@tangle-network/agent-eval 0.123.1 → 0.123.3

package/dist/multishot/index.d.ts

@@ -206,8 +206,8 @@ interface LlmCallResult {
     content: string;
     usage: LlmUsage;
     /**
-     * Cost in USD. Pulled from proxy's `_response_cost` field when present;
-     * `null` when neither the proxy nor the caller can derive it.
+     * Cost in USD. Uses the provider's reported cost when present, otherwise
+     * caller-supplied token pricing. `null` when neither is available.
      */
     costUsd: number | null;
     /** Model name actually used (echoed from response). */

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.123.1",
+    "version": "0.123.3",
     "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/rl.d.ts

@@ -1,7 +1,8 @@
 type AgentProfileCellSchemaVersion = 'agent-profile-cell/v1';
-type AgentProfileJson = string | number | boolean | null | AgentProfileJson[] | {
+type AgentProfileJsonObject = {
     [key: string]: AgentProfileJson;
 };
+type AgentProfileJson = string | number | boolean | null | AgentProfileJson[] | AgentProfileJsonObject;
 type AgentProfileDimensionValue = string | number | boolean | null;
 interface AgentProfileSource {
     /** Runtime/profile contract being fingerprinted, e.g. `agent-interface-profile`. */
@@ -2407,6 +2408,11 @@ interface DetectRewardHackingInput {
 declare function detectRewardHacking(input: DetectRewardHackingInput): RewardHackingReport;
 
 type CostChannel = 'agent' | 'judge' | 'verifier' | 'analyst' | 'driver' | (string & {});
+/** Per-million token rates for a model or endpoint not covered by package pricing. */
+interface CustomTokenPricing {
+    inputUsdPerMillion: number;
+    outputUsdPerMillion: number;
+}
 interface ChannelRollup {
     channel: CostChannel;
     calls: number;
@@ -2550,8 +2556,8 @@ interface LlmCallResult {
     content: string;
     usage: LlmUsage;
     /**
-     * Cost in USD. Pulled from proxy's `_response_cost` field when present;
-     * `null` when neither the proxy nor the caller can derive it.
+     * Cost in USD. Uses the provider's reported cost when present, otherwise
+     * caller-supplied token pricing. `null` when neither is available.
      */
     costUsd: number | null;
     /** Model name actually used (echoed from response). */
@@ -2610,6 +2616,21 @@ interface LlmClientOptions {
     deadlineMs?: number;
     /** Total provider attempts. Legacy option name; default 3 (1 initial + 2 retries). */
     maxRetries?: number;
+    /** Token rates used when the provider omits cost or package pricing does not cover the model. */
+    customTokenPricing?: CustomTokenPricing;
+    /**
+     * Transport for requests that declare `jsonSchema`. `native` sends
+     * `response_format: json_schema`; `json-object` sends the broadly supported
+     * JSON mode and relies on the caller to include the schema in model-visible
+     * instructions. Default: `native`.
+     */
+    jsonSchemaTransport?: 'native' | 'json-object';
+    /**
+     * JSON payload parsing policy. `extract` accepts fenced or prose-prefixed JSON.
+     * `exact` requires the complete response content to be one JSON value.
+     * Default: `extract`.
+     */
+    jsonPayloadMode?: 'extract' | 'exact';
     /** Fetch implementation — defaults to global `fetch`. Override for custom transport (e.g. tests). */
     fetch?: typeof fetch;
     /**

package/dist/rl.js

@@ -10,14 +10,14 @@ import {
 } from "./chunk-3RF76KTD.js";
 import {
   runEvalCampaign
-} from "./chunk-U5CHZ5M3.js";
+} from "./chunk-N7ZS6PEF.js";
 import {
   detectRewardHacking,
   extractVerifiableReward,
   extractVerifiableRewardsFromRecords,
   filterDeterministicallyRewarded
 } from "./chunk-ARU2PZFM.js";
-import "./chunk-NJC7U437.js";
+import "./chunk-MK3Q6GCL.js";
 import {
   rubricPredictiveValidity
 } from "./chunk-X4UCIOTZ.js";
@@ -34,12 +34,14 @@ import {
   thompsonCurriculum,
   varianceBasedCurriculum
 } from "./chunk-VZSRQ272.js";
+import "./chunk-GRCDRKII.js";
+import "./chunk-VI2UW6B6.js";
 import "./chunk-TT4KNT67.js";
 import "./chunk-PC4UYEBM.js";
 import "./chunk-VQMK5FMP.js";
-import "./chunk-S3UZOQ5Y.js";
+import "./chunk-LOW3U7JZ.js";
 import "./chunk-MA6HLL3S.js";
-import "./chunk-XJYR7XFV.js";
+import "./chunk-GC4ATIKK.js";
 import "./chunk-VSMTAMNK.js";
 import {
   ValidationError