@tangle-network/agent-eval 0.115.3 → 0.117.0

package/dist/traces.js

@@ -28,13 +28,14 @@ import {
   scoreTraceInsightReadiness,
   tokenizeDomainWords,
   traceAnalystOnRunComplete
-} from "./chunk-TLDB7WRY.js";
+} from "./chunk-YZPO4UHR.js";
 import {
   FAILURE_CLASSES,
   TRACE_SCHEMA_VERSION,
   aggregateLlm,
   argHash,
   groupBy,
+  hasCapturedToolArgs,
   isJudgeSpan,
   isLlmSpan,
   isRetrievalSpan,
@@ -45,13 +46,13 @@ import {
   runFailureClass,
   runsForScenario,
   toolSpans
-} from "./chunk-MHNQWM4I.js";
+} from "./chunk-LQUTGLOZ.js";
 import {
   TRACE_ANALYST_ACTOR_DESCRIPTION,
   TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION,
   TRACE_ANALYST_SUBAGENT_DESCRIPTION,
   analyzeTraces
-} from "./chunk-RPDDVKI7.js";
+} from "./chunk-4JLWXDYA.js";
 import {
   DEFAULT_REDACTION_RULES,
   REDACTION_VERSION,
@@ -60,6 +61,7 @@ import {
 } from "./chunk-GGE4NNQT.js";
 import {
   DEFAULT_TRACE_ANALYST_BUDGETS,
+  INPUT_VALUE,
   LLM_CACHED_TOKENS,
   LLM_CACHED_TOKEN_ATTR_KEYS,
   LLM_COST_ATTR_KEYS,
@@ -71,14 +73,18 @@ import {
   LLM_OUTPUT_TOKENS,
   LLM_OUTPUT_TOKEN_ATTR_KEYS,
   OPENINFERENCE_SPAN_KIND,
+  OUTPUT_VALUE,
   OtlpFileTraceStore,
   SPAN_KIND_ATTR_KEYS,
   SpanNotFoundError,
+  TOOL_ARGS_CAPTURED,
+  TOOL_LATENCY_MS,
   TOOL_NAME,
   TOOL_NAME_ATTR_KEYS,
   TRACE_ANALYST_TRUNCATION_MARKER_PREFIX,
   TraceFileMissingError,
   TraceNotFoundError,
+  applyToolSpanOtlpAttributes,
   asNumber,
   asString,
   buildTraceAnalystTools,
@@ -91,7 +97,7 @@ import {
   stringField,
   traceAnalystFunctionGroup,
   traceSpanKindToOpenInferenceKind
-} from "./chunk-LNQEP766.js";
+} from "./chunk-S2F4J57L.js";
 import {
   RunIntegrityError,
   assertRunCaptured,
@@ -119,6 +125,7 @@ export {
   FAILURE_CLASSES,
   FileSystemRawProviderSink,
   FileSystemTraceStore,
+  INPUT_VALUE,
   InMemoryRawProviderSink,
   InMemoryTraceStore,
   LLM_CACHED_TOKENS,
@@ -134,6 +141,7 @@ export {
   NoopRawProviderSink,
   OPENINFERENCE_SPAN_KIND,
   OTEL_AGENT_EVAL_SCOPE,
+  OUTPUT_VALUE,
   OtlpFileTraceStore,
   REDACTION_VERSION,
   ReplayCache,
@@ -141,6 +149,8 @@ export {
   RunIntegrityError,
   SPAN_KIND_ATTR_KEYS,
   SpanNotFoundError,
+  TOOL_ARGS_CAPTURED,
+  TOOL_LATENCY_MS,
   TOOL_NAME,
   TOOL_NAME_ATTR_KEYS,
   TRACE_ANALYST_ACTOR_DESCRIPTION,
@@ -153,6 +163,7 @@ export {
   TraceNotFoundError,
   aggregateLlm,
   analyzeTraces,
+  applyToolSpanOtlpAttributes,
   argHash,
   asNumber,
   asString,
@@ -178,6 +189,7 @@ export {
   firstStringAttr,
   flattenOtlpExportToNdjson,
   groupBy,
+  hasCapturedToolArgs,
   inferDomainKeywords,
   inferOtlpKind,
   isJudgeSpan,

package/dist/{types-C5gJrOVT.d.ts → types-BSw1rOUB.d.ts}

@@ -1,4 +1,7 @@
-import { d as RunTokenUsage } from './run-record-B7RTi_ix.js';
+import { P as PolicyEditCandidateRecord } from './policy-edit-wG9uFEFm.js';
+import { R as RunPaidCallInput, a as CostChannel, P as PaidCallResult, C as CostLedger, b as CostLedgerSummary } from './cost-ledger-DWy3XdJc.js';
+import { L as LlmCallMetadata } from './llm-client-qoDd18Qz.js';
+import { c as RunTokenUsage } from './run-record-BDH49H2E.js';
 
 /**
  * Pass A substrate types — `runCampaign` is the one primitive every
@@ -82,12 +85,20 @@ interface JudgeDimension {
 interface JudgeConfig<TArtifact, TScenario extends Scenario = Scenario> {
     name: string;
     dimensions: JudgeDimension[];
+    /** Stable scoring revision used by campaign resume and verdict caches.
+     * Built-in judges derive this from their prompt, model, and rubric. Custom
+     * judges should set it when closure state can change without changing code. */
+    judgeVersion?: string;
     /** Score one artifact. Throw on failure — a thrown judge is recorded as a
      *  failed cell, never silently folded into a zero. */
     score(input: {
         artifact: TArtifact;
         scenario: TScenario;
         signal: AbortSignal;
+        /** Shared run spend account and receipt attribution phase. */
+        costLedger?: CostLedger;
+        costPhase?: string;
+        costTags?: Record<string, string>;
     }): JudgeScore | Promise<JudgeScore>;
     appliesTo?: (scenario: TScenario) => boolean;
 }
@@ -104,6 +115,8 @@ interface JudgeScore {
     dimensions: Record<string, number>;
     composite: number;
     notes: string;
+    /** Provider metadata for display and diagnostics; accounting uses CostLedger receipts. */
+    llmCall?: LlmCallMetadata;
     /** Set when the judge itself failed (call error, unparseable output).
      *  `composite`/`dimensions` carry no signal — aggregators MUST exclude
      *  failed scores from means instead of folding them into zeros. */
@@ -168,6 +181,9 @@ interface ProposedCandidate {
      *  primitive it used. Survives to `GenerationCandidate.rationale` and the
      *  emitted provenance record. */
     rationale: string;
+    /** Structured, JSON-safe cause for this exact candidate when the proposer
+     *  can provide one. Policy edits retain the full validated edit here. */
+    candidateRecord?: PolicyEditCandidateRecord;
 }
 /** Type guard: a proposal carrying its rationale vs a bare
  *  surface. The loop branches on this to populate `GenerationCandidate`. */
@@ -194,6 +210,28 @@ interface ParetoParent {
     label?: string;
     rationale?: string;
 }
+/** Exact measured state for the surface an optimizer is learning from.
+ *  Unlike a model-authored expected gain, every value here comes from a
+ *  completed campaign over the designed denominator. */
+interface ScoredSurfaceOutcome {
+    /** Optimization/search evidence only. Held-out results must never flow back
+     *  into a proposer through this type. */
+    split: 'search';
+    /** Generation that actually measured this surface (`-1` for the baseline). */
+    generation: number;
+    surfaceHash: string;
+    composite: number;
+    dimensions: Record<string, number>;
+    scenarios: Array<{
+        scenarioId: string;
+        composite: number;
+        notes?: string;
+    }>;
+    coverage: {
+        expectedCells: number;
+        scorableCells: number;
+    };
+}
 /** Stateless surface mutation — given findings + current
  *  surface, return N candidate surfaces. Pure transform, no generation
  *  awareness. Reflective-mutation and `AxGEPA` mutators conform. Wrapped by
@@ -221,6 +259,12 @@ interface ProposeContext<TFindings = unknown> {
     populationSize: number;
     generation: number;
     signal: AbortSignal;
+    /** Measured baseline for this optimization run. `runOptimization` always
+     *  supplies it; optional for standalone proposer callers. */
+    baselineOutcome?: ScoredSurfaceOutcome;
+    /** Measured result for `currentSurface`, the complete global incumbent every
+     *  new candidate mutates. `runOptimization` always supplies it. */
+    incumbentOutcome?: ScoredSurfaceOutcome;
     /** Optional analysis report produced before proposal. Opaque to the substrate:
      *  the proposer that consumes it owns the shape. */
     report?: unknown;
@@ -238,6 +282,9 @@ interface ProposeContext<TFindings = unknown> {
      *  scenarios) into a merged candidate. Proposers doing pure single-parent
      *  reflection may ignore it. See {@link ParetoParent}. */
     paretoParents?: ParetoParent[];
+    /** Shared run spend account and receipt attribution phase. */
+    costLedger?: CostLedger;
+    costPhase?: string;
     /** FIREWALL (non-negotiable): the held-out judge is write-only — its verdicts
      *  score the chosen output and gate promotion, and are NEVER an input to
      *  proposal/steering (else the optimizer games the acceptance axis = an
@@ -319,6 +366,9 @@ interface GateContext<TArtifact, TScenario extends Scenario> {
         candidate: number;
         baseline: number;
     };
+    /** Shared run spend account and receipt attribution phase. */
+    costLedger?: CostLedger;
+    costPhase?: string;
     signal: AbortSignal;
 }
 interface GateResult {
@@ -357,35 +407,15 @@ interface CampaignArtifactWriter {
  *  backend-integrity guard with ONE source of truth — a field added to
  *  `RunTokenUsage` is a compile error here, not a silent drift. */
 type CampaignTokenUsage = RunTokenUsage;
-/** Cell-scoped cost meter. NOTHING is captured automatically —
- *  the substrate does not intercept the LLM call, so it cannot see cost or
- *  tokens unless the dispatch reports them. Every LLM cost MUST be reported via
- *  `observe` and every token count via `observeTokens`; a dispatch that reports
- *  neither yields a `{cost:0, tokens:0}` cell, which the backend-integrity
- *  guard (`assertRealBackend`) correctly reads as a stub. Also use `observe`
- *  for non-LLM spend (sandbox time, tool costs). */
+/** Cell-scoped paid-call entry point. The dispatch places every paid operation
+ *  inside `runPaidCall`; the returned provider result supplies one receipt with
+ *  cost, tokens, and resolved model. Calls made outside this method are not
+ *  admitted or captured. */
 interface CampaignCostMeter {
-    observe(amountUsd: number, source: string): void;
-    /** Record LLM token usage for this cell; accumulates across calls. A cell
-     *  has `costUsd` but no token counts unless the dispatch reports them here —
-     *  and the backend-integrity guard (`assertRealBackend`) keys on
-     *  `tokenUsage`, so a cell that never reports tokens reads as a stub. Any
-     *  dispatch that calls an LLM MUST report its usage. */
-    observeTokens(usage: CampaignTokenUsage): void;
-    /** Record the concrete model the backend RESOLVED this cell to at runtime.
-     *  The substrate cannot see the LLM call, so it cannot know which model a
-     *  vendor-locked harness actually served — only the dispatch, reading the
-     *  backend's usage/terminal events, can. A dispatch whose profile declares a
-     *  runtime-resolved model (the `HARNESS_NATIVE_MODEL` sentinel) MUST report
-     *  the resolved, snapshot-bearing id here so the RunRecord pins a real model
-     *  instead of the sentinel. Last write wins (a cell issues one logical run);
-     *  optional because most dispatches declare a concrete model up front. */
-    observeModel?(model: string): void;
-    current(): number;
-    /** Accumulated token usage for this cell (zeros if never observed). */
-    tokens(): CampaignTokenUsage;
-    /** The runtime-resolved model reported via `observeModel`, if any. */
-    resolvedModel?(): string | undefined;
+    /** The only paid-call path. Returns a typed result; callers must inspect it. */
+    runPaidCall<T>(input: Omit<RunPaidCallInput<T>, 'channel' | 'phase' | 'tags'> & {
+        channel?: CostChannel;
+    }): Promise<PaidCallResult<T>>;
 }
 /** Source tag — required on every store write. Used by the
  *  default training-source filter (production-trace samples NOT used as
@@ -478,15 +508,16 @@ interface CampaignCellResult<TArtifact> {
     artifact: TArtifact;
     judgeScores: Record<string, JudgeScore>;
     costUsd: number;
-    /** LLM token usage the dispatch reported via `ctx.cost.observeTokens`.
-     *  `{ input: 0, output: 0 }` when the dispatch reported none — which the
-     *  backend-integrity guard reads as a stub. */
+    /** True when at least one priced receipt used the model table instead of a provider bill. */
+    costEstimated?: boolean;
+    /** Exact durable receipts required to reuse this cached result. */
+    costCallIds?: string[];
+    /** Agent-call token usage committed by `ctx.cost.runPaidCall`.
+     *  `{ input: 0, output: 0 }` when no paid agent call was recorded. */
     tokenUsage: CampaignTokenUsage;
-    /** The concrete model the backend resolved this cell to at runtime, reported
-     *  by the dispatch via `ctx.cost.observeModel`. Set only when the dispatch
-     *  reported it — a profile that declares a concrete model up front has no
-     *  need to. Consumed by `buildRunRecord` to pin the real model when the
-     *  declared model is the `HARNESS_NATIVE_MODEL` sentinel. */
+    /** Concrete model from the latest committed agent receipt. Consumed by
+     *  `buildRunRecord` to pin the model when the declared profile uses a
+     *  runtime-resolved sentinel. */
     resolvedModel?: string;
     durationMs: number;
     seed: number;
@@ -517,6 +548,29 @@ interface GenerationCandidate {
     surfaceHash: string;
     composite: number;
     ci95: [number, number];
+    /** Exact surface this candidate mutated. */
+    parentSurfaceHash?: string;
+    /** Measured search-split composite of the exact parent surface. */
+    parentComposite?: number;
+    /** Candidate composite minus its parent's composite. Present only when the
+     *  candidate completed the designed denominator. */
+    observedDeltaFromParent?: number;
+    /** Whether this candidate had a scorable result for every designed campaign
+     *  cell and was therefore eligible for ranking, promotion, and Pareto
+     *  selection. Older externally-authored records may omit this field; loop
+     *  records always populate it. */
+    eligibleForPromotion?: boolean;
+    /** Exact denominator receipt for selection eligibility. Scores stay
+     *  descriptive: an incomplete candidate is retained with its observed score
+     *  and errors instead of receiving an invented penalty. */
+    coverage?: {
+        expectedCells: number;
+        scorableCells: number;
+        unscorableCells: Array<{
+            cellId: string;
+            reason: string;
+        }>;
+    };
     /** Mean score per judge dimension across all cells (scenarios × reps ×
      *  judges that reported the dimension). */
     dimensions: Record<string, number>;
@@ -538,10 +592,15 @@ interface GenerationCandidate {
      *  "because rationale Z" the audit requires to survive to the result.
      *  Present when the proposer returned a `ProposedCandidate`. */
     rationale?: string;
+    /** Exact structured cause threaded from the proposer, when available. */
+    candidateRecord?: PolicyEditCandidateRecord;
 }
 interface CampaignAggregates {
     byJudge: Record<string, JudgeAggregate>;
     byScenario: Record<string, ScenarioAggregate>;
+    /** Canonical campaign accounting, including worker and judge calls. */
+    cost: CostLedgerSummary;
+    /** Compatibility alias of `cost.totalCostUsd`. */
     totalCostUsd: number;
     cellsExecuted: number;
     cellsSkipped: number;
@@ -572,4 +631,4 @@ interface CampaignResult<TArtifact = unknown, TScenario extends Scenario = Scena
     scenarios: Array<Pick<TScenario, 'id' | 'kind'>>;
 }
 
-export { type JudgeAggregate as A, type ScenarioAggregate as B, type CampaignResult as C, type DispatchContext as D, isProposedCandidate as E, labelTrustRank as F, type GateResult as G, type JudgeScore as J, type LabeledScenarioStore as L, type MutableSurface as M, type OptimizationProposer as O, type ParetoParent as P, type RedactionStatus as R, type Scenario as S, type TraceSpan as T, type JudgeDimension as a, type JudgeConfig as b, type DispatchFn as c, type CampaignTraceWriter as d, type GenerationRecord as e, type SurfaceProposer as f, type Gate as g, type GateDecision as h, type CampaignAggregates as i, type CampaignArtifactWriter as j, type CampaignCellResult as k, type CampaignCostMeter as l, type CodeSurface as m, type GateContext as n, type GenerationCandidate as o, type Mutator as p, type OptimizerConfig as q, type SessionScript as r, type LabeledScenarioWrite as s, type LabeledScenarioSampleArgs as t, type LabeledScenarioRecord as u, type LabelTrust as v, type ProposedCandidate as w, type ProposeContext as x, type LabeledScenarioSource as y, type CampaignTokenUsage as z };
+export { type JudgeAggregate as A, type ScenarioAggregate as B, type CampaignResult as C, type DispatchContext as D, type ScoredSurfaceOutcome as E, isProposedCandidate as F, type Gate as G, labelTrustRank as H, type JudgeScore as J, type LabeledScenarioStore as L, type MutableSurface as M, type OptimizationProposer as O, type ParetoParent as P, type RedactionStatus as R, type Scenario as S, type TraceSpan as T, type JudgeDimension as a, type JudgeConfig as b, type SurfaceProposer as c, type GateDecision as d, type CampaignAggregates as e, type CampaignArtifactWriter as f, type CampaignCellResult as g, type CampaignCostMeter as h, type CampaignTraceWriter as i, type CodeSurface as j, type DispatchFn as k, type GateContext as l, type GateResult as m, type GenerationCandidate as n, type GenerationRecord as o, type Mutator as p, type OptimizerConfig as q, type SessionScript as r, type LabeledScenarioWrite as s, type LabeledScenarioSampleArgs as t, type LabeledScenarioRecord as u, type LabelTrust as v, type ProposedCandidate as w, type ProposeContext as x, type LabeledScenarioSource as y, type CampaignTokenUsage as z };

package/dist/{types-C7DGg5ex.d.ts → types-BkfcQnxV.d.ts}

@@ -1,3 +1,4 @@
+import { C as CostLedger, b as CostLedgerSummary } from './cost-ledger-DWy3XdJc.js';
 import { TCloud } from '@tangle-network/tcloud';
 
 interface Scenario {
@@ -50,6 +51,8 @@ interface ScenarioResult {
     overallScore: number;
     totalDurationMs: number;
     artifacts: CollectedArtifacts;
+    /** Agent and judge spend attributed to this scenario. */
+    cost?: CostLedgerSummary;
 }
 interface TurnResult {
     turnIndex: number;
@@ -96,6 +99,7 @@ interface BenchmarkReport {
     promptVersion: string;
     scenarioCount: number;
     results: ScenarioResult[];
+    cost?: CostLedgerSummary;
     summary: {
         overallAvg: number;
         byPersona: Record<string, {
@@ -266,11 +270,22 @@ interface BenchmarkRunnerConfig {
     passThreshold?: number;
     generation?: number;
     promptVersion?: string;
+    /** Shared ledger for agent and judge calls made by the benchmark. */
+    costLedger?: CostLedger;
+    /** Exact maximum provider attempts configured on the supplied TCloud client. */
+    tcloudMaximumAttempts?: number;
 }
 interface JudgeInput {
     scenario: Scenario;
     turns: TurnResult[];
     artifacts: CollectedArtifacts;
+    /** Shared ledger for paid built-in judges. Direct calls default to an uncapped ledger. */
+    costLedger?: CostLedger;
+    costPhase?: string;
+    costTags?: Record<string, string>;
+    signal?: AbortSignal;
+    /** Exact maximum provider attempts configured on the supplied TCloud client. */
+    tcloudMaximumAttempts?: number;
 }
 type JudgeFn = (tc: TCloud, input: JudgeInput) => Promise<JudgeScore[]>;
 

package/dist/wire/index.d.ts

@@ -1,14 +1,17 @@
-import { F as FeedbackTrajectoryStore } from '../feedback-trajectory-pDcz1lQ1.js';
-import { T as TraceStore } from '../store-BsVi7ncX.js';
+import { C as CostLedger } from '../cost-ledger-DWy3XdJc.js';
+import { F as FeedbackTrajectoryStore } from '../feedback-trajectory-BUnM58xL.js';
+import { a as LlmClientOptions } from '../llm-client-qoDd18Qz.js';
+import { T as TraceStore } from '../store-DGqD0Pyo.js';
 import { z } from 'zod';
 import { OpenAPIObject } from 'openapi3-ts/oas31';
 import * as hono_types from 'hono/types';
 import { ServerType } from '@hono/node-server';
 import { Hono } from 'hono';
-import '../emitter-BRchAAAx.js';
-import '../schema-SGWcK9wa.js';
-import '../dataset-NENEzRgk.js';
 import '../errors-oeQrLqXC.js';
+import '../emitter-CjD7vUwv.js';
+import '../schema-B3Q3l9Z_.js';
+import '../dataset-NENEzRgk.js';
+import '../raw-provider-sink-C46HDghv.js';
 
 declare const RubricDimensionSchema: z.ZodObject<{
     id: z.ZodString;
@@ -122,9 +125,9 @@ declare const TraceEventSchema: z.ZodObject<{
     runId: z.ZodString;
     spanId: z.ZodOptional<z.ZodString>;
     kind: z.ZodEnum<{
-        policy_violation: "policy_violation";
-        custom: "custom";
         error: "error";
+        custom: "custom";
+        policy_violation: "policy_violation";
         log: "log";
         budget_decrement: "budget_decrement";
         budget_breach: "budget_breach";
@@ -140,9 +143,9 @@ declare const TracesIngestRequestSchema: z.ZodObject<{
         runId: z.ZodString;
         spanId: z.ZodOptional<z.ZodString>;
         kind: z.ZodEnum<{
-            policy_violation: "policy_violation";
-            custom: "custom";
             error: "error";
+            custom: "custom";
+            policy_violation: "policy_violation";
             log: "log";
             budget_decrement: "budget_decrement";
             budget_breach: "budget_breach";
@@ -165,8 +168,8 @@ declare const FeedbackLabelSchema: z.ZodObject<{
     id: z.ZodOptional<z.ZodString>;
     source: z.ZodEnum<{
         judge: "judge";
-        system: "system";
         user: "user";
+        system: "system";
         policy: "policy";
         environment: "environment";
         metric: "metric";
@@ -198,10 +201,10 @@ declare const FeedbackAttemptSchema: z.ZodObject<{
     id: z.ZodString;
     stepIndex: z.ZodNumber;
     artifactType: z.ZodEnum<{
-        action: "action";
-        decision: "decision";
         text: "text";
         code: "code";
+        action: "action";
+        decision: "decision";
         plan: "plan";
         research: "research";
         ui: "ui";
@@ -226,8 +229,8 @@ declare const FeedbackAttemptSchema: z.ZodObject<{
         id: z.ZodOptional<z.ZodString>;
         source: z.ZodEnum<{
             judge: "judge";
-            system: "system";
             user: "user";
+            system: "system";
             policy: "policy";
             environment: "environment";
             metric: "metric";
@@ -270,10 +273,10 @@ declare const FeedbackTrajectorySchema: z.ZodObject<{
         id: z.ZodString;
         stepIndex: z.ZodNumber;
         artifactType: z.ZodEnum<{
-            action: "action";
-            decision: "decision";
             text: "text";
             code: "code";
+            action: "action";
+            decision: "decision";
             plan: "plan";
             research: "research";
             ui: "ui";
@@ -298,8 +301,8 @@ declare const FeedbackTrajectorySchema: z.ZodObject<{
             id: z.ZodOptional<z.ZodString>;
             source: z.ZodEnum<{
                 judge: "judge";
-                system: "system";
                 user: "user";
+                system: "system";
                 policy: "policy";
                 environment: "environment";
                 metric: "metric";
@@ -334,8 +337,8 @@ declare const FeedbackTrajectorySchema: z.ZodObject<{
         id: z.ZodOptional<z.ZodString>;
         source: z.ZodEnum<{
             judge: "judge";
-            system: "system";
             user: "user";
+            system: "system";
             policy: "policy";
             environment: "environment";
             metric: "metric";
@@ -438,7 +441,13 @@ declare class WireError extends Error {
     readonly details?: unknown | undefined;
     constructor(code: string, message: string, status?: number, details?: unknown | undefined);
 }
-declare function handleJudge(req: JudgeRequest): Promise<JudgeResult>;
+interface HandleJudgeOptions {
+    costLedger?: CostLedger;
+    costPhase?: string;
+    llm?: LlmClientOptions;
+    signal?: AbortSignal;
+}
+declare function handleJudge(req: JudgeRequest, options?: HandleJudgeOptions): Promise<JudgeResult>;
 declare function handleListRubrics(): ListRubricsResponse;
 declare function handleVersion(): VersionResponse;
 /**
@@ -569,4 +578,4 @@ interface StartedServer {
  */
 declare function startServerAsync(opts?: ServeOptions): Promise<StartedServer>;
 
-export { BUILTIN_RUBRICS, type ErrorResponse, ErrorResponseSchema, type FailureMode, FailureModeSchema, FeedbackAttemptSchema, type FeedbackIngestResponse, FeedbackIngestResponseSchema, FeedbackLabelSchema, type FeedbackTrajectory, FeedbackTrajectorySchema, HealthResponseSchema, type IngestionStores, type JudgeRequest, JudgeRequestSchema, type JudgeResult, JudgeResultSchema, type ListRubricsResponse, ListRubricsResponseSchema, type Rubric, type RubricDimension, RubricDimensionSchema, type RubricInfo, RubricInfoSchema, RubricSchema, type ServeOptions, type StartedServer, type TraceEvent, TraceEventSchema, type TracesIngestRequest, TracesIngestRequestSchema, type TracesIngestResponse, TracesIngestResponseSchema, type VersionResponse, VersionResponseSchema, WIRE_VERSION, WireError, buildOpenApi, createApp, dispatchRpc, getBuiltinRubric, handleFeedbackIngest, handleJudge, handleListRubrics, handleTracesIngest, handleVersion, hashRubric, listBuiltinRubrics, runRpcBatch, runRpcOnce, startServer, startServerAsync };
+export { BUILTIN_RUBRICS, type ErrorResponse, ErrorResponseSchema, type FailureMode, FailureModeSchema, FeedbackAttemptSchema, type FeedbackIngestResponse, FeedbackIngestResponseSchema, FeedbackLabelSchema, type FeedbackTrajectory, FeedbackTrajectorySchema, type HandleJudgeOptions, HealthResponseSchema, type IngestionStores, type JudgeRequest, JudgeRequestSchema, type JudgeResult, JudgeResultSchema, type ListRubricsResponse, ListRubricsResponseSchema, type Rubric, type RubricDimension, RubricDimensionSchema, type RubricInfo, RubricInfoSchema, RubricSchema, type ServeOptions, type StartedServer, type TraceEvent, TraceEventSchema, type TracesIngestRequest, TracesIngestRequestSchema, type TracesIngestResponse, TracesIngestResponseSchema, type VersionResponse, VersionResponseSchema, WIRE_VERSION, WireError, buildOpenApi, createApp, dispatchRpc, getBuiltinRubric, handleFeedbackIngest, handleJudge, handleListRubrics, handleTracesIngest, handleVersion, hashRubric, listBuiltinRubrics, runRpcBatch, runRpcOnce, startServer, startServerAsync };

package/dist/wire/index.js

@@ -34,8 +34,10 @@ import {
   runRpcOnce,
   startServer,
   startServerAsync
-} from "../chunk-4D5RVB3W.js";
-import "../chunk-GY4SYVPJ.js";
+} from "../chunk-LTVG32KX.js";
+import "../chunk-NJC7U437.js";
+import "../chunk-VCTY3W6J.js";
+import "../chunk-VI2UW6B6.js";
 import "../chunk-PC4UYEBM.js";
 import "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";

package/docs/design/loop-taxonomy.md

@@ -201,8 +201,7 @@ consume the **same dataset** the flywheel builds.
   agent-runtime.
 - **runCampaign** — a measurement: a surface scored over N scenarios × M reps.
   agent-eval. (A "campaign" = a coordinated batch of measurements.)
-- **runOptimization** — the improvement loop body: proposer suggests surfaces,
-  each measured by a campaign, top-K promoted per generation. agent-eval.
+- **runOptimization** — the improvement loop body: proposer suggests surfaces, each is measured, and only a candidate that beats the global incumbent is promoted. agent-eval.
 - **runImprovementLoop** — `runOptimization` + holdout re-score + release gate
   + optional PR. agent-eval.
 - **runAnalystLoop** — reflective autoresearch: findings + knowledge updates +

package/docs/distributed-driver.md

@@ -135,7 +135,7 @@ round-robin, region-affinity from a previous run, scheduling table).
 | **Auth** | Bearer token on `Authorization`; pluggable via `auth: string \| () => string \| Promise<string>` for rotation/refresh. |
 | **Payload size** | Server enforces `maxBodyBytes` (default 10 MB). |
 | **Traces** | Both ends emit OTel — if both point at the same OTLP collector, you get a unified trace per cell. See `docs/adapters-observability.md`. |
-| **Cost** | Worker's `ctx.cost.observe(usd, source)` is local to the worker process. Roll up server-side and attach to your worker-side telemetry; we don't (yet) forward cost back to the coordinator. Tracked as follow-up. |
+| **Cost** | Worker's `ctx.cost.runPaidCall(...)` writes durable receipts in the worker process. Roll up those receipts server-side and attach them to worker telemetry; they are not forwarded to the coordinator automatically. |
 
 ## Running the reference example
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.115.3",
+  "version": "0.117.0",
   "description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {
@@ -148,7 +148,7 @@
     "@hono/node-server": "^2.0.0",
     "@tangle-network/agent-interface": "^0.22.0",
     "@tangle-network/tcloud": "^0.4.14",
-    "hono": "^4.12.16",
+    "hono": "^4.12.30",
     "zod": "^4.3.6"
   },
   "devDependencies": {
@@ -165,7 +165,7 @@
     "minimumReleaseAge": 4320,
     "overrides": {
       "postcss@<8.5.10": "^8.5.10",
-      "ws@>=8.0.0 <8.20.1": "^8.20.1"
+      "ws@>=8.0.0 <8.21.0": "^8.21.0"
     }
   },
   "engines": {