@tangle-network/agent-eval 0.163.2 → 0.170.0

package/dist/{skillopt-optimization-method-x7TTF23P.d.ts → heldout-gate-Bn7_xWCv.d.ts}

@@ -1,10 +1,22 @@
 import { c as CostLedgerHandle } from "./cost-ledger-DbQdN3nO.js";
-import { a as RunRecord } from "./run-record-VVy4T9OW.js";
-import { Qn as ServedModelPolicy, V as SearchRunIdentity, W as RedTeamCase, at as OptimizationMethod, et as CanaryOptions, pt as OptimizationPackageSource } from "./transient-failure-DKF5Mofa.js";
+import { a as RunRecord } from "./run-record-DQjRcYwA.js";
+import { A as CanaryOptions, F as OptimizationMethod, Qn as ServedModelPolicy, U as OptimizationPackageSource, b as RedTeamCase, yt as SearchRunIdentity } from "./provenance-CIRUardl.js";
 import { C as LlmCallMetadata, p as ChatClient } from "./types-Bfk0uxRj.js";
+import { R as Scenario, S as JudgeConfig, i as CampaignCostMeter, j as MutableSurface, o as CampaignScenarioIdentity, p as Gate } from "./types-Dy237wiH.js";
 import { _ as ExternalTextCandidate, c as ExternalOptimizerModelBudget, h as ExternalOptimizerRunnerCommand, l as ExternalOptimizerModelCall, m as ExternalOptimizerResumeMode, r as ExternalOptimizerCallbackLimits, v as ExternalTextEvaluationRequest } from "./external-optimizer-contracts-szBJ_1vh.js";
-import { R as Scenario, S as JudgeConfig, i as CampaignCostMeter, j as MutableSurface, o as CampaignScenarioIdentity, p as Gate } from "./types-D4s7Z6nq.js";
-import { t as DetectRewardHackingInput } from "./reward-hacking-MBf7qpSB.js";
+import { t as DetectRewardHackingInput } from "./reward-hacking-CgPRUesA.js";
+//#region src/campaign/coverage.d.ts
+/** Reject campaign designs whose denominator cannot be identified exactly. */
+declare function assertCampaignDesign<TScenario extends Scenario>(scenarios: readonly TScenario[], reps: number): void;
+/** Redacted but independently verifiable identity of one complete scenario. */
+declare function campaignScenarioIdentity<TScenario extends Scenario>(scenario: TScenario): CampaignScenarioIdentity & Pick<TScenario, 'id' | 'kind'>;
+/** Canonical split identity reconstructed from redacted scenario identities. */
+declare function campaignSplitDigestFromIdentities(scenarios: readonly CampaignScenarioIdentity[], reps: number): `sha256:${string}`;
+/** Canonical identity of the exact scenario payloads and replicate count. */
+declare function campaignSplitDigest<TScenario extends Scenario>(scenarios: readonly TScenario[], reps: number): `sha256:${string}`;
+/** Refuse a campaign whose retained task identities contradict its split digest. */
+declare function assertCampaignSplitIdentity(scenarios: readonly CampaignScenarioIdentity[], reps: number, splitDigest: string): void;
+//#endregion
 //#region src/reference-equivalence-judge.d.ts
 declare const REFERENCE_EQUIVALENCE_JUDGE_VERSION = "reference-equivalence-judge-v1-2026-07-13";
 declare const REFERENCE_EQUIVALENCE_INPUT_LIMITS: {
@@ -42,18 +54,6 @@ declare function createReferenceEquivalenceJudge(options: ReferenceEquivalenceJu
 /** Direct-call adapter over the campaign judge for product callers. */
 declare function runReferenceEquivalenceJudge(input: ReferenceEquivalenceJudgeInput, options: ReferenceEquivalenceJudgeOptions): Promise<ReferenceEquivalenceJudgeResult>;
 //#endregion
-//#region src/campaign/coverage.d.ts
-/** Reject campaign designs whose denominator cannot be identified exactly. */
-declare function assertCampaignDesign<TScenario extends Scenario>(scenarios: readonly TScenario[], reps: number): void;
-/** Redacted but independently verifiable identity of one complete scenario. */
-declare function campaignScenarioIdentity<TScenario extends Scenario>(scenario: TScenario): CampaignScenarioIdentity & Pick<TScenario, 'id' | 'kind'>;
-/** Canonical split identity reconstructed from redacted scenario identities. */
-declare function campaignSplitDigestFromIdentities(scenarios: readonly CampaignScenarioIdentity[], reps: number): `sha256:${string}`;
-/** Canonical identity of the exact scenario payloads and replicate count. */
-declare function campaignSplitDigest<TScenario extends Scenario>(scenarios: readonly TScenario[], reps: number): `sha256:${string}`;
-/** Refuse a campaign whose retained task identities contradict its split digest. */
-declare function assertCampaignSplitIdentity(scenarios: readonly CampaignScenarioIdentity[], reps: number, splitDigest: string): void;
-//#endregion
 //#region src/campaign/external-text-evaluation.d.ts
 interface ExternalOptimizationExample {
   id: string;
@@ -145,99 +145,6 @@ interface ExternalTextOptimizationMethodConfig<TScenario extends Scenario, TArti
  */
 declare function externalTextOptimizationMethod<TScenario extends Scenario, TArtifact>(config: ExternalTextOptimizationMethodConfig<TScenario, TArtifact>): OptimizationMethod<TScenario, TArtifact>;
 //#endregion
-//#region src/campaign/gates/compose.d.ts
-/** Compose gates — all must `ship` for the composite to `ship`. First
- *  non-ship verdict short-circuits the composite verdict, but ALL gates run
- *  (so the result records every gate's reason — useful for diagnostics). */
-declare function composeGate<TArtifact = unknown, TScenario extends Scenario = Scenario>(...gates: Array<Gate<TArtifact, TScenario>>): Gate<TArtifact, TScenario>;
-//#endregion
-//#region src/campaign/gates/default-production-gate.d.ts
-type DefaultProductionGateCheck = 'dimension-regression' | 'budget' | 'red-team' | 'reward-hacking' | 'canary';
-type DefaultProductionRewardHackingOptions = Omit<DetectRewardHackingInput, 'runs' | 'truthOf'> & {
-  truthOf: NonNullable<DetectRewardHackingInput['truthOf']>;
-};
-interface DefaultProductionGateOptions {
-  /** Required: scenarios held out from training; substrate compares
-   *  candidate-on-holdout vs baseline-on-holdout. */
-  holdoutScenarios: Scenario[];
-  /** Minimum held-out lift the **paired-bootstrap CI lower bound** must clear
-   *  to ship — NOT a point estimate. Default 0 ⇒ "confidently positive at the
-   *  confidence level". Interpreted in the judge's native composite scale (set
-   *  e.g. 2 for a 0-100 rubric to require a ≥2-point significant gain). */
-  deltaThreshold?: number;
-  /** Confidence level for the held-out + dimension bootstraps. Default 0.95. */
-  confidence?: number;
-  /** Bootstrap resamples. Default 2000. */
-  bootstrapResamples?: number;
-  /** Fixed bootstrap seed for a deterministic verdict. Default 1337. */
-  bootstrapSeed?: number;
-  /** Minimum paired holdout observations (scenarios × reps) before a
-   *  significance claim is allowed. The exact small-sample test may require
-   *  more observations at the selected confidence. Default 3. */
-  minProductiveRuns?: number;
-  /** Ship statistic for the held-out significance test. Default `'mean'`
-   *  (tie-robust — see `heldoutSignificance`). Pass `'median'` for
-   *  outlier-robustness at the cost of tie-blindness. */
-  heldoutStatistic?: 'mean' | 'median';
-  /** Critical judge dimensions that must NOT significantly regress even when
-   *  the net composite rises (anti-Goodhart). The gate HOLDS if any listed
-   *  dimension's paired-delta CI lower bound < −`regressionTolerance`. E.g.
-   *  `['hallucination_free']` for a legal agent. */
-  criticalDimensions?: string[];
-  /** Tolerance for the per-dimension regression guard, in the dimension's
-   *  native scale. When omitted it auto-scales off observed magnitudes:
-   *  0.05 on [0,1], 5 on 0-100. */
-  regressionTolerance?: number;
-  /** Total $ budget for the complete improvement run. Requires
-   *  `GateContext.costLedger`; missing or incomplete accounting holds. */
-  budgetUsd?: number;
-  /** Static artifact-screening cases. Only `expected: 'ignore'` cases without
-   *  tool assertions are valid because this check does not dispatch case inputs
-   *  or observe tool calls. */
-  redTeamBattery?: RedTeamCase[];
-  /** Shared run history, oldest first. Supplying history does not enable either
-   *  monitoring check; configure `rewardHacking` and/or `canary` explicitly. */
-  recentRuns?: RunRecord[];
-  /** Enable reward-hacking monitoring with a caller-owned independent truth channel. */
-  rewardHacking?: DefaultProductionRewardHackingOptions;
-  /** Enable canary monitoring. Pass `{}` to use the canary defaults. */
-  canary?: CanaryOptions;
-  /** Optional checks that must be evaluated even when their normal input is
-   *  absent. Configuring a check's input also makes that check required.
-   *  Missing evidence always records `not_evaluated`; required unevaluated
-   *  checks hold the release decision. Held-out significance is always required. */
-  requiredChecks?: DefaultProductionGateCheck[];
-}
-/**
- * Opinionated production gate composing held-out significance, red-team, reward-hacking, and canary checks into a single `Gate.decide` decision.
- */
-declare function defaultProductionGate<TArtifact, TScenario extends Scenario>(options: DefaultProductionGateOptions): Gate<TArtifact, TScenario>;
-//#endregion
-//#region src/campaign/gates/heldout-gate.d.ts
-interface HeldOutGateOptions<TScenario extends Scenario = Scenario> {
-  scenarios: TScenario[];
-  /** Effect-size threshold the CI lower bound must clear, in the judge's native
-   *  scale. Default 0.5. Equality holds; CI.low must be greater than this value. */
-  deltaThreshold?: number;
-  /** Bootstrap CI confidence. Default 0.95. */
-  confidence?: number;
-  /** Minimum paired holdout observations to claim significance. The exact
-   *  small-sample test may require more observations at the selected
-   *  confidence. Default 3. */
-  minProductiveRuns?: number;
-  /** Bootstrap resamples. Default 2000. */
-  resamples?: number;
-  /** Fixed bootstrap seed for deterministic verdicts. Default 1337. */
-  bootstrapSeed?: number;
-}
-/**
- * Composable held-out gate: ships only when the lower bound of the DECIDING
- * paired interval on the candidate-minus-baseline composite delta clears
- * `deltaThreshold` — Tango's score interval on a pass/fail holdout, the mean
- * bootstrap otherwise. See {@link decidePairedPromotion}.
- */
-declare function heldOutGate<TArtifact, TScenario extends Scenario>(options: HeldOutGateOptions<TScenario>): Gate<TArtifact, TScenario>;
-//#endregion
 //#region src/campaign/optimizer-model.d.ts
 type OptimizerModelBudget = ExternalOptimizerModelBudget;
 /** One metered model path supplied by the package that owns execution. */
@@ -449,5 +356,98 @@ interface SkillOptOptimizationMethodConfig<TScenario extends Scenario, TArtifact
 /** Run Microsoft's SkillOpt trainer as a complete optimization method. */
 declare function skillOptOptimizationMethod<TScenario extends Scenario, TArtifact>(config: SkillOptOptimizationMethodConfig<TScenario, TArtifact>): OptimizationMethod<TScenario, TArtifact>;
 //#endregion
-export { campaignScenarioIdentity as A, runReferenceEquivalenceJudge as B, ExternalTextOptimizerContext as C, decodeExternalTextCandidate as D, ExternalTextEvaluationResponse as E, ReferenceEquivalenceJudgeInput as F, ReferenceEquivalenceJudgeOptions as I, ReferenceEquivalenceJudgeResult as L, campaignSplitDigestFromIdentities as M, REFERENCE_EQUIVALENCE_INPUT_LIMITS as N, assertCampaignDesign as O, REFERENCE_EQUIVALENCE_JUDGE_VERSION as P, ReferenceEquivalenceScenario as R, ExternalTextOptimizationMethodConfig as S, ExternalOptimizationExample as T, DefaultProductionGateOptions as _, GepaAdaptiveEngineRun as a, composeGate as b, GepaOptimizationMethodConfig as c, gepaOptimizationMethod as d, OpenAICompatibleOptimizerModel as f, DefaultProductionGateCheck as g, heldOutGate as h, skillOptOptimizationMethod as i, campaignSplitDigest as j, assertCampaignSplitIdentity as k, GepaOptimizationRecipe as l, HeldOutGateOptions as m, SkillOptRunnerCommand as n, GepaEngineOptions as o, OptimizerModelBudget as p, SkillOptTrainerConfig as r, GepaEngineRun as s, SkillOptOptimizationMethodConfig as t, GepaRunnerCommand as u, DefaultProductionRewardHackingOptions as v, ExternalTextOptimizerResult as w, externalTextOptimizationMethod as x, defaultProductionGate as y, createReferenceEquivalenceJudge as z };
-//# sourceMappingURL=skillopt-optimization-method-x7TTF23P.d.ts.map
+//#region src/campaign/gates/compose.d.ts
+/** Compose gates — all must `ship` for the composite to `ship`. First
+ *  non-ship verdict short-circuits the composite verdict, but ALL gates run
+ *  (so the result records every gate's reason — useful for diagnostics). */
+declare function composeGate<TArtifact = unknown, TScenario extends Scenario = Scenario>(...gates: Array<Gate<TArtifact, TScenario>>): Gate<TArtifact, TScenario>;
+//#endregion
+//#region src/campaign/gates/default-production-gate.d.ts
+type DefaultProductionGateCheck = 'dimension-regression' | 'budget' | 'red-team' | 'reward-hacking' | 'canary';
+type DefaultProductionRewardHackingOptions = Omit<DetectRewardHackingInput, 'runs' | 'truthOf'> & {
+  truthOf: NonNullable<DetectRewardHackingInput['truthOf']>;
+};
+interface DefaultProductionGateOptions {
+  /** Required: scenarios held out from training; substrate compares
+   *  candidate-on-holdout vs baseline-on-holdout. */
+  holdoutScenarios: Scenario[];
+  /** Minimum held-out lift the **paired-bootstrap CI lower bound** must clear
+   *  to ship — NOT a point estimate. Default 0 ⇒ "confidently positive at the
+   *  confidence level". Interpreted in the judge's native composite scale (set
+   *  e.g. 2 for a 0-100 rubric to require a ≥2-point significant gain). */
+  deltaThreshold?: number;
+  /** Confidence level for the held-out + dimension bootstraps. Default 0.95. */
+  confidence?: number;
+  /** Bootstrap resamples. Default 2000. */
+  bootstrapResamples?: number;
+  /** Fixed bootstrap seed for a deterministic verdict. Default 1337. */
+  bootstrapSeed?: number;
+  /** Minimum paired holdout observations (scenarios × reps) before a
+   *  significance claim is allowed. The exact small-sample test may require
+   *  more observations at the selected confidence. Default 3. */
+  minProductiveRuns?: number;
+  /** Ship statistic for the held-out significance test. Default `'mean'`
+   *  (tie-robust — see `heldoutSignificance`). Pass `'median'` for
+   *  outlier-robustness at the cost of tie-blindness. */
+  heldoutStatistic?: 'mean' | 'median';
+  /** Critical judge dimensions that must NOT significantly regress even when
+   *  the net composite rises (anti-Goodhart). The gate HOLDS if any listed
+   *  dimension's paired-delta CI lower bound < −`regressionTolerance`. E.g.
+   *  `['hallucination_free']` for a legal agent. */
+  criticalDimensions?: string[];
+  /** Tolerance for the per-dimension regression guard, in the dimension's
+   *  native scale. When omitted it auto-scales off observed magnitudes:
+   *  0.05 on [0,1], 5 on 0-100. */
+  regressionTolerance?: number;
+  /** Total $ budget for the complete improvement run. Requires
+   *  `GateContext.costLedger`; missing or incomplete accounting holds. */
+  budgetUsd?: number;
+  /** Static artifact-screening cases. Only `expected: 'ignore'` cases without
+   *  tool assertions are valid because this check does not dispatch case inputs
+   *  or observe tool calls. */
+  redTeamBattery?: RedTeamCase[];
+  /** Shared run history, oldest first. Supplying history does not enable either
+   *  monitoring check; configure `rewardHacking` and/or `canary` explicitly. */
+  recentRuns?: RunRecord[];
+  /** Enable reward-hacking monitoring with a caller-owned independent truth channel. */
+  rewardHacking?: DefaultProductionRewardHackingOptions;
+  /** Enable canary monitoring. Pass `{}` to use the canary defaults. */
+  canary?: CanaryOptions;
+  /** Optional checks that must be evaluated even when their normal input is
+   *  absent. Configuring a check's input also makes that check required.
+   *  Missing evidence always records `not_evaluated`; required unevaluated
+   *  checks hold the release decision. Held-out significance is always required. */
+  requiredChecks?: DefaultProductionGateCheck[];
+}
+/**
+ * Opinionated production gate composing held-out significance, red-team, reward-hacking, and canary checks into a single `Gate.decide` decision.
+ */
+declare function defaultProductionGate<TArtifact, TScenario extends Scenario>(options: DefaultProductionGateOptions): Gate<TArtifact, TScenario>;
+//#endregion
+//#region src/campaign/gates/heldout-gate.d.ts
+interface HeldOutGateOptions<TScenario extends Scenario = Scenario> {
+  scenarios: TScenario[];
+  /** Effect-size threshold the CI lower bound must clear, in the judge's native
+   *  scale. Default 0.5. Equality holds; CI.low must be greater than this value. */
+  deltaThreshold?: number;
+  /** Bootstrap CI confidence. Default 0.95. */
+  confidence?: number;
+  /** Minimum paired holdout observations to claim significance. The exact
+   *  small-sample test may require more observations at the selected
+   *  confidence. Default 3. */
+  minProductiveRuns?: number;
+  /** Bootstrap resamples. Default 2000. */
+  resamples?: number;
+  /** Fixed bootstrap seed for deterministic verdicts. Default 1337. */
+  bootstrapSeed?: number;
+}
+/**
+ * Composable held-out gate: ships only when the lower bound of the DECIDING
+ * paired interval on the candidate-minus-baseline composite delta clears
+ * `deltaThreshold` — Tango's score interval on a pass/fail holdout, the mean
+ * bootstrap otherwise. See {@link decidePairedPromotion}.
+ */
+declare function heldOutGate<TArtifact, TScenario extends Scenario>(options: HeldOutGateOptions<TScenario>): Gate<TArtifact, TScenario>;
+//#endregion
+export { ReferenceEquivalenceJudgeInput as A, campaignSplitDigestFromIdentities as B, ExternalTextOptimizerContext as C, decodeExternalTextCandidate as D, ExternalTextEvaluationResponse as E, runReferenceEquivalenceJudge as F, assertCampaignDesign as I, assertCampaignSplitIdentity as L, ReferenceEquivalenceJudgeResult as M, ReferenceEquivalenceScenario as N, REFERENCE_EQUIVALENCE_INPUT_LIMITS as O, createReferenceEquivalenceJudge as P, campaignScenarioIdentity as R, ExternalTextOptimizationMethodConfig as S, ExternalOptimizationExample as T, GepaRunnerCommand as _, DefaultProductionRewardHackingOptions as a, OptimizerModelBudget as b, SkillOptOptimizationMethodConfig as c, skillOptOptimizationMethod as d, GepaAdaptiveEngineRun as f, GepaOptimizationRecipe as g, GepaOptimizationMethodConfig as h, DefaultProductionGateOptions as i, ReferenceEquivalenceJudgeOptions as j, REFERENCE_EQUIVALENCE_JUDGE_VERSION as k, SkillOptRunnerCommand as l, GepaEngineRun as m, heldOutGate as n, defaultProductionGate as o, GepaEngineOptions as p, DefaultProductionGateCheck as r, composeGate as s, HeldOutGateOptions as t, SkillOptTrainerConfig as u, gepaOptimizationMethod as v, ExternalTextOptimizerResult as w, externalTextOptimizationMethod as x, OpenAICompatibleOptimizerModel as y, campaignSplitDigest as z };
+//# sourceMappingURL=heldout-gate-Bn7_xWCv.d.ts.map

package/dist/heldout-gate-Bn7_xWCv.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"heldout-gate-Bn7_xWCv.d.ts","names":[],"sources":["../src/campaign/coverage.ts","../src/reference-equivalence-judge.ts","../src/campaign/external-text-evaluation.ts","../src/campaign/external-text-optimization-contract.ts","../src/campaign/external-text-optimization.ts","../src/campaign/optimizer-model.ts","../src/campaign/gepa-optimization-method.ts","../src/campaign/skillopt-optimization-method.ts","../src/campaign/gates/compose.ts","../src/campaign/gates/default-production-gate.ts","../src/campaign/gates/heldout-gate.ts"],"mappings":";;;;;;;;;iBAWgB,qBAAqB,kBAAkB,UACrD,oBAAoB,aACpB;;iBA2Bc,yBAAyB,kBAAkB,UACzD,UAAU,YACT,2BAA2B,KAAK;;iBAUnB,kCACd,oBAAoB,4BACpB;;iBAgBc,oBAAoB,kBAAkB,UACpD,oBAAoB,aACpB;;iBAOc,4BACd,oBAAoB,4BACpB,cACA;;;cC3EW;cAEA;WACX;WACA;WACA;;UAGe,qCAAqC;EACpD;EACA;;UAGe;EACf;EACA;EACA;;UAGe;;EAEf,MAAM;;EAEN;;EAEA,SAAS;;EAET,aAAa;;UAGE,wCAAwC;EACvD;EACA;EACA;EACA;;;iBA0Bc,gCACd,SAAS,mCACR,oBAAoB;;iBA2CD,6BACpB,OAAO,gCACP,SAAS,mCACR,QAAQ;;;UCrGM;EACf;EACA;;UAGe;EACf;EACA;IACE;IACA,YAAY;IACZ;IACA;;;IAGA;IACA;MACE;MACA;;;;iBAwHU,4BAA4B,WAAW,wBAAwB;;;UC3I9D;WACN;WACA;WACA;WACA;WACA;WACA,eAAe;WACf,mBAAmB;WACnB,uBAAuB;WACvB;WACA;;WAEA;WACA;WACA;WACA,QAAQ;;WAER,MAAM;WACN,WACP,SAFa,kCAGV,QAAQ;;UAGE;EACf,eAAe;EACf;EACA;IACM;;IACA;;IACA;IAAkB;;;;;;;;;UAqBT,qCACf,kBAAkB,UAClB;EAEA;EACA,QAAQ,KAAK;EACb;EACA;EACA;EACA;;EAEA;;EAEA;;EAEA,SAAS;EACT;EACA;EACA,oBAAoB,UAAU;EAC9B,oBAAoB,UAAU,WAAW,UAAU;EACnD,MAAM,SAAS,iCAAiC,QAAQ;;;;;;;;;;iBCN1C,+BAA+B,kBAAkB,UAAU,WACzE,QAAQ,qCAAqC,WAAW,aACvD,mBAAmB,WAAW;;;KCvErB,uBAAuB;;UAGlB;EACf;EACA,QAAQ;;EAER,MAAM;;EAEN;;;;;;;EAOA,oBAAoB;;;;;;;EAOpB;;;;;UCyCe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;;;;EAKA,eAAe;;;UAIA,sBAAsB;;EAErC;;;KAIU,wBAAwB;;;;;;;KAQxB;EAEN;EACA,KAAK;;EAGL;EACA,eAAe;;EAGf;EACA,eAAe;;EAEf;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;;EAGA;EACA,eAAe;EACf;;EAGA;EACA,eAAe;EACf;;EAGA;EACA,kBAAkB;EAClB,cAAc;EACd;;;KAIM,oBAAoB;UAEf,6BAA6B,kBAAkB,UAAU;;EAExE;;EAEA,QAAQ;;EAER;;EAEA;;EAEA;;;;;EAKA;;EAEA;;EAEA;;EAEA,2BAA2B,QAAQ;;EAEnC;;;;;;EAMA,YAAY;;;;;;EAMZ,oBAAoB,UAAU;;EAE9B,oBAAoB,UAAU,WAAW,UAAU;;EAEnD,SAAS;;;;;EAKT;EACA,SAAS;;;;;;;;;;EAUT;IAAiB,UAAU;IAAmB;;;iBA6ChC,uBAAuB,kBAAkB,UAAU,WACjE,QAAQ,6BAA6B,WAAW,aAC/C,mBAAmB,WAAW;;;UC7LhB;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;;;;EAKA,YAAY;;KAGF,wBAAwB;UAEnB,iCAAiC,kBAAkB,UAAU;EAC5E;;EAEA;EACA;;EAEA;EACA,SAAS;;;;;EAKT,WAAW;;EAEX;;EAEA;EACA;;EAEA;;EAEA,2BAA2B,QAAQ;EACnC;EACA,oBAAoB,UAAU;EAC9B,oBAAoB,UAAU,WAAW,UAAU;EACnD,SAAS;EACT,SAAS;;;iBAIK,2BAA2B,kBAAkB,UAAU,WACrE,QAAQ,iCAAiC,WAAW,aACnD,mBAAmB,WAAW;;;;;;iBC1GjB,YAAY,qBAAqB,kBAAkB,WAAW,aACzE,OAAO,MAAM,KAAK,WAAW,cAC/B,KAAK,WAAW;;;KCaP;KAOA,wCAAwC,KAClD;EAGA,SAAS,YAAY;;UAGN;;;EAGf,kBAAkB;;;;;EAKlB;;EAEA;;EAEA;;EAEA;;;;EAIA;;;;EAIA;;;;;EAKA;;;;EAIA;;;EAGA;;;;EAIA,iBAAiB;;;EAGjB,aAAa;;EAEb,gBAAgB;;EAEhB,SAAS;;;;;EAKT,iBAAiB;;;;;iBAMH,sBAAsB,WAAW,kBAAkB,UACjE,SAAS,+BACR,KAAK,WAAW;;;UC/EF,mBAAmB,kBAAkB,WAAW;EAC/D,WAAW;;;EAGX;;EAEA;;;;EAIA;;EAEA;;EAEA;;;;;;;;iBASc,YAAY,WAAW,kBAAkB,UACvD,SAAS,mBAAmB,aAC3B,KAAK,WAAW"}

package/dist/hosted/index.d.ts

@@ -1,5 +1,5 @@
-import { o as InsightReport } from "../insight-report-DRe8LB6d.js";
-import { _ as UnixNanoTimestamp, a as hostedTenantFromEnv, c as EvalRunGenerationSnapshot, d as HostedIngestHeaders, f as HostedWireVersion, g as TraceSpanEvent, h as IngestTracesRequest, i as hostedClientFromEnv, l as EvalRunStatus, m as IngestResponse, n as HostedTenant, o as EvalRunCellScore, p as IngestEvalRunsRequest, r as createHostedClient, s as EvalRunEvent, t as HostedClient, u as HOSTED_WIRE_VERSION } from "../client-L9VVPkim.js";
+import { o as InsightReport } from "../insight-report-08F022xN.js";
+import { _ as UnixNanoTimestamp, a as hostedTenantFromEnv, c as EvalRunGenerationSnapshot, d as HostedIngestHeaders, f as HostedWireVersion, g as TraceSpanEvent, h as IngestTracesRequest, i as hostedClientFromEnv, l as EvalRunStatus, m as IngestResponse, n as HostedTenant, o as EvalRunCellScore, p as IngestEvalRunsRequest, r as createHostedClient, s as EvalRunEvent, t as HostedClient, u as HOSTED_WIRE_VERSION } from "../client-_Fsa5c2_.js";
 import { z } from "zod";
 //#region src/hosted/schemas.d.ts
 declare const UnixNanoTimestampSchema: z.ZodType<UnixNanoTimestamp>;

package/dist/hosted/index.js

@@ -1,2 +1,2 @@
-import { a as IngestEvalRunsRequestSchema, c as InsightReportSchema, d as HOSTED_WIRE_VERSION, i as EvalRunEventSchema, l as TraceSpanEventSchema, n as hostedClientFromEnv, o as IngestResponseSchema, r as hostedTenantFromEnv, s as IngestTracesRequestSchema, t as createHostedClient, u as UnixNanoTimestampSchema } from "../client-CX7KqIdB.js";
+import { a as IngestEvalRunsRequestSchema, c as InsightReportSchema, d as HOSTED_WIRE_VERSION, i as EvalRunEventSchema, l as TraceSpanEventSchema, n as hostedClientFromEnv, o as IngestResponseSchema, r as hostedTenantFromEnv, s as IngestTracesRequestSchema, t as createHostedClient, u as UnixNanoTimestampSchema } from "../client-BvwNkIRN.js";
 export { EvalRunEventSchema, HOSTED_WIRE_VERSION, IngestEvalRunsRequestSchema, IngestResponseSchema, IngestTracesRequestSchema, InsightReportSchema, TraceSpanEventSchema, UnixNanoTimestampSchema, createHostedClient, hostedClientFromEnv, hostedTenantFromEnv };