@tangle-network/agent-eval 0.170.0 → 0.171.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +13 -0
- package/dist/adapters/http.d.ts +2 -2
- package/dist/analyst/index.d.ts +5 -5
- package/dist/analyst/index.js +1 -1
- package/dist/{benchmark-command-BA7qOdWw.js → benchmark-command-D8k3Gf0J.js} +2 -2
- package/dist/{benchmark-command-BA7qOdWw.js.map → benchmark-command-D8k3Gf0J.js.map} +1 -1
- package/dist/benchmarks/index.d.ts +4 -4
- package/dist/benchmarks/index.js +2 -2
- package/dist/campaign/index.d.ts +6 -6
- package/dist/campaign/index.js +3 -3
- package/dist/{campaign-BeCbxFqs.js → campaign-B72njjHj.js} +4 -4
- package/dist/{campaign-BeCbxFqs.js.map → campaign-B72njjHj.js.map} +1 -1
- package/dist/cli.js +1 -1
- package/dist/{client-_Fsa5c2_.d.ts → client-CDtcZ3p9.d.ts} +2 -2
- package/dist/{client-_Fsa5c2_.d.ts.map → client-CDtcZ3p9.d.ts.map} +1 -1
- package/dist/contract/index.d.ts +9 -9
- package/dist/contract/index.js +3 -3
- package/dist/{default-registry-ovxrOP0_.d.ts → default-registry-B0s2zU-s.d.ts} +3 -3
- package/dist/{default-registry-ovxrOP0_.d.ts.map → default-registry-B0s2zU-s.d.ts.map} +1 -1
- package/dist/{define-agent-eval-DVJm8Xlh.d.ts → define-agent-eval-Cjy2yhqP.d.ts} +4 -4
- package/dist/{define-agent-eval-DVJm8Xlh.d.ts.map → define-agent-eval-Cjy2yhqP.d.ts.map} +1 -1
- package/dist/{define-agent-eval-Clj-8igZ.js → define-agent-eval-Dy8QgxAI.js} +2 -2
- package/dist/{define-agent-eval-Clj-8igZ.js.map → define-agent-eval-Dy8QgxAI.js.map} +1 -1
- package/dist/{engine-D12Rb6WB.d.ts → engine-CAmTUk52.d.ts} +2 -2
- package/dist/{engine-D12Rb6WB.d.ts.map → engine-CAmTUk52.d.ts.map} +1 -1
- package/dist/experiment/index.d.ts +2 -2
- package/dist/{heldout-gate-Bn7_xWCv.d.ts → heldout-gate-Dh2b62w8.d.ts} +3 -3
- package/dist/{heldout-gate-Bn7_xWCv.d.ts.map → heldout-gate-Dh2b62w8.d.ts.map} +1 -1
- package/dist/hosted/index.d.ts +1 -1
- package/dist/{index-DBbivBNs.d.ts → index-8VIogTyS.d.ts} +6 -6
- package/dist/{index-DBbivBNs.d.ts.map → index-8VIogTyS.d.ts.map} +1 -1
- package/dist/{index-CM-SM00y.d.ts → index-DT73JraI.d.ts} +3 -3
- package/dist/{index-CM-SM00y.d.ts.map → index-DT73JraI.d.ts.map} +1 -1
- package/dist/{index-Bfs5aufo.d.ts → index-fNXZMCzX.d.ts} +9 -9
- package/dist/{index-Bfs5aufo.d.ts.map → index-fNXZMCzX.d.ts.map} +1 -1
- package/dist/index.d.ts +138 -9
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +709 -7
- package/dist/index.js.map +1 -1
- package/dist/{llm-judge-DbJdo8Nj.js → llm-judge-BtJ2Sfk_.js} +5 -2
- package/dist/{llm-judge-DbJdo8Nj.js.map → llm-judge-BtJ2Sfk_.js.map} +1 -1
- package/dist/{matrix-BpI5Trmo.d.ts → matrix-DiHmUobV.d.ts} +2 -2
- package/dist/{matrix-BpI5Trmo.d.ts.map → matrix-DiHmUobV.d.ts.map} +1 -1
- package/dist/multishot/golden/index.d.ts +1 -1
- package/dist/multishot/index.d.ts +2 -2
- package/dist/openapi.json +1 -1
- package/dist/{produced-state-CtSIp5cQ.js → produced-state-Cm6DU_Ao.js} +2 -2
- package/dist/{produced-state-CtSIp5cQ.js.map → produced-state-Cm6DU_Ao.js.map} +1 -1
- package/dist/{promotion-policy-CkXSgKkF.d.ts → promotion-policy-WSXtBgBb.d.ts} +2 -2
- package/dist/{promotion-policy-CkXSgKkF.d.ts.map → promotion-policy-WSXtBgBb.d.ts.map} +1 -1
- package/dist/{provenance-CIRUardl.d.ts → provenance-CafMdZKM.d.ts} +7 -3
- package/dist/{provenance-CIRUardl.d.ts.map → provenance-CafMdZKM.d.ts.map} +1 -1
- package/dist/rl.d.ts +1 -1
- package/dist/{skillopt-optimization-method-B2R9C5aG.js → skillopt-optimization-method-DzlF2RM7.js} +2 -2
- package/dist/{skillopt-optimization-method-B2R9C5aG.js.map → skillopt-optimization-method-DzlF2RM7.js.map} +1 -1
- package/dist/{statistical-heldout-DFS7QGpS.d.ts → statistical-heldout-UhiexnjU.d.ts} +2 -2
- package/dist/{statistical-heldout-DFS7QGpS.d.ts.map → statistical-heldout-UhiexnjU.d.ts.map} +1 -1
- package/dist/{store-tool-spans-B2DJ_82T.d.ts → store-tool-spans-D_qMl2__.d.ts} +3 -3
- package/dist/{store-tool-spans-B2DJ_82T.d.ts.map → store-tool-spans-D_qMl2__.d.ts.map} +1 -1
- package/dist/{tool-groups-BnXlCJZQ.d.ts → tool-groups-RGYfVWpc.d.ts} +2 -2
- package/dist/tool-groups-RGYfVWpc.d.ts.map +1 -0
- package/dist/trace-repair/index.d.ts +1 -1
- package/dist/traces.d.ts +2 -2
- package/dist/{types-i21ccEkr.d.ts → types-CMyW4GnH.d.ts} +2 -2
- package/dist/{types-i21ccEkr.d.ts.map → types-CMyW4GnH.d.ts.map} +1 -1
- package/dist/{types-Dy237wiH.d.ts → types-JHMOqZI4.d.ts} +11 -1
- package/dist/types-JHMOqZI4.d.ts.map +1 -0
- package/dist/wire/index.d.ts +3 -3
- package/dist/wire/index.d.ts.map +1 -1
- package/docs/campaign-proposers.md +4 -0
- package/package.json +19 -18
- package/dist/tool-groups-BnXlCJZQ.d.ts.map +0 -1
- package/dist/types-Dy237wiH.d.ts.map +0 -1
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import { R as Scenario, b as GenerationRecord, p as Gate, w as JudgeScore } from "./types-
|
|
1
|
+
import { R as Scenario, b as GenerationRecord, p as Gate, w as JudgeScore } from "./types-JHMOqZI4.js";
|
|
2
2
|
import { T as EProcessState, r as SignedManifest } from "./pre-registration-DHz6P_6f.js";
|
|
3
3
|
import { a as PairedPromotionDecision, d as PairedBootstrapResult, i as PairedMcNemarEvidence, r as PairedDecisionStatistic, t as PairedDecisionMethod } from "./paired-promotion-decision-CGzg0cI_.js";
|
|
4
4
|
//#region src/campaign/gates/sequential.d.ts
|
|
@@ -292,4 +292,4 @@ declare function dimensionRegressions(candidate: Map<string, Record<string, Judg
|
|
|
292
292
|
}): DimensionRegression[];
|
|
293
293
|
//#endregion
|
|
294
294
|
export { sequentialPairedGate as _, detectScale as a, pairHoldout as c, SequentialDecision as d, SequentialObservation as f, sequentialDecide as g, SequentialStreamState as h, PairedHoldout as i, SequentialDecideFn as l, SequentialPairedGateOptions as m, HeldoutSignificance as n, dimensionRegressions as o, SequentialPairedGate as p, HeldoutSignificanceOptions as r, heldoutSignificance as s, DimensionRegression as t, SequentialDecideOptions as u };
|
|
295
|
-
//# sourceMappingURL=statistical-heldout-
|
|
295
|
+
//# sourceMappingURL=statistical-heldout-UhiexnjU.d.ts.map
|
package/dist/{statistical-heldout-DFS7QGpS.d.ts.map → statistical-heldout-UhiexnjU.d.ts.map}
RENAMED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"statistical-heldout-
|
|
1
|
+
{"version":3,"file":"statistical-heldout-UhiexnjU.d.ts","names":[],"sources":["../src/campaign/gates/sequential.ts","../src/campaign/gates/statistical-heldout.ts"],"mappings":";;;;KAsCY;UAEK;EACf,UAAU;;EAEV;;EAEA;;;EAGA;;UAGe;;;EAGf;;;;EAIA;;;EAGA;;EAEA;;;;EAIA;;;EAGA;;;EAGA,kBAAkB;;EAElB;;;;;;;EAOA,SAAS;;;;;;KAOC,wBAAwB;EAAkB,UAAU;;UAE/C,qBAAqB,qBAAqB,kBAAkB,WAAW,kBAC9E,KAAK,WAAW;;;;;;;;EAQxB,QAAQ,gBAAgB;;;EAGxB,SAAS;;;;;;;;;;;;;;;;iBAsPK,qBAAqB,qBAAqB,kBAAkB,WAAW,UACrF,SAAS,8BACR,qBAAqB,WAAW;UAsFlB;;EAEf;;EAEA;;EAEA;;EAEA;;UAGe;GACd;IAAQ,SAAS;;IAAyB;IAAe;;;EAE1D,SAAS;;;;;;;;;;;;;;;;;;;;;;;;;iBA0BK,iBAAiB,UAAS,0BAA+B;;;UC/axD;;EAEf;;EAEA;;EAEA;;;;;;;;;;;;;iBAcc,YACd,WAAW,YAAY,eAAe,cACtC,UAAU,YAAY,eAAe,cACrC,aAAa,aACb,SAAS,GAAG,oCACX;UAsDc;EACf,QAAQ;;;;;;;;;;;;EAYR,WAAW;;;;EAIX,iBAAiB;;;;;;;EAOjB,UAAU;;EAEV,mBAAmB;;EAEnB,SAAS;;;;EAIT;;EAEA;;EAEA;;EAEA,gBAAgB;;EAEhB;;;;EAIA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;;EAEA;EACA;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA6Bc,oBACd,QAAQ,eACR,OAAM,6BACL;UAkEc;EACf;;;EAGA,WAAW;;;;EAIX;;EAEA;IAAM;IAAa;;;EAEnB,mBAAmB;;EAEnB,SAAS;;EAET;;;;;EAKA;EACA;EACA;;;;;iBAMc,YAAY;;;;;;;;;;;;;;;;;;;iBAsBZ,qBACd,WAAW,YAAY,eAAe,cACtC,UAAU,YAAY,eAAe,cACrC,aAAa,aACb,8BACA;EACE;EACA;EACA;EACA;;;EAGA;IAED"}
|
|
@@ -5,8 +5,8 @@ import { u as RunTokenUsage } from "./run-record-DQjRcYwA.js";
|
|
|
5
5
|
import { A as SearchSpanResult, B as ViewSpansResult, D as DatasetOverview, F as TraceAnalystFilters, H as ViewTraceResult, S as BoundedTraceAnalysisStoreOptions, T as TraceAnalysisStoreContext, f as AnalystUsageReceipt, i as AnalystFinding, j as SearchTraceResult, k as QueryTracesPage, w as TraceAnalysisStore } from "./types-DMoNFDWi.js";
|
|
6
6
|
import { l as RawProviderSink, o as ProviderRedactor } from "./raw-provider-sink-BU29Sh8h.js";
|
|
7
7
|
import { s as TraceStore } from "./store-Cq9oOrI1.js";
|
|
8
|
-
import { i as TraceAnalystLimits, r as TraceAnalysisEngineResult, t as TraceAnalysisEngine } from "./engine-
|
|
9
|
-
import { t as TraceToolGroupName } from "./tool-groups-
|
|
8
|
+
import { i as TraceAnalystLimits, r as TraceAnalysisEngineResult, t as TraceAnalysisEngine } from "./engine-CAmTUk52.js";
|
|
9
|
+
import { t as TraceToolGroupName } from "./tool-groups-RGYfVWpc.js";
|
|
10
10
|
//#region src/trace-analyst/store-boundary.d.ts
|
|
11
11
|
/** Apply the public validation, cancellation, not-found, and size rules to any adapter. */
|
|
12
12
|
declare function createBoundedTraceAnalysisStore(source: TraceAnalysisStore, options?: BoundedTraceAnalysisStoreOptions): TraceAnalysisStore;
|
|
@@ -405,4 +405,4 @@ declare class ToolTraceMissingError extends CaptureIntegrityError {
|
|
|
405
405
|
declare function toolSpansToTraceAnalysisStore(spans: readonly ToolSpan[] | null | undefined, options?: ToolSpansToTraceAnalysisStoreOptions): TraceAnalysisStore;
|
|
406
406
|
//#endregion
|
|
407
407
|
export { REDACTION_VERSION as A, exportRunAsOtlp as B, scoreTraceInsightReadiness as C, AnalyzeTracesResult as D, AnalyzeTracesOptions as E, OtlpFlatLine as F, ExtractedUsage as G, CaptureFetchOptions as H, OTEL_AGENT_EVAL_SCOPE as I, extractUsageFromResponse as J, SseUsageMode as K, OtlpExport as L, RedactionRule as M, redactString as N, analyzeTraces as O, redactValue as P, OtlpResourceSpans as R, planTraceInsightQuestions as S, AnalyzeTracesInput as T, captureFetchToRawSink as U, CaptureFetchContext as V, ExtractUsageFromSseOptions as W, createBoundedTraceAnalysisStore as X, extractUsageFromSse as Y, buildTraceInsightPrompt as _, ToolSpansToTraceAnalysisStoreOptions as a, domainEvidencePattern as b, TraceInsightFinding as c, TraceInsightQualityGate as d, TraceInsightQuestion as f, buildTraceInsightContext as g, TraceInsightTask as h, OtlpFileTraceStoreOptions as i, RedactionReport as j, DEFAULT_REDACTION_RULES as k, TraceInsightPanelRole as l, TraceInsightSuite as m, toolSpansToTraceAnalysisStore as n, otlpTextToTraceAnalysisStore as o, TraceInsightReadiness as p, extractUsage as q, OtlpFileTraceStore as r, TraceInsightContext as s, ToolTraceMissingError as t, TraceInsightPromptInput as u, defaultTraceInsightPanel as v, tokenizeDomainWords as w, inferDomainKeywords as x, describeTraceInsightScope as y, OtlpSpan as z };
|
|
408
|
-
//# sourceMappingURL=store-tool-spans-
|
|
408
|
+
//# sourceMappingURL=store-tool-spans-D_qMl2__.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"store-tool-spans-
|
|
1
|
+
{"version":3,"file":"store-tool-spans-D_qMl2__.d.ts","names":[],"sources":["../src/trace-analyst/store-boundary.ts","../src/trace/extract-usage.ts","../src/trace/capture-fetch.ts","../src/trace/otel.ts","../src/trace/otlp-flat.ts","../src/trace/redact.ts","../src/trace-analyst/analyst.ts","../src/trace-analyst/insights.ts","../src/trace-analyst/store-otlp.ts","../src/trace-analyst/store-tool-spans.ts"],"mappings":";;;;;;;;;;;iBA6BgB,gCACd,QAAQ,oBACR,UAAS,mCACR;;;KChBS,iBAAiB;KACjB;UAEK;;EAEf,OAAO;;;;;;;iBAuDO,aAAa,gBAAgB;;;;;;iBA6C7B,oBACd,cACA,UAAS,6BACR;;;;;;;iBA4DmB,yBACpB,UAAU,UACV,aAAa,6BACZ,QAAQ;;;UCrKM;;;EAGf;;EAEA;;;EAGA;;EAEA;;EAEA;;UAGe;;EAEf,WAAW;;;EAGX;;;EAGA;;;;;;;;EAQA,WAAW,OAAO,gBAAgB,KAAK;;EAEvC,eAAe;;iBAkDD,sBACd,cAAc,WAAW,OACzB,MAAM,iBACN,KAAK,qBACL,OAAM,6BACE,WAAW;;;cCxFR;EAA0B;EAAoC;;UAE1D;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA,YAAY;IACV;IACA;MAAS;MAAsB;MAAmB;MAAsB;;;EAE1E,SAAS;IAAQ;IAAsB;IAAc,aAAa;;EAClE;IAAW;IAAc;;;UAGV;EACf;IAAY,YAAY;;EACxB,YAAY;IAAQ,cAAc;IAAuB,OAAO;;;UAGjD;EACf,eAAe;;;iBAIK,gBACpB,OAAO,YACP,eACA,gBAAe,4CACd,QAAQ;;;KClDC;UAEK;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;IACE,MAAM;IACN;;EAEF;IAAY,YAAY;;EACxB,YAAY;EACZ,SAAS;IAAQ;IAAc;IAAuB,aAAa;;;;;;;;;;;;;;;;;;UCNpD;EACf;EACA,SAAS;;EAET;;UAGe;EACf;EACA,QAAQ;;;cAIG,yBAAyB;cAezB;;;;;iBAMG,aACd,eACA,QAAO;EACJ;EAAgB,QAAQ;;;;;;;iBAuBb,YACd,gBACA,QAAO,iBACP,SAAQ;EACL;EAAgB,QAAQ;;;;UChEZ;EACf;EACA;EACA;EACA;;KAGU,sBAAsB;UAEjB;;EAEf,iBAAiB;;EAEjB,QAAQ;;EAER;;EAEA,YAAY;EACZ,SAAS,QAAQ;EACjB;EACA;EACA,aAAa;EACb;EACA,yBAAyB;EACzB,4BAA4B;EAC5B,eAAe,SAAS;EACxB,OAAO;EACP,OAAO,iBAAiB,SAAS;EACjC,SAAS;;;;;;;;iBASW,cACpB,OAAO,oBACP,SAAS,uBACR,QAAQ;;;UCtDM;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,OAAO;;UAGQ;EACf;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf,OAAO;EACP,WAAW;EACX,QAAQ;EACR,SAAS;EACT;;UAGe;EACf,OAAO;EACP;EACA;EACA,WAAW;EACX,OAAO;EACP,UAAU;EACV,OAAO;EACP,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,OAAO;;iBA0BO,oBAAoB;iBAMpB,oBAAoB,OAAO;iBAuB3B,sBAAsB,qBAAqB;iBAS3C,0BAA0B,OAAO;iBAsBjC,0BAA0B,OAAO,0BAA0B;iBAsD3D,yBAAyB,OAAO,0BAA0B;iBAa1D,2BAA2B,SAAS,sBAAsB;iBA2D1D,4BAA4B;iBA6B5B,wBAAwB,OAAO;;;UC1K9B;;EAEf;;EAEA;;EAEA;;EAEA;;KAGG,gCAAgC;UAEpB,kCAAkC;;EAEjD;;;;;;;;EAQA;;uBAMa,kCAAkC;mBAC9B;mBACA;mBACA;mBACA;UACT;UACA;UACA;UACA;EAER,YAAY,MAAM;EAyBZ,SAAS,kBAAkB,UAAU,4BAA4B;EAOjE,SACJ;IAAQ;IAAkB;KAC1B,UAAU,4BACT;EAeG,YACJ,UAAU,qBACV,UAAU,4BACT,QAAQ;EAoDL,YACJ;IAAQ,UAAU;IAAqB;IAAe;KACtD,UAAU,4BACT,QAAQ;EAmBL,YACJ,UAAU,qBACV,UAAU,4BACT;EAMG,UACJ;IAAQ;IAAkB;KAC1B,UAAU,4BACT,QAAQ;EAuCL,UACJ;IAAQ;IAAkB;IAA6B;KACvD,UAAU,4BACT,QAAQ;EAsFL,YACJ;IAAQ;IAAkB;IAAuB;KACjD,UAAU,4BACT,QAAQ;EA4CL,WACJ;IAAQ;IAAkB;IAAiB;IAAuB;KAClE,UAAU,4BACT,QAAQ;;;EA6CL,cAAc,UAAU,4BAA4B;UAI5C;qBAmBK,WAAW,SAAS,cAAc,QAAQ;qBAE1C;UAEL;UAmBA;UAsHA;UA+DN;UAkBA;UAwDA;UAmBM;;cA+CH,2BAA2B;mBACrB;mBACA;EAEjB,YAAY,MAAM;;;YAYF,WAAW,SAAS,cAAc,QAAQ;YAkBhD;;;iBAwBI,6BACd,cACA,OAAM,uCACL;;;;cC/7BU,8BAA8B;EACzC;;;;;;iBASc,8BACd,gBAAgB,+BAChB,UAAS,uCACR"}
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
import { w as TraceAnalysisStore } from "./types-DMoNFDWi.js";
|
|
2
|
-
import { m as TraceAnalysisToolDescriptor } from "./engine-
|
|
2
|
+
import { m as TraceAnalysisToolDescriptor } from "./engine-CAmTUk52.js";
|
|
3
3
|
//#region src/analyst/tool-groups.d.ts
|
|
4
4
|
/** Named tool sets. Kinds pass `tools: TRACE_TOOL_GROUPS.failureForensics` etc. */
|
|
5
5
|
type TraceToolGroupName =
|
|
@@ -25,4 +25,4 @@ type TraceToolGroupName =
|
|
|
25
25
|
declare function buildTraceToolsForGroup(group: TraceToolGroupName, store: TraceAnalysisStore): TraceAnalysisToolDescriptor[];
|
|
26
26
|
//#endregion
|
|
27
27
|
export { buildTraceToolsForGroup as n, TraceToolGroupName as t };
|
|
28
|
-
//# sourceMappingURL=tool-groups-
|
|
28
|
+
//# sourceMappingURL=tool-groups-RGYfVWpc.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"tool-groups-RGYfVWpc.d.ts","names":[],"sources":["../src/analyst/tool-groups.ts"],"mappings":";;;;KAmBY;;;;;;;;;;;;;;;;;;;;iBAgDI,wBACd,OAAO,oBACP,OAAO,qBACN"}
|
|
@@ -2,7 +2,7 @@ import { r as CaptureIntegrityError } from "../errors-DEE6u6ot.js";
|
|
|
2
2
|
import { b as CustomTokenPricing, c as CostLedgerHandle, p as CostProvenance } from "../cost-ledger-DbQdN3nO.js";
|
|
3
3
|
import { u as RunTokenUsage } from "../run-record-DQjRcYwA.js";
|
|
4
4
|
import { t as DefaultVerdict } from "../verdict-E4eRNf7-.js";
|
|
5
|
-
import { i as TraceAnalystLimits, t as TraceAnalysisEngine } from "../engine-
|
|
5
|
+
import { i as TraceAnalystLimits, t as TraceAnalysisEngine } from "../engine-CAmTUk52.js";
|
|
6
6
|
import { t as PrimeBridgeTransport } from "../prime-bridge-transport-6feEglLf.js";
|
|
7
7
|
import { a as RecordedTrajectoryStep, h as isRecordedTimeout } from "../steps-CiNVJry_.js";
|
|
8
8
|
//#region src/trace-repair/mini-swe-scaffold.d.ts
|
package/dist/traces.d.ts
CHANGED
|
@@ -2,10 +2,10 @@ import { c as ValidationError, o as LimitExceededError, r as CaptureIntegrityErr
|
|
|
2
2
|
import { C as TraceEvent, E as isToolSpan, S as ToolSpan, T as isLlmSpan, _ as Span, a as FAILURE_CLASSES, b as SpanStatus, c as JudgeSpan, d as RetrievalSpan, f as Run, g as SandboxSpan, h as RunStatus, i as EventKind, l as LlmSpan, m as RunOutcome, n as BudgetLedgerEntry, o as FailureClass, p as RunLayer, r as BudgetSpec, s as GenericSpan, t as Artifact, u as Message, v as SpanBase, w as isJudgeSpan, x as TRACE_SCHEMA_VERSION, y as SpanKind } from "./schema-DID1Cqct.js";
|
|
3
3
|
import { a as RunRecord, l as RunTerminalOutcome, s as RunSplitTag, u as RunTokenUsage } from "./run-record-DQjRcYwA.js";
|
|
4
4
|
import { A as SearchSpanResult, B as ViewSpansResult, C as TRACE_ANALYSIS_LIMITS, D as DatasetOverview, E as DEFAULT_TRACE_ANALYST_BUDGETS, F as TraceAnalystFilters, H as ViewTraceResult, I as TraceAnalystSpan, L as TraceAnalystSpanKind, M as SpanMatchRecord, N as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, O as ErrorCluster, P as TraceAnalystByteBudgets, R as TraceAnalystSpanStatus, S as BoundedTraceAnalysisStoreOptions, T as TraceAnalysisStoreContext, V as ViewTraceOversized, j as SearchTraceResult, k as QueryTracesPage, w as TraceAnalysisStore, z as TraceAnalystTraceSummary } from "./types-DMoNFDWi.js";
|
|
5
|
-
import { A as REDACTION_VERSION, B as exportRunAsOtlp, C as scoreTraceInsightReadiness, D as AnalyzeTracesResult, E as AnalyzeTracesOptions, F as OtlpFlatLine, G as ExtractedUsage, H as CaptureFetchOptions, I as OTEL_AGENT_EVAL_SCOPE, J as extractUsageFromResponse, K as SseUsageMode, L as OtlpExport, M as RedactionRule, N as redactString, O as analyzeTraces, P as redactValue, R as OtlpResourceSpans, S as planTraceInsightQuestions, T as AnalyzeTracesInput, U as captureFetchToRawSink, V as CaptureFetchContext, W as ExtractUsageFromSseOptions, X as createBoundedTraceAnalysisStore, Y as extractUsageFromSse, _ as buildTraceInsightPrompt, a as ToolSpansToTraceAnalysisStoreOptions, b as domainEvidencePattern, c as TraceInsightFinding, d as TraceInsightQualityGate, f as TraceInsightQuestion, g as buildTraceInsightContext, h as TraceInsightTask, i as OtlpFileTraceStoreOptions, j as RedactionReport, k as DEFAULT_REDACTION_RULES, l as TraceInsightPanelRole, m as TraceInsightSuite, n as toolSpansToTraceAnalysisStore, o as otlpTextToTraceAnalysisStore, p as TraceInsightReadiness, q as extractUsage, r as OtlpFileTraceStore, s as TraceInsightContext, t as ToolTraceMissingError, u as TraceInsightPromptInput, v as defaultTraceInsightPanel, w as tokenizeDomainWords, x as inferDomainKeywords, y as describeTraceInsightScope, z as OtlpSpan } from "./store-tool-spans-
|
|
5
|
+
import { A as REDACTION_VERSION, B as exportRunAsOtlp, C as scoreTraceInsightReadiness, D as AnalyzeTracesResult, E as AnalyzeTracesOptions, F as OtlpFlatLine, G as ExtractedUsage, H as CaptureFetchOptions, I as OTEL_AGENT_EVAL_SCOPE, J as extractUsageFromResponse, K as SseUsageMode, L as OtlpExport, M as RedactionRule, N as redactString, O as analyzeTraces, P as redactValue, R as OtlpResourceSpans, S as planTraceInsightQuestions, T as AnalyzeTracesInput, U as captureFetchToRawSink, V as CaptureFetchContext, W as ExtractUsageFromSseOptions, X as createBoundedTraceAnalysisStore, Y as extractUsageFromSse, _ as buildTraceInsightPrompt, a as ToolSpansToTraceAnalysisStoreOptions, b as domainEvidencePattern, c as TraceInsightFinding, d as TraceInsightQualityGate, f as TraceInsightQuestion, g as buildTraceInsightContext, h as TraceInsightTask, i as OtlpFileTraceStoreOptions, j as RedactionReport, k as DEFAULT_REDACTION_RULES, l as TraceInsightPanelRole, m as TraceInsightSuite, n as toolSpansToTraceAnalysisStore, o as otlpTextToTraceAnalysisStore, p as TraceInsightReadiness, q as extractUsage, r as OtlpFileTraceStore, s as TraceInsightContext, t as ToolTraceMissingError, u as TraceInsightPromptInput, v as defaultTraceInsightPanel, w as tokenizeDomainWords, x as inferDomainKeywords, y as describeTraceInsightScope, z as OtlpSpan } from "./store-tool-spans-D_qMl2__.js";
|
|
6
6
|
import { a as NoopRawProviderSink, c as RawProviderEvent, d as defaultProviderRedactor, f as providerFromBaseUrl, i as InMemoryRawProviderSinkOptions, l as RawProviderSink, n as FileSystemRawProviderSinkOptions, o as ProviderRedactor, r as InMemoryRawProviderSink, s as RawProviderDirection, t as FileSystemRawProviderSink, u as RawProviderSinkFilter } from "./raw-provider-sink-BU29Sh8h.js";
|
|
7
7
|
import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, r as FileSystemTraceStoreOptions, s as TraceStore, t as EventFilter } from "./store-Cq9oOrI1.js";
|
|
8
|
-
import { f as BuildTraceAnalysisToolsOptions, g as traceAnalystFunctionGroup, h as buildTraceAnalysisToolDescriptors, m as TraceAnalysisToolDescriptor, p as TRACE_ANALYST_TOOL_NAMESPACE } from "./engine-
|
|
8
|
+
import { f as BuildTraceAnalysisToolsOptions, g as traceAnalystFunctionGroup, h as buildTraceAnalysisToolDescriptors, m as TraceAnalysisToolDescriptor, p as TRACE_ANALYST_TOOL_NAMESPACE } from "./engine-CAmTUk52.js";
|
|
9
9
|
import { a as TraceEmitterOptions, i as TraceEmitter, n as RunCompleteHookContext, r as SpanHandle, t as RunCompleteHook } from "./emitter-Bvnu0VzL.js";
|
|
10
10
|
import { C as TOOL_LATENCY_MS, D as asNumber, E as applyLlmSpanOtlpAttributes, O as contextInputTokens, S as TOOL_ARGS_CAPTURED, T as TOOL_NAME_ATTR_KEYS, _ as LlmSpanOtlpInput, a as LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, b as RUN_COST_ATTR_KEYS, c as LLM_COST_USD, d as LLM_MODEL_ATTR_KEYS, f as LLM_MODEL_NAME, g as LLM_REASONING_TOKEN_ATTR_KEYS, h as LLM_REASONING_TOKENS, i as LLM_CACHE_WRITE_TOKENS, k as firstNumberAttr, l as LLM_INPUT_TOKENS, m as LLM_OUTPUT_TOKEN_ATTR_KEYS, n as LLM_CACHED_TOKENS, o as LLM_CONTEXT_TOKENS, p as LLM_OUTPUT_TOKENS, r as LLM_CACHED_TOKEN_ATTR_KEYS, s as LLM_COST_ATTR_KEYS, t as INPUT_VALUE, u as LLM_INPUT_TOKEN_ATTR_KEYS, v as OPENINFERENCE_SPAN_KIND, w as TOOL_NAME, x as SPAN_KIND_ATTR_KEYS, y as OUTPUT_VALUE } from "./attribute-vocabulary-DLJ6303h.js";
|
|
11
11
|
import { a as RunIntegrityReport, i as RunIntegrityIssueCode, n as RunIntegrityExpectations, o as assertRunCaptured, r as RunIntegrityIssue, s as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-B_EDELom.js";
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
import { s as RunSplitTag } from "./run-record-DQjRcYwA.js";
|
|
2
|
-
import { R as Scenario, d as DispatchContext } from "./types-
|
|
2
|
+
import { R as Scenario, d as DispatchContext } from "./types-JHMOqZI4.js";
|
|
3
3
|
//#region src/benchmarks/types.d.ts
|
|
4
4
|
type BenchmarkTaskKind = 'retrieval' | 'rag-answer' | 'hallucination' | 'kb-improvement' | 'routing' | 'custom';
|
|
5
5
|
type BenchmarkFamily = 'beir' | 'mteb-retrieval' | 'msmarco' | 'trec-dl' | 'miracl' | 'lotte' | 'bright' | 'crag' | 'hotpotqa' | 'kilt' | 'ragtruth' | 'faithbench' | 'first-party' | 'custom';
|
|
@@ -90,4 +90,4 @@ declare const BENCHMARK_SPLIT_SEED = "agent-eval-v1";
|
|
|
90
90
|
declare function deterministicSplit(itemId: string, seed?: string): RunSplitTag;
|
|
91
91
|
//#endregion
|
|
92
92
|
export { BenchmarkFamily as a, BenchmarkSource as c, BenchmarkEvaluation as i, BenchmarkTaskKind as l, BenchmarkAdapter as n, BenchmarkResponder as o, BenchmarkDatasetItem as r, BenchmarkScenario as s, BENCHMARK_SPLIT_SEED as t, deterministicSplit as u };
|
|
93
|
-
//# sourceMappingURL=types-
|
|
93
|
+
//# sourceMappingURL=types-CMyW4GnH.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types-
|
|
1
|
+
{"version":3,"file":"types-CMyW4GnH.d.ts","names":[],"sources":["../src/benchmarks/types.ts"],"mappings":";;;KASY;KAQA;UAgBK,qBAAqB;;;EAGpC;;EAEA,SAAS;;EAET,QAAQ;;EAER,SAAS;;EAET,WAAW;;EAEX;;EAEA,SAAS;EACT,WAAW;;UAGI;;;;EAIf;;EAEA;;EAEA,aAAa;;;EAGb,KAAK;EACL;;UAGe;EACf;EACA;EACA;EACA;EACA;;;UAQe,iBAAiB,kBAAkB,oBAAoB;;EAEtE;EACA,SAAS;EACT,WAAW;EACX;EACA,SAAS;EACT;;;;;EAKA,YAAY,OAAO,cAAc,QAAQ,qBAAqB;;EAE9D,SAAS,MAAM,qBAAqB,WAAW,UAAU,YAAY,QAAQ;;;;EAI7E,YAAY,iBAAiB;;UAGd,kBAAkB,4BAA4B;EAC7D;EACA;EACA,QAAQ;EACR,UAAU;EACV,UAAU;EACV,MAAM,qBAAqB;;KAGjB,mBAAmB,oBAAoB,uBAAuB;EACxE,UAAU,kBAAkB;EAC5B,MAAM,qBAAqB;EAC3B,SAAS;MACL,QAAQ,aAAa;;;cAoBd;;;;;;;;;iBAUG,mBACd,gBACA,gBACC"}
|
|
@@ -195,6 +195,12 @@ interface ProposedCandidate {
|
|
|
195
195
|
* primitive it used. Survives to `GenerationCandidate.rationale` and the
|
|
196
196
|
* emitted provenance record. */
|
|
197
197
|
rationale: string;
|
|
198
|
+
/** Typed, JSON-safe attribution the proposer wants measured alongside the
|
|
199
|
+
* candidate (for example a policy-edit candidate record). Opaque to the
|
|
200
|
+
* loop: it is carried byte-for-byte onto `GenerationCandidate.attribution`
|
|
201
|
+
* and never interpreted here — the producer tags it with its own schema
|
|
202
|
+
* field and validates it on readback. */
|
|
203
|
+
attribution?: Readonly<Record<string, unknown>>;
|
|
198
204
|
}
|
|
199
205
|
/** Type guard: a proposal carrying its rationale vs a bare
|
|
200
206
|
* surface. The loop branches on this to populate `GenerationCandidate`. */
|
|
@@ -604,6 +610,10 @@ interface GenerationCandidate {
|
|
|
604
610
|
* "because rationale Z" the audit requires to survive to the result.
|
|
605
611
|
* Present when the proposer returned a `ProposedCandidate`. */
|
|
606
612
|
rationale?: string;
|
|
613
|
+
/** Proposer-supplied typed attribution, carried unchanged from
|
|
614
|
+
* `ProposedCandidate.attribution`. Readers validate against the producer's
|
|
615
|
+
* schema tag; the loop never interprets it. */
|
|
616
|
+
attribution?: Readonly<Record<string, unknown>>;
|
|
607
617
|
}
|
|
608
618
|
interface CampaignAggregates {
|
|
609
619
|
byJudge: Record<string, JudgeAggregate>;
|
|
@@ -650,4 +660,4 @@ interface CampaignResult<TArtifact = unknown, TScenario extends Scenario = Scena
|
|
|
650
660
|
}
|
|
651
661
|
//#endregion
|
|
652
662
|
export { LabeledScenarioWrite as A, ScoredSurfaceOutcome as B, JudgeDimension as C, LabeledScenarioSampleArgs as D, LabeledScenarioRecord as E, ProposeContext as F, labelTrustRank as G, SurfaceProposer as H, ProposedCandidate as I, RedactionStatus as L, OptimizerConfig as M, ParetoParent as N, LabeledScenarioSource as O, ProposalTrackContext as P, Scenario as R, JudgeConfig as S, LabelTrust as T, TraceSpan as U, SessionScript as V, isProposedCandidate as W, GateDecision as _, CampaignResult as a, GenerationRecord as b, CampaignTraceWriter as c, DispatchContext as d, DispatchFn as f, GateContribution as g, GateContext as h, CampaignCostMeter as i, MutableSurface as j, LabeledScenarioStore as k, CodeSurface as l, GateCheckStatus as m, CampaignArtifactWriter as n, CampaignScenarioIdentity as o, Gate as p, CampaignCellResult as r, CampaignTokenUsage as s, CampaignAggregates as t, ComponentSurface as u, GateResult as v, JudgeScore as w, JudgeAggregate as x, GenerationCandidate as y, ScenarioAggregate as z };
|
|
653
|
-
//# sourceMappingURL=types-
|
|
663
|
+
//# sourceMappingURL=types-JHMOqZI4.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types-JHMOqZI4.d.ts","names":[],"sources":["../src/campaign/types.ts"],"mappings":";;;;;;;UAgCiB;EACf;EACA;EACA;;;;;EAKA;;;UAIe,iCAAiC,KAAK;EACrD;;;;;UAMe;EACf;;EAEA;EACA;EACA;EACA;EACA,QAAQ;EACR,OAAO;EACP,WAAW;EACX,MAAM;;EAEN;;EAEA;;;;;;;;EAQA;;;;KAKU,WAAW,kBAAkB,UAAU,cACjD,UAAU,WACV,KAAK,oBACF,QAAQ;;;;UAOI,cAAc,WAAW;EACxC;EACA;EACA;;EAEA;;;EAGA,sBAAsB,UAAU,WAAW,sBAAsB,UAAU,cAAc;;UAK1E;;EAEf;;EAEA;;;;;;;;;UAUe,YAAY,WAAW,kBAAkB,WAAW;EACnE;EACA,YAAY;;;;EAIZ;;;EAGA,MAAM;IACJ,UAAU;IACV,UAAU;IACV,QAAQ;;IAER,aAAa;IACb;IACA,WAAW;MACT,aAAa,QAAQ;EACzB,aAAa,UAAU;;;;;;;;;;;UAYR;EACf,YAAY;EACZ;EACA;;EAEA,UAAU;;;;;;;EAOV;;;;EAIA,eAAe,eAAe;IAAQ;IAAe;;;;;EAIrD;;;EAGA;;EAEA;;EAEA,WAAW,eAAe;;;;;;;UAUX;WACN;;;WAGA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;;WAGA;aACE;aACA;aACA;;;WAGF;;;UAIM;WACN;WACA,YAAY,SAAS;;;;;;;;;;;KAYpB,0BAA0B,mBAAmB;;;;;;;UAQxC;EACf,SAAS;;EAET;;;;EAIA;;;;;;EAMA,cAAc,SAAS;;;;iBAKT,oBACd,OAAO,iBAAiB,oBACvB,SAAS;;;;;;;;;UAkBK;EACf,SAAS;EACT;;;EAGA,YAAY;;;EAGZ;;EAEA;EACA;EACA;;;UAIe;EACf;EACA;EACA;EACA;EACA;;;;;UAMe;;;EAGf;;EAEA;EACA;EACA;EACA,YAAY;;;;;EAKZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;EAC1E;IACE;IACA;;;;;UAMa,eAAe,YAAY;;;WAGjC,gBAAgB;WAChB,SAAS,cAAc;WACvB,UAAU,cAAc;;WAExB;WACA;WACA,QAAQ;;WAER,QAAQ;;;WAGR,kBAAkB;;;WAGlB,mBAAmB;;;;WAInB,gBAAgB;;;;WAIhB;;;;;;;WAOA,gBAAgB,cAAc;;WAE9B,aAAa;WACb;;;;;;;;;;;;;;UAeM,gBAAgB,YAAY;EAC3C;;;;;EAKA,QAAQ,KAAK,eAAe,aAAa,QAAQ,MAAM,iBAAiB;;;EAGxE,QAAQ;IAAQ,SAAS,cAAc;;IAAwB;IAAe;;;UAG/D;EACf;EACA;EACA,mBAAmB,qBAAqB;;UAGzB,wBAAwB;EACvC,UAAU;;;KAMA;;KAGA;UAEK;EACf;EACA,QAAQ;EACR;;UAGe,YAAY,WAAW,kBAAkB;EACxD,oBAAoB,YAAY;EAChC,oBAAoB,YAAY;;EAEhC,aAAa,YAAY,eAAe;;;;;EAKxC,sBAAsB,YAAY,eAAe;;;;;;;EAOjD,yBAAyB,YAAY,eAAe;;;EAGpD,uBAAuB,YAAY;EACnC,WAAW;EACX;IAAQ;IAAmB;;;EAE3B,aAAa;EACb;EACA,QAAQ;;UAGO;EACf,UAAU;EACV;EACA,mBAAmB;EACnB;;;UAIe,KAAK,qBAAqB,kBAAkB,WAAW;EACtE;EACA,OAAO,KAAK,YAAY,WAAW,aAAa,QAAQ;;;;UAOzC;EACf,KAAK,cAAc,aAAa,0BAA0B;EAC1D,SAAS;;UAGM;EACf,IAAI,aAAa;EACjB,aAAa,aAAa;;;;UAKX;EACf,MAAM,cAAc,kBAAkB,aAAa;EACnD,UAAU,cAAc,iBAAiB;;;;;;KAO/B,qBAAqB;;;;;UAMhB;;EAEf,YAAY,GACV,OAAO,KAAK,iBAAiB;IAC3B,UAAU;MAEX,QAAQ,eAAe;;;;;KAQhB;KAOA;;;;;;;;;;;;;;;KAgBA;;iBASI,eAAe,OAAO;;;;UAOrB,qBAAqB,kBAAkB,WAAW,UAAU;EAC3E,UAAU;EACV,UAAU;EACV,aAAa,eAAe;EAC5B,QAAQ;EACR;EACA;EACA,iBAAiB;;;;;EAKjB,aAAa;;EAEb;;UAGe,sBAAsB,kBAAkB,WAAW,UAAU,6BACpE,qBAAqB,WAAW;;EAExC;;;EAGA;;UAGe;EACf;;EAEA;;;;EAIA;EACA;IACE;IACA,SAAS,wBAAwB;IACjC;IACA;;;;;IAKA,WAAW;;;UAIE;EACf,QAAQ,OAAO,uBAAuB;EACtC,OAAO,MAAM,4BAA4B,QAAQ;EACjD,QAAQ;IACN;IACA;IACA,UAAU;;;IAGV,SAAS,OAAO;;;UAMH,mBAAmB;;;EAGlC;EACA;EACA;EACA;EACA;EACA,UAAU;EACV,aAAa,eAAe;;;EAG5B;;EAEA,gBAAgB;;EAEhB;;;EAGA,YAAY;;;EAGZ;;;EAGA;EACA;EACA;EACA;;;;EAIA;;EAEA;;EAEA;EACA;;UAGe;EACf;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf;EACA,YAAY;EACZ;;;;;;UAOe;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;;EAIA;;;;EAIA;IACE;IACA;IACA,iBAAiB;MAAQ;MAAgB;;;;;EAI3C,YAAY;;;;;;;;;;EAUZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;;;EAG1E;;;;EAIA;;;;EAIA,cAAc,SAAS;;UAGR;EACf,SAAS,eAAe;EACxB,YAAY,eAAe;;EAE3B,MAAM;;EAEN;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;UAGe,eAAe,qBAAqB,kBAAkB,WAAW;;EAEhF;;EAEA;EACA;;EAEA;EACA;EACA;EACA;EACA,OAAO,MAAM,mBAAmB;EAChC,YAAY;EACZ;IACE,aAAa;IACb;;EAEF,OAAO;EACP;EACA;EACA,iBAAiB;;;EAGjB,WAAW,MAAM,2BAA2B,KAAK"}
|
package/dist/wire/index.d.ts
CHANGED
|
@@ -388,14 +388,14 @@ declare function handleTracesIngest(req: TracesIngestRequest, stores: IngestionS
|
|
|
388
388
|
*/
|
|
389
389
|
declare function handleFeedbackIngest(req: FeedbackTrajectory, stores: IngestionStores): Promise<FeedbackIngestResponse>;
|
|
390
390
|
//#endregion
|
|
391
|
-
//#region node_modules/.pnpm/openapi3-ts@4.6.
|
|
391
|
+
//#region node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/specification-extension.d.ts
|
|
392
392
|
type IExtensionName = `x-${string}`;
|
|
393
393
|
type IExtensionType = any;
|
|
394
394
|
type ISpecificationExtension = {
|
|
395
395
|
[extensionName: IExtensionName]: IExtensionType;
|
|
396
396
|
};
|
|
397
397
|
//#endregion
|
|
398
|
-
//#region node_modules/.pnpm/openapi3-ts@4.6.
|
|
398
|
+
//#region node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/oas-common.d.ts
|
|
399
399
|
interface ServerObject extends ISpecificationExtension {
|
|
400
400
|
url: string;
|
|
401
401
|
description?: string;
|
|
@@ -409,7 +409,7 @@ interface ServerVariableObject extends ISpecificationExtension {
|
|
|
409
409
|
description?: string;
|
|
410
410
|
}
|
|
411
411
|
//#endregion
|
|
412
|
-
//#region node_modules/.pnpm/openapi3-ts@4.6.
|
|
412
|
+
//#region node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/openapi31.d.ts
|
|
413
413
|
interface OpenAPIObject extends ISpecificationExtension {
|
|
414
414
|
openapi: string;
|
|
415
415
|
info: InfoObject;
|
package/dist/wire/index.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/wire/schemas.ts","../../src/wire/handlers.ts","../../node_modules/.pnpm/openapi3-ts@4.6.
|
|
1
|
+
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/wire/schemas.ts","../../src/wire/handlers.ts","../../node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/specification-extension.d.ts","../../node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/oas-common.d.ts","../../node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/openapi31.d.ts","../../src/wire/openapi.ts","../../src/wire/rpc.ts","../../src/wire/rubrics.ts","../../src/wire/server.ts"],"x_google_ignoreList":[2,3,4],"mappings":";;;;;;;;cAoBa,uBAAqB,EAAA;;;;;;GAkBL,EAAA,KAAA;cAEvB,mBAAiB,EAAA;;;GAKE,EAAA,KAAA;cAIZ,cAAY,EAAA;;;;;;;;;;;;;;;;;;;GA6BL,EAAA,KAAA;cAIP,oBAAkB,EAAA;;;;;;;;;;;;;;;;;;;;;;;;;GAwBL,EAAA,KAAA;cAEb,mBAAiB,EAAA;;;;;;;;;GA6BL,EAAA,KAAA;cAInB,kBAAgB,EAAA;;;;;;;;;;GAcE,EAAA,KAAA;cAEX,2BAAyB,EAAA;;;;;;;;;;;;GAIL,EAAA,KAAA;cAIpB,uBAAqB,EAAA;;;;;GAWL,EAAA,KAAA;cAEhB,sBAAoB,EAAA;;;GAKL,EAAA,KAAA;;;;;;;cAUf,kBAAgB,EAAA;;;;;;;;;;;;;;;;GA0BL,EAAA,KAAA;cAEX,2BAAyB,EAAA;;;;;;;;;;;;;;;;;;GAQL,EAAA,KAAA;cAEpB,4BAA0B,EAAA;;;;;;;GAiBL,EAAA,KAAA;cA2DrB,0BAAwB,EAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA4BL,EAAA,KAAA;cAEnB,8BAA4B,EAAA;;;GAKL,EAAA,KAAA;KAExB,aAAa,EAAE,aAAa;KAC5B,sBAAsB,EAAE,aAAa;KACrC,uBAAuB,EAAE,aAAa;KACtC,qBAAqB,EAAE,aAAa;KACpC,yBAAyB,EAAE,aAAa;cAIvC,qBAAmB,EAAA;;;;;;GAcL,EAAA,KAAA;KAIf,kBAAkB,EAAE,aAAa;KACjC,cAAc,EAAE,aAAa;KAC7B,SAAS,EAAE,aAAa;KACxB,eAAe,EAAE,aAAa;KAC9B,cAAc,EAAE,aAAa;KAC7B,aAAa,EAAE,aAAa;KAC5B,sBAAsB,EAAE,aAAa;KACrC,kBAAkB,EAAE,aAAa;KACjC,gBAAgB,EAAE,aAAa;;;;;;;cAU9B;;;;;;;;cASA;;;;;;;;;;iBAWG,WAAW,QAAQ;;;;cCpXtB,kBAAkB;WAEX;WAEA;WACA;EAJlB,YACkB,cAChB,iBACgB,iBACA;;UAgJH;EACf,aAAa;EACb;;;;;;EAMA,OAAO;EACP;;EAEA,UAAU;EACV,SAAS;;iBAGW,YACpB,KAAK,cACL,UAAS,qBACR,QAAQ;iBAuEK,qBAAqB;iBAMrB,iBAAiB;;;;;;;;;UAmBhB;EACf,aAAa;EACb,gBAAgB;;;;;;;;;;;;iBAaI,mBACpB,KAAK,qBACL,QAAQ,kBACP,QAAQ;;;;;;iBA8BW,qBACpB,KAAK,oBACL,QAAQ,kBACP,QAAQ;;;KC9VC;KACA;KACA;GACP,eAAe,iBAAiB;;;;UCFpB,qBAAqB;EAClC;EACA;EACA;KACK,YAAY;;;UAGJ,6BAA6B;EAC1C;EACA;EACA;;;;UCPa,sBAAsB;EACnC;EACA,MAAM;EACN,UAAU;EACV,QAAQ;EACR,aAAa;EACb,WAAW;EACX,OAAO;EACP,eAAe;EACf,WAAW;;UAEE,mBAAmB;EAChC;EACA;EACA;EACA,UAAU;EACV,UAAU;EACV;;UAEa,sBAAsB;EACnC;EACA;EACA;;UAEa,sBAAsB;EACnC;EACA;EACA;;UAEa,yBAAyB;EACtC;KACK,iBAAiB,eAAe;;EAErC;KACK,mBAAmB,iBAAiB;;EAEzC;KACK,oBAAoB,kBAAkB;;EAE3C;KACK,kBAAkB,gBAAgB;;EAEvC;KACK,kBAAkB,oBAAoB;;EAE3C;KACK,iBAAiB,eAAe;;EAErC;KACK,yBAAyB,uBAAuB;;EAErD;KACK,eAAe,aAAa;;EAEjC;KACK,mBAAmB,iBAAiB;;EAEzC;KACK,mBAAmB,iBAAiB;;;UAG5B,oBAAoB;GAChC,eAAe;;UAIH,uBAAuB;EACpC;EACA;EACA;EACA,MAAM;EACN,MAAM;EACN,OAAO;EACP,SAAS;EACT,UAAU;EACV,OAAO;EACP,QAAQ;EACR,QAAQ;EACR,UAAU;EACV,cAAc,kBAAkB;;UAEnB,wBAAwB;EACrC;EACA;EACA;EACA,eAAe;EACf;EACA,cAAc,kBAAkB;EAChC,cAAc,oBAAoB;EAClC,YAAY;EACZ,YAAY;EACZ;EACA,WAAW;EACX,UAAU;;UAEG,oCAAoC;EACjD;EACA;;KAEQ;KACA;UACK,4BAA4B;EACzC;EACA;EACA;EACA;EACA,QAAQ;EACR;EACA;EACA,SAAS,eAAe;EACxB;KACK,gBAAgB,gBAAgB;;EAErC;EACA,UAAU;;UAEG,wBAAwB;EACrC;EACA,IAAI;;UAES,0BAA0B;EACvC;EACA,SAAS;EACT;;UAEa;GACZ,oBAAoB;;UAER,wBAAwB;EACrC,SAAS,eAAe;EACxB,WAAW;EACX;EACA,WAAW;;UAEE,uBAAuB;GACnC,mBAAmB;;UAEP;EACb;EACA;KACK,cAAc,eAAe;;EAElC;EACA;EACA;GACC;;UAEY,wBAAwB;EACrC,UAAU,iBAAiB;GAC1B,qBAAqB,iBAAiB;;UAE1B,uBAAuB;EACpC;EACA,UAAU;EACV,UAAU;EACV,QAAQ;;UAEK,wBAAwB;GACpC,eAAe,iBAAiB;;UAEpB,uBAAuB;GACnC,eAAe;;UAEH;GACZ,eAAe,eAAe;;UAElB;EACb;EACA;EACA;EACA;GACC;;UAEY;GACZ,eAAe,aAAa;;UAEhB,mBAAmB;EAChC;EACA;EACA,aAAa;EACb;EACA;EACA,SAAS;GACR;;UAEY;GACZ;;UAEY,qBAAqB;EAClC;;UAEa,kBAAkB;EAC/B;EACA;EACA,eAAe;GACd;;UAEY;GACZ,eAAe,gBAAgB;;UAEnB;EACb;EACA;EACA;;KAGQ;UACK,qBAAqB;EAClC;EACA,gBAAgB;EAChB;EACA;EACA,MAAM;EACN,eAAe;EACf;EACA;EACA;EACA,OAAO,mBAAmB;EAC1B;EACA,SAAS,eAAe;EACxB,SAAS,eAAe;EACxB,SAAS,eAAe;EACxB,MAAM,eAAe;EACrB,QAAQ,eAAe;EACvB;KACK,uBAAuB,eAAe;;EAE3C,uBAAuB,eAAe;EACtC,gBAAgB,eAAe;EAC/B;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA,eAAe,eAAe;EAC9B;EACA;;UAMa;EACb;EACA;KACK;;;UAGQ,kBAAkB;EAC/B;EACA;EACA;EACA;EACA;;KAEQ;UACK,6BAA6B;EAC1C,MAAM;EACN;EACA;EACA;EACA;EACA;EACA,QAAQ;EACR;;UAEa,yBAAyB;EACtC,WAAW;EACX,WAAW;EACX,oBAAoB;EACpB,oBAAoB;;UAEP,wBAAwB;EACrC;EACA;EACA;EACA,QAAQ;;UAEK,qBAAqB;GACjC;;UAEY;GACZ;;;;iBChRW,aAAa,yBAAyB;;;UCX5C;EACR;EACA;;UAGQ;EACR;;UAGQ;EACR;IAAS;IAAc;IAAiB;;;UAGzB;;EAEf,OAAO;EACP;;iBAGoB,YACpB,KAAK,YACL,UAAS,aACR,QAAQ,aAAa;;iBAqDF,WAAW,iBAAiB,UAAS,aAAkB;;iBAuBvD,YAAY,iBAAiB,UAAS,aAAkB;;;;iBCjB9D,iBAAiB,eAAe;;iBAKhC;;;;;;;;;;;;;UCtEC;;EAEf,SAAS;;;;;;EAMT,OAAO;;EAEP;;;;;;;;;;EAUA;IACE,mBAAmB,4BAA4B;;;iBAMnC,UAAU,OAAM,mBAAqB,0BAAA,+BAAA;UA0IpC,qBAAqB;;EAEpC;;EAEA;;iBAGc,YAAY,OAAM,eAAoB;UAUrC;EACf,QAAQ;;;;EAIR;;EAEA;;EAEA,SAAS;;;;;;;iBAQK,iBAAiB,OAAM,eAAoB,QAAQ"}
|
|
@@ -531,6 +531,10 @@ const proposer: SurfaceProposer = {
|
|
|
531
531
|
Return a label and rationale when they will help later analysis.
|
|
532
532
|
Candidate creation must not read final test results.
|
|
533
533
|
|
|
534
|
+
A proposer may also attach `attribution`: one opaque, JSON-safe record the loop carries byte-for-byte onto `GenerationCandidate.attribution` and the loop provenance record.
|
|
535
|
+
The loop never interprets it.
|
|
536
|
+
Tag it with your own schema field and validate it on readback — `PolicyEdit` candidate records (`makePolicyEditCandidateRecord` from the analyst surface) are the first producer, which is what lets a forecast in an edit be scored against the measured delta later.
|
|
537
|
+
|
|
534
538
|
`runOptimization()` rejects a candidate whose `surfaceHash` was already admitted.
|
|
535
539
|
This includes the baseline, an earlier generation, and another candidate in the same proposal.
|
|
536
540
|
The complete proposal is checked before candidate dispatch, so duplicates cannot consume candidate cells.
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@tangle-network/agent-eval",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.171.0",
|
|
4
4
|
"description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
|
|
5
5
|
"homepage": "https://github.com/tangle-network/agent-eval#readme",
|
|
6
6
|
"repository": {
|
|
@@ -197,33 +197,33 @@
|
|
|
197
197
|
},
|
|
198
198
|
"dependencies": {
|
|
199
199
|
"@asteasolutions/zod-to-openapi": "^9.1.0",
|
|
200
|
-
"@hono/node-server": "^2.
|
|
201
|
-
"@tangle-network/agent-core": "^0.9.
|
|
202
|
-
"@tangle-network/agent-interface": "^
|
|
200
|
+
"@hono/node-server": "^2.1.1",
|
|
201
|
+
"@tangle-network/agent-core": "^0.9.6",
|
|
202
|
+
"@tangle-network/agent-interface": "^2.0.0",
|
|
203
203
|
"@tangle-network/agent-trace-contract": "^1.0.2",
|
|
204
|
-
"hono": "^4.
|
|
204
|
+
"hono": "^4.13.5",
|
|
205
205
|
"linear-sum-assignment": "1.0.9",
|
|
206
206
|
"re2js": "2.8.6",
|
|
207
|
-
"zod": "
|
|
207
|
+
"zod": "4.5.4"
|
|
208
208
|
},
|
|
209
209
|
"devDependencies": {
|
|
210
210
|
"@arethetypeswrong/cli": "^0.18.5",
|
|
211
|
-
"@arizeai/phoenix-evals": "^2.
|
|
212
|
-
"@biomejs/biome": "^2.5.
|
|
213
|
-
"@types/node": "^26.
|
|
211
|
+
"@arizeai/phoenix-evals": "^2.3.0",
|
|
212
|
+
"@biomejs/biome": "^2.5.11",
|
|
213
|
+
"@types/node": "^26.4.0",
|
|
214
214
|
"autoevals": "^0.3.0",
|
|
215
|
-
"esbuild": "^0.28.
|
|
215
|
+
"esbuild": "^0.28.2",
|
|
216
216
|
"fast-check": "^4.9.0",
|
|
217
217
|
"husky": "^9.1.7",
|
|
218
218
|
"lib-r-math.js": "^3.0.2",
|
|
219
|
-
"lint-staged": "^17.
|
|
220
|
-
"openapi3-ts": "^4.6.
|
|
221
|
-
"oxc-parser": "^0.
|
|
222
|
-
"publint": "^0.3.
|
|
219
|
+
"lint-staged": "^17.4.1",
|
|
220
|
+
"openapi3-ts": "^4.6.1",
|
|
221
|
+
"oxc-parser": "^0.147.0",
|
|
222
|
+
"publint": "^0.3.24",
|
|
223
223
|
"tsdown": "^0.22.14",
|
|
224
|
-
"tsx": "^4.23.
|
|
224
|
+
"tsx": "^4.23.12",
|
|
225
225
|
"typescript": "^7.0.2",
|
|
226
|
-
"vitest": "^4.1.
|
|
226
|
+
"vitest": "^4.1.11",
|
|
227
227
|
"yaml": "2.9.0"
|
|
228
228
|
},
|
|
229
229
|
"pnpm": {
|
|
@@ -231,7 +231,8 @@
|
|
|
231
231
|
"minimumReleaseAgeExclude": [
|
|
232
232
|
"@tangle-network/*",
|
|
233
233
|
"esbuild",
|
|
234
|
-
"vite"
|
|
234
|
+
"vite",
|
|
235
|
+
"zod@4.5.4"
|
|
235
236
|
],
|
|
236
237
|
"overrides": {
|
|
237
238
|
"@arizeai/openinference-core>@opentelemetry/core": "^2.10.0",
|
|
@@ -242,7 +243,7 @@
|
|
|
242
243
|
}
|
|
243
244
|
},
|
|
244
245
|
"engines": {
|
|
245
|
-
"node": ">=20"
|
|
246
|
+
"node": ">=20.19.0"
|
|
246
247
|
},
|
|
247
248
|
"lint-staged": {
|
|
248
249
|
"src/**/*.{ts,tsx}": [
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"tool-groups-BnXlCJZQ.d.ts","names":[],"sources":["../src/analyst/tool-groups.ts"],"mappings":";;;;KAmBY;;;;;;;;;;;;;;;;;;;;iBAgDI,wBACd,OAAO,oBACP,OAAO,qBACN"}
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"types-Dy237wiH.d.ts","names":[],"sources":["../src/campaign/types.ts"],"mappings":";;;;;;;UAgCiB;EACf;EACA;EACA;;;;;EAKA;;;UAIe,iCAAiC,KAAK;EACrD;;;;;UAMe;EACf;;EAEA;EACA;EACA;EACA;EACA,QAAQ;EACR,OAAO;EACP,WAAW;EACX,MAAM;;EAEN;;EAEA;;;;;;;;EAQA;;;;KAKU,WAAW,kBAAkB,UAAU,cACjD,UAAU,WACV,KAAK,oBACF,QAAQ;;;;UAOI,cAAc,WAAW;EACxC;EACA;EACA;;EAEA;;;EAGA,sBAAsB,UAAU,WAAW,sBAAsB,UAAU,cAAc;;UAK1E;;EAEf;;EAEA;;;;;;;;;UAUe,YAAY,WAAW,kBAAkB,WAAW;EACnE;EACA,YAAY;;;;EAIZ;;;EAGA,MAAM;IACJ,UAAU;IACV,UAAU;IACV,QAAQ;;IAER,aAAa;IACb;IACA,WAAW;MACT,aAAa,QAAQ;EACzB,aAAa,UAAU;;;;;;;;;;;UAYR;EACf,YAAY;EACZ;EACA;;EAEA,UAAU;;;;;;;EAOV;;;;EAIA,eAAe,eAAe;IAAQ;IAAe;;;;;EAIrD;;;EAGA;;EAEA;;EAEA,WAAW,eAAe;;;;;;;UAUX;WACN;;;WAGA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;;WAGA;aACE;aACA;aACA;;;WAGF;;;UAIM;WACN;WACA,YAAY,SAAS;;;;;;;;;;;KAYpB,0BAA0B,mBAAmB;;;;;;;UAQxC;EACf,SAAS;;EAET;;;;EAIA;;;;iBAKc,oBACd,OAAO,iBAAiB,oBACvB,SAAS;;;;;;;;;UAkBK;EACf,SAAS;EACT;;;EAGA,YAAY;;;EAGZ;;EAEA;EACA;EACA;;;UAIe;EACf;EACA;EACA;EACA;EACA;;;;;UAMe;;;EAGf;;EAEA;EACA;EACA;EACA,YAAY;;;;;EAKZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;EAC1E;IACE;IACA;;;;;UAMa,eAAe,YAAY;;;WAGjC,gBAAgB;WAChB,SAAS,cAAc;WACvB,UAAU,cAAc;;WAExB;WACA;WACA,QAAQ;;WAER,QAAQ;;;WAGR,kBAAkB;;;WAGlB,mBAAmB;;;;WAInB,gBAAgB;;;;WAIhB;;;;;;;WAOA,gBAAgB,cAAc;;WAE9B,aAAa;WACb;;;;;;;;;;;;;;UAeM,gBAAgB,YAAY;EAC3C;;;;;EAKA,QAAQ,KAAK,eAAe,aAAa,QAAQ,MAAM,iBAAiB;;;EAGxE,QAAQ;IAAQ,SAAS,cAAc;;IAAwB;IAAe;;;UAG/D;EACf;EACA;EACA,mBAAmB,qBAAqB;;UAGzB,wBAAwB;EACvC,UAAU;;;KAMA;;KAGA;UAEK;EACf;EACA,QAAQ;EACR;;UAGe,YAAY,WAAW,kBAAkB;EACxD,oBAAoB,YAAY;EAChC,oBAAoB,YAAY;;EAEhC,aAAa,YAAY,eAAe;;;;;EAKxC,sBAAsB,YAAY,eAAe;;;;;;;EAOjD,yBAAyB,YAAY,eAAe;;;EAGpD,uBAAuB,YAAY;EACnC,WAAW;EACX;IAAQ;IAAmB;;;EAE3B,aAAa;EACb;EACA,QAAQ;;UAGO;EACf,UAAU;EACV;EACA,mBAAmB;EACnB;;;UAIe,KAAK,qBAAqB,kBAAkB,WAAW;EACtE;EACA,OAAO,KAAK,YAAY,WAAW,aAAa,QAAQ;;;;UAOzC;EACf,KAAK,cAAc,aAAa,0BAA0B;EAC1D,SAAS;;UAGM;EACf,IAAI,aAAa;EACjB,aAAa,aAAa;;;;UAKX;EACf,MAAM,cAAc,kBAAkB,aAAa;EACnD,UAAU,cAAc,iBAAiB;;;;;;KAO/B,qBAAqB;;;;;UAMhB;;EAEf,YAAY,GACV,OAAO,KAAK,iBAAiB;IAC3B,UAAU;MAEX,QAAQ,eAAe;;;;;KAQhB;KAOA;;;;;;;;;;;;;;;KAgBA;;iBASI,eAAe,OAAO;;;;UAOrB,qBAAqB,kBAAkB,WAAW,UAAU;EAC3E,UAAU;EACV,UAAU;EACV,aAAa,eAAe;EAC5B,QAAQ;EACR;EACA;EACA,iBAAiB;;;;;EAKjB,aAAa;;EAEb;;UAGe,sBAAsB,kBAAkB,WAAW,UAAU,6BACpE,qBAAqB,WAAW;;EAExC;;;EAGA;;UAGe;EACf;;EAEA;;;;EAIA;EACA;IACE;IACA,SAAS,wBAAwB;IACjC;IACA;;;;;IAKA,WAAW;;;UAIE;EACf,QAAQ,OAAO,uBAAuB;EACtC,OAAO,MAAM,4BAA4B,QAAQ;EACjD,QAAQ;IACN;IACA;IACA,UAAU;;;IAGV,SAAS,OAAO;;;UAMH,mBAAmB;;;EAGlC;EACA;EACA;EACA;EACA;EACA,UAAU;EACV,aAAa,eAAe;;;EAG5B;;EAEA,gBAAgB;;EAEhB;;;EAGA,YAAY;;;EAGZ;;;EAGA;EACA;EACA;EACA;;;;EAIA;;EAEA;;EAEA;EACA;;UAGe;EACf;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf;EACA,YAAY;EACZ;;;;;;UAOe;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;;EAIA;;;;EAIA;IACE;IACA;IACA,iBAAiB;MAAQ;MAAgB;;;;;EAI3C,YAAY;;;;;;;;;;EAUZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;;;EAG1E;;;;EAIA;;UAGe;EACf,SAAS,eAAe;EACxB,YAAY,eAAe;;EAE3B,MAAM;;EAEN;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;UAGe,eAAe,qBAAqB,kBAAkB,WAAW;;EAEhF;;EAEA;EACA;;EAEA;EACA;EACA;EACA;EACA,OAAO,MAAM,mBAAmB;EAChC,YAAY;EACZ;IACE,aAAa;IACb;;EAEF,OAAO;EACP;EACA;EACA,iBAAiB;;;EAGjB,WAAW,MAAM,2BAA2B,KAAK"}
|