agentv 3.14.6 → 4.1.0

package/dist/{chunk-CQRWNXVG.js → chunk-2W5JKKXC.js}

@@ -10,6 +10,7 @@ import {
   loadManifestResults,
   loadRunCache,
   package_default,
+  parseResultManifest,
   resolveEvalPaths,
   resolveExistingRunPrimaryPath,
   resolveResultSourcePath,
@@ -23,9 +24,11 @@ import {
   validateFileReferences,
   validateTargetsFile,
   writeArtifactsFromResults
-} from "./chunk-Y25VL7PX.js";
+} from "./chunk-4Z326WWF.js";
 import {
+  DEFAULT_CATEGORY,
   createBuiltinRegistry,
+  deriveCategory,
   executeScript,
   getAgentvHome,
   getOutputFilenames,
@@ -40,7 +43,7 @@ import {
   toSnakeCaseDeep as toSnakeCaseDeep2,
   transpileEvalYamlFile,
   trimBaselineResult
-} from "./chunk-ELQEFMGO.js";
+} from "./chunk-XEAW7OQT.js";
 import {
   __commonJS,
   __esm,
@@ -3479,9 +3482,23 @@ var ASSERTION_TEMPLATES = {
   default: `#!/usr/bin/env bun
 import { defineAssertion } from '@agentv/eval';
 
-export default defineAssertion(({ outputText }) => {
+/** Extract text from the last message with the given role. */
+function getMessageText(messages: Array<{ role: string; content?: unknown }>, role = 'assistant'): string {
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const msg = messages[i];
+    if (msg.role !== role) continue;
+    if (typeof msg.content === 'string') return msg.content;
+    if (Array.isArray(msg.content)) {
+      return msg.content.filter((b: any) => b.type === 'text').map((b: any) => b.text).join('\\n');
+    }
+  }
+  return '';
+}
+
+export default defineAssertion(({ output }) => {
   // TODO: Implement your assertion logic
-  const pass = outputText.length > 0;
+  const text = getMessageText(output ?? []);
+  const pass = text.length > 0;
   return {
     pass,
     reasoning: pass ? 'Output has content' : 'Output is empty',
@@ -3491,9 +3508,23 @@ export default defineAssertion(({ outputText }) => {
   score: `#!/usr/bin/env bun
 import { defineAssertion } from '@agentv/eval';
 
-export default defineAssertion(({ outputText }) => {
+/** Extract text from the last message with the given role. */
+function getMessageText(messages: Array<{ role: string; content?: unknown }>, role = 'assistant'): string {
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const msg = messages[i];
+    if (msg.role !== role) continue;
+    if (typeof msg.content === 'string') return msg.content;
+    if (Array.isArray(msg.content)) {
+      return msg.content.filter((b: any) => b.type === 'text').map((b: any) => b.text).join('\\n');
+    }
+  }
+  return '';
+}
+
+export default defineAssertion(({ output }) => {
   // TODO: Implement your scoring logic (0.0 to 1.0)
-  const score = outputText.length > 0 ? 1.0 : 0.0;
+  const text = getMessageText(output ?? []);
+  const score = text.length > 0 ? 1.0 : 0.0;
   return {
     pass: score >= 0.5,
     score,
@@ -4186,7 +4217,7 @@ var evalRunCommand = command({
   },
   handler: async (args) => {
     if (args.evalPaths.length === 0 && process.stdin.isTTY) {
-      const { launchInteractiveWizard } = await import("./interactive-5ESM5DWV.js");
+      const { launchInteractiveWizard } = await import("./interactive-7ZYS6IOC.js");
       await launchInteractiveWizard();
       return;
     }
@@ -4421,7 +4452,8 @@ var evalBenchCommand = command({
     const manifest = JSON.parse(await readFile(join(exportDir, "manifest.json"), "utf8"));
     const testIds = manifest.test_ids;
     const targetName = manifest.target?.name ?? "unknown";
-    const evalSet = manifest.eval_set ?? "";
+    const evalSet = manifest.dataset ?? "";
+    const experiment = manifest.experiment;
     const safeEvalSet = evalSet ? evalSet.replace(/[\/\\:*?"<>|]/g, "_") : "";
     let stdinData;
     if (llmScoresPath) {
@@ -4531,7 +4563,8 @@ var evalBenchCommand = command({
         JSON.stringify({
           timestamp: manifest.timestamp,
           test_id: testId,
-          eval_set: evalSet || void 0,
+          dataset: evalSet || void 0,
+          experiment: experiment || void 0,
           score: Math.round(weightedScore * 1e3) / 1e3,
           target: targetName,
           scores,
@@ -4553,6 +4586,7 @@ var evalBenchCommand = command({
       metadata: {
         eval_file: manifest.eval_file,
         timestamp: manifest.timestamp,
+        experiment: experiment || void 0,
         targets: [targetName],
         tests_run: testIds
       },
@@ -4594,6 +4628,12 @@ function computeStats(values) {
 // src/commands/pipeline/grade.ts
 import { mkdir as mkdir2, readFile as readFile2, readdir as readdir2, writeFile as writeFile3 } from "node:fs/promises";
 import { join as join2 } from "node:path";
+function extractInputText(input) {
+  if (!input || input.length === 0) return "";
+  if (input.length === 1) return input[0].content;
+  return input.map((m) => `@[${m.role}]:
+${m.content}`).join("\n\n");
+}
 var evalGradeCommand = command({
   name: "grade",
   description: "Run code-grader assertions on responses in an export directory",
@@ -4608,7 +4648,7 @@ var evalGradeCommand = command({
     const manifestPath = join2(exportDir, "manifest.json");
     const manifest = JSON.parse(await readFile2(manifestPath, "utf8"));
     const testIds = manifest.test_ids;
-    const evalSet = manifest.eval_set ?? "";
+    const evalSet = manifest.dataset ?? "";
     const safeEvalSet = evalSet ? evalSet.replace(/[\/\\:*?"<>|]/g, "_") : "";
     let totalGraders = 0;
     let totalPassed = 0;
@@ -4630,14 +4670,13 @@ var evalGradeCommand = command({
       for (const graderFile of graderFiles) {
         const graderConfig = JSON.parse(await readFile2(join2(codeGradersDir, graderFile), "utf8"));
         const graderName = graderConfig.name;
+        const inputText = extractInputText(inputData.input);
         const payload = JSON.stringify({
           output: [{ role: "assistant", content: responseText }],
-          input: inputData.input_messages,
-          question: inputData.input_text,
+          input: inputData.input,
           criteria: "",
           expected_output: [],
-          reference_answer: "",
-          input_files: [],
+          input_files: inputData.input_files ?? [],
           trace: null,
           token_usage: null,
           cost_usd: null,
@@ -4647,8 +4686,8 @@ var evalGradeCommand = command({
           file_changes: null,
           workspace_path: null,
           config: graderConfig.config ?? null,
-          metadata: {},
-          input_text: inputData.input_text,
+          metadata: inputData.metadata ?? {},
+          input_text: inputText,
           output_text: responseText,
           expected_output_text: ""
         });
@@ -4706,7 +4745,7 @@ var evalGradeCommand = command({
 // src/commands/pipeline/input.ts
 import { readFile as readFile3 } from "node:fs/promises";
 import { mkdir as mkdir3, writeFile as writeFile4 } from "node:fs/promises";
-import { dirname, join as join3, resolve } from "node:path";
+import { dirname, join as join3, relative, resolve } from "node:path";
 var evalInputCommand = command({
   name: "input",
   description: "Extract eval inputs, target commands, and grader prompts for subagent-mode runs",
@@ -4720,14 +4759,20 @@ var evalInputCommand = command({
       type: optional(string),
       long: "out",
       description: "Output directory for extracted inputs (default: .agentv/results/runs/<timestamp>)"
+    }),
+    experiment: option({
+      type: optional(string),
+      long: "experiment",
+      description: "Experiment label (e.g. with_skills, without_skills)"
     })
   },
-  handler: async ({ evalPath, out }) => {
+  handler: async ({ evalPath, out, experiment }) => {
     const resolvedEvalPath = resolve(evalPath);
     const outDir = resolve(out ?? buildDefaultRunDir(process.cwd()));
     const repoRoot = await findRepoRoot(dirname(resolvedEvalPath));
     const evalDir = dirname(resolvedEvalPath);
-    const suite = await loadTestSuite(resolvedEvalPath, repoRoot);
+    const category = deriveCategory(relative(process.cwd(), resolvedEvalPath));
+    const suite = await loadTestSuite(resolvedEvalPath, repoRoot, { category });
     const tests = suite.tests;
     if (tests.length === 0) {
       console.error("No tests found in eval file.");
@@ -4736,6 +4781,7 @@ var evalInputCommand = command({
     let targetInfo = null;
     let targetName = "agent";
     let targetKind = "agent";
+    let subagentModeAllowed = true;
     try {
       const selection = await selectTarget({
         testFilePath: resolvedEvalPath,
@@ -4748,15 +4794,20 @@ var evalInputCommand = command({
         env: process.env
       });
       targetName = selection.targetName;
-      if (selection.resolvedTarget.kind === "cli") {
+      const resolved = selection.resolvedTarget;
+      subagentModeAllowed = resolved.subagentModeAllowed !== false;
+      if (resolved.kind === "cli") {
         targetKind = "cli";
-        const config = selection.resolvedTarget.config;
+        subagentModeAllowed = false;
+        const config = resolved.config;
         targetInfo = {
           kind: "cli",
           command: config.command,
           cwd: config.cwd ?? evalDir,
           timeoutMs: config.timeoutMs ?? 3e4
         };
+      } else {
+        targetKind = resolved.kind;
       }
     } catch {
     }
@@ -4768,15 +4819,13 @@ var evalInputCommand = command({
       const testDir = join3(outDir, ...subpath);
       await mkdir3(testDir, { recursive: true });
       testIds.push(test.id);
-      const inputText = test.question;
       const inputMessages = test.input.map((m) => ({
         role: m.role,
         content: typeof m.content === "string" ? m.content : m.content
       }));
       await writeJson(join3(testDir, "input.json"), {
-        input_text: inputText,
-        input_messages: inputMessages,
-        file_paths: test.file_paths,
+        input: inputMessages,
+        input_files: test.file_paths,
         metadata: test.metadata ?? {}
       });
       if (targetInfo) {
@@ -4804,11 +4853,13 @@ var evalInputCommand = command({
     }
     await writeJson(join3(outDir, "manifest.json"), {
       eval_file: resolvedEvalPath,
-      eval_set: evalSetName || void 0,
+      dataset: evalSetName || void 0,
+      experiment: experiment || void 0,
       timestamp: (/* @__PURE__ */ new Date()).toISOString(),
       target: {
         name: targetName,
-        kind: targetKind
+        kind: targetKind,
+        subagent_mode_allowed: subagentModeAllowed
       },
       test_ids: testIds
     });
@@ -4870,7 +4921,13 @@ import { execSync } from "node:child_process";
 import { existsSync as existsSync3, readFileSync as readFileSync4, unlinkSync } from "node:fs";
 import { mkdir as mkdir4, readFile as readFile4, readdir as readdir3, writeFile as writeFile5 } from "node:fs/promises";
 import { tmpdir } from "node:os";
-import { dirname as dirname2, join as join4, resolve as resolve2 } from "node:path";
+import { dirname as dirname2, join as join4, relative as relative2, resolve as resolve2 } from "node:path";
+function extractInputText2(input) {
+  if (!input || input.length === 0) return "";
+  if (input.length === 1) return input[0].content;
+  return input.map((m) => `@[${m.role}]:
+${m.content}`).join("\n\n");
+}
 function loadEnvFile(dir) {
   let current = resolve2(dir);
   while (true) {
@@ -4910,14 +4967,25 @@ var evalRunCommand2 = command({
       type: optional(number),
       long: "workers",
       description: "Parallel workers for target invocation (default: all tests)"
+    }),
+    experiment: option({
+      type: optional(string),
+      long: "experiment",
+      description: "Experiment label (e.g. with_skills, without_skills)"
+    }),
+    graderType: option({
+      type: optional(oneOf(["code", "none"])),
+      long: "grader-type",
+      description: 'Which grading phase to run: "code" runs code-graders inline, omit to skip grading (use pipeline grade separately)'
     })
   },
-  handler: async ({ evalPath, out, workers }) => {
+  handler: async ({ evalPath, out, workers, experiment, graderType }) => {
     const resolvedEvalPath = resolve2(evalPath);
     const outDir = resolve2(out ?? buildDefaultRunDir(process.cwd()));
     const repoRoot = await findRepoRoot(dirname2(resolvedEvalPath));
     const evalDir = dirname2(resolvedEvalPath);
-    const suite = await loadTestSuite(resolvedEvalPath, repoRoot);
+    const category = deriveCategory(relative2(process.cwd(), resolvedEvalPath));
+    const suite = await loadTestSuite(resolvedEvalPath, repoRoot, { category });
     const tests = suite.tests;
     if (tests.length === 0) {
       console.error("No tests found in eval file.");
@@ -4958,15 +5026,13 @@ var evalRunCommand2 = command({
       const testDir = join4(outDir, ...subpath);
       await mkdir4(testDir, { recursive: true });
       testIds.push(test.id);
-      const inputText = test.question;
       const inputMessages = test.input.map((m) => ({
         role: m.role,
         content: typeof m.content === "string" ? m.content : m.content
       }));
       await writeJson2(join4(testDir, "input.json"), {
-        input_text: inputText,
-        input_messages: inputMessages,
-        file_paths: test.file_paths,
+        input: inputMessages,
+        input_files: test.file_paths,
         metadata: test.metadata ?? {}
       });
       if (targetInfo) {
@@ -4994,7 +5060,8 @@ var evalRunCommand2 = command({
     }
     await writeJson2(join4(outDir, "manifest.json"), {
       eval_file: resolvedEvalPath,
-      eval_set: evalSetName || void 0,
+      dataset: evalSetName || void 0,
+      experiment: experiment || void 0,
       timestamp: (/* @__PURE__ */ new Date()).toISOString(),
       target: { name: targetName, kind: targetKind },
       test_ids: testIds
@@ -5019,11 +5086,12 @@ var evalRunCommand2 = command({
         const timeoutMs = invoke.timeout_ms ?? 12e4;
         const promptFile = join4(tmpdir(), `agentv-prompt-${testId}-${Date.now()}.txt`);
         const outputFile = join4(tmpdir(), `agentv-output-${testId}-${Date.now()}.txt`);
-        await writeFile5(promptFile, inputData.input_text, "utf8");
+        const inputText = extractInputText2(inputData.input);
+        await writeFile5(promptFile, inputText, "utf8");
         let rendered = template;
         rendered = rendered.replace("{PROMPT_FILE}", promptFile);
         rendered = rendered.replace("{OUTPUT_FILE}", outputFile);
-        rendered = rendered.replace("{PROMPT}", inputData.input_text);
+        rendered = rendered.replace("{PROMPT}", inputText);
         const start = performance.now();
         try {
           execSync(rendered, {
@@ -5080,6 +5148,12 @@ var evalRunCommand2 = command({
     } else {
       console.log("Subagent-as-target mode \u2014 skipping CLI invocation.");
     }
+    if (graderType !== "code") {
+      console.log(`
+Done. Results in ${outDir}`);
+      console.log("To run code graders: agentv pipeline grade <run-dir>  (or re-run with --grader-type code)");
+      return;
+    }
     let totalGraders = 0;
     let totalPassed = 0;
     for (const testId of testIds) {
@@ -5100,14 +5174,13 @@ var evalRunCommand2 = command({
       for (const graderFile of graderFiles) {
         const graderConfig = JSON.parse(await readFile4(join4(codeGradersDir, graderFile), "utf8"));
         const graderName = graderConfig.name;
+        const inputText = extractInputText2(inputData.input);
         const payload = JSON.stringify({
           output: [{ role: "assistant", content: responseText }],
-          input: inputData.input_messages,
-          question: inputData.input_text,
+          input: inputData.input,
           criteria: "",
           expected_output: [],
-          reference_answer: "",
-          input_files: [],
+          input_files: inputData.input_files ?? [],
           trace: null,
           token_usage: null,
           cost_usd: null,
@@ -5117,8 +5190,8 @@ var evalRunCommand2 = command({
           file_changes: null,
           workspace_path: null,
           config: graderConfig.config ?? null,
-          metadata: {},
-          input_text: inputData.input_text,
+          metadata: inputData.metadata ?? {},
+          input_text: inputText,
           output_text: responseText,
           expected_output_text: ""
         });
@@ -5306,7 +5379,7 @@ function toRawResult(result) {
   return {
     timestamp: result.timestamp,
     test_id: result.testId,
-    eval_set: result.eval_set,
+    dataset: result.dataset,
     conversation_id: result.conversationId,
     score: result.score,
     assertions: result.assertions?.map((assertion) => ({
@@ -5429,7 +5502,7 @@ function loadOtlpTraceFile(filePath) {
     }
     return {
       test_id: stringAttr(rootAttrs.agentv_test_id) ?? stringAttr(rootAttrs.agentv_eval_id) ?? `trace-${index + 1}`,
-      eval_set: stringAttr(rootAttrs.agentv_eval_set),
+      dataset: stringAttr(rootAttrs.agentv_dataset),
       target: stringAttr(rootAttrs.agentv_target),
       score,
       error: root.status?.code === 2 ? root.status.message : void 0,
@@ -6173,8 +6246,9 @@ var resultsCommand = subcommands({
 });
 
 // src/commands/results/serve.ts
-import { existsSync as existsSync7, readFileSync as readFileSync8, writeFileSync as writeFileSync3 } from "node:fs";
+import { existsSync as existsSync7, readFileSync as readFileSync8, readdirSync as readdirSync3, statSync as statSync4, writeFileSync as writeFileSync3 } from "node:fs";
 import path9 from "node:path";
+import { fileURLToPath as fileURLToPath2 } from "node:url";
 import { Hono } from "hono";
 function feedbackPath(resultDir) {
   return path9.join(resultDir, "feedback.json");
@@ -6195,24 +6269,46 @@ function writeFeedback(cwd, data) {
   writeFileSync3(feedbackPath(cwd), `${JSON.stringify(data, null, 2)}
 `, "utf8");
 }
-function createApp(results, resultDir, cwd, sourceFile) {
+function createApp(results, resultDir, cwd, sourceFile, options) {
   const searchDir = cwd ?? resultDir;
   const app2 = new Hono();
+  const studioDistPath = options?.studioDir ?? resolveStudioDistDir();
+  if (!studioDistPath || !existsSync7(path9.join(studioDistPath, "index.html"))) {
+    throw new Error('Studio dist not found. Run "bun run build" in apps/studio/ to build the SPA.');
+  }
   app2.get("/", (c3) => {
-    return c3.html(generateServeHtml(results, sourceFile));
+    const indexPath = path9.join(studioDistPath, "index.html");
+    if (existsSync7(indexPath)) {
+      return c3.html(readFileSync8(indexPath, "utf8"));
+    }
+    return c3.notFound();
   });
   app2.get("/api/runs", (c3) => {
     const metas = listResultFiles(searchDir);
     return c3.json({
-      runs: metas.map((m) => ({
-        filename: m.filename,
-        path: m.path,
-        timestamp: m.timestamp,
-        test_count: m.testCount,
-        pass_rate: m.passRate,
-        avg_score: m.avgScore,
-        size_bytes: m.sizeBytes
-      }))
+      runs: metas.map((m) => {
+        let target;
+        let experiment;
+        try {
+          const records = loadLightweightResults(m.path);
+          if (records.length > 0) {
+            target = records[0].target;
+            experiment = records[0].experiment;
+          }
+        } catch {
+        }
+        return {
+          filename: m.filename,
+          path: m.path,
+          timestamp: m.timestamp,
+          test_count: m.testCount,
+          pass_rate: m.passRate,
+          avg_score: m.avgScore,
+          size_bytes: m.sizeBytes,
+          ...target && { target },
+          ...experiment && { experiment }
+        };
+      })
     });
   });
   app2.get("/api/runs/:filename", (c3) => {
@@ -6272,692 +6368,406 @@ function createApp(results, resultDir, cwd, sourceFile) {
     writeFeedback(resultDir, existing);
     return c3.json(existing);
   });
-  return app2;
-}
-function stripHeavyFields(results) {
-  return results.map((r) => {
-    const { requests, trace, ...rest } = r;
-    const toolCalls = trace?.toolCalls && Object.keys(trace.toolCalls).length > 0 ? trace.toolCalls : void 0;
-    const graderDurationMs = (r.scores ?? []).reduce((sum, s) => sum + (s.durationMs ?? 0), 0);
-    return {
-      ...rest,
-      ...toolCalls && { _toolCalls: toolCalls },
-      ...graderDurationMs > 0 && { _graderDurationMs: graderDurationMs }
-    };
+  app2.get("/api/runs/:filename/datasets", (c3) => {
+    const filename = c3.req.param("filename");
+    const metas = listResultFiles(searchDir);
+    const meta = metas.find((m) => m.filename === filename);
+    if (!meta) {
+      return c3.json({ error: "Run not found" }, 404);
+    }
+    try {
+      const loaded = patchTestIds(loadManifestResults(meta.path));
+      const datasetMap = /* @__PURE__ */ new Map();
+      for (const r of loaded) {
+        const ds = r.dataset ?? r.target ?? "default";
+        const entry = datasetMap.get(ds) ?? { total: 0, passed: 0, scoreSum: 0 };
+        entry.total++;
+        if (r.score >= 1) entry.passed++;
+        entry.scoreSum += r.score;
+        datasetMap.set(ds, entry);
+      }
+      const datasets = [...datasetMap.entries()].map(([name, entry]) => ({
+        name,
+        total: entry.total,
+        passed: entry.passed,
+        failed: entry.total - entry.passed,
+        avg_score: entry.total > 0 ? entry.scoreSum / entry.total : 0
+      }));
+      return c3.json({ datasets });
+    } catch {
+      return c3.json({ error: "Failed to load datasets" }, 500);
+    }
   });
-}
-function escapeHtml(s) {
-  return s.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;").replace(/"/g, "&quot;");
-}
-function generateServeHtml(results, sourceFile) {
-  const lightResults = stripHeavyFields(results);
-  const dataJson = JSON.stringify(lightResults).replace(/</g, "\\u003c").replace(/>/g, "\\u003e").replace(/\u2028/g, "\\u2028").replace(/\u2029/g, "\\u2029");
-  return `<!DOCTYPE html>
-<html lang="en">
-<head>
-    <meta charset="utf-8">
-    <meta name="viewport" content="width=device-width, initial-scale=1">
-    <title>AgentV Results Review</title>
-    <style>
-${SERVE_STYLES}
-    </style>
-</head>
-<body>
-    <header class="header">
-        <div class="header-left">
-            <h1 class="header-title">AgentV</h1>
-            <span class="header-subtitle">Results Review</span>
-        </div>
-        <div class="header-center">
-            <select id="run-picker" class="run-picker" title="Switch result file">
-                <option value="">Loading runs...</option>
-            </select>
-        </div>
-        <div class="header-right">
-            <span class="timestamp">${escapeHtml((/* @__PURE__ */ new Date()).toISOString())}</span>
-        </div>
-    </header>
-    <nav class="tabs" id="tabs">
-        <button class="tab active" data-tab="overview">Overview</button>
-        <button class="tab" data-tab="tests">Test Cases</button>
-    </nav>
-    <main id="app"></main>
-    <script>
-    var DATA = ${dataJson};
-    var INITIAL_SOURCE = ${sourceFile ? JSON.stringify(path9.basename(sourceFile)).replace(/</g, "\\u003c").replace(/>/g, "\\u003e") : "null"};
-${SERVE_SCRIPT}
-    </script>
-</body>
-</html>`;
-}
-var SERVE_STYLES = `
-*{margin:0;padding:0;box-sizing:border-box}
-:root{
-  --bg:#f6f8fa;--surface:#fff;--border:#d0d7de;--border-light:#e8ebee;
-  --text:#1f2328;--text-muted:#656d76;
-  --primary:#0969da;--primary-bg:#ddf4ff;
-  --success:#1a7f37;--success-bg:#dafbe1;
-  --danger:#cf222e;--danger-bg:#ffebe9;
-  --warning:#9a6700;--warning-bg:#fff8c5;
-  --radius:6px;
-  --shadow:0 1px 3px rgba(31,35,40,.04),0 1px 2px rgba(31,35,40,.06);
-  --font:-apple-system,BlinkMacSystemFont,"Segoe UI","Noto Sans",Helvetica,Arial,sans-serif;
-  --mono:ui-monospace,SFMono-Regular,"SF Mono",Menlo,Consolas,monospace;
-}
-body{font-family:var(--font);background:var(--bg);color:var(--text);line-height:1.5;font-size:14px}
-
-/* Header */
-.header{background:var(--surface);border-bottom:1px solid var(--border);padding:12px 24px;display:flex;align-items:center;justify-content:space-between}
-.header-left{display:flex;align-items:baseline;gap:12px}
-.header-title{font-size:18px;font-weight:600}
-.header-subtitle{font-size:14px;color:var(--text-muted)}
-.header-center{flex:1;display:flex;justify-content:center;padding:0 16px}
-.run-picker{padding:6px 10px;border:1px solid var(--border);border-radius:var(--radius);font-size:13px;background:var(--surface);color:var(--text);font-family:var(--font);max-width:400px;width:100%;cursor:pointer}
-.run-picker:hover{border-color:var(--primary)}
-.run-picker:focus{outline:none;border-color:var(--primary);box-shadow:0 0 0 3px var(--primary-bg)}
-.timestamp{font-size:12px;color:var(--text-muted);font-family:var(--mono)}
-
-/* Tabs */
-.tabs{background:var(--surface);border-bottom:1px solid var(--border);padding:0 24px;display:flex}
-.tab{background:none;border:none;padding:10px 16px;font-size:14px;color:var(--text-muted);cursor:pointer;border-bottom:2px solid transparent;font-family:var(--font);transition:color .15s,border-color .15s}
-.tab:hover{color:var(--text)}
-.tab.active{color:var(--text);font-weight:600;border-bottom-color:var(--primary)}
-
-#app{max-width:1280px;margin:0 auto;padding:24px}
-
-/* Stat cards */
-.stats-grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(140px,1fr));gap:12px;margin-bottom:24px}
-.stat-card{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);padding:16px;text-align:center;box-shadow:var(--shadow)}
-.stat-card.pass .stat-value{color:var(--success)}
-.stat-card.fail .stat-value{color:var(--danger)}
-.stat-card.error .stat-value{color:var(--danger)}
-.stat-card.warn .stat-value{color:var(--warning)}
-.stat-card.total .stat-value{color:var(--primary)}
-.stat-value{font-size:28px;font-weight:700;line-height:1.2}
-.stat-label{font-size:12px;color:var(--text-muted);text-transform:uppercase;letter-spacing:.5px;margin-top:4px}
-
-/* Sections */
-.section{margin-bottom:24px}
-.section-title{font-size:16px;font-weight:600;margin-bottom:12px}
-
-/* Tables */
-.table-wrap{overflow-x:auto;background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);box-shadow:var(--shadow)}
-.data-table{width:100%;border-collapse:collapse;font-size:13px}
-.data-table th{background:var(--bg);border-bottom:1px solid var(--border);padding:8px 12px;text-align:left;font-weight:600;font-size:12px;color:var(--text-muted);text-transform:uppercase;letter-spacing:.3px;white-space:nowrap}
-.data-table th.sortable{cursor:pointer;user-select:none}
-.data-table th.sortable:hover{color:var(--text)}
-.data-table td{padding:8px 12px;border-bottom:1px solid var(--border-light);vertical-align:middle}
-.data-table tbody tr:last-child td{border-bottom:none}
-
-/* Status icons */
-.status-icon{display:inline-flex;align-items:center;justify-content:center;width:22px;height:22px;border-radius:50%;font-size:12px;font-weight:700}
-.status-icon.pass{background:var(--success-bg);color:var(--success)}
-.status-icon.fail{background:var(--danger-bg);color:var(--danger)}
-.status-icon.error{background:var(--warning-bg);color:var(--warning)}
-
-/* Score colors */
-.score-high{color:var(--success);font-weight:600}
-.score-mid{color:var(--warning);font-weight:600}
-.score-low{color:var(--danger);font-weight:600}
-
-/* Pass-rate bar */
-.bar-bg{width:100px;height:8px;background:var(--border-light);border-radius:4px;overflow:hidden}
-.bar-fill{height:100%;border-radius:4px;transition:width .3s}
-.bar-fill.score-high{background:var(--success)}
-.bar-fill.score-mid{background:var(--warning)}
-.bar-fill.score-low{background:var(--danger)}
-
-/* Histogram */
-.histogram{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);padding:16px;box-shadow:var(--shadow)}
-.hist-row{display:flex;align-items:center;gap:12px;margin-bottom:8px}
-.hist-row:last-child{margin-bottom:0}
-.hist-label{width:60px;font-size:12px;color:var(--text-muted);text-align:right;flex-shrink:0}
-.hist-bar-bg{flex:1;height:20px;background:var(--border-light);border-radius:3px;overflow:hidden}
-.hist-bar{height:100%;border-radius:3px;transition:width .3s}
-.hist-count{width:30px;font-size:12px;color:var(--text-muted);text-align:right;flex-shrink:0}
-
-/* Filters */
-.filter-bar{display:flex;gap:8px;margin-bottom:16px;align-items:center;flex-wrap:wrap}
-.filter-select,.filter-search{padding:6px 10px;border:1px solid var(--border);border-radius:var(--radius);font-size:13px;background:var(--surface);color:var(--text);font-family:var(--font)}
-.filter-search{flex:1;min-width:200px}
-.filter-count{font-size:12px;color:var(--text-muted);margin-left:auto}
-
-/* Test rows */
-.test-row{cursor:pointer;transition:background .1s}
-.test-row:hover{background:var(--bg)!important}
-.test-row.expanded{background:var(--primary-bg)!important}
-.expand-col{width:32px;text-align:center}
-.expand-icon{color:var(--text-muted);font-size:12px}
-.fw-medium{font-weight:500}
-.text-pass{color:var(--success)}.text-fail{color:var(--danger)}.text-error{color:var(--warning)}
-
-/* Detail panel */
-.detail-row td{padding:0!important;background:var(--bg)!important}
-.detail-panel{padding:16px 24px}
-.detail-grid{display:grid;grid-template-columns:1fr 1fr;gap:16px;margin-bottom:16px}
-.detail-block h4{font-size:12px;color:var(--text-muted);text-transform:uppercase;letter-spacing:.3px;margin-bottom:6px}
-.detail-pre{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);padding:12px;font-family:var(--mono);font-size:12px;white-space:pre-wrap;word-break:break-word;max-height:300px;overflow-y:auto;line-height:1.6}
-.detail-panel h4{font-size:13px;font-weight:600;margin:16px 0 8px}
-.eval-table{width:100%;border-collapse:collapse;font-size:13px;background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);margin-bottom:12px}
-.eval-table th{background:var(--bg);padding:6px 10px;text-align:left;font-size:11px;font-weight:600;color:var(--text-muted);text-transform:uppercase;border-bottom:1px solid var(--border)}
-.eval-table td{padding:8px 10px;border-bottom:1px solid var(--border-light)}
-.reasoning-cell{max-width:500px;font-size:12px;color:var(--text-muted)}
-.expect-list{list-style:none;padding:0;margin-bottom:12px}
-.expect-list li{padding:4px 8px 4px 24px;position:relative;font-size:13px}
-.expect-list.pass li::before{content:"\\2713";position:absolute;left:4px;color:var(--success);font-weight:700}
-.expect-list.fail li::before{content:"\\2717";position:absolute;left:4px;color:var(--danger);font-weight:700}
-.error-box{background:var(--danger-bg);border:1px solid var(--danger);border-radius:var(--radius);padding:12px;margin-bottom:12px}
-.error-box h4{color:var(--danger);margin:0 0 6px}
-.error-box pre{font-family:var(--mono);font-size:12px;white-space:pre-wrap;word-break:break-word}
-.detail-meta{font-size:12px;color:var(--text-muted);margin-top:12px;padding-top:12px;border-top:1px solid var(--border-light)}
-.tool-calls{display:flex;flex-wrap:wrap;gap:6px;margin-bottom:12px}
-.tool-tag{display:inline-block;padding:2px 10px;font-size:12px;font-family:var(--mono);background:var(--primary-bg);color:var(--primary);border:1px solid var(--border);border-radius:12px}
-.empty-state{text-align:center;padding:48px 24px;color:var(--text-muted)}
-.empty-state h3{font-size:16px;margin-bottom:8px;color:var(--text)}
-.welcome-state{text-align:center;padding:80px 24px;color:var(--text-muted)}
-.welcome-state h2{font-size:24px;margin-bottom:12px;color:var(--text);font-weight:600}
-.welcome-state p{font-size:15px;margin-bottom:8px;max-width:500px;margin-left:auto;margin-right:auto}
-.welcome-state code{font-family:var(--mono);background:var(--surface);border:1px solid var(--border);border-radius:3px;padding:2px 6px;font-size:13px}
-.welcome-state .hint{margin-top:24px;font-size:13px;color:var(--text-muted)}
-
-/* Feedback */
-.feedback-section{margin-top:16px;padding-top:16px;border-top:1px solid var(--border-light)}
-.feedback-input{width:100%;min-height:80px;padding:8px 12px;border:1px solid var(--border);border-radius:var(--radius);font-family:var(--font);font-size:13px;resize:vertical;background:var(--surface);color:var(--text)}
-.feedback-input:focus{outline:none;border-color:var(--primary);box-shadow:0 0 0 3px var(--primary-bg)}
-.feedback-submit{margin-top:8px;padding:6px 16px;background:var(--primary);color:#fff;border:none;border-radius:var(--radius);font-size:13px;cursor:pointer;font-family:var(--font)}
-.feedback-submit:hover{opacity:.9}
-.feedback-submit:disabled{opacity:.5;cursor:default}
-.feedback-status{margin-left:8px;font-size:12px;color:var(--success)}
-`;
-var SERVE_SCRIPT = `
-(function(){
-  /* ---- helpers ---- */
-  function esc(s){
-    if(s==null)return"";
-    return String(s).replace(/&/g,"&amp;").replace(/</g,"&lt;").replace(/>/g,"&gt;").replace(/"/g,"&quot;");
-  }
-  function getStatus(r){
-    if(r.executionStatus==="execution_error")return"error";
-    if(r.executionStatus==="quality_failure")return"fail";
-    if(r.executionStatus==="ok")return"pass";
-    if(r.error)return"error";
-    return r.score>=0.5?"pass":"fail";
-  }
-  function sIcon(s){
-    if(s==="pass")return'<span class="status-icon pass">\\u2713</span>';
-    if(s==="fail")return'<span class="status-icon fail">\\u2717</span>';
-    return'<span class="status-icon error">!</span>';
-  }
-  function fmtDur(ms){
-    if(ms==null)return"\\u2014";
-    if(ms<1000)return ms+"ms";
-    if(ms<60000)return(ms/1000).toFixed(1)+"s";
-    return Math.floor(ms/60000)+"m "+Math.round((ms%60000)/1000)+"s";
-  }
-  function fmtTok(n){
-    if(n==null)return"\\u2014";
-    if(n>=1e6)return(n/1e6).toFixed(1)+"M";
-    if(n>=1e3)return(n/1e3).toFixed(1)+"K";
-    return String(n);
-  }
-  function fmtCost(u){if(u==null)return"\\u2014";if(u<0.01)return"<$0.01";return"$"+u.toFixed(2);}
-  function fmtPct(v){if(v==null)return"\\u2014";return(v*100).toFixed(1)+"%";}
-  function sCls(v){if(v==null)return"";if(v>=0.9)return"score-high";if(v>=0.5)return"score-mid";return"score-low";}
-
-  /* ---- feedback state ---- */
-  var feedbackCache={};
-
-  function loadFeedback(){
-    fetch("/api/feedback").then(function(r){return r.json();}).then(function(d){
-      if(d&&d.reviews){
-        for(var i=0;i<d.reviews.length;i++){
-          feedbackCache[d.reviews[i].test_id]=d.reviews[i].comment;
-        }
-        populateFeedbackTextareas();
-      }
-    }).catch(function(){});
-  }
-
-  function populateFeedbackTextareas(){
-    var areas=document.querySelectorAll(".feedback-input");
-    for(var i=0;i<areas.length;i++){
-      var tid=areas[i].getAttribute("data-test-id");
-      if(tid&&feedbackCache[tid]!=null){
-        areas[i].value=feedbackCache[tid];
-      }
+  app2.get("/api/runs/:filename/categories", (c3) => {
+    const filename = c3.req.param("filename");
+    const metas = listResultFiles(searchDir);
+    const meta = metas.find((m) => m.filename === filename);
+    if (!meta) {
+      return c3.json({ error: "Run not found" }, 404);
     }
-  }
-
-  function saveFeedback(testId,comment,statusEl,btn){
-    btn.disabled=true;
-    statusEl.textContent="Saving...";
-    statusEl.style.color="var(--text-muted)";
-    fetch("/api/feedback",{
-      method:"POST",
-      headers:{"Content-Type":"application/json"},
-      body:JSON.stringify({reviews:[{test_id:testId,comment:comment}]})
-    }).then(function(r){return r.json();}).then(function(){
-      feedbackCache[testId]=comment;
-      statusEl.textContent="Saved";
-      statusEl.style.color="var(--success)";
-      btn.disabled=false;
-      setTimeout(function(){statusEl.textContent="";},2000);
-    }).catch(function(){
-      statusEl.textContent="Error saving";
-      statusEl.style.color="var(--danger)";
-      btn.disabled=false;
-    });
-  }
-
-  /* ---- compute stats ---- */
-  function computeStats(d){
-    var t=d.length,p=0,f=0,e=0,dur=0,ti=0,to=0,cost=0,sc=[],tc=0;
-    for(var i=0;i<d.length;i++){
-      var r=d[i],s=getStatus(r);
-      if(s==="pass")p++;else if(s==="fail")f++;else e++;
-      if(r.durationMs)dur+=r.durationMs;
-      if(r.tokenUsage){ti+=(r.tokenUsage.input||0);to+=(r.tokenUsage.output||0);}
-      if(r.costUsd)cost+=r.costUsd;
-      if(s!=="error")sc.push(r.score);
-      if(r._toolCalls){for(var k in r._toolCalls)tc+=r._toolCalls[k];}
-    }
-    var g=t-e;
-    return{total:t,passed:p,failed:f,errors:e,passRate:g>0?p/g:0,dur:dur,tokens:ti+to,inTok:ti,outTok:to,cost:cost,scores:sc,toolCalls:tc};
-  }
-  function computeTargets(d){
-    var m={};
-    for(var i=0;i<d.length;i++){
-      var r=d[i],tgt=r.target||"unknown";
-      if(!m[tgt])m[tgt]={target:tgt,results:[],p:0,f:0,e:0,ts:0,sc:0,dur:0,tok:0,cost:0};
-      var o=m[tgt];o.results.push(r);
-      var s=getStatus(r);
-      if(s==="pass")o.p++;else if(s==="fail")o.f++;else o.e++;
-      if(s!=="error"){o.ts+=r.score;o.sc++;}
-      if(r.durationMs)o.dur+=r.durationMs;
-      if(r.tokenUsage)o.tok+=(r.tokenUsage.input||0)+(r.tokenUsage.output||0);
-      if(r.costUsd)o.cost+=r.costUsd;
-    }
-    var a=[];for(var k in m)a.push(m[k]);return a;
-  }
-  function getEvalNames(){
-    var n={};
-    for(var i=0;i<DATA.length;i++){
-      var sc=DATA[i].scores;
-      if(sc)for(var j=0;j<sc.length;j++)n[sc[j].name]=true;
-    }
-    return Object.keys(n);
-  }
-  function getEvalScore(r,name){
-    if(!r.scores)return null;
-    for(var i=0;i<r.scores.length;i++)if(r.scores[i].name===name)return r.scores[i].score;
-    return null;
-  }
-
-  var stats=computeStats(DATA);
-  var tgtStats=computeTargets(DATA);
-  var tgtNames=tgtStats.map(function(t){return t.target;});
-
-  /* ---- state ---- */
-  var state={tab:"overview",filter:{status:"all",target:"all",search:""},sort:{col:"testId",dir:"asc"},expanded:{}};
-
-  /* ---- DOM refs ---- */
-  var app=document.getElementById("app");
-  var tabBtns=document.querySelectorAll(".tab");
-
-  /* ---- tabs ---- */
-  function setTab(t){
-    state.tab=t;
-    for(var i=0;i<tabBtns.length;i++)tabBtns[i].classList.toggle("active",tabBtns[i].getAttribute("data-tab")===t);
-    render();
-  }
-  for(var i=0;i<tabBtns.length;i++){
-    tabBtns[i].addEventListener("click",(function(b){return function(){setTab(b.getAttribute("data-tab"));};})(tabBtns[i]));
-  }
-
-  /* ---- render ---- */
-  function render(){
-    if(DATA.length===0){
-      app.innerHTML='<div class="welcome-state">'
-        +'<h2>No results yet</h2>'
-        +'<p>Run an evaluation or mount a results directory to see results here.</p>'
-        +'<p><code>agentv eval &lt;eval-file&gt;</code></p>'
-        +'<p class="hint">The dashboard will automatically detect new result files.</p>'
-        +'</div>';
-      return;
+    try {
+      const loaded = patchTestIds(loadManifestResults(meta.path));
+      const categoryMap = /* @__PURE__ */ new Map();
+      for (const r of loaded) {
+        const cat = r.category ?? DEFAULT_CATEGORY;
+        const entry = categoryMap.get(cat) ?? {
+          total: 0,
+          passed: 0,
+          scoreSum: 0,
+          datasets: /* @__PURE__ */ new Set()
+        };
+        entry.total++;
+        if (r.score >= 1) entry.passed++;
+        entry.scoreSum += r.score;
+        entry.datasets.add(r.dataset ?? r.target ?? "default");
+        categoryMap.set(cat, entry);
+      }
+      const categories = [...categoryMap.entries()].map(([name, entry]) => ({
+        name,
+        total: entry.total,
+        passed: entry.passed,
+        failed: entry.total - entry.passed,
+        avg_score: entry.total > 0 ? entry.scoreSum / entry.total : 0,
+        dataset_count: entry.datasets.size
+      }));
+      return c3.json({ categories });
+    } catch {
+      return c3.json({ error: "Failed to load categories" }, 500);
     }
-    if(state.tab==="overview")renderOverview();else renderTests();
-  }
-
-  /* ---- stat card helper ---- */
-  function card(label,value,type){
-    return'<div class="stat-card '+type+'"><div class="stat-value">'+value+'</div><div class="stat-label">'+label+"</div></div>";
-  }
-
-  /* ---- overview ---- */
-  function renderOverview(){
-    var h='<div class="stats-grid">';
-    h+=card("Total Tests",stats.total,"total");
-    h+=card("Passed",stats.passed,"pass");
-    h+=card("Failed",stats.failed,"fail");
-    h+=card("Errors",stats.errors,"error");
-    var prCls=stats.passRate>=0.9?"pass":stats.passRate>=0.5?"warn":"fail";
-    h+=card("Pass Rate",fmtPct(stats.passRate),prCls);
-    h+=card("Duration",fmtDur(stats.dur),"neutral");
-    h+=card("Tokens",fmtTok(stats.tokens),"neutral");
-    h+=card("Est. Cost",fmtCost(stats.cost),"neutral");
-    if(stats.toolCalls>0)h+=card("Tool Calls",fmtTok(stats.toolCalls),"neutral");
-    h+="</div>";
-
-    /* targets table */
-    if(tgtStats.length>1){
-      h+='<div class="section"><h2 class="section-title">Targets</h2><div class="table-wrap"><table class="data-table">';
-      h+="<thead><tr><th>Target</th><th>Pass Rate</th><th></th><th>Passed</th><th>Failed</th><th>Errors</th><th>Avg Score</th><th>Duration</th><th>Tokens</th><th>Cost</th></tr></thead><tbody>";
-      for(var i=0;i<tgtStats.length;i++){
-        var t=tgtStats[i],g=t.p+t.f,pr=g>0?t.p/g:0,avg=t.sc>0?t.ts/t.sc:0;
-        h+="<tr><td class=\\"fw-medium\\">"+esc(t.target)+"</td><td>"+fmtPct(pr)+'</td><td><div class="bar-bg"><div class="bar-fill '+sCls(pr)+'" style="width:'+(pr*100)+'%"></div></div></td>';
-        h+='<td class="text-pass">'+t.p+'</td><td class="text-fail">'+t.f+'</td><td class="text-error">'+t.e+"</td>";
-        h+='<td class="'+sCls(avg)+'">'+fmtPct(avg)+"</td><td>"+fmtDur(t.dur)+"</td><td>"+fmtTok(t.tok)+"</td><td>"+fmtCost(t.cost)+"</td></tr>";
-      }
-      h+="</tbody></table></div></div>";
+  });
+  app2.get("/api/runs/:filename/categories/:category/datasets", (c3) => {
+    const filename = c3.req.param("filename");
+    const category = decodeURIComponent(c3.req.param("category"));
+    const metas = listResultFiles(searchDir);
+    const meta = metas.find((m) => m.filename === filename);
+    if (!meta) {
+      return c3.json({ error: "Run not found" }, 404);
     }
-
-    /* histogram */
-    if(stats.scores.length>0){
-      var bk=[0,0,0,0,0];
-      for(var i=0;i<stats.scores.length;i++){var idx=Math.min(Math.floor(stats.scores[i]*5),4);bk[idx]++;}
-      var mx=Math.max.apply(null,bk);
-      var lb=["0\\u201320%","20\\u201340%","40\\u201360%","60\\u201380%","80\\u2013100%"];
-      h+='<div class="section"><h2 class="section-title">Score Distribution</h2><div class="histogram">';
-      for(var i=0;i<bk.length;i++){
-        var pct=mx>0?(bk[i]/mx*100):0;
-        h+='<div class="hist-row"><span class="hist-label">'+lb[i]+'</span><div class="hist-bar-bg"><div class="hist-bar '+(i>=4?"score-high":i>=2?"score-mid":"score-low")+'" style="width:'+pct+'%"></div></div><span class="hist-count">'+bk[i]+"</span></div>";
+    try {
+      const loaded = patchTestIds(loadManifestResults(meta.path));
+      const filtered = loaded.filter((r) => (r.category ?? DEFAULT_CATEGORY) === category);
+      const datasetMap = /* @__PURE__ */ new Map();
+      for (const r of filtered) {
+        const ds = r.dataset ?? r.target ?? "default";
+        const entry = datasetMap.get(ds) ?? { total: 0, passed: 0, scoreSum: 0 };
+        entry.total++;
+        if (r.score >= 1) entry.passed++;
+        entry.scoreSum += r.score;
+        datasetMap.set(ds, entry);
+      }
+      const datasets = [...datasetMap.entries()].map(([name, entry]) => ({
+        name,
+        total: entry.total,
+        passed: entry.passed,
+        failed: entry.total - entry.passed,
+        avg_score: entry.total > 0 ? entry.scoreSum / entry.total : 0
+      }));
+      return c3.json({ datasets });
+    } catch {
+      return c3.json({ error: "Failed to load datasets" }, 500);
+    }
+  });
+  app2.get("/api/runs/:filename/evals/:evalId", (c3) => {
+    const filename = c3.req.param("filename");
+    const evalId = c3.req.param("evalId");
+    const metas = listResultFiles(searchDir);
+    const meta = metas.find((m) => m.filename === filename);
+    if (!meta) {
+      return c3.json({ error: "Run not found" }, 404);
+    }
+    try {
+      const loaded = patchTestIds(loadManifestResults(meta.path));
+      const result = loaded.find((r) => r.testId === evalId);
+      if (!result) {
+        return c3.json({ error: "Eval not found" }, 404);
       }
-      h+="</div></div>";
+      return c3.json({ eval: result });
+    } catch {
+      return c3.json({ error: "Failed to load eval" }, 500);
     }
-    app.innerHTML=h;
-  }
-
-  /* ---- test cases ---- */
-  function renderTests(){
-    var evalNames=getEvalNames();
-    var h='<div class="filter-bar">';
-    h+='<select id="flt-status" class="filter-select"><option value="all">All Status</option><option value="pass">Passed</option><option value="fail">Failed</option><option value="error">Errors</option></select>';
-    if(tgtNames.length>1){
-      h+='<select id="flt-target" class="filter-select"><option value="all">All Targets</option>';
-      for(var i=0;i<tgtNames.length;i++)h+='<option value="'+esc(tgtNames[i])+'">'+esc(tgtNames[i])+"</option>";
-      h+="</select>";
-    }
-    h+='<input type="text" id="flt-search" class="filter-search" placeholder="Search tests..." value="'+esc(state.filter.search)+'">';
-    h+='<span class="filter-count" id="flt-count"></span></div>';
-
-    h+='<div class="table-wrap"><table class="data-table" id="test-tbl"><thead><tr>';
-    h+='<th class="expand-col"></th>';
-    h+=sHdr("Status","status");
-    h+=sHdr("Test ID","testId");
-    if(tgtNames.length>1)h+=sHdr("Target","target");
-    h+=sHdr("Score","score");
-    for(var i=0;i<evalNames.length;i++)h+="<th>"+esc(evalNames[i])+"</th>";
-    h+=sHdr("Duration","durationMs");
-    h+=sHdr("Cost","costUsd");
-    h+="</tr></thead><tbody id=\\"test-body\\"></tbody></table></div>";
-    app.innerHTML=h;
-
-    /* wire events */
-    var selS=document.getElementById("flt-status");
-    selS.value=state.filter.status;
-    selS.addEventListener("change",function(e){state.filter.status=e.target.value;renderRows();});
-    var selT=document.getElementById("flt-target");
-    if(selT){selT.value=state.filter.target;selT.addEventListener("change",function(e){state.filter.target=e.target.value;renderRows();});}
-    document.getElementById("flt-search").addEventListener("input",function(e){state.filter.search=e.target.value;renderRows();});
-    var ths=document.querySelectorAll("th[data-sort]");
-    for(var i=0;i<ths.length;i++){
-      ths[i].addEventListener("click",(function(th){return function(){
-        var c=th.getAttribute("data-sort");
-        if(state.sort.col===c)state.sort.dir=state.sort.dir==="asc"?"desc":"asc";
-        else{state.sort.col=c;state.sort.dir="asc";}
-        renderTests();
-      };})(ths[i]));
-    }
-    renderRows();
-  }
-
-  function sHdr(label,col){
-    var arrow="";
-    if(state.sort.col===col)arrow=state.sort.dir==="asc"?" \\u2191":" \\u2193";
-    return'<th class="sortable" data-sort="'+col+'">'+label+arrow+"</th>";
-  }
-
-  function filtered(){
-    var out=[];
-    for(var i=0;i<DATA.length;i++){
-      var r=DATA[i],s=getStatus(r);
-      if(state.filter.status!=="all"&&s!==state.filter.status)continue;
-      if(state.filter.target!=="all"&&r.target!==state.filter.target)continue;
-      if(state.filter.search&&(r.testId||"").toLowerCase().indexOf(state.filter.search.toLowerCase())===-1)continue;
-      out.push(r);
-    }
-    var col=state.sort.col,dir=state.sort.dir==="asc"?1:-1;
-    out.sort(function(a,b){
-      var va=col==="status"?getStatus(a):a[col],vb=col==="status"?getStatus(b):b[col];
-      if(va==null&&vb==null)return 0;if(va==null)return 1;if(vb==null)return-1;
-      if(typeof va==="string")return va.localeCompare(vb)*dir;
-      return(va-vb)*dir;
+  });
+  app2.get("/api/index", (c3) => {
+    const metas = listResultFiles(searchDir);
+    const entries2 = metas.map((m) => {
+      let totalCostUsd = 0;
+      try {
+        const loaded = patchTestIds(loadManifestResults(m.path));
+        totalCostUsd = loaded.reduce((sum, r) => sum + (r.costUsd ?? 0), 0);
+      } catch {
+      }
+      return {
+        run_filename: m.filename,
+        test_count: m.testCount,
+        pass_rate: m.passRate,
+        avg_score: m.avgScore,
+        total_cost_usd: totalCostUsd,
+        timestamp: m.timestamp
+      };
+    });
+    return c3.json({ entries: entries2 });
+  });
+  function buildFileTree(dirPath, relativeTo) {
+    if (!existsSync7(dirPath) || !statSync4(dirPath).isDirectory()) {
+      return [];
+    }
+    const entries2 = readdirSync3(dirPath, { withFileTypes: true });
+    return entries2.sort((a, b) => {
+      if (a.isDirectory() !== b.isDirectory()) return a.isDirectory() ? -1 : 1;
+      return a.name.localeCompare(b.name);
+    }).map((entry) => {
+      const fullPath = path9.join(dirPath, entry.name);
+      const relPath = path9.relative(relativeTo, fullPath);
+      if (entry.isDirectory()) {
+        return {
+          name: entry.name,
+          path: relPath,
+          type: "dir",
+          children: buildFileTree(fullPath, relativeTo)
+        };
+      }
+      return { name: entry.name, path: relPath, type: "file" };
     });
-    return out;
   }
-
-  function renderRows(){
-    var rows=filtered(),evalNames=getEvalNames();
-    var tbody=document.getElementById("test-body");
-    var colSpan=5+evalNames.length+(tgtNames.length>1?1:0);
-    document.getElementById("flt-count").textContent=rows.length+" of "+DATA.length+" tests";
-    var h="";
-    for(var i=0;i<rows.length;i++){
-      var r=rows[i],s=getStatus(r),key=r.testId+":"+r.target,exp=!!state.expanded[key];
-      h+='<tr class="test-row '+s+(exp?" expanded":"")+'" data-key="'+esc(key)+'" data-test-id="'+esc(r.testId)+'">';
-      h+='<td class="expand-col"><span class="expand-icon">'+(exp?"\\u25BE":"\\u25B8")+"</span></td>";
-      h+="<td>"+sIcon(s)+"</td>";
-      h+='<td class="fw-medium">'+esc(r.testId)+"</td>";
-      if(tgtNames.length>1)h+="<td>"+esc(r.target)+"</td>";
-      h+='<td class="'+sCls(r.score)+'">'+fmtPct(r.score)+"</td>";
-      for(var j=0;j<evalNames.length;j++){
-        var es=getEvalScore(r,evalNames[j]);
-        h+='<td class="'+sCls(es)+'">'+(es!=null?fmtPct(es):"\\u2014")+"</td>";
-      }
-      h+="<td>"+fmtDur(r.durationMs)+"</td><td>"+fmtCost(r.costUsd)+"</td></tr>";
-      if(exp)h+='<tr class="detail-row"><td colspan="'+colSpan+'">'+renderDetail(r)+"</td></tr>";
-    }
-    if(rows.length===0)h+='<tr><td colspan="'+colSpan+'" class="empty-state">No matching tests</td></tr>';
-    tbody.innerHTML=h;
-
-    /* row click */
-    var trs=tbody.querySelectorAll(".test-row");
-    for(var k=0;k<trs.length;k++){
-      trs[k].addEventListener("click",(function(tr){return function(){
-        var key=tr.getAttribute("data-key");
-        state.expanded[key]=!state.expanded[key];
-        renderRows();
-      };})(trs[k]));
+  function inferLanguage(filePath) {
+    const ext = path9.extname(filePath).toLowerCase();
+    const langMap = {
+      ".json": "json",
+      ".jsonl": "json",
+      ".ts": "typescript",
+      ".tsx": "typescript",
+      ".js": "javascript",
+      ".jsx": "javascript",
+      ".md": "markdown",
+      ".yaml": "yaml",
+      ".yml": "yaml",
+      ".log": "plaintext",
+      ".txt": "plaintext",
+      ".py": "python",
+      ".sh": "shell",
+      ".bash": "shell",
+      ".css": "css",
+      ".html": "html",
+      ".xml": "xml",
+      ".svg": "xml",
+      ".toml": "toml",
+      ".diff": "diff",
+      ".patch": "diff"
+    };
+    return langMap[ext] ?? "plaintext";
+  }
+  app2.get("/api/runs/:filename/evals/:evalId/files", (c3) => {
+    const filename = c3.req.param("filename");
+    const evalId = c3.req.param("evalId");
+    const metas = listResultFiles(searchDir);
+    const meta = metas.find((m) => m.filename === filename);
+    if (!meta) {
+      return c3.json({ error: "Run not found" }, 404);
     }
-
-    /* wire feedback buttons */
-    var btns=tbody.querySelectorAll(".feedback-submit");
-    for(var k=0;k<btns.length;k++){
-      btns[k].addEventListener("click",(function(btn){return function(ev){
-        ev.stopPropagation();
-        var tid=btn.getAttribute("data-test-id");
-        var sec=btn.closest(".feedback-section");
-        var ta=sec.querySelector(".feedback-input");
-        var st=sec.querySelector(".feedback-status");
-        saveFeedback(tid,ta.value,st,btn);
-      };})(btns[k]));
+    try {
+      const content = readFileSync8(meta.path, "utf8");
+      const records = parseResultManifest(content);
+      const record = records.find((r) => (r.test_id ?? r.eval_id) === evalId);
+      if (!record) {
+        return c3.json({ error: "Eval not found" }, 404);
+      }
+      const baseDir = path9.dirname(meta.path);
+      const knownPaths = [
+        record.grading_path,
+        record.timing_path,
+        record.input_path,
+        record.output_path,
+        record.response_path
+      ].filter((p) => !!p);
+      if (knownPaths.length === 0) {
+        return c3.json({ files: [] });
+      }
+      const artifactDirs = knownPaths.map((p) => path9.dirname(p));
+      let commonDir = artifactDirs[0];
+      for (const dir of artifactDirs) {
+        while (!dir.startsWith(commonDir)) {
+          commonDir = path9.dirname(commonDir);
+        }
+      }
+      const artifactAbsDir = path9.join(baseDir, commonDir);
+      const files = buildFileTree(artifactAbsDir, baseDir);
+      return c3.json({ files });
+    } catch {
+      return c3.json({ error: "Failed to load file tree" }, 500);
     }
-
-    /* prevent textarea clicks from toggling row */
-    var tas=tbody.querySelectorAll(".feedback-input");
-    for(var k=0;k<tas.length;k++){
-      tas[k].addEventListener("click",function(ev){ev.stopPropagation();});
+  });
+  app2.get("/api/runs/:filename/evals/:evalId/files/*", (c3) => {
+    const filename = c3.req.param("filename");
+    const evalId = c3.req.param("evalId");
+    const metas = listResultFiles(searchDir);
+    const meta = metas.find((m) => m.filename === filename);
+    if (!meta) {
+      return c3.json({ error: "Run not found" }, 404);
     }
-
-    populateFeedbackTextareas();
-  }
-
-  /* ---- detail panel ---- */
-  function renderDetail(r){
-    var h='<div class="detail-panel">';
-
-    /* input / output */
-    h+='<div class="detail-grid">';
-    if(r.input!=null){
-      h+='<div class="detail-block"><h4>Input</h4><pre class="detail-pre">'+esc(JSON.stringify(r.input,null,2))+"</pre></div>";
+    const requestPath = c3.req.path;
+    const prefix = `/api/runs/${filename}/evals/${evalId}/files/`;
+    const filePath = requestPath.slice(prefix.length);
+    if (!filePath) {
+      return c3.json({ error: "No file path specified" }, 400);
     }
-    h+='<div class="detail-block"><h4>Output</h4><pre class="detail-pre">'+esc(r.output?JSON.stringify(r.output,null,2):"")+"</pre></div>";
-    h+="</div>";
-
-    /* evaluator results */
-    if(r.scores&&r.scores.length>0){
-      h+="<h4>Evaluator Results</h4>";
-      h+='<table class="eval-table"><thead><tr><th>Evaluator</th><th>Score</th><th>Status</th><th>Assertions</th></tr></thead><tbody>';
-      for(var i=0;i<r.scores.length;i++){
-        var ev=r.scores[i],evS=ev.score>=0.5?"pass":"fail";
-        var evAssertions=ev.assertions||[];
-        var evSummary=evAssertions.map(function(a){return (a.passed?"\\u2713 ":"\\u2717 ")+a.text;}).join("; ");
-        h+="<tr><td class=\\"fw-medium\\">"+esc(ev.name)+'</td><td class="'+sCls(ev.score)+'">'+fmtPct(ev.score)+"</td><td>"+sIcon(evS)+'</td><td class="reasoning-cell">'+esc(evSummary)+"</td></tr>";
-      }
-      h+="</tbody></table>";
+    const baseDir = path9.dirname(meta.path);
+    const absolutePath = path9.resolve(baseDir, filePath);
+    if (!absolutePath.startsWith(path9.resolve(baseDir) + path9.sep) && absolutePath !== path9.resolve(baseDir)) {
+      return c3.json({ error: "Path traversal not allowed" }, 403);
     }
-
-    /* assertions */
-    var passedA=r.assertions?r.assertions.filter(function(a){return a.passed;}):[];
-    var failedA=r.assertions?r.assertions.filter(function(a){return !a.passed;}):[];
-    if(passedA.length>0){
-      h+='<h4>Passed Assertions</h4><ul class="expect-list pass">';
-      for(var i=0;i<passedA.length;i++)h+="<li>"+esc(passedA[i].text)+(passedA[i].evidence?" <span class=\\"reasoning-cell\\">("+esc(passedA[i].evidence)+")</span>":"")+"</li>";
-      h+="</ul>";
-    }
-    if(failedA.length>0){
-      h+='<h4>Failed Assertions</h4><ul class="expect-list fail">';
-      for(var i=0;i<failedA.length;i++)h+="<li>"+esc(failedA[i].text)+(failedA[i].evidence?" <span class=\\"reasoning-cell\\">("+esc(failedA[i].evidence)+")</span>":"")+"</li>";
-      h+="</ul>";
+    if (!existsSync7(absolutePath) || !statSync4(absolutePath).isFile()) {
+      return c3.json({ error: "File not found" }, 404);
     }
-
-    /* tool calls */
-    if(r._toolCalls){
-      var tc=r._toolCalls,tcArr=[];
-      for(var k in tc)tcArr.push({name:k,count:tc[k]});
-      tcArr.sort(function(a,b){return b.count-a.count;});
-      h+='<h4>Tool Calls</h4><div class="tool-calls">';
-      for(var i=0;i<tcArr.length;i++)h+='<span class="tool-tag">'+esc(tcArr[i].name)+": "+tcArr[i].count+"</span>";
-      h+="</div>";
+    try {
+      const fileContent = readFileSync8(absolutePath, "utf8");
+      const language = inferLanguage(absolutePath);
+      return c3.json({ content: fileContent, language });
+    } catch {
+      return c3.json({ error: "Failed to read file" }, 500);
     }
-
-    /* error */
-    if(r.error)h+='<div class="error-box"><h4>Error</h4><pre>'+esc(r.error)+"</pre></div>";
-
-    /* metadata */
-    h+='<div class="detail-meta">';
-    var m=[];
-    if(r.tokenUsage)m.push(fmtTok(r.tokenUsage.input)+" in / "+fmtTok(r.tokenUsage.output)+" out tokens");
-    if(r.durationMs){
-      if(r._graderDurationMs>0){
-        var execMs=r.durationMs-r._graderDurationMs;
-        m.push(fmtDur(execMs>0?execMs:0)+" executor + "+fmtDur(r._graderDurationMs)+" grader");
-      }else{
-        m.push(fmtDur(r.durationMs));
-      }
-    }
-    if(r.target)m.push(r.target);
-    if(r.costUsd)m.push(fmtCost(r.costUsd));
-    if(r.timestamp)m.push(r.timestamp);
-    h+=esc(m.join(" \\u00B7 "));
-    h+="</div>";
-
-    /* feedback section */
-    var tid=r.testId||"";
-    var existingComment=feedbackCache[tid]||"";
-    h+='<div class="feedback-section">';
-    h+='<h4>Feedback</h4>';
-    h+='<textarea class="feedback-input" data-test-id="'+esc(tid)+'" placeholder="Add feedback for this test..." onclick="event.stopPropagation()">'+esc(existingComment)+'</textarea>';
-    h+='<div style="display:flex;align-items:center">';
-    h+='<button class="feedback-submit" data-test-id="'+esc(tid)+'">Save Feedback</button>';
-    h+='<span class="feedback-status"></span>';
-    h+='</div></div>';
-
-    h+="</div>";
-    return h;
-  }
-
-  /* ---- run picker ---- */
-  var runPicker=document.getElementById("run-picker");
-  var knownRunFilenames=[];
-
-  function refreshRunList(){
-    fetch("/api/runs").then(function(r){return r.json();}).then(function(d){
-      if(!d||!d.runs)return;
-      var runs=d.runs;
-      var newFilenames=runs.map(function(r){return r.filename;});
-
-      /* Detect new runs that appeared since last poll */
-      if(knownRunFilenames.length>0){
-        var hasNew=newFilenames.some(function(f){return knownRunFilenames.indexOf(f)===-1;});
-        if(hasNew&&DATA.length===0){
-          /* Auto-load the first (most recent) run when starting from empty state */
-          loadRun(runs[0].filename);
+  });
+  app2.get("/api/experiments", (c3) => {
+    const metas = listResultFiles(searchDir);
+    const experimentMap = /* @__PURE__ */ new Map();
+    for (const m of metas) {
+      try {
+        const records = loadLightweightResults(m.path);
+        for (const r of records) {
+          const experiment = r.experiment ?? "default";
+          const entry = experimentMap.get(experiment) ?? {
+            targets: /* @__PURE__ */ new Set(),
+            runFilenames: /* @__PURE__ */ new Set(),
+            evalCount: 0,
+            passedCount: 0,
+            lastTimestamp: ""
+          };
+          entry.runFilenames.add(m.filename);
+          if (r.target) entry.targets.add(r.target);
+          entry.evalCount++;
+          if (r.score >= 1) entry.passedCount++;
+          if (r.timestamp && r.timestamp > entry.lastTimestamp) {
+            entry.lastTimestamp = r.timestamp;
+          }
+          experimentMap.set(experiment, entry);
         }
+      } catch {
       }
-      knownRunFilenames=newFilenames;
-
-      /* Rebuild picker options */
-      var h='<option value="">Select a result file...</option>';
-      if(runs.length===0){
-        h='<option value="">No result files</option>';
-      }
-      for(var i=0;i<runs.length;i++){
-        var r=runs[i];
-        var label=r.filename+" ("+r.test_count+" tests, "+(r.pass_rate*100).toFixed(0)+"% pass)";
-        h+='<option value="'+esc(r.filename)+'">'+esc(label)+"</option>";
+    }
+    const experiments = [...experimentMap.entries()].map(([name, entry]) => ({
+      name,
+      run_count: entry.runFilenames.size,
+      target_count: entry.targets.size,
+      eval_count: entry.evalCount,
+      passed_count: entry.passedCount,
+      pass_rate: entry.evalCount > 0 ? entry.passedCount / entry.evalCount : 0,
+      last_run: entry.lastTimestamp || null
+    }));
+    return c3.json({ experiments });
+  });
+  app2.get("/api/targets", (c3) => {
+    const metas = listResultFiles(searchDir);
+    const targetMap = /* @__PURE__ */ new Map();
+    for (const m of metas) {
+      try {
+        const records = loadLightweightResults(m.path);
+        for (const r of records) {
+          const target = r.target ?? "default";
+          const entry = targetMap.get(target) ?? {
+            experiments: /* @__PURE__ */ new Set(),
+            runFilenames: /* @__PURE__ */ new Set(),
+            evalCount: 0,
+            passedCount: 0
+          };
+          entry.runFilenames.add(m.filename);
+          if (r.experiment) entry.experiments.add(r.experiment);
+          entry.evalCount++;
+          if (r.score >= 1) entry.passedCount++;
+          targetMap.set(target, entry);
+        }
+      } catch {
       }
-      runPicker.innerHTML=h;
-      /* Pre-select the initially loaded run */
-      if(INITIAL_SOURCE&&runs.length>0){
-        runPicker.value=INITIAL_SOURCE;
+    }
+    const targets = [...targetMap.entries()].map(([name, entry]) => ({
+      name,
+      run_count: entry.runFilenames.size,
+      experiment_count: entry.experiments.size,
+      eval_count: entry.evalCount,
+      passed_count: entry.passedCount,
+      pass_rate: entry.evalCount > 0 ? entry.passedCount / entry.evalCount : 0
+    }));
+    return c3.json({ targets });
+  });
+  app2.get("/assets/*", (c3) => {
+    const assetPath = c3.req.path;
+    const filePath = path9.join(studioDistPath, assetPath);
+    if (!existsSync7(filePath)) {
+      return c3.notFound();
+    }
+    const content = readFileSync8(filePath);
+    const ext = path9.extname(filePath);
+    const mimeTypes = {
+      ".js": "application/javascript",
+      ".css": "text/css",
+      ".html": "text/html",
+      ".json": "application/json",
+      ".svg": "image/svg+xml",
+      ".png": "image/png",
+      ".woff2": "font/woff2",
+      ".woff": "font/woff"
+    };
+    const contentType = mimeTypes[ext] ?? "application/octet-stream";
+    return new Response(content, {
+      headers: {
+        "Content-Type": contentType,
+        "Cache-Control": "public, max-age=31536000, immutable"
       }
-    }).catch(function(err){console.warn("Failed to refresh run list:",err);});
-  }
-
-  function loadRun(filename){
-    fetch("/api/runs/"+encodeURIComponent(filename)).then(function(r){return r.json();}).then(function(d){
-      if(d.error){console.error(d.error);return;}
-      DATA=d.results;
-      stats=computeStats(DATA);
-      tgtStats=computeTargets(DATA);
-      tgtNames=tgtStats.map(function(t){return t.target;});
-      state.expanded={};
-      feedbackCache={};
-      loadFeedback();
-      render();
-      /* Update picker selection */
-      runPicker.value=filename;
-    }).catch(function(err){console.error("Failed to load run:",err);});
+    });
+  });
+  app2.get("*", (c3) => {
+    if (c3.req.path.startsWith("/api/")) {
+      return c3.json({ error: "Not found" }, 404);
+    }
+    const indexPath = path9.join(studioDistPath, "index.html");
+    if (existsSync7(indexPath)) {
+      return c3.html(readFileSync8(indexPath, "utf8"));
+    }
+    return c3.notFound();
+  });
+  return app2;
+}
+function resolveStudioDistDir() {
+  const currentDir = typeof __dirname !== "undefined" ? __dirname : path9.dirname(fileURLToPath2(import.meta.url));
+  const candidates = [
+    // From src/commands/results/ → sibling apps/studio/dist
+    path9.resolve(currentDir, "../../../../studio/dist"),
+    // From dist/ → sibling apps/studio/dist (monorepo dev)
+    path9.resolve(currentDir, "../../studio/dist"),
+    // Bundled inside CLI dist (published package)
+    path9.resolve(currentDir, "../studio"),
+    // From dist/ in monorepo root context
+    path9.resolve(currentDir, "../../../apps/studio/dist")
+  ];
+  for (const candidate of candidates) {
+    if (existsSync7(candidate) && existsSync7(path9.join(candidate, "index.html"))) {
+      return candidate;
+    }
   }
-
-  runPicker.addEventListener("change",function(){
-    var val=runPicker.value;
-    if(val)loadRun(val);
+  return void 0;
+}
+function stripHeavyFields(results) {
+  return results.map((r) => {
+    const { requests, trace, ...rest } = r;
+    const toolCalls = trace?.toolCalls && Object.keys(trace.toolCalls).length > 0 ? trace.toolCalls : void 0;
+    const graderDurationMs = (r.scores ?? []).reduce((sum, s) => sum + (s.durationMs ?? 0), 0);
+    return {
+      ...rest,
+      ...toolCalls && { _toolCalls: toolCalls },
+      ...graderDurationMs > 0 && { _graderDurationMs: graderDurationMs }
+    };
   });
-
-  /* Poll for new result files every 5 seconds */
-  refreshRunList();
-  setInterval(refreshRunList,5000);
-
-  /* ---- init ---- */
-  loadFeedback();
-  render();
-})();
-`;
+}
 var resultsServeCommand = command({
-  name: "serve",
-  description: "Start a local HTTP server to review evaluation results",
+  name: "studio",
+  description: "Start AgentV Studio \u2014 a local dashboard for reviewing evaluation results",
   args: {
     source: positional({
       type: optional(string),
@@ -7594,7 +7404,7 @@ 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.eval_set ? `  ${c2.dim}eval-set: ${result.eval_set}${c2.reset}` : ""}`
+    `${c2.bold}${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}` : ""}`
   );
   if (result.error) {
     lines.push(`  ${c2.red}Error: ${result.error}${c2.reset}`);
@@ -7768,8 +7578,8 @@ function groupResults(results, groupBy2) {
       case "target":
         key = result.target ?? "unknown";
         break;
-      case "eval-set":
-        key = result.eval_set ?? "unknown";
+      case "dataset":
+        key = result.dataset ?? "unknown";
         break;
       case "test-id":
         key = result.test_id ?? result.eval_id ?? "unknown";
@@ -8482,7 +8292,7 @@ var app = subcommands({
     pipeline: pipelineCommand,
     results: resultsCommand,
     self: selfCommand,
-    serve: resultsServeCommand,
+    studio: resultsServeCommand,
     trace: traceCommand,
     transpile: transpileCommand,
     trim: trimCommand,
@@ -8500,7 +8310,7 @@ var TOP_LEVEL_COMMANDS = /* @__PURE__ */ new Set([
   "pipeline",
   "results",
   "self",
-  "serve",
+  "studio",
   "trace",
   "transpile",
   "trim",
@@ -8547,4 +8357,4 @@ export {
   preprocessArgv,
   runCli
 };
-//# sourceMappingURL=chunk-CQRWNXVG.js.map
+//# sourceMappingURL=chunk-2W5JKKXC.js.map