agentv 4.41.4 → 4.41.6-next.1

package/dist/{chunk-QCF3LUI7.js → chunk-CGMO3QA7.js}

@@ -6156,7 +6156,7 @@ async function expandFileReferences(tests, evalFileDir) {
   return expanded;
 }
 
-// ../../packages/core/dist/chunk-7MDLZQ2R.js
+// ../../packages/core/dist/chunk-WKBRUYYT.js
 import path50 from "node:path";
 import { pathToFileURL as pathToFileURL2 } from "node:url";
 import { existsSync as existsSync7 } from "node:fs";
@@ -18820,7 +18820,7 @@ var RequestError = class _RequestError extends Error {
   }
 };
 
-// ../../packages/core/dist/chunk-7MDLZQ2R.js
+// ../../packages/core/dist/chunk-WKBRUYYT.js
 import { exec as execCallback } from "node:child_process";
 import { readdirSync, statSync } from "node:fs";
 import { readFile as readFile32, readdir as readdir2, stat as stat2 } from "node:fs/promises";
@@ -22990,15 +22990,15 @@ var SkillTriggerGrader = class {
     for (const toolCall of allToolCalls) {
       const toolName = toolCall.tool ?? "";
       const input = toolCall.input ?? {};
-      if (toolName === "Skill") {
+      if (toolName.toLowerCase() === "skill") {
         const skillArg = String(input.skill ?? "");
         if (skillArg.includes(skillName)) {
           triggered = true;
           evidence = `Skill tool invoked with skill="${skillArg}"`;
           break;
         }
-      } else if (toolName === "Read") {
-        const filePath = String(input.file_path ?? "");
+      } else if (toolName.toLowerCase() === "read") {
+        const filePath = String(input.file_path ?? input.path ?? "");
         if (filePath.includes(`skills/${skillName}/`)) {
           triggered = true;
           evidence = `Read tool loaded skill file: ${filePath}`;
@@ -38727,7 +38727,7 @@ async function loadTestSuite(evalFilePath, repoRoot, options) {
     return { tests: await loadTestsFromAgentSkills(evalFilePath) };
   }
   if (format === "typescript") {
-    const { loadTsEvalSuite: loadTsEvalSuite2 } = await import("./ts-eval-loader-UPFWPMQO-SKZNUXPB.js");
+    const { loadTsEvalSuite: loadTsEvalSuite2 } = await import("./ts-eval-loader-IVMSOI7K-JUPHBJM7.js");
     return loadTsEvalSuite2(evalFilePath, resolveToAbsolutePath(repoRoot), options);
   }
   const { tests, parsed, suiteWorkspacePath } = await loadTestsFromYaml(
@@ -38735,24 +38735,18 @@ async function loadTestSuite(evalFilePath, repoRoot, options) {
     repoRoot,
     options
   );
-  const metadata = parseMetadata(parsed);
-  const failOnError = extractFailOnError(parsed);
-  const threshold = extractThreshold(parsed);
-  return {
-    tests,
-    trials: extractTrialsConfig(parsed),
-    targets: extractTargetsFromSuite(parsed),
-    targetRefs: extractTargetRefsFromSuite(parsed),
-    workers: extractWorkersFromSuite(parsed),
-    cacheConfig: extractCacheConfig(parsed),
-    budgetUsd: extractBudgetUsd(parsed),
-    ...metadata !== void 0 && { metadata },
-    ...failOnError !== void 0 && { failOnError },
-    ...threshold !== void 0 && { threshold },
-    ...suiteWorkspacePath !== void 0 && { workspacePath: suiteWorkspacePath }
-  };
+  return buildEvalSuiteResult(parsed, tests, suiteWorkspacePath);
 }
 var loadEvalSuite = loadTestSuite;
+async function loadTestSuiteFromYamlObject(evalFilePath, suiteObject, repoRoot, options) {
+  const { tests, parsed, suiteWorkspacePath } = await loadTestsFromParsedYamlValue(
+    suiteObject,
+    evalFilePath,
+    repoRoot,
+    options
+  );
+  return buildEvalSuiteResult(parsed, tests, suiteWorkspacePath);
+}
 async function loadTests(evalFilePath, repoRoot, options) {
   const format = detectFormat(evalFilePath);
   if (format === "jsonl") {
@@ -38762,7 +38756,7 @@ async function loadTests(evalFilePath, repoRoot, options) {
     return loadTestsFromAgentSkills(evalFilePath);
   }
   if (format === "typescript") {
-    const { loadTsEvalSuite: loadTsEvalSuite2 } = await import("./ts-eval-loader-UPFWPMQO-SKZNUXPB.js");
+    const { loadTsEvalSuite: loadTsEvalSuite2 } = await import("./ts-eval-loader-IVMSOI7K-JUPHBJM7.js");
     const suite = await loadTsEvalSuite2(evalFilePath, resolveToAbsolutePath(repoRoot), options);
     return suite.tests;
   }
@@ -38771,14 +38765,17 @@ async function loadTests(evalFilePath, repoRoot, options) {
 }
 var loadEvalCases = loadTests;
 async function loadTestsFromYaml(evalFilePath, repoRoot, options) {
+  const absoluteTestPath = path46.resolve(evalFilePath);
+  const rawFile = await readFile18(absoluteTestPath, "utf8");
+  return loadTestsFromParsedYamlValue(parseYamlValue(rawFile), evalFilePath, repoRoot, options);
+}
+async function loadTestsFromParsedYamlValue(rawParsed, evalFilePath, repoRoot, options) {
   const verbose = options?.verbose ?? false;
   const filterPattern = options?.filter;
   const absoluteTestPath = path46.resolve(evalFilePath);
   const repoRootPath = resolveToAbsolutePath(repoRoot);
   const searchRoots = buildSearchRoots2(absoluteTestPath, repoRootPath);
   const config2 = await loadConfig(absoluteTestPath, repoRootPath);
-  const rawFile = await readFile18(absoluteTestPath, "utf8");
-  const rawParsed = parseYamlValue(rawFile);
   const rawCaseSnapshots = buildRawInlineTestSnapshots(rawParsed);
   const interpolated = interpolateEnv(rawParsed, process.env);
   if (!isJsonObject(interpolated)) {
@@ -38993,6 +38990,24 @@ async function loadTestsFromYaml(evalFilePath, repoRoot, options) {
   }
   return { tests: results, parsed: suite, suiteWorkspacePath: suiteWorkspace?.path };
 }
+function buildEvalSuiteResult(parsed, tests, suiteWorkspacePath) {
+  const metadata = parseMetadata(parsed);
+  const failOnError = extractFailOnError(parsed);
+  const threshold = extractThreshold(parsed);
+  return {
+    tests,
+    trials: extractTrialsConfig(parsed),
+    targets: extractTargetsFromSuite(parsed),
+    targetRefs: extractTargetRefsFromSuite(parsed),
+    workers: extractWorkersFromSuite(parsed),
+    cacheConfig: extractCacheConfig(parsed),
+    budgetUsd: extractBudgetUsd(parsed),
+    ...metadata !== void 0 && { metadata },
+    ...failOnError !== void 0 && { failOnError },
+    ...threshold !== void 0 && { threshold },
+    ...suiteWorkspacePath !== void 0 && { workspacePath: suiteWorkspacePath }
+  };
+}
 var SOURCE_SECRET_KEY_PATTERN = /(api[_-]?key|authorization|bearer|credential|password|private[_-]?key|secret|token)/i;
 var REDACTED_SOURCE_VALUE = "[redacted]";
 function buildRawInlineTestSnapshots(rawParsed) {
@@ -43629,6 +43644,8 @@ async function loadEnvHierarchy(repoRoot, startPath) {
   }
 }
 var EXPORT_NAMES = ["default", "config", "evalConfig"];
+var SDK_EVAL_SUITE_SYMBOL = Symbol.for("@agentv/sdk/eval-suite");
+var SDK_TO_EVAL_YAML_OBJECT_SYMBOL = Symbol.for("@agentv/sdk/to-eval-yaml-object");
 async function loadTsEvalFile(filePath) {
   const absolutePath = path50.resolve(filePath);
   const moduleUrl = pathToFileURL2(absolutePath).href;
@@ -43636,20 +43653,28 @@ async function loadTsEvalFile(filePath) {
   let config2;
   for (const name of EXPORT_NAMES) {
     const candidate = module[name];
-    if (isEvalConfigLike(candidate)) {
+    if (isSupportedTsEvalExport(candidate)) {
       config2 = candidate;
       break;
     }
   }
   if (!config2) {
     throw new Error(
-      `${filePath}: no EvalConfig export found. Export an EvalConfig as default, 'config', or 'evalConfig'.`
+      `${filePath}: no supported eval export found. Export defineEval(...) or an EvalConfig as default, 'config', or 'evalConfig'.`
     );
   }
   return { config: config2, filePath: absolutePath };
 }
 async function loadTsEvalSuite(filePath, repoRoot, options) {
   const { config: config2, filePath: absolutePath } = await loadTsEvalFile(filePath);
+  if (isSdkEvalSuiteExport(config2)) {
+    return loadTestSuiteFromYamlObject(
+      absolutePath,
+      config2[SDK_TO_EVAL_YAML_OBJECT_SYMBOL](),
+      repoRoot,
+      options
+    );
+  }
   const materialized = await materializeEvalConfig(config2, {
     repoRoot,
     baseDir: path50.dirname(absolutePath),
@@ -43680,6 +43705,12 @@ async function loadTsEvalSuite(filePath, repoRoot, options) {
     }
   };
 }
+function isSdkEvalSuiteExport(value) {
+  return !!value && typeof value === "object" && value[SDK_EVAL_SUITE_SYMBOL] === true && typeof value[SDK_TO_EVAL_YAML_OBJECT_SYMBOL] === "function";
+}
+function isSupportedTsEvalExport(value) {
+  return isSdkEvalSuiteExport(value) || isEvalConfigLike(value);
+}
 function isEvalConfigLike(value) {
   if (!value || typeof value !== "object") return false;
   const obj = value;
@@ -43920,6 +43951,7 @@ export {
   readTestSuiteMetadata,
   loadTestSuite,
   loadEvalSuite,
+  loadTestSuiteFromYamlObject,
   loadTests,
   loadEvalCases,
   loadTestById,
@@ -43954,4 +43986,4 @@ export {
   loadTsEvalFile,
   loadTsEvalSuite
 };
-//# sourceMappingURL=chunk-QCF3LUI7.js.map
+//# sourceMappingURL=chunk-CGMO3QA7.js.map