@tangle-network/agent-eval 0.175.0 → 0.176.0

package/dist/experiment/index.js

@@ -1,16 +1,16 @@
 import { n as CaptureIntegrityError, s as ValidationError } from "../errors-Dngq5h35.js";
 import { r as canonicalString } from "../canonical-DPyQ_rpt.js";
-import { a as verifyManifest, i as signManifest, n as hashJson, r as manifestContentDigest, t as evaluateHypothesis } from "../pre-registration-D94b7Of5.js";
+import { _ as verifyManifest, g as signManifest, h as manifestContentDigest, m as hashJson, p as evaluateHypothesis } from "../agent-profile-cell-0gSi5ffD.js";
 import { t as mulberry32 } from "../random-Dn5fPWkt.js";
 import { a as requiredSampleSize, c as holm, i as requiredPairedSampleSize, n as mcnemarRequiredN, o as benjaminiHochberg, r as pairedMde, s as bonferroni, t as mcnemarPower } from "../power-and-mde-B8F2RdcD.js";
 import { f as mcnemar, g as pairedRiskDifferenceScore, h as pairedRiskDifferenceExact, m as pairedRiskDifference, n as pairArms, r as pairRunRecords, t as comparePairedArms, v as wilson } from "../paired-arms-D4aeIHUy.js";
 import { r as pairedBootstrap, t as BOOTSTRAP_GATE_MIN_N } from "../paired-tests-C8iCsioC.js";
 import { t as eProcess } from "../sequential-eprocess-D1jKoihe.js";
-import { a as heldoutSignificance, o as pairHoldout, t as powerPreflight } from "../power-preflight-CFXm0Vjo.js";
-import { c as inMemoryExperimentStore, i as ExperimentTracker, n as attest, o as fileExperimentStore, r as verifyAttestation } from "../attestation-XSUpbc4o.js";
+import { M as heldoutSignificance, N as pairHoldout, O as powerPreflight, a as createEvidenceReceipt, i as INDEPENDENT_EVIDENCE_AUTHORITY_KINDS, n as EVIDENCE_AUTHORITY_KINDS, o as isIndependentEvidence, r as EVIDENCE_RECEIPT_VERSION, s as verifyEvidenceReceipt, t as createCampaignEvidenceReceipt } from "../campaign-evidence-D8DBLqLI.js";
+import { a as inMemoryExperimentStore, r as fileExperimentStore, t as ExperimentTracker } from "../experiment-tracker-BKEumQug.js";
 import { n as pairedEvalueSequence, r as sequentialCrossingHorizon } from "../sequential-CzK5DarL.js";
-import { n as paretoPolicy, r as paretoSignificanceGate, t as buildEvidenceVector } from "../promotion-policy-LY9mVQ7W.js";
-import { n as sequentialPairedGate, t as sequentialDecide } from "../sequential-B51qAYE4.js";
+import { n as paretoPolicy, r as paretoSignificanceGate, t as buildEvidenceVector } from "../promotion-policy-DWOm70gx.js";
+import { n as sequentialPairedGate, t as sequentialDecide } from "../sequential-B5gXgcyp.js";
 import { createHash } from "node:crypto";
 import { z } from "zod";
 //#region src/experiment/ast.ts
@@ -972,128 +972,6 @@ async function openSealedExperiment(sealed) {
 	};
 }
 //#endregion
-//#region src/experiment/evidence-receipt.ts
-/**
-* Evidence receipts are the join between Runtime execution and Eval promotion.
-*
-* Runtime says what actually ran. Eval says what independently measured it.
-* This receipt binds those worlds without making either package import the other:
-* stable pursuit/run identity, exact candidate/evaluator/environment/input/output
-* content identities, and the authority class that produced the observation.
-*
-* The payload is attested with agent-eval's existing canonical report attestation.
-* Signing/key management deliberately remains outside this substrate; consumers may
-* sign the byte-stable receipt or anchor its attestation in a transparency log.
-*/
-const EVIDENCE_RECEIPT_VERSION = "1.0.0";
-/** Closed promotion vocabulary. A typo or unknown future kind is never independent by default. */
-const EVIDENCE_AUTHORITY_KINDS = [
-	"candidate-self-report",
-	"independent-evaluator",
-	"independent-replication",
-	"human-review",
-	"production-canary"
-];
-const INDEPENDENT_EVIDENCE_AUTHORITY_KINDS = [
-	"independent-evaluator",
-	"independent-replication",
-	"human-review",
-	"production-canary"
-];
-/**
-* Mint a content-attested evidence receipt. Required identity fields are deliberately
-* non-optional: unknown evidence stays unknown and cannot accidentally look certified.
-* The attestation's input provenance must equal the receipt commitment so two competing
-* descriptions of the evaluated population cannot coexist inside one valid receipt.
-*/
-function createEvidenceReceipt(input, provenance) {
-	const inputSetCommitment = requiredIdentity(input.inputSetCommitment, "inputSetCommitment");
-	assertInputCommitment(provenance, inputSetCommitment);
-	const binding = Object.freeze({
-		schemaVersion: EVIDENCE_RECEIPT_VERSION,
-		pursuitId: requiredIdentity(input.pursuitId, "pursuitId"),
-		runId: requiredIdentity(input.runId, "runId"),
-		candidateDigest: requiredIdentity(input.candidateDigest, "candidateDigest"),
-		evaluatorDigest: requiredIdentity(input.evaluatorDigest, "evaluatorDigest"),
-		environmentDigest: requiredIdentity(input.environmentDigest, "environmentDigest"),
-		inputSetCommitment,
-		outputDigest: requiredIdentity(input.outputDigest, "outputDigest"),
-		resultDigest: requiredIdentity(input.resultDigest, "resultDigest"),
-		authority: requiredAuthority(input.authority),
-		...input.experimentDigest === void 0 ? {} : { experimentDigest: requiredIdentity(input.experimentDigest, "experimentDigest") },
-		...input.observerDigest === void 0 ? {} : { observerDigest: requiredIdentity(input.observerDigest, "observerDigest") }
-	});
-	return Object.freeze({
-		binding,
-		attestation: attest(binding, provenance)
-	});
-}
-/**
-* Verify promotion-grade evidence. Generic report attestation keeps a legacy read path,
-* but an EvidenceReceipt never accepts unbound provenance: changing the evaluator code,
-* model versions, input commitment provenance, or creation record must invalidate the
-* evidence rather than merely annotating it as legacy.
-*/
-function verifyEvidenceReceipt(receipt) {
-	if (receipt.binding.schemaVersion !== "1.0.0") return {
-		valid: false,
-		reason: `unsupported evidence receipt version '${receipt.binding.schemaVersion}'`
-	};
-	try {
-		for (const [field, value] of Object.entries({
-			pursuitId: receipt.binding.pursuitId,
-			runId: receipt.binding.runId,
-			candidateDigest: receipt.binding.candidateDigest,
-			evaluatorDigest: receipt.binding.evaluatorDigest,
-			environmentDigest: receipt.binding.environmentDigest,
-			inputSetCommitment: receipt.binding.inputSetCommitment,
-			outputDigest: receipt.binding.outputDigest,
-			resultDigest: receipt.binding.resultDigest
-		})) requiredIdentity(value, field);
-		requiredAuthority(receipt.binding.authority);
-		assertInputCommitment(receipt.attestation.provenance, receipt.binding.inputSetCommitment);
-	} catch (error) {
-		return {
-			valid: false,
-			reason: error instanceof Error ? error.message : String(error)
-		};
-	}
-	const verification = verifyAttestation(receipt.binding, receipt.attestation);
-	if (!verification.valid) return verification;
-	if (verification.legacyUnboundProvenance === true || receipt.attestation.envelopeHash === void 0) return {
-		valid: false,
-		reason: "evidence receipt provenance is not bound by an attestation envelope"
-	};
-	return { valid: true };
-}
-/**
-* Promotion may choose a stricter policy, but this primitive makes the basic separation
-* explicit: only a recognized independent authority is independent. Unknown/forged kinds
-* and candidate self-reports both return false.
-*/
-function isIndependentEvidence(receipt) {
-	return INDEPENDENT_EVIDENCE_AUTHORITY_KINDS.includes(receipt.binding.authority?.kind);
-}
-function requiredAuthority(value) {
-	if (typeof value !== "object" || value === null || Array.isArray(value)) throw new TypeError("evidence receipt: authority must be an object");
-	const authority = value;
-	if (typeof authority.kind !== "string" || !EVIDENCE_AUTHORITY_KINDS.includes(authority.kind)) throw new TypeError(`evidence receipt: unknown authority kind '${String(authority.kind)}'`);
-	if (typeof authority.id !== "string") throw new TypeError("evidence receipt: authority.id must be a string");
-	return Object.freeze({
-		kind: authority.kind,
-		id: requiredIdentity(authority.id, "authority.id")
-	});
-}
-function assertInputCommitment(provenance, inputSetCommitment) {
-	if (typeof provenance?.inputsHash !== "string" || provenance.inputsHash.trim().length === 0) throw new TypeError("evidence receipt: provenance.inputsHash is required");
-	if (provenance.inputsHash.trim() !== inputSetCommitment) throw new TypeError("evidence receipt: provenance.inputsHash must equal binding.inputSetCommitment");
-}
-function requiredIdentity(value, field) {
-	const normalized = value.trim();
-	if (normalized.length === 0) throw new TypeError(`evidence receipt: ${field} must be non-empty`);
-	return normalized;
-}
-//#endregion
 //#region src/experiment/evidence-record.ts
 /**
 * The evidence registry record — the canonical, machine-readable unit for a
@@ -1435,6 +1313,6 @@ function mixSeed(seed, effect) {
 	return (seed ^ Math.round(effect * 1000003) * 2654435769) >>> 0 | 0;
 }
 //#endregion
-export { BOOTSTRAP_GATE_MIN_N, DesignRefusalError, EVIDENCE_AUTHORITY_KINDS, EVIDENCE_RECEIPT_VERSION, EVIDENCE_STATES, ExperimentTracker, FunnelIntegrityError, INDEPENDENT_EVIDENCE_AUTHORITY_KINDS, MatchedBudgetError, SealIntegrityError, amendExperiment, assertDesignAdequate, assertFunnelReconciles, assertMatchedBudgets, benjaminiHochberg, bonferroni, buildEvidenceVector, buildFunnel, classifyReissue, clusteredPower, comparePairedArms, composeFunnels, computeEstimand, computeInterval, createEvidenceReceipt, defineExperiment, eProcess, evaluateCondition, evaluateHaltRule, evaluateHypothesis, evaluateIdentityGate, evaluateOracleDeterminismGate, evaluatePopulationReproducibilityGate, evaluatePowerFloorGate, evaluatePredicate, evaluateProvenanceGate, evidenceRegistryRecordSchema, executeAdmissionRule, executeDecisionRule, fileExperimentStore, hashJson, heldoutSignificance, holm, inMemoryExperimentStore, isIndependentEvidence, manifestContentDigest, mcnemar, mcnemarPower, mcnemarRequiredN, mulberry32, openSealedExperiment, pairArms, pairHoldout, pairRunRecords, pairedBootstrap, pairedEvalueSequence, pairedMde, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, paretoPolicy, paretoSignificanceGate, parseEvidenceRegistryRecord, powerPreflight, projectNLadderBudget, readField, renderEvidenceIndex, renderFunnelTable, requiredPairedSampleSize, requiredSampleSize, runSelectionRule, runUniformPassBudget, sealExperiment, sequentialCrossingHorizon, sequentialDecide, sequentialPairedGate, signManifest, validateEvidenceRegistry, verifyEvidenceReceipt, verifyManifest, verifyMatchedBudgets, verifySealedExperiment, wilson };
+export { BOOTSTRAP_GATE_MIN_N, DesignRefusalError, EVIDENCE_AUTHORITY_KINDS, EVIDENCE_RECEIPT_VERSION, EVIDENCE_STATES, ExperimentTracker, FunnelIntegrityError, INDEPENDENT_EVIDENCE_AUTHORITY_KINDS, MatchedBudgetError, SealIntegrityError, amendExperiment, assertDesignAdequate, assertFunnelReconciles, assertMatchedBudgets, benjaminiHochberg, bonferroni, buildEvidenceVector, buildFunnel, classifyReissue, clusteredPower, comparePairedArms, composeFunnels, computeEstimand, computeInterval, createCampaignEvidenceReceipt, createEvidenceReceipt, defineExperiment, eProcess, evaluateCondition, evaluateHaltRule, evaluateHypothesis, evaluateIdentityGate, evaluateOracleDeterminismGate, evaluatePopulationReproducibilityGate, evaluatePowerFloorGate, evaluatePredicate, evaluateProvenanceGate, evidenceRegistryRecordSchema, executeAdmissionRule, executeDecisionRule, fileExperimentStore, hashJson, heldoutSignificance, holm, inMemoryExperimentStore, isIndependentEvidence, manifestContentDigest, mcnemar, mcnemarPower, mcnemarRequiredN, mulberry32, openSealedExperiment, pairArms, pairHoldout, pairRunRecords, pairedBootstrap, pairedEvalueSequence, pairedMde, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, paretoPolicy, paretoSignificanceGate, parseEvidenceRegistryRecord, powerPreflight, projectNLadderBudget, readField, renderEvidenceIndex, renderFunnelTable, requiredPairedSampleSize, requiredSampleSize, runSelectionRule, runUniformPassBudget, sealExperiment, sequentialCrossingHorizon, sequentialDecide, sequentialPairedGate, signManifest, validateEvidenceRegistry, verifyEvidenceReceipt, verifyManifest, verifyMatchedBudgets, verifySealedExperiment, wilson };
 
 //# sourceMappingURL=index.js.map