaign/search-state.ts`: `SearchState`, the invariants and read model, and the id functions.
187
244
  - `src/campaign/estimate-node.ts`: `estimateNode`, `estimateNodeFromCells` and `searchPosterior`.
188
245
  - `src/campaign/search-ledger-recording.ts`: `SearchRecorder` and the surface helpers.
246
+ - `src/campaign/search-kernel.ts`: `runSearch`, the executor, proposer and codec ports, and `searchPolicyView`.
247
+ - `src/campaign/search-policy.ts`: `SearchPolicy`, `incumbent` and `crowdedFrontierParent`.
248
+ - `src/campaign/allocation.ts`: `SearchAllocator` and `uniform`.
249
+ - `src/campaign/presets/run-optimization.ts`: `runOptimization` as a search on the kernel.
189
250
  - `src/campaign/gepa-search-import.ts`: the GEPA population and evaluation importers.
190
251
  - `src/campaign/search-history-receipt.ts`: receipts and admission.
191
252
  - `src/ledger-core/`: hashing, locking, durable appends, chain verification, and the trusted-head pin.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@tangle-network/agent-eval",
3
- "version": "0.191.0",
3
+ "version": "0.192.0",
4
4
  "description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
5
5
  "homepage": "https://github.com/tangle-network/agent-eval#readme",
6
6
  "repository": {