agentv 3.7.0 → 3.9.0

package/dist/{chunk-TR6H437M.js → chunk-Q2YWV4QM.js}

@@ -16,7 +16,7 @@ import {
   validateEvalFile,
   validateFileReferences,
   validateTargetsFile
-} from "./chunk-7YS6YNJZ.js";
+} from "./chunk-GC5P5HHZ.js";
 import {
   createBuiltinRegistry,
   createProvider,
@@ -34,7 +34,7 @@ import {
   toSnakeCaseDeep as toSnakeCaseDeep2,
   transpileEvalYamlFile,
   trimBaselineResult
-} from "./chunk-XGG64VIY.js";
+} from "./chunk-TXDPYXHY.js";
 import {
   __commonJS,
   __esm,
@@ -3714,7 +3714,6 @@ async function getPromptEvalInput(evalPath, testId) {
   return {
     test_id: evalCase.id,
     input: resolveMessages(evalCase.input, fileMap),
-    guideline_paths: evalCase.guideline_paths,
     criteria: evalCase.criteria
   };
 }
@@ -3739,9 +3738,8 @@ async function getPromptEvalGradingBrief(evalPath, testId) {
   if (inputText) {
     lines.push(`Input: "${inputText}"`);
   }
-  const filePaths = evalCase.file_paths.filter((p) => !evalCase.guideline_paths.includes(p));
-  if (filePaths.length > 0) {
-    lines.push(`Files: ${filePaths.join(", ")}`);
+  if (evalCase.file_paths.length > 0) {
+    lines.push(`Files: ${evalCase.file_paths.join(", ")}`);
   }
   if (evalCase.reference_answer) {
     lines.push(`Expected: "${evalCase.reference_answer}"`);
@@ -3973,7 +3971,6 @@ var evalAssertCommand = command({
         criteria: "",
         expected_output: [],
         reference_answer: "",
-        guideline_files: [],
         input_files: [],
         trace: null,
         token_usage: null,
@@ -4180,11 +4177,16 @@ var evalRunCommand = command({
       type: optional(string),
       long: "model",
       description: 'Override model for the grader target (e.g., "openai:gpt-5-mini")'
+    }),
+    outputMessages: option({
+      type: optional(string),
+      long: "output-messages",
+      description: 'Number of trailing messages to include in results output (default: 1, or "all")'
     })
   },
   handler: async (args) => {
     if (args.evalPaths.length === 0 && process.stdin.isTTY) {
-      const { launchInteractiveWizard } = await import("./interactive-F6XECJ33.js");
+      const { launchInteractiveWizard } = await import("./interactive-3VTDK5NX.js");
       await launchInteractiveWizard();
       return;
     }
@@ -4220,7 +4222,8 @@ var evalRunCommand = command({
       benchmarkJson: args.benchmarkJson,
       artifacts: args.artifacts,
       graderTarget: args.graderTarget,
-      model: args.model
+      model: args.model,
+      outputMessages: args.outputMessages
     };
     await runEvalCommand({ testFiles: resolvedPaths, rawOptions });
   }
@@ -4706,10 +4709,9 @@ function exportResults(sourceFile, content, outputDir) {
   const outputsDir = path8.join(outputDir, "outputs");
   mkdirSync2(outputsDir, { recursive: true });
   for (const result of patched) {
-    const outputText = result.outputText;
-    if (outputText) {
+    if (result.output && result.output.length > 0) {
       const id = safeTestId(result);
-      writeFileSync3(path8.join(outputsDir, `${id}.txt`), outputText);
+      writeFileSync3(path8.join(outputsDir, `${id}.txt`), JSON.stringify(result.output, null, 2));
     }
   }
 }
@@ -5021,7 +5023,6 @@ function toTraceSummary(raw) {
   return toCamelCaseDeep(raw.trace);
 }
 function extractCandidate(raw) {
-  if (raw.output_text !== void 0) return raw.output_text;
   if (raw.output !== void 0)
     return typeof raw.output === "string" ? raw.output : JSON.stringify(raw.output);
   return "";
@@ -5033,7 +5034,6 @@ function buildEvalTest(raw) {
     input: [],
     input_segments: [],
     expected_output: [],
-    guideline_paths: [],
     file_paths: [],
     criteria: ""
   };
@@ -5071,7 +5071,7 @@ async function runScore(results, evaluatorConfig, testIdFilter) {
       target: { kind: "custom", name: raw.target ?? "unknown", config: {} },
       provider: stubProvider,
       attempt: 1,
-      promptInputs: { question: "", guidelines: "" },
+      promptInputs: { question: "" },
       now: /* @__PURE__ */ new Date(),
       output: Array.isArray(output) ? output : void 0,
       trace,
@@ -5325,7 +5325,7 @@ function formatResultDetail(result, index, tree) {
   }
   const scoreColor = result.score >= 0.9 ? c2.green : result.score >= 0.5 ? c2.yellow : c2.red;
   lines.push(
-    `${c2.bold}${testId}${c2.reset}  ${scoreColor}${formatScore(result.score)}${c2.reset}${result.target ? `  ${c2.dim}target: ${result.target}${c2.reset}` : ""}${result.dataset ? `  ${c2.dim}dataset: ${result.dataset}${c2.reset}` : ""}`
+    `${c2.bold}${testId}${c2.reset}  ${scoreColor}${formatScore(result.score)}${c2.reset}${result.target ? `  ${c2.dim}target: ${result.target}${c2.reset}` : ""}${result.eval_set ? `  ${c2.dim}eval-set: ${result.eval_set}${c2.reset}` : ""}`
   );
   if (result.error) {
     lines.push(`  ${c2.red}Error: ${result.error}${c2.reset}`);
@@ -5499,8 +5499,8 @@ function groupResults(results, groupBy2) {
       case "target":
         key = result.target ?? "unknown";
         break;
-      case "dataset":
-        key = result.dataset ?? "unknown";
+      case "eval-set":
+        key = result.eval_set ?? "unknown";
         break;
       case "test-id":
         key = result.test_id ?? result.eval_id ?? "unknown";
@@ -5582,10 +5582,10 @@ var traceStatsCommand = command({
       description: "Path to JSONL result file"
     }),
     groupBy: option({
-      type: optional(oneOf(["target", "dataset", "test-id"])),
+      type: optional(oneOf(["target", "eval-set", "test-id"])),
       long: "group-by",
       short: "g",
-      description: "Group statistics by: target, dataset, or test-id"
+      description: "Group statistics by: target, eval-set, or test-id"
     }),
     format: option({
       type: optional(oneOf(["table", "json"])),
@@ -6276,4 +6276,4 @@ export {
   preprocessArgv,
   runCli
 };
-//# sourceMappingURL=chunk-TR6H437M.js.map
+//# sourceMappingURL=chunk-Q2YWV4QM.js.map