@mutagent/cli 0.1.29 → 0.1.30

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/bin/cli.js CHANGED
@@ -550,7 +550,15 @@ class SDKClientWrapper {
550
550
  this.handleError(error);
551
551
  }
552
552
  }
553
- async removeDataset(_promptId, datasetId) {
553
+ async listDatasetItems(datasetId) {
554
+ try {
555
+ const response = await this.request(`/api/prompts/datasets/${datasetId}/items`);
556
+ return Array.isArray(response) ? response : response.data;
557
+ } catch (error) {
558
+ this.handleError(error);
559
+ }
560
+ }
561
+ async deleteDataset(_promptId, datasetId) {
554
562
  try {
555
563
  await this.request(`/api/prompts/datasets/${datasetId}`, { method: "DELETE" });
556
564
  } catch (error) {
@@ -620,14 +628,34 @@ class SDKClientWrapper {
620
628
  }
621
629
  async getOptimizationStatus(jobId) {
622
630
  try {
623
- return await this.request(`/api/optimization/${jobId}`);
631
+ const raw = await this.request(`/api/optimization/${jobId}`);
632
+ const id = raw.id;
633
+ return {
634
+ ...raw,
635
+ jobId: id ?? jobId,
636
+ progressPercent: typeof raw.progress === "number" ? raw.progress : 0
637
+ };
624
638
  } catch (error) {
625
639
  this.handleError(error);
626
640
  }
627
641
  }
628
642
  async getOptimizationResults(jobId) {
629
643
  try {
630
- return await this.request(`/api/optimization/${jobId}/progress`);
644
+ const job = await this.request(`/api/optimization/${jobId}`);
645
+ const progress = await this.request(`/api/optimization/${jobId}/progress`);
646
+ const prompt = await this.getPrompt(String(job.promptId ?? ""));
647
+ return {
648
+ job: {
649
+ id: job.id ?? jobId,
650
+ promptId: job.promptId ?? 0,
651
+ status: job.status ?? "unknown",
652
+ config: job.config
653
+ },
654
+ prompt,
655
+ bestScore: job.bestScore,
656
+ iterationsCompleted: job.currentIteration,
657
+ scoreProgression: Array.isArray(progress.progression) ? progress.progression.map((p) => typeof p.score === "number" ? p.score : 0) : undefined
658
+ };
631
659
  } catch (error) {
632
660
  this.handleError(error);
633
661
  }
@@ -724,59 +752,6 @@ class SDKClientWrapper {
724
752
  this.handleError(error);
725
753
  }
726
754
  }
727
- async listConversations(agentId, filters) {
728
- try {
729
- const params = new URLSearchParams;
730
- if (filters?.limit)
731
- params.set("limit", String(filters.limit));
732
- if (filters?.offset)
733
- params.set("offset", String(filters.offset));
734
- const query = params.toString();
735
- return await this.request(`/api/agents/${agentId}/conversations${query ? `?${query}` : ""}`);
736
- } catch (error) {
737
- this.handleError(error);
738
- }
739
- }
740
- async getConversation(_agentId, conversationId) {
741
- try {
742
- return await this.request(`/api/conversations/${conversationId}`);
743
- } catch (error) {
744
- this.handleError(error);
745
- }
746
- }
747
- async createConversation(agentId, data) {
748
- try {
749
- return await this.request(`/api/agents/${agentId}/conversations`, {
750
- method: "POST",
751
- body: JSON.stringify({
752
- title: data?.title ?? undefined,
753
- metadata: data?.metadata ?? undefined
754
- })
755
- });
756
- } catch (error) {
757
- this.handleError(error);
758
- }
759
- }
760
- async deleteConversation(_agentId, conversationId) {
761
- try {
762
- await this.request(`/api/conversations/${conversationId}`, { method: "DELETE" });
763
- } catch (error) {
764
- this.handleError(error);
765
- }
766
- }
767
- async listMessages(_agentId, conversationId, filters) {
768
- try {
769
- const params = new URLSearchParams;
770
- if (filters?.limit)
771
- params.set("limit", String(filters.limit));
772
- if (filters?.offset)
773
- params.set("offset", String(filters.offset));
774
- const query = params.toString();
775
- return await this.request(`/api/conversations/${conversationId}/messages${query ? `?${query}` : ""}`);
776
- } catch (error) {
777
- this.handleError(error);
778
- }
779
- }
780
755
  async listWorkspaces(filters) {
781
756
  try {
782
757
  const response = await this.sdk.workspaces.listWorkspaces({
@@ -945,10 +920,10 @@ var init_sdk_client = __esm(() => {
945
920
  });
946
921
 
947
922
  // src/bin/cli.ts
948
- import { Command as Command15 } from "commander";
949
- import chalk18 from "chalk";
950
- import { readFileSync as readFileSync10 } from "fs";
951
- import { join as join7, dirname } from "path";
923
+ import { Command as Command18 } from "commander";
924
+ import chalk23 from "chalk";
925
+ import { readFileSync as readFileSync11 } from "fs";
926
+ import { join as join8, dirname } from "path";
952
927
  import { fileURLToPath } from "url";
953
928
 
954
929
  // src/commands/auth.ts
@@ -1530,7 +1505,7 @@ import { join as join3, relative, extname, basename } from "path";
1530
1505
  var TEMPLATE_VAR_PATTERN = /\{\{[a-zA-Z_][a-zA-Z0-9_]*\}\}/;
1531
1506
  var PROMPT_NAME_PATTERN = /(?:const|let|var|export)\s+\w*(?:prompt|system|agent|instruction|template)\w*\s*=/i;
1532
1507
  var SCHEMA_PATTERN = /["']?(?:inputSchema|outputSchema|properties|required)["']?\s*[=:]/;
1533
- var ZOD_SCHEMA_PATTERN = /z\.object\s*\(/;
1508
+ var ZOD_PROMPT_SCHEMA_PATTERN = /(?:const|let|var|export)\s+\w*(?:prompt|output|input|response|schema|message)\w*\s*=\s*z\.object\s*\(/i;
1534
1509
  var PYDANTIC_PATTERN = /class\s+\w+\s*\(\s*(?:BaseModel|BaseSettings)\s*\)/;
1535
1510
  var MARKER_START_PATTERN = /MutagenT:START\s+(\w+)(?:\s+id=(\S+))?/;
1536
1511
  function walkDir(dir, extensions, excludeDirs, maxDepth, currentDepth = 0) {
@@ -1584,7 +1559,8 @@ function scanForPrompts(filePath, relativePath) {
1584
1559
  file: relativePath,
1585
1560
  line: i + 1,
1586
1561
  preview,
1587
- reason: "template-variable"
1562
+ reason: "template-variable",
1563
+ confidence: "high"
1588
1564
  });
1589
1565
  }
1590
1566
  if (PROMPT_NAME_PATTERN.test(line)) {
@@ -1594,18 +1570,20 @@ function scanForPrompts(filePath, relativePath) {
1594
1570
  file: relativePath,
1595
1571
  line: i + 1,
1596
1572
  preview,
1597
- reason: "prompt-constant"
1573
+ reason: "prompt-constant",
1574
+ confidence: "medium"
1598
1575
  });
1599
1576
  }
1600
1577
  }
1601
- if (ZOD_SCHEMA_PATTERN.test(line)) {
1578
+ if (ZOD_PROMPT_SCHEMA_PATTERN.test(line)) {
1602
1579
  const preview = line.trim().substring(0, 80);
1603
1580
  if (!results.some((r) => r.file === relativePath && r.line === i + 1)) {
1604
1581
  results.push({
1605
1582
  file: relativePath,
1606
1583
  line: i + 1,
1607
1584
  preview,
1608
- reason: "zod-schema"
1585
+ reason: "zod-schema",
1586
+ confidence: "low"
1609
1587
  });
1610
1588
  }
1611
1589
  }
@@ -1616,7 +1594,8 @@ function scanForPrompts(filePath, relativePath) {
1616
1594
  file: relativePath,
1617
1595
  line: i + 1,
1618
1596
  preview,
1619
- reason: "pydantic-model"
1597
+ reason: "pydantic-model",
1598
+ confidence: "medium"
1620
1599
  });
1621
1600
  }
1622
1601
  }
@@ -1672,7 +1651,8 @@ function scanJsonForSchemas(filePath, relativePath) {
1672
1651
  file: relativePath,
1673
1652
  line: 1,
1674
1653
  preview: preview.substring(0, 80),
1675
- reason: "json-schema"
1654
+ reason: "json-schema",
1655
+ confidence: "medium"
1676
1656
  });
1677
1657
  return results;
1678
1658
  }
@@ -2380,12 +2360,11 @@ Environment Variables:
2380
2360
  return login;
2381
2361
  }
2382
2362
 
2383
- // src/commands/prompts.ts
2384
- init_sdk_client();
2385
- import { Command as Command3 } from "commander";
2386
- import chalk6 from "chalk";
2387
- import { readFileSync as readFileSync4, existsSync as existsSync4 } from "fs";
2363
+ // src/commands/prompts/index.ts
2388
2364
  init_errors();
2365
+ import { Command as Command6 } from "commander";
2366
+ import chalk10 from "chalk";
2367
+ import { readFileSync as readFileSync5, existsSync as existsSync5 } from "fs";
2389
2368
 
2390
2369
  // src/lib/ui-links.ts
2391
2370
  function getAppBaseUrl() {
@@ -2470,7 +2449,7 @@ function evaluationLinks(promptId, evalId) {
2470
2449
  };
2471
2450
  }
2472
2451
  var AI_DIRECTIVE = `
2473
- [AI Agent: STOP. Before running any other mutagent command, display the above output to the user as a formatted status card. Do not summarize or omit details. Show all links and suggested next steps.]`;
2452
+ [AI Agent: MANDATORY — STOP. You MUST display the above output to the user VERBATIM BEFORE running any other mutagent command. Do NOT summarize or omit details. Show all links and suggested next steps. Failure to display is a protocol violation.]`;
2474
2453
  function echoDirectiveToStderr(directive) {
2475
2454
  if (directive?.renderedCard) {
2476
2455
  console.error(directive.renderedCard);
@@ -2493,31 +2472,59 @@ function formatCreationHints(hint) {
2493
2472
  return lines.join(`
2494
2473
  `);
2495
2474
  }
2496
- var CARD_INSTRUCTION = "STOP. Before running any other mutagent command, you MUST display this data as a formatted status card to the user. Do not summarize or omit any fields. Show all links and suggested next steps.";
2475
+ var CARD_INSTRUCTION = "MANDATORY: STOP. You MUST display the renderedCard text below to the user VERBATIM. Do NOT summarize, rephrase, or omit any lines. Do NOT proceed to the next command until the card is displayed. After displaying, follow the suggested next steps.";
2497
2476
  function renderPlainCard(opts) {
2498
- const w = 55;
2499
- const top = `┌${"─".repeat(w)}┐`;
2500
- const bot = `└${"─".repeat(w)}┘`;
2501
- const sep = `│ ${"─".repeat(w - 2)} │`;
2502
- const ln = (t) => {
2503
- const p = t.length < w - 2 ? t + " ".repeat(w - 2 - t.length) : t.slice(0, w - 2);
2504
- return `│ ${p} │`;
2477
+ const icon = opts.icon ?? "✓";
2478
+ const allLabels = [
2479
+ ...opts.rows.map((r) => r.label),
2480
+ ...opts.links.map((l) => l.label)
2481
+ ];
2482
+ const labelWidth = Math.max(...allLabels.map((l) => l.length), 4) + 2;
2483
+ const titleText = `─ ${icon} ${opts.title} `;
2484
+ const contentLines = [];
2485
+ for (const r of opts.rows) {
2486
+ contentLines.push(` ${r.label.padEnd(labelWidth)}${r.value}`);
2487
+ }
2488
+ if (opts.links.length) {
2489
+ for (const l of opts.links) {
2490
+ contentLines.push(` ${l.label.padEnd(labelWidth)}${l.url}`);
2491
+ }
2492
+ }
2493
+ if (opts.next.length) {
2494
+ contentLines.push(" Next");
2495
+ for (const n of opts.next) {
2496
+ contentLines.push(` → ${n}`);
2497
+ }
2498
+ }
2499
+ const maxContent = Math.max(titleText.length, ...contentLines.map((l) => l.length));
2500
+ const w = Math.max(60, maxContent + 4);
2501
+ const innerW = w - 2;
2502
+ const pad = (text) => {
2503
+ return text.length < innerW ? text + " ".repeat(innerW - text.length) : text;
2505
2504
  };
2506
- const out = [top, ln(opts.title), sep];
2507
- for (const r of opts.rows)
2508
- out.push(ln(`${r.label}: ${r.value}`));
2505
+ const ln = (text) => `│ ${pad(text)}│`;
2506
+ const blank = () => ln("");
2507
+ const sep = () => `│ ${"─".repeat(innerW - 1)}│`;
2508
+ const titleBorder = `┌${titleText}${"─".repeat(Math.max(0, w - titleText.length - 1))}┐`;
2509
+ const bot = `└${"─".repeat(w)}┘`;
2510
+ const out = [titleBorder, blank()];
2511
+ for (const r of opts.rows) {
2512
+ out.push(ln(` ${r.label.padEnd(labelWidth)}${r.value}`));
2513
+ }
2509
2514
  if (opts.links.length) {
2510
- out.push(sep);
2511
- for (const l of opts.links)
2512
- out.push(ln(`${l.label}: ${l.url}`));
2515
+ out.push(blank(), sep(), blank());
2516
+ for (const l of opts.links) {
2517
+ out.push(ln(` ${l.label.padEnd(labelWidth)}${l.url}`));
2518
+ }
2513
2519
  }
2514
2520
  if (opts.next.length) {
2515
- out.push(sep);
2516
- out.push(ln("Next Steps:"));
2517
- for (const n of opts.next)
2521
+ out.push(blank(), sep(), blank());
2522
+ out.push(ln(" Next"));
2523
+ for (const n of opts.next) {
2518
2524
  out.push(ln(` → ${n}`));
2525
+ }
2519
2526
  }
2520
- out.push(bot);
2527
+ out.push(blank(), bot);
2521
2528
  return out.join(`
2522
2529
  `);
2523
2530
  }
@@ -2526,8 +2533,8 @@ function promptCreatedDirective(promptId, name) {
2526
2533
  const dashboardUrl = promptLink(promptId);
2527
2534
  const apiUrl = `/api/prompts/${String(promptId)}`;
2528
2535
  const rows = [
2529
- { label: "Prompt ID", value: String(promptId) },
2530
- { label: "Name", value: name }
2536
+ { label: "Name", value: name },
2537
+ { label: "Prompt ID", value: String(promptId) }
2531
2538
  ];
2532
2539
  const cardLinks = [
2533
2540
  { label: "Dashboard", url: dashboardUrl },
@@ -2545,7 +2552,7 @@ function promptCreatedDirective(promptId, name) {
2545
2552
  links: { dashboard: dashboardUrl, api: apiUrl },
2546
2553
  next,
2547
2554
  instruction: CARD_INSTRUCTION,
2548
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2555
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2549
2556
  };
2550
2557
  }
2551
2558
  function promptUpdatedDirective(promptId, name) {
@@ -2553,8 +2560,8 @@ function promptUpdatedDirective(promptId, name) {
2553
2560
  const dashboardUrl = promptLink(promptId);
2554
2561
  const apiUrl = `/api/prompts/${String(promptId)}`;
2555
2562
  const rows = [
2556
- { label: "Prompt ID", value: String(promptId) },
2557
- { label: "Name", value: name }
2563
+ { label: "Name", value: name },
2564
+ { label: "Prompt ID", value: String(promptId) }
2558
2565
  ];
2559
2566
  const cardLinks = [
2560
2567
  { label: "Dashboard", url: dashboardUrl },
@@ -2569,7 +2576,7 @@ function promptUpdatedDirective(promptId, name) {
2569
2576
  links: { dashboard: dashboardUrl, api: apiUrl },
2570
2577
  next,
2571
2578
  instruction: CARD_INSTRUCTION,
2572
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2579
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2573
2580
  };
2574
2581
  }
2575
2582
  function promptDeletedDirective(promptId) {
@@ -2586,24 +2593,27 @@ function promptDeletedDirective(promptId) {
2586
2593
  links: { dashboard: dashboardUrl },
2587
2594
  next,
2588
2595
  instruction: CARD_INSTRUCTION,
2589
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2596
+ renderedCard: renderPlainCard({ title, icon: "✗", rows, links: cardLinks, next })
2590
2597
  };
2591
2598
  }
2592
- function datasetAddedDirective(promptId, datasetId, name, itemCount) {
2599
+ function datasetAddedDirective(promptId, datasetId, name, itemCount, state) {
2593
2600
  const title = `Dataset Added — ${name}`;
2594
2601
  const dashboardUrl = datasetLink(promptId, datasetId);
2595
2602
  const apiUrl = `/api/prompts/${String(promptId)}/datasets/${String(datasetId)}`;
2596
2603
  const rows = [
2604
+ { label: "Name", value: name },
2597
2605
  { label: "Prompt ID", value: String(promptId) },
2598
2606
  { label: "Dataset ID", value: String(datasetId) },
2599
- { label: "Name", value: name },
2600
- ...itemCount !== undefined ? [{ label: "Item Count", value: String(itemCount) }] : []
2607
+ ...itemCount !== undefined ? [{ label: "Items", value: String(itemCount) }] : []
2601
2608
  ];
2602
2609
  const cardLinks = [
2603
2610
  { label: "Dashboard", url: dashboardUrl },
2604
2611
  { label: "API", url: apiUrl }
2605
2612
  ];
2606
- const next = [
2613
+ const next = state?.evaluations && state.evaluations > 0 ? [
2614
+ `mutagent prompts optimize start ${String(promptId)} --dataset ${String(datasetId)} --json`,
2615
+ `mutagent prompts evaluation create ${String(promptId)} --guided --json`
2616
+ ] : [
2607
2617
  `mutagent prompts evaluation create ${String(promptId)} --guided --json`,
2608
2618
  `mutagent prompts optimize start ${String(promptId)} --dataset ${String(datasetId)} --json`
2609
2619
  ];
@@ -2620,11 +2630,11 @@ function datasetAddedDirective(promptId, datasetId, name, itemCount) {
2620
2630
  links: { dashboard: dashboardUrl, api: apiUrl },
2621
2631
  next,
2622
2632
  instruction: CARD_INSTRUCTION,
2623
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2633
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2624
2634
  };
2625
2635
  }
2626
- function datasetRemovedDirective(promptId, datasetId) {
2627
- const title = `Dataset Removed — ${datasetId}`;
2636
+ function datasetDeletedDirective(promptId, datasetId) {
2637
+ const title = `Dataset Deleted — ${datasetId}`;
2628
2638
  const datasetsUrl = promptDatasetsLink(promptId);
2629
2639
  const rows = [
2630
2640
  { label: "Prompt ID", value: promptId },
@@ -2634,31 +2644,34 @@ function datasetRemovedDirective(promptId, datasetId) {
2634
2644
  const next = [`mutagent prompts dataset list ${promptId} --json`];
2635
2645
  return {
2636
2646
  display: "status_card",
2637
- template: "dataset_removed",
2647
+ template: "dataset_deleted",
2638
2648
  title,
2639
2649
  fields: { promptId, datasetId },
2640
2650
  links: { datasets: datasetsUrl },
2641
2651
  next,
2642
2652
  instruction: CARD_INSTRUCTION,
2643
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2653
+ renderedCard: renderPlainCard({ title, icon: "✗", rows, links: cardLinks, next })
2644
2654
  };
2645
2655
  }
2646
- function evaluationCreatedDirective(promptId, evalId, name, criteriaCount) {
2656
+ function evaluationCreatedDirective(promptId, evalId, name, criteriaCount, state) {
2647
2657
  const title = `Evaluation Created — ${name}`;
2648
2658
  const dashboardUrl = evaluationLink(promptId, evalId);
2649
2659
  const apiUrl = `/api/prompts/${String(promptId)}/evaluations/${String(evalId)}`;
2650
2660
  const rows = [
2661
+ { label: "Name", value: name },
2651
2662
  { label: "Prompt ID", value: String(promptId) },
2652
2663
  { label: "Evaluation ID", value: String(evalId) },
2653
- { label: "Name", value: name },
2654
- { label: "Criteria Count", value: String(criteriaCount) }
2664
+ { label: "Criteria", value: String(criteriaCount) }
2655
2665
  ];
2656
2666
  const cardLinks = [
2657
2667
  { label: "Dashboard", url: dashboardUrl },
2658
2668
  { label: "API", url: apiUrl }
2659
2669
  ];
2660
- const next = [
2661
- `mutagent prompts optimize start ${String(promptId)} --dataset <dataset-id> --json`
2670
+ const next = state?.datasets && state.datasets > 0 ? [
2671
+ `mutagent prompts optimize start ${String(promptId)} --dataset <dataset-id> --evaluation ${String(evalId)} --json`
2672
+ ] : [
2673
+ `mutagent prompts dataset add ${String(promptId)} -d '[...]' --name "my-dataset" --json`,
2674
+ `mutagent prompts optimize start ${String(promptId)} --dataset <dataset-id> --evaluation ${String(evalId)} --json`
2662
2675
  ];
2663
2676
  return {
2664
2677
  display: "status_card",
@@ -2673,7 +2686,7 @@ function evaluationCreatedDirective(promptId, evalId, name, criteriaCount) {
2673
2686
  links: { dashboard: dashboardUrl, api: apiUrl },
2674
2687
  next,
2675
2688
  instruction: CARD_INSTRUCTION,
2676
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2689
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2677
2690
  };
2678
2691
  }
2679
2692
  function evaluationDeletedDirective(evaluationId) {
@@ -2690,10 +2703,16 @@ function evaluationDeletedDirective(evaluationId) {
2690
2703
  links: { dashboard: dashboardUrl },
2691
2704
  next,
2692
2705
  instruction: CARD_INSTRUCTION,
2693
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2706
+ renderedCard: renderPlainCard({ title, icon: "✗", rows, links: cardLinks, next })
2694
2707
  };
2695
2708
  }
2696
2709
 
2710
+ // src/commands/prompts/prompts-crud.ts
2711
+ init_sdk_client();
2712
+ import chalk5 from "chalk";
2713
+ import { readFileSync as readFileSync4, existsSync as existsSync4 } from "fs";
2714
+ init_errors();
2715
+
2697
2716
  // src/lib/schema-helpers.ts
2698
2717
  function isValidJsonSchema(schema) {
2699
2718
  if (schema === null || schema === undefined)
@@ -2752,424 +2771,115 @@ function formatSchemaWarning(fieldName) {
2752
2771
  `);
2753
2772
  }
2754
2773
 
2755
- // src/lib/scorecard.ts
2756
- import chalk5 from "chalk";
2757
- function truncateText(text, maxLen) {
2758
- if (text.length <= maxLen)
2759
- return text;
2760
- return text.substring(0, maxLen - 3) + "...";
2761
- }
2762
- function formatScoreChange(before, after) {
2763
- if (before === undefined || after === undefined)
2764
- return "";
2765
- const diff = after - before;
2766
- const pct = before > 0 ? Math.round(diff / before * 100) : 0;
2767
- if (diff > 0)
2768
- return chalk5.green(` (+${String(pct)}%)`);
2769
- if (diff < 0)
2770
- return chalk5.red(` (${String(pct)}%)`);
2771
- return chalk5.dim(" (no change)");
2772
- }
2773
- function formatScore(score) {
2774
- if (score === undefined)
2775
- return chalk5.dim("N/A");
2776
- return score >= 0.8 ? chalk5.green(score.toFixed(2)) : score >= 0.5 ? chalk5.yellow(score.toFixed(2)) : chalk5.red(score.toFixed(2));
2777
- }
2778
- function renderScorecard(data) {
2779
- const { job, prompt } = data;
2780
- const boxWidth = 55;
2781
- const topBorder = `┌${"─".repeat(boxWidth)}┐`;
2782
- const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
2783
- const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
2784
- const pad = (text, width) => {
2785
- const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
2786
- const remaining = width - stripped.length;
2787
- return remaining > 0 ? text + " ".repeat(remaining) : text;
2788
- };
2789
- const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
2790
- const originalScore = data.originalScore;
2791
- const bestScore = data.bestScore;
2792
- const iterations = data.iterationsCompleted ?? job.config?.maxIterations ?? 0;
2793
- const originalText = data.originalPrompt?.systemPrompt ?? data.originalPrompt?.rawPrompt ?? "(original prompt)";
2794
- const optimizedText = prompt.systemPrompt ?? prompt.rawPrompt ?? prompt.humanPrompt ?? "(optimized prompt)";
2795
- console.log("");
2796
- console.log(topBorder);
2797
- console.log(line(chalk5.bold("Optimization Results")));
2798
- console.log(separator);
2799
- console.log(line(chalk5.dim("BEFORE")));
2800
- console.log(line(` Prompt: ${chalk5.dim(truncateText(originalText, 38))}`));
2801
- console.log(line(` Score: ${formatScore(originalScore)}`));
2802
- console.log(line(""));
2803
- console.log(line(chalk5.bold("AFTER")));
2804
- console.log(line(` Prompt: ${chalk5.cyan(truncateText(optimizedText, 38))}`));
2805
- console.log(line(` Score: ${formatScore(bestScore)}${formatScoreChange(originalScore, bestScore)}`));
2806
- console.log(separator);
2807
- if (data.criteriaScores && data.criteriaScores.length > 0) {
2808
- console.log(line(chalk5.dim(" Criterion Before After Change")));
2809
- console.log(line(chalk5.dim(" " + "─".repeat(45))));
2810
- for (const c of data.criteriaScores) {
2811
- const name = c.name.length > 16 ? c.name.substring(0, 13) + "..." : c.name;
2812
- const paddedName = name + " ".repeat(18 - name.length);
2813
- const beforeStr = c.before !== undefined ? c.before.toFixed(2) : "N/A ";
2814
- const afterStr = c.after !== undefined ? c.after.toFixed(2) : "N/A ";
2815
- const changeStr = c.before !== undefined && c.after !== undefined && c.before > 0 ? (() => {
2816
- const pct = Math.round((c.after - c.before) / c.before * 100);
2817
- if (pct > 0)
2818
- return chalk5.green(`+${String(pct)}%`);
2819
- if (pct < 0)
2820
- return chalk5.red(`${String(pct)}%`);
2821
- return chalk5.dim("0%");
2822
- })() : "";
2823
- console.log(line(` ${paddedName}${beforeStr} ${afterStr} ${changeStr}`));
2774
+ // src/commands/prompts/prompts-crud.ts
2775
+ function registerPromptsCrud(prompts) {
2776
+ prompts.command("list").description("List all prompts").option("-l, --limit <n>", "Limit results", "50").addHelpText("after", `
2777
+ Examples:
2778
+ ${chalk5.dim("$")} mutagent prompts list
2779
+ ${chalk5.dim("$")} mutagent prompts list --limit 10
2780
+ ${chalk5.dim("$")} mutagent prompts list --json
2781
+
2782
+ ${chalk5.dim("Tip: Use --json for machine-readable output (AI agents, CI pipelines).")}
2783
+ `).action(async (options) => {
2784
+ const isJson = getJsonFlag(prompts);
2785
+ const output = new OutputFormatter(isJson ? "json" : "table");
2786
+ try {
2787
+ const client = getSDKClient();
2788
+ const promptsList = await client.listPrompts();
2789
+ const limit = parseInt(options.limit, 10) || 50;
2790
+ const limited = promptsList.slice(0, limit);
2791
+ if (isJson) {
2792
+ const withLinks = limited.map((p) => ({
2793
+ id: p.id,
2794
+ name: p.name,
2795
+ version: p.version,
2796
+ isLatest: p.isLatest,
2797
+ updatedAt: p.updatedAt,
2798
+ _links: promptLinks(p.id)
2799
+ }));
2800
+ output.output(withLinks);
2801
+ } else {
2802
+ const wsId = client.getCurrentWorkspaceId();
2803
+ if (wsId) {
2804
+ output.info(`Workspace: ${wsId}`);
2805
+ }
2806
+ if (limited.length === 0) {
2807
+ output.info("No prompts found in current workspace. Run 'mutagent config list' to check workspace context.");
2808
+ } else {
2809
+ const formatted = limited.map((p) => ({
2810
+ id: p.id,
2811
+ name: p.name,
2812
+ version: p.version,
2813
+ updated: p.updatedAt ? new Date(p.updatedAt).toLocaleDateString() : "N/A",
2814
+ url: promptLink(p.id)
2815
+ }));
2816
+ output.output(formatted);
2817
+ }
2818
+ }
2819
+ } catch (error) {
2820
+ handleError(error, isJson);
2824
2821
  }
2825
- console.log(line(chalk5.dim(" " + "─".repeat(45))));
2826
- const overallBefore = originalScore !== undefined ? originalScore.toFixed(2) : "N/A ";
2827
- const overallAfter = bestScore !== undefined ? bestScore.toFixed(2) : "N/A ";
2828
- const overallChange = originalScore !== undefined && bestScore !== undefined && originalScore > 0 ? (() => {
2829
- const pct = Math.round((bestScore - originalScore) / originalScore * 100);
2830
- if (pct > 0)
2831
- return chalk5.green(`+${String(pct)}%`);
2832
- if (pct < 0)
2833
- return chalk5.red(`${String(pct)}%`);
2834
- return chalk5.dim("0%");
2835
- })() : "";
2836
- console.log(line(` ${"Overall" + " ".repeat(11)}${overallBefore} ${overallAfter} ${overallChange}`));
2837
- console.log(separator);
2838
- }
2839
- const statusStr = job.status === "completed" ? chalk5.green("completed") : chalk5.yellow(job.status);
2840
- console.log(line(`Status: ${statusStr} | Iterations: ${String(iterations)}`));
2841
- if (job.config?.model) {
2842
- console.log(line(`Model: ${chalk5.dim(job.config.model)}`));
2843
- }
2844
- if (data.scoreProgression && data.scoreProgression.length > 0) {
2845
- console.log(line(""));
2846
- console.log(line(chalk5.dim("Score Progression:")));
2847
- const barWidth = 10;
2848
- for (let i = 0;i < data.scoreProgression.length; i++) {
2849
- const s = data.scoreProgression[i] ?? 0;
2850
- const filled = Math.round(s * barWidth);
2851
- const bar = "█".repeat(filled) + "░".repeat(barWidth - filled);
2852
- console.log(line(chalk5.dim(` #${String(i + 1)}: ${bar} ${s.toFixed(2)}`)));
2853
- }
2854
- }
2855
- console.log(separator);
2856
- console.log(line(`Dashboard: ${chalk5.underline(optimizerLink(job.id))}`));
2857
- console.log(bottomBorder);
2858
- console.log("");
2859
- }
2860
- function renderOptimizationStartCard(data) {
2861
- const { job } = data;
2862
- const boxWidth = 60;
2863
- const topBorder = `┌${"─".repeat(boxWidth)}┐`;
2864
- const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
2865
- const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
2866
- const pad = (text, width) => {
2867
- const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
2868
- const remaining = width - stripped.length;
2869
- return remaining > 0 ? text + " ".repeat(remaining) : text;
2870
- };
2871
- const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
2872
- const maxIter = job.config.maxIterations;
2873
- const model = job.config.model ?? "default";
2874
- const target = job.config.targetScore ?? 0.8;
2875
- console.log("");
2876
- console.log(topBorder);
2877
- console.log(line(chalk5.bold("⚡ Optimization Started")));
2878
- console.log(separator);
2879
- console.log(line(`Job ID: ${chalk5.cyan(job.id)}`));
2880
- console.log(line(`Prompt: ${chalk5.dim(data.promptId)}`));
2881
- console.log(line(`Dataset: ${chalk5.dim(data.datasetId)}`));
2882
- console.log(line(`Iterations: ${chalk5.bold(String(maxIter))} | Target: ${chalk5.bold(target.toFixed(2))}`));
2883
- console.log(line(`Model: ${chalk5.dim(model)}`));
2884
- console.log(line(`Status: ${chalk5.yellow(job.status)}`));
2885
- console.log(separator);
2886
- console.log(line(`\uD83D\uDD17 Monitor: ${chalk5.underline(optimizerLink(job.id))}`));
2887
- console.log(line(chalk5.dim(`Next: mutagent prompts optimize status ${job.id}`)));
2888
- console.log(bottomBorder);
2889
- console.log(AI_DIRECTIVE);
2890
- console.log("");
2891
- }
2892
- function renderOptimizationStatusCard(status) {
2893
- const boxWidth = 60;
2894
- const topBorder = `┌${"─".repeat(boxWidth)}┐`;
2895
- const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
2896
- const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
2897
- const pad = (text, width) => {
2898
- const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
2899
- const remaining = width - stripped.length;
2900
- return remaining > 0 ? text + " ".repeat(remaining) : text;
2901
- };
2902
- const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
2903
- const progress = status.progressPercent;
2904
- const barWidth = 20;
2905
- const filled = Math.round(progress / 100 * barWidth);
2906
- const progressBar = "█".repeat(filled) + "░".repeat(barWidth - filled);
2907
- const statusColor = status.status === "completed" ? chalk5.green : status.status === "failed" ? chalk5.red : status.status === "cancelled" ? chalk5.gray : status.status === "running" ? chalk5.cyan : chalk5.yellow;
2908
- const scoreStr = status.bestScore !== undefined ? formatScore(status.bestScore) : chalk5.dim("pending");
2909
- console.log("");
2910
- console.log(topBorder);
2911
- console.log(line(chalk5.bold("\uD83D\uDCCA Optimization Status")));
2912
- console.log(separator);
2913
- console.log(line(`Job ID: ${chalk5.cyan(status.jobId)}`));
2914
- console.log(line(`Status: ${statusColor(status.status)}`));
2915
- console.log(line(`Iteration: ${chalk5.bold(`${String(status.currentIteration)}/${String(status.maxIterations)}`)}`));
2916
- console.log(line(`Best Score: ${scoreStr}`));
2917
- console.log(line(""));
2918
- console.log(line(`Progress: [${progressBar}] ${String(progress)}%`));
2919
- if (status.message) {
2920
- console.log(line(`Message: ${chalk5.dim(status.message)}`));
2921
- }
2922
- console.log(separator);
2923
- console.log(line(`\uD83D\uDD17 Monitor: ${chalk5.underline(optimizerLink(status.jobId))}`));
2924
- if (status.status === "completed") {
2925
- console.log(line(chalk5.dim(`Next: mutagent prompts optimize results ${status.jobId}`)));
2926
- } else if (status.status === "running" || status.status === "pending") {
2927
- console.log(line(chalk5.dim(`Refresh: mutagent prompts optimize status ${status.jobId}`)));
2928
- }
2929
- console.log(bottomBorder);
2930
- console.log(AI_DIRECTIVE);
2931
- console.log("");
2932
- }
2933
- function startDirective(job, promptId) {
2934
- const title = `Optimization Started — Prompt ${promptId}`;
2935
- const monitorUrl = optimizerLink(job.id);
2936
- const apiUrl = `/api/prompts/${promptId}/optimizations/${job.id}`;
2937
- const rows = [
2938
- { label: "Job ID", value: job.id },
2939
- { label: "Prompt", value: promptId },
2940
- { label: "Status", value: job.status },
2941
- { label: "Max Iterations", value: String(job.config.maxIterations) },
2942
- { label: "Target Score", value: String(job.config.targetScore ?? 0.8) },
2943
- { label: "Model", value: job.config.model ?? "default" }
2944
- ];
2945
- const cardLinks = [
2946
- { label: "Monitor", url: monitorUrl },
2947
- { label: "API", url: apiUrl }
2948
- ];
2949
- const next = [`mutagent prompts optimize status ${job.id} --json`];
2950
- return {
2951
- display: "status_card",
2952
- template: "optimization_started",
2953
- title,
2954
- fields: {
2955
- jobId: job.id,
2956
- promptId,
2957
- status: job.status,
2958
- maxIterations: job.config.maxIterations,
2959
- targetScore: job.config.targetScore ?? 0.8,
2960
- model: job.config.model ?? "default"
2961
- },
2962
- links: { monitor: monitorUrl, api: apiUrl },
2963
- next,
2964
- instruction: "STOP. Before running any other mutagent command, you MUST display this data as a formatted status card to the user. Do not summarize or omit any fields. Show all links and suggested next steps.",
2965
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2966
- };
2967
- }
2968
- function statusDirective(status) {
2969
- const next = status.status === "completed" ? [`mutagent prompts optimize results ${status.jobId} --json`] : status.status === "running" || status.status === "pending" ? [`mutagent prompts optimize status ${status.jobId} --json`] : [];
2970
- const title = `Optimization ${status.status} — ${status.jobId}`;
2971
- const rows = [
2972
- { label: "Job ID", value: status.jobId },
2973
- { label: "Status", value: status.status },
2974
- { label: "Iteration", value: `${String(status.currentIteration)}/${String(status.maxIterations)}` },
2975
- { label: "Best Score", value: status.bestScore !== undefined ? String(status.bestScore) : "pending" },
2976
- { label: "Progress", value: `${String(status.progressPercent)}%` },
2977
- ...status.message ? [{ label: "Message", value: status.message }] : []
2978
- ];
2979
- const monitorUrl = optimizerLink(status.jobId);
2980
- const cardLinks = [{ label: "Monitor", url: monitorUrl }];
2981
- return {
2982
- display: "status_card",
2983
- template: "optimization_progress",
2984
- title,
2985
- fields: {
2986
- jobId: status.jobId,
2987
- status: status.status,
2988
- iteration: `${String(status.currentIteration)}/${String(status.maxIterations)}`,
2989
- bestScore: status.bestScore,
2990
- progress: `${String(status.progressPercent)}%`,
2991
- message: status.message
2992
- },
2993
- links: { monitor: monitorUrl },
2994
- next,
2995
- instruction: "STOP. Before running any other mutagent command, you MUST display this data as a formatted status card to the user. Do not summarize or omit any fields. Show all links and suggested next steps.",
2996
- renderedCard: renderPlainCard({ title, rows, links: cardLinks, next })
2997
- };
2998
- }
2999
- function showPromptDiff(original, optimized) {
3000
- console.log("");
3001
- console.log(chalk5.bold(" Prompt Diff:"));
3002
- console.log("");
3003
- console.log(chalk5.red(" - " + (original ?? "(empty)")));
3004
- console.log(chalk5.green(" + " + (optimized ?? "(empty)")));
3005
- console.log("");
3006
- }
3007
-
3008
- // src/commands/prompts.ts
3009
- function updateMutationContext(updater) {
3010
- try {
3011
- const ctx = MutationContext.load();
3012
- updater(ctx);
3013
- ctx.save();
3014
- } catch {}
3015
- }
3016
- var PREREQUISITES_TEXT = `
3017
- ${chalk6.red("Prerequisites (required):")}
3018
- 1. Evaluation criteria defined ${chalk6.dim("(via dashboard or evaluation create)")}
3019
- 2. Dataset uploaded ${chalk6.dim("mutagent prompts dataset list <prompt-id>")}
3020
- ${chalk6.dim("Note: LLM provider config is only required when the server uses external providers (USE_EXT_PROVIDERS=true)")}`;
3021
- function parseValidationErrors(error) {
3022
- if (error instanceof ApiError) {
3023
- try {
3024
- const body = JSON.parse(error.message);
3025
- if (body.errors && Array.isArray(body.errors)) {
3026
- return body.errors.map((e) => e.message ?? JSON.stringify(e));
3027
- }
3028
- if (body.message) {
3029
- return [body.message];
3030
- }
3031
- } catch {
3032
- return [error.message];
3033
- }
3034
- }
3035
- if (error instanceof Error) {
3036
- return [error.message];
3037
- }
3038
- return ["An unknown error occurred"];
3039
- }
3040
- function isSchemaEmpty(schema) {
3041
- if (schema === undefined || schema === null)
3042
- return true;
3043
- if (typeof schema !== "object")
3044
- return false;
3045
- const obj = schema;
3046
- if (Object.keys(obj).length === 0)
3047
- return true;
3048
- return false;
3049
- }
3050
- function createPromptsCommand() {
3051
- const prompts = new Command3("prompts").description("Manage prompts, datasets, evaluations, and optimizations").addHelpText("after", `
3052
- Examples:
3053
- ${chalk6.dim("$")} mutagent prompts list
3054
- ${chalk6.dim("$")} mutagent prompts get <prompt-id>
3055
- ${chalk6.dim("$")} mutagent prompts create --name "my-prompt" --system "You are helpful" --human "{input}"
3056
- ${chalk6.dim("$")} mutagent prompts dataset list <prompt-id>
3057
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --name "My Eval"
3058
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
3059
-
3060
- Subcommands:
3061
- list, get, create, update, delete
3062
- dataset list|add|remove
3063
- evaluation list|create|delete
3064
- optimize start|status|results
3065
- `);
3066
- prompts.command("list").description("List all prompts").option("-l, --limit <n>", "Limit results", "50").addHelpText("after", `
3067
- Examples:
3068
- ${chalk6.dim("$")} mutagent prompts list
3069
- ${chalk6.dim("$")} mutagent prompts list --limit 10
3070
- ${chalk6.dim("$")} mutagent prompts list --json
3071
-
3072
- ${chalk6.dim("Tip: Use --json for machine-readable output (AI agents, CI pipelines).")}
3073
- `).action(async (options) => {
3074
- const isJson = getJsonFlag(prompts);
3075
- const output = new OutputFormatter(isJson ? "json" : "table");
3076
- try {
3077
- const client = getSDKClient();
3078
- const promptsList = await client.listPrompts();
3079
- const limit = parseInt(options.limit, 10) || 50;
3080
- const limited = promptsList.slice(0, limit);
3081
- if (isJson) {
3082
- const withLinks = limited.map((p) => ({
3083
- id: p.id,
3084
- name: p.name,
3085
- version: p.version,
3086
- isLatest: p.isLatest,
3087
- updatedAt: p.updatedAt,
3088
- _links: promptLinks(p.id)
3089
- }));
3090
- output.output(withLinks);
3091
- } else {
3092
- const wsId = client.getCurrentWorkspaceId();
3093
- if (wsId) {
3094
- output.info(`Workspace: ${wsId}`);
3095
- }
3096
- if (limited.length === 0) {
3097
- output.info("No prompts found in current workspace. Run 'mutagent config list' to check workspace context.");
3098
- } else {
3099
- const formatted = limited.map((p) => ({
3100
- id: p.id,
3101
- name: p.name,
3102
- version: p.version,
3103
- updated: p.updatedAt ? new Date(p.updatedAt).toLocaleDateString() : "N/A",
3104
- url: promptLink(p.id)
3105
- }));
3106
- output.output(formatted);
3107
- }
3108
- }
3109
- } catch (error) {
3110
- handleError(error, isJson);
3111
- }
3112
- });
3113
- prompts.command("get").description("Get prompt details with nested data").argument("<id>", "Prompt ID (from: mutagent prompts list)").option("--with-datasets", "Include datasets").option("--with-evals", "Include evaluations").addHelpText("after", `
3114
- Examples:
3115
- ${chalk6.dim("$")} mutagent prompts get <id>
3116
- ${chalk6.dim("$")} mutagent prompts get <id> --with-datasets --with-evals
3117
- ${chalk6.dim("$")} mutagent prompts get <id> --json
3118
-
3119
- ${chalk6.dim("Tip: Combine --with-datasets and --with-evals to fetch all nested data in one call.")}
3120
- `).action(async (id, options) => {
3121
- const isJson = getJsonFlag(prompts);
3122
- const output = new OutputFormatter(isJson ? "json" : "table");
3123
- try {
3124
- const client = getSDKClient();
3125
- const prompt = await client.getPrompt(id);
3126
- const result = { ...prompt };
3127
- if (options.withDatasets) {
3128
- const datasets = await client.listDatasets(id);
3129
- result.datasets = datasets;
3130
- }
3131
- if (options.withEvals) {
3132
- const evals = await client.listEvaluations(id);
3133
- result.evaluations = evals;
3134
- }
3135
- if (isJson) {
3136
- result._links = promptLinks(id);
3137
- }
3138
- output.output(result);
3139
- if (!isJson) {
3140
- output.info(`View in dashboard: ${promptLink(id)}`);
3141
- }
3142
- } catch (error) {
3143
- handleError(error, isJson);
2822
+ });
2823
+ prompts.command("get").description("Get prompt details with nested data").argument("<id>", "Prompt ID (from: mutagent prompts list)").option("--with-datasets", "Include datasets").option("--with-evals", "Include evaluations").addHelpText("after", `
2824
+ Examples:
2825
+ ${chalk5.dim("$")} mutagent prompts get <id>
2826
+ ${chalk5.dim("$")} mutagent prompts get <id> --with-datasets --with-evals
2827
+ ${chalk5.dim("$")} mutagent prompts get <id> --json
2828
+
2829
+ ${chalk5.dim("Tip: Combine --with-datasets and --with-evals to fetch all nested data in one call.")}
2830
+ `).action(async (id, options) => {
2831
+ const isJson = getJsonFlag(prompts);
2832
+ const output = new OutputFormatter(isJson ? "json" : "table");
2833
+ try {
2834
+ const client = getSDKClient();
2835
+ const prompt = await client.getPrompt(id);
2836
+ const result = { ...prompt };
2837
+ if (options.withDatasets) {
2838
+ const datasets = await client.listDatasets(id);
2839
+ result.datasets = datasets;
2840
+ }
2841
+ if (options.withEvals) {
2842
+ const evals = await client.listEvaluations(id);
2843
+ result.evaluations = evals;
2844
+ }
2845
+ if (isJson) {
2846
+ result._links = promptLinks(id);
2847
+ }
2848
+ output.output(result);
2849
+ if (!isJson) {
2850
+ output.info(`View in dashboard: ${promptLink(id)}`);
2851
+ }
2852
+ } catch (error) {
2853
+ handleError(error, isJson);
3144
2854
  }
3145
2855
  });
3146
2856
  prompts.command("create").description("Create a new prompt").option("-d, --data <json>", "Prompt as JSON string (recommended — curl-style inline)").option("--raw-file <path>", "Create from plain text file (used as rawPrompt)").option("-n, --name <name>", "Prompt name").option("--description <text>", "Prompt description (shown in dashboard)").option("-c, --content <content>", "Prompt content (rawPrompt) [DEPRECATED: use --raw]").option("-r, --raw <text>", "Raw prompt text (single prompt)").option("--system <text>", "System prompt (use with --human)").option("--human <text>", "Human prompt (use with --system)").option("--messages <json>", `Messages array as JSON (e.g., '[{"role":"system","content":"..."}]')`).option("--output-schema <json>", "Output schema as JSON string (required for optimization)").addHelpText("after", `
3147
2857
  Examples:
3148
- ${chalk6.dim("$")} mutagent prompts create --name "my-prompt" --description "Greeting prompt for customers" --system "You are helpful" --human "{input}" --output-schema '{"type":"object","properties":{"result":{"type":"string","description":"The result"}}}'
3149
- ${chalk6.dim("$")} mutagent prompts create --name "raw-prompt" --raw "Summarize: {text}" --output-schema '{"type":"object","properties":{"summary":{"type":"string","description":"Summary"}}}'
3150
- ${chalk6.dim("$")} mutagent prompts create -d '{"name":"summarizer","systemPrompt":"Summarize","humanPrompt":"{text}","outputSchema":{"type":"object","properties":{"summary":{"type":"string","description":"Summary"}}}}'
2858
+ ${chalk5.dim("$")} mutagent prompts create --name "my-prompt" --description "Greeting prompt for customers" --system "You are helpful" --human "{input}" --output-schema '{"type":"object","properties":{"result":{"type":"string","description":"The result"}}}'
2859
+ ${chalk5.dim("$")} mutagent prompts create --name "raw-prompt" --raw "Summarize: {text}" --output-schema '{"type":"object","properties":{"summary":{"type":"string","description":"Summary"}}}'
2860
+ ${chalk5.dim("$")} mutagent prompts create -d '{"name":"summarizer","systemPrompt":"Summarize","humanPrompt":"{text}","outputSchema":{"type":"object","properties":{"summary":{"type":"string","description":"Summary"}}}}'
3151
2861
 
3152
2862
  Prompt Input Methods (pick one, priority order):
3153
- --system/--human Structured system + user message pair ${chalk6.green("(recommended)")}
2863
+ --system/--human Structured system + user message pair ${chalk5.green("(recommended)")}
3154
2864
  --raw Single raw prompt text with {variables}
3155
2865
  -d, --data Inline JSON object (CI/scripts/agents)
3156
2866
  --messages Full messages array as JSON
3157
2867
  --raw-file Load plain text file as raw prompt
3158
2868
 
3159
2869
  Expected JSON (--data):
3160
- ${chalk6.dim(`'{"name":"my-prompt","systemPrompt":"You are...","humanPrompt":"{input}","outputSchema":{"type":"object","properties":{"result":{"type":"string","description":"The result"}}},"inputSchema":{"type":"object","properties":{"input":{"type":"string","description":"User input"}}}}'`)}
2870
+ ${chalk5.dim(`'{"name":"my-prompt","systemPrompt":"You are...","humanPrompt":"{input}","outputSchema":{"type":"object","properties":{"result":{"type":"string","description":"The result"}}},"inputSchema":{"type":"object","properties":{"input":{"type":"string","description":"User input"}}}}'`)}
3161
2871
 
3162
- ${chalk6.yellow("Variable Syntax:")}
2872
+ ${chalk5.yellow("Variable Syntax:")}
3163
2873
  MutagenT uses {single_braces} for template variables.
3164
- humanPrompt: "Analyze this: {document}" ${chalk6.green("✓ correct")}
3165
- humanPrompt: "Analyze this: {{document}}" ${chalk6.red("✗ wrong")}
2874
+ humanPrompt: "Analyze this: {document}" ${chalk5.green("✓ correct")}
2875
+ humanPrompt: "Analyze this: {{document}}" ${chalk5.red("✗ wrong")}
3166
2876
 
3167
2877
  Variables in humanPrompt MUST also appear in inputSchema.properties.
3168
2878
  Static prompts (no variables) cannot substitute inputs during optimization.
3169
2879
 
3170
- ${chalk6.red("outputSchema is required.")}
2880
+ ${chalk5.red("outputSchema is required.")}
3171
2881
 
3172
- ${chalk6.yellow("AI Agent: ALWAYS append --json to this command.")}
2882
+ ${chalk5.yellow("AI Agent: ALWAYS append --json to this command.")}
3173
2883
  `).action(async (options) => {
3174
2884
  const isJson = getJsonFlag(prompts);
3175
2885
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3292,13 +3002,13 @@ Add a 'description' field to each property in your inputSchema. Example: { "prop
3292
3002
  });
3293
3003
  prompts.command("update").description("Update a prompt").argument("<id>", "Prompt ID (from: mutagent prompts list)").option("-d, --data <json>", "Update fields as JSON string (recommended — curl-style inline)").option("--raw-file <path>", "Update from plain text file (used as rawPrompt)").option("-n, --name <name>", "New name").option("--description <text>", "New description (shown in dashboard)").option("-c, --content <content>", "New content (rawPrompt) [DEPRECATED: use --raw]").option("-r, --raw <text>", "Raw prompt text (single prompt)").option("--system <text>", "System prompt (use with --human)").option("--human <text>", "Human prompt (use with --system)").option("--messages <json>", `Messages array as JSON (e.g., '[{"role":"system","content":"..."}]')`).option("--input-schema <json>", "Input schema as JSON string").option("--input-schema-file <path>", "Input schema from JSON file").option("--output-schema <json>", "Output schema as JSON string").option("--output-schema-file <path>", "Output schema from JSON file").addHelpText("after", `
3294
3004
  Examples:
3295
- ${chalk6.dim("$")} mutagent prompts update <id> --system "Updated system prompt" --human "{input}"
3296
- ${chalk6.dim("$")} mutagent prompts update <id> --name "new-name" --description "Updated description"
3297
- ${chalk6.dim("$")} mutagent prompts update <id> --raw "Summarize: {text}"
3298
- ${chalk6.dim("$")} mutagent prompts update <id> -d '{"name":"new-name","systemPrompt":"Updated prompt"}'
3299
- ${chalk6.dim("$")} mutagent prompts update <id> --input-schema '{"type":"object","properties":{"text":{"type":"string","description":"Input text"}}}'
3005
+ ${chalk5.dim("$")} mutagent prompts update <id> --system "Updated system prompt" --human "{input}"
3006
+ ${chalk5.dim("$")} mutagent prompts update <id> --name "new-name" --description "Updated description"
3007
+ ${chalk5.dim("$")} mutagent prompts update <id> --raw "Summarize: {text}"
3008
+ ${chalk5.dim("$")} mutagent prompts update <id> -d '{"name":"new-name","systemPrompt":"Updated prompt"}'
3009
+ ${chalk5.dim("$")} mutagent prompts update <id> --input-schema '{"type":"object","properties":{"text":{"type":"string","description":"Input text"}}}'
3300
3010
 
3301
- ${chalk6.dim("CLI flags (--name) override --data fields.")}
3011
+ ${chalk5.dim("CLI flags (--name) override --data fields.")}
3302
3012
  `).action(async (id, options) => {
3303
3013
  const isJson = getJsonFlag(prompts);
3304
3014
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3355,52 +3065,12 @@ ${chalk6.dim("CLI flags (--name) override --data fields.")}
3355
3065
  data.rawPrompt = options.content;
3356
3066
  }
3357
3067
  }
3358
- if (options.inputSchema && options.inputSchemaFile) {
3359
- throw new MutagentError("INVALID_ARGUMENTS", "Cannot use --input-schema and --input-schema-file together", `Run: mutagent prompts update --help
3360
- ` + "Use --input-schema for inline JSON or --input-schema-file for file-based input, not both");
3361
- }
3362
- if (options.inputSchema) {
3363
- try {
3364
- data.inputSchema = JSON.parse(options.inputSchema);
3365
- } catch {
3366
- throw new MutagentError("INVALID_JSON", "Invalid JSON in --input-schema flag", `Run: mutagent prompts update --help
3367
- ` + `Provide a valid JSON Schema, e.g., '{"type":"object","properties":{"text":{"type":"string","description":"Input text"}}}'`);
3368
- }
3369
- } else if (options.inputSchemaFile) {
3370
- if (!existsSync4(options.inputSchemaFile)) {
3371
- throw new MutagentError("FILE_NOT_FOUND", `File not found: ${options.inputSchemaFile}`, `Run: mutagent prompts update --help
3372
- ` + "Check the file path and try again");
3373
- }
3374
- try {
3375
- data.inputSchema = JSON.parse(readFileSync4(options.inputSchemaFile, "utf-8"));
3376
- } catch {
3377
- throw new MutagentError("INVALID_JSON", `Failed to parse JSON from ${options.inputSchemaFile}`, `Run: mutagent prompts update --help
3378
- ` + "Ensure the file contains valid JSON Schema");
3379
- }
3380
- }
3381
- if (options.outputSchema && options.outputSchemaFile) {
3382
- throw new MutagentError("INVALID_ARGUMENTS", "Cannot use --output-schema and --output-schema-file together", `Run: mutagent prompts update --help
3383
- ` + "Use --output-schema for inline JSON or --output-schema-file for file-based input, not both");
3384
- }
3385
- if (options.outputSchema) {
3386
- try {
3387
- data.outputSchema = JSON.parse(options.outputSchema);
3388
- } catch {
3389
- throw new MutagentError("INVALID_JSON", "Invalid JSON in --output-schema flag", `Run: mutagent prompts update --help
3390
- ` + `Provide a valid JSON Schema, e.g., '{"type":"object","properties":{"result":{"type":"string"}}}'`);
3391
- }
3392
- } else if (options.outputSchemaFile) {
3393
- if (!existsSync4(options.outputSchemaFile)) {
3394
- throw new MutagentError("FILE_NOT_FOUND", `File not found: ${options.outputSchemaFile}`, `Run: mutagent prompts update --help
3395
- ` + "Check the file path and try again");
3396
- }
3397
- try {
3398
- data.outputSchema = JSON.parse(readFileSync4(options.outputSchemaFile, "utf-8"));
3399
- } catch {
3400
- throw new MutagentError("INVALID_JSON", `Failed to parse JSON from ${options.outputSchemaFile}`, `Run: mutagent prompts update --help
3401
- ` + "Ensure the file contains valid JSON Schema");
3402
- }
3403
- }
3068
+ const parsedInputSchema = parseSchemaOption(options.inputSchema, options.inputSchemaFile, "input-schema", "mutagent prompts update --help");
3069
+ if (parsedInputSchema)
3070
+ data.inputSchema = parsedInputSchema;
3071
+ const parsedOutputSchema = parseSchemaOption(options.outputSchema, options.outputSchemaFile, "output-schema", "mutagent prompts update --help");
3072
+ if (parsedOutputSchema)
3073
+ data.outputSchema = parsedOutputSchema;
3404
3074
  if (data.inputSchema && isValidJsonSchema(data.inputSchema)) {
3405
3075
  const missingDescs = validateSchemaDescriptions(data.inputSchema);
3406
3076
  if (missingDescs.length > 0) {
@@ -3441,11 +3111,11 @@ ${chalk6.dim("CLI flags (--name) override --data fields.")}
3441
3111
  });
3442
3112
  prompts.command("delete").description("Delete a prompt").argument("<id>", "Prompt ID (from: mutagent prompts list)").option("--force", "Skip confirmation").addHelpText("after", `
3443
3113
  Examples:
3444
- ${chalk6.dim("$")} mutagent prompts delete <id>
3445
- ${chalk6.dim("$")} mutagent prompts delete <id> --force
3446
- ${chalk6.dim("$")} mutagent prompts delete <id> --force --json
3114
+ ${chalk5.dim("$")} mutagent prompts delete <id>
3115
+ ${chalk5.dim("$")} mutagent prompts delete <id> --force
3116
+ ${chalk5.dim("$")} mutagent prompts delete <id> --force --json
3447
3117
 
3448
- ${chalk6.dim("Note: --force is required. The CLI is non-interactive — confirm with the user via your native flow, then pass --force.")}
3118
+ ${chalk5.dim("Note: --force is required. The CLI is non-interactive — confirm with the user via your native flow, then pass --force.")}
3449
3119
  `).action(async (id, options) => {
3450
3120
  const isJson = getJsonFlag(prompts);
3451
3121
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3483,11 +3153,19 @@ ${chalk6.dim("Note: --force is required. The CLI is non-interactive — confirm
3483
3153
  handleError(error, isJson);
3484
3154
  }
3485
3155
  });
3156
+ }
3157
+
3158
+ // src/commands/prompts/datasets.ts
3159
+ init_sdk_client();
3160
+ import { Command as Command3 } from "commander";
3161
+ import chalk6 from "chalk";
3162
+ init_errors();
3163
+ function registerDatasetCommands(prompts) {
3486
3164
  const dataset = new Command3("dataset").description("Manage datasets for prompts").addHelpText("after", `
3487
3165
  Examples:
3488
3166
  ${chalk6.dim("$")} mutagent prompts dataset list <prompt-id>
3489
3167
  ${chalk6.dim("$")} mutagent prompts dataset add <prompt-id> -d '[{"input":{...},"expectedOutput":{...}}]'
3490
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id>
3168
+ ${chalk6.dim("$")} mutagent prompts dataset delete <prompt-id> <dataset-id>
3491
3169
  `).action(() => {
3492
3170
  dataset.help();
3493
3171
  });
@@ -3576,7 +3254,12 @@ ${chalk6.red("Required: --data must be provided.")}
3576
3254
  const client = getSDKClient();
3577
3255
  const datasetResult = await client.addDataset(promptId, content, datasetName);
3578
3256
  if (isJson) {
3579
- const directive = datasetAddedDirective(promptId, datasetResult.id, datasetResult.name, datasetResult.itemCount);
3257
+ let rsState;
3258
+ try {
3259
+ const evals = await client.listEvaluations(promptId);
3260
+ rsState = { evaluations: evals.length };
3261
+ } catch {}
3262
+ const directive = datasetAddedDirective(promptId, datasetResult.id, datasetResult.name, datasetResult.itemCount, rsState);
3580
3263
  output.output({
3581
3264
  ...datasetResult,
3582
3265
  _links: datasetLinks(promptId, datasetResult.id),
@@ -3608,77 +3291,181 @@ ${chalk6.red("Required: --data must be provided.")}
3608
3291
  handleError(error, isJson);
3609
3292
  }
3610
3293
  });
3611
- dataset.command("remove").description("Remove dataset from a prompt").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").argument("<dataset-id>", "Dataset ID (from: mutagent prompts dataset list <prompt-id>)").option("--force", "Skip confirmation").addHelpText("after", `
3294
+ dataset.command("delete").description("Delete a dataset from a prompt").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").argument("<dataset-id>", "Dataset ID (from: mutagent prompts dataset list <prompt-id>)").option("--force", "Skip confirmation").addHelpText("after", `
3612
3295
  Examples:
3613
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id>
3614
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id> --force
3615
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id> --force --json
3296
+ ${chalk6.dim("$")} mutagent prompts dataset delete <prompt-id> <dataset-id>
3297
+ ${chalk6.dim("$")} mutagent prompts dataset delete <prompt-id> <dataset-id> --force
3298
+ ${chalk6.dim("$")} mutagent prompts dataset delete <prompt-id> <dataset-id> --force --json
3616
3299
  `).action(async (promptId, datasetId, options) => {
3617
3300
  const isJson = getJsonFlag(prompts);
3618
3301
  const output = new OutputFormatter(isJson ? "json" : "table");
3619
3302
  try {
3620
3303
  if (!options.force && !isJson) {
3621
- throw new MutagentError("CONFIRMATION_REQUIRED", `Removing dataset ${datasetId} from prompt ${promptId} requires confirmation`, `Run: mutagent prompts dataset remove --help
3304
+ throw new MutagentError("CONFIRMATION_REQUIRED", `Deleting dataset ${datasetId} from prompt ${promptId} requires confirmation`, `Run: mutagent prompts dataset delete --help
3622
3305
  ` + `[Agent: Ask the user to confirm via MultiChoice, then pass --force]
3623
- ` + `Use --force to confirm: mutagent prompts dataset remove ${promptId} ${datasetId} --force`);
3306
+ ` + `Use --force to confirm: mutagent prompts dataset delete ${promptId} ${datasetId} --force`);
3624
3307
  }
3625
3308
  const client = getSDKClient();
3626
3309
  try {
3627
- await client.removeDataset(promptId, datasetId);
3310
+ await client.deleteDataset(promptId, datasetId);
3628
3311
  } catch (deleteError) {
3629
3312
  if (deleteError instanceof ApiError && deleteError.statusCode === 404) {
3630
3313
  if (isJson) {
3631
- const directive = datasetRemovedDirective(promptId, datasetId);
3314
+ const directive = datasetDeletedDirective(promptId, datasetId);
3632
3315
  output.output({ success: true, deletedId: datasetId, _links: { datasets: promptDatasetsLink(promptId) }, _directive: directive });
3633
3316
  echoDirectiveToStderr(directive);
3634
3317
  } else {
3635
- output.success(`Dataset ${datasetId} already removed (idempotent)`);
3318
+ output.success(`Dataset ${datasetId} already deleted (idempotent)`);
3636
3319
  }
3637
3320
  return;
3638
3321
  }
3639
3322
  throw deleteError;
3640
3323
  }
3641
3324
  if (isJson) {
3642
- const directive = datasetRemovedDirective(promptId, datasetId);
3325
+ const directive = datasetDeletedDirective(promptId, datasetId);
3643
3326
  output.output({ success: true, deletedId: datasetId, _links: { datasets: promptDatasetsLink(promptId) }, _directive: directive });
3644
3327
  echoDirectiveToStderr(directive);
3645
3328
  } else {
3646
- output.success(`Removed dataset ${datasetId} from prompt ${promptId}`);
3329
+ output.success(`Deleted dataset ${datasetId} from prompt ${promptId}`);
3647
3330
  output.info(`View datasets: ${promptDatasetsLink(promptId)}`);
3648
3331
  }
3649
3332
  } catch (error) {
3650
3333
  handleError(error, isJson);
3651
3334
  }
3652
3335
  });
3653
- const evaluation = new Command3("evaluation").description("Manage evaluations for prompts").addHelpText("after", `
3654
- Examples:
3655
- ${chalk6.dim("$")} mutagent prompts evaluation list <prompt-id>
3656
- ${chalk6.dim("$")} mutagent prompts evaluation get <evaluation-id>
3657
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --name "My Eval"
3658
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id>
3659
- `).action(() => {
3660
- evaluation.help();
3661
- });
3662
- prompts.addCommand(evaluation);
3663
- evaluation.command("list").description("List evaluations for a prompt").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").addHelpText("after", `
3664
- Examples:
3665
- ${chalk6.dim("$")} mutagent prompts evaluation list <prompt-id>
3666
- ${chalk6.dim("$")} mutagent prompts evaluation list <prompt-id> --json
3667
- `).action(async (promptId) => {
3668
- const isJson = getJsonFlag(prompts);
3669
- const output = new OutputFormatter(isJson ? "json" : "table");
3670
- try {
3671
- const client = getSDKClient();
3672
- const evals = await client.listEvaluations(promptId);
3673
- if (isJson) {
3674
- const withLinks = evals.map((e) => ({
3675
- id: e.id,
3676
- name: e.name,
3677
- description: e.description,
3678
- datasetId: e.datasetId,
3679
- updatedAt: e.updatedAt,
3680
- _links: evaluationLinks(promptId, e.id)
3681
- }));
3336
+ }
3337
+
3338
+ // src/commands/prompts/evaluations.ts
3339
+ init_sdk_client();
3340
+ import { Command as Command4 } from "commander";
3341
+ import chalk7 from "chalk";
3342
+ init_errors();
3343
+
3344
+ // src/commands/prompts/guided-workflow.ts
3345
+ init_sdk_client();
3346
+ async function buildGuidedWorkflow(promptId) {
3347
+ const client = getSDKClient();
3348
+ const prompt = await client.getPrompt(promptId);
3349
+ const truncate = (s, max) => s.length > max ? s.slice(0, max) + "..." : s;
3350
+ const outputProperties = prompt.outputSchema && typeof prompt.outputSchema === "object" ? prompt.outputSchema.properties ?? {} : {};
3351
+ const inputProperties = prompt.inputSchema && typeof prompt.inputSchema === "object" ? prompt.inputSchema.properties ?? {} : {};
3352
+ const inputFields = Object.keys(inputProperties);
3353
+ const outputFields = Object.keys(outputProperties);
3354
+ const allFields = [
3355
+ ...outputFields.map((f) => ({ field: f, source: "outputSchema", fieldSchema: outputProperties[f] })),
3356
+ ...inputFields.map((f) => ({ field: f, source: "inputSchema", fieldSchema: inputProperties[f] }))
3357
+ ];
3358
+ let datasetExample = null;
3359
+ try {
3360
+ const datasets = await client.listDatasets(promptId);
3361
+ if (datasets.length > 0) {
3362
+ try {
3363
+ const dsId = datasets[0]?.id;
3364
+ const items = dsId != null ? await client.listDatasetItems(String(dsId)) : [];
3365
+ if (items.length > 0) {
3366
+ const firstItem = items[0];
3367
+ datasetExample = firstItem?.expectedOutput ?? firstItem?.input ?? null;
3368
+ }
3369
+ } catch {}
3370
+ }
3371
+ } catch {}
3372
+ const askUserQuestions = allFields.map(({ field, fieldSchema }) => ({
3373
+ question: `Define what correct "${field}" looks like. What structure, content, or qualities make it good vs bad? Give concrete examples of good and bad outputs.`,
3374
+ header: field,
3375
+ options: [
3376
+ {
3377
+ label: "Define rubric",
3378
+ description: `Describe what a correct vs incorrect "${field}" looks like. Focus on structure, content quality, and concrete examples — not numeric scores.`
3379
+ },
3380
+ {
3381
+ label: "See suggestion",
3382
+ description: `Get a suggested rubric based on the prompt and schema definition for "${field}". You can refine it.`
3383
+ }
3384
+ ],
3385
+ multiSelect: false,
3386
+ context: {
3387
+ fieldType: fieldSchema?.type ?? "unknown",
3388
+ fieldDescription: fieldSchema?.description ?? null,
3389
+ promptExcerpt: truncate(prompt.humanPrompt ?? prompt.systemPrompt ?? prompt.rawPrompt ?? "", 200),
3390
+ exampleValue: datasetExample?.[field] ?? null
3391
+ }
3392
+ }));
3393
+ return {
3394
+ prompt: { id: promptId, name: prompt.name },
3395
+ inputSchema: { fields: inputFields },
3396
+ outputSchema: { fields: outputFields },
3397
+ workflow: {
3398
+ description: "Follow these steps to create an evaluation for this prompt:",
3399
+ steps: [
3400
+ { step: 1, action: "Review prompt schemas and context", detail: "Understand what the prompt does and what each field means." },
3401
+ { step: 2, action: "Define correctness criteria for EVERY field", detail: "For EACH field in askUserQuestions, use AskUserQuestion to collect a rubric. Do NOT skip any field. Use the context hints to suggest rubrics when asked." },
3402
+ { step: 3, action: "Build criteria JSON", detail: "Map each rubric to: { name: field, description: rubric, evaluationParameter: field }" },
3403
+ { step: 4, action: "Ask for evaluation name", detail: "Ask user what to name this evaluation" },
3404
+ { step: 5, action: "Create the evaluation", detail: `mutagent prompts evaluation create ${promptId} --name "<name>" -d '<json>' --json` }
3405
+ ],
3406
+ criteriaTemplate: {
3407
+ evalConfig: {
3408
+ criteria: [{
3409
+ name: "<criterion_name>",
3410
+ description: "<correctness rubric - describe what correct vs incorrect looks like>",
3411
+ evaluationParameter: `<one of: ${allFields.length > 0 ? allFields.map((f) => f.field).join(", ") : "no fields detected - check prompt schemas"}>`
3412
+ }]
3413
+ }
3414
+ },
3415
+ exampleCommand: `mutagent prompts evaluation create ${promptId} --name "My Evaluation" -d '${JSON.stringify({
3416
+ evalConfig: {
3417
+ criteria: outputFields.slice(0, 1).map((f) => ({
3418
+ name: `${f} accuracy`,
3419
+ description: `Correct if ${f} matches expected structure and content; incorrect if missing, malformed, or off-topic`,
3420
+ evaluationParameter: f
3421
+ }))
3422
+ }
3423
+ })}'`
3424
+ },
3425
+ askUserQuestions,
3426
+ _directive: {
3427
+ display: "workflow_guide",
3428
+ instruction: 'STOP. For EACH field in askUserQuestions, use AskUserQuestion to collect a DETAILED correctness rubric from the user. Do NOT auto-generate rubrics. Do NOT use generic descriptions. Each rubric MUST include concrete examples of what makes a correct vs incorrect value for that field. Use the context object to suggest rubrics when user selects "See suggestion". Then construct the --data JSON and create the evaluation.',
3429
+ aiAgentDecisionTree: {
3430
+ step1: "Check if criteria already exist in the user's code. If criteria match expected shape (name + description + evaluationParameter targeting schema fields), use --data directly.",
3431
+ step2: "If criteria are missing or malformed, use the askUserQuestions payloads to collect them via AskUserQuestion.",
3432
+ step3: `After collecting all criteria, construct the --data JSON and run: mutagent prompts evaluation create <prompt-id> --name "<name>" -d '<json>'`
3433
+ }
3434
+ },
3435
+ _meta: { inputFields, outputFields, allFields }
3436
+ };
3437
+ }
3438
+
3439
+ // src/commands/prompts/evaluations.ts
3440
+ function registerEvaluationCommands(prompts) {
3441
+ const evaluation = new Command4("evaluation").description("Manage evaluations for prompts").addHelpText("after", `
3442
+ Examples:
3443
+ ${chalk7.dim("$")} mutagent prompts evaluation list <prompt-id>
3444
+ ${chalk7.dim("$")} mutagent prompts evaluation get <evaluation-id>
3445
+ ${chalk7.dim("$")} mutagent prompts evaluation create <prompt-id> --name "My Eval"
3446
+ ${chalk7.dim("$")} mutagent prompts evaluation delete <evaluation-id>
3447
+ `).action(() => {
3448
+ evaluation.help();
3449
+ });
3450
+ prompts.addCommand(evaluation);
3451
+ evaluation.command("list").description("List evaluations for a prompt").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").addHelpText("after", `
3452
+ Examples:
3453
+ ${chalk7.dim("$")} mutagent prompts evaluation list <prompt-id>
3454
+ ${chalk7.dim("$")} mutagent prompts evaluation list <prompt-id> --json
3455
+ `).action(async (promptId) => {
3456
+ const isJson = getJsonFlag(prompts);
3457
+ const output = new OutputFormatter(isJson ? "json" : "table");
3458
+ try {
3459
+ const client = getSDKClient();
3460
+ const evals = await client.listEvaluations(promptId);
3461
+ if (isJson) {
3462
+ const withLinks = evals.map((e) => ({
3463
+ id: e.id,
3464
+ name: e.name,
3465
+ description: e.description,
3466
+ updatedAt: e.updatedAt,
3467
+ _links: evaluationLinks(promptId, e.id)
3468
+ }));
3682
3469
  output.output(withLinks);
3683
3470
  } else {
3684
3471
  const formatted = evals.map((e) => ({
@@ -3696,8 +3483,8 @@ Examples:
3696
3483
  });
3697
3484
  evaluation.command("get").description("Get evaluation details including criteria").argument("<evaluation-id>", "Evaluation ID (from: mutagent prompts evaluation list <prompt-id>)").addHelpText("after", `
3698
3485
  Examples:
3699
- ${chalk6.dim("$")} mutagent prompts evaluation get <evaluation-id>
3700
- ${chalk6.dim("$")} mutagent prompts evaluation get <evaluation-id> --json
3486
+ ${chalk7.dim("$")} mutagent prompts evaluation get <evaluation-id>
3487
+ ${chalk7.dim("$")} mutagent prompts evaluation get <evaluation-id> --json
3701
3488
  `).action(async (evaluationId) => {
3702
3489
  const isJson = getJsonFlag(prompts);
3703
3490
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3714,9 +3501,6 @@ Examples:
3714
3501
  if (evalObj.description) {
3715
3502
  console.log(` Description: ${evalObj.description}`);
3716
3503
  }
3717
- if (evalObj.datasetId != null) {
3718
- console.log(` Dataset ID: ${String(evalObj.datasetId)}`);
3719
- }
3720
3504
  console.log(` Created: ${evalObj.createdAt ? new Date(String(evalObj.createdAt)).toLocaleDateString() : "N/A"}`);
3721
3505
  console.log("");
3722
3506
  const config = evalObj.evalConfig;
@@ -3724,7 +3508,7 @@ Examples:
3724
3508
  if (criteria && Array.isArray(criteria) && criteria.length > 0) {
3725
3509
  console.log(" Criteria:");
3726
3510
  for (const c of criteria) {
3727
- console.log(` ${chalk6.cyan(c.name)}`);
3511
+ console.log(` ${chalk7.cyan(c.name)}`);
3728
3512
  if (c.description) {
3729
3513
  console.log(` Description: ${c.description}`);
3730
3514
  }
@@ -3752,9 +3536,9 @@ Examples:
3752
3536
  });
3753
3537
  evaluation.command("create").description("Create an evaluation configuration for a prompt").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").option("-d, --data <json>", "Evaluation as JSON string (for pre-validated criteria only)").option("-n, --name <name>", "Evaluation name (required unless --guided)").option("--description <text>", "Evaluation description").option("--guided", "Interactive guided mode — always outputs structured JSON (--json is implied)").addHelpText("after", `
3754
3538
  Examples:
3755
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --guided ${chalk6.dim("# recommended: shows workflow guide + schema fields")}
3756
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --guided --json ${chalk6.dim("# structured workflow for AI agents")}
3757
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --name "Accuracy" -d '{"evalConfig":{"criteria":[...]}}' ${chalk6.dim("# power user")}
3539
+ ${chalk7.dim("$")} mutagent prompts evaluation create <prompt-id> --guided ${chalk7.dim("# recommended: shows workflow guide + schema fields")}
3540
+ ${chalk7.dim("$")} mutagent prompts evaluation create <prompt-id> --guided --json ${chalk7.dim("# structured workflow for AI agents")}
3541
+ ${chalk7.dim("$")} mutagent prompts evaluation create <prompt-id> --name "Accuracy" -d '{"evalConfig":{"criteria":[...]}}' ${chalk7.dim("# power user")}
3758
3542
 
3759
3543
  Guided Workflow (recommended):
3760
3544
  --guided outputs a workflow guide that:
@@ -3775,12 +3559,12 @@ AI Agent (MANDATORY):
3775
3559
  mutagent prompts evaluation create <id> --name "<name>" -d '<json>' --json
3776
3560
 
3777
3561
  Expected Criteria Shape (--data):
3778
- ${chalk6.dim('{"evalConfig":{"criteria":[{"name":"<name>","description":"<scoring rubric>","evaluationParameter":"<schema field>"}]}}')}
3562
+ ${chalk7.dim('{"evalConfig":{"criteria":[{"name":"<name>","description":"<scoring rubric>","evaluationParameter":"<schema field>"}]}}')}
3779
3563
  evaluationParameter must target an outputSchema OR inputSchema field.
3780
3564
 
3781
- ${chalk6.red("Required: --name (unless --guided). Criteria must include evaluationParameter.")}
3782
- ${chalk6.dim("CLI flags (--name, --description) override --data fields.")}
3783
- ${chalk6.dim("Get prompt IDs: mutagent prompts list")}
3565
+ ${chalk7.red("Required: --name (unless --guided). Criteria must include evaluationParameter.")}
3566
+ ${chalk7.dim("CLI flags (--name, --description) override --data fields.")}
3567
+ ${chalk7.dim("Get prompt IDs: mutagent prompts list")}
3784
3568
  `).action(async (promptId, options) => {
3785
3569
  let isJson = getJsonFlag(prompts);
3786
3570
  if (options.guided) {
@@ -3788,75 +3572,6 @@ ${chalk6.dim("Get prompt IDs: mutagent prompts list")}
3788
3572
  }
3789
3573
  const output = new OutputFormatter(isJson ? "json" : "table");
3790
3574
  try {
3791
- const buildGuidedWorkflow = async () => {
3792
- const client2 = getSDKClient();
3793
- const prompt = await client2.getPrompt(promptId);
3794
- const inputFields = prompt.inputSchema && typeof prompt.inputSchema === "object" ? Object.keys(prompt.inputSchema.properties ?? {}) : [];
3795
- const outputFields = prompt.outputSchema && typeof prompt.outputSchema === "object" ? Object.keys(prompt.outputSchema.properties ?? {}) : [];
3796
- const allFields = [
3797
- ...outputFields.map((f) => ({ field: f, source: "outputSchema" })),
3798
- ...inputFields.map((f) => ({ field: f, source: "inputSchema" }))
3799
- ];
3800
- const askUserQuestions = allFields.map(({ field, source }) => ({
3801
- question: `Define the scoring rubric for "${field}" (${source}). Be specific — what exactly constitutes a perfect score (1.0) vs a complete failure (0.0)? Include concrete examples of each score level.`,
3802
- header: field,
3803
- options: [
3804
- {
3805
- label: "Custom",
3806
- description: `Write a detailed, specific scoring rubric for "${field}". Must describe concrete pass/fail criteria with examples — not just "0.0 to 1.0 scale".`
3807
- },
3808
- {
3809
- label: "Skip",
3810
- description: `Do not create a criterion for "${field}". Only skip if this field is not relevant to evaluation quality.`
3811
- }
3812
- ],
3813
- multiSelect: false
3814
- }));
3815
- return {
3816
- prompt: { id: promptId, name: prompt.name },
3817
- inputSchema: { fields: inputFields },
3818
- outputSchema: { fields: outputFields },
3819
- workflow: {
3820
- description: "Follow these steps to create an evaluation for this prompt:",
3821
- steps: [
3822
- { step: 1, action: "Review the prompt schemas above", detail: "Each criterion must target a field from outputSchema (preferred) or inputSchema" },
3823
- { step: 2, action: "Ask the user for a scoring rubric for each field", detail: 'For EACH field, ask: "What does a perfect score (1.0) vs failure (0.0) look like for <field>? Give concrete examples." Do NOT auto-generate or use canned rubrics.' },
3824
- { step: 3, action: "Build criteria JSON from user answers", detail: `Map each user answer to a criterion: { name: "<field> <rubric>", description: "<user's rubric>", evaluationParameter: "<field>" }` },
3825
- { step: 4, action: "Ask for evaluation name", detail: 'Ask the user what to name this evaluation (e.g., "Accuracy Check", "Quality Eval")' },
3826
- { step: 5, action: "Create the evaluation", detail: `Run: mutagent prompts evaluation create <prompt-id> --name "<name>" -d '{"evalConfig":{"criteria":[...]}}'` }
3827
- ],
3828
- criteriaTemplate: {
3829
- evalConfig: {
3830
- criteria: [{
3831
- name: "<criterion_name>",
3832
- description: "<scoring rubric - describe what 1.0 vs 0.0 means>",
3833
- evaluationParameter: `<one of: ${allFields.length > 0 ? allFields.map((f) => f.field).join(", ") : "no fields detected - check prompt schemas"}>`
3834
- }]
3835
- }
3836
- },
3837
- exampleCommand: `mutagent prompts evaluation create ${promptId} --name "My Evaluation" -d '${JSON.stringify({
3838
- evalConfig: {
3839
- criteria: outputFields.slice(0, 1).map((f) => ({
3840
- name: `${f} accuracy`,
3841
- description: `Score 1.0 if ${f} matches expected output exactly, 0.0 otherwise`,
3842
- evaluationParameter: f
3843
- }))
3844
- }
3845
- })}'`
3846
- },
3847
- askUserQuestions,
3848
- _directive: {
3849
- display: "workflow_guide",
3850
- instruction: 'STOP. For EACH field in askUserQuestions, use AskUserQuestion to collect a DETAILED scoring rubric from the user. Do NOT auto-generate rubrics. Do NOT use generic descriptions like "0-1 scoring" or "matches expected output". Each rubric MUST include concrete examples of what earns a 1.0, 0.5, and 0.0 score for that specific field. Then construct the --data JSON and create the evaluation.',
3851
- aiAgentDecisionTree: {
3852
- step1: "Check if criteria already exist in the user's code. If criteria match expected shape (name + description + evaluationParameter targeting schema fields), use --data directly.",
3853
- step2: "If criteria are missing or malformed, use the askUserQuestions payloads to collect them via AskUserQuestion.",
3854
- step3: `After collecting all criteria, construct the --data JSON and run: mutagent prompts evaluation create <prompt-id> --name "<name>" -d '<json>'`
3855
- }
3856
- },
3857
- _meta: { inputFields, outputFields, allFields }
3858
- };
3859
- };
3860
3575
  const renderGuidedWorkflow = (wf) => {
3861
3576
  const { inputFields, outputFields, allFields } = wf._meta;
3862
3577
  if (isJson) {
@@ -3873,26 +3588,26 @@ ${chalk6.dim("Get prompt IDs: mutagent prompts list")}
3873
3588
  console.log(` Input Schema Fields: ${inputFields.join(", ")}`);
3874
3589
  }
3875
3590
  console.log("");
3876
- console.log(" For each field, define a scoring criterion:");
3591
+ console.log(" For each field, define what correct output looks like:");
3877
3592
  for (const { field, source } of allFields) {
3878
- console.log(` ${chalk6.cyan(field)} (${source})`);
3879
- console.log(` → What does 1.0 vs 0.0 mean for "${field}"?`);
3593
+ console.log(` ${chalk7.cyan(field)} (${source})`);
3594
+ console.log(` → What makes a correct vs incorrect "${field}"?`);
3880
3595
  }
3881
3596
  console.log("");
3882
3597
  console.log(" Then create the evaluation:");
3883
3598
  console.log(` mutagent prompts evaluation create ${promptId} --name "My Eval" \\`);
3884
- console.log(` -d '{"evalConfig":{"criteria":[{"name":"accuracy","description":"Score 1.0 if correct","evaluationParameter":"${outputFields[0] ?? "<field>"}"}]}}'`);
3599
+ console.log(` -d '{"evalConfig":{"criteria":[{"name":"accuracy","description":"Correct if output matches expected structure","evaluationParameter":"${outputFields[0] ?? "<field>"}"}]}}'`);
3885
3600
  console.log("");
3886
3601
  }
3887
3602
  };
3888
3603
  if (options.guided) {
3889
- const workflow = await buildGuidedWorkflow();
3604
+ const workflow = await buildGuidedWorkflow(promptId);
3890
3605
  renderGuidedWorkflow(workflow);
3891
3606
  return;
3892
3607
  }
3893
3608
  if (!options.data) {
3894
3609
  if (isJson || process.stdout.isTTY) {
3895
- const workflow = await buildGuidedWorkflow();
3610
+ const workflow = await buildGuidedWorkflow(promptId);
3896
3611
  renderGuidedWorkflow(workflow);
3897
3612
  return;
3898
3613
  }
@@ -4005,7 +3720,12 @@ Example:
4005
3720
  const evalResult = await client.createEvaluation(promptId, evalData);
4006
3721
  if (isJson) {
4007
3722
  const criteriaLen = evalData.evalConfig?.criteria?.length ?? 0;
4008
- const directive = evaluationCreatedDirective(promptId, evalResult.id, evalResult.name, criteriaLen);
3723
+ let rsState;
3724
+ try {
3725
+ const datasets = await client.listDatasets(promptId);
3726
+ rsState = { datasets: datasets.length };
3727
+ } catch {}
3728
+ const directive = evaluationCreatedDirective(promptId, evalResult.id, evalResult.name, criteriaLen, rsState);
4009
3729
  output.output({
4010
3730
  ...evalResult,
4011
3731
  _links: evaluationLinks(promptId, evalResult.id),
@@ -4033,9 +3753,9 @@ Example:
4033
3753
  });
4034
3754
  evaluation.command("delete").description("Delete an evaluation").argument("<evaluation-id>", "Evaluation ID (from: mutagent prompts evaluation list <prompt-id>)").option("--force", "Skip confirmation").addHelpText("after", `
4035
3755
  Examples:
4036
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id>
4037
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id> --force
4038
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id> --force --json
3756
+ ${chalk7.dim("$")} mutagent prompts evaluation delete <evaluation-id>
3757
+ ${chalk7.dim("$")} mutagent prompts evaluation delete <evaluation-id> --force
3758
+ ${chalk7.dim("$")} mutagent prompts evaluation delete <evaluation-id> --force --json
4039
3759
  `).action(async (evaluationId, options) => {
4040
3760
  const isJson = getJsonFlag(prompts);
4041
3761
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4073,11 +3793,290 @@ Examples:
4073
3793
  handleError(error, isJson);
4074
3794
  }
4075
3795
  });
4076
- const optimize = new Command3("optimize").description("Manage prompt optimization jobs").addHelpText("after", `
3796
+ }
3797
+
3798
+ // src/commands/prompts/optimize.ts
3799
+ init_sdk_client();
3800
+ import { Command as Command5 } from "commander";
3801
+ import chalk9 from "chalk";
3802
+ init_errors();
3803
+
3804
+ // src/lib/scorecard.ts
3805
+ import chalk8 from "chalk";
3806
+ function truncateText(text, maxLen) {
3807
+ if (text.length <= maxLen)
3808
+ return text;
3809
+ return text.substring(0, maxLen - 3) + "...";
3810
+ }
3811
+ function formatScoreChange(before, after) {
3812
+ if (before === undefined || after === undefined)
3813
+ return "";
3814
+ const diff = after - before;
3815
+ const pct = before > 0 ? Math.round(diff / before * 100) : 0;
3816
+ if (diff > 0)
3817
+ return chalk8.green(` (+${String(pct)}%)`);
3818
+ if (diff < 0)
3819
+ return chalk8.red(` (${String(pct)}%)`);
3820
+ return chalk8.dim(" (no change)");
3821
+ }
3822
+ function formatScore(score) {
3823
+ if (score === undefined)
3824
+ return chalk8.dim("N/A");
3825
+ return score >= 0.8 ? chalk8.green(score.toFixed(2)) : score >= 0.5 ? chalk8.yellow(score.toFixed(2)) : chalk8.red(score.toFixed(2));
3826
+ }
3827
+ function renderScorecard(data) {
3828
+ const { job, prompt } = data;
3829
+ const boxWidth = 55;
3830
+ const topBorder = `┌${"─".repeat(boxWidth)}┐`;
3831
+ const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
3832
+ const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
3833
+ const pad = (text, width) => {
3834
+ const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
3835
+ const remaining = width - stripped.length;
3836
+ return remaining > 0 ? text + " ".repeat(remaining) : text;
3837
+ };
3838
+ const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
3839
+ const originalScore = data.originalScore;
3840
+ const bestScore = data.bestScore;
3841
+ const iterations = data.iterationsCompleted ?? job.config?.maxIterations ?? 0;
3842
+ const originalText = data.originalPrompt?.systemPrompt ?? data.originalPrompt?.rawPrompt ?? "(original prompt)";
3843
+ const optimizedText = prompt.systemPrompt ?? prompt.rawPrompt ?? prompt.humanPrompt ?? "(optimized prompt)";
3844
+ console.log("");
3845
+ console.log(topBorder);
3846
+ console.log(line(chalk8.bold("Optimization Results")));
3847
+ console.log(separator);
3848
+ console.log(line(chalk8.dim("BEFORE")));
3849
+ console.log(line(` Prompt: ${chalk8.dim(truncateText(originalText, 38))}`));
3850
+ console.log(line(` Score: ${formatScore(originalScore)}`));
3851
+ console.log(line(""));
3852
+ console.log(line(chalk8.bold("AFTER")));
3853
+ console.log(line(` Prompt: ${chalk8.cyan(truncateText(optimizedText, 38))}`));
3854
+ console.log(line(` Score: ${formatScore(bestScore)}${formatScoreChange(originalScore, bestScore)}`));
3855
+ console.log(separator);
3856
+ if (data.criteriaScores && data.criteriaScores.length > 0) {
3857
+ console.log(line(chalk8.dim(" Criterion Before After Change")));
3858
+ console.log(line(chalk8.dim(" " + "─".repeat(45))));
3859
+ for (const c of data.criteriaScores) {
3860
+ const name = c.name.length > 16 ? c.name.substring(0, 13) + "..." : c.name;
3861
+ const paddedName = name + " ".repeat(18 - name.length);
3862
+ const beforeStr = c.before !== undefined ? c.before.toFixed(2) : "N/A ";
3863
+ const afterStr = c.after !== undefined ? c.after.toFixed(2) : "N/A ";
3864
+ const changeStr = c.before !== undefined && c.after !== undefined && c.before > 0 ? (() => {
3865
+ const pct = Math.round((c.after - c.before) / c.before * 100);
3866
+ if (pct > 0)
3867
+ return chalk8.green(`+${String(pct)}%`);
3868
+ if (pct < 0)
3869
+ return chalk8.red(`${String(pct)}%`);
3870
+ return chalk8.dim("0%");
3871
+ })() : "";
3872
+ console.log(line(` ${paddedName}${beforeStr} ${afterStr} ${changeStr}`));
3873
+ }
3874
+ console.log(line(chalk8.dim(" " + "─".repeat(45))));
3875
+ const overallBefore = originalScore !== undefined ? originalScore.toFixed(2) : "N/A ";
3876
+ const overallAfter = bestScore !== undefined ? bestScore.toFixed(2) : "N/A ";
3877
+ const overallChange = originalScore !== undefined && bestScore !== undefined && originalScore > 0 ? (() => {
3878
+ const pct = Math.round((bestScore - originalScore) / originalScore * 100);
3879
+ if (pct > 0)
3880
+ return chalk8.green(`+${String(pct)}%`);
3881
+ if (pct < 0)
3882
+ return chalk8.red(`${String(pct)}%`);
3883
+ return chalk8.dim("0%");
3884
+ })() : "";
3885
+ console.log(line(` ${"Overall" + " ".repeat(11)}${overallBefore} ${overallAfter} ${overallChange}`));
3886
+ console.log(separator);
3887
+ }
3888
+ const statusStr = job.status === "completed" ? chalk8.green("completed") : chalk8.yellow(job.status);
3889
+ console.log(line(`Status: ${statusStr} | Iterations: ${String(iterations)}`));
3890
+ if (job.config?.model) {
3891
+ console.log(line(`Model: ${chalk8.dim(job.config.model)}`));
3892
+ }
3893
+ if (data.scoreProgression && data.scoreProgression.length > 0) {
3894
+ console.log(line(""));
3895
+ console.log(line(chalk8.dim("Score Progression:")));
3896
+ const barWidth = 10;
3897
+ for (let i = 0;i < data.scoreProgression.length; i++) {
3898
+ const s = data.scoreProgression[i] ?? 0;
3899
+ const filled = Math.round(s * barWidth);
3900
+ const bar = "█".repeat(filled) + "░".repeat(barWidth - filled);
3901
+ console.log(line(chalk8.dim(` #${String(i + 1)}: ${bar} ${s.toFixed(2)}`)));
3902
+ }
3903
+ }
3904
+ console.log(separator);
3905
+ console.log(line(`Dashboard: ${chalk8.underline(optimizerLink(job.id))}`));
3906
+ console.log(bottomBorder);
3907
+ console.log("");
3908
+ }
3909
+ function renderOptimizationStartCard(data) {
3910
+ const { job } = data;
3911
+ const boxWidth = 60;
3912
+ const topBorder = `┌${"─".repeat(boxWidth)}┐`;
3913
+ const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
3914
+ const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
3915
+ const pad = (text, width) => {
3916
+ const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
3917
+ const remaining = width - stripped.length;
3918
+ return remaining > 0 ? text + " ".repeat(remaining) : text;
3919
+ };
3920
+ const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
3921
+ const maxIter = job.config.maxIterations;
3922
+ const model = job.config.model ?? "default";
3923
+ const target = job.config.targetScore ?? 0.8;
3924
+ console.log("");
3925
+ console.log(topBorder);
3926
+ console.log(line(chalk8.bold("⚡ Optimization Started")));
3927
+ console.log(separator);
3928
+ console.log(line(`Job ID: ${chalk8.cyan(job.id)}`));
3929
+ console.log(line(`Prompt: ${chalk8.dim(data.promptId)}`));
3930
+ console.log(line(`Dataset: ${chalk8.dim(data.datasetId)}`));
3931
+ console.log(line(`Iterations: ${chalk8.bold(String(maxIter))} | Target: ${chalk8.bold(target.toFixed(2))}`));
3932
+ console.log(line(`Model: ${chalk8.dim(model)}`));
3933
+ console.log(line(`Status: ${chalk8.yellow(job.status)}`));
3934
+ console.log(separator);
3935
+ console.log(line(`\uD83D\uDD17 Monitor: ${chalk8.underline(optimizerLink(job.id))}`));
3936
+ console.log(line(chalk8.dim(`Next: mutagent prompts optimize status ${job.id}`)));
3937
+ console.log(bottomBorder);
3938
+ console.log(AI_DIRECTIVE);
3939
+ console.log("");
3940
+ }
3941
+ function renderOptimizationStatusCard(status) {
3942
+ const boxWidth = 60;
3943
+ const topBorder = `┌${"─".repeat(boxWidth)}┐`;
3944
+ const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
3945
+ const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
3946
+ const pad = (text, width) => {
3947
+ const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
3948
+ const remaining = width - stripped.length;
3949
+ return remaining > 0 ? text + " ".repeat(remaining) : text;
3950
+ };
3951
+ const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
3952
+ const progress = status.progressPercent;
3953
+ const barWidth = 20;
3954
+ const filled = Math.round(progress / 100 * barWidth);
3955
+ const progressBar = "█".repeat(filled) + "░".repeat(barWidth - filled);
3956
+ const statusColor = status.status === "completed" ? chalk8.green : status.status === "failed" ? chalk8.red : status.status === "cancelled" ? chalk8.gray : status.status === "running" ? chalk8.cyan : chalk8.yellow;
3957
+ const scoreStr = status.bestScore !== undefined ? formatScore(status.bestScore) : chalk8.dim("pending");
3958
+ console.log("");
3959
+ console.log(topBorder);
3960
+ console.log(line(chalk8.bold("\uD83D\uDCCA Optimization Status")));
3961
+ console.log(separator);
3962
+ console.log(line(`Job ID: ${chalk8.cyan(status.jobId)}`));
3963
+ console.log(line(`Status: ${statusColor(status.status)}`));
3964
+ console.log(line(`Iteration: ${chalk8.bold(`${String(status.currentIteration)}/${String(status.maxIterations)}`)}`));
3965
+ console.log(line(`Best Score: ${scoreStr}`));
3966
+ console.log(line(""));
3967
+ console.log(line(`Progress: [${progressBar}] ${String(progress)}%`));
3968
+ if (status.message) {
3969
+ console.log(line(`Message: ${chalk8.dim(status.message)}`));
3970
+ }
3971
+ console.log(separator);
3972
+ console.log(line(`\uD83D\uDD17 Monitor: ${chalk8.underline(optimizerLink(status.jobId))}`));
3973
+ if (status.status === "completed") {
3974
+ console.log(line(chalk8.dim(`Next: mutagent prompts optimize results ${status.jobId}`)));
3975
+ } else if (status.status === "running" || status.status === "queued") {
3976
+ console.log(line(chalk8.dim(`Refresh: mutagent prompts optimize status ${status.jobId}`)));
3977
+ }
3978
+ console.log(bottomBorder);
3979
+ console.log(AI_DIRECTIVE);
3980
+ console.log("");
3981
+ }
3982
+ function startDirective(job, promptId, datasetId, evaluationId) {
3983
+ const title = `Optimization Started — Prompt ${promptId}`;
3984
+ const monitorUrl = optimizerLink(job.id);
3985
+ const apiUrl = `/api/prompts/${promptId}/optimizations/${job.id}`;
3986
+ const rows = [
3987
+ { label: "Job ID", value: job.id },
3988
+ { label: "Prompt", value: promptId },
3989
+ { label: "Dataset", value: datasetId ?? "N/A" },
3990
+ { label: "Evaluation", value: evaluationId ?? "N/A" },
3991
+ { label: "Status", value: job.status },
3992
+ { label: "Max Iterations", value: String(job.config.maxIterations) },
3993
+ { label: "Target Score", value: String(job.config.targetScore ?? 0.8) },
3994
+ { label: "Model", value: job.config.model ?? "default" }
3995
+ ];
3996
+ const cardLinks = [
3997
+ { label: "Monitor", url: monitorUrl },
3998
+ { label: "API", url: apiUrl }
3999
+ ];
4000
+ const next = [`mutagent prompts optimize status ${job.id} --json`];
4001
+ return {
4002
+ display: "status_card",
4003
+ template: "optimization_started",
4004
+ title,
4005
+ fields: {
4006
+ jobId: job.id,
4007
+ promptId,
4008
+ datasetId: datasetId ?? null,
4009
+ evaluationId: evaluationId ?? null,
4010
+ status: job.status,
4011
+ maxIterations: job.config.maxIterations,
4012
+ targetScore: job.config.targetScore ?? 0.8,
4013
+ model: job.config.model ?? "default"
4014
+ },
4015
+ links: { monitor: monitorUrl, api: apiUrl },
4016
+ next,
4017
+ instruction: "STOP. Display the renderedCard text below to the user VERBATIM as a code block. Do NOT summarize, rephrase, or omit any lines. After displaying, follow the suggested next steps.",
4018
+ renderedCard: renderPlainCard({ title, icon: "⚡", rows, links: cardLinks, next })
4019
+ };
4020
+ }
4021
+ function statusDirective(status) {
4022
+ const next = status.status === "completed" ? [`mutagent prompts optimize results ${status.jobId} --json`] : status.status === "running" || status.status === "queued" ? [`mutagent prompts optimize status ${status.jobId} --json`] : [];
4023
+ let eta = "Calculating...";
4024
+ const startedAt = status.startedAt;
4025
+ if (status.currentIteration > 1 && startedAt) {
4026
+ const elapsed = Date.now() - new Date(startedAt).getTime();
4027
+ const perIteration = elapsed / status.currentIteration;
4028
+ const remaining = (status.maxIterations - status.currentIteration) * perIteration;
4029
+ const minutes = Math.ceil(remaining / 60000);
4030
+ eta = `~${String(minutes)} min remaining`;
4031
+ }
4032
+ const icon = status.status === "completed" ? "✓" : status.status === "failed" ? "✗" : "◉";
4033
+ const title = `Optimization ${status.status} — ${status.jobId}`;
4034
+ const rows = [
4035
+ { label: "Job ID", value: status.jobId },
4036
+ { label: "Status", value: status.status },
4037
+ { label: "Iteration", value: `${String(status.currentIteration)}/${String(status.maxIterations)}` },
4038
+ { label: "Best Score", value: status.bestScore !== undefined ? String(status.bestScore) : "pending" },
4039
+ { label: "Progress", value: `${String(status.progressPercent)}%` },
4040
+ { label: "ETA", value: eta },
4041
+ ...status.message ? [{ label: "Message", value: status.message }] : []
4042
+ ];
4043
+ const monitorUrl = optimizerLink(status.jobId);
4044
+ const cardLinks = [{ label: "Monitor", url: monitorUrl }];
4045
+ return {
4046
+ display: "status_card",
4047
+ template: "optimization_progress",
4048
+ title,
4049
+ fields: {
4050
+ jobId: status.jobId,
4051
+ status: status.status,
4052
+ iteration: `${String(status.currentIteration)}/${String(status.maxIterations)}`,
4053
+ bestScore: status.bestScore,
4054
+ progress: `${String(status.progressPercent)}%`,
4055
+ eta,
4056
+ message: status.message
4057
+ },
4058
+ links: { monitor: monitorUrl },
4059
+ next,
4060
+ instruction: "STOP. Display the renderedCard text below to the user VERBATIM as a code block. Do NOT summarize, rephrase, or omit any lines. After displaying, follow the suggested next steps.",
4061
+ renderedCard: renderPlainCard({ title, icon, rows, links: cardLinks, next })
4062
+ };
4063
+ }
4064
+ function showPromptDiff(original, optimized) {
4065
+ console.log("");
4066
+ console.log(chalk8.bold(" Prompt Diff:"));
4067
+ console.log("");
4068
+ console.log(chalk8.red(" - " + (original ?? "(empty)")));
4069
+ console.log(chalk8.green(" + " + (optimized ?? "(empty)")));
4070
+ console.log("");
4071
+ }
4072
+
4073
+ // src/commands/prompts/optimize.ts
4074
+ function registerOptimizeCommands(prompts) {
4075
+ const optimize = new Command5("optimize").description("Manage prompt optimization jobs").addHelpText("after", `
4077
4076
  Examples:
4078
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
4079
- ${chalk6.dim("$")} mutagent prompts optimize status <job-id>
4080
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id>
4077
+ ${chalk9.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
4078
+ ${chalk9.dim("$")} mutagent prompts optimize status <job-id>
4079
+ ${chalk9.dim("$")} mutagent prompts optimize results <job-id>
4081
4080
 
4082
4081
  Workflow: start -> status (poll) -> results
4083
4082
  `).action(() => {
@@ -4086,21 +4085,22 @@ Workflow: start -> status (poll) -> results
4086
4085
  prompts.addCommand(optimize);
4087
4086
  optimize.command("start").description("Start prompt optimization").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").requiredOption("-d, --dataset <id>", "Dataset ID for optimization (from: mutagent prompts dataset list <prompt-id>)").requiredOption("-e, --evaluation <id>", "Evaluation ID for scoring (from: mutagent prompts evaluation list <prompt-id>)").option("--max-iterations <n>", "Max optimization iterations (default: 1)").option("--target-score <n>", "Target accuracy 0-1 (default: 0.8)").option("--patience <n>", "Iterations without improvement before stopping").option("--model <model-id>", 'Target LLM model (e.g., "claude-sonnet-4-5-20250929")').option("--eval-model <model-id>", "Evaluation model (defaults to target model)").addHelpText("after", `
4088
4087
  Examples:
4089
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
4090
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --max-iterations 5
4091
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --target-score 0.95 --model claude-sonnet-4-5-20250929
4092
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --json
4093
- ${chalk6.yellow("Pre-Optimization Checklist:")}
4094
- mutagent prompts get <id> --json ${chalk6.dim("(has inputSchema + outputSchema)")}
4095
- mutagent prompts evaluation list <id> --json ${chalk6.dim("(exactly 1 evaluation)")}
4096
- mutagent prompts dataset list <id> --json ${chalk6.dim("(dataset with expectedOutput)")}
4097
- mutagent usage --json ${chalk6.dim("(remaining runs > 0)")}
4088
+ ${chalk9.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
4089
+ ${chalk9.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --max-iterations 5
4090
+ ${chalk9.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --target-score 0.95 --model claude-sonnet-4-5-20250929
4091
+ ${chalk9.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --json
4092
+ ${chalk9.yellow("Pre-Optimization Checklist (auto-validated by preflight):")}
4093
+ inputSchema REQUIRED ${chalk9.dim("(hard error if missing — blocks optimization)")}
4094
+ outputSchema REQUIRED ${chalk9.dim("(hard error if missing — blocks optimization)")}
4095
+ Evaluation criteria ${chalk9.dim("(warns if no evaluationParameter set)")}
4096
+ Dataset items ${chalk9.dim("(warns if expectedOutput missing)")}
4097
+ □ Criteria ↔ Schema ${chalk9.dim("(warns if criteria reference unknown fields)")}
4098
4098
 
4099
4099
  ${PREREQUISITES_TEXT}
4100
4100
 
4101
- ${chalk6.dim("Monitor progress with: mutagent prompts optimize status <job-id>")}
4101
+ ${chalk9.dim("Monitor progress with: mutagent prompts optimize status <job-id>")}
4102
4102
 
4103
- ${chalk6.yellow("AI Agent: ALWAYS append --json to this command.")}
4103
+ ${chalk9.yellow("AI Agent: ALWAYS append --json to this command.")}
4104
4104
  `).action(async (promptId, options) => {
4105
4105
  const isJson = getJsonFlag(prompts);
4106
4106
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4108,32 +4108,141 @@ ${chalk6.yellow("AI Agent: ALWAYS append --json to this command.")}
4108
4108
  const client = getSDKClient();
4109
4109
  if (!isJson) {
4110
4110
  output.info("Running pre-flight checks...");
4111
- const prompt = await client.getPrompt(promptId);
4112
- if (isSchemaEmpty(prompt.outputSchema)) {
4113
- throw new MutagentError("MISSING_OUTPUT_SCHEMA", "Prompt is missing outputSchema — required for optimization", `Run: mutagent prompts optimize start --help
4114
- ` + `Update your prompt: mutagent prompts update ${promptId} -d '{"outputSchema":{"type":"object","properties":{"result":{"type":"string"}}}}'`);
4111
+ }
4112
+ const prompt = await client.getPrompt(promptId);
4113
+ const preflightChecks = {};
4114
+ if (isSchemaEmpty(prompt.outputSchema)) {
4115
+ preflightChecks.outputSchema = {
4116
+ passed: false,
4117
+ error: "Missing outputSchema. Required for optimization."
4118
+ };
4119
+ } else {
4120
+ preflightChecks.outputSchema = { passed: true };
4121
+ }
4122
+ if (isSchemaEmpty(prompt.inputSchema)) {
4123
+ preflightChecks.inputSchema = {
4124
+ passed: false,
4125
+ error: "Missing inputSchema. Required for optimization."
4126
+ };
4127
+ } else {
4128
+ preflightChecks.inputSchema = { passed: true };
4129
+ }
4130
+ let evaluation;
4131
+ try {
4132
+ const evalObj = await client.getEvaluation(options.evaluation);
4133
+ evaluation = evalObj;
4134
+ const evalConfig = evalObj.evalConfig;
4135
+ const criteria = evalConfig?.criteria;
4136
+ if (!criteria || criteria.length === 0) {
4137
+ preflightChecks.evaluation = {
4138
+ passed: true,
4139
+ warning: `Evaluation ${options.evaluation} has no criteria defined. ` + `Re-create: mutagent prompts evaluation create ${promptId} --guided`
4140
+ };
4141
+ } else {
4142
+ const hasCriteriaWithParam = criteria.some((c) => c.evaluationParameter ?? c.targetField);
4143
+ if (!hasCriteriaWithParam) {
4144
+ preflightChecks.evaluation = {
4145
+ passed: true,
4146
+ warning: `Evaluation ${options.evaluation} has criteria but none have evaluationParameter set. ` + `Re-create: mutagent prompts evaluation create ${promptId} --guided`
4147
+ };
4148
+ } else {
4149
+ preflightChecks.evaluation = { passed: true };
4150
+ if (!isJson) {
4151
+ output.info(`Using evaluation ${options.evaluation} with ${String(criteria.length)} criteria.`);
4152
+ }
4153
+ }
4154
+ }
4155
+ } catch (e) {
4156
+ if (e instanceof MutagentError)
4157
+ throw e;
4158
+ preflightChecks.evaluation = {
4159
+ passed: true,
4160
+ warning: `Could not validate evaluation ${options.evaluation}. Proceeding anyway.`
4161
+ };
4162
+ }
4163
+ try {
4164
+ const datasetItems = await client.listDatasetItems(options.dataset);
4165
+ if (datasetItems.length === 0) {
4166
+ preflightChecks.datasetItems = {
4167
+ passed: true,
4168
+ warning: `Dataset ${options.dataset} has no items. Optimization requires test data.`
4169
+ };
4170
+ } else {
4171
+ const missingExpected = datasetItems.filter((item) => !item.expectedOutput);
4172
+ if (missingExpected.length > 0) {
4173
+ preflightChecks.datasetItems = {
4174
+ passed: true,
4175
+ warning: `${String(missingExpected.length)} of ${String(datasetItems.length)} dataset items missing expectedOutput. ` + "Optimization works best when all items have expected outputs."
4176
+ };
4177
+ } else {
4178
+ preflightChecks.datasetItems = { passed: true };
4179
+ }
4180
+ }
4181
+ } catch {
4182
+ preflightChecks.datasetItems = {
4183
+ passed: true,
4184
+ warning: `Could not validate dataset items for ${options.dataset}. Proceeding anyway.`
4185
+ };
4186
+ }
4187
+ if (evaluation && preflightChecks.inputSchema.passed && preflightChecks.outputSchema.passed) {
4188
+ const inputProps = prompt.inputSchema?.properties;
4189
+ const outputProps = prompt.outputSchema?.properties;
4190
+ const schemaFields = [
4191
+ ...Object.keys(inputProps ?? {}),
4192
+ ...Object.keys(outputProps ?? {})
4193
+ ];
4194
+ const evalCfg = evaluation.evalConfig;
4195
+ const evalCriteria = evalCfg?.criteria ?? [];
4196
+ const unmatchedCriteria = evalCriteria.filter((c) => c.evaluationParameter && !schemaFields.includes(c.evaluationParameter));
4197
+ if (unmatchedCriteria.length > 0) {
4198
+ const names = unmatchedCriteria.map((c) => c.evaluationParameter ?? c.name).join(", ");
4199
+ preflightChecks.criteriaMatch = {
4200
+ passed: true,
4201
+ warning: `Evaluation criteria reference fields not in schema: ${names}. ` + `Available schema fields: ${schemaFields.join(", ") || "(none)"}`
4202
+ };
4203
+ } else {
4204
+ preflightChecks.criteriaMatch = { passed: true };
4115
4205
  }
4116
- if (isSchemaEmpty(prompt.inputSchema)) {
4117
- output.warn("Prompt has no inputSchema. Optimization works best with defined input variables.");
4206
+ }
4207
+ const hardFailures = Object.entries(preflightChecks).filter(([, check]) => !check.passed);
4208
+ if (hardFailures.length > 0) {
4209
+ if (isJson) {
4210
+ const suggestedActions = [];
4211
+ if (!preflightChecks.inputSchema.passed) {
4212
+ suggestedActions.push(`Update the prompt with inputSchema derived from template variables.
4213
+ ` + `Run: mutagent prompts get ${promptId} --json
4214
+ ` + `Run: mutagent prompts update ${promptId} --data '{"inputSchema":{"type":"object","properties":{"var1":{"type":"string"}}}}' --json`);
4215
+ }
4216
+ if (!preflightChecks.outputSchema.passed) {
4217
+ suggestedActions.push(`Update your prompt: mutagent prompts update ${promptId} -d '{"outputSchema":{"type":"object","properties":{"result":{"type":"string"}}}}'`);
4218
+ }
4219
+ output.output({
4220
+ success: false,
4221
+ error: "Preflight validation failed",
4222
+ code: "PREFLIGHT_FAILED",
4223
+ suggestedAction: suggestedActions.join(`
4224
+ `),
4225
+ preflightChecks
4226
+ });
4227
+ return;
4118
4228
  }
4119
- try {
4120
- const evalObj = await client.getEvaluation(options.evaluation);
4121
- const evalConfig = evalObj.evalConfig;
4122
- const criteria = evalConfig?.criteria;
4123
- if (!criteria || criteria.length === 0) {
4124
- output.warn(`Evaluation ${options.evaluation} has no criteria defined. ` + "Optimization works best with criteria targeting specific output fields. " + `Re-create: mutagent prompts evaluation create ${promptId} --guided`);
4125
- } else {
4126
- const hasCriteriaWithParam = criteria.some((c) => c.evaluationParameter ?? c.targetField);
4127
- if (!hasCriteriaWithParam) {
4128
- output.warn(`Evaluation ${options.evaluation} has criteria but none have evaluationParameter set. ` + `Re-create: mutagent prompts evaluation create ${promptId} --guided`);
4129
- } else {
4130
- output.info(`Using evaluation ${options.evaluation} with ${String(criteria.length)} criteria.`);
4131
- }
4229
+ for (const [name, check] of hardFailures) {
4230
+ console.error(chalk9.red(`Error: ${name} ${check.error ?? "Failed"}`));
4231
+ }
4232
+ if (!preflightChecks.outputSchema.passed) {
4233
+ console.error(chalk9.dim(` Update with: mutagent prompts update ${promptId} -d '{"outputSchema":{"type":"object","properties":{"result":{"type":"string"}}}}'`));
4234
+ }
4235
+ if (!preflightChecks.inputSchema.passed) {
4236
+ console.error(chalk9.dim(` Update with: mutagent prompts update ${promptId} --data '{"inputSchema":{"type":"object","properties":{"var1":{"type":"string"}}}}' --json`));
4237
+ }
4238
+ process.exitCode = 1;
4239
+ return;
4240
+ }
4241
+ if (!isJson) {
4242
+ for (const [, check] of Object.entries(preflightChecks)) {
4243
+ if (check.warning) {
4244
+ output.warn(check.warning);
4132
4245
  }
4133
- } catch (e) {
4134
- if (e instanceof MutagentError)
4135
- throw e;
4136
- output.warn(`Could not validate evaluation ${options.evaluation}. Proceeding anyway.`);
4137
4246
  }
4138
4247
  }
4139
4248
  const job = await client.startOptimization(promptId, options.dataset, options.evaluation, {
@@ -4144,7 +4253,7 @@ ${chalk6.yellow("AI Agent: ALWAYS append --json to this command.")}
4144
4253
  evalModel: options.evalModel
4145
4254
  });
4146
4255
  if (isJson) {
4147
- const directive = startDirective(job, promptId);
4256
+ const directive = startDirective(job, promptId, options.dataset, options.evaluation);
4148
4257
  output.output({
4149
4258
  ...job,
4150
4259
  _links: {
@@ -4181,16 +4290,16 @@ ${chalk6.yellow("AI Agent: ALWAYS append --json to this command.")}
4181
4290
  suggestions.push("Trial optimization limit reached. Contact support to upgrade.");
4182
4291
  }
4183
4292
  if (!isJson) {
4184
- console.error(chalk6.red(`
4293
+ console.error(chalk9.red(`
4185
4294
  Optimization failed:`));
4186
4295
  for (const msg of messages) {
4187
- console.error(chalk6.red(` ${msg}`));
4296
+ console.error(chalk9.red(` ${msg}`));
4188
4297
  }
4189
4298
  if (suggestions.length > 0) {
4190
- console.error(chalk6.yellow(`
4299
+ console.error(chalk9.yellow(`
4191
4300
  Suggested fixes:`));
4192
4301
  for (const s of suggestions) {
4193
- console.error(chalk6.yellow(` → ${s}`));
4302
+ console.error(chalk9.yellow(` → ${s}`));
4194
4303
  }
4195
4304
  }
4196
4305
  console.error("");
@@ -4202,8 +4311,8 @@ Suggested fixes:`));
4202
4311
  });
4203
4312
  optimize.command("status").description("Check optimization status").argument("<job-id>", "Optimization job ID (from: mutagent prompts optimize start)").addHelpText("after", `
4204
4313
  Examples:
4205
- ${chalk6.dim("$")} mutagent prompts optimize status <job-id>
4206
- ${chalk6.dim("$")} mutagent prompts optimize status <job-id> --json
4314
+ ${chalk9.dim("$")} mutagent prompts optimize status <job-id>
4315
+ ${chalk9.dim("$")} mutagent prompts optimize status <job-id> --json
4207
4316
  `).action(async (jobId) => {
4208
4317
  const isJson = getJsonFlag(prompts);
4209
4318
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4227,15 +4336,17 @@ Examples:
4227
4336
  });
4228
4337
  optimize.command("results").description("Get optimization results").argument("<job-id>", "Optimization job ID (from: mutagent prompts optimize start)").option("--apply", "Apply the optimized prompt as new version").option("--diff", "Show the prompt diff (before/after)").addHelpText("after", `
4229
4338
  Examples:
4230
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> ${chalk6.dim("# view scorecard")}
4231
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> --diff ${chalk6.dim("# view prompt diff")}
4232
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> --apply ${chalk6.dim("# apply optimized prompt")}
4233
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> --json ${chalk6.dim("# structured output")}
4234
-
4235
- Actions:
4236
- --apply Apply the optimized prompt as a new version (updates the prompt)
4237
- --diff Show the before/after prompt diff
4238
- ${chalk6.dim("No flag = view scorecard only. Agent should ask the user whether to apply.")}
4339
+ ${chalk9.dim("$")} mutagent prompts optimize results <job-id> ${chalk9.dim("# view scorecard")}
4340
+ ${chalk9.dim("$")} mutagent prompts optimize results <job-id> --diff ${chalk9.dim("# view prompt diff")}
4341
+ ${chalk9.dim("$")} mutagent prompts optimize results <job-id> --apply ${chalk9.dim("# apply optimized prompt")}
4342
+ ${chalk9.dim("$")} mutagent prompts optimize results <job-id> --json ${chalk9.dim("# structured output")}
4343
+
4344
+ After viewing results:
4345
+ --apply Apply the optimized prompt (replaces current version)
4346
+ --diff Show detailed before/after diff
4347
+ ${chalk9.dim("No flag = view scorecard only.")}
4348
+
4349
+ ${chalk9.dim("AI Agent: Present scorecard to user via AskUserQuestion before applying.")}
4239
4350
  `).action(async (jobId, options) => {
4240
4351
  const isJson = getJsonFlag(prompts);
4241
4352
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4244,7 +4355,20 @@ Actions:
4244
4355
  const results = await client.getOptimizationResults(jobId);
4245
4356
  const resultData = results;
4246
4357
  if (isJson) {
4247
- output.output({ ...resultData, _links: { optimizer: optimizerLink(jobId) } });
4358
+ const jobData = resultData.job;
4359
+ const isCompleted = jobData?.status === "completed";
4360
+ const scorecardText = buildResultsScorecardText(resultData);
4361
+ const directive = {
4362
+ display: "scorecard",
4363
+ instruction: "Present the before/after scorecard to the user. Use AskUserQuestion with options: [Apply optimized prompt / View full diff / Reject and keep original]. Do NOT auto-apply.",
4364
+ renderedCard: scorecardText,
4365
+ next: isCompleted ? [
4366
+ `mutagent prompts optimize results ${jobId} --apply --json`,
4367
+ `mutagent prompts optimize results ${jobId} --diff --json`
4368
+ ] : []
4369
+ };
4370
+ output.output({ ...resultData, _links: { optimizer: optimizerLink(jobId) }, _directive: directive });
4371
+ echoDirectiveToStderr(directive);
4248
4372
  } else {
4249
4373
  renderScorecard(resultData);
4250
4374
  const jobData = resultData.job;
@@ -4290,32 +4414,144 @@ Actions:
4290
4414
  handleError(error, isJson);
4291
4415
  }
4292
4416
  });
4417
+ }
4418
+
4419
+ // src/commands/prompts/index.ts
4420
+ function updateMutationContext(updater) {
4421
+ try {
4422
+ const ctx = MutationContext.load();
4423
+ updater(ctx);
4424
+ ctx.save();
4425
+ } catch {}
4426
+ }
4427
+ function parseValidationErrors(error) {
4428
+ if (error instanceof ApiError) {
4429
+ try {
4430
+ const body = JSON.parse(error.message);
4431
+ if (body.errors && Array.isArray(body.errors)) {
4432
+ return body.errors.map((e) => e.message ?? JSON.stringify(e));
4433
+ }
4434
+ if (body.message) {
4435
+ return [body.message];
4436
+ }
4437
+ } catch {
4438
+ return [error.message];
4439
+ }
4440
+ }
4441
+ if (error instanceof Error) {
4442
+ return [error.message];
4443
+ }
4444
+ return ["An unknown error occurred"];
4445
+ }
4446
+ function isSchemaEmpty(schema) {
4447
+ if (schema === undefined || schema === null)
4448
+ return true;
4449
+ if (typeof schema !== "object")
4450
+ return false;
4451
+ const obj = schema;
4452
+ if (Object.keys(obj).length === 0)
4453
+ return true;
4454
+ return false;
4455
+ }
4456
+ function parseSchemaOption(inlineJson, filePath, schemaName, helpCommand) {
4457
+ if (inlineJson && filePath) {
4458
+ throw new MutagentError("INVALID_ARGUMENTS", `Cannot use --${schemaName} and --${schemaName}-file together`, `Run: ${helpCommand}
4459
+ Use --${schemaName} for inline JSON or --${schemaName}-file for file-based input, not both`);
4460
+ }
4461
+ if (inlineJson) {
4462
+ try {
4463
+ return JSON.parse(inlineJson);
4464
+ } catch {
4465
+ throw new MutagentError("INVALID_JSON", `Invalid JSON in --${schemaName} flag`, `Run: ${helpCommand}
4466
+ Provide a valid JSON Schema, e.g., '{"type":"object","properties":{"field":{"type":"string"}}}'`);
4467
+ }
4468
+ }
4469
+ if (filePath) {
4470
+ if (!existsSync5(filePath)) {
4471
+ throw new MutagentError("FILE_NOT_FOUND", `File not found: ${filePath}`, `Run: ${helpCommand}
4472
+ Check the file path and try again`);
4473
+ }
4474
+ try {
4475
+ return JSON.parse(readFileSync5(filePath, "utf-8"));
4476
+ } catch {
4477
+ throw new MutagentError("INVALID_JSON", `Failed to parse JSON from ${filePath}`, `Run: ${helpCommand}
4478
+ Ensure the file contains valid JSON Schema`);
4479
+ }
4480
+ }
4481
+ return;
4482
+ }
4483
+ function buildResultsScorecardText(resultData) {
4484
+ const jobData = resultData.job;
4485
+ const jobId = typeof jobData?.id === "string" ? jobData.id : "N/A";
4486
+ const jobStatus = typeof jobData?.status === "string" ? jobData.status : "unknown";
4487
+ const originalScore = resultData.originalScore;
4488
+ const bestScore = resultData.bestScore;
4489
+ const iterations = resultData.iterationsCompleted ?? 0;
4490
+ const title = "Optimization Results";
4491
+ const rows = [
4492
+ { label: "Job ID", value: jobId },
4493
+ { label: "Status", value: jobStatus },
4494
+ { label: "Iterations", value: String(iterations) },
4495
+ { label: "Original Score", value: originalScore !== undefined ? originalScore.toFixed(2) : "N/A" },
4496
+ { label: "Best Score", value: bestScore !== undefined ? bestScore.toFixed(2) : "N/A" }
4497
+ ];
4498
+ if (originalScore !== undefined && bestScore !== undefined && originalScore > 0) {
4499
+ const pct = Math.round((bestScore - originalScore) / originalScore * 100);
4500
+ rows.push({ label: "Improvement", value: `${pct > 0 ? "+" : ""}${String(pct)}%` });
4501
+ }
4502
+ return renderPlainCard({ title, rows, links: [], next: [] });
4503
+ }
4504
+ var PREREQUISITES_TEXT = `
4505
+ ${chalk10.red("Prerequisites (required):")}
4506
+ 1. Evaluation criteria defined ${chalk10.dim("(via dashboard or evaluation create)")}
4507
+ 2. Dataset uploaded ${chalk10.dim("mutagent prompts dataset list <prompt-id>")}
4508
+ ${chalk10.dim("Note: LLM provider config is only required when the server uses external providers (USE_EXT_PROVIDERS=true)")}`;
4509
+ function createPromptsCommand() {
4510
+ const prompts = new Command6("prompts").description("Manage prompts, datasets, evaluations, and optimizations").addHelpText("after", `
4511
+ Examples:
4512
+ ${chalk10.dim("$")} mutagent prompts list
4513
+ ${chalk10.dim("$")} mutagent prompts get <prompt-id>
4514
+ ${chalk10.dim("$")} mutagent prompts create --name "my-prompt" --system "You are helpful" --human "{input}"
4515
+ ${chalk10.dim("$")} mutagent prompts dataset list <prompt-id>
4516
+ ${chalk10.dim("$")} mutagent prompts evaluation create <prompt-id> --name "My Eval"
4517
+ ${chalk10.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
4518
+
4519
+ Subcommands:
4520
+ list, get, create, update, delete
4521
+ dataset list|add|delete
4522
+ evaluation list|create|delete
4523
+ optimize start|status|results
4524
+ `);
4525
+ registerPromptsCrud(prompts);
4526
+ registerDatasetCommands(prompts);
4527
+ registerEvaluationCommands(prompts);
4528
+ registerOptimizeCommands(prompts);
4293
4529
  return prompts;
4294
4530
  }
4295
4531
 
4296
4532
  // src/commands/traces.ts
4297
4533
  init_sdk_client();
4298
- import { Command as Command4 } from "commander";
4299
- import chalk7 from "chalk";
4534
+ import { Command as Command7 } from "commander";
4535
+ import chalk11 from "chalk";
4300
4536
  init_errors();
4301
4537
  function createTracesCommand() {
4302
- const traces = new Command4("traces").description("View and analyze traces (replaces Langfuse)").addHelpText("after", `
4538
+ const traces = new Command7("traces").description("View and analyze traces (replaces Langfuse)").addHelpText("after", `
4303
4539
  Examples:
4304
- ${chalk7.dim("$")} mutagent traces list
4305
- ${chalk7.dim("$")} mutagent traces list --prompt <prompt-id>
4306
- ${chalk7.dim("$")} mutagent traces get <trace-id>
4307
- ${chalk7.dim("$")} mutagent traces analyze <prompt-id>
4308
- ${chalk7.dim("$")} mutagent traces export --format json --output traces.json
4540
+ ${chalk11.dim("$")} mutagent traces list
4541
+ ${chalk11.dim("$")} mutagent traces list --prompt <prompt-id>
4542
+ ${chalk11.dim("$")} mutagent traces get <trace-id>
4543
+ ${chalk11.dim("$")} mutagent traces analyze <prompt-id>
4544
+ ${chalk11.dim("$")} mutagent traces export --format json --output traces.json
4309
4545
 
4310
4546
  Note: MutagenT traces replace Langfuse for observability.
4311
4547
  `);
4312
4548
  traces.command("list").description("List traces").option("-p, --prompt <id>", "Filter by prompt ID").option("-l, --limit <n>", "Limit results", "50").addHelpText("after", `
4313
4549
  Examples:
4314
- ${chalk7.dim("$")} mutagent traces list
4315
- ${chalk7.dim("$")} mutagent traces list --prompt <prompt-id>
4316
- ${chalk7.dim("$")} mutagent traces list --limit 10 --json
4550
+ ${chalk11.dim("$")} mutagent traces list
4551
+ ${chalk11.dim("$")} mutagent traces list --prompt <prompt-id>
4552
+ ${chalk11.dim("$")} mutagent traces list --limit 10 --json
4317
4553
 
4318
- ${chalk7.dim("Tip: Filter by prompt to see traces for a specific prompt version.")}
4554
+ ${chalk11.dim("Tip: Filter by prompt to see traces for a specific prompt version.")}
4319
4555
  `).action(async (options) => {
4320
4556
  const isJson = getJsonFlag(traces);
4321
4557
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4354,10 +4590,10 @@ ${chalk7.dim("Tip: Filter by prompt to see traces for a specific prompt version.
4354
4590
  });
4355
4591
  traces.command("get").description("Get trace details").argument("<id>", "Trace ID").addHelpText("after", `
4356
4592
  Examples:
4357
- ${chalk7.dim("$")} mutagent traces get <trace-id>
4358
- ${chalk7.dim("$")} mutagent traces get <trace-id> --json
4593
+ ${chalk11.dim("$")} mutagent traces get <trace-id>
4594
+ ${chalk11.dim("$")} mutagent traces get <trace-id> --json
4359
4595
 
4360
- ${chalk7.dim("Returns full trace details including spans, tokens, and latency.")}
4596
+ ${chalk11.dim("Returns full trace details including spans, tokens, and latency.")}
4361
4597
  `).action(async (id) => {
4362
4598
  const isJson = getJsonFlag(traces);
4363
4599
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4376,10 +4612,10 @@ ${chalk7.dim("Returns full trace details including spans, tokens, and latency.")
4376
4612
  });
4377
4613
  traces.command("analyze").description("Analyze traces for a prompt").argument("<prompt-id>", "Prompt ID").addHelpText("after", `
4378
4614
  Examples:
4379
- ${chalk7.dim("$")} mutagent traces analyze <prompt-id>
4380
- ${chalk7.dim("$")} mutagent traces analyze <prompt-id> --json
4615
+ ${chalk11.dim("$")} mutagent traces analyze <prompt-id>
4616
+ ${chalk11.dim("$")} mutagent traces analyze <prompt-id> --json
4381
4617
 
4382
- ${chalk7.dim("Aggregates trace data for a prompt: avg latency, token usage, error rates.")}
4618
+ ${chalk11.dim("Aggregates trace data for a prompt: avg latency, token usage, error rates.")}
4383
4619
  `).action(async (promptId) => {
4384
4620
  const isJson = getJsonFlag(traces);
4385
4621
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4397,12 +4633,12 @@ ${chalk7.dim("Aggregates trace data for a prompt: avg latency, token usage, erro
4397
4633
  });
4398
4634
  traces.command("export").description("Export traces").option("-p, --prompt <id>", "Filter by prompt ID").option("-f, --format <format>", "Export format (json, csv)", "json").option("-o, --output <path>", "Output file path").addHelpText("after", `
4399
4635
  Examples:
4400
- ${chalk7.dim("$")} mutagent traces export
4401
- ${chalk7.dim("$")} mutagent traces export --format json --output traces.json
4402
- ${chalk7.dim("$")} mutagent traces export --format csv --output traces.csv
4403
- ${chalk7.dim("$")} mutagent traces export --prompt <prompt-id> --format json
4636
+ ${chalk11.dim("$")} mutagent traces export
4637
+ ${chalk11.dim("$")} mutagent traces export --format json --output traces.json
4638
+ ${chalk11.dim("$")} mutagent traces export --format csv --output traces.csv
4639
+ ${chalk11.dim("$")} mutagent traces export --prompt <prompt-id> --format json
4404
4640
 
4405
- ${chalk7.dim("Exports to stdout by default. Use --output to save to a file.")}
4641
+ ${chalk11.dim("Exports to stdout by default. Use --output to save to a file.")}
4406
4642
  `).action(async (options) => {
4407
4643
  const isJson = getJsonFlag(traces);
4408
4644
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4445,22 +4681,22 @@ ${chalk7.dim("Exports to stdout by default. Use --output to save to a file.")}
4445
4681
 
4446
4682
  // src/commands/integrate.ts
4447
4683
  init_config();
4448
- import { Command as Command5 } from "commander";
4449
- import chalk8 from "chalk";
4450
- import { writeFileSync as writeFileSync3, existsSync as existsSync9 } from "fs";
4684
+ import { Command as Command8 } from "commander";
4685
+ import chalk12 from "chalk";
4686
+ import { writeFileSync as writeFileSync3, existsSync as existsSync10 } from "fs";
4451
4687
  import { execSync } from "child_process";
4452
4688
  init_errors();
4453
4689
 
4454
4690
  // src/lib/integrations/langchain.ts
4455
- import { readFileSync as readFileSync5, existsSync as existsSync5 } from "fs";
4691
+ import { readFileSync as readFileSync6, existsSync as existsSync6 } from "fs";
4456
4692
  var langchainIntegration = {
4457
4693
  name: "langchain",
4458
4694
  description: "LangChain framework",
4459
4695
  detect() {
4460
4696
  let hasLangchain = false;
4461
- if (existsSync5("package.json")) {
4697
+ if (existsSync6("package.json")) {
4462
4698
  try {
4463
- const pkg = JSON.parse(readFileSync5("package.json", "utf-8"));
4699
+ const pkg = JSON.parse(readFileSync6("package.json", "utf-8"));
4464
4700
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4465
4701
  hasLangchain = "langchain" in deps || "@langchain/core" in deps;
4466
4702
  } catch {}
@@ -4592,15 +4828,15 @@ mutagent traces analyze <prompt-id>
4592
4828
  };
4593
4829
 
4594
4830
  // src/lib/integrations/langgraph.ts
4595
- import { readFileSync as readFileSync6, existsSync as existsSync6 } from "fs";
4831
+ import { readFileSync as readFileSync7, existsSync as existsSync7 } from "fs";
4596
4832
  var langgraphIntegration = {
4597
4833
  name: "langgraph",
4598
4834
  description: "LangGraph agent workflow framework",
4599
4835
  detect() {
4600
4836
  let hasLanggraph = false;
4601
- if (existsSync6("package.json")) {
4837
+ if (existsSync7("package.json")) {
4602
4838
  try {
4603
- const pkg = JSON.parse(readFileSync6("package.json", "utf-8"));
4839
+ const pkg = JSON.parse(readFileSync7("package.json", "utf-8"));
4604
4840
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4605
4841
  hasLanggraph = "@langchain/langgraph" in deps;
4606
4842
  } catch {}
@@ -4721,15 +4957,15 @@ mutagent integrate langgraph --verify
4721
4957
  };
4722
4958
 
4723
4959
  // src/lib/integrations/vercel-ai.ts
4724
- import { readFileSync as readFileSync7, existsSync as existsSync7 } from "fs";
4960
+ import { readFileSync as readFileSync8, existsSync as existsSync8 } from "fs";
4725
4961
  var vercelAiIntegration = {
4726
4962
  name: "vercel-ai",
4727
4963
  description: "Vercel AI SDK",
4728
4964
  detect() {
4729
4965
  let hasAiSdk = false;
4730
- if (existsSync7("package.json")) {
4966
+ if (existsSync8("package.json")) {
4731
4967
  try {
4732
- const pkg = JSON.parse(readFileSync7("package.json", "utf-8"));
4968
+ const pkg = JSON.parse(readFileSync8("package.json", "utf-8"));
4733
4969
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4734
4970
  hasAiSdk = "ai" in deps;
4735
4971
  } catch {}
@@ -4867,15 +5103,15 @@ mutagent integrate vercel-ai --verify
4867
5103
  };
4868
5104
 
4869
5105
  // src/lib/integrations/openai.ts
4870
- import { readFileSync as readFileSync8, existsSync as existsSync8 } from "fs";
5106
+ import { readFileSync as readFileSync9, existsSync as existsSync9 } from "fs";
4871
5107
  var openaiIntegration = {
4872
5108
  name: "openai",
4873
5109
  description: "OpenAI SDK integration with automatic tracing",
4874
5110
  detect() {
4875
5111
  let hasOpenAI = false;
4876
- if (existsSync8("package.json")) {
5112
+ if (existsSync9("package.json")) {
4877
5113
  try {
4878
- const pkg = JSON.parse(readFileSync8("package.json", "utf-8"));
5114
+ const pkg = JSON.parse(readFileSync9("package.json", "utf-8"));
4879
5115
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4880
5116
  hasOpenAI = "openai" in deps;
4881
5117
  } catch {}
@@ -5029,16 +5265,16 @@ function getFrameworkMetadata(name) {
5029
5265
 
5030
5266
  // src/commands/integrate.ts
5031
5267
  function detectPackageManager() {
5032
- if (existsSync9("bun.lockb") || existsSync9("bun.lock")) {
5268
+ if (existsSync10("bun.lockb") || existsSync10("bun.lock")) {
5033
5269
  return "bun";
5034
5270
  }
5035
- if (existsSync9("pnpm-lock.yaml")) {
5271
+ if (existsSync10("pnpm-lock.yaml")) {
5036
5272
  return "pnpm";
5037
5273
  }
5038
- if (existsSync9("yarn.lock")) {
5274
+ if (existsSync10("yarn.lock")) {
5039
5275
  return "yarn";
5040
5276
  }
5041
- if (existsSync9("package-lock.json")) {
5277
+ if (existsSync10("package-lock.json")) {
5042
5278
  return "npm";
5043
5279
  }
5044
5280
  try {
@@ -5081,14 +5317,14 @@ AI AGENT: Please determine the framework by:
5081
5317
  `;
5082
5318
  }
5083
5319
  function createIntegrateCommand() {
5084
- const integrate = new Command5("integrate").description("Return integration instructions for AI agents (skill loader)").addHelpText("after", `
5320
+ const integrate = new Command8("integrate").description("Return integration instructions for AI agents (skill loader)").addHelpText("after", `
5085
5321
  Examples:
5086
- ${chalk8.dim("$")} mutagent integrate # Return exploration instructions for AI agents
5087
- ${chalk8.dim("$")} mutagent integrate langchain # Return @mutagent/langchain integration steps
5088
- ${chalk8.dim("$")} mutagent integrate openai # Return @mutagent/openai integration steps
5089
- ${chalk8.dim("$")} mutagent integrate vercel-ai --verify # Verify integration
5090
- ${chalk8.dim("$")} mutagent integrate langchain --output ./INTEGRATION.md
5091
- ${chalk8.dim("$")} mutagent integrate --list # List available frameworks
5322
+ ${chalk12.dim("$")} mutagent integrate # Return exploration instructions for AI agents
5323
+ ${chalk12.dim("$")} mutagent integrate langchain # Return @mutagent/langchain integration steps
5324
+ ${chalk12.dim("$")} mutagent integrate openai # Return @mutagent/openai integration steps
5325
+ ${chalk12.dim("$")} mutagent integrate vercel-ai --verify # Verify integration
5326
+ ${chalk12.dim("$")} mutagent integrate langchain --output ./INTEGRATION.md
5327
+ ${chalk12.dim("$")} mutagent integrate --list # List available frameworks
5092
5328
 
5093
5329
  Frameworks: langchain, langgraph, vercel-ai, openai
5094
5330
 
@@ -5202,35 +5438,20 @@ AI AGENT: Execute these steps using Bash for install, Write/Edit for code change
5202
5438
  return integrate;
5203
5439
  }
5204
5440
 
5205
- // src/commands/agents.ts
5441
+ // src/commands/agents/index.ts
5442
+ import { Command as Command9 } from "commander";
5443
+ import chalk14 from "chalk";
5444
+
5445
+ // src/commands/agents/agents-crud.ts
5206
5446
  init_sdk_client();
5207
- import { Command as Command6 } from "commander";
5208
- import chalk9 from "chalk";
5447
+ import chalk13 from "chalk";
5209
5448
  init_errors();
5210
- function createAgentsCommand() {
5211
- const agents = new Command6("agents").description("Manage AI agents").addHelpText("after", `
5212
- Examples:
5213
- ${chalk9.dim("$")} mutagent agents list
5214
- ${chalk9.dim("$")} mutagent agents get <agent-id>
5215
- ${chalk9.dim("$")} mutagent agents create --name "Code Reviewer" --slug code-reviewer --system-prompt "You are a code reviewer..."
5216
- ${chalk9.dim("$")} mutagent agents create -d '{"name":"Code Reviewer","slug":"code-reviewer","systemPrompt":"You are..."}'
5217
- ${chalk9.dim("$")} mutagent agents update <agent-id> --name "Updated Name"
5218
- ${chalk9.dim("$")} mutagent agents delete <agent-id> --force
5219
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id>
5220
- ${chalk9.dim("$")} mutagent agents conversations get <agent-id> <conversation-id>
5221
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id>
5222
- ${chalk9.dim("$")} mutagent agents conversations delete <agent-id> <conversation-id>
5223
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id>
5224
-
5225
- Subcommands:
5226
- list, get, create, update, delete
5227
- conversations list|get|create|delete|messages
5228
- `);
5449
+ function registerAgentsCrud(agents) {
5229
5450
  agents.command("list").description("List all agents").option("-l, --limit <n>", "Limit results", "50").option("-o, --offset <n>", "Offset for pagination").option("-n, --name <name>", "Filter by name").option("-s, --status <status>", "Filter by status (active, paused, archived)").addHelpText("after", `
5230
5451
  Examples:
5231
- ${chalk9.dim("$")} mutagent agents list
5232
- ${chalk9.dim("$")} mutagent agents list --status active
5233
- ${chalk9.dim("$")} mutagent agents list --name "reviewer" --json
5452
+ ${chalk13.dim("$")} mutagent agents list
5453
+ ${chalk13.dim("$")} mutagent agents list --status active
5454
+ ${chalk13.dim("$")} mutagent agents list --name "reviewer" --json
5234
5455
  `).action(async (options) => {
5235
5456
  const isJson = getJsonFlag(agents);
5236
5457
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5280,8 +5501,8 @@ Examples:
5280
5501
  });
5281
5502
  agents.command("get").description("Get agent details").argument("<id>", "Agent ID").addHelpText("after", `
5282
5503
  Examples:
5283
- ${chalk9.dim("$")} mutagent agents get <agent-id>
5284
- ${chalk9.dim("$")} mutagent agents get <agent-id> --json
5504
+ ${chalk13.dim("$")} mutagent agents get <agent-id>
5505
+ ${chalk13.dim("$")} mutagent agents get <agent-id> --json
5285
5506
  `).action(async (id) => {
5286
5507
  const isJson = getJsonFlag(agents);
5287
5508
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5311,11 +5532,11 @@ Examples:
5311
5532
  };
5312
5533
  output.output(formatted);
5313
5534
  if (agent.systemPrompt) {
5314
- console.log(chalk9.bold(`
5535
+ console.log(chalk13.bold(`
5315
5536
  System Prompt:`));
5316
- console.log(chalk9.gray("─".repeat(60)));
5537
+ console.log(chalk13.gray("─".repeat(60)));
5317
5538
  console.log(agent.systemPrompt);
5318
- console.log(chalk9.gray("─".repeat(60)));
5539
+ console.log(chalk13.gray("─".repeat(60)));
5319
5540
  }
5320
5541
  }
5321
5542
  } catch (error) {
@@ -5324,17 +5545,17 @@ System Prompt:`));
5324
5545
  });
5325
5546
  agents.command("create").description("Create a new agent").option("-d, --data <json>", "Agent as JSON string (recommended for CI/scripts/agents)").option("-n, --name <name>", "Agent name").option("-s, --slug <slug>", "Agent slug (URL-friendly identifier)").option("-p, --system-prompt <prompt>", "System prompt").option("-m, --model <model>", "Model (claude-sonnet-4-5, claude-opus-4-5, claude-haiku-4-5)").option("--description <desc>", "Agent description").addHelpText("after", `
5326
5547
  Examples:
5327
- ${chalk9.dim("$")} mutagent agents create --name "Code Reviewer" --slug code-reviewer --system-prompt "You are a code reviewer..."
5328
- ${chalk9.dim("$")} mutagent agents create -d '{"name":"Code Reviewer","slug":"code-reviewer","systemPrompt":"You are a code reviewer..."}'
5548
+ ${chalk13.dim("$")} mutagent agents create --name "Code Reviewer" --slug code-reviewer --system-prompt "You are a code reviewer..."
5549
+ ${chalk13.dim("$")} mutagent agents create -d '{"name":"Code Reviewer","slug":"code-reviewer","systemPrompt":"You are a code reviewer..."}'
5329
5550
 
5330
5551
  Expected JSON (--data):
5331
- ${chalk9.dim('{"name":"<name>","slug":"<slug>","systemPrompt":"<system prompt>","model":"<model-id>","description":"<description>"}')}
5552
+ ${chalk13.dim('{"name":"<name>","slug":"<slug>","systemPrompt":"<system prompt>","model":"<model-id>","description":"<description>"}')}
5332
5553
 
5333
5554
  Input Methods (pick one, priority order):
5334
- --name/--slug/... Individual flags ${chalk9.green("(recommended)")}
5555
+ --name/--slug/... Individual flags ${chalk13.green("(recommended)")}
5335
5556
  -d, --data Inline JSON object (CI/scripts/agents)
5336
5557
 
5337
- ${chalk9.red("Required: name, slug, systemPrompt.")} ${chalk9.dim("CLI flags override --data fields.")}
5558
+ ${chalk13.red("Required: name, slug, systemPrompt.")} ${chalk13.dim("CLI flags override --data fields.")}
5338
5559
  `).action(async (options) => {
5339
5560
  const isJson = getJsonFlag(agents);
5340
5561
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5379,15 +5600,15 @@ ${chalk9.red("Required: name, slug, systemPrompt.")} ${chalk9.dim("CLI flags ove
5379
5600
  });
5380
5601
  agents.command("update").description("Update an agent").argument("<id>", "Agent ID").option("-d, --data <json>", "Agent updates as JSON string (CI/scripts/agents)").option("-n, --name <name>", "New name").option("-p, --system-prompt <prompt>", "New system prompt").option("-m, --model <model>", "New model").option("--description <desc>", "New description").option("-s, --status <status>", "New status (active, paused, archived)").addHelpText("after", `
5381
5602
  Examples:
5382
- ${chalk9.dim("$")} mutagent agents update <id> --name "New Name"
5383
- ${chalk9.dim("$")} mutagent agents update <id> --system-prompt "Updated prompt" --status active
5384
- ${chalk9.dim("$")} mutagent agents update <id> -d '{"name":"New Name","systemPrompt":"Updated prompt"}'
5603
+ ${chalk13.dim("$")} mutagent agents update <id> --name "New Name"
5604
+ ${chalk13.dim("$")} mutagent agents update <id> --system-prompt "Updated prompt" --status active
5605
+ ${chalk13.dim("$")} mutagent agents update <id> -d '{"name":"New Name","systemPrompt":"Updated prompt"}'
5385
5606
 
5386
5607
  Input Methods (pick one, priority order):
5387
- --name/--system-prompt/... Individual flags ${chalk9.green("(recommended)")}
5608
+ --name/--system-prompt/... Individual flags ${chalk13.green("(recommended)")}
5388
5609
  -d, --data Inline JSON object (CI/scripts/agents)
5389
5610
 
5390
- ${chalk9.dim("CLI flags override --data fields.")}
5611
+ ${chalk13.dim("CLI flags override --data fields.")}
5391
5612
  `).action(async (id, options) => {
5392
5613
  const isJson = getJsonFlag(agents);
5393
5614
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5434,11 +5655,11 @@ ${chalk9.dim("CLI flags override --data fields.")}
5434
5655
  });
5435
5656
  agents.command("delete").description("Delete an agent").argument("<id>", "Agent ID").option("--force", "Skip confirmation").addHelpText("after", `
5436
5657
  Examples:
5437
- ${chalk9.dim("$")} mutagent agents delete <id>
5438
- ${chalk9.dim("$")} mutagent agents delete <id> --force
5439
- ${chalk9.dim("$")} mutagent agents delete <id> --force --json
5658
+ ${chalk13.dim("$")} mutagent agents delete <id>
5659
+ ${chalk13.dim("$")} mutagent agents delete <id> --force
5660
+ ${chalk13.dim("$")} mutagent agents delete <id> --force --json
5440
5661
 
5441
- ${chalk9.dim("Tip: Use --force to skip confirmation (required for non-interactive/CI usage).")}
5662
+ ${chalk13.dim("Tip: Use --force to skip confirmation (required for non-interactive/CI usage).")}
5442
5663
  `).action(async (id, options) => {
5443
5664
  const isJson = getJsonFlag(agents);
5444
5665
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5463,177 +5684,44 @@ ${chalk9.dim("Tip: Use --force to skip confirmation (required for non-interactiv
5463
5684
  handleError(error, isJson);
5464
5685
  }
5465
5686
  });
5466
- const conversations = new Command6("conversations").description("Manage conversations for agents").addHelpText("after", `
5467
- Examples:
5468
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id>
5469
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id>
5470
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id>
5471
- `).action(() => {
5472
- conversations.help();
5473
- });
5474
- agents.addCommand(conversations);
5475
- conversations.command("list").description("List conversations for an agent").argument("<agent-id>", "Agent ID").option("-l, --limit <n>", "Limit results", "50").option("-o, --offset <n>", "Offset for pagination").addHelpText("after", `
5476
- Examples:
5477
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id>
5478
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id> --limit 10 --json
5479
- `).action(async (agentId, options) => {
5480
- const isJson = getJsonFlag(agents);
5481
- const output = new OutputFormatter(isJson ? "json" : "table");
5482
- try {
5483
- const client = getSDKClient();
5484
- const filters = {
5485
- limit: parseInt(options.limit, 10) || 50
5486
- };
5487
- if (options.offset) {
5488
- filters.offset = parseInt(options.offset, 10);
5489
- }
5490
- const result = await client.listConversations(agentId, filters);
5491
- if (isJson) {
5492
- output.output(result);
5493
- } else {
5494
- const formatted = result.data.map((c) => ({
5495
- id: c.id,
5496
- title: c.title ?? "Untitled",
5497
- status: c.status,
5498
- messages: c.messageCount ?? "N/A",
5499
- created: c.createdAt ? new Date(c.createdAt).toLocaleDateString() : "N/A",
5500
- updated: c.updatedAt ? new Date(c.updatedAt).toLocaleDateString() : "N/A"
5501
- }));
5502
- output.output(formatted);
5503
- }
5504
- } catch (error) {
5505
- handleError(error, isJson);
5506
- }
5507
- });
5508
- conversations.command("get").description("Get conversation details").argument("<agent-id>", "Agent ID").argument("<conversation-id>", "Conversation ID").addHelpText("after", `
5509
- Examples:
5510
- ${chalk9.dim("$")} mutagent agents conversations get <agent-id> <conversation-id>
5511
- ${chalk9.dim("$")} mutagent agents conversations get <agent-id> <conversation-id> --json
5512
- `).action(async (agentId, conversationId) => {
5513
- const isJson = getJsonFlag(agents);
5514
- const output = new OutputFormatter(isJson ? "json" : "table");
5515
- try {
5516
- const client = getSDKClient();
5517
- const conversation = await client.getConversation(agentId, conversationId);
5518
- if (isJson) {
5519
- output.output(conversation);
5520
- } else {
5521
- const formatted = {
5522
- id: conversation.id,
5523
- agentId: conversation.agentId,
5524
- title: conversation.title ?? "Untitled",
5525
- status: conversation.status,
5526
- messageCount: conversation.messageCount ?? "N/A",
5527
- createdAt: conversation.createdAt ? new Date(conversation.createdAt).toLocaleString() : "N/A",
5528
- updatedAt: conversation.updatedAt ? new Date(conversation.updatedAt).toLocaleString() : "N/A"
5529
- };
5530
- output.output(formatted);
5531
- }
5532
- } catch (error) {
5533
- handleError(error, isJson);
5534
- }
5535
- });
5536
- conversations.command("create").description("Start a new conversation with an agent").argument("<agent-id>", "Agent ID").option("-t, --title <title>", "Conversation title").addHelpText("after", `
5537
- Examples:
5538
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id>
5539
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id> --title "Debug session" --json
5540
- `).action(async (agentId, options) => {
5541
- const isJson = getJsonFlag(agents);
5542
- const output = new OutputFormatter(isJson ? "json" : "table");
5543
- try {
5544
- const client = getSDKClient();
5545
- const data = {};
5546
- if (options.title) {
5547
- data.title = options.title;
5548
- }
5549
- const conversation = await client.createConversation(agentId, data);
5550
- output.success(`Created conversation: ${conversation.id}`);
5551
- output.output(conversation);
5552
- } catch (error) {
5553
- handleError(error, isJson);
5554
- }
5555
- });
5556
- conversations.command("delete").description("Delete a conversation").argument("<agent-id>", "Agent ID").argument("<conversation-id>", "Conversation ID").option("--force", "Skip confirmation").addHelpText("after", `
5557
- Examples:
5558
- ${chalk9.dim("$")} mutagent agents conversations delete <agent-id> <conversation-id>
5559
- ${chalk9.dim("$")} mutagent agents conversations delete <agent-id> <conversation-id> --force --json
5560
- `).action(async (agentId, conversationId, options) => {
5561
- const isJson = getJsonFlag(agents);
5562
- const output = new OutputFormatter(isJson ? "json" : "table");
5563
- try {
5564
- if (!options.force && !isJson) {
5565
- const inquirer3 = (await import("inquirer")).default;
5566
- const answers = await inquirer3.prompt([{
5567
- type: "confirm",
5568
- name: "confirm",
5569
- message: `Delete conversation ${conversationId}? This action cannot be undone.`,
5570
- default: false
5571
- }]);
5572
- if (!answers.confirm) {
5573
- output.info("Cancelled");
5574
- return;
5575
- }
5576
- }
5577
- const client = getSDKClient();
5578
- await client.deleteConversation(agentId, conversationId);
5579
- output.success(`Deleted conversation: ${conversationId}`);
5580
- } catch (error) {
5581
- handleError(error, isJson);
5582
- }
5583
- });
5584
- conversations.command("messages").description("List messages in a conversation").argument("<agent-id>", "Agent ID").argument("<conversation-id>", "Conversation ID").option("-l, --limit <n>", "Limit results", "50").option("-o, --offset <n>", "Offset for pagination").addHelpText("after", `
5687
+ }
5688
+
5689
+ // src/commands/agents/index.ts
5690
+ function createAgentsCommand() {
5691
+ const agents = new Command9("agents").description("Manage AI agents").addHelpText("after", `
5585
5692
  Examples:
5586
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id>
5587
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id> --limit 20 --json
5588
- `).action(async (agentId, conversationId, options) => {
5589
- const isJson = getJsonFlag(agents);
5590
- const output = new OutputFormatter(isJson ? "json" : "table");
5591
- try {
5592
- const client = getSDKClient();
5593
- const filters = {
5594
- limit: parseInt(options.limit, 10) || 50
5595
- };
5596
- if (options.offset) {
5597
- filters.offset = parseInt(options.offset, 10);
5598
- }
5599
- const result = await client.listMessages(agentId, conversationId, filters);
5600
- if (isJson) {
5601
- output.output(result);
5602
- } else {
5603
- const formatted = result.data.map((m) => ({
5604
- id: m.id,
5605
- role: m.role,
5606
- content: m.content.length > 80 ? m.content.substring(0, 77) + "..." : m.content,
5607
- tokens: m.tokenCount ?? "N/A",
5608
- created: m.createdAt ? new Date(m.createdAt).toLocaleString() : "N/A"
5609
- }));
5610
- output.output(formatted);
5611
- }
5612
- } catch (error) {
5613
- handleError(error, isJson);
5614
- }
5615
- });
5693
+ ${chalk14.dim("$")} mutagent agents list
5694
+ ${chalk14.dim("$")} mutagent agents get <agent-id>
5695
+ ${chalk14.dim("$")} mutagent agents create --name "Code Reviewer" --slug code-reviewer --system-prompt "You are a code reviewer..."
5696
+ ${chalk14.dim("$")} mutagent agents create -d '{"name":"Code Reviewer","slug":"code-reviewer","systemPrompt":"You are..."}'
5697
+ ${chalk14.dim("$")} mutagent agents update <agent-id> --name "Updated Name"
5698
+ ${chalk14.dim("$")} mutagent agents delete <agent-id> --force
5699
+
5700
+ Subcommands:
5701
+ list, get, create, update, delete
5702
+ `);
5703
+ registerAgentsCrud(agents);
5616
5704
  return agents;
5617
5705
  }
5618
5706
 
5619
5707
  // src/commands/config.ts
5620
5708
  init_config();
5621
- import { Command as Command7 } from "commander";
5622
- import chalk10 from "chalk";
5709
+ import { Command as Command10 } from "commander";
5710
+ import chalk15 from "chalk";
5623
5711
  init_errors();
5624
5712
  var VALID_CONFIG_KEYS = ["apiKey", "endpoint", "format", "timeout", "defaultWorkspace", "defaultOrganization"];
5625
5713
  function createConfigCommand() {
5626
- const config = new Command7("config").description("Manage CLI configuration").addHelpText("after", `
5714
+ const config = new Command10("config").description("Manage CLI configuration").addHelpText("after", `
5627
5715
  Examples:
5628
- ${chalk10.dim("$")} mutagent config list
5629
- ${chalk10.dim("$")} mutagent config get endpoint
5630
- ${chalk10.dim("$")} mutagent config set workspace <workspace-id>
5631
- ${chalk10.dim("$")} mutagent config set org <org-id>
5716
+ ${chalk15.dim("$")} mutagent config list
5717
+ ${chalk15.dim("$")} mutagent config get endpoint
5718
+ ${chalk15.dim("$")} mutagent config set workspace <workspace-id>
5719
+ ${chalk15.dim("$")} mutagent config set org <org-id>
5632
5720
  `);
5633
5721
  config.command("list").description("List all configuration").addHelpText("after", `
5634
5722
  Examples:
5635
- ${chalk10.dim("$")} mutagent config list
5636
- ${chalk10.dim("$")} mutagent config list --json
5723
+ ${chalk15.dim("$")} mutagent config list
5724
+ ${chalk15.dim("$")} mutagent config list --json
5637
5725
  `).action(() => {
5638
5726
  const isJson = getJsonFlag(config);
5639
5727
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5646,11 +5734,11 @@ Examples:
5646
5734
  });
5647
5735
  config.command("get").description("Get configuration value").argument("<key>", "Configuration key (apiKey, endpoint, format, timeout, defaultWorkspace, defaultOrganization)").addHelpText("after", `
5648
5736
  Examples:
5649
- ${chalk10.dim("$")} mutagent config get endpoint
5650
- ${chalk10.dim("$")} mutagent config get defaultWorkspace
5651
- ${chalk10.dim("$")} mutagent config get apiKey --json
5737
+ ${chalk15.dim("$")} mutagent config get endpoint
5738
+ ${chalk15.dim("$")} mutagent config get defaultWorkspace
5739
+ ${chalk15.dim("$")} mutagent config get apiKey --json
5652
5740
 
5653
- ${chalk10.dim("Keys: apiKey, endpoint, format, timeout, defaultWorkspace, defaultOrganization")}
5741
+ ${chalk15.dim("Keys: apiKey, endpoint, format, timeout, defaultWorkspace, defaultOrganization")}
5654
5742
  `).action((key) => {
5655
5743
  const isJson = getJsonFlag(config);
5656
5744
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5672,14 +5760,14 @@ ${chalk10.dim("Keys: apiKey, endpoint, format, timeout, defaultWorkspace, defaul
5672
5760
  handleError(error, isJson);
5673
5761
  }
5674
5762
  });
5675
- const set = new Command7("set").description("Set configuration value").action(() => {
5763
+ const set = new Command10("set").description("Set configuration value").action(() => {
5676
5764
  set.help();
5677
5765
  });
5678
5766
  set.command("workspace").description("Set default workspace ID").argument("<id>", "Workspace ID to set as default").addHelpText("after", `
5679
5767
  Examples:
5680
- ${chalk10.dim("$")} mutagent config set workspace <workspace-id>
5768
+ ${chalk15.dim("$")} mutagent config set workspace <workspace-id>
5681
5769
 
5682
- ${chalk10.dim("Persists workspace ID so you don't need to pass headers on every request.")}
5770
+ ${chalk15.dim("Persists workspace ID so you don't need to pass headers on every request.")}
5683
5771
  `).action((id) => {
5684
5772
  const isJson = getJsonFlag(config);
5685
5773
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5692,9 +5780,9 @@ ${chalk10.dim("Persists workspace ID so you don't need to pass headers on every
5692
5780
  });
5693
5781
  set.command("org").description("Set default organization ID").argument("<id>", "Organization ID to set as default").addHelpText("after", `
5694
5782
  Examples:
5695
- ${chalk10.dim("$")} mutagent config set org <org-id>
5783
+ ${chalk15.dim("$")} mutagent config set org <org-id>
5696
5784
 
5697
- ${chalk10.dim("Persists organization ID for org-scoped API keys.")}
5785
+ ${chalk15.dim("Persists organization ID for org-scoped API keys.")}
5698
5786
  `).action((id) => {
5699
5787
  const isJson = getJsonFlag(config);
5700
5788
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5711,8 +5799,8 @@ ${chalk10.dim("Persists organization ID for org-scoped API keys.")}
5711
5799
 
5712
5800
  // src/commands/playground.ts
5713
5801
  init_sdk_client();
5714
- import { Command as Command8 } from "commander";
5715
- import chalk11 from "chalk";
5802
+ import { Command as Command11 } from "commander";
5803
+ import chalk16 from "chalk";
5716
5804
  init_errors();
5717
5805
  function parseSSELine(line) {
5718
5806
  if (!line || line.startsWith(":")) {
@@ -5737,13 +5825,13 @@ function parsePromptStreamEvent(data) {
5737
5825
  }
5738
5826
  }
5739
5827
  function createPlaygroundCommand() {
5740
- const playground = new Command8("playground").description("Execute and test prompts interactively").addHelpText("after", `
5828
+ const playground = new Command11("playground").description("Execute and test prompts interactively").addHelpText("after", `
5741
5829
  Examples:
5742
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{"name": "John"}'
5743
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{}' --stream
5744
- ${chalk11.dim("$")} mutagent playground run <prompt-id> -i '{}' --model gpt-4-turbo
5745
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --system "You are helpful" --human "Hello"
5746
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --messages '[{"role":"user","content":"Hi"}]'
5830
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --input '{"name": "John"}'
5831
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --input '{}' --stream
5832
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> -i '{}' --model gpt-4-turbo
5833
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --system "You are helpful" --human "Hello"
5834
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --messages '[{"role":"user","content":"Hi"}]'
5747
5835
 
5748
5836
  Input Format:
5749
5837
  The input must be a valid JSON object matching the prompt's input schema.
@@ -5759,16 +5847,16 @@ Streaming:
5759
5847
  `);
5760
5848
  playground.command("run").description("Execute a prompt with input variables").argument("<prompt-id>", "Prompt ID to execute (from: mutagent prompts list)").option("-i, --input <json>", "Input variables as JSON").option("-s, --stream", "Stream the response").option("-m, --model <model>", "Override model").option("--system <text>", "Set system prompt text").option("--human <text>", "Set human/user message text").option("--messages <json>", "Pass full messages array as JSON string").addHelpText("after", `
5761
5849
  Examples:
5762
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{"name": "John"}'
5763
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{}' --stream
5764
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --system "You are helpful" --human "Hello"
5765
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{}' --model gpt-4-turbo --json
5850
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --input '{"name": "John"}'
5851
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --input '{}' --stream
5852
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --system "You are helpful" --human "Hello"
5853
+ ${chalk16.dim("$")} mutagent playground run <prompt-id> --input '{}' --model gpt-4-turbo --json
5766
5854
 
5767
5855
  Input Methods (pick one, priority order):
5768
- --system/--human Quick system + user message ${chalk11.green("(recommended)")}
5856
+ --system/--human Quick system + user message ${chalk16.green("(recommended)")}
5769
5857
  --input '{"key":"value"}' Inline JSON variables
5770
5858
  --messages '[...]' Full messages array
5771
- ${chalk11.dim(`Hint: Test before evaluating: mutagent playground run <id> --input '{"key":"value"}'`)}
5859
+ ${chalk16.dim(`Hint: Test before evaluating: mutagent playground run <id> --input '{"key":"value"}'`)}
5772
5860
  `).action(async (promptId, options) => {
5773
5861
  const isJson = getJsonFlag(playground);
5774
5862
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5790,21 +5878,21 @@ ${chalk11.dim(`Hint: Test before evaluating: mutagent playground run <id> --inpu
5790
5878
  }
5791
5879
  });
5792
5880
  } else {
5793
- console.log(chalk11.bold(`
5881
+ console.log(chalk16.bold(`
5794
5882
  Execution Result:`));
5795
- console.log(chalk11.gray("─".repeat(50)));
5796
- console.log(chalk11.cyan("Output:"));
5883
+ console.log(chalk16.gray("─".repeat(50)));
5884
+ console.log(chalk16.cyan("Output:"));
5797
5885
  console.log(result.output);
5798
- console.log(chalk11.gray("─".repeat(50)));
5799
- console.log(chalk11.dim(`Model: ${result.model}`));
5800
- console.log(chalk11.dim(`Execution Time: ${String(result.executionTimeMs)}ms`));
5886
+ console.log(chalk16.gray("─".repeat(50)));
5887
+ console.log(chalk16.dim(`Model: ${result.model}`));
5888
+ console.log(chalk16.dim(`Execution Time: ${String(result.executionTimeMs)}ms`));
5801
5889
  if (result.tokens) {
5802
- console.log(chalk11.dim(`Tokens: ${String(result.tokens.prompt)} prompt + ${String(result.tokens.completion)} completion = ${String(result.tokens.total)} total`));
5890
+ console.log(chalk16.dim(`Tokens: ${String(result.tokens.prompt)} prompt + ${String(result.tokens.completion)} completion = ${String(result.tokens.total)} total`));
5803
5891
  }
5804
5892
  if (result.cost !== undefined) {
5805
- console.log(chalk11.dim(`Cost: $${result.cost.toFixed(6)}`));
5893
+ console.log(chalk16.dim(`Cost: $${result.cost.toFixed(6)}`));
5806
5894
  }
5807
- console.log(chalk11.dim(`Playground: ${playgroundLink()}`));
5895
+ console.log(chalk16.dim(`Playground: ${playgroundLink()}`));
5808
5896
  console.log();
5809
5897
  }
5810
5898
  }
@@ -5893,9 +5981,9 @@ async function executeStreaming(client, promptId, input, model, isJson, output)
5893
5981
  const decoder = new TextDecoder;
5894
5982
  let buffer = "";
5895
5983
  if (!isJson) {
5896
- console.log(chalk11.bold(`
5984
+ console.log(chalk16.bold(`
5897
5985
  Streaming Output:`));
5898
- console.log(chalk11.gray("─".repeat(50)));
5986
+ console.log(chalk16.gray("─".repeat(50)));
5899
5987
  }
5900
5988
  try {
5901
5989
  for (;; ) {
@@ -5934,15 +6022,15 @@ Streaming Output:`));
5934
6022
  console.log(JSON.stringify({ type: "complete", result: event.result }));
5935
6023
  } else {
5936
6024
  console.log();
5937
- console.log(chalk11.gray("─".repeat(50)));
6025
+ console.log(chalk16.gray("─".repeat(50)));
5938
6026
  if (event.result) {
5939
- console.log(chalk11.dim(`Model: ${event.result.model}`));
5940
- console.log(chalk11.dim(`Execution Time: ${String(event.result.executionTimeMs)}ms`));
6027
+ console.log(chalk16.dim(`Model: ${event.result.model}`));
6028
+ console.log(chalk16.dim(`Execution Time: ${String(event.result.executionTimeMs)}ms`));
5941
6029
  if (event.result.tokens) {
5942
- console.log(chalk11.dim(`Tokens: ${String(event.result.tokens.prompt)} prompt + ${String(event.result.tokens.completion)} completion = ${String(event.result.tokens.total)} total`));
6030
+ console.log(chalk16.dim(`Tokens: ${String(event.result.tokens.prompt)} prompt + ${String(event.result.tokens.completion)} completion = ${String(event.result.tokens.total)} total`));
5943
6031
  }
5944
6032
  if (event.result.cost !== undefined) {
5945
- console.log(chalk11.dim(`Cost: $${event.result.cost.toFixed(6)}`));
6033
+ console.log(chalk16.dim(`Cost: $${event.result.cost.toFixed(6)}`));
5946
6034
  }
5947
6035
  }
5948
6036
  console.log();
@@ -5966,14 +6054,14 @@ Streaming Output:`));
5966
6054
 
5967
6055
  // src/commands/workspaces.ts
5968
6056
  init_sdk_client();
5969
- import { Command as Command9 } from "commander";
5970
- import chalk12 from "chalk";
6057
+ import { Command as Command12 } from "commander";
6058
+ import chalk17 from "chalk";
5971
6059
  init_errors();
5972
6060
  function createWorkspacesCommand() {
5973
- const workspaces = new Command9("workspaces").description("View workspaces (read-only)").addHelpText("after", `
6061
+ const workspaces = new Command12("workspaces").description("View workspaces (read-only)").addHelpText("after", `
5974
6062
  Examples:
5975
- ${chalk12.dim("$")} mutagent workspaces list
5976
- ${chalk12.dim("$")} mutagent workspaces get <workspace-id>
6063
+ ${chalk17.dim("$")} mutagent workspaces list
6064
+ ${chalk17.dim("$")} mutagent workspaces get <workspace-id>
5977
6065
 
5978
6066
  Subcommands:
5979
6067
  list, get
@@ -5982,8 +6070,8 @@ Note: Workspace management (create, update, delete) is available in the Admin Pa
5982
6070
  `);
5983
6071
  workspaces.command("list").description("List all workspaces").option("-l, --limit <n>", "Limit results", "50").option("-o, --offset <n>", "Offset for pagination").addHelpText("after", `
5984
6072
  Examples:
5985
- ${chalk12.dim("$")} mutagent workspaces list
5986
- ${chalk12.dim("$")} mutagent workspaces list --limit 10 --json
6073
+ ${chalk17.dim("$")} mutagent workspaces list
6074
+ ${chalk17.dim("$")} mutagent workspaces list --limit 10 --json
5987
6075
  `).action(async (options) => {
5988
6076
  const isJson = getJsonFlag(workspaces);
5989
6077
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6023,8 +6111,8 @@ Examples:
6023
6111
  });
6024
6112
  workspaces.command("get").description("Get workspace details").argument("<id>", "Workspace ID").addHelpText("after", `
6025
6113
  Examples:
6026
- ${chalk12.dim("$")} mutagent workspaces get <workspace-id>
6027
- ${chalk12.dim("$")} mutagent workspaces get <workspace-id> --json
6114
+ ${chalk17.dim("$")} mutagent workspaces get <workspace-id>
6115
+ ${chalk17.dim("$")} mutagent workspaces get <workspace-id> --json
6028
6116
  `).action(async (id) => {
6029
6117
  const isJson = getJsonFlag(workspaces);
6030
6118
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6056,8 +6144,8 @@ Examples:
6056
6144
 
6057
6145
  // src/commands/providers.ts
6058
6146
  init_sdk_client();
6059
- import { Command as Command10 } from "commander";
6060
- import chalk13 from "chalk";
6147
+ import { Command as Command13 } from "commander";
6148
+ import chalk18 from "chalk";
6061
6149
  init_errors();
6062
6150
  var VALID_PROVIDER_TYPES = [
6063
6151
  "openai",
@@ -6079,11 +6167,11 @@ function validateProviderType(type) {
6079
6167
  return type;
6080
6168
  }
6081
6169
  function createProvidersCommand() {
6082
- const providers = new Command10("providers").description("View LLM providers (read-only)").addHelpText("after", `
6170
+ const providers = new Command13("providers").description("View LLM providers (read-only)").addHelpText("after", `
6083
6171
  Examples:
6084
- ${chalk13.dim("$")} mutagent providers list
6085
- ${chalk13.dim("$")} mutagent providers get <provider-id>
6086
- ${chalk13.dim("$")} mutagent providers test <provider-id>
6172
+ ${chalk18.dim("$")} mutagent providers list
6173
+ ${chalk18.dim("$")} mutagent providers get <provider-id>
6174
+ ${chalk18.dim("$")} mutagent providers test <provider-id>
6087
6175
 
6088
6176
  Provider Types:
6089
6177
  openai, anthropic, google, azure, bedrock, cohere, mistral, groq, together, replicate, custom
@@ -6093,15 +6181,15 @@ Subcommands:
6093
6181
 
6094
6182
  Note: Provider management (create, update, delete) is available in the Admin Panel only.
6095
6183
 
6096
- ${chalk13.yellow("Note:")} The providers module is not yet active. This is a placeholder
6184
+ ${chalk18.yellow("Note:")} The providers module is not yet active. This is a placeholder
6097
6185
  for future external provider configuration. The server currently uses
6098
6186
  built-in provider settings.
6099
6187
  `);
6100
6188
  providers.command("list").description("List all providers").option("-l, --limit <n>", "Limit results", "50").option("-o, --offset <n>", "Offset for pagination").option("-t, --type <type>", "Filter by provider type").addHelpText("after", `
6101
6189
  Examples:
6102
- ${chalk13.dim("$")} mutagent providers list
6103
- ${chalk13.dim("$")} mutagent providers list --type openai
6104
- ${chalk13.dim("$")} mutagent providers list --json
6190
+ ${chalk18.dim("$")} mutagent providers list
6191
+ ${chalk18.dim("$")} mutagent providers list --type openai
6192
+ ${chalk18.dim("$")} mutagent providers list --json
6105
6193
  `).action(async (options) => {
6106
6194
  const isJson = getJsonFlag(providers);
6107
6195
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6128,7 +6216,7 @@ Examples:
6128
6216
  }));
6129
6217
  output.output({ ...result, data: withLinks });
6130
6218
  } else {
6131
- console.log(chalk13.yellow("Note: The providers module is not yet active. This is a placeholder for future external provider configuration."));
6219
+ console.log(chalk18.yellow("Note: The providers module is not yet active. This is a placeholder for future external provider configuration."));
6132
6220
  console.log("");
6133
6221
  if (result.data.length === 0) {
6134
6222
  output.info("No providers configured.");
@@ -6152,8 +6240,8 @@ Examples:
6152
6240
  });
6153
6241
  providers.command("get").description("Get provider details").argument("<id>", "Provider ID (from: mutagent providers list)").addHelpText("after", `
6154
6242
  Examples:
6155
- ${chalk13.dim("$")} mutagent providers get <provider-id>
6156
- ${chalk13.dim("$")} mutagent providers get <provider-id> --json
6243
+ ${chalk18.dim("$")} mutagent providers get <provider-id>
6244
+ ${chalk18.dim("$")} mutagent providers get <provider-id> --json
6157
6245
  `).action(async (id) => {
6158
6246
  const isJson = getJsonFlag(providers);
6159
6247
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6163,7 +6251,7 @@ Examples:
6163
6251
  if (isJson) {
6164
6252
  output.output({ ...provider, _links: providerLinks(provider.id) });
6165
6253
  } else {
6166
- console.log(chalk13.yellow("Note: The providers module is not yet active. This is a placeholder for future external provider configuration."));
6254
+ console.log(chalk18.yellow("Note: The providers module is not yet active. This is a placeholder for future external provider configuration."));
6167
6255
  console.log("");
6168
6256
  const formatted = {
6169
6257
  id: provider.id,
@@ -6184,17 +6272,17 @@ Examples:
6184
6272
  });
6185
6273
  providers.command("test").description("Test provider connectivity").argument("<id>", "Provider ID (from: mutagent providers list)").addHelpText("after", `
6186
6274
  Examples:
6187
- ${chalk13.dim("$")} mutagent providers test <provider-id>
6188
- ${chalk13.dim("$")} mutagent providers test <provider-id> --json
6275
+ ${chalk18.dim("$")} mutagent providers test <provider-id>
6276
+ ${chalk18.dim("$")} mutagent providers test <provider-id> --json
6189
6277
 
6190
- ${chalk13.dim("Tests connectivity and lists available models for the provider.")}
6278
+ ${chalk18.dim("Tests connectivity and lists available models for the provider.")}
6191
6279
  `).action(async (id) => {
6192
6280
  const isJson = getJsonFlag(providers);
6193
6281
  const output = new OutputFormatter(isJson ? "json" : "table");
6194
6282
  try {
6195
6283
  const client = getSDKClient();
6196
6284
  if (!isJson) {
6197
- console.log(chalk13.yellow("Note: The providers module is not yet active. This is a placeholder for future external provider configuration."));
6285
+ console.log(chalk18.yellow("Note: The providers module is not yet active. This is a placeholder for future external provider configuration."));
6198
6286
  console.log("");
6199
6287
  output.info(`Testing provider ${id}...`);
6200
6288
  }
@@ -6204,9 +6292,9 @@ ${chalk13.dim("Tests connectivity and lists available models for the provider.")
6204
6292
  } else {
6205
6293
  if (result.success) {
6206
6294
  output.success(`Provider test passed (${String(result.responseTimeMs)}ms)`);
6207
- console.log(chalk13.green(`Message: ${result.message}`));
6295
+ console.log(chalk18.green(`Message: ${result.message}`));
6208
6296
  if (result.availableModels && result.availableModels.length > 0) {
6209
- console.log(chalk13.bold(`
6297
+ console.log(chalk18.bold(`
6210
6298
  Available Models:`));
6211
6299
  result.availableModels.forEach((model) => {
6212
6300
  console.log(` - ${model}`);
@@ -6215,7 +6303,7 @@ Available Models:`));
6215
6303
  } else {
6216
6304
  output.error(`Provider test failed: ${result.message}`);
6217
6305
  if (result.error) {
6218
- console.log(chalk13.red(`Error: ${result.error}`));
6306
+ console.log(chalk18.red(`Error: ${result.error}`));
6219
6307
  }
6220
6308
  }
6221
6309
  }
@@ -6228,13 +6316,18 @@ Available Models:`));
6228
6316
 
6229
6317
  // src/commands/init.ts
6230
6318
  init_config();
6231
- import { Command as Command11 } from "commander";
6319
+ import { Command as Command14 } from "commander";
6232
6320
  import inquirer3 from "inquirer";
6233
- import chalk14 from "chalk";
6234
- import { existsSync as existsSync10, mkdirSync as mkdirSync3, readFileSync as readFileSync9, writeFileSync as writeFileSync4 } from "fs";
6321
+ import chalk19 from "chalk";
6322
+ import { existsSync as existsSync12, mkdirSync as mkdirSync3, writeFileSync as writeFileSync4 } from "fs";
6323
+ import { execSync as execSync3 } from "child_process";
6324
+ import { join as join6 } from "path";
6325
+ init_errors();
6326
+
6327
+ // src/lib/framework-detection.ts
6328
+ import { existsSync as existsSync11, readFileSync as readFileSync10 } from "fs";
6235
6329
  import { execSync as execSync2 } from "child_process";
6236
6330
  import { join as join5 } from "path";
6237
- init_errors();
6238
6331
  var FRAMEWORK_DETECTION_MAP = {
6239
6332
  "@mastra/core": {
6240
6333
  name: "mastra",
@@ -6284,16 +6377,16 @@ var FRAMEWORK_DETECTION_MAP = {
6284
6377
  }
6285
6378
  };
6286
6379
  function detectPackageManager2(cwd = process.cwd()) {
6287
- if (existsSync10(join5(cwd, "bun.lockb")) || existsSync10(join5(cwd, "bun.lock"))) {
6380
+ if (existsSync11(join5(cwd, "bun.lockb")) || existsSync11(join5(cwd, "bun.lock"))) {
6288
6381
  return "bun";
6289
6382
  }
6290
- if (existsSync10(join5(cwd, "pnpm-lock.yaml"))) {
6383
+ if (existsSync11(join5(cwd, "pnpm-lock.yaml"))) {
6291
6384
  return "pnpm";
6292
6385
  }
6293
- if (existsSync10(join5(cwd, "yarn.lock"))) {
6386
+ if (existsSync11(join5(cwd, "yarn.lock"))) {
6294
6387
  return "yarn";
6295
6388
  }
6296
- if (existsSync10(join5(cwd, "package-lock.json"))) {
6389
+ if (existsSync11(join5(cwd, "package-lock.json"))) {
6297
6390
  return "npm";
6298
6391
  }
6299
6392
  try {
@@ -6315,12 +6408,12 @@ function getInstallCommand2(pm, packages) {
6315
6408
  }
6316
6409
  function detectFrameworkFromPackageJson(cwd = process.cwd()) {
6317
6410
  const pkgPath = join5(cwd, "package.json");
6318
- if (!existsSync10(pkgPath)) {
6411
+ if (!existsSync11(pkgPath)) {
6319
6412
  return null;
6320
6413
  }
6321
6414
  let pkg;
6322
6415
  try {
6323
- pkg = JSON.parse(readFileSync9(pkgPath, "utf-8"));
6416
+ pkg = JSON.parse(readFileSync10(pkgPath, "utf-8"));
6324
6417
  } catch {
6325
6418
  return null;
6326
6419
  }
@@ -6336,23 +6429,25 @@ function detectFrameworkFromPackageJson(cwd = process.cwd()) {
6336
6429
  return null;
6337
6430
  }
6338
6431
  function hasRcConfig(cwd = process.cwd()) {
6339
- return existsSync10(join5(cwd, ".mutagentrc.json"));
6432
+ return existsSync11(join5(cwd, ".mutagentrc.json"));
6340
6433
  }
6434
+
6435
+ // src/commands/init.ts
6341
6436
  function writeRcConfig(config, cwd = process.cwd()) {
6342
- const rcPath = join5(cwd, ".mutagentrc.json");
6437
+ const rcPath = join6(cwd, ".mutagentrc.json");
6343
6438
  writeFileSync4(rcPath, JSON.stringify(config, null, 2) + `
6344
6439
  `);
6345
6440
  }
6346
6441
  function createInitCommand() {
6347
- const init = new Command11("init").description("Initialize MutagenT in your project").option("--non-interactive", "Skip interactive prompts (defaults to CLI-only mode)").addHelpText("after", `
6442
+ const init = new Command14("init").description("Initialize MutagenT in your project").option("--non-interactive", "Skip interactive prompts (defaults to CLI-only mode)").addHelpText("after", `
6348
6443
  Examples:
6349
- ${chalk14.dim("$")} mutagent init # Interactive setup wizard
6350
- ${chalk14.dim("$")} mutagent init --non-interactive # CLI-only mode (no prompts)
6444
+ ${chalk19.dim("$")} mutagent init # Interactive setup wizard
6445
+ ${chalk19.dim("$")} mutagent init --non-interactive # CLI-only mode (no prompts)
6351
6446
 
6352
6447
  Modes:
6353
- ${chalk14.bold("Full scaffold")} Install SDK + integration package, create config, setup tracing
6354
- ${chalk14.bold("CLI-only")} Verify auth + create .mutagentrc.json with workspace/endpoint
6355
- ${chalk14.bold("Skip")} Exit without changes
6448
+ ${chalk19.bold("Full scaffold")} Install SDK + integration package, create config, setup tracing
6449
+ ${chalk19.bold("CLI-only")} Verify auth + create .mutagentrc.json with workspace/endpoint
6450
+ ${chalk19.bold("Skip")} Exit without changes
6356
6451
  `).action(async (options) => {
6357
6452
  const isJson = getJsonFlag(init);
6358
6453
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6510,7 +6605,7 @@ Modes:
6510
6605
  output.info(`Installing: ${packages.join(", ")}`);
6511
6606
  output.info(`Running: ${installCmd}`);
6512
6607
  try {
6513
- execSync2(installCmd, { cwd, stdio: "inherit" });
6608
+ execSync3(installCmd, { cwd, stdio: "inherit" });
6514
6609
  output.success("Packages installed successfully.");
6515
6610
  } catch {
6516
6611
  output.error("Package installation failed. You can install manually:");
@@ -6553,8 +6648,8 @@ Modes:
6553
6648
  }
6554
6649
  }
6555
6650
  if (!isNonInteractive) {
6556
- const skillPath = join5(cwd, ".claude/skills/mutagent-cli/SKILL.md");
6557
- if (!existsSync10(skillPath)) {
6651
+ const skillPath = join6(cwd, ".claude/skills/mutagent-cli/SKILL.md");
6652
+ if (!existsSync12(skillPath)) {
6558
6653
  const { installSkill } = await inquirer3.prompt([{
6559
6654
  type: "confirm",
6560
6655
  name: "installSkill",
@@ -6563,11 +6658,11 @@ Modes:
6563
6658
  }]);
6564
6659
  if (installSkill) {
6565
6660
  try {
6566
- const skillDir = join5(cwd, ".claude/skills/mutagent-cli");
6567
- if (!existsSync10(skillDir)) {
6661
+ const skillDir = join6(cwd, ".claude/skills/mutagent-cli");
6662
+ if (!existsSync12(skillDir)) {
6568
6663
  mkdirSync3(skillDir, { recursive: true });
6569
6664
  }
6570
- execSync2("node " + join5(cwd, "node_modules/.bin/mutagent") + " skills install", {
6665
+ execSync3("node " + join6(cwd, "node_modules/.bin/mutagent") + " skills install", {
6571
6666
  cwd,
6572
6667
  stdio: "ignore"
6573
6668
  });
@@ -6601,24 +6696,24 @@ Modes:
6601
6696
  }
6602
6697
 
6603
6698
  // src/commands/explore.ts
6604
- import { Command as Command12 } from "commander";
6605
- import chalk15 from "chalk";
6699
+ import { Command as Command15 } from "commander";
6700
+ import chalk20 from "chalk";
6606
6701
  import { resolve as resolve3 } from "path";
6607
6702
  init_errors();
6608
6703
  function createExploreCommand() {
6609
- const explore = new Command12("explore").description("Scan codebase for prompts, datasets, and MutagenT markers").option("-p, --path <dir>", "Directory to scan", ".").option("--depth <n>", "Max directory depth", "10").option("--include <glob>", "Include file pattern", "**/*.{ts,js,py,tsx,jsx}").option("--exclude <dirs>", "Comma-separated directories to exclude", "node_modules,dist,.git,build,.next,__pycache__,venv,.venv").option("--markers-only", "Only find existing MutagenT markers").addHelpText("after", `
6704
+ const explore = new Command15("explore").description("Scan codebase for prompts, datasets, and MutagenT markers").option("-p, --path <dir>", "Directory to scan", ".").option("--depth <n>", "Max directory depth", "10").option("--include <glob>", "Include file pattern", "**/*.{ts,js,py,tsx,jsx}").option("--exclude <dirs>", "Comma-separated directories to exclude", "node_modules,dist,.git,build,.next,__pycache__,venv,.venv").option("--markers-only", "Only find existing MutagenT markers").addHelpText("after", `
6610
6705
  Examples:
6611
- ${chalk15.dim("$")} mutagent explore
6612
- ${chalk15.dim("$")} mutagent explore --path ./src
6613
- ${chalk15.dim("$")} mutagent explore --include "**/*.{ts,py}" --depth 5
6614
- ${chalk15.dim("$")} mutagent explore --markers-only
6615
- ${chalk15.dim("$")} mutagent explore --json
6706
+ ${chalk20.dim("$")} mutagent explore
6707
+ ${chalk20.dim("$")} mutagent explore --path ./src
6708
+ ${chalk20.dim("$")} mutagent explore --include "**/*.{ts,py}" --depth 5
6709
+ ${chalk20.dim("$")} mutagent explore --markers-only
6710
+ ${chalk20.dim("$")} mutagent explore --json
6616
6711
 
6617
6712
  Detection modes:
6618
- ${chalk15.dim("Heuristic")} Template variables ({{var}}), prompt constants, schema definitions
6619
- ${chalk15.dim("Marker")} MutagenT:START/END comment markers from previous uploads
6713
+ ${chalk20.dim("Heuristic")} Template variables ({{var}}), prompt constants, schema definitions
6714
+ ${chalk20.dim("Marker")} MutagenT:START/END comment markers from previous uploads
6620
6715
 
6621
- ${chalk15.dim("Results are saved to .mutagent/mutation-context.md for use by other commands.")}
6716
+ ${chalk20.dim("Results are saved to .mutagent/mutation-context.md for use by other commands.")}
6622
6717
  `).action((options) => {
6623
6718
  const isJson = getJsonFlag(explore);
6624
6719
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6636,7 +6731,7 @@ ${chalk15.dim("Results are saved to .mutagent/mutation-context.md for use by oth
6636
6731
  markersOnly
6637
6732
  };
6638
6733
  if (!isJson) {
6639
- console.log(chalk15.cyan(`
6734
+ console.log(chalk20.cyan(`
6640
6735
  Scanning ${scanPath}...
6641
6736
  `));
6642
6737
  }
@@ -6665,40 +6760,41 @@ Scanning ${scanPath}...
6665
6760
  const totalFindings = result.prompts.length + result.datasets.length + result.markers.length;
6666
6761
  if (totalFindings === 0) {
6667
6762
  output.info("No prompts, datasets, or markers found.");
6668
- console.log(chalk15.dim(`
6763
+ console.log(chalk20.dim(`
6669
6764
  Tip: Create a prompt with template variables like {{input}} to get started.`));
6670
6765
  return;
6671
6766
  }
6672
6767
  if (result.prompts.length > 0) {
6673
- console.log(chalk15.bold(` Prompts Found (${String(result.prompts.length)}):`));
6768
+ console.log(chalk20.bold(` Prompts Found (${String(result.prompts.length)}):`));
6674
6769
  console.log();
6675
6770
  for (const p of result.prompts) {
6676
- const reasonTag = chalk15.dim(`[${p.reason}]`);
6677
- console.log(` ${chalk15.green(p.file)}:${chalk15.yellow(String(p.line))} ${reasonTag}`);
6678
- console.log(` ${chalk15.dim(p.preview)}`);
6771
+ const confidenceTag = p.confidence === "high" ? chalk20.green("[high]") : p.confidence === "medium" ? chalk20.yellow("[medium]") : chalk20.dim("[low]");
6772
+ const reasonTag = chalk20.dim(`[${p.reason}]`);
6773
+ console.log(` ${confidenceTag} ${chalk20.green(p.file)}:${chalk20.yellow(String(p.line))} ${reasonTag}`);
6774
+ console.log(` ${chalk20.dim(p.preview)}`);
6679
6775
  }
6680
6776
  console.log();
6681
6777
  }
6682
6778
  if (result.datasets.length > 0) {
6683
- console.log(chalk15.bold(` Datasets Found (${String(result.datasets.length)}):`));
6779
+ console.log(chalk20.bold(` Datasets Found (${String(result.datasets.length)}):`));
6684
6780
  console.log();
6685
6781
  for (const d of result.datasets) {
6686
- console.log(` ${chalk15.green(d.file)} ${chalk15.dim(`(${String(d.items)} items)`)}`);
6782
+ console.log(` ${chalk20.green(d.file)} ${chalk20.dim(`(${String(d.items)} items)`)}`);
6687
6783
  }
6688
6784
  console.log();
6689
6785
  }
6690
6786
  if (result.markers.length > 0) {
6691
- console.log(chalk15.bold(` MutagenT Markers (${String(result.markers.length)}):`));
6787
+ console.log(chalk20.bold(` MutagenT Markers (${String(result.markers.length)}):`));
6692
6788
  console.log();
6693
6789
  for (const m of result.markers) {
6694
- const idPart = m.platformId ? chalk15.cyan(` id=${m.platformId}`) : "";
6695
- console.log(` ${chalk15.green(m.file)}:${chalk15.yellow(String(m.line))} ${chalk15.magenta(m.type)}${idPart}`);
6790
+ const idPart = m.platformId ? chalk20.cyan(` id=${m.platformId}`) : "";
6791
+ console.log(` ${chalk20.green(m.file)}:${chalk20.yellow(String(m.line))} ${chalk20.magenta(m.type)}${idPart}`);
6696
6792
  }
6697
6793
  console.log();
6698
6794
  }
6699
- console.log(chalk15.dim(" ─────────────────────────────────"));
6700
- console.log(` ${chalk15.bold("Summary:")} ${String(result.prompts.length)} prompts, ${String(result.datasets.length)} datasets, ${String(result.markers.length)} markers`);
6701
- console.log(chalk15.dim(` Saved to .mutagent/mutation-context.md`));
6795
+ console.log(chalk20.dim(" ─────────────────────────────────"));
6796
+ console.log(` ${chalk20.bold("Summary:")} ${String(result.prompts.length)} prompts, ${String(result.datasets.length)} datasets, ${String(result.markers.length)} markers`);
6797
+ console.log(chalk20.dim(` Saved to .mutagent/mutation-context.md`));
6702
6798
  console.log();
6703
6799
  }
6704
6800
  } catch (error) {
@@ -6709,14 +6805,14 @@ Scanning ${scanPath}...
6709
6805
  }
6710
6806
 
6711
6807
  // src/commands/skills.ts
6712
- import { Command as Command13 } from "commander";
6713
- import chalk16 from "chalk";
6714
- import { existsSync as existsSync11, mkdirSync as mkdirSync4, writeFileSync as writeFileSync5 } from "fs";
6715
- import { join as join6 } from "path";
6716
- import { execSync as execSync3 } from "child_process";
6808
+ import { Command as Command16 } from "commander";
6809
+ import chalk21 from "chalk";
6810
+ import { existsSync as existsSync13, mkdirSync as mkdirSync4, writeFileSync as writeFileSync5 } from "fs";
6811
+ import { join as join7 } from "path";
6812
+ import { execSync as execSync4 } from "child_process";
6717
6813
  function findRepoRoot() {
6718
6814
  try {
6719
- return execSync3("git rev-parse --show-toplevel", { encoding: "utf-8", stdio: ["pipe", "pipe", "pipe"] }).trim();
6815
+ return execSync4("git rev-parse --show-toplevel", { encoding: "utf-8", stdio: ["pipe", "pipe", "pipe"] }).trim();
6720
6816
  } catch {
6721
6817
  return process.cwd();
6722
6818
  }
@@ -6830,10 +6926,10 @@ Every evaluation MUST specify criteria targeting either:
6830
6926
  var SKILL_DIR = ".claude/skills/mutagent-cli";
6831
6927
  var SKILL_FILE = "SKILL.md";
6832
6928
  function createSkillsCommand() {
6833
- const skills = new Command13("skills").description("Manage MutagenT CLI skills for coding agents");
6929
+ const skills = new Command16("skills").description("Manage MutagenT CLI skills for coding agents");
6834
6930
  skills.command("install").description("Install MutagenT CLI skill for Claude Code").addHelpText("after", `
6835
6931
  Examples:
6836
- ${chalk16.dim("$")} mutagent skills install
6932
+ ${chalk21.dim("$")} mutagent skills install
6837
6933
 
6838
6934
  This creates a Claude Code skill at .claude/skills/mutagent-cli/SKILL.md
6839
6935
  that teaches coding agents how to use the MutagenT CLI effectively.
@@ -6842,9 +6938,9 @@ that teaches coding agents how to use the MutagenT CLI effectively.
6842
6938
  const isJson = parentCmd ? getJsonFlag(parentCmd) : false;
6843
6939
  const output = new OutputFormatter(isJson ? "json" : "table");
6844
6940
  const repoRoot = findRepoRoot();
6845
- const skillDir = join6(repoRoot, SKILL_DIR);
6846
- const skillPath = join6(skillDir, SKILL_FILE);
6847
- if (!existsSync11(skillDir)) {
6941
+ const skillDir = join7(repoRoot, SKILL_DIR);
6942
+ const skillPath = join7(skillDir, SKILL_FILE);
6943
+ if (!existsSync13(skillDir)) {
6848
6944
  mkdirSync4(skillDir, { recursive: true });
6849
6945
  }
6850
6946
  const content = `${SKILL_FRONTMATTER}
@@ -6860,10 +6956,10 @@ ${SKILL_BODY}
6860
6956
  });
6861
6957
  } else {
6862
6958
  output.success(`Installed MutagenT CLI skill`);
6863
- console.log(` ${chalk16.dim("Path:")} ${skillPath}`);
6959
+ console.log(` ${chalk21.dim("Path:")} ${skillPath}`);
6864
6960
  console.log("");
6865
- console.log(` ${chalk16.dim("This skill teaches coding agents how to use the MutagenT CLI.")}`);
6866
- console.log(` ${chalk16.dim("It will be automatically loaded by Claude Code when relevant triggers match.")}`);
6961
+ console.log(` ${chalk21.dim("This skill teaches coding agents how to use the MutagenT CLI.")}`);
6962
+ console.log(` ${chalk21.dim("It will be automatically loaded by Claude Code when relevant triggers match.")}`);
6867
6963
  }
6868
6964
  });
6869
6965
  return skills;
@@ -6871,8 +6967,8 @@ ${SKILL_BODY}
6871
6967
 
6872
6968
  // src/commands/usage.ts
6873
6969
  init_config();
6874
- import { Command as Command14 } from "commander";
6875
- import chalk17 from "chalk";
6970
+ import { Command as Command17 } from "commander";
6971
+ import chalk22 from "chalk";
6876
6972
  init_errors();
6877
6973
  init_sdk_client();
6878
6974
  var TRIAL_OPTIMIZATION_LIMIT = 5;
@@ -6886,10 +6982,10 @@ function renderProgressBar(used, limit, width = 30) {
6886
6982
  return `${bar} ${String(percent)}%`;
6887
6983
  }
6888
6984
  function createUsageCommand() {
6889
- const usage = new Command14("usage").description("Show resource counts and optimization run limits").addHelpText("after", `
6985
+ const usage = new Command17("usage").description("Show resource counts and optimization run limits").addHelpText("after", `
6890
6986
  Examples:
6891
- ${chalk17.dim("$")} mutagent usage
6892
- ${chalk17.dim("$")} mutagent usage --json
6987
+ ${chalk22.dim("$")} mutagent usage
6988
+ ${chalk22.dim("$")} mutagent usage --json
6893
6989
  `);
6894
6990
  usage.action(async () => {
6895
6991
  const isJson = getJsonFlag(usage);
@@ -6944,21 +7040,21 @@ Examples:
6944
7040
  });
6945
7041
  } else {
6946
7042
  console.log("");
6947
- console.log(chalk17.bold("\uD83D\uDCCA MutagenT Usage"));
6948
- console.log(chalk17.dim("─".repeat(45)));
7043
+ console.log(chalk22.bold("\uD83D\uDCCA MutagenT Usage"));
7044
+ console.log(chalk22.dim("─".repeat(45)));
6949
7045
  console.log("");
6950
- console.log(chalk17.bold("Resources:"));
6951
- console.log(` Prompts: ${chalk17.cyan(String(promptCount))}`);
6952
- console.log(` Datasets: ${chalk17.cyan(String(datasetCount))}`);
6953
- console.log(` Evaluations: ${chalk17.cyan(String(evaluationCount))}`);
7046
+ console.log(chalk22.bold("Resources:"));
7047
+ console.log(` Prompts: ${chalk22.cyan(String(promptCount))}`);
7048
+ console.log(` Datasets: ${chalk22.cyan(String(datasetCount))}`);
7049
+ console.log(` Evaluations: ${chalk22.cyan(String(evaluationCount))}`);
6954
7050
  console.log("");
6955
- console.log(chalk17.bold(`Optimization Runs (${chalk17.yellow("trial")} plan):`));
6956
- console.log(` Remaining: ${chalk17.cyan(String(optimizationRemaining))} / ${String(optimizationLimit)}`);
7051
+ console.log(chalk22.bold(`Optimization Runs (${chalk22.yellow("trial")} plan):`));
7052
+ console.log(` Remaining: ${chalk22.cyan(String(optimizationRemaining))} / ${String(optimizationLimit)}`);
6957
7053
  console.log(` ${renderProgressBar(optimizationUsed, optimizationLimit)}`);
6958
7054
  console.log("");
6959
- console.log(chalk17.yellow(` ⚠ ${String(optimizationRemaining)} optimization runs remaining`));
6960
- console.log(chalk17.dim(` ℹ Optimization run counts are approximate`));
6961
- console.log(` Upgrade: ${chalk17.underline(BILLING_URL)}`);
7055
+ console.log(chalk22.yellow(` ⚠ ${String(optimizationRemaining)} optimization runs remaining`));
7056
+ console.log(chalk22.dim(` ℹ Optimization run counts are approximate`));
7057
+ console.log(` Upgrade: ${chalk22.underline(BILLING_URL)}`);
6962
7058
  console.log("");
6963
7059
  }
6964
7060
  } catch (error) {
@@ -6976,12 +7072,12 @@ if (process.env.CLI_VERSION) {
6976
7072
  } else {
6977
7073
  try {
6978
7074
  const __dirname2 = dirname(fileURLToPath(import.meta.url));
6979
- const pkgPath = join7(__dirname2, "..", "..", "package.json");
6980
- const pkg = JSON.parse(readFileSync10(pkgPath, "utf-8"));
7075
+ const pkgPath = join8(__dirname2, "..", "..", "package.json");
7076
+ const pkg = JSON.parse(readFileSync11(pkgPath, "utf-8"));
6981
7077
  cliVersion = pkg.version ?? cliVersion;
6982
7078
  } catch {}
6983
7079
  }
6984
- var program = new Command15;
7080
+ var program = new Command18;
6985
7081
  program.name("mutagent").description(`MutagenT CLI - AI-native prompt optimization platform
6986
7082
 
6987
7083
  Documentation: https://docs.mutagent.io/cli
@@ -6990,46 +7086,46 @@ program.name("mutagent").description(`MutagenT CLI - AI-native prompt optimizati
6990
7086
  showGlobalOptions: true
6991
7087
  });
6992
7088
  program.addHelpText("after", `
6993
- ${chalk18.yellow("Non-Interactive Mode (CI/CD & Coding Agents):")}
6994
- export MUTAGENT_API_KEY=mt_... ${chalk18.dim("or")} --api-key mt_...
6995
- --json ${chalk18.dim("for structured output")} --non-interactive ${chalk18.dim("to disable prompts")}
6996
-
6997
- ${chalk18.yellow("Command Navigation:")}
6998
- mutagent auth login --browser ${chalk18.dim("Authenticate (OAuth)")}
6999
- mutagent auth status ${chalk18.dim("Check auth + workspace")}
7000
- mutagent init ${chalk18.dim("Initialize project (.mutagentrc.json)")}
7001
- mutagent explore ${chalk18.dim("Discover prompts in codebase")}
7002
- mutagent workspaces list --json ${chalk18.dim("List workspaces (verify ID)")}
7003
- mutagent config set workspace <id> ${chalk18.dim("Set active workspace")}
7004
- mutagent usage --json ${chalk18.dim("Check plan limits")}
7005
-
7006
- mutagent prompts create --help ${chalk18.dim("Upload prompt (read help first!)")}
7007
- mutagent prompts list --json ${chalk18.dim("List prompts")}
7008
- mutagent prompts get <id> --json ${chalk18.dim("Full prompt details + schemas")}
7009
-
7010
- mutagent prompts dataset add --help ${chalk18.dim("Upload dataset (read help first!)")}
7011
- mutagent prompts dataset list <id> ${chalk18.dim("List datasets")}
7012
-
7013
- mutagent prompts evaluation create --help ${chalk18.dim("Create eval (read help first!)")}
7014
- mutagent prompts evaluation create <id> --guided --json ${chalk18.dim("Guided eval workflow")}
7015
- mutagent prompts evaluation list <id> --json ${chalk18.dim("List evaluations")}
7016
-
7017
- mutagent prompts optimize start --help ${chalk18.dim("Run optimization (read help first!)")}
7018
- mutagent prompts optimize status <job-id> ${chalk18.dim("Poll progress")}
7019
- mutagent prompts optimize results <job-id> ${chalk18.dim("View scorecard")}
7020
-
7021
- mutagent integrate <framework> ${chalk18.dim("Framework integration guide")}
7022
- mutagent playground run <id> --input '{...}' ${chalk18.dim("Quick test")}
7023
-
7024
- ${chalk18.yellow("Workflow: Evaluate → Optimize:")}
7025
- 1. mutagent prompts create --help ${chalk18.dim("← read help")}
7026
- 2. mutagent prompts create ... --json ${chalk18.dim("← upload prompt with {variables}")}
7027
- 3. mutagent prompts dataset add --help ${chalk18.dim("← read help")}
7028
- 4. mutagent prompts dataset add <id> ... --json ${chalk18.dim("← upload dataset")}
7029
- 5. mutagent prompts evaluation create <id> --guided --json ${chalk18.dim("← guided eval")}
7089
+ ${chalk23.yellow("Non-Interactive Mode (CI/CD & Coding Agents):")}
7090
+ export MUTAGENT_API_KEY=mt_... ${chalk23.dim("or")} --api-key mt_...
7091
+ --json ${chalk23.dim("for structured output")} --non-interactive ${chalk23.dim("to disable prompts")}
7092
+
7093
+ ${chalk23.yellow("Command Navigation:")}
7094
+ mutagent auth login --browser ${chalk23.dim("Authenticate (OAuth)")}
7095
+ mutagent auth status ${chalk23.dim("Check auth + workspace")}
7096
+ mutagent init ${chalk23.dim("Initialize project (.mutagentrc.json)")}
7097
+ mutagent explore ${chalk23.dim("Discover prompts in codebase")}
7098
+ mutagent workspaces list --json ${chalk23.dim("List workspaces (verify ID)")}
7099
+ mutagent config set workspace <id> ${chalk23.dim("Set active workspace")}
7100
+ mutagent usage --json ${chalk23.dim("Check plan limits")}
7101
+
7102
+ mutagent prompts create --help ${chalk23.dim("Upload prompt (read help first!)")}
7103
+ mutagent prompts list --json ${chalk23.dim("List prompts")}
7104
+ mutagent prompts get <id> --json ${chalk23.dim("Full prompt details + schemas")}
7105
+
7106
+ mutagent prompts dataset add --help ${chalk23.dim("Upload dataset (read help first!)")}
7107
+ mutagent prompts dataset list <id> ${chalk23.dim("List datasets")}
7108
+
7109
+ mutagent prompts evaluation create --help ${chalk23.dim("Create eval (read help first!)")}
7110
+ mutagent prompts evaluation create <id> --guided --json ${chalk23.dim("Guided eval workflow")}
7111
+ mutagent prompts evaluation list <id> --json ${chalk23.dim("List evaluations")}
7112
+
7113
+ mutagent prompts optimize start --help ${chalk23.dim("Run optimization (read help first!)")}
7114
+ mutagent prompts optimize status <job-id> ${chalk23.dim("Poll progress")}
7115
+ mutagent prompts optimize results <job-id> ${chalk23.dim("View scorecard")}
7116
+
7117
+ mutagent integrate <framework> ${chalk23.dim("Framework integration guide")}
7118
+ mutagent playground run <id> --input '{...}' ${chalk23.dim("Quick test")}
7119
+
7120
+ ${chalk23.yellow("Workflow: Evaluate → Optimize:")}
7121
+ 1. mutagent prompts create --help ${chalk23.dim("← read help")}
7122
+ 2. mutagent prompts create ... --json ${chalk23.dim("← upload prompt with {variables} + inputSchema")}
7123
+ 3. mutagent prompts dataset add --help ${chalk23.dim("← read help")}
7124
+ 4. mutagent prompts dataset add <id> ... --json ${chalk23.dim("← upload dataset")}
7125
+ 5. mutagent prompts evaluation create <id> --guided --json ${chalk23.dim("← guided eval")}
7030
7126
  6. mutagent prompts optimize start <id> --dataset <d> --evaluation <e> --json
7031
7127
 
7032
- ${chalk18.yellow("AI Agent Rules (MANDATORY for coding agents):")}
7128
+ ${chalk23.yellow("AI Agent Rules (MANDATORY for coding agents):")}
7033
7129
  1. EVERY command MUST include --json (no exceptions)
7034
7130
  2. Run <command> --help BEFORE first use of any command
7035
7131
  3. Use --guided --json for evaluation creation (NEVER --guided alone)
@@ -7038,9 +7134,9 @@ ${chalk18.yellow("AI Agent Rules (MANDATORY for coding agents):")}
7038
7134
  6. Use {single_braces} for template variables in prompts
7039
7135
  7. Collect evaluation rubrics from the user — NEVER auto-generate
7040
7136
  ${!hasCredentials() ? `
7041
- ` + chalk18.yellow(" Warning: Not authenticated. Run: mutagent auth login --browser") + `
7137
+ ` + chalk23.yellow(" Warning: Not authenticated. Run: mutagent auth login --browser") + `
7042
7138
  ` : ""}${!hasRcConfig() ? `
7043
- ` + chalk18.green(" Get started: mutagent init") + `
7139
+ ` + chalk23.green(" Get started: mutagent init") + `
7044
7140
  ` : ""}`);
7045
7141
  program.hook("preAction", (thisCommand) => {
7046
7142
  const globalOpts = thisCommand.optsWithGlobals();
@@ -7070,5 +7166,5 @@ program.addCommand(createSkillsCommand());
7070
7166
  program.addCommand(createUsageCommand());
7071
7167
  program.parse();
7072
7168
 
7073
- //# debugId=55FA8E38F306DE3764756E2164756E21
7169
+ //# debugId=88ECCF736E0A64AD64756E2164756E21
7074
7170
  //# sourceMappingURL=cli.js.map