@tangle-network/agent-eval 0.58.2 → 0.59.1

package/dist/index.d.ts

@@ -12,12 +12,12 @@ import { L as LlmClientOptions } from './llm-client-BXVRUZyX.js';
 export { d as LlmCallError, b as LlmCallRequest, c as LlmCallResult, e as LlmClient, f as LlmMessage, g as LlmRouteAssertionError, a as LlmRouteRequirements, h as LlmUsage, i as assertLlmRoute, j as backoffMs, k as callLlm, l as callLlmJson, m as isTransientLlmError, p as probeLlm, s as stripFencedJson } from './llm-client-BXVRUZyX.js';
 import { AnalyzeTracesOptions, OtelExporter, OtelExportConfig, AnalyzeTracesInput, AnalyzeTracesResult } from './traces.js';
 export { AnalyzeTracesTurnSnapshot, CaptureFetchContext, CaptureFetchOptions, DEFAULT_REDACTION_RULES, ExportableSpan, FlattenOtlpOptions, OTEL_AGENT_EVAL_SCOPE, OtlpExport, OtlpFileTraceStore, OtlpFileTraceStoreOptions, OtlpFlatLine, OtlpResourceSpans, OtlpSpan, REDACTION_VERSION, RedactionReport, RedactionRule, ReplayCache, ReplayCacheEntry, ReplayCacheMissError, ReplayCacheStats, ReplayFetchOptions, SpanNotFoundError, TRACE_ANALYST_ACTOR_DESCRIPTION, TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, TRACE_ANALYST_SUBAGENT_DESCRIPTION, TraceAnalystHookOptions, TraceFileMissingError, TraceInsightContext, TraceInsightFinding, TraceInsightPanelRole, TraceInsightPromptInput, TraceInsightQualityGate, TraceInsightQuestion, TraceInsightReadiness, TraceInsightSuite, TraceInsightTask, TraceNotFoundError, analyzeTraces, buildTraceAnalystTools, buildTraceInsightContext, buildTraceInsightPrompt, captureFetchToRawSink, createOtelExporter, createOtelTracingStore, createReplayFetch, defaultTraceInsightPanel, describeTraceInsightScope, domainEvidencePattern, exportRunAsOtlp, flattenOtlpExportToNdjson, inferDomainKeywords, iterateRawCalls, otelRunCompleteHook, planTraceInsightQuestions, redactString, redactValue, scoreTraceInsightReadiness, tokenizeDomainWords, traceAnalystFunctionGroup, traceAnalystOnRunComplete } from './traces.js';
-import { T as TraceAnalysisStore } from './store-CJbzDxZ2.js';
-export { D as DEFAULT_TRACE_ANALYST_BUDGETS, a as DatasetOverview, Q as QueryTracesPage, S as SearchSpanResult, b as SearchTraceResult, c as SpanMatchRecord, d as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, e as TraceAnalystByteBudgets, f as TraceAnalystFilters, g as TraceAnalystSpan, h as TraceAnalystSpanKind, i as TraceAnalystSpanStatus, j as TraceAnalystTraceSummary, V as ViewSpansResult, k as ViewTraceOversized, l as ViewTraceResult } from './store-CJbzDxZ2.js';
+import { T as TraceAnalysisStore } from './store-jzKpMl16.js';
+export { D as DEFAULT_TRACE_ANALYST_BUDGETS, a as DatasetOverview, Q as QueryTracesPage, S as SearchSpanResult, b as SearchTraceResult, c as SpanMatchRecord, d as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, e as TraceAnalystByteBudgets, f as TraceAnalystFilters, g as TraceAnalystSpan, h as TraceAnalystSpanKind, i as TraceAnalystSpanStatus, j as TraceAnalystTraceSummary, V as ViewSpansResult, k as ViewTraceOversized, l as ViewTraceResult } from './store-jzKpMl16.js';
 import { b as JudgeFn, J as JudgeInput, B as BenchmarkRunnerConfig, S as Scenario, c as BenchmarkReport, P as ProductClientConfig, C as CheckResult, T as TestResult, d as PersonaConfig, D as DriverResult, e as DriverState, f as CollectedArtifacts, g as ScenarioResult, h as TurnMetrics, i as ScenarioFile, j as CompletionCriterion } from './types-DhqpAi_z.js';
 export { A as ArtifactCheck, k as ArtifactResult, E as EvalResult, F as FeedbackPattern, l as JudgeConfig, m as JudgeRubric, a as JudgeScore, n as PersonaRigor, R as RouteMap, o as RubricDimension, p as Turn, q as TurnResult } from './types-DhqpAi_z.js';
-import { a as Analyst, b as AnalystSeverity, c as AnalystFinding, d as AnalystCost, e as AnalystContext } from './registry-BSWy0rvH.js';
-export { f as AnalystHooks, g as AnalystInputKind, A as AnalystRegistry, h as AnalystRegistryOptions, i as AnalystRequirements, j as AnalystRunEvent, k as AnalystRunInputs, l as AnalystRunResult, m as AnalystRunSummary, B as BudgetPolicy, C as ChatCallOpts, n as ChatClient, o as ChatRequest, p as ChatResponse, q as ChatTransport, r as CliBridgeTransportOpts, s as CreateChatClientOpts, D as DirectProviderTransportOpts, E as EvidenceRef, M as MockTransportOpts, R as RegistryRunOpts, t as RouterTransportOpts, S as SandboxSdkTransportOpts, u as computeFindingId, v as createChatClient, w as makeFinding } from './registry-BSWy0rvH.js';
+import { a as Analyst, b as AnalystSeverity, c as AnalystFinding, d as AnalystCost, e as AnalystContext } from './registry-DK9kqXvb.js';
+export { f as AnalystHooks, g as AnalystInputKind, A as AnalystRegistry, h as AnalystRegistryOptions, i as AnalystRequirements, j as AnalystRunEvent, k as AnalystRunInputs, l as AnalystRunResult, m as AnalystRunSummary, B as BudgetPolicy, C as ChatCallOpts, n as ChatClient, o as ChatRequest, p as ChatResponse, q as ChatTransport, r as CliBridgeTransportOpts, s as CreateChatClientOpts, D as DirectProviderTransportOpts, E as EvidenceRef, M as MockTransportOpts, R as RegistryRunOpts, t as RouterTransportOpts, S as SandboxSdkTransportOpts, u as computeFindingId, v as createChatClient, w as makeFinding } from './registry-DK9kqXvb.js';
 import { TCloud } from '@tangle-network/tcloud';
 import { z } from 'zod';
 export { c as ControlActionFailureMode, d as ControlActionOutcome, e as ControlBudget, f as ControlContext, g as ControlDecision, C as ControlEvalResult, a as ControlRunResult, h as ControlRuntimeConfig, i as ControlRuntimeError, j as ControlSeverity, b as ControlStep, k as ControlStopPolicies, S as StopDecision, l as allCriticalPassed, o as objectiveEval, r as runAgentControlLoop, s as stopOnNoProgress, m as stopOnRepeatedAction, n as subjectiveEval } from './control-runtime-DuFBYg7A.js';

package/dist/index.js

@@ -189,7 +189,7 @@ import {
   tokenizeDomainWords,
   traceAnalystFunctionGroup,
   traceAnalystOnRunComplete
-} from "./chunk-5GLYP2IQ.js";
+} from "./chunk-PIEAE33T.js";
 import {
   DEFAULT_REDACTION_RULES,
   REDACTION_VERSION,

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.58.2",
+    "version": "0.59.1",
     "description": "HTTP and stdio RPC interface to agent-eval. The TypeScript runtime is the source of truth; this spec is the contract that cross-language clients (Python, Rust, Go) generate from.\n\nWire-protocol version: 1.0.0. Bumps on breaking changes to request/response schemas.",
     "contact": {
       "name": "Tangle Network",

package/dist/{registry-BSWy0rvH.d.ts → registry-DK9kqXvb.d.ts}

@@ -1,6 +1,6 @@
 import { b as LlmCallRequest, c as LlmCallResult } from './llm-client-BXVRUZyX.js';
 import { R as RunRecord } from './run-record-etiCMsUq.js';
-import { T as TraceAnalysisStore } from './store-CJbzDxZ2.js';
+import { T as TraceAnalysisStore } from './store-jzKpMl16.js';
 import { J as JudgeInput } from './types-DhqpAi_z.js';
 
 /**

package/dist/{store-CJbzDxZ2.d.ts → store-jzKpMl16.d.ts}

@@ -69,6 +69,29 @@ interface TraceAnalystFilters {
      *  expensive on large datasets. Use the indexed filters above first. */
     regex_pattern?: string;
 }
+/** One distinct error signature across the dataset — the deterministic unit of
+ *  failure coverage. Signatures normalize volatile tokens (digits, hex/uuids,
+ *  paths, durations) out of the span `status_message` so semantically identical
+ *  failures collapse into one cluster. An analyst that accounts for every
+ *  cluster has, by construction, covered every distinct failure mode. */
+interface ErrorCluster {
+    /** Normalized status_message — the cluster key. */
+    signature: string;
+    /** A verbatim, un-normalized exemplar message (for exact-string citation). */
+    status_message_sample: string;
+    /** The span name that most often carries this signature, if any. */
+    span_name: string | null;
+    /** The tool that most often carries this signature, if any. */
+    tool_name: string | null;
+    trace_count: number;
+    span_count: number;
+    /** trace_count / total error traces in the matched set (0..1). */
+    prevalence: number;
+    /** Real trace ids carrying this signature (capped), passable to view/search. */
+    exemplar_trace_ids: string[];
+    /** Real span ids carrying this signature (capped). */
+    exemplar_span_ids: string[];
+}
 interface DatasetOverview {
     total_traces: number;
     raw_jsonl_bytes: number;
@@ -82,6 +105,11 @@ interface DatasetOverview {
         trace_count: number;
         span_count: number;
     };
+    /** The COMPLETE deterministic error-signature population, sorted by
+     *  trace_count desc. This is the failure-coverage checklist: an analysis is
+     *  complete only when every cluster here is accounted for. Empty when the
+     *  matched set has no error spans. */
+    error_clusters: ErrorCluster[];
     time_range: {
         earliest: string;
         latest: string;

package/dist/traces.d.ts

@@ -9,8 +9,8 @@ export { E as EventFilter, F as FileSystemTraceStore, a as FileSystemTraceStoreO
 export { a as aggregateLlm, b as argHash, g as groupBy, j as judgeSpans, l as llmSpans, r as runFailureClass, c as runsForScenario, t as toolSpans } from './query-CqTxMwDw.js';
 export { A as Artifact, B as BudgetLedgerEntry, h as BudgetSpec, E as EventKind, i as FAILURE_CLASSES, F as FailureClass, G as GenericSpan, J as JudgeSpan, L as LlmSpan, M as Message, d as RetrievalSpan, R as Run, g as RunLayer, b as RunOutcome, f as RunStatus, e as SandboxSpan, S as Span, j as SpanBase, c as SpanKind, k as SpanStatus, l as TRACE_SCHEMA_VERSION, T as ToolSpan, a as TraceEvent, m as isJudgeSpan, n as isLlmSpan, o as isRetrievalSpan, p as isSandboxSpan, q as isToolSpan } from './schema-m0gsnbt3.js';
 import { AxAIService, AxFunction } from '@ax-llm/ax';
-import { T as TraceAnalysisStore, f as TraceAnalystFilters, a as DatasetOverview, Q as QueryTracesPage, l as ViewTraceResult, V as ViewSpansResult, b as SearchTraceResult, S as SearchSpanResult } from './store-CJbzDxZ2.js';
-export { D as DEFAULT_TRACE_ANALYST_BUDGETS, c as SpanMatchRecord, d as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, e as TraceAnalystByteBudgets, g as TraceAnalystSpan, h as TraceAnalystSpanKind, i as TraceAnalystSpanStatus, j as TraceAnalystTraceSummary, k as ViewTraceOversized } from './store-CJbzDxZ2.js';
+import { T as TraceAnalysisStore, f as TraceAnalystFilters, a as DatasetOverview, Q as QueryTracesPage, l as ViewTraceResult, V as ViewSpansResult, b as SearchTraceResult, S as SearchSpanResult } from './store-jzKpMl16.js';
+export { D as DEFAULT_TRACE_ANALYST_BUDGETS, c as SpanMatchRecord, d as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, e as TraceAnalystByteBudgets, g as TraceAnalystSpan, h as TraceAnalystSpanKind, i as TraceAnalystSpanStatus, j as TraceAnalystTraceSummary, k as ViewTraceOversized } from './store-jzKpMl16.js';
 
 /**
  * `captureFetchToRawSink` — wrap a `fetch` so every request / response / error

package/dist/traces.js

@@ -34,7 +34,7 @@ import {
   tokenizeDomainWords,
   traceAnalystFunctionGroup,
   traceAnalystOnRunComplete
-} from "./chunk-5GLYP2IQ.js";
+} from "./chunk-PIEAE33T.js";
 import {
   DEFAULT_REDACTION_RULES,
   REDACTION_VERSION,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.58.2",
+  "version": "0.59.1",
   "description": "Substrate for self-improving agents: traces, verifiable rewards, preferences, GEPA / reflective mutation, auto-research, replay, sequential anytime-valid stats, and release gates.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {
@@ -153,7 +153,7 @@
     "zod": "^4.3.6"
   },
   "peerDependencies": {
-    "@tangle-network/sandbox": ">=0.2.1 <0.4.0"
+    "@tangle-network/sandbox": ">=0.2.1 <0.5.0"
   },
   "peerDependenciesMeta": {
     "@tangle-network/sandbox": {
@@ -162,7 +162,7 @@
   },
   "devDependencies": {
     "@biomejs/biome": "^2.4.15",
-    "@tangle-network/sandbox": "0.3.0",
+    "@tangle-network/sandbox": "0.4.0",
     "@types/node": "^25.6.0",
     "husky": "^9.1.7",
     "lint-staged": "^17.0.5",