openpond 0.2.61 → 0.2.62

package/dist/web/assets/{training-managed-adapter-81RrdT1k.js → training-managed-adapter-BXNwgd3P.js}

@@ -1 +1 @@
-import{C as e,T as t,a as n,b as r,c as i,d as a,g as o,h as s,i as ee,m as c,o as l,p as u,r as d,v as f,x as p}from"./schemas-lO3MoWQr.js";import{a as te,c as m,d as h,i as g,l as _,n as ne,o as v,s as y,t as re}from"./common-BMjyKlX-.js";import{D as ie,E as b,r as ae}from"./settings-BKmRueaw.js";import{A as oe,D as se,R as x,U as ce,W as le,_ as S,d as C,h as ue,i as w,p as de,t as fe,v as pe}from"./chunk-H3JZ3P5T-B0SCoPw4.js";import{A as T,D as E,E as D,H as me,I as he,M as ge,O as _e,R as ve,U as ye}from"./contracts-BHTOD1nd.js";import{C as be,K as xe,O as Se,T as Ce,W as O,Y as we,b as Te,f as Ee,i as De,l as Oe,n as ke,r as Ae,t as je,y as Me}from"./chunk-GAIDBPFY-BLuHpurI.js";var Ne=f().trim().min(1).max(200),Pe=c({sessionId:Ne,turnId:Ne,source:de,admissionHash:v,modelConfigurationHash:v,traceHash:v,runtimeEventRefs:n(Ne).max(1e4),startedAt:m,completedAt:m,outputHash:v,partialOutputAvailable:l()}).strict().superRefine((e,t)=>{new Set(e.runtimeEventRefs).size!==e.runtimeEventRefs.length&&t.addIssue({code:`custom`,path:[`runtimeEventRefs`],message:`Native event references must be unique.`}),Date.parse(e.completedAt)<Date.parse(e.startedAt)&&t.addIssue({code:`custom`,path:[`completedAt`],message:`Native completion precedes its start.`})});c({schemaVersion:a(`openpond.nativeHarnessExperimentAttempt.v1`),taskId:f().min(1).max(500),status:d([`completed`,`cancelled`,`timed_out`,`policy_failure`]),output:f().max(8388608).nullable(),error:f().max(2e3).nullable(),messages:n(p()).max(4002),collected:l(),environmentCleanupComplete:a(!0),snapshot:ee(),native:Pe}).strict().superRefine((e,t)=>{let n=e.status===`completed`;(n!==e.collected||n!==(e.output!==null)||n!==(e.error===null))&&t.addIssue({code:`custom`,message:`Native status, output, collection and error must agree.`}),n&&(e.native.partialOutputAvailable||e.native.outputHash!==h(e.output))&&t.addIssue({code:`custom`,path:[`native`,`outputHash`],message:`Native output differs from its retained evidence.`})}).extend({contentHash:v}).strict();var Fe=c({schemaVersion:a(`openpond.humanResultView.v1`),executionId:f().min(1),executionHash:f().regex(/^[a-f0-9]{64}$/),selection:n(x).max(1e4),coverage:c({pending:u().int().nonnegative(),submitted:u().int().nonnegative(),accepted:u().int().nonnegative(),rejected:u().int().nonnegative(),cancelled:u().int().nonnegative()}).strict(),results:n(c({review:x,decision:x.nullable(),attemptIds:n(f().min(1)).min(1).max(2),mode:d([`individual`,`pairwise`]),origin:d([`hosted_retained`,`recorded_evidence`,`owner_attested_local`,`native_local`]),status:d([`pending`,`submitted`,`accepted`,`rejected`,`cancelled`]),grader:x,answers:n(oe).max(100)}).strict()).max(1e4)}).strict().extend({contentHash:f().regex(/^[a-f0-9]{64}$/)}).strict(),Ie=f().regex(/^[a-f0-9]{64}$/),Le=f().min(1).max(240),Re={feedbackKey:Le,criterionId:Le},ze=i(`kind`,[c({...Re,kind:a(`normalized_score`),minimum:u().finite(),maximum:u().finite(),passThreshold:u().min(0).max(1).optional()}).strict(),c({...Re,kind:a(`boolean`)}).strict(),c({...Re,kind:a(`category`)}).strict(),c({...Re,kind:a(`preference_score`),tieScore:u().min(0).max(1).nullable()}).strict()]),Be=c({execution:g,resultHash:Ie,reviews:n(c({review:x,decision:x}).strict()).min(1).max(1e4)}).strict(),Ve=c({rules:n(ze).min(1).max(100),baseline:Be,candidate:Be}).strict().superRefine((e,t)=>{new Set(e.rules.map(e=>e.feedbackKey)).size!==e.rules.length&&t.addIssue({code:`custom`,message:`Select each Human feedback key once.`})}),He=c({manifest:D,result:E}).strict(),Ue=c({id:Le,manifestHash:Ie,resultHash:Ie}).strict(),We=c({schemaVersion:a(`openpond.humanComparisonProjection.v1`),teamId:Le,projectId:Le,selections:Ve,original:c({baseline:Ue,candidate:Ue}).strict(),humanResults:c({baseline:Fe,candidate:Fe}).strict(),baseline:He,candidate:He}).strict().extend({contentHash:Ie}).strict();function Ge(e){throw new ye(e,409)}function Ke(e,t){let n=We.parse(e),{contentHash:r,...i}=n;r!==h(i)&&Ge(`human_comparison_projection_invalid`);for(let e of[`baseline`,`candidate`]){let r=T(n[e]),i=r.manifest.lineage,a=i?.humanProjection,o=n.original[e],s=n.selections[e];(!i||!a||a.originalManifest.id!==o.id||a.originalManifest.contentHash!==o.manifestHash||a.selectionHash!==h(s)||a.rulesHash!==h(n.selections.rules)||h(i.execution)!==h(s.execution)||r.manifest.teamId!==n.teamId||n.humanResults[e].contentHash!==s.resultHash)&&Ge(`human_comparison_projection_lineage_changed`);let{contentHash:ee,...c}=n.humanResults[e];if((h(c)!==ee||c.executionId!==s.execution.id||c.executionHash!==s.execution.contentHash||h(c.selection)!==h(s.reviews.map(e=>e.review)))&&Ge(`human_comparison_human_result_changed`),t){let n=T(t[e]);(n.manifest.lineage?.humanProjection||n.manifest.id!==o.id||n.manifest.contentHash!==o.manifestHash||n.result.contentHash!==o.resultHash||h(n.manifest.lineage?.execution)!==h(s.execution))&&Ge(`human_comparison_original_changed`)}}return n}var qe=c({adapterId:y,placement:d([`local`,`remote`,`colocated`,`provider_native`]),runtimeVersion:f().trim().min(1).max(200),capabilityReceipt:v}).strict(),Je=c({maxTurns:u().int().positive().max(1e5),timeoutMs:u().int().positive().max(864e5),maxOutputBytes:u().int().positive().max(25e7),maximumSpendUsd:u().nonnegative().nullable()}).strict();c({schemaVersion:a(`openpond.runManifest.v1`),id:y,harnessRelease:g,tasksetRelease:g,model:ie,runtimeTarget:qe,limits:Je,approval:c({id:y,contentHash:v}).strict().nullable(),createdAt:m,metadata:te}).strict().extend({contentHash:v}).strict();var Ye=c({schemaVersion:a(`openpond.attemptReceipt.v1`),id:y,runManifest:g,taskId:y,seed:f().trim().min(1).max(500),terminal:l(),failureClass:re.nullable(),outputHash:v.nullable(),traceHash:v,artifactRefs:n(ne).max(1e5),graderEvidenceRefs:n(ne).max(1e4),startedAt:m,completedAt:m,latencyMs:u().int().nonnegative(),costUsd:u().nonnegative().nullable(),legacyAttemptRef:y.nullable().default(null),metadata:te}).strict().extend({contentHash:v}).strict(),Xe=c({schemaVersion:a(`openpond.evaluationResult.v1`),id:y,runManifest:g,harnessRelease:g,tasksetRelease:g,model:ie,receiptRefs:n(g).min(1).max(1e6),attemptCount:u().int().positive(),rewardEligibleCount:u().int().nonnegative(),terminalCount:u().int().nonnegative(),meanScore:u().min(0).max(1).nullable(),failureCounts:o(re,u().int().nonnegative()),authoredMetric:le.optional(),metadata:te}).strict().extend({contentHash:v}).strict();c({schemaVersion:a(`openpond.harnessCompatibility.v1`),id:y,baseHarnessRelease:g,candidateHarnessRelease:g,tasksetRelease:g,environmentHash:v,toolContractHash:v,policyHash:v,graderInterfaceHash:v,metadata:te}).strict().extend({contentHash:v}).strict();var Ze=i(`kind`,[c({kind:a(`model`),model:ie,configurationHash:v}).strict(),c({kind:a(`fixture`)}).strict()]),Qe=i(`kind`,[c({kind:a(`taskset`),environmentRelease:g,verifierSetRelease:g,policyHash:v}).strict(),c({kind:a(`harness`),harnessRelease:g}).strict()]),$e=c({receiptId:y,taskId:y,seed:f().trim().min(1).max(500),fixtureId:y.nullable()}).strict(),et=c({schemaVersion:a(`openpond.tasksetRunManifest.v1`),id:y,tasksetRelease:g,packageHash:v,execution:Qe,profileEvaluation:ue.optional(),policy:Ze,gradingRole:a(`evaluation`),metricPolicy:ce,population:n($e).min(1).max(1e5),runtimeTarget:qe,limits:Je,createdAt:m,metadata:te}).strict().superRefine((e,t)=>{new Set(e.population.map(e=>e.receiptId)).size!==e.population.length&&t.addIssue({code:`custom`,path:[`population`],message:`Run receipt identities must be unique.`}),e.population.some(t=>t.fixtureId!==null!=(e.policy.kind===`fixture`))&&t.addIssue({code:`custom`,path:[`population`],message:`Only fixture runs require a fixture identity for every member.`}),e.profileEvaluation&&(e.execution.kind!==`harness`||e.execution.harnessRelease.id!==e.profileEvaluation.harnessRelease.id||e.execution.harnessRelease.contentHash!==e.profileEvaluation.harnessRelease.contentHash)&&t.addIssue({code:`custom`,path:[`profileEvaluation`],message:`Profile evaluation must execute its exact bound Harness release.`})}).safeExtend({contentHash:v}),k=f().trim().min(1).max(240),A=f().trim().min(1).max(5e3),tt=d([`no_training`,`prompting`,`retrieval`,`sft`,`preference`,`grpo_rft`,`sdft_opsd`,`sdpo`,`agentic_rl`]),nt=c({id:k,taskIndex:u().int().nonnegative(),label:Me,output:o(f(),p()),infrastructureError:f().trim().min(1).max(1e4).nullable(),expectedPassed:l(),expectedRewardEligible:l(),metadata:o(f(),p()).default({})}),rt=c({schemaVersion:a(`openpond.capabilityDiagnosis.v1`),summary:f().trim().min(1).max(1e4),stableBehavior:n(A).max(100).default([]),changingKnowledge:n(A).max(100).default([]),requiredContext:n(A).max(100).default([]),requiredTools:n(k).max(100).default([]),intervention:tt,trainingEligible:l(),rationale:n(A).min(1).max(100),confidence:u().min(0).max(1)}),it=c({id:k,sourceId:k,sourceTurnId:k.nullable(),split:Se,origin:d([`extracted`,`corrected`,`synthetic`,`expert_authored`]),inputPrompt:f().trim().min(1).max(1e5),expectedOutputText:f().trim().min(1).max(2e5).nullable(),rationale:A}),at=c({schemaVersion:a(`openpond.taskDesignProposal.v1`),id:k,name:f().trim().min(1).max(500),objective:f().trim().min(1).max(2e4),diagnosis:rt.default({schemaVersion:`openpond.capabilityDiagnosis.v1`,summary:`Reproduce the selected approved behavior.`,stableBehavior:[],changingKnowledge:[],requiredContext:[],requiredTools:[],intervention:`sft`,trainingEligible:!0,rationale:[`The selected examples were supplied as demonstrations.`],confidence:.5}),taskKind:Ce.shape.taskKind,sourceIds:n(k).min(1).max(1e5),assumptions:n(A).max(1e3),successCriteria:n(A).min(1).max(1e3),proposedGraders:n(Te).max(1e3).default([]),graderFixtures:n(nt).max(1e5).default([]),generatedFiles:n(De).max(1e3).default([]),proposedExamples:n(it).max(1e5).default([]),proposedMethod:Oe.shape.recommendedMethod,trainingPath:Ee.nullable().default(null),policy:be,warnings:n(f().trim().min(1)).default([]),createdAt:f().trim().min(1)}),j=f().regex(/^[a-f0-9]{64}$/),ot=c({providerId:a(`claude-code`),configurationHash:j,executableHash:j,version:f().regex(/^2\.\d+\.\d+$/),capabilityHash:j,maximumRequestCostUsd:u().finite().positive().max(1e4),requestTimeoutMs:u().int().min(1e3).max(3e5),maximumTurns:u().int().min(1).max(100),holdOpen:l()}).strict();c({id:f().min(1).max(200),receiptId:f().min(1).max(200),sessionId:e(),operationId:f().min(1).max(200),action:d([`message`,`seal`]),text:f().max(262144).optional()}).strict().superRefine((e,t)=>{e.action===`message`&&!e.text?.trim()&&t.addIssue({code:`custom`,message:`A human intervention requires a message.`}),e.action===`seal`&&e.text!==void 0&&t.addIssue({code:`custom`,message:`Sealing does not add a prompt.`})});var st=c({provider:a(`claude-code`),sessionId:e(),requestId:f().min(1).max(200),modelId:f().min(1).max(500),runtime:ot,sourceHash:j,outputHash:j,traceHash:j,runtimeEventRefs:n(f().min(1).max(200)).min(1).max(2e4),eventCount:u().int().positive().max(2e4),startedAt:t(),completedAt:t(),cleanupComplete:a(!0)}).strict().superRefine((e,t)=>{(e.runtimeEventRefs.length!==e.eventCount||new Set(e.runtimeEventRefs).size!==e.eventCount)&&t.addIssue({code:`custom`,message:`Process event references must match the exact retained trace population.`})}),ct=c({providerId:a(`codex`),configurationHash:f().regex(/^[a-f0-9]{64}$/),accountHash:f().regex(/^[a-f0-9]{64}$/),reasoningEffort:d([`low`,`medium`,`high`,`xhigh`]),maximumRequestCostUsd:u().finite().positive().max(1e4),requestTimeoutMs:u().int().min(1e3).max(3e5)}).strict(),lt=c({id:f().trim().min(1).max(240),revision:u().int().positive(),contentHash:f().regex(/^[a-f0-9]{64}$/),targetId:f().trim().min(1).max(240).nullable()}).strict(),ut=n(c({role:d([`system`,`user`,`assistant`]),content:f().max(1e5)}).strict()).max(100).superRefine((e,t)=>{e.reduce((e,t)=>e+t.content.length,0)>1e5&&t.addIssue({code:`custom`,message:`Evaluation prompt exceeds 100,000 characters.`})}),dt=c({schemaVersion:a(`openpond.experimentAttemptGrade.v1`),selectionHash:f().regex(/^[a-f0-9]{64}$/),grades:n(me).min(1).max(100),score:u().min(0).max(1).nullable(),passed:l(),rewardEligible:l(),gradingStatus:d([`scored`,`pending`,`unscorable`,`not_configured`]),failureClass:fe.nullable(),contentHash:f().regex(/^[a-f0-9]{64}$/)}).strict().superRefine((e,t)=>{let{contentHash:n,...r}=e;C(r)!==n&&t.addIssue({code:`custom`,message:`Experiment grading integrity failed.`});let i=e.grades.flatMap(e=>e.components);(i.length!==e.grades.length||new Set(i.map(e=>e.graderId)).size!==i.length)&&t.addIssue({code:`custom`,message:`Each selected criterion requires exactly one distinct grader result.`});for(let n of e.grades){let{contentHash:e,...r}=n;C(r)!==e&&t.addIssue({code:`custom`,message:`A retained criterion changed.`});for(let e of n.components){let{contentHash:n,...r}=e;C(r)!==n&&t.addIssue({code:`custom`,message:`A retained grader result changed.`})}}}),M=f().trim().min(1).max(200),N=f().regex(/^[a-f0-9]{64}$/),ft=i(`kind`,[c({kind:a(`hosted_chat`),modelId:M,localRuntime:r([c({providerId:a(`custom-openai-compatible`),configurationHash:N,maximumRequestCostUsd:u().finite().positive().max(1e4),requestTimeoutMs:u().int().min(1e3).max(3e5)}).strict(),ot,ct]).optional(),candidate:c({artifactId:M,contentHash:N,preparationId:M,dispatchId:M,baseProfileId:M,workerImageDigest:f().regex(/^sha256:[a-f0-9]{64}$/)}).strict().optional(),harness:de.optional(),messages:ut.optional(),maxOutputTokens:u().int().min(1).max(1e6).default(1e5),temperature:u().min(0).max(2).default(0),topP:u().gt(0).max(1).default(1)}).strict(),c({kind:a(`fixture`),fixtureId:M}).strict()]),pt=c({schemaVersion:a(`openpond.modelStarterAttemptRequest.v1`),operationId:M,teamId:M,modelProjectId:M.nullable(),project:lt.optional(),taskset:O,taskId:M,environmentSeed:u().int().min(0).max(2147483647).default(0),policy:ft}).strict();c({modelProjectId:M.nullable(),taskset:O,afterTaskId:M.optional(),limit:u().int().min(1).max(100).default(25)}).strict(),c({modelProjectId:M.nullable(),taskset:O,available:l(),unavailableReason:f().max(1e3).nullable(),models:n(c({id:M,name:f().max(500),maximumOutputTokens:u().int().min(1).max(1e6).optional()}).strict()).max(200),tasks:n(c({id:M,split:se,inputPreview:f().max(500),fixtures:n(c({id:M,label:f().max(200)}).strict()).max(1e3)}).strict()).max(100),nextCursor:M.nullable()}).strict();var mt=c({schemaVersion:a(`openpond.modelStarterAttemptSummary.v1`),id:M,revision:u().int().positive(),request:pt,status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]),policySnapshot:c({modelId:M,provider:M,upstreamModelId:f().min(1).max(500),configurationHash:N}).strict().nullable(),createdAt:t(),startedAt:t().nullable(),completedAt:t().nullable(),cleanupComplete:l(),resultAvailable:l(),score:u().min(0).max(1).nullable(),gradingStatus:d([`not_started`,`pending`,`scored`,`unscorable`,`not_configured`]).default(`not_started`),passed:l().nullable(),outputPreview:f().max(500).nullable(),error:c({code:M,message:f().max(2e3)}).strict().nullable()}).strict().superRefine((e,t)=>{e.request.policy.kind===`fixture`!=(e.policySnapshot===null)&&t.addIssue({code:`custom`,path:[`policySnapshot`],message:`Fixture execution must not claim a model identity.`}),e.status===`completed`&&(!e.cleanupComplete||!e.resultAvailable||!e.completedAt)&&t.addIssue({code:`custom`,message:`Completed attempts require retained results and confirmed cleanup.`})});c({modelProjectId:M,afterId:M.optional(),limit:u().int().min(1).max(100).default(25)}).strict(),c({items:n(mt).max(100),nextCursor:M.nullable()}).strict(),c({schemaVersion:a(`openpond.modelStarterAttemptResult.v1`),attempt:mt,output:f().max(1048576).nullable(),messages:n(o(f(),p())).max(4002),composition:he.nullable(),grade:me.nullable().optional(),experimentGrade:dt.nullable().optional(),environment:c({status:d([`completed`,`budget_exhausted`,`cancelled`,`timed_out`,`policy_failure`,`environment_failure`]),collected:l(),definition:O,initialStateHash:N.nullable(),finalStateHash:N.nullable(),attemptHash:N}).strict(),native:Pe.optional(),providerRequestIds:n(M).max(1001),contentHash:N}).strict().superRefine((e,t)=>{let n=e.attempt.request.policy.kind===`hosted_chat`?e.attempt.request.policy.harness:void 0;!!n!=!!e.native&&t.addIssue({code:`custom`,path:[`native`],message:`Standalone Harness results require their exact native evidence.`}),e.native&&(!n||C(n)!==C(e.native.source)||e.native.modelConfigurationHash!==e.attempt.policySnapshot?.configurationHash||e.environment.initialStateHash!==null||e.environment.finalStateHash!==null||e.environment.status===`completed`&&(!e.environment.collected||e.output===null||e.native.partialOutputAvailable||e.native.outputHash!==C(e.output)))&&t.addIssue({code:`custom`,path:[`native`],message:`Native evidence differs from the admitted source, model or output.`}),e.experimentGrade&&((e.grade||e.composition)&&t.addIssue({code:`custom`,message:`Selected Experiment grading cannot claim another grade population.`}),(![`completed`,`failed`].includes(e.attempt.status)||!e.attempt.resultAvailable||e.attempt.score!==e.experimentGrade.score||e.attempt.gradingStatus!==e.experimentGrade.gradingStatus||e.attempt.passed!==(e.experimentGrade.score===null?null:e.experimentGrade.passed))&&t.addIssue({code:`custom`,message:`Experiment grading differs from its retained attempt summary.`})),e.grade&&(e.composition&&t.addIssue({code:`custom`,path:[`grade`],message:`An ordinary grade cannot also claim a Reward composition.`}),(![`completed`,`failed`].includes(e.attempt.status)||!e.attempt.resultAvailable||e.attempt.score!==e.grade.score||e.attempt.gradingStatus!==e.grade.gradingStatus||e.attempt.passed!==(e.grade.score===null?null:e.grade.passed))&&t.addIssue({code:`custom`,path:[`grade`],message:`Ordinary grade differs from its terminal attempt summary.`}))});var P=f().trim().min(1).max(240),ht=f().regex(/^[a-f0-9]{64}$/),gt=c({maxOutputTokens:u().int().min(1).max(1e6).default(4096),temperature:u().min(0).max(2).default(0),topP:u().gt(0).max(1).default(1)}).strict(),F=c({kind:a(`hosted_harness`),modelId:P,maxOutputTokens:gt.shape.maxOutputTokens.removeDefault().optional(),temperature:gt.shape.temperature.removeDefault().optional(),topP:gt.shape.topP.removeDefault().optional(),localRuntime:ct.optional(),candidate:c({artifactId:P,contentHash:ht,preparationId:P,dispatchId:P,baseProfileId:P,workerImageDigest:f().regex(/^sha256:[a-f0-9]{64}$/)}).strict().optional(),sourceCandidate:ue.shape.sourceCandidate,reasoningEffort:d([`off`,`low`,`medium`,`high`,`xhigh`,`max`]).optional(),profileRepositoryId:P,source:ue,modelConfigurationHash:ht,packageHash:ht,externalDatasetBinding:S.optional()}).strict().superRefine((e,t)=>{JSON.stringify(e.externalDatasetBinding??null)!==JSON.stringify(e.source.externalDatasetBinding??null)&&t.addIssue({code:`custom`,message:`The policy and source must bind the same external Dataset recipe.`}),JSON.stringify(e.sourceCandidate??null)!==JSON.stringify(e.source.sourceCandidate??null)&&t.addIssue({code:`custom`,message:`The policy and prepared source must bind the same actual source candidate.`})}),I=f().trim().min(1).max(200),_t=f().regex(/^[a-f0-9]{64}$/),vt=i(`kind`,[ft.options[0],c({kind:a(`fixture`)}).strict(),F]),yt=c({schemaVersion:a(`openpond.modelTasksetRunRequest.v1`),operationId:I,teamId:I,modelProjectId:I.nullable(),name:f().trim().min(1).max(200).optional(),project:lt.optional(),taskset:O,policy:vt,population:n($e).min(1).max(1e4)}).strict().superRefine((e,t)=>{new Set(e.population.map(e=>e.receiptId)).size!==e.population.length&&t.addIssue({code:`custom`,path:[`population`],message:`Run receipt identities must be unique.`}),e.population.some(t=>t.fixtureId!==null!=(e.policy.kind===`fixture`))&&t.addIssue({code:`custom`,path:[`population`],message:`Only fixture runs require a fixture identity for every member.`}),e.population.some(e=>!/^(0|[1-9][0-9]*)$/.test(e.seed)||Number(e.seed)>2147483647)&&t.addIssue({code:`custom`,path:[`population`],message:`Hosted environment seeds must be integers from 0 to 2147483647.`}),e.policy.kind===`hosted_harness`&&e.modelProjectId!==null&&t.addIssue({code:`custom`,path:[`modelProjectId`],message:`A Harness target retains its released Profile binding separately from Model artifacts.`}),e.policy.kind===`hosted_harness`&&new Set(e.population.map(e=>xe([e.taskId,e.seed]))).size!==e.population.length&&t.addIssue({code:`custom`,path:[`population`],message:`Harness case identities must be unique for each task and seed.`})}),bt=c({schemaVersion:a(`openpond.modelTasksetRunSummary.v1`),id:I,revision:u().int().positive(),teamId:I,modelProjectId:I.nullable(),name:f().trim().min(1).max(200).optional(),project:lt.optional(),operationId:I,taskset:O,policyKind:d([`hosted_chat`,`fixture`,`hosted_harness`]),manifestHash:_t,status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]),totalCount:u().int().min(1).max(1e4),counts:c({pending:u().int().nonnegative(),running:u().int().nonnegative(),completed:u().int().nonnegative(),failed:u().int().nonnegative(),cancelled:u().int().nonnegative()}).strict(),createdAt:t(),startedAt:t().nullable(),completedAt:t().nullable(),cleanupComplete:l(),resultAvailable:l(),score:u().min(0).max(1).nullable(),metricName:f().min(1).max(240),error:c({code:I,message:f().max(2e3)}).strict().nullable()}).strict().superRefine((e,t)=>{Object.values(e.counts).reduce((e,t)=>e+t,0)!==e.totalCount&&t.addIssue({code:`custom`,path:[`counts`],message:`Run counts must account for the complete population.`}),e.status!==`completed`&&e.score!==null&&t.addIssue({code:`custom`,path:[`score`],message:`Only completed evaluations publish a metric value.`}),[`completed`,`failed`,`cancelled`].includes(e.status)&&(!e.completedAt||e.counts.pending||e.counts.running)&&t.addIssue({code:`custom`,message:`Terminal runs require complete population accounting.`}),e.status===`cancelled`&&!e.cleanupComplete&&t.addIssue({code:`custom`,message:`Cancelled runs require confirmed cleanup.`}),e.resultAvailable&&![`completed`,`failed`,`cancelled`].includes(e.status)&&t.addIssue({code:`custom`,message:`Run results require terminal population accounting.`}),e.status===`completed`&&(!e.cleanupComplete||!e.resultAvailable)&&t.addIssue({code:`custom`,message:`Completed evaluations require retained results and confirmed cleanup.`})}),xt=c({summary:bt,request:yt,manifest:et,policySnapshot:mt.shape.policySnapshot}).strict();c({modelProjectId:I.optional(),projectId:lt.shape.id.optional(),afterId:I.optional(),limit:u().int().min(1).max(100).default(25)}).strict(),c({items:n(bt).max(100),nextCursor:I.nullable()}).strict(),c({schemaVersion:a(`openpond.modelTasksetRunResult.v1`),run:xt,receipts:n(Ye).min(1).max(1e4),metric:le.nullable(),contentHash:_t}).strict();var St=f().min(1).max(200).refine(e=>![`__proto__`,`prototype`,`constructor`].includes(e),`Use an ordinary field name.`),Ct=n(St).min(1).max(20),wt=d([`input`,`output`,`expectedOutput`,`evaluatorContext`]),Tt=c({destination:wt,path:Ct,source:wt,sourcePath:n(St).max(20)}).strict(),Et=n(Tt).max(100).superRefine((e,t)=>{for(let[n,r]of e.entries())e.slice(0,n).some(e=>e.destination===r.destination&&(Dt(e.path,r.path)||Dt(r.path,e.path)))&&t.addIssue({code:`custom`,path:[n,`path`],message:`Mapped destination paths must not overlap.`})});function Dt(e,t){return e.length<=t.length&&e.every((e,n)=>e===t[n])}var L=f().trim().min(1).max(200),R=f().regex(/^[a-f0-9]{64}$/),Ot=c({id:L,revision:u().int().positive(),contentHash:R}).strict(),kt=c({id:L,version:f().min(1).max(200),contentHash:R,name:f().trim().min(1).max(500).optional(),feedbackKey:ge,release:Ot.nullable(),mappings:Et.optional()}).strict(),At=c({id:L,version:f().min(1).max(200),contentHash:R,mappings:Et}).strict(),jt=c({operationId:L,id:L.optional(),expectedRevision:u().int().nonnegative(),request:yt,maximumCostUsd:u().finite().positive().max(1e4),graders:n(At).min(1).max(100).optional()}).strict().superRefine((e,t)=>{e.request.name||t.addIssue({code:`custom`,path:[`request`,`name`],message:`Name this Experiment before saving it.`}),!!e.id!=e.expectedRevision>0&&t.addIssue({code:`custom`,path:[`expectedRevision`],message:`An update requires its saved definition revision.`}),Math.round(e.maximumCostUsd*1e6)<1&&t.addIssue({code:`custom`,path:[`maximumCostUsd`],message:`The budget must be at least $0.000001.`}),e.graders&&new Set(e.graders.map(e=>e.id)).size!==e.graders.length&&t.addIssue({code:`custom`,path:[`graders`],message:`Select each grader once.`})});c({schemaVersion:a(`openpond.experimentDefinition.v1`),id:L,teamId:L,revision:u().int().positive(),request:yt,maximumCostUsd:u().finite().positive().max(1e4),graders:n(kt).min(1).max(100),createdAt:t(),updatedAt:t()}).strict().extend({contentHash:R}),c({operationId:L,definition:Ot}).strict(),c({projectId:L.optional(),datasetHash:R.optional(),search:f().trim().max(200).optional(),status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]).optional(),afterId:L.optional(),limit:u().int().min(1).max(100).default(30)}).strict(),c({definition:Ot.nullable(),configurationHash:R.optional(),admissionRequestHash:R.optional(),sourceExperimentId:L.optional(),maximumCostUsd:u().finite().positive().max(1e4),graders:n(kt).min(1).max(100)}).strict().superRefine((e,t)=>{e.definition===null&&(!e.configurationHash||!e.admissionRequestHash)&&t.addIssue({code:`custom`,path:[`configurationHash`],message:`A flat Experiment requires its sealed configuration hash.`})}),c({operationId:L,profileRepositoryId:L,definitionId:L,modelId:L,localRuntime:F.shape.localRuntime,maxOutputTokens:F.shape.maxOutputTokens,temperature:F.shape.temperature,topP:F.shape.topP,reasoningEffort:d([`off`,`low`,`medium`,`high`,`xhigh`,`max`]).optional(),maximumCostUsd:u().finite().positive().max(1e4),externalDatasetBinding:S.optional(),sourceCandidate:F.shape.sourceCandidate}).strict(),c({request:yt,manifest:et,maximumCostUsd:u().finite().positive().max(1e4),graders:n(kt).min(1).max(100)}).strict();var z=f().trim().min(1).max(240),B=f().trim().min(1),Mt=f().trim().min(8).max(256),Nt=o(f(),p()).default({}),V=z.nullable(),Pt=d([`slash_train`,`session_menu`,`bulk_selection`,`training_page`,`task_candidate`]),Ft=d([`defaults`,`customize`]),It=d([`automated`,`manual`]);c({schemaVersion:a(`openpond.trainingSourceEstimate.v1`),sessionId:z,messageCount:u().int().nonnegative(),estimatedTokens:u().int().nonnegative(),textBytes:u().int().nonnegative()}),c({query:f().max(500).default(``),offset:u().int().nonnegative().default(0),limit:u().int().min(1).max(100).default(20),candidates:n(c({sessionId:z,title:f().trim().min(1).max(500),updatedAt:B})).max(500).default([])});var Lt=c({sessionId:z,title:f().trim().min(1).max(500),updatedAt:B,snippet:f().max(2e3).nullable()});c({schemaVersion:a(`openpond.trainingChatSearchResult.v1`),query:f(),offset:u().int().nonnegative(),limit:u().int().positive(),total:u().int().nonnegative(),hasMore:l(),indexedChats:u().int().nonnegative(),totalChats:u().int().nonnegative(),indexing:l(),entries:n(Lt)});var Rt=d([`policy_failure`,`grader_failure`,`environment_failure`,`infrastructure_failure`,`timeout`,`cancelled`]),zt=c({graderId:z.nullable(),path:f().trim().min(1).max(1e3),language:d([`javascript`,`typescript`,`python`,`json`,`text`]),content:f().max(2e6),sha256:Mt,declaredSha256:Mt.nullable(),integrity:d([`verified`,`unverified`,`mismatch`])}),Bt=c({protocolVersion:f().trim().min(1).max(200),module:f().trim().min(1).max(1e3),moduleSha256:Mt,command:n(f().max(2e3)).max(100),cwd:f().trim().min(1).max(4e3),maxTurns:u().int().positive().max(100)});c({schemaVersion:a(`openpond.tasksetGraderDetails.v1`),taskset:_,sourceTasksetId:z,graders:n(Te).max(1e3),runtime:Bt.nullable(),sources:n(zt).max(1e3),unavailableReason:f().max(2e3).nullable()});var Vt=c({schemaVersion:a(`openpond.taskAttempt.v1`),id:z,tasksetId:z,taskId:z,split:Se,attempt:u().int().nonnegative(),seed:u().int(),modelRef:b.nullable(),startedAt:B,completedAt:B,output:o(f(),p()),runtimeEventRefs:n(z).max(1e4).default([]),artifactRefs:n(z).max(1e4).default([]),privilegedOutcomeRef:V,infrastructureError:f().trim().min(1).max(1e4).nullable(),costUsd:u().nonnegative().nullable(),latencyMs:u().int().nonnegative(),userInterventions:u().int().nonnegative().default(0),metadata:Nt}),Ht=c({schemaVersion:a(`openpond.taskAttemptArtifact.v1`),id:z,tasksetId:z,taskId:z,attemptId:z,kind:d([`raw_model_response`,`runtime_trace`,`environment_state`,`grader_evidence`,`output_artifact`]),path:f().trim().min(1).max(4e3),mediaType:f().trim().min(1).max(200).nullable().optional(),sha256:Mt,sizeBytes:u().int().nonnegative(),createdAt:B,metadata:Nt}),Ut=c({graderId:z,graderVersion:f().trim().min(1).max(100),score:u().min(0).max(1).nullable(),passed:l(),hardGate:l(),rewardEligible:l(),feedback:f().trim().max(2e4).nullable(),evidenceRefs:n(z).max(1e4).default([]),judge:b.nullable().optional().default(null),calibrationStatus:d([`not_applicable`,`pending`,`passed`,`failed`])}),Wt=c({schemaVersion:a(`openpond.gradeResult.v1`),id:z,attemptId:z,graderSetHash:Mt,score:u().min(0).max(1).nullable(),passed:l(),components:n(Ut).min(1).max(1e3),failureClass:Rt.nullable(),feedback:n(f().trim().min(1).max(2e4)).max(1e3).default([]),rewardEligible:l(),rewardComposition:he.optional(),createdAt:B});d([`stable_hash_top_n`,`rft_easy_curriculum_v1`]);var Gt=c({schemaVersion:a(`openpond.graderAuditReport.v1`),id:z,tasksetId:z,tasksetHash:Mt,fixtureRefs:n(z).min(1).max(1e5),gradeRefs:n(z).min(1).max(1e5),passed:l(),hackingChecksPassed:l(),leakageChecksPassed:l(),infrastructureSafetyPassed:l(),failures:n(c({fixtureId:z,label:Me,gradeId:z,reason:f().trim().min(1).max(5e3)})).max(1e5),createdAt:B});function Kt(e,t=`New model`){let n=e?.trim().split(/\s+/).filter(Boolean)??[];return n.length===0?t:n.slice(0,5).join(` `)}var qt=c({schemaVersion:a(`openpond.baseModelPreference.v1`),modelId:z,revision:f().trim().min(1).max(256).nullable(),tokenizerRevision:f().trim().min(1).max(256).nullable(),chatTemplateHash:f().trim().min(8).max(256).nullable(),modelAssetId:V,source:d([`managed`,`local`,`builtin`])}),Jt=c({schemaVersion:a(`openpond.taskCreationRequest.v1`),id:z,profileId:z,surface:Pt,mode:Ft,entryMode:It.default(`manual`),resourceIntent:d([`workproduct`,`dataset`]).default(`workproduct`),buildIntent:ke.default(`demonstrations`),buildSpecification:Ae.nullable().default(null),objective:f().trim().min(1).max(2e4).nullable(),methodHint:d([`sft`,`dpo`,`grpo`,`ppo`]).nullable().default(null),preferredBaseModelId:z.nullable().default(null),preferredBaseModel:qt.nullable().default(null),sourceIds:n(z).max(1e5),candidateId:V,analysisModel:b.nullable(),analysisReasoningEffort:ae.nullable().default(null),createImproveRunId:V.default(null),targetIntent:c({kind:d([`agent`,`skill`,`extension`,`model`,`configuration`]).nullable(),id:V,displayName:f().trim().min(1).max(500).nullable(),operation:d([`create`,`improve`])}).default({kind:`model`,id:null,displayName:null,operation:`create`}),disclosure:c({status:d([`not_required`,`pending`,`approved`,`declined`]),content:a(`raw_excerpts`),sourceIds:n(z).max(1e5),providerModel:b.nullable(),approvalId:V,approvedAt:B.nullable()}).default({status:`not_required`,content:`raw_excerpts`,sourceIds:[],providerModel:null,approvalId:null,approvedAt:null}),createdAt:B}),Yt=c({schemaVersion:a(`openpond.taskCreationSnapshot.v1`),id:z,request:Jt,state:d([`planning`,`awaiting_disclosure_approval`,`awaiting_questions`,`recommendation_ready`,`awaiting_materialization_approval`,`materializing`,`validating`,`ready`,`blocked`,`failed`,`cancelled`]),proposal:at.nullable(),materializedTasksetId:V,disclosureApprovalId:V,materializationApprovalId:V,blockingQuestions:n(c({id:z,kind:d([`objective`,`consent`,`success_signal`,`privacy_licensing`,`interpretation`]),prompt:f().trim().min(1).max(5e3),answer:f().trim().min(1).max(2e4).nullable()})).default([]),transcript:n(c({id:z,role:d([`user`,`assistant`,`system`,`tool`]),text:f().max(1e5),createdAt:B})).max(1e4).default([]),repairHistory:n(je).max(1e3).default([]),blockedReason:f().trim().min(1).max(1e4).nullable(),createdAt:B,updatedAt:B});c({schemaVersion:a(`openpond.taskCreationTranscript.v1`),creationId:z,profileId:z,messages:Yt.shape.transcript,updatedAt:B});var Xt=f().trim().min(1).max(200),Zt=c({operationId:Xt,execution:w,executionKind:a(`recorded_evidence`).optional(),graders:n(ve).min(1).max(100),maximumCostUsd:u().finite().min(1e-6).max(1e4),mappings:n(c({graderId:Xt,fields:Et}).strict()).max(100).optional()}).strict().superRefine((e,t)=>{new Set(e.graders.map(e=>e.id)).size!==e.graders.length&&t.addIssue({code:`custom`,path:[`graders`],message:`Select each grader once.`}),e.mappings&&(new Set(e.mappings.map(e=>e.graderId)).size!==e.mappings.length||e.mappings.some(t=>!e.graders.some(e=>e.id===t.graderId)))&&t.addIssue({code:`custom`,path:[`mappings`],message:`Map each selected grader at most once.`})});c({schemaVersion:a(`openpond.experimentScoringPass.v1`),id:Xt,teamId:Xt,revision:u().int().positive(),request:Zt,graders:n(kt).min(1).max(100),status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]),totalCount:u().int().min(1).max(1e4),counts:c({pending:u().int().nonnegative(),running:u().int().nonnegative(),scored:u().int().nonnegative(),unavailable:u().int().nonnegative(),failed:u().int().nonnegative(),cancelled:u().int().nonnegative()}).strict(),resultAvailable:l(),createdAt:t(),startedAt:t().nullable(),completedAt:t().nullable(),error:c({code:Xt,message:f().max(2e3)}).strict().nullable()}).strict().extend({contentHash:f().regex(/^[a-f0-9]{64}$/)});var Qt=c({sourcePass:w,graders:Zt.shape.graders,mappings:Zt.shape.mappings,maximumCostUsd:u().finite().positive().max(1e4)}).strict(),$t=c({externalDatasetBinding:S,scoring:Qt.optional(),modelId:f().min(1).max(300),maximumCostUsd:u().finite().positive().max(1e4)}).strict(),en=c({source:d([`local`,`github`,`openpond_git`]),repositoryId:f().trim().min(1),profileId:f().trim().min(1)}).strict(),H=f().min(1).max(240),U=f().regex(/^[a-f0-9]{64}$/),W=c({schemaVersion:a(`openpond.experimentImprovement.v1`),id:H,operationId:H,requestHash:U,actorId:H,teamId:H,projectId:H.nullable(),ownerWorkspaceId:H,ownerRevision:u().int().nonnegative(),ownerSourceRevision:f().min(1).max(500),baseRelease:w,profileRef:en,baseProfileSourceRevision:f().min(1).max(500),evidence:c({manifest:D,result:E}).strict(),component:c({kind:d([`skill`,`workflow`,`agent`,`instruction`]),path:f().min(1).max(2e3),workflowId:H.optional()}).strict(),mode:d([`manual`,`llm_assisted`]),status:d([`draft`,`authoring_queued`,`authoring`,`frozen`,`testing`,`ready_for_review`,`adopting`,`accepted`,`rolled_back`,`declined`,`failed`,`cancelled`]),revision:u().int().positive(),limits:c({maximumCostUsd:u().finite().positive(),maximumAuthoringSteps:u().int().min(1).max(500),maximumDurationMs:u().int().min(1e3).max(864e5)}).strict(),authoringSteps:u().int().nonnegative().max(500),work:c({sessionId:H,turnId:H,revision:u().int().positive(),startedAt:f().datetime()}).strict().nullable(),frozen:c({release:w,sourceRevision:f().min(1).max(500),profileSourceRevision:f().min(1).max(500),authoringHash:U,partitionHash:U,executableSourceHash:U,protectedClosureHash:U,candidateRevision:u().int().positive(),diff:n(c({path:f(),before:U.nullable(),after:U.nullable()}).strict()).max(1e4)}).strict().nullable(),test:c({recipe:$t,recipeHash:f().regex(/^[a-f0-9]{64}$/),baselineOperationId:f(),candidateOperationId:f(),cancelRequestedAt:f().datetime().nullable(),dispatching:c({kind:d([`baseline`,`candidate`]),token:H,pid:u().int().positive(),processStart:f().nullable(),startedAt:f().datetime()}).strict().nullable()}).strict().nullable(),comparison:c({baseline:c({manifest:D,result:E}).strict(),candidate:c({manifest:D,result:E}).strict(),frozenHash:U,humanProjection:We.optional()}).strict().nullable(),adoptionIntent:c({operationId:H,expectedOwnerRevision:u().int().nonnegative(),frozenHash:U,humanProjection:We.optional()}).strict().nullable(),acceptance:w.nullable(),rollback:w.nullable(),error:f().max(2e3).nullable(),createdAt:f().datetime(),updatedAt:f().datetime(),contentHash:U}).strict();function tn(e){let t=W.parse(e),{contentHash:n,...r}=t;if(C(r)!==n)throw Error(`Improvement state integrity failed.`);if(T(t.evidence),t.evidence.manifest.teamId!==t.teamId||t.evidence.result.status!==`completed`||!t.evidence.result.cases.some(e=>e.status===`completed`))throw Error(`Improvement requires usable completed owner evidence.`);if(t.test&&(C(t.test.recipe)!==t.test.recipeHash||pe(t.test.recipe.externalDatasetBinding).profileId!==t.profileRef.profileId))throw Error(`The test recipe changed its exact selected Profile.`);if(t.comparison?.humanProjection){let e=Ke(t.comparison.humanProjection);if(e.teamId!==t.teamId||e.projectId!==t.projectId||C(e.baseline)!==C(t.comparison.baseline)||C(e.candidate)!==C(t.comparison.candidate))throw Error(`Human comparison projection differs from its exact owner evidence.`)}if(t.comparison&&(T(t.comparison.baseline),T(t.comparison.candidate),!t.frozen||t.comparison.frozenHash!==C(t.frozen)||!_e(t.comparison.baseline,t.comparison.candidate).comparable))throw Error(`Improvement comparison is not bound to a comparable frozen candidate.`);if(t.status===`ready_for_review`&&(!t.frozen||!t.comparison))throw Error(`Improvement is missing qualification.`);if(t.status===`adopting`&&(!t.adoptionIntent||!t.frozen||t.adoptionIntent.frozenHash!==C(t.frozen)))throw Error(`Improvement adoption intent is not bound to its tested revision.`);if(t.status===`accepted`&&(!t.acceptance||!t.comparison||!t.frozen))throw Error(`Improvement acceptance is missing its exact qualification receipt.`);if(t.status===`rolled_back`&&(!t.acceptance||!t.rollback))throw Error(`Profile rollback is missing its durable owner receipt.`);return t}var nn=c({states:n(W).max(50),nextCursor:H.nullable()}).strict(),rn=c({operationId:H,ownerWorkspaceId:H,expectedOwnerRevision:u().int().nonnegative(),baseRelease:w,profileRef:en,profileSourceRevision:f().min(1).max(500),projectId:H.nullable(),evidence:w,component:W.shape.component,mode:W.shape.mode,limits:W.shape.limits}).strict(),an=i(`operation`,[c({operation:a(`list`),limit:u().int().min(1).max(50).default(30),cursor:H.optional()}).strict(),c({operation:a(`options`),evidence:w}).strict(),c({operation:a(`start`),request:rn}).strict(),c({operation:d([`read`,`humanComparisonOptions`]),id:H}).strict(),c({operation:a(`instructions`),id:H}).strict(),c({operation:a(`edit`),id:H,revision:u().int().positive(),expectedFileHash:U,text:f().max(25e4)}).strict(),c({operation:a(`mode`),id:H,revision:u().int().positive(),mode:d([`manual`,`llm_assisted`])}).strict(),c({operation:a(`author`),id:H,revision:u().int().positive(),modelId:H,prompt:f().min(1).max(2e4)}).strict(),c({operation:d([`freeze`,`finishAuthoring`,`useCandidate`,`discard`,`cancel`,`rollback`]),id:H,revision:u().int().positive()}).strict(),c({operation:a(`test`),id:H,revision:u().int().positive(),recipe:$t}).strict(),c({operation:a(`compare`),id:H,revision:u().int().positive(),baseline:w,candidate:w,humanSelections:Ve.optional()}).strict()]),on=c({teamId:H,evidence:w,scoring:Qt.optional(),profiles:n(c({id:U,name:f(),profileRef:en,ownerWorkspaceId:H,ownerRevision:u().int().nonnegative(),baseRelease:w,profileSourceRevision:f(),components:n(c({kind:d([`skill`,`instruction`,`agent`,`workflow`]),path:f(),agentId:f().optional(),workflowId:f().optional(),label:f(),binding:S}).strict()).max(500)}).strict()).max(100)}).strict(),sn=c({candidateId:H,actorId:H,teamId:H,projectId:H.nullable(),candidateRevision:u().int().positive(),recipeHash:U,baseline:c({manifest:D,result:E}).strict(),candidate:c({manifest:D,result:E}).strict(),contentHash:U}).strict();function cn(e){let t=sn.parse(e),{contentHash:n,...r}=t;if(C(r)!==n)throw Error(`Accepted Human discovery integrity failed.`);for(let e of[t.baseline,t.candidate])if(T(e),e.manifest.teamId!==t.teamId||e.manifest.lineage?.humanProjection)throw Error(`Discovery must retain original owner evidence.`);return t}var G=f().trim().min(1).max(240),K=f().regex(/^[a-f0-9]{64}$/),ln=c({schemaVersion:a(`openpond.advancedRefinerEvaluation.v1`),operationId:G,actorId:G,teamId:G,projectId:G.nullable(),mode:d([`review_quality`,`downstream_adaptation`]),profileRef:c({source:d([`local`,`github`,`openpond_git`]),repositoryId:G,profileId:G}).strict(),sourceRevision:f().min(1).max(500),baselineRelease:w,refinerRelease:w,evidence:w,externalDatasetBinding:S,adaptationDatasetBinding:S,privateClosureHash:K,adaptationTaskIds:n(G).min(1).max(1e4),holdoutTaskIds:n(G).min(1).max(1e4),adaptationSplit:d([`train`,`validation`,`test`,`frozen_eval`]),holdoutSplit:d([`train`,`validation`,`test`,`frozen_eval`]),environmentPolicy:i(`kind`,[c({kind:a(`resettable`),environmentHash:K}).strict(),c({kind:a(`frozen_observations`),environmentHash:K}).strict()]),maximumCostUsd:u().finite().positive().max(1e4),maximumDurationMs:u().int().min(1e3).max(864e5),maximumModelSteps:u().int().min(1).max(1e3),contentHash:K}).strict();function un(e){let t=ln.parse(e),{contentHash:n,...r}=t;if(C(r)!==n)throw Error(`Advanced Refiner evaluation admission changed.`);let i=pe(t.externalDatasetBinding);if(i.profileId!==t.profileRef.profileId||i.protectedProfileClosureHash!==t.privateClosureHash)throw Error(`Advanced evaluation changed its Profile or protected evaluator closure.`);if(t.adaptationSplit===t.holdoutSplit||new Set([...t.adaptationTaskIds,...t.holdoutTaskIds]).size!==t.adaptationTaskIds.length+t.holdoutTaskIds.length)throw Error(`Adaptation and private holdout must be disjoint ordered cases.`);let a=pe(t.adaptationDatasetBinding);if(a.profileId!==i.profileId||a.packageHash!==i.packageHash||C(a.dataset)!==C(i.dataset)||C(a.target)!==C(i.target)||a.protectedProfileClosureHash!==i.protectedProfileClosureHash)throw Error(`This native advanced adapter needs separate adaptation and holdout recipes from the same exact released package and target.`);if(a.split!==t.adaptationSplit||i.split!==t.holdoutSplit)throw Error(`Advanced split policies differ from their exact admitted recipes.`);let o=new Set(i.population.map(e=>e.taskId)),s=new Set(a.population.map(e=>e.taskId));if(t.holdoutTaskIds.some(e=>!o.has(e))||t.adaptationTaskIds.some(e=>!s.has(e)))throw Error(`Advanced evaluation selected cases outside its exact Dataset recipe.`);return t}var dn=c({pin:ln,modelId:G,profileId:G,model:we,seed:u().int(),reasoningEffort:d([`none`,`off`,`low`,`medium`,`high`,`xhigh`,`max`]).nullable()}).strict(),fn=i(`operation`,[c({operation:a(`start`),request:dn}).strict(),c({operation:d([`read`,`cancel`,`resume`]),id:G}).strict(),c({operation:a(`list`)}).strict()]),pn=c({taskId:G,sourceCaseHash:K,reviewSessionId:G,reviewTurnId:G,inputHarness:w,trigger:w,decision:d([`no_action`,`route_deterministically`,`queue_refiner`]),outcome:w.nullable(),proposal:w.nullable(),validations:n(w).max(100),grade:c({id:G,contentHash:K,score:u().min(0).max(1).nullable(),passed:l(),failureClass:f().nullable()}).strict(),resetComplete:l()}).strict(),mn=c({schemaVersion:a(`openpond.advancedRefinerReviewReceipt.v1`),id:G,runId:G,pinHash:K,ownerActorId:G,teamId:G,privateClosureHash:K,refinerRelease:w,cases:n(pn).min(1).max(1e4),reviewScore:u().min(0).max(1).nullable(),completedCases:u().int().nonnegative(),noActionCases:u().int().nonnegative(),routedCases:u().int().nonnegative(),proposalCases:u().int().nonnegative(),costUsd:u().finite().nonnegative().nullable(),cleanupComplete:l(),createdAt:f().datetime(),contentHash:K}).strict(),hn=on.shape.profiles.element,gn=on.extend({profiles:n(hn.extend({components:n(hn.shape.components.element.extend({adaptationSplits:n(c({split:d([`train`,`validation`,`test`,`frozen_eval`]),taskIds:n(G)})),environmentHash:K,resettable:l(),reviewQuality:l()}))})),refinerReleases:n(w),activeRefiner:w.nullable(),models:n(c({id:G,profileId:G,name:f(),location:d([`local`,`hosted`])}))}),_n=c({operation:a(`prepare`),evidence:w,profileOptionId:G,componentLabel:G,mode:d([`review_quality`,`downstream_adaptation`]),adaptationSplit:d([`train`,`validation`,`test`,`frozen_eval`]),operationId:G,refinerRelease:w,maximumCostUsd:u().positive(),maximumDurationMs:u().int().positive(),maximumModelSteps:u().int().positive()}).strict();r([fn,_n,c({operation:a(`options`),evidence:w}).strict()]);var vn=c({modelRequests:u().int().nonnegative(),knownSpendUsd:u().nonnegative(),totalSpendUsd:u().nonnegative().nullable(),heldUsd:u().nonnegative(),uncertainRequests:u().int().nonnegative()}).strict(),q=y.nullable(),J=u().int().nonnegative(),yn=f().regex(/^[a-f0-9]{40}$/),bn=c({name:f().trim().min(1).max(160),startedAt:m,completedAt:m,durationMs:u().nonnegative(),clock:d([`monotonic`,`provider`,`wall`]),outcome:d([`succeeded`,`failed`,`cancelled`])}).strict(),xn=c({schemaVersion:a(`openpond.correlatedTelemetryReceipt.v1`),stage:d([`training`,`evaluation`,`deployment`,`inference`]),correlation:c({modelRunId:q,modelVersionId:q,policyVersion:J.nullable(),taskId:q,rolloutGroupId:q,providerResourceId:q,deploymentId:q,inferenceRequestId:q}).strict(),spans:n(bn).max(1e4),usage:c({promptTokens:J.nullable(),generatedTokens:J.nullable(),gpuSeconds:u().nonnegative().nullable(),workerActiveSeconds:u().nonnegative().nullable(),optimizerSteps:J.nullable(),rolloutGroups:J.nullable(),successfulTrajectories:J.nullable(),failedTrajectories:J.nullable(),peakGpuMemoryBytes:J.nullable(),peakGpuUtilizationPercent:u().min(0).max(100).nullable()}).strict(),resource:c({provider:y,resourceIds:n(y).max(1e3),gpuType:f().trim().min(1).max(300).nullable(),gpuCount:J.nullable(),baseProfileId:q,baseRepository:f().trim().min(1).max(512).nullable(),baseRevision:yn.nullable(),adapterContentHash:v.nullable()}).strict(),cost:c({currency:a(`USD`),providerReportedUsd:u().nonnegative().nullable(),quotedHourlyUsd:u().nonnegative().nullable(),estimatedUsd:u().nonnegative().nullable(),methodologyVersion:f().trim().min(1).max(160),pricingInputs:o(f().trim().min(1).max(160),u().finite()),unitEstimates:o(f().trim().min(1).max(160),u().nonnegative())}).strict(),recordedAt:m,contentHash:v}).strict(),Y=f().trim().min(1).max(240),X=f().regex(/^[a-f0-9]{64}$/),Z=f().datetime({offset:!0}),Q=c({id:Y,contentHash:X}).strict(),Sn=Q.extend({revision:u().int().positive()}).strict(),Cn=c({providerId:Y,modelId:f().trim().min(1).max(300)}).strict(),wn=d([`correction`,`sibling_verification`,`cumulative_known`,`development`,`retained`,`frozen_final`,`training_eligible`]),Tn=c({taskId:Y,taskContentHash:X,panelRole:wn,passLabel:Y.nullable(),optimizerEligible:l(),criticalInvariantIds:n(Y).max(100).default([])}).strict().superRefine((e,t)=>{e.optimizerEligible!==(e.panelRole===`correction`)&&t.addIssue({code:`custom`,path:[`optimizerEligible`],message:`Only correction cases may be optimizer eligible.`})}),En=c({schemaVersion:a(`openpond.continualBenchIssuePacket.v1`),id:Y,revision:u().int().positive(),contentHash:X,familyId:Y,familyLabel:f().trim().min(1).max(500),severity:d([`low`,`medium`,`high`,`critical`]),cases:n(Tn).min(2).max(1e4),duplicateEvidence:Q,leakageEvidence:Q,priorExposureEvidence:Q,observedAt:Z,reviewedAt:Z,queuedAt:Z.nullable(),runAt:Z.nullable(),createdBy:Y,createdAt:Z}).strict().superRefine((e,t)=>{e.cases.some(e=>e.panelRole===`correction`)||t.addIssue({code:`custom`,path:[`cases`],message:`An issue packet requires a correction case.`}),e.cases.some(e=>e.panelRole===`sibling_verification`)||t.addIssue({code:`custom`,path:[`cases`],message:`An issue packet requires a sibling-verification case.`}),new Set(e.cases.map(e=>e.taskId)).size!==e.cases.length&&t.addIssue({code:`custom`,path:[`cases`],message:`Issue-packet task ids must be unique.`})}),Dn=c({id:Y,role:wn,passLabel:Y.nullable(),taskset:Sn,familyIds:n(Y).min(1).max(1e5),taskCount:u().int().positive().max(1e6),sealedAt:Z}).strict().superRefine((e,t)=>{let n=e.role===`correction`||e.role===`sibling_verification`||e.role===`cumulative_known`;n!==(e.passLabel!==null)&&t.addIssue({code:`custom`,path:[`passLabel`],message:n?`This panel role requires a pass label.`:`This panel role is series-wide and cannot carry a pass label.`})}),On=i(`kind`,[c({kind:a(`base_model`),id:Y,revision:f().trim().min(1).max(500),contentHash:X}).strict(),c({kind:a(`model_version`),id:Y,contentHash:X}).strict(),c({kind:a(`external_reference`),id:Y,model:Cn,providerRevision:f().trim().min(1).max(500),contentHash:X}).strict()]),kn=c({scheduleEntryId:Y,ordinal:u().int().nonnegative(),label:Y,role:d([`seed`,`daily_residual`,`rank_candidate`,`weekly_rollup`,`full_refresh`]),parentRule:d([`base_model`,`previous_release`,`seed_release`,`accepted_daily_head`,`accepted_seed`]),trainableRank:u().int().positive().max(1024),correctionPanelIds:n(Y).min(1).max(1e3),optimizerGroupsPerTask:u().int().positive().max(1e3),trajectoriesPerGroup:u().int().min(4).max(1e3)}).strict(),An=c({revision:u().int().positive(),requireAllAttemptsTerminal:a(!0),criticalCorrectionPassRate:u().min(0).max(1),siblingPassRate:u().min(0).max(1),behavioralRetentionRate:u().min(0).max(1),maximumRetainedRegressionPoints:u().min(0).max(100),blockCriticalPriorRegression:l(),authorizedClaims:n(d([`systems_complete`,`correction_absorbed`,`issue_generalized`,`continually_current`,`frontier_pareto_result`])).min(1).max(5)}).strict(),jn=c({schemaVersion:a(`openpond.continualBenchmarkProtocol.v1`),id:Y,revision:u().int().positive(),contentHash:X,ownerId:Y,createdAt:Z,sealedAt:Z,creationReceipt:Q,predecessorSeries:c({id:Y,revision:u().int().positive(),contentHash:X}).strict().nullable(),scenarioPack:Q,issueFamilyLedger:Q,issuePacketRelease:Q,panels:n(Dn).min(6).max(1e4),grader:Q,judge:c({release:Q,rubricRelease:Q,calibrationRelease:Q,model:Cn}).strict().nullable(),policies:c({base:On,master:On.nullable(),externalReferences:n(On).min(1).max(20)}).strict(),invariants:c({systemPromptHash:X,toolSchema:Q,application:Q,harness:Q,runtime:Q,workerImage:Q,autoRefiner:c({enabled:l(),release:Q.nullable()}).strict()}).strict(),schedule:n(kn).min(1).max(1e3),evaluation:c({seeds:n(u().int()).min(3).max(100),repetitions:u().int().min(3).max(20),powerRule:Q,pairedBootstrapSamples:u().int().min(1e3).max(1e6),confidenceLevel:a(.95)}).strict(),resources:c({maximumTrainingGpuSeconds:u().int().positive(),maximumEvaluationGpuSeconds:u().int().positive(),maximumProviderSpendUsd:u().positive(),maximumTotalSpendUsd:u().positive(),maximumConcurrentGroups:u().int().positive().max(1e3)}).strict(),currencyThresholds:An}).strict().superRefine((e,t)=>{let n=e.panels.map(e=>e.id);new Set(n).size!==n.length&&t.addIssue({code:`custom`,path:[`panels`],message:`Protocol panel ids must be unique.`});for(let n of[`development`,`retained`,`frozen_final`,`training_eligible`])e.panels.filter(e=>e.role===n).length!==1&&t.addIssue({code:`custom`,path:[`panels`],message:`A protocol requires exactly one ${n} panel.`});[...e.schedule].sort((e,t)=>e.ordinal-t.ordinal).some((e,t)=>e.ordinal!==t)&&t.addIssue({code:`custom`,path:[`schedule`],message:`Protocol schedule ordinals must be contiguous from zero.`});let r=new Set(n);e.schedule.some(e=>e.correctionPanelIds.some(e=>!r.has(e)))&&t.addIssue({code:`custom`,path:[`schedule`],message:`Every scheduled correction panel must belong to the protocol.`})}),Mn=d([`draft`,`active`,`completed`,`archived`]),Nn=d([`draft`,`ready`,`queued`,`running`,`candidate`,`accepted`,`rejected`,`no_signal`,`failed`,`cancelled`]),Pn=d([`seed`,`daily_residual`,`rank_candidate`,`weekly_rollup`,`full_refresh`]),Fn=d([`base_model`,`previous_release`,`seed_release`,`accepted_daily_head`,`accepted_seed`]),In=d([`seed_taskset`,`nightly_selection`,`daily_cohort_union`,`eligible_task_pool`]),Ln=d([`live_client`,`evaluation_attempt`,`replay_evidence`,`manual`]),Rn=c({id:y,ordinal:u().int().nonnegative(),label:f().trim().min(1).max(100),role:Pn,parentRule:Fn,taskSource:In,trainableRank:u().int().positive().max(1024),minimumTasks:u().int().nonnegative().max(1e5),maximumTasks:u().int().positive().max(1e5)}).strict().superRefine((e,t)=>{e.minimumTasks>e.maximumTasks&&t.addIssue({code:`custom`,path:[`minimumTasks`],message:`The minimum task count cannot exceed the maximum task count.`});let{parentRules:n,taskSource:r}={seed:{parentRules:[`base_model`],taskSource:`seed_taskset`},daily_residual:{parentRules:[`previous_release`,`accepted_daily_head`],taskSource:`nightly_selection`},rank_candidate:{parentRules:[`seed_release`],taskSource:`eligible_task_pool`},weekly_rollup:{parentRules:[`seed_release`,`accepted_seed`],taskSource:`daily_cohort_union`},full_refresh:{parentRules:[`base_model`],taskSource:`eligible_task_pool`}}[e.role];n.includes(e.parentRule)||t.addIssue({code:`custom`,path:[`parentRule`],message:`${e.role} requires parent rule ${n.join(` or `)}.`}),e.taskSource!==r&&t.addIssue({code:`custom`,path:[`taskSource`],message:`${e.role} requires task source ${r}.`}),e.ordinal===0&&e.role!==`seed`&&t.addIssue({code:`custom`,path:[`role`],message:`The first series entry must be the seed.`})}),zn=c({id:y,version:u().int().positive(),requireCheckpoint:l(),requireAppliedOptimizerUpdate:l(),minimumCurrentCohortMeanImprovement:u().finite(),maximumRetainedMeanRegression:u().finite().nonnegative(),blockCriticalInvariantRegression:l(),automaticDailyAdvancement:l()}).strict(),Bn=c({schemaVersion:a(`openpond.modelComparisonSeries.v1`),id:y,profileId:y,modelProjectId:y,name:f().trim().min(1).max(500),objective:f().trim().min(1).max(2e4),status:Mn,revision:u().int().positive(),productionBinding:c({role:a(`chat_manual`),roleTargetId:y}).strict(),baseModel:c({id:y,revision:f().trim().min(1).max(256)}).strict(),seedTaskset:_,eligibleTaskPool:_,evaluationTasksets:c({development:_,retained:_,frozenFinal:_}).strict(),grader:g,benchmarkProtocol:jn.nullable().default(null),automaticEvaluation:c({enabled:l()}).strict().default({enabled:!1}),residualProfile:c({profileId:y,serializedEnvelopeRank:u().int().positive().max(1024),maximumEnabledRank:u().int().positive().max(1024),topology:a(`uniform_block_masked`)}).strict(),schedule:n(Rn).min(1).max(1e3),scheduleSealedAt:m.nullable(),advancementPolicy:zn,executionPolicy:c({startWhenReady:a(!1)}).strict(),acceptedSeedEntryId:y.nullable(),acceptedDailyHeadEntryId:y.nullable(),promotedBindingId:y.nullable(),createdBy:y,createdAt:m,updatedAt:m}).strict().superRefine((e,t)=>{[...e.schedule].sort((e,t)=>e.ordinal-t.ordinal).some((e,t)=>e.ordinal!==t)&&t.addIssue({code:`custom`,path:[`schedule`],message:`Schedule ordinals must be unique and contiguous from zero.`}),(new Set(e.schedule.map(e=>e.id)).size!==e.schedule.length||new Set(e.schedule.map(e=>e.label)).size!==e.schedule.length)&&t.addIssue({code:`custom`,path:[`schedule`],message:`Schedule ids and labels must be unique.`}),e.residualProfile.maximumEnabledRank>e.residualProfile.serializedEnvelopeRank&&t.addIssue({code:`custom`,path:[`residualProfile`],message:`Maximum enabled rank cannot exceed the serialized envelope.`});let n=Object.values(e.evaluationTasksets);new Set(n.map(e=>`${e.id}:${e.revision}:${e.contentHash}`)).size!==n.length&&t.addIssue({code:`custom`,path:[`evaluationTasksets`],message:`Development, retained, and frozen-final cohorts require distinct immutable Taskset releases.`});let r=e.schedule.find(e=>e.role===`seed`);if(!r||e.schedule.filter(e=>e.role===`seed`).length!==1){t.addIssue({code:`custom`,path:[`schedule`],message:`A series requires exactly one seed entry.`});return}let i=e.schedule.filter(e=>e.role===`rank_candidate`);i.length===1&&t.addIssue({code:`custom`,path:[`schedule`],message:`A causal rank comparison requires at least two rank candidates.`}),new Set(i.map(e=>e.trainableRank)).size!==i.length&&t.addIssue({code:`custom`,path:[`schedule`],message:`Causal rank candidates must use distinct trainable ranks.`}),[r.trainableRank+e.schedule.filter(e=>e.role===`daily_residual`).reduce((e,t)=>e+t.trainableRank,0),...e.schedule.filter(e=>e.role===`weekly_rollup`).map(e=>r.trainableRank+e.trainableRank),...e.schedule.filter(e=>e.role===`rank_candidate`).map(e=>r.trainableRank+e.trainableRank),...e.schedule.filter(e=>e.role===`full_refresh`).map(e=>e.trainableRank)].some(t=>t>e.residualProfile.maximumEnabledRank)&&t.addIssue({code:`custom`,path:[`schedule`],message:`A declared branch exceeds the maximum enabled rank.`})}),Vn=c({source:Ln,taskIds:n(y).min(1).max(1e5),observedFrom:m,observedTo:m,reviewedAt:m,reviewedBy:y,sourceTaskset:_}).strict(),Hn=Vn.extend({derivedFamilyKeys:n(y).min(1).max(1e5),familyDerivation:c({algorithm:y,sourceTasksetHash:v}).strict(),selectionHash:v}).strict(),Un=i(`kind`,[c({kind:a(`base_model`),id:y,revision:f().trim().min(1).max(256)}).strict(),c({kind:a(`model_version`),id:y,contentHash:v}).strict()]),Wn=c({id:y,branchOrdinal:u().int().nonnegative(),rank:u().int().positive().max(1024),offsetStart:u().int().nonnegative(),offsetEnd:u().int().positive(),optimizationRole:d([`frozen`,`trainable`]),artifactLineageId:y.nullable()}).strict().superRefine((e,t)=>{e.offsetEnd-e.offsetStart!==e.rank&&t.addIssue({code:`custom`,path:[`offsetEnd`],message:`Block offsets must span exactly the declared rank.`})}),Gn=c({disposition:d([`advance`,`hold`,`no_signal`]),policy:c({id:y,version:u().int().positive()}).strict(),reasonCodes:n(y).min(1).max(100),summary:f().trim().min(1).max(5e3),decidedBy:y,decidedAt:m}).strict(),Kn=c({evaluationRunId:y,modelVersionId:y,taskset:_,grader:g,cohortRole:d([`current`,`correction`,`sibling_verification`,`cumulative_known`,`development`,`retained`,`prior_disclosed`,`frozen_final`]),panelId:y.nullable().default(null),protocol:c({id:y,revision:u().int().positive(),contentHash:v}).strict().nullable().default(null)}).strict(),qn=c({attemptOrdinal:u().int().positive(),trainingPlanId:y.nullable(),modelRunId:y.nullable(),terminalStatus:d([`failed`,`cancelled`]),queuedAt:m.nullable(),startedAt:m.nullable(),completedAt:m}).strict(),Jn=c({seriesId:y,entryId:y,scheduleEntryId:y,ordinal:u().int().nonnegative(),releaseHash:v}).strict(),Yn=c({schemaVersion:a(`openpond.modelComparisonSeriesEntry.v1`),id:y,seriesId:y,profileId:y,modelProjectId:y,scheduleEntryId:y,releaseHash:v,ordinal:u().int().nonnegative(),label:f().trim().min(1).max(100),role:Pn,branch:d([`daily`,`rank_candidate`,`weekly_rollup`,`full_refresh`]),status:Nn,parent:Un,taskset:_,sourceTasksets:n(_).min(1).max(1e3),taskSelection:Hn.nullable(),trainableRank:u().int().positive().max(1024),serializedEnvelopeRank:u().int().positive().max(1024),enabledCumulativeRank:u().int().positive().max(1024),trainableBlockId:y,residualBlocks:n(Wn).min(1).max(1024),attemptOrdinal:u().int().positive().default(1),priorRunAttempts:n(qn).max(100).default([]),trainingPlanId:y.nullable(),modelRunId:y.nullable(),modelVersionId:y.nullable(),evaluations:n(Kn).max(1e5),decision:Gn.nullable(),promotionBindingId:y.nullable(),queuedAt:m.nullable(),startedAt:m.nullable(),completedAt:m.nullable(),createdAt:m,updatedAt:m}).strict().superRefine((e,t)=>{let n=[...e.residualBlocks].sort((e,t)=>e.offsetStart-t.offsetStart);n.some((e,t)=>e.offsetStart!==(t===0?0:n[t-1].offsetEnd))&&t.addIssue({code:`custom`,path:[`residualBlocks`],message:`Residual block offsets must be contiguous from zero.`});let r=n.at(-1)?.offsetEnd??0;(r!==e.enabledCumulativeRank||r>e.serializedEnvelopeRank)&&t.addIssue({code:`custom`,path:[`enabledCumulativeRank`],message:`Enabled rank must match the block inventory and fit the envelope.`});let i=e.residualBlocks.filter(e=>e.optimizationRole===`trainable`);(i.length!==1||i[0]?.id!==e.trainableBlockId||i[0]?.rank!==e.trainableRank)&&t.addIssue({code:`custom`,path:[`trainableBlockId`],message:`Exactly one block must be trainable during the Run and match trainableRank.`});let a=e.priorRunAttempts.map(e=>e.attemptOrdinal);(e.priorRunAttempts.length!==e.attemptOrdinal-1||a.some((e,t)=>e!==t+1))&&t.addIssue({code:`custom`,path:[`priorRunAttempts`],message:`Prior Run attempts must be a contiguous history before the active attempt.`})});c({seriesId:y,scheduleEntryId:y,taskSelection:Vn.nullable(),expectedSeriesRevision:u().int().positive()}).strict();var Xn=d([`base_reference`,`lora_adapter`]),Zn=c({baseModel:c({source:a(`huggingface`),repoId:y,revision:f().regex(/^[a-f0-9]{40}$/),configHash:v,tokenizerHash:v,licenseId:y,gated:l()}).strict(),processor:c({repository:y,revision:f().regex(/^[a-f0-9]{40}$/),configHash:v}).strict()}).strict().superRefine((e,t)=>{(e.baseModel.repoId!==e.processor.repository||e.baseModel.revision!==e.processor.revision)&&t.addIssue({code:`custom`,path:[`processor`],message:`Reward Model processor must match the pinned scorer base model.`})}),Qn=c({schemaVersion:a(`openpond.modelVersion.v1`),id:y,modelId:y,profileId:y,version:u().int().nonnegative(),kind:Xn,status:d([`available`,`failed`]),baseModel:qt,taskset:_,comparisonSeriesEntry:Jn.nullable().optional(),evaluationTasksetRef:_.nullable().optional(),releaseGraph:c({resolvedBundleHash:v,profileRelease:_,harnessRelease:g,agentRelease:g.nullable(),grader:c({id:y,contentHash:v}).strict()}).strict(),artifactLineageId:y.nullable(),adapterStatus:d([`not_trained`,`trained`]),createdAt:m,contentHash:v}).strict().superRefine((e,t)=>{e.kind===`base_reference`&&(e.version!==0||e.adapterStatus!==`not_trained`||e.artifactLineageId!==null)&&t.addIssue({code:`custom`,message:`A base-reference Model Version must be version 0 without adapter lineage.`}),e.kind===`lora_adapter`&&(e.version===0||e.adapterStatus!==`trained`||!e.artifactLineageId)&&t.addIssue({code:`custom`,message:`A LoRA Model Version must be version 1+ with adapter lineage.`})}),$n=c({schemaVersion:a(`openpond.rewardModelVersion.v1`),id:y,modelId:y,profileId:y,version:u().int().positive(),role:a(`reward`),status:d([`available`,`failed`]),scope:d([`synthetic_smoke`,`human_preference`]),baseModel:qt,runtime:Zn.nullable().default(null),taskset:_,preferenceDatasetRelease:g,releaseGraph:c({resolvedBundleHash:v,profileRelease:_,harnessRelease:g,grader:c({id:y,contentHash:v}).strict()}).strict(),artifacts:c({checkpoint:g.extend({objectRef:f().regex(/^r2:\/\/[A-Za-z0-9._/-]+$/),files:n(c({path:f().trim().min(1).max(512),sizeBytes:u().int().positive(),sha256:v}).strict()).min(4).max(128)}).strict(),adapter:g,scalarHead:g,bucketHead:g.nullable(),processorRelease:g}).strict(),qualificationReport:g.nullable(),createdAt:m,contentHash:v}).strict(),er=c({schemaVersion:a(`openpond.rewardModelRunReceipt.v1`),provider:y,providerRunId:y,resolvedBundleHash:v,finalCheckpoint:g.extend({objectRef:f().regex(/^r2:\/\/[A-Za-z0-9._/-]+$/)}).strict(),adapter:g,scalarHead:g,bucketHead:g.nullable(),processorRelease:g,optimizerEvidence:g,managedExecutionReceipt:g.nullable().default(null),parameterDeltaHash:v,cleanup:c({computeReleased:l(),providerTerminalObserved:l()}).strict(),contentHash:v}).strict(),tr=s(e=>{if(!e||typeof e!=`object`||Array.isArray(e))return e;let t=e;if(t.tasksetRelease!==void 0)return e;let n=t.taskset;if(!n||typeof n!=`object`||Array.isArray(n))return e;let r=n;return typeof r.id!=`string`||typeof r.contentHash!=`string`?e:{...t,tasksetRelease:{id:r.id,contentHash:r.contentHash}}},c({schemaVersion:a(`openpond.rewardModelRun.v1`),id:y,rewardModelId:y,rewardModelVersionId:y.nullable(),profileId:y,role:a(`reward`),scope:d([`synthetic_smoke`,`human_preference`]),status:d([`prepared`,`running`,`succeeded`,`failed`,`cancelled`]),taskset:_,tasksetRelease:g,preferenceDatasetRelease:g,recipeRelease:g,destinationId:y,quote:c({maximumSpendUsd:u().positive(),hourlyCostUsd:u().nonnegative().nullable()}).strict(),managedRunId:y.nullable(),progress:c({completedSteps:u().int().nonnegative(),totalSteps:u().int().positive(),latestLoss:u().finite().nullable()}).strict(),receipt:er.nullable(),qualificationReport:g.nullable(),accruedSpendUsd:u().nonnegative().nullable(),failureOwner:d([`authoring`,`admission`,`provider`,`runner`,`artifact`,`qualification`,`cleanup`]).nullable(),failure:f().trim().min(1).max(5e3).nullable(),startedAt:m,completedAt:m.nullable(),updatedAt:m}).strict()).superRefine((e,t)=>{e.status===`succeeded`&&!e.receipt&&t.addIssue({code:`custom`,message:`A successful Reward Model Run requires a canonical receipt.`}),e.status!==`succeeded`&&e.rewardModelVersionId!==null&&t.addIssue({code:`custom`,message:`Only a successful Reward Model Run can publish a Reward Model Version.`})}),nr=c({schemaVersion:a(`openpond.modelRunReceipt.v1`),provider:y,providerRunId:y,assignmentHash:v,resultHash:v,transcriptHash:v,traceHash:v.nullable(),resolvedBundleHash:v,artifactPath:f().trim().min(1).max(2e3),cleanup:c({computeReleased:l(),tunnelClosed:l()}).strict(),telemetry:xn.nullable().default(null),contentHash:v}).strict(),$=c({inputTokens:u().int().nonnegative(),outputTokens:u().int().nonnegative(),totalTokens:u().int().nonnegative(),costUsd:u().nonnegative().nullable()}).strict(),rr=f().regex(/^(?:[a-f0-9]{40}|[a-f0-9]{64})$/),ir=c({phase:d([`baseline`,`adaptation`,`candidate_adaptation`,`candidate`]),taskId:y,attemptId:y,sessionId:y.nullable(),turnId:y.nullable(),passed:l(),score:u().min(0).max(1).nullable(),failureClass:Rt.nullable(),inputTokens:u().int().nonnegative(),outputTokens:u().int().nonnegative(),totalTokens:u().int().nonnegative(),latencyMs:u().int().nonnegative(),costUsd:u().nonnegative().nullable(),startedAt:m}).strict(),ar=c({targetTaskCount:u().int().nonnegative(),comparedTaskCount:u().int().nonnegative(),passedTaskCount:u().int().nonnegative(),failedTaskCount:u().int().nonnegative(),lowerTaskCount:u().int().nonnegative(),higherTaskCount:u().int().nonnegative(),unchangedTaskCount:u().int().nonnegative()}).strict(),or=c({target:a(`all_tasks_lower`),targetTaskCount:u().int().nonnegative(),comparedTaskCount:u().int().nonnegative(),passedTaskCount:u().int().nonnegative(),failedTaskCount:u().int().nonnegative(),lowerTaskCount:u().int().nonnegative(),higherTaskCount:u().int().nonnegative(),unchangedTaskCount:u().int().nonnegative(),baselineTokens:u().int().nonnegative(),refinedTokens:u().int().nonnegative(),tokenDelta:u().int(),tokenDeltaPercent:u().finite().nullable(),complete:l(),passed:l(),cohorts:c({adaptation:ar,heldOut:ar}).strict()}).strict(),sr=c({baseline:$,adaptation:$,candidateAdaptation:$,candidate:$,refiner:$,grader:$}).strict(),cr=c({usage:sr,observedSpendUsd:u().nonnegative(),attempts:n(ir).max(1e4)}).strict(),lr=c({id:y,contentHash:v,artifactPath:f().trim().min(1).max(2e3)}).strict(),ur=c({schemaVersion:a(`openpond.modelEvaluationReceipt.v1`),benchmarkId:y,resultManifest:c({id:y,contentHash:v,artifactPath:f().trim().min(1).max(2e3)}).strict(),stages:c({baseline:g,adaptation:g,refiner:g,candidateAdaptation:g,candidate:g,comparison:g}).strict(),usage:sr,quality:c({baselinePassRate:u().min(0).max(1),candidatePassRate:u().min(0).max(1),adaptationBaselinePassRate:u().min(0).max(1),adaptationCandidatePassRate:u().min(0).max(1),adaptationCandidatePassed:l(),heldOutCandidatePassed:l(),passed:l()}).strict(),foregroundTokenDelta:u().int(),foregroundTokenDeltaPercent:u().finite().nullable(),taskEfficiency:or.optional(),efficiency:c({grossForegroundTokenSavings:u().int(),overheadTokens:u().int().nonnegative(),firstPassNetTokenSavings:u().int(),breakEvenReuseCount:u().int().nonnegative().nullable(),amortizedTokenSavings:u().int(),amortizedReuseCount:u().int().positive()}).strict(),budget:c({maximumSpendUsd:u().nonnegative(),observedSpendUsd:u().nonnegative(),enforced:l()}).strict(),evidenceSnapshot:g,lineage:c({adaptationEvidenceHash:v,refinerInputHash:v,refinerOutcomeHash:v,validationHash:v,applyReceiptHash:v,candidateRelease:g,valid:l()}).strict(),invalidReasons:n(f().trim().min(1).max(2e3)).max(100),attempts:n(ir).max(1e4).default([]),terminalClassification:d([`improved`,`no_improvement`,`regressed`,`inconclusive`,`cancelled`,`infrastructure_failure`]),profileGit:c({ref:f().trim().min(1).max(2e3),commit:rr,baseCommit:rr}).strict().nullable(),contentHash:v}).strict(),dr=c({schemaVersion:a(`openpond.modelEvaluationStopReceipt.v1`),benchmarkId:y,terminalClassification:a(`inconclusive`),stopReason:d([`candidate_harness_unchanged`,`candidate_lineage_invalid`]),reason:f().trim().min(1).max(5e3),stoppedAfter:a(`refiner`),baselineHarness:g,candidateHarness:g,refiner:c({id:y,contentHash:v,outcomeCount:u().int().nonnegative()}).strict(),usage:sr,budget:c({maximumSpendUsd:u().nonnegative(),observedSpendUsd:u().nonnegative(),enforced:l()}).strict(),evidenceSnapshot:g,attempts:n(ir).max(1e4),contentHash:v}).strict(),fr=c({level:a(.95),lower:u().finite(),upper:u().finite()}).strict().superRefine((e,t)=>{e.lower>e.upper&&t.addIssue({code:`custom`,message:`A confidence interval's lower bound cannot exceed its upper bound.`})}),pr=c({schemaVersion:a(`openpond.modelComparisonBenchmarkReceipt.v1`),benchmarkId:a(`model-comparison`),target:c({kind:d([`base_model`,`model_version`,`external_reference`]),label:f().trim().min(1).max(300),modelVersionId:y.nullable(),model:b.nullable()}).strict(),taskset:_,grader:g,sampling:c({seeds:n(u().int()).min(1).max(100),repetitions:u().int().positive().max(20)}).strict(),deterministic:c({attemptedTaskCount:u().int().nonnegative(),completedTaskCount:u().int().nonnegative(),passedTaskCount:u().int().nonnegative(),failedTaskCount:u().int().nonnegative(),meanScore:u().finite().nullable(),passRate:u().min(0).max(1).nullable(),passRateCi95:fr.nullable()}).strict(),judge:c({judgeModel:b,judgeRelease:g,rubricRelease:g,calibrationRelease:g,score:u().min(0).max(100),scoreCi95:fr,comparisonCount:u().int().positive(),wins:u().int().nonnegative(),ties:u().int().nonnegative(),losses:u().int().nonnegative(),referenceLabel:f().trim().min(1).max(300)}).strict().superRefine((e,t)=>{e.wins+e.ties+e.losses!==e.comparisonCount&&t.addIssue({code:`custom`,message:`Judge win, tie, and loss counts must equal comparisonCount.`})}).nullable(),attempts:n(c({attemptId:y.nullable().default(null),taskId:y,seed:u().int(),repetition:u().int().nonnegative(),status:d([`succeeded`,`failed`]),deterministicScore:u().finite().nullable(),passed:l().nullable(),judgeScore:u().min(0).max(100).nullable(),judgePreference:d([`win`,`tie`,`loss`]).nullable(),transcriptHash:v.nullable(),traceHash:v.nullable(),transcriptArtifact:c({artifactPath:f().trim().min(1).max(2e3),jsonPointer:f().trim().min(1).max(1e3)}).strict().nullable().default(null),traceArtifact:c({artifactPath:f().trim().min(1).max(2e3),jsonPointer:f().trim().min(1).max(1e3)}).strict().nullable().default(null),latencyMs:u().int().nonnegative().nullable().default(null),failureClass:Rt.nullable()}).strict()).max(1e5),usage:c({policy:$.nullable(),judge:$.nullable(),observedSpendUsd:u().nonnegative().nullable(),evaluationGpuSeconds:u().nonnegative().nullable().default(null)}).strict(),managedServing:c({jobId:y,terminalState:d([`completed`,`cancelled`,`failed`,`budget_exhausted`]),sourcePolicyVersion:u().int().nonnegative(),sourceAdapterSha256:v.nullable(),servedPolicyVersion:u().int().nonnegative().nullable(),servedAdapterSha256:v.nullable(),accruedSpendUsd:u().nonnegative(),cleanupAttestationHash:v,resourceCount:u().int().nonnegative(),activeResourceCount:a(0)}).strict().nullable().default(null),evidenceSnapshot:c({id:y,contentHash:v,artifactPath:f().trim().min(1).max(2e3)}).strict(),completedAt:m,contentHash:v}).strict().superRefine((e,t)=>{let n=e.deterministic;n.completedTaskCount>n.attemptedTaskCount&&t.addIssue({code:`custom`,message:`Completed task count cannot exceed attempted task count.`}),n.passedTaskCount+n.failedTaskCount!==n.completedTaskCount&&t.addIssue({code:`custom`,message:`Passed and failed task counts must equal completed task count.`}),n.completedTaskCount===0!=(n.meanScore===null)&&t.addIssue({code:`custom`,message:`Deterministic mean score is available exactly when tasks completed.`}),n.completedTaskCount===0!=(n.passRate===null)&&t.addIssue({code:`custom`,message:`Deterministic pass rate is available exactly when tasks completed.`})}),mr=c({advancedEvaluation:ln.optional(),benchmarkId:a(`harness-refiner`),model:b,upstreamModel:c({providerId:y,modelId:f().trim().min(1).max(300),revision:f().trim().min(1).max(300),pricing:c({version:f().trim().min(1).max(300),source:f().trim().min(1).max(300),effectiveAt:m,inputUsdPerMillionTokens:u().nonnegative(),cachedInputUsdPerMillionTokens:u().nonnegative(),outputUsdPerMillionTokens:u().nonnegative()}).strict().optional()}).strict(),reasoningEffort:f().trim().min(1).max(100).nullable(),seeds:n(u().int()).min(1).max(100),repetitions:u().int().positive().max(20),maximumSpendUsd:u().nonnegative(),attemptPlan:n(c({stage:d([`baseline`,`adaptation`,`candidate_adaptation`,`candidate`]),split:f().trim().min(1).max(100),taskIds:n(y).min(1).max(1e5),attemptCount:u().int().positive()}).strict()).length(4)}).strict(),hr=c({benchmarkId:a(`model-comparison`),target:c({kind:d([`base_model`,`model_version`,`external_reference`]),label:f().trim().min(1).max(300),modelVersionId:y.nullable(),model:b.nullable()}).strict(),grader:g,judge:c({model:b,release:g,rubricRelease:g,calibrationRelease:g,referenceLabel:f().trim().min(1).max(300)}).strict().nullable(),seeds:n(u().int()).min(1).max(100),repetitions:u().int().positive().max(20),maximumSpendUsd:u().nonnegative(),series:c({id:y,protocol:c({id:y,revision:u().int().positive(),contentHash:v}).strict()}).strict().nullable().default(null),panel:c({id:y,role:d([`correction`,`sibling_verification`,`cumulative_known`,`development`,`retained`,`frozen_final`]),passLabel:y.nullable()}).strict().nullable().default(null),comparisonPair:c({entryId:y,parent:Un,candidateModelVersionId:y}).strict().nullable().default(null),attemptPlan:n(c({stage:a(`comparison`),split:f().trim().min(1).max(100),taskIds:n(y).min(1).max(1e5),attemptCount:u().int().positive()}).strict()).length(1)}).strict(),gr=r([mr,hr]),_r=c({schemaVersion:a(`openpond.modelRun.v1`),id:y,modelId:y,modelVersionId:y.nullable(),profileId:y,kind:d([`rollout_smoke`,`training`,`evaluation`]),status:d([`prepared`,`running`,`succeeded`,`failed`,`cancelled`]),method:d([`sft`,`dpo`,`grpo`,`ppo`,`sdft`,`opd`,`opsd`,`sdpo`]).nullable(),destinationId:y.nullable(),taskset:_,comparisonSeriesEntry:Jn.nullable().optional(),evaluationTasksetRef:_.nullable().optional(),harnessRelease:g.nullable().optional(),quote:c({maximumSpendUsd:u().nonnegative(),hourlyCostUsd:u().nonnegative().nullable()}).strict().nullable(),evaluation:gr.nullable().default(null),evaluationProgress:c({stage:d([`baseline`,`adaptation`,`refiner`,`candidate_adaptation`,`candidate`,`comparison`]),completedAttempts:u().int().nonnegative(),totalAttempts:u().int().positive(),accounting:cr.nullable().default(null),advancedAccounting:vn.optional(),evidenceSnapshot:lr.nullable().default(null)}).strict().nullable().default(null),reward:c({raw:u(),components:o(f(),u())}).strict().nullable(),receipt:r([nr,ur,dr,pr,mn]).nullable(),adapterArtifactLineageId:y.nullable(),failure:f().trim().min(1).max(5e3).nullable(),startedAt:m,completedAt:m.nullable(),updatedAt:m}).strict().superRefine((e,t)=>{e.status===`succeeded`&&!e.receipt&&t.addIssue({code:`custom`,message:`A successful Model Run requires a canonical receipt.`}),e.kind===`evaluation`&&(e.method!==null||e.destinationId!==null||!e.evaluation||e.adapterArtifactLineageId!==null)&&t.addIssue({code:`custom`,message:`An evaluation Model Run requires evaluation configuration and cannot claim training method, destination, or adapter lineage.`}),e.kind===`evaluation`&&e.receipt&&e.receipt.schemaVersion!==`openpond.modelEvaluationReceipt.v1`&&e.receipt.schemaVersion!==`openpond.modelEvaluationStopReceipt.v1`&&e.receipt.schemaVersion!==`openpond.modelComparisonBenchmarkReceipt.v1`&&e.receipt.schemaVersion!==`openpond.advancedRefinerReviewReceipt.v1`&&t.addIssue({code:`custom`,message:`An evaluation Model Run requires an evaluation receipt.`}),e.kind!==`evaluation`&&(e.method===null||e.destinationId===null||e.quote===null)&&t.addIssue({code:`custom`,message:`A training or rollout Model Run requires method, destination, and quote.`}),e.kind===`rollout_smoke`&&e.adapterArtifactLineageId!==null&&t.addIssue({code:`custom`,message:`A rollout smoke cannot claim adapter artifacts.`})});function vr(e){let t=new Map;for(let n of e.attempts){let e=`${n.phase}\u0000${n.taskId}`,r=t.get(e);(!r||n.startedAt>r.startedAt||n.startedAt===r.startedAt&&n.attemptId>r.attemptId)&&t.set(e,n)}let n=[...t.values()],r=[[`adaptation`,`adaptation`,`candidate_adaptation`],[`held_out`,`baseline`,`candidate`]].flatMap(([e,t,r])=>{let i=new Map(n.filter(e=>e.phase===r).map(e=>[e.taskId,e]));return n.filter(e=>e.phase===t).flatMap(t=>{let n=i.get(t.taskId);return n?[{cohort:e,taskId:t.taskId,baseline:t,refined:n,tokenDelta:n.totalTokens-t.totalTokens}]:[]})}),i=new Set(n.filter(e=>e.phase===`baseline`||e.phase===`adaptation`).map(e=>e.taskId)).size,a=e.targetTaskCount??i;if(!Number.isInteger(a)||a<0)throw Error(`Benchmark task-efficiency target must be a non-negative integer.`);let o={adaptation:new Set(n.filter(e=>e.phase===`adaptation`).map(e=>e.taskId)).size,held_out:new Set(n.filter(e=>e.phase===`baseline`).map(e=>e.taskId)).size},s=(e,t)=>({targetTaskCount:t,comparedTaskCount:e.length,passedTaskCount:e.filter(e=>e.tokenDelta<0).length,failedTaskCount:e.filter(e=>e.tokenDelta>=0).length,lowerTaskCount:e.filter(e=>e.tokenDelta<0).length,higherTaskCount:e.filter(e=>e.tokenDelta>0).length,unchangedTaskCount:e.filter(e=>e.tokenDelta===0).length}),ee=s(r,a),c=r.reduce((e,t)=>e+t.baseline.totalTokens,0),l=r.reduce((e,t)=>e+t.refined.totalTokens,0),u=l-c,d=r.length===a;return{summary:or.parse({target:`all_tasks_lower`,...ee,baselineTokens:c,refinedTokens:l,tokenDelta:u,tokenDeltaPercent:c>0?u/c*100:null,complete:d,passed:d&&ee.passedTaskCount===a,cohorts:{adaptation:s(r.filter(e=>e.cohort===`adaptation`),o.adaptation),heldOut:s(r.filter(e=>e.cohort===`held_out`),o.held_out)}}),pairs:r}}var yr=f().trim().min(1).max(240),br=f().trim().min(1),xr=f().trim().min(8).max(256),Sr=o(f(),p()).default({}),Cr=c({schemaVersion:a(`openpond.localModelChatConfiguration.v1`).default(`openpond.localModelChatConfiguration.v1`),profile:d([`efficient`,`full_harness`,`custom`]).default(`efficient`),systemPromptMode:d([`lean`,`full_harness`,`custom`]).default(`lean`),customSystemPrompt:f().max(2e4).nullable().default(null),contextWindowTokens:u().int().min(128).max(32768).default(1024),maxOutputTokens:u().int().min(1).max(512).default(64),temperature:u().min(0).max(2).default(0),repetitionPenalty:u().min(.5).max(2).default(1.1),noRepeatNgramSize:u().int().min(0).max(10).default(3),compaction:d([`off`,`when_needed`]).default(`when_needed`),keepWarmSeconds:u().int().min(0).max(3600).default(300),updatedAt:br.nullable().default(null)}),wr=Cr.parse({}),Tr=c({schemaVersion:a(`openpond.managedAdapterServingProjection.v1`),teamId:yr.nullable().default(null),source:a(`sandbox_managed_rl`),sourceRef:yr,canonicalArtifactId:yr.nullable(),canonicalModelName:f().trim().min(1).max(256).nullable().default(null),canonicalArtifactState:d([`imported_unvalidated`,`evaluating`,`promotable`,`rejected`,`deleted`]).nullable(),canonicalDeploymentId:yr.nullable(),canonicalDeploymentState:d([`requested`,`deploying`,`ready`,`degraded`,`deleting`,`deleted`,`failed`]).nullable(),state:d([`pending`,`imported`,`ready`,`failed`]),customerBindingAllowed:l().default(!1),artifactContentHash:xr.nullable().default(null),baseProfileId:yr.nullable().default(null),publishedAt:br.nullable(),lastSyncedAt:br,lastError:f().trim().min(1).max(5e3).nullable()});function Er(e){let t=e.managedServing;return!!(t?.source===`sandbox_managed_rl`&&t.canonicalArtifactState===`promotable`&&t.customerBindingAllowed)}function Dr(e){return e.state===`ready`&&e.canonicalArtifactState===`promotable`&&e.canonicalDeploymentState===`ready`&&e.customerBindingAllowed&&!!e.teamId&&!!e.canonicalArtifactId&&!!e.canonicalDeploymentId}function Or(e){if(e.promotable&&e.frozenEvaluationArtifactId)return{kind:`source_frozen_evaluation`,evaluationArtifactId:e.frozenEvaluationArtifactId,canonicalArtifactId:null,canonicalDeploymentId:null};let t=e.managedServing;return t?.source===`sandbox_managed_rl`&&Dr(t)&&t.canonicalArtifactId&&t.canonicalDeploymentId?{kind:`sandbox_customer_binding`,evaluationArtifactId:null,canonicalArtifactId:t.canonicalArtifactId,canonicalDeploymentId:t.canonicalDeploymentId}:null}function kr(e){let t=e.managedServing;return t?Dr(t)&&t.canonicalArtifactId&&t.canonicalDeploymentId?{kind:`sandbox_serving`,canonicalArtifactId:t.canonicalArtifactId,canonicalDeploymentId:t.canonicalDeploymentId}:null:{kind:`source_artifact`,canonicalArtifactId:null,canonicalDeploymentId:null}}export{Pe as $,cn as A,Et as B,gn as C,on as D,nn as E,Ht as F,yt as G,At as H,Vt as I,at as J,st as K,Yt as L,qt as M,Wt as N,W as O,Gt as P,Fe as Q,Kt as R,fn as S,an as T,jt as U,kt as V,xt as W,et as X,tt as Y,Xe as Z,Jn as _,kr as a,En as b,yr as c,_r as d,Qn as f,vr as g,$n as h,Er as i,tn as j,$t as k,Sr as l,Zn as m,Cr as n,Or as o,tr as p,dt as q,Tr as r,xr as s,wr as t,br as u,Yn as v,un as w,wn as x,Bn as y,Ot as z};
+import{C as e,T as t,a as n,b as r,c as i,d as a,g as o,h as s,i as ee,m as c,o as l,p as u,r as d,v as f,x as p}from"./schemas-lO3MoWQr.js";import{a as te,c as m,d as h,i as g,l as _,n as ne,o as v,s as y,t as re}from"./common-BMjyKlX-.js";import{D as ie,E as b,r as ae}from"./settings-BKmRueaw.js";import{A as oe,D as se,R as x,U as ce,W as le,_ as S,d as C,h as ue,i as w,p as de,t as fe,v as pe}from"./chunk-H3JZ3P5T-B0SCoPw4.js";import{A as T,D as E,E as D,H as me,I as he,M as ge,O as _e,R as ve,U as ye}from"./contracts-BHTOD1nd.js";import{C as be,K as xe,O as Se,T as Ce,W as O,Y as we,b as Te,f as Ee,i as De,l as Oe,n as ke,r as Ae,t as je,y as Me}from"./chunk-ZFYI4P6C-BmsNtSkR.js";var Ne=f().trim().min(1).max(200),Pe=c({sessionId:Ne,turnId:Ne,source:de,admissionHash:v,modelConfigurationHash:v,traceHash:v,runtimeEventRefs:n(Ne).max(1e4),startedAt:m,completedAt:m,outputHash:v,partialOutputAvailable:l()}).strict().superRefine((e,t)=>{new Set(e.runtimeEventRefs).size!==e.runtimeEventRefs.length&&t.addIssue({code:`custom`,path:[`runtimeEventRefs`],message:`Native event references must be unique.`}),Date.parse(e.completedAt)<Date.parse(e.startedAt)&&t.addIssue({code:`custom`,path:[`completedAt`],message:`Native completion precedes its start.`})});c({schemaVersion:a(`openpond.nativeHarnessExperimentAttempt.v1`),taskId:f().min(1).max(500),status:d([`completed`,`cancelled`,`timed_out`,`policy_failure`]),output:f().max(8388608).nullable(),error:f().max(2e3).nullable(),messages:n(p()).max(4002),collected:l(),environmentCleanupComplete:a(!0),snapshot:ee(),native:Pe}).strict().superRefine((e,t)=>{let n=e.status===`completed`;(n!==e.collected||n!==(e.output!==null)||n!==(e.error===null))&&t.addIssue({code:`custom`,message:`Native status, output, collection and error must agree.`}),n&&(e.native.partialOutputAvailable||e.native.outputHash!==h(e.output))&&t.addIssue({code:`custom`,path:[`native`,`outputHash`],message:`Native output differs from its retained evidence.`})}).extend({contentHash:v}).strict();var Fe=c({schemaVersion:a(`openpond.humanResultView.v1`),executionId:f().min(1),executionHash:f().regex(/^[a-f0-9]{64}$/),selection:n(x).max(1e4),coverage:c({pending:u().int().nonnegative(),submitted:u().int().nonnegative(),accepted:u().int().nonnegative(),rejected:u().int().nonnegative(),cancelled:u().int().nonnegative()}).strict(),results:n(c({review:x,decision:x.nullable(),attemptIds:n(f().min(1)).min(1).max(2),mode:d([`individual`,`pairwise`]),origin:d([`hosted_retained`,`recorded_evidence`,`owner_attested_local`,`native_local`]),status:d([`pending`,`submitted`,`accepted`,`rejected`,`cancelled`]),grader:x,answers:n(oe).max(100)}).strict()).max(1e4)}).strict().extend({contentHash:f().regex(/^[a-f0-9]{64}$/)}).strict(),Ie=f().regex(/^[a-f0-9]{64}$/),Le=f().min(1).max(240),Re={feedbackKey:Le,criterionId:Le},ze=i(`kind`,[c({...Re,kind:a(`normalized_score`),minimum:u().finite(),maximum:u().finite(),passThreshold:u().min(0).max(1).optional()}).strict(),c({...Re,kind:a(`boolean`)}).strict(),c({...Re,kind:a(`category`)}).strict(),c({...Re,kind:a(`preference_score`),tieScore:u().min(0).max(1).nullable()}).strict()]),Be=c({execution:g,resultHash:Ie,reviews:n(c({review:x,decision:x}).strict()).min(1).max(1e4)}).strict(),Ve=c({rules:n(ze).min(1).max(100),baseline:Be,candidate:Be}).strict().superRefine((e,t)=>{new Set(e.rules.map(e=>e.feedbackKey)).size!==e.rules.length&&t.addIssue({code:`custom`,message:`Select each Human feedback key once.`})}),He=c({manifest:D,result:E}).strict(),Ue=c({id:Le,manifestHash:Ie,resultHash:Ie}).strict(),We=c({schemaVersion:a(`openpond.humanComparisonProjection.v1`),teamId:Le,projectId:Le,selections:Ve,original:c({baseline:Ue,candidate:Ue}).strict(),humanResults:c({baseline:Fe,candidate:Fe}).strict(),baseline:He,candidate:He}).strict().extend({contentHash:Ie}).strict();function Ge(e){throw new ye(e,409)}function Ke(e,t){let n=We.parse(e),{contentHash:r,...i}=n;r!==h(i)&&Ge(`human_comparison_projection_invalid`);for(let e of[`baseline`,`candidate`]){let r=T(n[e]),i=r.manifest.lineage,a=i?.humanProjection,o=n.original[e],s=n.selections[e];(!i||!a||a.originalManifest.id!==o.id||a.originalManifest.contentHash!==o.manifestHash||a.selectionHash!==h(s)||a.rulesHash!==h(n.selections.rules)||h(i.execution)!==h(s.execution)||r.manifest.teamId!==n.teamId||n.humanResults[e].contentHash!==s.resultHash)&&Ge(`human_comparison_projection_lineage_changed`);let{contentHash:ee,...c}=n.humanResults[e];if((h(c)!==ee||c.executionId!==s.execution.id||c.executionHash!==s.execution.contentHash||h(c.selection)!==h(s.reviews.map(e=>e.review)))&&Ge(`human_comparison_human_result_changed`),t){let n=T(t[e]);(n.manifest.lineage?.humanProjection||n.manifest.id!==o.id||n.manifest.contentHash!==o.manifestHash||n.result.contentHash!==o.resultHash||h(n.manifest.lineage?.execution)!==h(s.execution))&&Ge(`human_comparison_original_changed`)}}return n}var qe=c({adapterId:y,placement:d([`local`,`remote`,`colocated`,`provider_native`]),runtimeVersion:f().trim().min(1).max(200),capabilityReceipt:v}).strict(),Je=c({maxTurns:u().int().positive().max(1e5),timeoutMs:u().int().positive().max(864e5),maxOutputBytes:u().int().positive().max(25e7),maximumSpendUsd:u().nonnegative().nullable()}).strict();c({schemaVersion:a(`openpond.runManifest.v1`),id:y,harnessRelease:g,tasksetRelease:g,model:ie,runtimeTarget:qe,limits:Je,approval:c({id:y,contentHash:v}).strict().nullable(),createdAt:m,metadata:te}).strict().extend({contentHash:v}).strict();var Ye=c({schemaVersion:a(`openpond.attemptReceipt.v1`),id:y,runManifest:g,taskId:y,seed:f().trim().min(1).max(500),terminal:l(),failureClass:re.nullable(),outputHash:v.nullable(),traceHash:v,artifactRefs:n(ne).max(1e5),graderEvidenceRefs:n(ne).max(1e4),startedAt:m,completedAt:m,latencyMs:u().int().nonnegative(),costUsd:u().nonnegative().nullable(),legacyAttemptRef:y.nullable().default(null),metadata:te}).strict().extend({contentHash:v}).strict(),Xe=c({schemaVersion:a(`openpond.evaluationResult.v1`),id:y,runManifest:g,harnessRelease:g,tasksetRelease:g,model:ie,receiptRefs:n(g).min(1).max(1e6),attemptCount:u().int().positive(),rewardEligibleCount:u().int().nonnegative(),terminalCount:u().int().nonnegative(),meanScore:u().min(0).max(1).nullable(),failureCounts:o(re,u().int().nonnegative()),authoredMetric:le.optional(),metadata:te}).strict().extend({contentHash:v}).strict();c({schemaVersion:a(`openpond.harnessCompatibility.v1`),id:y,baseHarnessRelease:g,candidateHarnessRelease:g,tasksetRelease:g,environmentHash:v,toolContractHash:v,policyHash:v,graderInterfaceHash:v,metadata:te}).strict().extend({contentHash:v}).strict();var Ze=i(`kind`,[c({kind:a(`model`),model:ie,configurationHash:v}).strict(),c({kind:a(`fixture`)}).strict()]),Qe=i(`kind`,[c({kind:a(`taskset`),environmentRelease:g,verifierSetRelease:g,policyHash:v}).strict(),c({kind:a(`harness`),harnessRelease:g}).strict()]),$e=c({receiptId:y,taskId:y,seed:f().trim().min(1).max(500),fixtureId:y.nullable()}).strict(),et=c({schemaVersion:a(`openpond.tasksetRunManifest.v1`),id:y,tasksetRelease:g,packageHash:v,execution:Qe,profileEvaluation:ue.optional(),policy:Ze,gradingRole:a(`evaluation`),metricPolicy:ce,population:n($e).min(1).max(1e5),runtimeTarget:qe,limits:Je,createdAt:m,metadata:te}).strict().superRefine((e,t)=>{new Set(e.population.map(e=>e.receiptId)).size!==e.population.length&&t.addIssue({code:`custom`,path:[`population`],message:`Run receipt identities must be unique.`}),e.population.some(t=>t.fixtureId!==null!=(e.policy.kind===`fixture`))&&t.addIssue({code:`custom`,path:[`population`],message:`Only fixture runs require a fixture identity for every member.`}),e.profileEvaluation&&(e.execution.kind!==`harness`||e.execution.harnessRelease.id!==e.profileEvaluation.harnessRelease.id||e.execution.harnessRelease.contentHash!==e.profileEvaluation.harnessRelease.contentHash)&&t.addIssue({code:`custom`,path:[`profileEvaluation`],message:`Profile evaluation must execute its exact bound Harness release.`})}).safeExtend({contentHash:v}),k=f().trim().min(1).max(240),A=f().trim().min(1).max(5e3),tt=d([`no_training`,`prompting`,`retrieval`,`sft`,`preference`,`grpo_rft`,`sdft_opsd`,`sdpo`,`agentic_rl`]),nt=c({id:k,taskIndex:u().int().nonnegative(),label:Me,output:o(f(),p()),infrastructureError:f().trim().min(1).max(1e4).nullable(),expectedPassed:l(),expectedRewardEligible:l(),metadata:o(f(),p()).default({})}),rt=c({schemaVersion:a(`openpond.capabilityDiagnosis.v1`),summary:f().trim().min(1).max(1e4),stableBehavior:n(A).max(100).default([]),changingKnowledge:n(A).max(100).default([]),requiredContext:n(A).max(100).default([]),requiredTools:n(k).max(100).default([]),intervention:tt,trainingEligible:l(),rationale:n(A).min(1).max(100),confidence:u().min(0).max(1)}),it=c({id:k,sourceId:k,sourceTurnId:k.nullable(),split:Se,origin:d([`extracted`,`corrected`,`synthetic`,`expert_authored`]),inputPrompt:f().trim().min(1).max(1e5),expectedOutputText:f().trim().min(1).max(2e5).nullable(),rationale:A}),at=c({schemaVersion:a(`openpond.taskDesignProposal.v1`),id:k,name:f().trim().min(1).max(500),objective:f().trim().min(1).max(2e4),diagnosis:rt.default({schemaVersion:`openpond.capabilityDiagnosis.v1`,summary:`Reproduce the selected approved behavior.`,stableBehavior:[],changingKnowledge:[],requiredContext:[],requiredTools:[],intervention:`sft`,trainingEligible:!0,rationale:[`The selected examples were supplied as demonstrations.`],confidence:.5}),taskKind:Ce.shape.taskKind,sourceIds:n(k).min(1).max(1e5),assumptions:n(A).max(1e3),successCriteria:n(A).min(1).max(1e3),proposedGraders:n(Te).max(1e3).default([]),graderFixtures:n(nt).max(1e5).default([]),generatedFiles:n(De).max(1e3).default([]),proposedExamples:n(it).max(1e5).default([]),proposedMethod:Oe.shape.recommendedMethod,trainingPath:Ee.nullable().default(null),policy:be,warnings:n(f().trim().min(1)).default([]),createdAt:f().trim().min(1)}),j=f().regex(/^[a-f0-9]{64}$/),ot=c({providerId:a(`claude-code`),configurationHash:j,executableHash:j,version:f().regex(/^2\.\d+\.\d+$/),capabilityHash:j,maximumRequestCostUsd:u().finite().positive().max(1e4),requestTimeoutMs:u().int().min(1e3).max(3e5),maximumTurns:u().int().min(1).max(100),holdOpen:l()}).strict();c({id:f().min(1).max(200),receiptId:f().min(1).max(200),sessionId:e(),operationId:f().min(1).max(200),action:d([`message`,`seal`]),text:f().max(262144).optional()}).strict().superRefine((e,t)=>{e.action===`message`&&!e.text?.trim()&&t.addIssue({code:`custom`,message:`A human intervention requires a message.`}),e.action===`seal`&&e.text!==void 0&&t.addIssue({code:`custom`,message:`Sealing does not add a prompt.`})});var st=c({provider:a(`claude-code`),sessionId:e(),requestId:f().min(1).max(200),modelId:f().min(1).max(500),runtime:ot,sourceHash:j,outputHash:j,traceHash:j,runtimeEventRefs:n(f().min(1).max(200)).min(1).max(2e4),eventCount:u().int().positive().max(2e4),startedAt:t(),completedAt:t(),cleanupComplete:a(!0)}).strict().superRefine((e,t)=>{(e.runtimeEventRefs.length!==e.eventCount||new Set(e.runtimeEventRefs).size!==e.eventCount)&&t.addIssue({code:`custom`,message:`Process event references must match the exact retained trace population.`})}),ct=c({providerId:a(`codex`),configurationHash:f().regex(/^[a-f0-9]{64}$/),accountHash:f().regex(/^[a-f0-9]{64}$/),reasoningEffort:d([`low`,`medium`,`high`,`xhigh`]),maximumRequestCostUsd:u().finite().positive().max(1e4),requestTimeoutMs:u().int().min(1e3).max(3e5)}).strict(),lt=c({id:f().trim().min(1).max(240),revision:u().int().positive(),contentHash:f().regex(/^[a-f0-9]{64}$/),targetId:f().trim().min(1).max(240).nullable()}).strict(),ut=n(c({role:d([`system`,`user`,`assistant`]),content:f().max(1e5)}).strict()).max(100).superRefine((e,t)=>{e.reduce((e,t)=>e+t.content.length,0)>1e5&&t.addIssue({code:`custom`,message:`Evaluation prompt exceeds 100,000 characters.`})}),dt=c({schemaVersion:a(`openpond.experimentAttemptGrade.v1`),selectionHash:f().regex(/^[a-f0-9]{64}$/),grades:n(me).min(1).max(100),score:u().min(0).max(1).nullable(),passed:l(),rewardEligible:l(),gradingStatus:d([`scored`,`pending`,`unscorable`,`not_configured`]),failureClass:fe.nullable(),contentHash:f().regex(/^[a-f0-9]{64}$/)}).strict().superRefine((e,t)=>{let{contentHash:n,...r}=e;C(r)!==n&&t.addIssue({code:`custom`,message:`Experiment grading integrity failed.`});let i=e.grades.flatMap(e=>e.components);(i.length!==e.grades.length||new Set(i.map(e=>e.graderId)).size!==i.length)&&t.addIssue({code:`custom`,message:`Each selected criterion requires exactly one distinct grader result.`});for(let n of e.grades){let{contentHash:e,...r}=n;C(r)!==e&&t.addIssue({code:`custom`,message:`A retained criterion changed.`});for(let e of n.components){let{contentHash:n,...r}=e;C(r)!==n&&t.addIssue({code:`custom`,message:`A retained grader result changed.`})}}}),M=f().trim().min(1).max(200),N=f().regex(/^[a-f0-9]{64}$/),ft=i(`kind`,[c({kind:a(`hosted_chat`),modelId:M,localRuntime:r([c({providerId:a(`custom-openai-compatible`),configurationHash:N,maximumRequestCostUsd:u().finite().positive().max(1e4),requestTimeoutMs:u().int().min(1e3).max(3e5)}).strict(),ot,ct]).optional(),candidate:c({artifactId:M,contentHash:N,preparationId:M,dispatchId:M,baseProfileId:M,workerImageDigest:f().regex(/^sha256:[a-f0-9]{64}$/)}).strict().optional(),harness:de.optional(),messages:ut.optional(),maxOutputTokens:u().int().min(1).max(1e6).default(1e5),temperature:u().min(0).max(2).default(0),topP:u().gt(0).max(1).default(1)}).strict(),c({kind:a(`fixture`),fixtureId:M}).strict()]),pt=c({schemaVersion:a(`openpond.modelStarterAttemptRequest.v1`),operationId:M,teamId:M,modelProjectId:M.nullable(),project:lt.optional(),taskset:O,taskId:M,environmentSeed:u().int().min(0).max(2147483647).default(0),policy:ft}).strict();c({modelProjectId:M.nullable(),taskset:O,afterTaskId:M.optional(),limit:u().int().min(1).max(100).default(25)}).strict(),c({modelProjectId:M.nullable(),taskset:O,available:l(),unavailableReason:f().max(1e3).nullable(),models:n(c({id:M,name:f().max(500),maximumOutputTokens:u().int().min(1).max(1e6).optional()}).strict()).max(200),tasks:n(c({id:M,split:se,inputPreview:f().max(500),fixtures:n(c({id:M,label:f().max(200)}).strict()).max(1e3)}).strict()).max(100),nextCursor:M.nullable()}).strict();var mt=c({schemaVersion:a(`openpond.modelStarterAttemptSummary.v1`),id:M,revision:u().int().positive(),request:pt,status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]),policySnapshot:c({modelId:M,provider:M,upstreamModelId:f().min(1).max(500),configurationHash:N}).strict().nullable(),createdAt:t(),startedAt:t().nullable(),completedAt:t().nullable(),cleanupComplete:l(),resultAvailable:l(),score:u().min(0).max(1).nullable(),gradingStatus:d([`not_started`,`pending`,`scored`,`unscorable`,`not_configured`]).default(`not_started`),passed:l().nullable(),outputPreview:f().max(500).nullable(),error:c({code:M,message:f().max(2e3)}).strict().nullable()}).strict().superRefine((e,t)=>{e.request.policy.kind===`fixture`!=(e.policySnapshot===null)&&t.addIssue({code:`custom`,path:[`policySnapshot`],message:`Fixture execution must not claim a model identity.`}),e.status===`completed`&&(!e.cleanupComplete||!e.resultAvailable||!e.completedAt)&&t.addIssue({code:`custom`,message:`Completed attempts require retained results and confirmed cleanup.`})});c({modelProjectId:M,afterId:M.optional(),limit:u().int().min(1).max(100).default(25)}).strict(),c({items:n(mt).max(100),nextCursor:M.nullable()}).strict(),c({schemaVersion:a(`openpond.modelStarterAttemptResult.v1`),attempt:mt,output:f().max(1048576).nullable(),messages:n(o(f(),p())).max(4002),composition:he.nullable(),grade:me.nullable().optional(),experimentGrade:dt.nullable().optional(),environment:c({status:d([`completed`,`budget_exhausted`,`cancelled`,`timed_out`,`policy_failure`,`environment_failure`]),collected:l(),definition:O,initialStateHash:N.nullable(),finalStateHash:N.nullable(),attemptHash:N}).strict(),native:Pe.optional(),providerRequestIds:n(M).max(1001),contentHash:N}).strict().superRefine((e,t)=>{let n=e.attempt.request.policy.kind===`hosted_chat`?e.attempt.request.policy.harness:void 0;!!n!=!!e.native&&t.addIssue({code:`custom`,path:[`native`],message:`Standalone Harness results require their exact native evidence.`}),e.native&&(!n||C(n)!==C(e.native.source)||e.native.modelConfigurationHash!==e.attempt.policySnapshot?.configurationHash||e.environment.initialStateHash!==null||e.environment.finalStateHash!==null||e.environment.status===`completed`&&(!e.environment.collected||e.output===null||e.native.partialOutputAvailable||e.native.outputHash!==C(e.output)))&&t.addIssue({code:`custom`,path:[`native`],message:`Native evidence differs from the admitted source, model or output.`}),e.experimentGrade&&((e.grade||e.composition)&&t.addIssue({code:`custom`,message:`Selected Experiment grading cannot claim another grade population.`}),(![`completed`,`failed`].includes(e.attempt.status)||!e.attempt.resultAvailable||e.attempt.score!==e.experimentGrade.score||e.attempt.gradingStatus!==e.experimentGrade.gradingStatus||e.attempt.passed!==(e.experimentGrade.score===null?null:e.experimentGrade.passed))&&t.addIssue({code:`custom`,message:`Experiment grading differs from its retained attempt summary.`})),e.grade&&(e.composition&&t.addIssue({code:`custom`,path:[`grade`],message:`An ordinary grade cannot also claim a Reward composition.`}),(![`completed`,`failed`].includes(e.attempt.status)||!e.attempt.resultAvailable||e.attempt.score!==e.grade.score||e.attempt.gradingStatus!==e.grade.gradingStatus||e.attempt.passed!==(e.grade.score===null?null:e.grade.passed))&&t.addIssue({code:`custom`,path:[`grade`],message:`Ordinary grade differs from its terminal attempt summary.`}))});var P=f().trim().min(1).max(240),ht=f().regex(/^[a-f0-9]{64}$/),gt=c({maxOutputTokens:u().int().min(1).max(1e6).default(4096),temperature:u().min(0).max(2).default(0),topP:u().gt(0).max(1).default(1)}).strict(),F=c({kind:a(`hosted_harness`),modelId:P,maxOutputTokens:gt.shape.maxOutputTokens.removeDefault().optional(),temperature:gt.shape.temperature.removeDefault().optional(),topP:gt.shape.topP.removeDefault().optional(),localRuntime:ct.optional(),candidate:c({artifactId:P,contentHash:ht,preparationId:P,dispatchId:P,baseProfileId:P,workerImageDigest:f().regex(/^sha256:[a-f0-9]{64}$/)}).strict().optional(),sourceCandidate:ue.shape.sourceCandidate,reasoningEffort:d([`off`,`low`,`medium`,`high`,`xhigh`,`max`]).optional(),profileRepositoryId:P,source:ue,modelConfigurationHash:ht,packageHash:ht,externalDatasetBinding:S.optional()}).strict().superRefine((e,t)=>{JSON.stringify(e.externalDatasetBinding??null)!==JSON.stringify(e.source.externalDatasetBinding??null)&&t.addIssue({code:`custom`,message:`The policy and source must bind the same external Dataset recipe.`}),JSON.stringify(e.sourceCandidate??null)!==JSON.stringify(e.source.sourceCandidate??null)&&t.addIssue({code:`custom`,message:`The policy and prepared source must bind the same actual source candidate.`})}),I=f().trim().min(1).max(200),_t=f().regex(/^[a-f0-9]{64}$/),vt=i(`kind`,[ft.options[0],c({kind:a(`fixture`)}).strict(),F]),yt=c({schemaVersion:a(`openpond.modelTasksetRunRequest.v1`),operationId:I,teamId:I,modelProjectId:I.nullable(),name:f().trim().min(1).max(200).optional(),project:lt.optional(),taskset:O,policy:vt,population:n($e).min(1).max(1e4)}).strict().superRefine((e,t)=>{new Set(e.population.map(e=>e.receiptId)).size!==e.population.length&&t.addIssue({code:`custom`,path:[`population`],message:`Run receipt identities must be unique.`}),e.population.some(t=>t.fixtureId!==null!=(e.policy.kind===`fixture`))&&t.addIssue({code:`custom`,path:[`population`],message:`Only fixture runs require a fixture identity for every member.`}),e.population.some(e=>!/^(0|[1-9][0-9]*)$/.test(e.seed)||Number(e.seed)>2147483647)&&t.addIssue({code:`custom`,path:[`population`],message:`Hosted environment seeds must be integers from 0 to 2147483647.`}),e.policy.kind===`hosted_harness`&&e.modelProjectId!==null&&t.addIssue({code:`custom`,path:[`modelProjectId`],message:`A Harness target retains its released Profile binding separately from Model artifacts.`}),e.policy.kind===`hosted_harness`&&new Set(e.population.map(e=>xe([e.taskId,e.seed]))).size!==e.population.length&&t.addIssue({code:`custom`,path:[`population`],message:`Harness case identities must be unique for each task and seed.`})}),bt=c({schemaVersion:a(`openpond.modelTasksetRunSummary.v1`),id:I,revision:u().int().positive(),teamId:I,modelProjectId:I.nullable(),name:f().trim().min(1).max(200).optional(),project:lt.optional(),operationId:I,taskset:O,policyKind:d([`hosted_chat`,`fixture`,`hosted_harness`]),manifestHash:_t,status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]),totalCount:u().int().min(1).max(1e4),counts:c({pending:u().int().nonnegative(),running:u().int().nonnegative(),completed:u().int().nonnegative(),failed:u().int().nonnegative(),cancelled:u().int().nonnegative()}).strict(),createdAt:t(),startedAt:t().nullable(),completedAt:t().nullable(),cleanupComplete:l(),resultAvailable:l(),score:u().min(0).max(1).nullable(),metricName:f().min(1).max(240),error:c({code:I,message:f().max(2e3)}).strict().nullable()}).strict().superRefine((e,t)=>{Object.values(e.counts).reduce((e,t)=>e+t,0)!==e.totalCount&&t.addIssue({code:`custom`,path:[`counts`],message:`Run counts must account for the complete population.`}),e.status!==`completed`&&e.score!==null&&t.addIssue({code:`custom`,path:[`score`],message:`Only completed evaluations publish a metric value.`}),[`completed`,`failed`,`cancelled`].includes(e.status)&&(!e.completedAt||e.counts.pending||e.counts.running)&&t.addIssue({code:`custom`,message:`Terminal runs require complete population accounting.`}),e.status===`cancelled`&&!e.cleanupComplete&&t.addIssue({code:`custom`,message:`Cancelled runs require confirmed cleanup.`}),e.resultAvailable&&![`completed`,`failed`,`cancelled`].includes(e.status)&&t.addIssue({code:`custom`,message:`Run results require terminal population accounting.`}),e.status===`completed`&&(!e.cleanupComplete||!e.resultAvailable)&&t.addIssue({code:`custom`,message:`Completed evaluations require retained results and confirmed cleanup.`})}),xt=c({summary:bt,request:yt,manifest:et,policySnapshot:mt.shape.policySnapshot}).strict();c({modelProjectId:I.optional(),projectId:lt.shape.id.optional(),afterId:I.optional(),limit:u().int().min(1).max(100).default(25)}).strict(),c({items:n(bt).max(100),nextCursor:I.nullable()}).strict(),c({schemaVersion:a(`openpond.modelTasksetRunResult.v1`),run:xt,receipts:n(Ye).min(1).max(1e4),metric:le.nullable(),contentHash:_t}).strict();var St=f().min(1).max(200).refine(e=>![`__proto__`,`prototype`,`constructor`].includes(e),`Use an ordinary field name.`),Ct=n(St).min(1).max(20),wt=d([`input`,`output`,`expectedOutput`,`evaluatorContext`]),Tt=c({destination:wt,path:Ct,source:wt,sourcePath:n(St).max(20)}).strict(),Et=n(Tt).max(100).superRefine((e,t)=>{for(let[n,r]of e.entries())e.slice(0,n).some(e=>e.destination===r.destination&&(Dt(e.path,r.path)||Dt(r.path,e.path)))&&t.addIssue({code:`custom`,path:[n,`path`],message:`Mapped destination paths must not overlap.`})});function Dt(e,t){return e.length<=t.length&&e.every((e,n)=>e===t[n])}var L=f().trim().min(1).max(200),R=f().regex(/^[a-f0-9]{64}$/),Ot=c({id:L,revision:u().int().positive(),contentHash:R}).strict(),kt=c({id:L,version:f().min(1).max(200),contentHash:R,name:f().trim().min(1).max(500).optional(),feedbackKey:ge,release:Ot.nullable(),mappings:Et.optional()}).strict(),At=c({id:L,version:f().min(1).max(200),contentHash:R,mappings:Et}).strict(),jt=c({operationId:L,id:L.optional(),expectedRevision:u().int().nonnegative(),request:yt,maximumCostUsd:u().finite().positive().max(1e4),graders:n(At).min(1).max(100).optional()}).strict().superRefine((e,t)=>{e.request.name||t.addIssue({code:`custom`,path:[`request`,`name`],message:`Name this Experiment before saving it.`}),!!e.id!=e.expectedRevision>0&&t.addIssue({code:`custom`,path:[`expectedRevision`],message:`An update requires its saved definition revision.`}),Math.round(e.maximumCostUsd*1e6)<1&&t.addIssue({code:`custom`,path:[`maximumCostUsd`],message:`The budget must be at least $0.000001.`}),e.graders&&new Set(e.graders.map(e=>e.id)).size!==e.graders.length&&t.addIssue({code:`custom`,path:[`graders`],message:`Select each grader once.`})});c({schemaVersion:a(`openpond.experimentDefinition.v1`),id:L,teamId:L,revision:u().int().positive(),request:yt,maximumCostUsd:u().finite().positive().max(1e4),graders:n(kt).min(1).max(100),createdAt:t(),updatedAt:t()}).strict().extend({contentHash:R}),c({operationId:L,definition:Ot}).strict(),c({projectId:L.optional(),datasetHash:R.optional(),search:f().trim().max(200).optional(),status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]).optional(),afterId:L.optional(),limit:u().int().min(1).max(100).default(30)}).strict(),c({definition:Ot.nullable(),configurationHash:R.optional(),admissionRequestHash:R.optional(),sourceExperimentId:L.optional(),maximumCostUsd:u().finite().positive().max(1e4),graders:n(kt).min(1).max(100)}).strict().superRefine((e,t)=>{e.definition===null&&(!e.configurationHash||!e.admissionRequestHash)&&t.addIssue({code:`custom`,path:[`configurationHash`],message:`A flat Experiment requires its sealed configuration hash.`})}),c({operationId:L,profileRepositoryId:L,definitionId:L,modelId:L,localRuntime:F.shape.localRuntime,maxOutputTokens:F.shape.maxOutputTokens,temperature:F.shape.temperature,topP:F.shape.topP,reasoningEffort:d([`off`,`low`,`medium`,`high`,`xhigh`,`max`]).optional(),maximumCostUsd:u().finite().positive().max(1e4),externalDatasetBinding:S.optional(),sourceCandidate:F.shape.sourceCandidate}).strict(),c({request:yt,manifest:et,maximumCostUsd:u().finite().positive().max(1e4),graders:n(kt).min(1).max(100)}).strict();var z=f().trim().min(1).max(240),B=f().trim().min(1),Mt=f().trim().min(8).max(256),Nt=o(f(),p()).default({}),V=z.nullable(),Pt=d([`slash_train`,`session_menu`,`bulk_selection`,`training_page`,`task_candidate`]),Ft=d([`defaults`,`customize`]),It=d([`automated`,`manual`]);c({schemaVersion:a(`openpond.trainingSourceEstimate.v1`),sessionId:z,messageCount:u().int().nonnegative(),estimatedTokens:u().int().nonnegative(),textBytes:u().int().nonnegative()}),c({query:f().max(500).default(``),offset:u().int().nonnegative().default(0),limit:u().int().min(1).max(100).default(20),candidates:n(c({sessionId:z,title:f().trim().min(1).max(500),updatedAt:B})).max(500).default([])});var Lt=c({sessionId:z,title:f().trim().min(1).max(500),updatedAt:B,snippet:f().max(2e3).nullable()});c({schemaVersion:a(`openpond.trainingChatSearchResult.v1`),query:f(),offset:u().int().nonnegative(),limit:u().int().positive(),total:u().int().nonnegative(),hasMore:l(),indexedChats:u().int().nonnegative(),totalChats:u().int().nonnegative(),indexing:l(),entries:n(Lt)});var Rt=d([`policy_failure`,`grader_failure`,`environment_failure`,`infrastructure_failure`,`timeout`,`cancelled`]),zt=c({graderId:z.nullable(),path:f().trim().min(1).max(1e3),language:d([`javascript`,`typescript`,`python`,`json`,`text`]),content:f().max(2e6),sha256:Mt,declaredSha256:Mt.nullable(),integrity:d([`verified`,`unverified`,`mismatch`])}),Bt=c({protocolVersion:f().trim().min(1).max(200),module:f().trim().min(1).max(1e3),moduleSha256:Mt,command:n(f().max(2e3)).max(100),cwd:f().trim().min(1).max(4e3),maxTurns:u().int().positive().max(100)});c({schemaVersion:a(`openpond.tasksetGraderDetails.v1`),taskset:_,sourceTasksetId:z,graders:n(Te).max(1e3),runtime:Bt.nullable(),sources:n(zt).max(1e3),unavailableReason:f().max(2e3).nullable()});var Vt=c({schemaVersion:a(`openpond.taskAttempt.v1`),id:z,tasksetId:z,taskId:z,split:Se,attempt:u().int().nonnegative(),seed:u().int(),modelRef:b.nullable(),startedAt:B,completedAt:B,output:o(f(),p()),runtimeEventRefs:n(z).max(1e4).default([]),artifactRefs:n(z).max(1e4).default([]),privilegedOutcomeRef:V,infrastructureError:f().trim().min(1).max(1e4).nullable(),costUsd:u().nonnegative().nullable(),latencyMs:u().int().nonnegative(),userInterventions:u().int().nonnegative().default(0),metadata:Nt}),Ht=c({schemaVersion:a(`openpond.taskAttemptArtifact.v1`),id:z,tasksetId:z,taskId:z,attemptId:z,kind:d([`raw_model_response`,`runtime_trace`,`environment_state`,`grader_evidence`,`output_artifact`]),path:f().trim().min(1).max(4e3),mediaType:f().trim().min(1).max(200).nullable().optional(),sha256:Mt,sizeBytes:u().int().nonnegative(),createdAt:B,metadata:Nt}),Ut=c({graderId:z,graderVersion:f().trim().min(1).max(100),score:u().min(0).max(1).nullable(),passed:l(),hardGate:l(),rewardEligible:l(),feedback:f().trim().max(2e4).nullable(),evidenceRefs:n(z).max(1e4).default([]),judge:b.nullable().optional().default(null),calibrationStatus:d([`not_applicable`,`pending`,`passed`,`failed`])}),Wt=c({schemaVersion:a(`openpond.gradeResult.v1`),id:z,attemptId:z,graderSetHash:Mt,score:u().min(0).max(1).nullable(),passed:l(),components:n(Ut).min(1).max(1e3),failureClass:Rt.nullable(),feedback:n(f().trim().min(1).max(2e4)).max(1e3).default([]),rewardEligible:l(),rewardComposition:he.optional(),createdAt:B});d([`stable_hash_top_n`,`rft_easy_curriculum_v1`]);var Gt=c({schemaVersion:a(`openpond.graderAuditReport.v1`),id:z,tasksetId:z,tasksetHash:Mt,fixtureRefs:n(z).min(1).max(1e5),gradeRefs:n(z).min(1).max(1e5),passed:l(),hackingChecksPassed:l(),leakageChecksPassed:l(),infrastructureSafetyPassed:l(),failures:n(c({fixtureId:z,label:Me,gradeId:z,reason:f().trim().min(1).max(5e3)})).max(1e5),createdAt:B});function Kt(e,t=`New model`){let n=e?.trim().split(/\s+/).filter(Boolean)??[];return n.length===0?t:n.slice(0,5).join(` `)}var qt=c({schemaVersion:a(`openpond.baseModelPreference.v1`),modelId:z,revision:f().trim().min(1).max(256).nullable(),tokenizerRevision:f().trim().min(1).max(256).nullable(),chatTemplateHash:f().trim().min(8).max(256).nullable(),modelAssetId:V,source:d([`managed`,`local`,`builtin`])}),Jt=c({schemaVersion:a(`openpond.taskCreationRequest.v1`),id:z,profileId:z,surface:Pt,mode:Ft,entryMode:It.default(`manual`),resourceIntent:d([`workproduct`,`dataset`]).default(`workproduct`),buildIntent:ke.default(`demonstrations`),buildSpecification:Ae.nullable().default(null),objective:f().trim().min(1).max(2e4).nullable(),methodHint:d([`sft`,`dpo`,`grpo`,`ppo`]).nullable().default(null),preferredBaseModelId:z.nullable().default(null),preferredBaseModel:qt.nullable().default(null),sourceIds:n(z).max(1e5),candidateId:V,analysisModel:b.nullable(),analysisReasoningEffort:ae.nullable().default(null),createImproveRunId:V.default(null),targetIntent:c({kind:d([`agent`,`skill`,`extension`,`model`,`configuration`]).nullable(),id:V,displayName:f().trim().min(1).max(500).nullable(),operation:d([`create`,`improve`])}).default({kind:`model`,id:null,displayName:null,operation:`create`}),disclosure:c({status:d([`not_required`,`pending`,`approved`,`declined`]),content:a(`raw_excerpts`),sourceIds:n(z).max(1e5),providerModel:b.nullable(),approvalId:V,approvedAt:B.nullable()}).default({status:`not_required`,content:`raw_excerpts`,sourceIds:[],providerModel:null,approvalId:null,approvedAt:null}),createdAt:B}),Yt=c({schemaVersion:a(`openpond.taskCreationSnapshot.v1`),id:z,request:Jt,state:d([`planning`,`awaiting_disclosure_approval`,`awaiting_questions`,`recommendation_ready`,`awaiting_materialization_approval`,`materializing`,`validating`,`ready`,`blocked`,`failed`,`cancelled`]),proposal:at.nullable(),materializedTasksetId:V,disclosureApprovalId:V,materializationApprovalId:V,blockingQuestions:n(c({id:z,kind:d([`objective`,`consent`,`success_signal`,`privacy_licensing`,`interpretation`]),prompt:f().trim().min(1).max(5e3),answer:f().trim().min(1).max(2e4).nullable()})).default([]),transcript:n(c({id:z,role:d([`user`,`assistant`,`system`,`tool`]),text:f().max(1e5),createdAt:B})).max(1e4).default([]),repairHistory:n(je).max(1e3).default([]),blockedReason:f().trim().min(1).max(1e4).nullable(),createdAt:B,updatedAt:B});c({schemaVersion:a(`openpond.taskCreationTranscript.v1`),creationId:z,profileId:z,messages:Yt.shape.transcript,updatedAt:B});var Xt=f().trim().min(1).max(200),Zt=c({operationId:Xt,execution:w,executionKind:a(`recorded_evidence`).optional(),graders:n(ve).min(1).max(100),maximumCostUsd:u().finite().min(1e-6).max(1e4),mappings:n(c({graderId:Xt,fields:Et}).strict()).max(100).optional()}).strict().superRefine((e,t)=>{new Set(e.graders.map(e=>e.id)).size!==e.graders.length&&t.addIssue({code:`custom`,path:[`graders`],message:`Select each grader once.`}),e.mappings&&(new Set(e.mappings.map(e=>e.graderId)).size!==e.mappings.length||e.mappings.some(t=>!e.graders.some(e=>e.id===t.graderId)))&&t.addIssue({code:`custom`,path:[`mappings`],message:`Map each selected grader at most once.`})});c({schemaVersion:a(`openpond.experimentScoringPass.v1`),id:Xt,teamId:Xt,revision:u().int().positive(),request:Zt,graders:n(kt).min(1).max(100),status:d([`queued`,`running`,`cancelling`,`completed`,`failed`,`cancelled`]),totalCount:u().int().min(1).max(1e4),counts:c({pending:u().int().nonnegative(),running:u().int().nonnegative(),scored:u().int().nonnegative(),unavailable:u().int().nonnegative(),failed:u().int().nonnegative(),cancelled:u().int().nonnegative()}).strict(),resultAvailable:l(),createdAt:t(),startedAt:t().nullable(),completedAt:t().nullable(),error:c({code:Xt,message:f().max(2e3)}).strict().nullable()}).strict().extend({contentHash:f().regex(/^[a-f0-9]{64}$/)});var Qt=c({sourcePass:w,graders:Zt.shape.graders,mappings:Zt.shape.mappings,maximumCostUsd:u().finite().positive().max(1e4)}).strict(),$t=c({externalDatasetBinding:S,scoring:Qt.optional(),modelId:f().min(1).max(300),maximumCostUsd:u().finite().positive().max(1e4)}).strict(),en=c({source:d([`local`,`github`,`openpond_git`]),repositoryId:f().trim().min(1),profileId:f().trim().min(1)}).strict(),H=f().min(1).max(240),U=f().regex(/^[a-f0-9]{64}$/),W=c({schemaVersion:a(`openpond.experimentImprovement.v1`),id:H,operationId:H,requestHash:U,actorId:H,teamId:H,projectId:H.nullable(),ownerWorkspaceId:H,ownerRevision:u().int().nonnegative(),ownerSourceRevision:f().min(1).max(500),baseRelease:w,profileRef:en,baseProfileSourceRevision:f().min(1).max(500),evidence:c({manifest:D,result:E}).strict(),component:c({kind:d([`skill`,`workflow`,`agent`,`instruction`]),path:f().min(1).max(2e3),workflowId:H.optional()}).strict(),mode:d([`manual`,`llm_assisted`]),status:d([`draft`,`authoring_queued`,`authoring`,`frozen`,`testing`,`ready_for_review`,`adopting`,`accepted`,`rolled_back`,`declined`,`failed`,`cancelled`]),revision:u().int().positive(),limits:c({maximumCostUsd:u().finite().positive(),maximumAuthoringSteps:u().int().min(1).max(500),maximumDurationMs:u().int().min(1e3).max(864e5)}).strict(),authoringSteps:u().int().nonnegative().max(500),work:c({sessionId:H,turnId:H,revision:u().int().positive(),startedAt:f().datetime()}).strict().nullable(),frozen:c({release:w,sourceRevision:f().min(1).max(500),profileSourceRevision:f().min(1).max(500),authoringHash:U,partitionHash:U,executableSourceHash:U,protectedClosureHash:U,candidateRevision:u().int().positive(),diff:n(c({path:f(),before:U.nullable(),after:U.nullable()}).strict()).max(1e4)}).strict().nullable(),test:c({recipe:$t,recipeHash:f().regex(/^[a-f0-9]{64}$/),baselineOperationId:f(),candidateOperationId:f(),cancelRequestedAt:f().datetime().nullable(),dispatching:c({kind:d([`baseline`,`candidate`]),token:H,pid:u().int().positive(),processStart:f().nullable(),startedAt:f().datetime()}).strict().nullable()}).strict().nullable(),comparison:c({baseline:c({manifest:D,result:E}).strict(),candidate:c({manifest:D,result:E}).strict(),frozenHash:U,humanProjection:We.optional()}).strict().nullable(),adoptionIntent:c({operationId:H,expectedOwnerRevision:u().int().nonnegative(),frozenHash:U,humanProjection:We.optional()}).strict().nullable(),acceptance:w.nullable(),rollback:w.nullable(),error:f().max(2e3).nullable(),createdAt:f().datetime(),updatedAt:f().datetime(),contentHash:U}).strict();function tn(e){let t=W.parse(e),{contentHash:n,...r}=t;if(C(r)!==n)throw Error(`Improvement state integrity failed.`);if(T(t.evidence),t.evidence.manifest.teamId!==t.teamId||t.evidence.result.status!==`completed`||!t.evidence.result.cases.some(e=>e.status===`completed`))throw Error(`Improvement requires usable completed owner evidence.`);if(t.test&&(C(t.test.recipe)!==t.test.recipeHash||pe(t.test.recipe.externalDatasetBinding).profileId!==t.profileRef.profileId))throw Error(`The test recipe changed its exact selected Profile.`);if(t.comparison?.humanProjection){let e=Ke(t.comparison.humanProjection);if(e.teamId!==t.teamId||e.projectId!==t.projectId||C(e.baseline)!==C(t.comparison.baseline)||C(e.candidate)!==C(t.comparison.candidate))throw Error(`Human comparison projection differs from its exact owner evidence.`)}if(t.comparison&&(T(t.comparison.baseline),T(t.comparison.candidate),!t.frozen||t.comparison.frozenHash!==C(t.frozen)||!_e(t.comparison.baseline,t.comparison.candidate).comparable))throw Error(`Improvement comparison is not bound to a comparable frozen candidate.`);if(t.status===`ready_for_review`&&(!t.frozen||!t.comparison))throw Error(`Improvement is missing qualification.`);if(t.status===`adopting`&&(!t.adoptionIntent||!t.frozen||t.adoptionIntent.frozenHash!==C(t.frozen)))throw Error(`Improvement adoption intent is not bound to its tested revision.`);if(t.status===`accepted`&&(!t.acceptance||!t.comparison||!t.frozen))throw Error(`Improvement acceptance is missing its exact qualification receipt.`);if(t.status===`rolled_back`&&(!t.acceptance||!t.rollback))throw Error(`Profile rollback is missing its durable owner receipt.`);return t}var nn=c({states:n(W).max(50),nextCursor:H.nullable()}).strict(),rn=c({operationId:H,ownerWorkspaceId:H,expectedOwnerRevision:u().int().nonnegative(),baseRelease:w,profileRef:en,profileSourceRevision:f().min(1).max(500),projectId:H.nullable(),evidence:w,component:W.shape.component,mode:W.shape.mode,limits:W.shape.limits}).strict(),an=i(`operation`,[c({operation:a(`list`),limit:u().int().min(1).max(50).default(30),cursor:H.optional()}).strict(),c({operation:a(`options`),evidence:w}).strict(),c({operation:a(`start`),request:rn}).strict(),c({operation:d([`read`,`humanComparisonOptions`]),id:H}).strict(),c({operation:a(`instructions`),id:H}).strict(),c({operation:a(`edit`),id:H,revision:u().int().positive(),expectedFileHash:U,text:f().max(25e4)}).strict(),c({operation:a(`mode`),id:H,revision:u().int().positive(),mode:d([`manual`,`llm_assisted`])}).strict(),c({operation:a(`author`),id:H,revision:u().int().positive(),modelId:H,prompt:f().min(1).max(2e4)}).strict(),c({operation:d([`freeze`,`finishAuthoring`,`useCandidate`,`discard`,`cancel`,`rollback`]),id:H,revision:u().int().positive()}).strict(),c({operation:a(`test`),id:H,revision:u().int().positive(),recipe:$t}).strict(),c({operation:a(`compare`),id:H,revision:u().int().positive(),baseline:w,candidate:w,humanSelections:Ve.optional()}).strict()]),on=c({teamId:H,evidence:w,scoring:Qt.optional(),profiles:n(c({id:U,name:f(),profileRef:en,ownerWorkspaceId:H,ownerRevision:u().int().nonnegative(),baseRelease:w,profileSourceRevision:f(),components:n(c({kind:d([`skill`,`instruction`,`agent`,`workflow`]),path:f(),agentId:f().optional(),workflowId:f().optional(),label:f(),binding:S}).strict()).max(500)}).strict()).max(100)}).strict(),sn=c({candidateId:H,actorId:H,teamId:H,projectId:H.nullable(),candidateRevision:u().int().positive(),recipeHash:U,baseline:c({manifest:D,result:E}).strict(),candidate:c({manifest:D,result:E}).strict(),contentHash:U}).strict();function cn(e){let t=sn.parse(e),{contentHash:n,...r}=t;if(C(r)!==n)throw Error(`Accepted Human discovery integrity failed.`);for(let e of[t.baseline,t.candidate])if(T(e),e.manifest.teamId!==t.teamId||e.manifest.lineage?.humanProjection)throw Error(`Discovery must retain original owner evidence.`);return t}var G=f().trim().min(1).max(240),K=f().regex(/^[a-f0-9]{64}$/),ln=c({schemaVersion:a(`openpond.advancedRefinerEvaluation.v1`),operationId:G,actorId:G,teamId:G,projectId:G.nullable(),mode:d([`review_quality`,`downstream_adaptation`]),profileRef:c({source:d([`local`,`github`,`openpond_git`]),repositoryId:G,profileId:G}).strict(),sourceRevision:f().min(1).max(500),baselineRelease:w,refinerRelease:w,evidence:w,externalDatasetBinding:S,adaptationDatasetBinding:S,privateClosureHash:K,adaptationTaskIds:n(G).min(1).max(1e4),holdoutTaskIds:n(G).min(1).max(1e4),adaptationSplit:d([`train`,`validation`,`test`,`frozen_eval`]),holdoutSplit:d([`train`,`validation`,`test`,`frozen_eval`]),environmentPolicy:i(`kind`,[c({kind:a(`resettable`),environmentHash:K}).strict(),c({kind:a(`frozen_observations`),environmentHash:K}).strict()]),maximumCostUsd:u().finite().positive().max(1e4),maximumDurationMs:u().int().min(1e3).max(864e5),maximumModelSteps:u().int().min(1).max(1e3),contentHash:K}).strict();function un(e){let t=ln.parse(e),{contentHash:n,...r}=t;if(C(r)!==n)throw Error(`Advanced Refiner evaluation admission changed.`);let i=pe(t.externalDatasetBinding);if(i.profileId!==t.profileRef.profileId||i.protectedProfileClosureHash!==t.privateClosureHash)throw Error(`Advanced evaluation changed its Profile or protected evaluator closure.`);if(t.adaptationSplit===t.holdoutSplit||new Set([...t.adaptationTaskIds,...t.holdoutTaskIds]).size!==t.adaptationTaskIds.length+t.holdoutTaskIds.length)throw Error(`Adaptation and private holdout must be disjoint ordered cases.`);let a=pe(t.adaptationDatasetBinding);if(a.profileId!==i.profileId||a.packageHash!==i.packageHash||C(a.dataset)!==C(i.dataset)||C(a.target)!==C(i.target)||a.protectedProfileClosureHash!==i.protectedProfileClosureHash)throw Error(`This native advanced adapter needs separate adaptation and holdout recipes from the same exact released package and target.`);if(a.split!==t.adaptationSplit||i.split!==t.holdoutSplit)throw Error(`Advanced split policies differ from their exact admitted recipes.`);let o=new Set(i.population.map(e=>e.taskId)),s=new Set(a.population.map(e=>e.taskId));if(t.holdoutTaskIds.some(e=>!o.has(e))||t.adaptationTaskIds.some(e=>!s.has(e)))throw Error(`Advanced evaluation selected cases outside its exact Dataset recipe.`);return t}var dn=c({pin:ln,modelId:G,profileId:G,model:we,seed:u().int(),reasoningEffort:d([`none`,`off`,`low`,`medium`,`high`,`xhigh`,`max`]).nullable()}).strict(),fn=i(`operation`,[c({operation:a(`start`),request:dn}).strict(),c({operation:d([`read`,`cancel`,`resume`]),id:G}).strict(),c({operation:a(`list`)}).strict()]),pn=c({taskId:G,sourceCaseHash:K,reviewSessionId:G,reviewTurnId:G,inputHarness:w,trigger:w,decision:d([`no_action`,`route_deterministically`,`queue_refiner`]),outcome:w.nullable(),proposal:w.nullable(),validations:n(w).max(100),grade:c({id:G,contentHash:K,score:u().min(0).max(1).nullable(),passed:l(),failureClass:f().nullable()}).strict(),resetComplete:l()}).strict(),mn=c({schemaVersion:a(`openpond.advancedRefinerReviewReceipt.v1`),id:G,runId:G,pinHash:K,ownerActorId:G,teamId:G,privateClosureHash:K,refinerRelease:w,cases:n(pn).min(1).max(1e4),reviewScore:u().min(0).max(1).nullable(),completedCases:u().int().nonnegative(),noActionCases:u().int().nonnegative(),routedCases:u().int().nonnegative(),proposalCases:u().int().nonnegative(),costUsd:u().finite().nonnegative().nullable(),cleanupComplete:l(),createdAt:f().datetime(),contentHash:K}).strict(),hn=on.shape.profiles.element,gn=on.extend({profiles:n(hn.extend({components:n(hn.shape.components.element.extend({adaptationSplits:n(c({split:d([`train`,`validation`,`test`,`frozen_eval`]),taskIds:n(G)})),environmentHash:K,resettable:l(),reviewQuality:l()}))})),refinerReleases:n(w),activeRefiner:w.nullable(),models:n(c({id:G,profileId:G,name:f(),location:d([`local`,`hosted`])}))}),_n=c({operation:a(`prepare`),evidence:w,profileOptionId:G,componentLabel:G,mode:d([`review_quality`,`downstream_adaptation`]),adaptationSplit:d([`train`,`validation`,`test`,`frozen_eval`]),operationId:G,refinerRelease:w,maximumCostUsd:u().positive(),maximumDurationMs:u().int().positive(),maximumModelSteps:u().int().positive()}).strict();r([fn,_n,c({operation:a(`options`),evidence:w}).strict()]);var vn=c({modelRequests:u().int().nonnegative(),knownSpendUsd:u().nonnegative(),totalSpendUsd:u().nonnegative().nullable(),heldUsd:u().nonnegative(),uncertainRequests:u().int().nonnegative()}).strict(),q=y.nullable(),J=u().int().nonnegative(),yn=f().regex(/^[a-f0-9]{40}$/),bn=c({name:f().trim().min(1).max(160),startedAt:m,completedAt:m,durationMs:u().nonnegative(),clock:d([`monotonic`,`provider`,`wall`]),outcome:d([`succeeded`,`failed`,`cancelled`])}).strict(),xn=c({schemaVersion:a(`openpond.correlatedTelemetryReceipt.v1`),stage:d([`training`,`evaluation`,`deployment`,`inference`]),correlation:c({modelRunId:q,modelVersionId:q,policyVersion:J.nullable(),taskId:q,rolloutGroupId:q,providerResourceId:q,deploymentId:q,inferenceRequestId:q}).strict(),spans:n(bn).max(1e4),usage:c({promptTokens:J.nullable(),generatedTokens:J.nullable(),gpuSeconds:u().nonnegative().nullable(),workerActiveSeconds:u().nonnegative().nullable(),optimizerSteps:J.nullable(),rolloutGroups:J.nullable(),successfulTrajectories:J.nullable(),failedTrajectories:J.nullable(),peakGpuMemoryBytes:J.nullable(),peakGpuUtilizationPercent:u().min(0).max(100).nullable()}).strict(),resource:c({provider:y,resourceIds:n(y).max(1e3),gpuType:f().trim().min(1).max(300).nullable(),gpuCount:J.nullable(),baseProfileId:q,baseRepository:f().trim().min(1).max(512).nullable(),baseRevision:yn.nullable(),adapterContentHash:v.nullable()}).strict(),cost:c({currency:a(`USD`),providerReportedUsd:u().nonnegative().nullable(),quotedHourlyUsd:u().nonnegative().nullable(),estimatedUsd:u().nonnegative().nullable(),methodologyVersion:f().trim().min(1).max(160),pricingInputs:o(f().trim().min(1).max(160),u().finite()),unitEstimates:o(f().trim().min(1).max(160),u().nonnegative())}).strict(),recordedAt:m,contentHash:v}).strict(),Y=f().trim().min(1).max(240),X=f().regex(/^[a-f0-9]{64}$/),Z=f().datetime({offset:!0}),Q=c({id:Y,contentHash:X}).strict(),Sn=Q.extend({revision:u().int().positive()}).strict(),Cn=c({providerId:Y,modelId:f().trim().min(1).max(300)}).strict(),wn=d([`correction`,`sibling_verification`,`cumulative_known`,`development`,`retained`,`frozen_final`,`training_eligible`]),Tn=c({taskId:Y,taskContentHash:X,panelRole:wn,passLabel:Y.nullable(),optimizerEligible:l(),criticalInvariantIds:n(Y).max(100).default([])}).strict().superRefine((e,t)=>{e.optimizerEligible!==(e.panelRole===`correction`)&&t.addIssue({code:`custom`,path:[`optimizerEligible`],message:`Only correction cases may be optimizer eligible.`})}),En=c({schemaVersion:a(`openpond.continualBenchIssuePacket.v1`),id:Y,revision:u().int().positive(),contentHash:X,familyId:Y,familyLabel:f().trim().min(1).max(500),severity:d([`low`,`medium`,`high`,`critical`]),cases:n(Tn).min(2).max(1e4),duplicateEvidence:Q,leakageEvidence:Q,priorExposureEvidence:Q,observedAt:Z,reviewedAt:Z,queuedAt:Z.nullable(),runAt:Z.nullable(),createdBy:Y,createdAt:Z}).strict().superRefine((e,t)=>{e.cases.some(e=>e.panelRole===`correction`)||t.addIssue({code:`custom`,path:[`cases`],message:`An issue packet requires a correction case.`}),e.cases.some(e=>e.panelRole===`sibling_verification`)||t.addIssue({code:`custom`,path:[`cases`],message:`An issue packet requires a sibling-verification case.`}),new Set(e.cases.map(e=>e.taskId)).size!==e.cases.length&&t.addIssue({code:`custom`,path:[`cases`],message:`Issue-packet task ids must be unique.`})}),Dn=c({id:Y,role:wn,passLabel:Y.nullable(),taskset:Sn,familyIds:n(Y).min(1).max(1e5),taskCount:u().int().positive().max(1e6),sealedAt:Z}).strict().superRefine((e,t)=>{let n=e.role===`correction`||e.role===`sibling_verification`||e.role===`cumulative_known`;n!==(e.passLabel!==null)&&t.addIssue({code:`custom`,path:[`passLabel`],message:n?`This panel role requires a pass label.`:`This panel role is series-wide and cannot carry a pass label.`})}),On=i(`kind`,[c({kind:a(`base_model`),id:Y,revision:f().trim().min(1).max(500),contentHash:X}).strict(),c({kind:a(`model_version`),id:Y,contentHash:X}).strict(),c({kind:a(`external_reference`),id:Y,model:Cn,providerRevision:f().trim().min(1).max(500),contentHash:X}).strict()]),kn=c({scheduleEntryId:Y,ordinal:u().int().nonnegative(),label:Y,role:d([`seed`,`daily_residual`,`rank_candidate`,`weekly_rollup`,`full_refresh`]),parentRule:d([`base_model`,`previous_release`,`seed_release`,`accepted_daily_head`,`accepted_seed`]),trainableRank:u().int().positive().max(1024),correctionPanelIds:n(Y).min(1).max(1e3),optimizerGroupsPerTask:u().int().positive().max(1e3),trajectoriesPerGroup:u().int().min(4).max(1e3)}).strict(),An=c({revision:u().int().positive(),requireAllAttemptsTerminal:a(!0),criticalCorrectionPassRate:u().min(0).max(1),siblingPassRate:u().min(0).max(1),behavioralRetentionRate:u().min(0).max(1),maximumRetainedRegressionPoints:u().min(0).max(100),blockCriticalPriorRegression:l(),authorizedClaims:n(d([`systems_complete`,`correction_absorbed`,`issue_generalized`,`continually_current`,`frontier_pareto_result`])).min(1).max(5)}).strict(),jn=c({schemaVersion:a(`openpond.continualBenchmarkProtocol.v1`),id:Y,revision:u().int().positive(),contentHash:X,ownerId:Y,createdAt:Z,sealedAt:Z,creationReceipt:Q,predecessorSeries:c({id:Y,revision:u().int().positive(),contentHash:X}).strict().nullable(),scenarioPack:Q,issueFamilyLedger:Q,issuePacketRelease:Q,panels:n(Dn).min(6).max(1e4),grader:Q,judge:c({release:Q,rubricRelease:Q,calibrationRelease:Q,model:Cn}).strict().nullable(),policies:c({base:On,master:On.nullable(),externalReferences:n(On).min(1).max(20)}).strict(),invariants:c({systemPromptHash:X,toolSchema:Q,application:Q,harness:Q,runtime:Q,workerImage:Q,autoRefiner:c({enabled:l(),release:Q.nullable()}).strict()}).strict(),schedule:n(kn).min(1).max(1e3),evaluation:c({seeds:n(u().int()).min(3).max(100),repetitions:u().int().min(3).max(20),powerRule:Q,pairedBootstrapSamples:u().int().min(1e3).max(1e6),confidenceLevel:a(.95)}).strict(),resources:c({maximumTrainingGpuSeconds:u().int().positive(),maximumEvaluationGpuSeconds:u().int().positive(),maximumProviderSpendUsd:u().positive(),maximumTotalSpendUsd:u().positive(),maximumConcurrentGroups:u().int().positive().max(1e3)}).strict(),currencyThresholds:An}).strict().superRefine((e,t)=>{let n=e.panels.map(e=>e.id);new Set(n).size!==n.length&&t.addIssue({code:`custom`,path:[`panels`],message:`Protocol panel ids must be unique.`});for(let n of[`development`,`retained`,`frozen_final`,`training_eligible`])e.panels.filter(e=>e.role===n).length!==1&&t.addIssue({code:`custom`,path:[`panels`],message:`A protocol requires exactly one ${n} panel.`});[...e.schedule].sort((e,t)=>e.ordinal-t.ordinal).some((e,t)=>e.ordinal!==t)&&t.addIssue({code:`custom`,path:[`schedule`],message:`Protocol schedule ordinals must be contiguous from zero.`});let r=new Set(n);e.schedule.some(e=>e.correctionPanelIds.some(e=>!r.has(e)))&&t.addIssue({code:`custom`,path:[`schedule`],message:`Every scheduled correction panel must belong to the protocol.`})}),Mn=d([`draft`,`active`,`completed`,`archived`]),Nn=d([`draft`,`ready`,`queued`,`running`,`candidate`,`accepted`,`rejected`,`no_signal`,`failed`,`cancelled`]),Pn=d([`seed`,`daily_residual`,`rank_candidate`,`weekly_rollup`,`full_refresh`]),Fn=d([`base_model`,`previous_release`,`seed_release`,`accepted_daily_head`,`accepted_seed`]),In=d([`seed_taskset`,`nightly_selection`,`daily_cohort_union`,`eligible_task_pool`]),Ln=d([`live_client`,`evaluation_attempt`,`replay_evidence`,`manual`]),Rn=c({id:y,ordinal:u().int().nonnegative(),label:f().trim().min(1).max(100),role:Pn,parentRule:Fn,taskSource:In,trainableRank:u().int().positive().max(1024),minimumTasks:u().int().nonnegative().max(1e5),maximumTasks:u().int().positive().max(1e5)}).strict().superRefine((e,t)=>{e.minimumTasks>e.maximumTasks&&t.addIssue({code:`custom`,path:[`minimumTasks`],message:`The minimum task count cannot exceed the maximum task count.`});let{parentRules:n,taskSource:r}={seed:{parentRules:[`base_model`],taskSource:`seed_taskset`},daily_residual:{parentRules:[`previous_release`,`accepted_daily_head`],taskSource:`nightly_selection`},rank_candidate:{parentRules:[`seed_release`],taskSource:`eligible_task_pool`},weekly_rollup:{parentRules:[`seed_release`,`accepted_seed`],taskSource:`daily_cohort_union`},full_refresh:{parentRules:[`base_model`],taskSource:`eligible_task_pool`}}[e.role];n.includes(e.parentRule)||t.addIssue({code:`custom`,path:[`parentRule`],message:`${e.role} requires parent rule ${n.join(` or `)}.`}),e.taskSource!==r&&t.addIssue({code:`custom`,path:[`taskSource`],message:`${e.role} requires task source ${r}.`}),e.ordinal===0&&e.role!==`seed`&&t.addIssue({code:`custom`,path:[`role`],message:`The first series entry must be the seed.`})}),zn=c({id:y,version:u().int().positive(),requireCheckpoint:l(),requireAppliedOptimizerUpdate:l(),minimumCurrentCohortMeanImprovement:u().finite(),maximumRetainedMeanRegression:u().finite().nonnegative(),blockCriticalInvariantRegression:l(),automaticDailyAdvancement:l()}).strict(),Bn=c({schemaVersion:a(`openpond.modelComparisonSeries.v1`),id:y,profileId:y,modelProjectId:y,name:f().trim().min(1).max(500),objective:f().trim().min(1).max(2e4),status:Mn,revision:u().int().positive(),productionBinding:c({role:a(`chat_manual`),roleTargetId:y}).strict(),baseModel:c({id:y,revision:f().trim().min(1).max(256)}).strict(),seedTaskset:_,eligibleTaskPool:_,evaluationTasksets:c({development:_,retained:_,frozenFinal:_}).strict(),grader:g,benchmarkProtocol:jn.nullable().default(null),automaticEvaluation:c({enabled:l()}).strict().default({enabled:!1}),residualProfile:c({profileId:y,serializedEnvelopeRank:u().int().positive().max(1024),maximumEnabledRank:u().int().positive().max(1024),topology:a(`uniform_block_masked`)}).strict(),schedule:n(Rn).min(1).max(1e3),scheduleSealedAt:m.nullable(),advancementPolicy:zn,executionPolicy:c({startWhenReady:a(!1)}).strict(),acceptedSeedEntryId:y.nullable(),acceptedDailyHeadEntryId:y.nullable(),promotedBindingId:y.nullable(),createdBy:y,createdAt:m,updatedAt:m}).strict().superRefine((e,t)=>{[...e.schedule].sort((e,t)=>e.ordinal-t.ordinal).some((e,t)=>e.ordinal!==t)&&t.addIssue({code:`custom`,path:[`schedule`],message:`Schedule ordinals must be unique and contiguous from zero.`}),(new Set(e.schedule.map(e=>e.id)).size!==e.schedule.length||new Set(e.schedule.map(e=>e.label)).size!==e.schedule.length)&&t.addIssue({code:`custom`,path:[`schedule`],message:`Schedule ids and labels must be unique.`}),e.residualProfile.maximumEnabledRank>e.residualProfile.serializedEnvelopeRank&&t.addIssue({code:`custom`,path:[`residualProfile`],message:`Maximum enabled rank cannot exceed the serialized envelope.`});let n=Object.values(e.evaluationTasksets);new Set(n.map(e=>`${e.id}:${e.revision}:${e.contentHash}`)).size!==n.length&&t.addIssue({code:`custom`,path:[`evaluationTasksets`],message:`Development, retained, and frozen-final cohorts require distinct immutable Taskset releases.`});let r=e.schedule.find(e=>e.role===`seed`);if(!r||e.schedule.filter(e=>e.role===`seed`).length!==1){t.addIssue({code:`custom`,path:[`schedule`],message:`A series requires exactly one seed entry.`});return}let i=e.schedule.filter(e=>e.role===`rank_candidate`);i.length===1&&t.addIssue({code:`custom`,path:[`schedule`],message:`A causal rank comparison requires at least two rank candidates.`}),new Set(i.map(e=>e.trainableRank)).size!==i.length&&t.addIssue({code:`custom`,path:[`schedule`],message:`Causal rank candidates must use distinct trainable ranks.`}),[r.trainableRank+e.schedule.filter(e=>e.role===`daily_residual`).reduce((e,t)=>e+t.trainableRank,0),...e.schedule.filter(e=>e.role===`weekly_rollup`).map(e=>r.trainableRank+e.trainableRank),...e.schedule.filter(e=>e.role===`rank_candidate`).map(e=>r.trainableRank+e.trainableRank),...e.schedule.filter(e=>e.role===`full_refresh`).map(e=>e.trainableRank)].some(t=>t>e.residualProfile.maximumEnabledRank)&&t.addIssue({code:`custom`,path:[`schedule`],message:`A declared branch exceeds the maximum enabled rank.`})}),Vn=c({source:Ln,taskIds:n(y).min(1).max(1e5),observedFrom:m,observedTo:m,reviewedAt:m,reviewedBy:y,sourceTaskset:_}).strict(),Hn=Vn.extend({derivedFamilyKeys:n(y).min(1).max(1e5),familyDerivation:c({algorithm:y,sourceTasksetHash:v}).strict(),selectionHash:v}).strict(),Un=i(`kind`,[c({kind:a(`base_model`),id:y,revision:f().trim().min(1).max(256)}).strict(),c({kind:a(`model_version`),id:y,contentHash:v}).strict()]),Wn=c({id:y,branchOrdinal:u().int().nonnegative(),rank:u().int().positive().max(1024),offsetStart:u().int().nonnegative(),offsetEnd:u().int().positive(),optimizationRole:d([`frozen`,`trainable`]),artifactLineageId:y.nullable()}).strict().superRefine((e,t)=>{e.offsetEnd-e.offsetStart!==e.rank&&t.addIssue({code:`custom`,path:[`offsetEnd`],message:`Block offsets must span exactly the declared rank.`})}),Gn=c({disposition:d([`advance`,`hold`,`no_signal`]),policy:c({id:y,version:u().int().positive()}).strict(),reasonCodes:n(y).min(1).max(100),summary:f().trim().min(1).max(5e3),decidedBy:y,decidedAt:m}).strict(),Kn=c({evaluationRunId:y,modelVersionId:y,taskset:_,grader:g,cohortRole:d([`current`,`correction`,`sibling_verification`,`cumulative_known`,`development`,`retained`,`prior_disclosed`,`frozen_final`]),panelId:y.nullable().default(null),protocol:c({id:y,revision:u().int().positive(),contentHash:v}).strict().nullable().default(null)}).strict(),qn=c({attemptOrdinal:u().int().positive(),trainingPlanId:y.nullable(),modelRunId:y.nullable(),terminalStatus:d([`failed`,`cancelled`]),queuedAt:m.nullable(),startedAt:m.nullable(),completedAt:m}).strict(),Jn=c({seriesId:y,entryId:y,scheduleEntryId:y,ordinal:u().int().nonnegative(),releaseHash:v}).strict(),Yn=c({schemaVersion:a(`openpond.modelComparisonSeriesEntry.v1`),id:y,seriesId:y,profileId:y,modelProjectId:y,scheduleEntryId:y,releaseHash:v,ordinal:u().int().nonnegative(),label:f().trim().min(1).max(100),role:Pn,branch:d([`daily`,`rank_candidate`,`weekly_rollup`,`full_refresh`]),status:Nn,parent:Un,taskset:_,sourceTasksets:n(_).min(1).max(1e3),taskSelection:Hn.nullable(),trainableRank:u().int().positive().max(1024),serializedEnvelopeRank:u().int().positive().max(1024),enabledCumulativeRank:u().int().positive().max(1024),trainableBlockId:y,residualBlocks:n(Wn).min(1).max(1024),attemptOrdinal:u().int().positive().default(1),priorRunAttempts:n(qn).max(100).default([]),trainingPlanId:y.nullable(),modelRunId:y.nullable(),modelVersionId:y.nullable(),evaluations:n(Kn).max(1e5),decision:Gn.nullable(),promotionBindingId:y.nullable(),queuedAt:m.nullable(),startedAt:m.nullable(),completedAt:m.nullable(),createdAt:m,updatedAt:m}).strict().superRefine((e,t)=>{let n=[...e.residualBlocks].sort((e,t)=>e.offsetStart-t.offsetStart);n.some((e,t)=>e.offsetStart!==(t===0?0:n[t-1].offsetEnd))&&t.addIssue({code:`custom`,path:[`residualBlocks`],message:`Residual block offsets must be contiguous from zero.`});let r=n.at(-1)?.offsetEnd??0;(r!==e.enabledCumulativeRank||r>e.serializedEnvelopeRank)&&t.addIssue({code:`custom`,path:[`enabledCumulativeRank`],message:`Enabled rank must match the block inventory and fit the envelope.`});let i=e.residualBlocks.filter(e=>e.optimizationRole===`trainable`);(i.length!==1||i[0]?.id!==e.trainableBlockId||i[0]?.rank!==e.trainableRank)&&t.addIssue({code:`custom`,path:[`trainableBlockId`],message:`Exactly one block must be trainable during the Run and match trainableRank.`});let a=e.priorRunAttempts.map(e=>e.attemptOrdinal);(e.priorRunAttempts.length!==e.attemptOrdinal-1||a.some((e,t)=>e!==t+1))&&t.addIssue({code:`custom`,path:[`priorRunAttempts`],message:`Prior Run attempts must be a contiguous history before the active attempt.`})});c({seriesId:y,scheduleEntryId:y,taskSelection:Vn.nullable(),expectedSeriesRevision:u().int().positive()}).strict();var Xn=d([`base_reference`,`lora_adapter`]),Zn=c({baseModel:c({source:a(`huggingface`),repoId:y,revision:f().regex(/^[a-f0-9]{40}$/),configHash:v,tokenizerHash:v,licenseId:y,gated:l()}).strict(),processor:c({repository:y,revision:f().regex(/^[a-f0-9]{40}$/),configHash:v}).strict()}).strict().superRefine((e,t)=>{(e.baseModel.repoId!==e.processor.repository||e.baseModel.revision!==e.processor.revision)&&t.addIssue({code:`custom`,path:[`processor`],message:`Reward Model processor must match the pinned scorer base model.`})}),Qn=c({schemaVersion:a(`openpond.modelVersion.v1`),id:y,modelId:y,profileId:y,version:u().int().nonnegative(),kind:Xn,status:d([`available`,`failed`]),baseModel:qt,taskset:_,comparisonSeriesEntry:Jn.nullable().optional(),evaluationTasksetRef:_.nullable().optional(),releaseGraph:c({resolvedBundleHash:v,profileRelease:_,harnessRelease:g,agentRelease:g.nullable(),grader:c({id:y,contentHash:v}).strict()}).strict(),artifactLineageId:y.nullable(),adapterStatus:d([`not_trained`,`trained`]),createdAt:m,contentHash:v}).strict().superRefine((e,t)=>{e.kind===`base_reference`&&(e.version!==0||e.adapterStatus!==`not_trained`||e.artifactLineageId!==null)&&t.addIssue({code:`custom`,message:`A base-reference Model Version must be version 0 without adapter lineage.`}),e.kind===`lora_adapter`&&(e.version===0||e.adapterStatus!==`trained`||!e.artifactLineageId)&&t.addIssue({code:`custom`,message:`A LoRA Model Version must be version 1+ with adapter lineage.`})}),$n=c({schemaVersion:a(`openpond.rewardModelVersion.v1`),id:y,modelId:y,profileId:y,version:u().int().positive(),role:a(`reward`),status:d([`available`,`failed`]),scope:d([`synthetic_smoke`,`human_preference`]),baseModel:qt,runtime:Zn.nullable().default(null),taskset:_,preferenceDatasetRelease:g,releaseGraph:c({resolvedBundleHash:v,profileRelease:_,harnessRelease:g,grader:c({id:y,contentHash:v}).strict()}).strict(),artifacts:c({checkpoint:g.extend({objectRef:f().regex(/^r2:\/\/[A-Za-z0-9._/-]+$/),files:n(c({path:f().trim().min(1).max(512),sizeBytes:u().int().positive(),sha256:v}).strict()).min(4).max(128)}).strict(),adapter:g,scalarHead:g,bucketHead:g.nullable(),processorRelease:g}).strict(),qualificationReport:g.nullable(),createdAt:m,contentHash:v}).strict(),er=c({schemaVersion:a(`openpond.rewardModelRunReceipt.v1`),provider:y,providerRunId:y,resolvedBundleHash:v,finalCheckpoint:g.extend({objectRef:f().regex(/^r2:\/\/[A-Za-z0-9._/-]+$/)}).strict(),adapter:g,scalarHead:g,bucketHead:g.nullable(),processorRelease:g,optimizerEvidence:g,managedExecutionReceipt:g.nullable().default(null),parameterDeltaHash:v,cleanup:c({computeReleased:l(),providerTerminalObserved:l()}).strict(),contentHash:v}).strict(),tr=s(e=>{if(!e||typeof e!=`object`||Array.isArray(e))return e;let t=e;if(t.tasksetRelease!==void 0)return e;let n=t.taskset;if(!n||typeof n!=`object`||Array.isArray(n))return e;let r=n;return typeof r.id!=`string`||typeof r.contentHash!=`string`?e:{...t,tasksetRelease:{id:r.id,contentHash:r.contentHash}}},c({schemaVersion:a(`openpond.rewardModelRun.v1`),id:y,rewardModelId:y,rewardModelVersionId:y.nullable(),profileId:y,role:a(`reward`),scope:d([`synthetic_smoke`,`human_preference`]),status:d([`prepared`,`running`,`succeeded`,`failed`,`cancelled`]),taskset:_,tasksetRelease:g,preferenceDatasetRelease:g,recipeRelease:g,destinationId:y,quote:c({maximumSpendUsd:u().positive(),hourlyCostUsd:u().nonnegative().nullable()}).strict(),managedRunId:y.nullable(),progress:c({completedSteps:u().int().nonnegative(),totalSteps:u().int().positive(),latestLoss:u().finite().nullable()}).strict(),receipt:er.nullable(),qualificationReport:g.nullable(),accruedSpendUsd:u().nonnegative().nullable(),failureOwner:d([`authoring`,`admission`,`provider`,`runner`,`artifact`,`qualification`,`cleanup`]).nullable(),failure:f().trim().min(1).max(5e3).nullable(),startedAt:m,completedAt:m.nullable(),updatedAt:m}).strict()).superRefine((e,t)=>{e.status===`succeeded`&&!e.receipt&&t.addIssue({code:`custom`,message:`A successful Reward Model Run requires a canonical receipt.`}),e.status!==`succeeded`&&e.rewardModelVersionId!==null&&t.addIssue({code:`custom`,message:`Only a successful Reward Model Run can publish a Reward Model Version.`})}),nr=c({schemaVersion:a(`openpond.modelRunReceipt.v1`),provider:y,providerRunId:y,assignmentHash:v,resultHash:v,transcriptHash:v,traceHash:v.nullable(),resolvedBundleHash:v,artifactPath:f().trim().min(1).max(2e3),cleanup:c({computeReleased:l(),tunnelClosed:l()}).strict(),telemetry:xn.nullable().default(null),contentHash:v}).strict(),$=c({inputTokens:u().int().nonnegative(),outputTokens:u().int().nonnegative(),totalTokens:u().int().nonnegative(),costUsd:u().nonnegative().nullable()}).strict(),rr=f().regex(/^(?:[a-f0-9]{40}|[a-f0-9]{64})$/),ir=c({phase:d([`baseline`,`adaptation`,`candidate_adaptation`,`candidate`]),taskId:y,attemptId:y,sessionId:y.nullable(),turnId:y.nullable(),passed:l(),score:u().min(0).max(1).nullable(),failureClass:Rt.nullable(),inputTokens:u().int().nonnegative(),outputTokens:u().int().nonnegative(),totalTokens:u().int().nonnegative(),latencyMs:u().int().nonnegative(),costUsd:u().nonnegative().nullable(),startedAt:m}).strict(),ar=c({targetTaskCount:u().int().nonnegative(),comparedTaskCount:u().int().nonnegative(),passedTaskCount:u().int().nonnegative(),failedTaskCount:u().int().nonnegative(),lowerTaskCount:u().int().nonnegative(),higherTaskCount:u().int().nonnegative(),unchangedTaskCount:u().int().nonnegative()}).strict(),or=c({target:a(`all_tasks_lower`),targetTaskCount:u().int().nonnegative(),comparedTaskCount:u().int().nonnegative(),passedTaskCount:u().int().nonnegative(),failedTaskCount:u().int().nonnegative(),lowerTaskCount:u().int().nonnegative(),higherTaskCount:u().int().nonnegative(),unchangedTaskCount:u().int().nonnegative(),baselineTokens:u().int().nonnegative(),refinedTokens:u().int().nonnegative(),tokenDelta:u().int(),tokenDeltaPercent:u().finite().nullable(),complete:l(),passed:l(),cohorts:c({adaptation:ar,heldOut:ar}).strict()}).strict(),sr=c({baseline:$,adaptation:$,candidateAdaptation:$,candidate:$,refiner:$,grader:$}).strict(),cr=c({usage:sr,observedSpendUsd:u().nonnegative(),attempts:n(ir).max(1e4)}).strict(),lr=c({id:y,contentHash:v,artifactPath:f().trim().min(1).max(2e3)}).strict(),ur=c({schemaVersion:a(`openpond.modelEvaluationReceipt.v1`),benchmarkId:y,resultManifest:c({id:y,contentHash:v,artifactPath:f().trim().min(1).max(2e3)}).strict(),stages:c({baseline:g,adaptation:g,refiner:g,candidateAdaptation:g,candidate:g,comparison:g}).strict(),usage:sr,quality:c({baselinePassRate:u().min(0).max(1),candidatePassRate:u().min(0).max(1),adaptationBaselinePassRate:u().min(0).max(1),adaptationCandidatePassRate:u().min(0).max(1),adaptationCandidatePassed:l(),heldOutCandidatePassed:l(),passed:l()}).strict(),foregroundTokenDelta:u().int(),foregroundTokenDeltaPercent:u().finite().nullable(),taskEfficiency:or.optional(),efficiency:c({grossForegroundTokenSavings:u().int(),overheadTokens:u().int().nonnegative(),firstPassNetTokenSavings:u().int(),breakEvenReuseCount:u().int().nonnegative().nullable(),amortizedTokenSavings:u().int(),amortizedReuseCount:u().int().positive()}).strict(),budget:c({maximumSpendUsd:u().nonnegative(),observedSpendUsd:u().nonnegative(),enforced:l()}).strict(),evidenceSnapshot:g,lineage:c({adaptationEvidenceHash:v,refinerInputHash:v,refinerOutcomeHash:v,validationHash:v,applyReceiptHash:v,candidateRelease:g,valid:l()}).strict(),invalidReasons:n(f().trim().min(1).max(2e3)).max(100),attempts:n(ir).max(1e4).default([]),terminalClassification:d([`improved`,`no_improvement`,`regressed`,`inconclusive`,`cancelled`,`infrastructure_failure`]),profileGit:c({ref:f().trim().min(1).max(2e3),commit:rr,baseCommit:rr}).strict().nullable(),contentHash:v}).strict(),dr=c({schemaVersion:a(`openpond.modelEvaluationStopReceipt.v1`),benchmarkId:y,terminalClassification:a(`inconclusive`),stopReason:d([`candidate_harness_unchanged`,`candidate_lineage_invalid`]),reason:f().trim().min(1).max(5e3),stoppedAfter:a(`refiner`),baselineHarness:g,candidateHarness:g,refiner:c({id:y,contentHash:v,outcomeCount:u().int().nonnegative()}).strict(),usage:sr,budget:c({maximumSpendUsd:u().nonnegative(),observedSpendUsd:u().nonnegative(),enforced:l()}).strict(),evidenceSnapshot:g,attempts:n(ir).max(1e4),contentHash:v}).strict(),fr=c({level:a(.95),lower:u().finite(),upper:u().finite()}).strict().superRefine((e,t)=>{e.lower>e.upper&&t.addIssue({code:`custom`,message:`A confidence interval's lower bound cannot exceed its upper bound.`})}),pr=c({schemaVersion:a(`openpond.modelComparisonBenchmarkReceipt.v1`),benchmarkId:a(`model-comparison`),target:c({kind:d([`base_model`,`model_version`,`external_reference`]),label:f().trim().min(1).max(300),modelVersionId:y.nullable(),model:b.nullable()}).strict(),taskset:_,grader:g,sampling:c({seeds:n(u().int()).min(1).max(100),repetitions:u().int().positive().max(20)}).strict(),deterministic:c({attemptedTaskCount:u().int().nonnegative(),completedTaskCount:u().int().nonnegative(),passedTaskCount:u().int().nonnegative(),failedTaskCount:u().int().nonnegative(),meanScore:u().finite().nullable(),passRate:u().min(0).max(1).nullable(),passRateCi95:fr.nullable()}).strict(),judge:c({judgeModel:b,judgeRelease:g,rubricRelease:g,calibrationRelease:g,score:u().min(0).max(100),scoreCi95:fr,comparisonCount:u().int().positive(),wins:u().int().nonnegative(),ties:u().int().nonnegative(),losses:u().int().nonnegative(),referenceLabel:f().trim().min(1).max(300)}).strict().superRefine((e,t)=>{e.wins+e.ties+e.losses!==e.comparisonCount&&t.addIssue({code:`custom`,message:`Judge win, tie, and loss counts must equal comparisonCount.`})}).nullable(),attempts:n(c({attemptId:y.nullable().default(null),taskId:y,seed:u().int(),repetition:u().int().nonnegative(),status:d([`succeeded`,`failed`]),deterministicScore:u().finite().nullable(),passed:l().nullable(),judgeScore:u().min(0).max(100).nullable(),judgePreference:d([`win`,`tie`,`loss`]).nullable(),transcriptHash:v.nullable(),traceHash:v.nullable(),transcriptArtifact:c({artifactPath:f().trim().min(1).max(2e3),jsonPointer:f().trim().min(1).max(1e3)}).strict().nullable().default(null),traceArtifact:c({artifactPath:f().trim().min(1).max(2e3),jsonPointer:f().trim().min(1).max(1e3)}).strict().nullable().default(null),latencyMs:u().int().nonnegative().nullable().default(null),failureClass:Rt.nullable()}).strict()).max(1e5),usage:c({policy:$.nullable(),judge:$.nullable(),observedSpendUsd:u().nonnegative().nullable(),evaluationGpuSeconds:u().nonnegative().nullable().default(null)}).strict(),managedServing:c({jobId:y,terminalState:d([`completed`,`cancelled`,`failed`,`budget_exhausted`]),sourcePolicyVersion:u().int().nonnegative(),sourceAdapterSha256:v.nullable(),servedPolicyVersion:u().int().nonnegative().nullable(),servedAdapterSha256:v.nullable(),accruedSpendUsd:u().nonnegative(),cleanupAttestationHash:v,resourceCount:u().int().nonnegative(),activeResourceCount:a(0)}).strict().nullable().default(null),evidenceSnapshot:c({id:y,contentHash:v,artifactPath:f().trim().min(1).max(2e3)}).strict(),completedAt:m,contentHash:v}).strict().superRefine((e,t)=>{let n=e.deterministic;n.completedTaskCount>n.attemptedTaskCount&&t.addIssue({code:`custom`,message:`Completed task count cannot exceed attempted task count.`}),n.passedTaskCount+n.failedTaskCount!==n.completedTaskCount&&t.addIssue({code:`custom`,message:`Passed and failed task counts must equal completed task count.`}),n.completedTaskCount===0!=(n.meanScore===null)&&t.addIssue({code:`custom`,message:`Deterministic mean score is available exactly when tasks completed.`}),n.completedTaskCount===0!=(n.passRate===null)&&t.addIssue({code:`custom`,message:`Deterministic pass rate is available exactly when tasks completed.`})}),mr=c({advancedEvaluation:ln.optional(),benchmarkId:a(`harness-refiner`),model:b,upstreamModel:c({providerId:y,modelId:f().trim().min(1).max(300),revision:f().trim().min(1).max(300),pricing:c({version:f().trim().min(1).max(300),source:f().trim().min(1).max(300),effectiveAt:m,inputUsdPerMillionTokens:u().nonnegative(),cachedInputUsdPerMillionTokens:u().nonnegative(),outputUsdPerMillionTokens:u().nonnegative()}).strict().optional()}).strict(),reasoningEffort:f().trim().min(1).max(100).nullable(),seeds:n(u().int()).min(1).max(100),repetitions:u().int().positive().max(20),maximumSpendUsd:u().nonnegative(),attemptPlan:n(c({stage:d([`baseline`,`adaptation`,`candidate_adaptation`,`candidate`]),split:f().trim().min(1).max(100),taskIds:n(y).min(1).max(1e5),attemptCount:u().int().positive()}).strict()).length(4)}).strict(),hr=c({benchmarkId:a(`model-comparison`),target:c({kind:d([`base_model`,`model_version`,`external_reference`]),label:f().trim().min(1).max(300),modelVersionId:y.nullable(),model:b.nullable()}).strict(),grader:g,judge:c({model:b,release:g,rubricRelease:g,calibrationRelease:g,referenceLabel:f().trim().min(1).max(300)}).strict().nullable(),seeds:n(u().int()).min(1).max(100),repetitions:u().int().positive().max(20),maximumSpendUsd:u().nonnegative(),series:c({id:y,protocol:c({id:y,revision:u().int().positive(),contentHash:v}).strict()}).strict().nullable().default(null),panel:c({id:y,role:d([`correction`,`sibling_verification`,`cumulative_known`,`development`,`retained`,`frozen_final`]),passLabel:y.nullable()}).strict().nullable().default(null),comparisonPair:c({entryId:y,parent:Un,candidateModelVersionId:y}).strict().nullable().default(null),attemptPlan:n(c({stage:a(`comparison`),split:f().trim().min(1).max(100),taskIds:n(y).min(1).max(1e5),attemptCount:u().int().positive()}).strict()).length(1)}).strict(),gr=r([mr,hr]),_r=c({schemaVersion:a(`openpond.modelRun.v1`),id:y,modelId:y,modelVersionId:y.nullable(),profileId:y,kind:d([`rollout_smoke`,`training`,`evaluation`]),status:d([`prepared`,`running`,`succeeded`,`failed`,`cancelled`]),method:d([`sft`,`dpo`,`grpo`,`ppo`,`sdft`,`opd`,`opsd`,`sdpo`]).nullable(),destinationId:y.nullable(),taskset:_,comparisonSeriesEntry:Jn.nullable().optional(),evaluationTasksetRef:_.nullable().optional(),harnessRelease:g.nullable().optional(),quote:c({maximumSpendUsd:u().nonnegative(),hourlyCostUsd:u().nonnegative().nullable()}).strict().nullable(),evaluation:gr.nullable().default(null),evaluationProgress:c({stage:d([`baseline`,`adaptation`,`refiner`,`candidate_adaptation`,`candidate`,`comparison`]),completedAttempts:u().int().nonnegative(),totalAttempts:u().int().positive(),accounting:cr.nullable().default(null),advancedAccounting:vn.optional(),evidenceSnapshot:lr.nullable().default(null)}).strict().nullable().default(null),reward:c({raw:u(),components:o(f(),u())}).strict().nullable(),receipt:r([nr,ur,dr,pr,mn]).nullable(),adapterArtifactLineageId:y.nullable(),failure:f().trim().min(1).max(5e3).nullable(),startedAt:m,completedAt:m.nullable(),updatedAt:m}).strict().superRefine((e,t)=>{e.status===`succeeded`&&!e.receipt&&t.addIssue({code:`custom`,message:`A successful Model Run requires a canonical receipt.`}),e.kind===`evaluation`&&(e.method!==null||e.destinationId!==null||!e.evaluation||e.adapterArtifactLineageId!==null)&&t.addIssue({code:`custom`,message:`An evaluation Model Run requires evaluation configuration and cannot claim training method, destination, or adapter lineage.`}),e.kind===`evaluation`&&e.receipt&&e.receipt.schemaVersion!==`openpond.modelEvaluationReceipt.v1`&&e.receipt.schemaVersion!==`openpond.modelEvaluationStopReceipt.v1`&&e.receipt.schemaVersion!==`openpond.modelComparisonBenchmarkReceipt.v1`&&e.receipt.schemaVersion!==`openpond.advancedRefinerReviewReceipt.v1`&&t.addIssue({code:`custom`,message:`An evaluation Model Run requires an evaluation receipt.`}),e.kind!==`evaluation`&&(e.method===null||e.destinationId===null||e.quote===null)&&t.addIssue({code:`custom`,message:`A training or rollout Model Run requires method, destination, and quote.`}),e.kind===`rollout_smoke`&&e.adapterArtifactLineageId!==null&&t.addIssue({code:`custom`,message:`A rollout smoke cannot claim adapter artifacts.`})});function vr(e){let t=new Map;for(let n of e.attempts){let e=`${n.phase}\u0000${n.taskId}`,r=t.get(e);(!r||n.startedAt>r.startedAt||n.startedAt===r.startedAt&&n.attemptId>r.attemptId)&&t.set(e,n)}let n=[...t.values()],r=[[`adaptation`,`adaptation`,`candidate_adaptation`],[`held_out`,`baseline`,`candidate`]].flatMap(([e,t,r])=>{let i=new Map(n.filter(e=>e.phase===r).map(e=>[e.taskId,e]));return n.filter(e=>e.phase===t).flatMap(t=>{let n=i.get(t.taskId);return n?[{cohort:e,taskId:t.taskId,baseline:t,refined:n,tokenDelta:n.totalTokens-t.totalTokens}]:[]})}),i=new Set(n.filter(e=>e.phase===`baseline`||e.phase===`adaptation`).map(e=>e.taskId)).size,a=e.targetTaskCount??i;if(!Number.isInteger(a)||a<0)throw Error(`Benchmark task-efficiency target must be a non-negative integer.`);let o={adaptation:new Set(n.filter(e=>e.phase===`adaptation`).map(e=>e.taskId)).size,held_out:new Set(n.filter(e=>e.phase===`baseline`).map(e=>e.taskId)).size},s=(e,t)=>({targetTaskCount:t,comparedTaskCount:e.length,passedTaskCount:e.filter(e=>e.tokenDelta<0).length,failedTaskCount:e.filter(e=>e.tokenDelta>=0).length,lowerTaskCount:e.filter(e=>e.tokenDelta<0).length,higherTaskCount:e.filter(e=>e.tokenDelta>0).length,unchangedTaskCount:e.filter(e=>e.tokenDelta===0).length}),ee=s(r,a),c=r.reduce((e,t)=>e+t.baseline.totalTokens,0),l=r.reduce((e,t)=>e+t.refined.totalTokens,0),u=l-c,d=r.length===a;return{summary:or.parse({target:`all_tasks_lower`,...ee,baselineTokens:c,refinedTokens:l,tokenDelta:u,tokenDeltaPercent:c>0?u/c*100:null,complete:d,passed:d&&ee.passedTaskCount===a,cohorts:{adaptation:s(r.filter(e=>e.cohort===`adaptation`),o.adaptation),heldOut:s(r.filter(e=>e.cohort===`held_out`),o.held_out)}}),pairs:r}}var yr=f().trim().min(1).max(240),br=f().trim().min(1),xr=f().trim().min(8).max(256),Sr=o(f(),p()).default({}),Cr=c({schemaVersion:a(`openpond.localModelChatConfiguration.v1`).default(`openpond.localModelChatConfiguration.v1`),profile:d([`efficient`,`full_harness`,`custom`]).default(`efficient`),systemPromptMode:d([`lean`,`full_harness`,`custom`]).default(`lean`),customSystemPrompt:f().max(2e4).nullable().default(null),contextWindowTokens:u().int().min(128).max(32768).default(1024),maxOutputTokens:u().int().min(1).max(512).default(64),temperature:u().min(0).max(2).default(0),repetitionPenalty:u().min(.5).max(2).default(1.1),noRepeatNgramSize:u().int().min(0).max(10).default(3),compaction:d([`off`,`when_needed`]).default(`when_needed`),keepWarmSeconds:u().int().min(0).max(3600).default(300),updatedAt:br.nullable().default(null)}),wr=Cr.parse({}),Tr=c({schemaVersion:a(`openpond.managedAdapterServingProjection.v1`),teamId:yr.nullable().default(null),source:a(`sandbox_managed_rl`),sourceRef:yr,canonicalArtifactId:yr.nullable(),canonicalModelName:f().trim().min(1).max(256).nullable().default(null),canonicalArtifactState:d([`imported_unvalidated`,`evaluating`,`promotable`,`rejected`,`deleted`]).nullable(),canonicalDeploymentId:yr.nullable(),canonicalDeploymentState:d([`requested`,`deploying`,`ready`,`degraded`,`deleting`,`deleted`,`failed`]).nullable(),state:d([`pending`,`imported`,`ready`,`failed`]),customerBindingAllowed:l().default(!1),artifactContentHash:xr.nullable().default(null),baseProfileId:yr.nullable().default(null),publishedAt:br.nullable(),lastSyncedAt:br,lastError:f().trim().min(1).max(5e3).nullable()});function Er(e){let t=e.managedServing;return!!(t?.source===`sandbox_managed_rl`&&t.canonicalArtifactState===`promotable`&&t.customerBindingAllowed)}function Dr(e){return e.state===`ready`&&e.canonicalArtifactState===`promotable`&&e.canonicalDeploymentState===`ready`&&e.customerBindingAllowed&&!!e.teamId&&!!e.canonicalArtifactId&&!!e.canonicalDeploymentId}function Or(e){if(e.promotable&&e.frozenEvaluationArtifactId)return{kind:`source_frozen_evaluation`,evaluationArtifactId:e.frozenEvaluationArtifactId,canonicalArtifactId:null,canonicalDeploymentId:null};let t=e.managedServing;return t?.source===`sandbox_managed_rl`&&Dr(t)&&t.canonicalArtifactId&&t.canonicalDeploymentId?{kind:`sandbox_customer_binding`,evaluationArtifactId:null,canonicalArtifactId:t.canonicalArtifactId,canonicalDeploymentId:t.canonicalDeploymentId}:null}function kr(e){let t=e.managedServing;return t?Dr(t)&&t.canonicalArtifactId&&t.canonicalDeploymentId?{kind:`sandbox_serving`,canonicalArtifactId:t.canonicalArtifactId,canonicalDeploymentId:t.canonicalDeploymentId}:null:{kind:`source_artifact`,canonicalArtifactId:null,canonicalDeploymentId:null}}export{Pe as $,cn as A,Et as B,gn as C,on as D,nn as E,Ht as F,yt as G,At as H,Vt as I,at as J,st as K,Yt as L,qt as M,Wt as N,W as O,Gt as P,Fe as Q,Kt as R,fn as S,an as T,jt as U,kt as V,xt as W,et as X,tt as Y,Xe as Z,Jn as _,kr as a,En as b,yr as c,_r as d,Qn as f,vr as g,$n as h,Er as i,tn as j,$t as k,Sr as l,Zn as m,Cr as n,Or as o,tr as p,dt as q,Tr as r,xr as s,wr as t,br as u,Yn as v,un as w,wn as x,Bn as y,Ot as z};