@tangle-network/agent-eval 0.120.3 → 0.120.4

package/dist/{chunk-P5MGQ2FY.js → chunk-ZTJPIIVI.js}

@@ -16,7 +16,7 @@ import {
   runImprovementLoop,
   surfaceContentHash,
   surfaceHash
-} from "./chunk-3CDFMEMO.js";
+} from "./chunk-7QBFOJ3E.js";
 import {
   SearchLedgerConflictError,
   SearchLedgerError,
@@ -7955,4 +7955,4 @@ export {
   verifyCodeSurface,
   resolveWorktreePath
 };
-//# sourceMappingURL=chunk-P5MGQ2FY.js.map
+//# sourceMappingURL=chunk-ZTJPIIVI.js.map

package/dist/contract/index.d.ts

@@ -2226,9 +2226,24 @@ declare function openAutoPr<TArtifact, TScenario extends Scenario>(options: Open
  * re-score + release gate + optional PR.
  */
 
+interface PremeasuredOptimizationBaseline<TArtifact, TScenario extends Scenario> {
+    /** Hash of the exact surface that produced `campaign`. */
+    surfaceHash: string;
+    /** Complete prior measurement reused by identity, including artifactsByPath. */
+    campaign: CampaignResult<TArtifact, TScenario>;
+}
 interface RunOptimizationBaseOptions<TScenario extends Scenario, TArtifact> extends Omit<RunCampaignOptions<TScenario, TArtifact>, 'dispatch'> {
     /** Initial mutable surface (typically system prompt or addendum). */
     baselineSurface: MutableSurface;
+    /**
+     * Complete prior measurement of `baselineSurface`. When present,
+     * `runOptimization` validates its surface, scenario split, seed, reps, and
+     * normal campaign coverage, then skips the baseline campaign entirely — no
+     * dispatch or resumability-cache lookup. Candidate campaigns still run
+     * normally. Prior spend remains in the imported campaign aggregates and is
+     * not added again to this continuation's CostLedger.
+     */
+    premeasuredBaseline?: PremeasuredOptimizationBaseline<TArtifact, TScenario>;
     /** Dispatcher that takes the CURRENT surface + scenario → artifact. */
     dispatchWithSurface: (surface: MutableSurface, scenario: TScenario, ctx: Parameters<RunCampaignOptions<TScenario, TArtifact>['dispatch']>[1]) => Promise<TArtifact>;
     /** The candidate-generation strategy. Wrap a population `Mutator` via

package/dist/contract/index.js

@@ -41,7 +41,7 @@ import {
   surfaceContentHash,
   surfaceHash,
   verifyLoopProvenanceRecord
-} from "../chunk-3CDFMEMO.js";
+} from "../chunk-7QBFOJ3E.js";
 import {
   assertCampaignSplitIdentity,
   assertRealAgentReceipts,

package/dist/index.js

@@ -51,7 +51,7 @@ import {
   harnessAxisOf,
   parseCorrectnessResponse,
   verifyCompletion
-} from "./chunk-P5MGQ2FY.js";
+} from "./chunk-ZTJPIIVI.js";
 import {
   DEFAULT_MUTATION_PRIMITIVES,
   DEFAULT_RED_TEAM_CORPUS,
@@ -91,7 +91,7 @@ import {
   scoreRedTeamOutput,
   surfaceContentHash,
   toolNamesForRun
-} from "./chunk-3CDFMEMO.js";
+} from "./chunk-7QBFOJ3E.js";
 import {
   BackendIntegrityError,
   assertRealAgentReceipts,

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.120.3",
+    "version": "0.120.4",
     "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/docs/eval-surface-map.md

@@ -12,7 +12,7 @@ composition point already exists (see Produced-state grading below).**
 | `runCampaign` | The measurement primitive. Run a dispatch over scenarios × seeds × reps, score each with judges, aggregate. Caller owns the dispatch. | `CampaignResult` |
 | `runEval` | The simplest preset over `runCampaign` — just score, no loop, no gate. The 80% "I want a scorecard" case. | `CampaignResult` |
 | `runProfileMatrix` | Factor the SAME scenarios across N agent **profiles** (models / prompt variants), with RunRecord stamping + a real-backend integrity guard. | `RunRecord[]` |
-| `runOptimization` | GENERATE: baseline → N generations of propose → measure → rank → promote. No release gate. | generations + winner |
+| `runOptimization` | GENERATE: measured or validated premeasured baseline → N generations of propose → measure → rank → promote. No release gate. | generations + winner |
 | `runImprovementLoop` | The release-gate shell around `runOptimization`: adds a held-out re-score + a promotion gate (+ optional auto-PR). | gate decision + winner |
 | `runEvalCampaign` | Inversion-of-control variant of `runCampaign` — the runner is handed a pre-wired trace/sink/emitter and integrity gating as a precondition. Use when you need full capture by construction. | `CampaignResult` + records |
 | `runAgentMatrix` | The bare N-axis cartesian scheduler with concurrency control. The layer beneath the eval surface — reach for it only when you need raw scheduling, not eval semantics. | cell results |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.120.3",
+  "version": "0.120.4",
   "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": {