@agentv/core 2.12.0 → 2.13.0

package/dist/evaluation/validation/index.js

@@ -6,7 +6,7 @@ import {
   findGitRoot,
   isEvaluatorKind,
   resolveFileReference
-} from "../../chunk-7HPKTRFZ.js";
+} from "../../chunk-JHER2LQ5.js";
 
 // src/evaluation/validation/file-type.ts
 import { readFile } from "node:fs/promises";

package/dist/index.cjs

@@ -1492,6 +1492,7 @@ __export(index_exports, {
   executeWorkspaceScript: () => executeWorkspaceScript,
   explorationRatio: () => explorationRatio,
   extractCacheConfig: () => extractCacheConfig,
+  extractFailOnError: () => extractFailOnError,
   extractJsonBlob: () => extractJsonBlob,
   extractTargetFromSuite: () => extractTargetFromSuite,
   extractTargetsFromSuite: () => extractTargetsFromSuite,
@@ -2014,6 +2015,11 @@ async function loadConfig(evalFilePath, repoRoot) {
         continue;
       }
       const config = parsed;
+      const requiredVersion = parsed.required_version;
+      if (requiredVersion !== void 0 && typeof requiredVersion !== "string") {
+        logWarning(`Invalid required_version in ${configPath}, expected string`);
+        continue;
+      }
       const guidelinePatterns = config.guideline_patterns;
       if (guidelinePatterns !== void 0 && !Array.isArray(guidelinePatterns)) {
         logWarning(`Invalid guideline_patterns in ${configPath}, expected array`);
@@ -2037,6 +2043,7 @@ async function loadConfig(evalFilePath, repoRoot) {
         configPath
       );
       return {
+        required_version: requiredVersion,
         guideline_patterns: guidelinePatterns,
         eval_patterns: evalPatterns,
         execution: executionDefaults
@@ -2180,6 +2187,22 @@ function extractTotalBudgetUsd(suite) {
   );
   return void 0;
 }
+function extractFailOnError(suite) {
+  const execution = suite.execution;
+  if (!execution || typeof execution !== "object" || Array.isArray(execution)) {
+    return void 0;
+  }
+  const executionObj = execution;
+  const raw = executionObj.fail_on_error ?? executionObj.failOnError;
+  if (raw === void 0 || raw === null) {
+    return void 0;
+  }
+  if (typeof raw === "boolean") {
+    return raw;
+  }
+  logWarning(`Invalid execution.fail_on_error: ${raw}. Must be true or false. Ignoring.`);
+  return void 0;
+}
 function parseExecutionDefaults(raw, configPath) {
   if (!raw || typeof raw !== "object" || Array.isArray(raw)) {
     return void 0;
@@ -4375,13 +4398,15 @@ async function loadTestSuite(evalFilePath, repoRoot, options) {
   }
   const { tests, parsed } = await loadTestsFromYaml(evalFilePath, repoRoot, options);
   const metadata = parseMetadata(parsed);
+  const failOnError = extractFailOnError(parsed);
   return {
     tests,
     trials: extractTrialsConfig(parsed),
     targets: extractTargetsFromSuite(parsed),
     cacheConfig: extractCacheConfig(parsed),
     totalBudgetUsd: extractTotalBudgetUsd(parsed),
-    ...metadata !== void 0 && { metadata }
+    ...metadata !== void 0 && { metadata },
+    ...failOnError !== void 0 && { failOnError }
   };
 }
 var loadEvalSuite = loadTestSuite;
@@ -15780,7 +15805,8 @@ async function runEvaluation(options) {
     cleanupWorkspaces,
     trials,
     streamCallbacks,
-    totalBudgetUsd
+    totalBudgetUsd,
+    failOnError
   } = options;
   let useCache = options.useCache;
   if (trials && trials.count > 1 && useCache) {
@@ -15976,6 +16002,7 @@ async function runEvaluation(options) {
   let beforeAllOutputAttached = false;
   let cumulativeBudgetCost = 0;
   let budgetExhausted = false;
+  let failOnErrorTriggered = false;
   const promises = filteredEvalCases.map(
     (evalCase) => limit(async () => {
       const workerId = nextWorkerId++;
@@ -16014,6 +16041,37 @@ async function runEvaluation(options) {
         }
         return budgetResult;
       }
+      if (failOnError === true && failOnErrorTriggered) {
+        const errorMsg = "Halted: execution error encountered with fail_on_error enabled";
+        const haltResult = {
+          timestamp: (now ?? (() => /* @__PURE__ */ new Date()))().toISOString(),
+          testId: evalCase.id,
+          dataset: evalCase.dataset,
+          score: 0,
+          hits: [],
+          misses: [],
+          answer: "",
+          target: target.name,
+          error: errorMsg,
+          executionStatus: "execution_error",
+          failureStage: "setup",
+          failureReasonCode: "error_threshold_exceeded",
+          executionError: { message: errorMsg, stage: "setup" }
+        };
+        if (onProgress) {
+          await onProgress({
+            workerId,
+            testId: evalCase.id,
+            status: "failed",
+            completedAt: Date.now(),
+            error: haltResult.error
+          });
+        }
+        if (onResult) {
+          await onResult(haltResult);
+        }
+        return haltResult;
+      }
       if (onProgress) {
         await onProgress({
           workerId,
@@ -16066,6 +16124,9 @@ async function runEvaluation(options) {
             }
           }
         }
+        if (failOnError === true && result.executionStatus === "execution_error") {
+          failOnErrorTriggered = true;
+        }
         if (beforeAllOutput && !beforeAllOutputAttached) {
           result = { ...result, beforeAllOutput };
           beforeAllOutputAttached = true;
@@ -18132,6 +18193,7 @@ function createAgentKernel() {
   executeWorkspaceScript,
   explorationRatio,
   extractCacheConfig,
+  extractFailOnError,
   extractJsonBlob,
   extractTargetFromSuite,
   extractTargetsFromSuite,