@tangle-network/agent-eval 0.122.1 → 0.122.3

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.122.1",
+    "version": "0.122.3",
     "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/{run-campaign-HNFPJET4.js → run-campaign-2YJ3Z6JS.js}

@@ -1,9 +1,9 @@
 import {
   planCampaignRun,
   runCampaign
-} from "./chunk-XDIRG3TO.js";
+} from "./chunk-N3QPYIVG.js";
 import "./chunk-PJQFMIOX.js";
-import "./chunk-JHOJHHU7.js";
+import "./chunk-ZMXDQ4K7.js";
 import "./chunk-VI2UW6B6.js";
 import "./chunk-ONWEPEDO.js";
 import "./chunk-PZ5AY32C.js";
@@ -11,4 +11,4 @@ export {
   planCampaignRun,
   runCampaign
 };
-//# sourceMappingURL=run-campaign-HNFPJET4.js.map
+//# sourceMappingURL=run-campaign-2YJ3Z6JS.js.map

package/dist/wire/index.js

@@ -34,9 +34,9 @@ import {
   runRpcOnce,
   startServer,
   startServerAsync
-} from "../chunk-4I2E3LLO.js";
+} from "../chunk-NMN4WGSJ.js";
 import "../chunk-NJC7U437.js";
-import "../chunk-JHOJHHU7.js";
+import "../chunk-ZMXDQ4K7.js";
 import "../chunk-VI2UW6B6.js";
 import "../chunk-PC4UYEBM.js";
 import "../chunk-ONWEPEDO.js";

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": "0.122.1",
+  "version": "0.122.3",
   "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": {

package/dist/chunk-3YYRZDON.js

@@ -1,45 +0,0 @@
-// src/concurrency.ts
-var Mutex = class {
-  locked = false;
-  waiters = [];
-  async acquire() {
-    if (!this.locked) {
-      this.locked = true;
-      return () => this.release();
-    }
-    return new Promise((resolve) => {
-      this.waiters.push(() => {
-        resolve(() => this.release());
-      });
-    });
-  }
-  release() {
-    const next = this.waiters.shift();
-    if (next) {
-      next();
-    } else {
-      this.locked = false;
-    }
-  }
-  async runExclusive(fn) {
-    const release = await this.acquire();
-    try {
-      return await fn();
-    } finally {
-      release();
-    }
-  }
-  /** True iff someone holds the lock right now. Diagnostics only. */
-  get isLocked() {
-    return this.locked;
-  }
-  /** Pending waiter count. Diagnostics only. */
-  get pending() {
-    return this.waiters.length;
-  }
-};
-
-export {
-  Mutex
-};
-//# sourceMappingURL=chunk-3YYRZDON.js.map

package/dist/chunk-3YYRZDON.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/concurrency.ts"],"sourcesContent":["/**\n * concurrency — small primitives the evolution loop needs.\n *\n * `Mutex` is a zero-dep async lock with FIFO fairness. The evolution loop\n * uses it to serialise checkout/build/commit sequences inside a single\n * pool slot, and to gate concurrent JSONL writers (see\n * `lockedJsonlReferenceReplayStore`).\n *\n * Deliberately minimal — no priority queue, no timeouts. If you need\n * those, swap to `async-mutex` at the call site.\n */\n\nexport class Mutex {\n  private locked = false\n  private readonly waiters: Array<() => void> = []\n\n  async acquire(): Promise<() => void> {\n    if (!this.locked) {\n      this.locked = true\n      return () => this.release()\n    }\n    return new Promise<() => void>((resolve) => {\n      this.waiters.push(() => {\n        resolve(() => this.release())\n      })\n    })\n  }\n\n  private release(): void {\n    const next = this.waiters.shift()\n    if (next) {\n      next()\n    } else {\n      this.locked = false\n    }\n  }\n\n  async runExclusive<T>(fn: () => Promise<T> | T): Promise<T> {\n    const release = await this.acquire()\n    try {\n      return await fn()\n    } finally {\n      release()\n    }\n  }\n\n  /** True iff someone holds the lock right now. Diagnostics only. */\n  get isLocked(): boolean {\n    return this.locked\n  }\n\n  /** Pending waiter count. Diagnostics only. */\n  get pending(): number {\n    return this.waiters.length\n  }\n}\n"],"mappings":";AAYO,IAAM,QAAN,MAAY;AAAA,EACT,SAAS;AAAA,EACA,UAA6B,CAAC;AAAA,EAE/C,MAAM,UAA+B;AACnC,QAAI,CAAC,KAAK,QAAQ;AAChB,WAAK,SAAS;AACd,aAAO,MAAM,KAAK,QAAQ;AAAA,IAC5B;AACA,WAAO,IAAI,QAAoB,CAAC,YAAY;AAC1C,WAAK,QAAQ,KAAK,MAAM;AACtB,gBAAQ,MAAM,KAAK,QAAQ,CAAC;AAAA,MAC9B,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AAAA,EAEQ,UAAgB;AACtB,UAAM,OAAO,KAAK,QAAQ,MAAM;AAChC,QAAI,MAAM;AACR,WAAK;AAAA,IACP,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,aAAgB,IAAsC;AAC1D,UAAM,UAAU,MAAM,KAAK,QAAQ;AACnC,QAAI;AACF,aAAO,MAAM,GAAG;AAAA,IAClB,UAAE;AACA,cAAQ;AAAA,IACV;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,UAAkB;AACpB,WAAO,KAAK,QAAQ;AAAA,EACtB;AACF;","names":[]}