@tangle-network/agent-eval 0.198.0 → 0.199.0

package/CHANGELOG.md

@@ -6,6 +6,26 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ## Unreleased
 
+## [0.199.0] — 2026-09-27
+
+### Added
+
+- `incumbent({ patience, minImprovement })` (`/campaign`): a node takes the hill climb's lead only when its mean beats the leader's by more than `minImprovement`, in the metric's units (default 0). The policy name records it, for example `incumbent(patience=2,minImprovement=1)`, so a resumed search with another margin is refused. Runtime's pursuit version chain runs on the kernel with it.
+
+### Fixed
+
+- A leader with no scored unit kept the lead forever: a root whose every cell ended unscored could not be displaced, so the search expanded it until patience and selected nothing. The first node that dodged no unit now takes the lead from it, under every built-in policy.
+
+## [0.198.1] — 2026-09-27
+
+### Fixed
+
+- `operatorYield` counted a node once per incoming edge instead of once; on a ledger with re-proposals, invalid nodes and nodes sharing only one unit with the root inflated an operator’s measured-outcome count. Those nodes now give no sample.
+- `taskMatrix` ignored whether the objective is maximized or minimized, and reported a specialist gain from as little as one shared unit while comparing nodes measured on different units. It now reports ‘insufficient’ below 2 nodes or 6 units, and sorts cluster ids with the same code-unit order every other lens uses (not `localeCompare`, which is locale-dependent).
+- `front` could place an invalid node on the frontier, and measured its cost axis by how far the allocator had run a node rather than by known spend. It now excludes invalid nodes and uses known cost per attempted cell.
+- `tree` showed a node with only unknown-cost cells as `$0.00` instead of the known spend plus the minimum over those cells.
+- `summarizeSamples` kept its own copy of the sample-size thresholds instead of using `searchEstimateMethod`, so it called a single sample ‘insufficient’ and printed a root’s yield against itself as `0.0000/$` instead of `unknown`.
+
 ## [0.198.0] — 2026-09-27
 
 ### Added

package/README.md

@@ -96,7 +96,7 @@ The [existing-agent example](./examples/foreign-agent-quickstart/) shows how to
 | Register evidence and decision rules | [`defineEvaluationClaim()` and `sealExperiment()`](./docs/evaluation-integrity.md) from `/experiment` | A declared population, independent unit, optional practical effect, and sealed rules. |
 | Check the evaluator | [`auditEvaluator()`](./docs/evaluation-integrity.md) and [calibration tools](./docs/outcome-validity.md) from `/meta-eval` | Error rates, admission evidence, bias diagnostics, and outcome associations. |
 | Analyze completed work | [`analyzeRuns()`](./examples/analyze-existing-runs/) from `/contract`; [trace analysts](./docs/trace-analysis.md) from `/analyst` | Comparisons and findings with links to recorded evidence. |
-| Gate on the path a run took | A [trace contract](./docs/trace-contracts.md) (`compileTraceContractSpec()` from the root), or `traces check` in CI (unreleased — `/traces` `feat/traces-check`, pending its redaction-core migration) | `pass`, `fail`, or `error` per rule, with the spans that broke it. |
+| Gate on the path a run took | A [trace contract](./docs/trace-contracts.md) (`compileTraceContractSpec()` from the root), or `traces check` in CI | `pass`, `fail`, or `error` per rule, with the spans that broke it. |
 
 `defineAgentEval()` also exposes `improve()` when the same agent, cases, judge, and baseline should share configuration.
 Use direct [campaign controls](./docs/eval-surface-map.md) for scheduling, durable caches, model matrices, or custom release rules.
@@ -133,6 +133,7 @@ The [historical evidence audit](./docs/design/self-improvement-evidence-audit.md
 
 Set `searchHistoryPolicy: 'require-complete'` when a method's search must be closed, with every cell, operation and node accounted for, before final evidence is exposed.
 The [search ledger](./docs/search-ledger.md) records every search as nodes, edges and cells, and its receipt is the bounded proof.
+Eight [lenses](./docs/search-ledger.md#all-eight-lenses-one-page) read that ledger and each expose one named signal a `SearchPolicy` can read — `operatorYield` for operator choice, `landscape` for a plateau-triggered draft, `skillManifold` for adaptive unit selection, `metaSearch` for tuning the climber itself — so `agent-eval search show <ledger> --operator-yield` prints the same numbers a policy or Intelligence would use.
 
 A `gateDecision` is `ship`, `hold`, `need_more_work`, `model_ceiling`, or `arch_ceiling`.
 Gate contributions distinguish missing evidence from measured failures and successful checks.
@@ -180,6 +181,8 @@ The [benchmark-book review](./docs/design/mlbenchmarks-book-review.md) records t
 
 ## Development
 
+Read [CLAUDE.md](./CLAUDE.md) (or [AGENTS.md](./AGENTS.md)) first: dependency boundaries, ledger canonicalization, and local conventions apply to every change.
+
 ```sh
 pnpm install
 pnpm build

package/dist/adapters/http.d.ts

@@ -1,6 +1,6 @@
 import { m as CostReceipt } from "../cost-ledger-DbQdN3nO.js";
 import { R as Scenario, d as DispatchContext, f as DispatchFn } from "../types-EBUav9Bu.js";
-import "../index-D-tI1vgf.js";
+import "../index-J8BB_NYv.js";
 //#region src/adapters/http.d.ts
 interface HttpDispatchOptions<TScenario extends Scenario, _TArtifact> {
   /** Static endpoint URL. Mutually exclusive with `resolveUrl`. */

package/dist/analyst/index.d.ts

@@ -1181,7 +1181,7 @@ declare function runAnalystBenchmarkCommand(argv: readonly string[], env?: NodeJ
 declare const ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM = "sha256-canonical-source-manifest";
 declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM = "sha256-canonical-file-manifest";
 declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES: readonly string[];
-declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "6a4ac13d561bde646277270c5861b6f1021ae91d582d6eff2047af6e52f2dce9";
+declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "ff82a54b5a6a00b28b428d38b991bbda1eb00277791567c6e63dde7fdc87330c";
 declare const ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 = "1e03f2daed356d60316aabefb407ec1e437ac94d408d61eea4ae096e9c6fbb5b";
 declare const ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 = "4dba263b6256a30d56c7fdb2d992d3a953c0035d731f359b704db806f68f75ac";
 declare const ANALYST_BENCHMARK_IMPLEMENTATION_FILES: readonly string[];

package/dist/analyst/index.js

@@ -8,7 +8,7 @@ import { t as createDspyRlmTraceEngine } from "../dspy-rlm-engine-B5T43BkZ.js";
 import { a as diffFindings, i as defaultIsMaterial, n as runSemanticConceptJudge, o as createChatTraceEngine, r as FindingsStore, t as SEMANTIC_CONCEPT_JUDGE_VERSION } from "../semantic-concept-judge-2SE_kXGF.js";
 import { t as createChatClient } from "../chat-client-CJORSG_J.js";
 import { a as scoreAnalystFindings, i as summarizeAnalystBenchmarkRunner, n as runAnalystBenchmark, r as traceStoreEvidenceResolver, t as registryBenchmarkRunner } from "../benchmark-BeQJyrRC.js";
-import { $ as ANALYST_BENCHMARK_MANIFEST_FILE, A as readAnalystInstructionsOverride, B as parseVerificationOutcome, C as emptyPublicBenchmarkRunner, D as summarizeCodeTraceCalibration, E as renderCodeTraceCalibrationMarkdown, F as publicBenchmarkRlmInstructions, G as ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, H as ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, I as publicBenchmarkSystemPrompt, J as ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, K as ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, L as DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, M as MAX_INCORRECT_BLOCKS, N as MAX_INCORRECT_BLOCK_STEPS, O as analystInstructionsOverrideFromText, P as publicBenchmarkProtocolSha256, Q as ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, R as appendVerificationArtifactsToOtlp, S as adaptPublicBenchmarkFindings, T as readAnalystBenchmarkArtifact, U as ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, V as ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, W as ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, X as analystBenchmarkImplementationDigest, Y as analystBenchmarkDependencyLockDigest, Z as ANALYST_BENCHMARK_COST_LEDGER_FILE, _ as publicDirectAnalystDefinition, a as primeCodeTraceAnalystDefinition, at as codeTraceBenchCase, b as analystDefinitionAsymmetries, c as preparePublicAnalystBenchmark, ct as agentRxPredictionsToFindings, d as selectPublicBenchmarkRows, dt as normalizeBenchmarkLabel, et as ANALYST_BENCHMARK_OBSERVATIONS_FILE, f as createPublicBenchmarkRlmRunner, g as createPublicBenchmarkDirectRunner, h as runReplVariableAnalystDefinition, i as primeAnalystProtocolSha256, it as summarizeAgentRxCalibration, j as CODE_TRACE_BENCH_ANALYST_PROMPT, k as effectiveAnalystProtocolSha256, l as publicBenchmarkDistributions, lt as normalizeAgentRxCategory, m as rlmEngineLimits, n as renderAnalystBenchmarkMarkdown, nt as AGENT_RX_UPSTREAM_REVISION, o as runInlineAnalystDefinition, ot as codeTracerPredictionsToFindings, p as publicRlmAnalystDefinition, q as ANALYST_BENCHMARK_IMPLEMENTATION_FILES, r as createPrimeBenchmarkRunner, rt as renderAgentRxCalibrationMarkdown, s as loadPublicBenchmarkRows, st as agentRxBenchmarkCase, t as runAnalystBenchmarkCommand, tt as compareAnalystRunners, u as publicBenchmarkSelectionReport, ut as roundAgentRxStep, v as runChunkedAnalystDefinition, w as expandCodeTraceFailureBlocks, x as analystDefinitionProtocolSha256, y as AnalystExpressivenessError, z as loadCodeTraceVerificationArtifacts } from "../benchmark-command-D8E5TaPC.js";
+import { $ as ANALYST_BENCHMARK_MANIFEST_FILE, A as readAnalystInstructionsOverride, B as parseVerificationOutcome, C as emptyPublicBenchmarkRunner, D as summarizeCodeTraceCalibration, E as renderCodeTraceCalibrationMarkdown, F as publicBenchmarkRlmInstructions, G as ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, H as ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, I as publicBenchmarkSystemPrompt, J as ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, K as ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, L as DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, M as MAX_INCORRECT_BLOCKS, N as MAX_INCORRECT_BLOCK_STEPS, O as analystInstructionsOverrideFromText, P as publicBenchmarkProtocolSha256, Q as ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, R as appendVerificationArtifactsToOtlp, S as adaptPublicBenchmarkFindings, T as readAnalystBenchmarkArtifact, U as ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, V as ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, W as ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, X as analystBenchmarkImplementationDigest, Y as analystBenchmarkDependencyLockDigest, Z as ANALYST_BENCHMARK_COST_LEDGER_FILE, _ as publicDirectAnalystDefinition, a as primeCodeTraceAnalystDefinition, at as codeTraceBenchCase, b as analystDefinitionAsymmetries, c as preparePublicAnalystBenchmark, ct as agentRxPredictionsToFindings, d as selectPublicBenchmarkRows, dt as normalizeBenchmarkLabel, et as ANALYST_BENCHMARK_OBSERVATIONS_FILE, f as createPublicBenchmarkRlmRunner, g as createPublicBenchmarkDirectRunner, h as runReplVariableAnalystDefinition, i as primeAnalystProtocolSha256, it as summarizeAgentRxCalibration, j as CODE_TRACE_BENCH_ANALYST_PROMPT, k as effectiveAnalystProtocolSha256, l as publicBenchmarkDistributions, lt as normalizeAgentRxCategory, m as rlmEngineLimits, n as renderAnalystBenchmarkMarkdown, nt as AGENT_RX_UPSTREAM_REVISION, o as runInlineAnalystDefinition, ot as codeTracerPredictionsToFindings, p as publicRlmAnalystDefinition, q as ANALYST_BENCHMARK_IMPLEMENTATION_FILES, r as createPrimeBenchmarkRunner, rt as renderAgentRxCalibrationMarkdown, s as loadPublicBenchmarkRows, st as agentRxBenchmarkCase, t as runAnalystBenchmarkCommand, tt as compareAnalystRunners, u as publicBenchmarkSelectionReport, ut as roundAgentRxStep, v as runChunkedAnalystDefinition, w as expandCodeTraceFailureBlocks, x as analystDefinitionProtocolSha256, y as AnalystExpressivenessError, z as loadCodeTraceVerificationArtifacts } from "../benchmark-command-CoBwYvau.js";
 import { a as extractPrimeJsonObject, c as primeProtocolSha256, d as runPrimeExchange, f as decodeReplyRows, i as emptyPrimeRawUsage, l as primeReplyDefect, n as buildPrimePrompt, o as mergePrimeRawUsage, r as buildPrimeRepairPrompt, s as normalizePrimeUsage, t as analystUsageReceiptFromPrimeUsage, u as projectPrimeTrajectory } from "../prime-protocol-DcAK9pC8.js";
 import { t as nodeHttpPrimeBridgeTransport } from "../prime-bridge-transport-Z3cpKGoV.js";
 import { existsSync, readFileSync, readdirSync, statSync } from "node:fs";

package/dist/{benchmark-command-D8E5TaPC.js → benchmark-command-CoBwYvau.js}

@@ -1416,7 +1416,7 @@ const ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES = Object.freeze([
 	"package.json",
 	"pnpm-lock.yaml"
 ]);
-const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "6a4ac13d561bde646277270c5861b6f1021ae91d582d6eff2047af6e52f2dce9";
+const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "ff82a54b5a6a00b28b428d38b991bbda1eb00277791567c6e63dde7fdc87330c";
 const ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 = "1e03f2daed356d60316aabefb407ec1e437ac94d408d61eea4ae096e9c6fbb5b";
 const ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 = "4dba263b6256a30d56c7fdb2d992d3a953c0035d731f359b704db806f68f75ac";
 const ANALYST_BENCHMARK_IMPLEMENTATION_FILES = Object.freeze([
@@ -6317,4 +6317,4 @@ function shellQuote(value) {
 //#endregion
 export { ANALYST_BENCHMARK_MANIFEST_FILE as $, readAnalystInstructionsOverride as A, parseVerificationOutcome as B, emptyPublicBenchmarkRunner as C, summarizeCodeTraceCalibration as D, renderCodeTraceCalibrationMarkdown as E, publicBenchmarkRlmInstructions as F, ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 as G, ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES as H, publicBenchmarkSystemPrompt as I, ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 as J, ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM as K, DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES as L, MAX_INCORRECT_BLOCKS as M, MAX_INCORRECT_BLOCK_STEPS as N, analystInstructionsOverrideFromText as O, publicBenchmarkProtocolSha256 as P, ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE as Q, appendVerificationArtifactsToOtlp as R, adaptPublicBenchmarkFindings as S, readAnalystBenchmarkArtifact as T, ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 as U, ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM as V, ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 as W, analystBenchmarkImplementationDigest as X, analystBenchmarkDependencyLockDigest as Y, ANALYST_BENCHMARK_COST_LEDGER_FILE as Z, publicDirectAnalystDefinition as _, primeCodeTraceAnalystDefinition as a, codeTraceBenchCase as at, analystDefinitionAsymmetries as b, preparePublicAnalystBenchmark as c, agentRxPredictionsToFindings as ct, selectPublicBenchmarkRows as d, normalizeBenchmarkLabel as dt, ANALYST_BENCHMARK_OBSERVATIONS_FILE as et, createPublicBenchmarkRlmRunner as f, createPublicBenchmarkDirectRunner as g, runReplVariableAnalystDefinition as h, primeAnalystProtocolSha256 as i, summarizeAgentRxCalibration as it, CODE_TRACE_BENCH_ANALYST_PROMPT as j, effectiveAnalystProtocolSha256 as k, publicBenchmarkDistributions as l, normalizeAgentRxCategory as lt, rlmEngineLimits as m, renderAnalystBenchmarkMarkdown as n, AGENT_RX_UPSTREAM_REVISION as nt, runInlineAnalystDefinition as o, codeTracerPredictionsToFindings as ot, publicRlmAnalystDefinition as p, ANALYST_BENCHMARK_IMPLEMENTATION_FILES as q, createPrimeBenchmarkRunner as r, renderAgentRxCalibrationMarkdown as rt, loadPublicBenchmarkRows as s, agentRxBenchmarkCase as st, runAnalystBenchmarkCommand as t, compareAnalystRunners as tt, publicBenchmarkSelectionReport as u, roundAgentRxStep as ut, runChunkedAnalystDefinition as v, expandCodeTraceFailureBlocks as w, analystDefinitionProtocolSha256 as x, AnalystExpressivenessError as y, loadCodeTraceVerificationArtifacts as z };
 
-//# sourceMappingURL=benchmark-command-D8E5TaPC.js.map
+//# sourceMappingURL=benchmark-command-CoBwYvau.js.map