@mutagent/cli 0.1.28 → 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() {
@@ -2470,9 +2449,11 @@ 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
- if (directive?.instruction) {
2454
+ if (directive?.renderedCard) {
2455
+ console.error(directive.renderedCard);
2456
+ } else if (directive?.instruction) {
2476
2457
  console.error(`
2477
2458
  ⚠️ AI AGENT: ${directive.instruction}
2478
2459
  `);
@@ -2491,115 +2472,247 @@ function formatCreationHints(hint) {
2491
2472
  return lines.join(`
2492
2473
  `);
2493
2474
  }
2494
- 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.";
2476
+ function renderPlainCard(opts) {
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;
2504
+ };
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
+ }
2514
+ if (opts.links.length) {
2515
+ out.push(blank(), sep(), blank());
2516
+ for (const l of opts.links) {
2517
+ out.push(ln(` ${l.label.padEnd(labelWidth)}${l.url}`));
2518
+ }
2519
+ }
2520
+ if (opts.next.length) {
2521
+ out.push(blank(), sep(), blank());
2522
+ out.push(ln(" Next"));
2523
+ for (const n of opts.next) {
2524
+ out.push(ln(` → ${n}`));
2525
+ }
2526
+ }
2527
+ out.push(blank(), bot);
2528
+ return out.join(`
2529
+ `);
2530
+ }
2495
2531
  function promptCreatedDirective(promptId, name) {
2532
+ const title = `Prompt Created — ${name}`;
2533
+ const dashboardUrl = promptLink(promptId);
2534
+ const apiUrl = `/api/prompts/${String(promptId)}`;
2535
+ const rows = [
2536
+ { label: "Name", value: name },
2537
+ { label: "Prompt ID", value: String(promptId) }
2538
+ ];
2539
+ const cardLinks = [
2540
+ { label: "Dashboard", url: dashboardUrl },
2541
+ { label: "API", url: apiUrl }
2542
+ ];
2543
+ const next = [
2544
+ `mutagent prompts evaluation create ${String(promptId)} --guided --json`,
2545
+ `mutagent prompts dataset add ${String(promptId)} -d '[...]' --name "my-dataset" --json`
2546
+ ];
2496
2547
  return {
2497
2548
  display: "status_card",
2498
2549
  template: "prompt_created",
2499
- title: `Prompt Created — ${name}`,
2550
+ title,
2500
2551
  fields: { promptId: String(promptId), name },
2501
- links: {
2502
- dashboard: promptLink(promptId),
2503
- api: `/api/prompts/${String(promptId)}`
2504
- },
2505
- next: [
2506
- `mutagent prompts evaluation create ${String(promptId)} --guided`,
2507
- `mutagent prompts dataset add ${String(promptId)} -d '[...]' --name "my-dataset"`
2508
- ],
2509
- instruction: CARD_INSTRUCTION
2552
+ links: { dashboard: dashboardUrl, api: apiUrl },
2553
+ next,
2554
+ instruction: CARD_INSTRUCTION,
2555
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2510
2556
  };
2511
2557
  }
2512
2558
  function promptUpdatedDirective(promptId, name) {
2559
+ const title = `Prompt Updated — ${name}`;
2560
+ const dashboardUrl = promptLink(promptId);
2561
+ const apiUrl = `/api/prompts/${String(promptId)}`;
2562
+ const rows = [
2563
+ { label: "Name", value: name },
2564
+ { label: "Prompt ID", value: String(promptId) }
2565
+ ];
2566
+ const cardLinks = [
2567
+ { label: "Dashboard", url: dashboardUrl },
2568
+ { label: "API", url: apiUrl }
2569
+ ];
2570
+ const next = [`mutagent prompts get ${String(promptId)} --json`];
2513
2571
  return {
2514
2572
  display: "status_card",
2515
2573
  template: "prompt_updated",
2516
- title: `Prompt Updated — ${name}`,
2574
+ title,
2517
2575
  fields: { promptId: String(promptId), name },
2518
- links: {
2519
- dashboard: promptLink(promptId),
2520
- api: `/api/prompts/${String(promptId)}`
2521
- },
2522
- next: [`mutagent prompts get ${String(promptId)} --json`],
2523
- instruction: CARD_INSTRUCTION
2576
+ links: { dashboard: dashboardUrl, api: apiUrl },
2577
+ next,
2578
+ instruction: CARD_INSTRUCTION,
2579
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2524
2580
  };
2525
2581
  }
2526
2582
  function promptDeletedDirective(promptId) {
2583
+ const title = `Prompt Deleted — ${promptId}`;
2584
+ const dashboardUrl = promptsDashboardLink();
2585
+ const rows = [{ label: "Prompt ID", value: promptId }];
2586
+ const cardLinks = [{ label: "Dashboard", url: dashboardUrl }];
2587
+ const next = [`mutagent prompts list --json`];
2527
2588
  return {
2528
2589
  display: "status_card",
2529
2590
  template: "prompt_deleted",
2530
- title: `Prompt Deleted — ${promptId}`,
2591
+ title,
2531
2592
  fields: { promptId },
2532
- links: { dashboard: promptsDashboardLink() },
2533
- next: [`mutagent prompts list --json`],
2534
- instruction: CARD_INSTRUCTION
2593
+ links: { dashboard: dashboardUrl },
2594
+ next,
2595
+ instruction: CARD_INSTRUCTION,
2596
+ renderedCard: renderPlainCard({ title, icon: "✗", rows, links: cardLinks, next })
2535
2597
  };
2536
2598
  }
2537
- function datasetAddedDirective(promptId, datasetId, name, itemCount) {
2599
+ function datasetAddedDirective(promptId, datasetId, name, itemCount, state) {
2600
+ const title = `Dataset Added — ${name}`;
2601
+ const dashboardUrl = datasetLink(promptId, datasetId);
2602
+ const apiUrl = `/api/prompts/${String(promptId)}/datasets/${String(datasetId)}`;
2603
+ const rows = [
2604
+ { label: "Name", value: name },
2605
+ { label: "Prompt ID", value: String(promptId) },
2606
+ { label: "Dataset ID", value: String(datasetId) },
2607
+ ...itemCount !== undefined ? [{ label: "Items", value: String(itemCount) }] : []
2608
+ ];
2609
+ const cardLinks = [
2610
+ { label: "Dashboard", url: dashboardUrl },
2611
+ { label: "API", url: apiUrl }
2612
+ ];
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
+ ] : [
2617
+ `mutagent prompts evaluation create ${String(promptId)} --guided --json`,
2618
+ `mutagent prompts optimize start ${String(promptId)} --dataset ${String(datasetId)} --json`
2619
+ ];
2538
2620
  return {
2539
2621
  display: "status_card",
2540
2622
  template: "dataset_added",
2541
- title: `Dataset Added — ${name}`,
2623
+ title,
2542
2624
  fields: {
2543
2625
  promptId: String(promptId),
2544
2626
  datasetId: String(datasetId),
2545
2627
  name,
2546
2628
  ...itemCount !== undefined ? { itemCount } : {}
2547
2629
  },
2548
- links: {
2549
- dashboard: datasetLink(promptId, datasetId),
2550
- api: `/api/prompts/${String(promptId)}/datasets/${String(datasetId)}`
2551
- },
2552
- next: [
2553
- `mutagent prompts evaluation create ${String(promptId)} --guided`,
2554
- `mutagent prompts optimize start ${String(promptId)} --dataset ${String(datasetId)}`
2555
- ],
2556
- instruction: CARD_INSTRUCTION
2630
+ links: { dashboard: dashboardUrl, api: apiUrl },
2631
+ next,
2632
+ instruction: CARD_INSTRUCTION,
2633
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2557
2634
  };
2558
2635
  }
2559
- function datasetRemovedDirective(promptId, datasetId) {
2636
+ function datasetDeletedDirective(promptId, datasetId) {
2637
+ const title = `Dataset Deleted — ${datasetId}`;
2638
+ const datasetsUrl = promptDatasetsLink(promptId);
2639
+ const rows = [
2640
+ { label: "Prompt ID", value: promptId },
2641
+ { label: "Dataset ID", value: datasetId }
2642
+ ];
2643
+ const cardLinks = [{ label: "Datasets", url: datasetsUrl }];
2644
+ const next = [`mutagent prompts dataset list ${promptId} --json`];
2560
2645
  return {
2561
2646
  display: "status_card",
2562
- template: "dataset_removed",
2563
- title: `Dataset Removed — ${datasetId}`,
2647
+ template: "dataset_deleted",
2648
+ title,
2564
2649
  fields: { promptId, datasetId },
2565
- links: { datasets: promptDatasetsLink(promptId) },
2566
- next: [`mutagent prompts dataset list ${promptId} --json`],
2567
- instruction: CARD_INSTRUCTION
2650
+ links: { datasets: datasetsUrl },
2651
+ next,
2652
+ instruction: CARD_INSTRUCTION,
2653
+ renderedCard: renderPlainCard({ title, icon: "✗", rows, links: cardLinks, next })
2568
2654
  };
2569
2655
  }
2570
- function evaluationCreatedDirective(promptId, evalId, name, criteriaCount) {
2656
+ function evaluationCreatedDirective(promptId, evalId, name, criteriaCount, state) {
2657
+ const title = `Evaluation Created — ${name}`;
2658
+ const dashboardUrl = evaluationLink(promptId, evalId);
2659
+ const apiUrl = `/api/prompts/${String(promptId)}/evaluations/${String(evalId)}`;
2660
+ const rows = [
2661
+ { label: "Name", value: name },
2662
+ { label: "Prompt ID", value: String(promptId) },
2663
+ { label: "Evaluation ID", value: String(evalId) },
2664
+ { label: "Criteria", value: String(criteriaCount) }
2665
+ ];
2666
+ const cardLinks = [
2667
+ { label: "Dashboard", url: dashboardUrl },
2668
+ { label: "API", url: apiUrl }
2669
+ ];
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`
2675
+ ];
2571
2676
  return {
2572
2677
  display: "status_card",
2573
2678
  template: "evaluation_created",
2574
- title: `Evaluation Created — ${name}`,
2679
+ title,
2575
2680
  fields: {
2576
2681
  promptId: String(promptId),
2577
2682
  evaluationId: String(evalId),
2578
2683
  name,
2579
2684
  criteriaCount
2580
2685
  },
2581
- links: {
2582
- dashboard: evaluationLink(promptId, evalId),
2583
- api: `/api/prompts/${String(promptId)}/evaluations/${String(evalId)}`
2584
- },
2585
- next: [
2586
- `mutagent prompts optimize start ${String(promptId)} --dataset <dataset-id>`
2587
- ],
2588
- instruction: CARD_INSTRUCTION
2686
+ links: { dashboard: dashboardUrl, api: apiUrl },
2687
+ next,
2688
+ instruction: CARD_INSTRUCTION,
2689
+ renderedCard: renderPlainCard({ title, icon: "✓", rows, links: cardLinks, next })
2589
2690
  };
2590
2691
  }
2591
2692
  function evaluationDeletedDirective(evaluationId) {
2693
+ const title = `Evaluation Deleted — ${evaluationId}`;
2694
+ const dashboardUrl = promptsDashboardLink();
2695
+ const rows = [{ label: "Evaluation ID", value: evaluationId }];
2696
+ const cardLinks = [{ label: "Dashboard", url: dashboardUrl }];
2697
+ const next = [];
2592
2698
  return {
2593
2699
  display: "status_card",
2594
2700
  template: "evaluation_deleted",
2595
- title: `Evaluation Deleted — ${evaluationId}`,
2701
+ title,
2596
2702
  fields: { evaluationId },
2597
- links: { dashboard: promptsDashboardLink() },
2598
- next: [],
2599
- instruction: CARD_INSTRUCTION
2703
+ links: { dashboard: dashboardUrl },
2704
+ next,
2705
+ instruction: CARD_INSTRUCTION,
2706
+ renderedCard: renderPlainCard({ title, icon: "✗", rows, links: cardLinks, next })
2600
2707
  };
2601
2708
  }
2602
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
+
2603
2716
  // src/lib/schema-helpers.ts
2604
2717
  function isValidJsonSchema(schema) {
2605
2718
  if (schema === null || schema === undefined)
@@ -2658,396 +2771,115 @@ function formatSchemaWarning(fieldName) {
2658
2771
  `);
2659
2772
  }
2660
2773
 
2661
- // src/lib/scorecard.ts
2662
- import chalk5 from "chalk";
2663
- function truncateText(text, maxLen) {
2664
- if (text.length <= maxLen)
2665
- return text;
2666
- return text.substring(0, maxLen - 3) + "...";
2667
- }
2668
- function formatScoreChange(before, after) {
2669
- if (before === undefined || after === undefined)
2670
- return "";
2671
- const diff = after - before;
2672
- const pct = before > 0 ? Math.round(diff / before * 100) : 0;
2673
- if (diff > 0)
2674
- return chalk5.green(` (+${String(pct)}%)`);
2675
- if (diff < 0)
2676
- return chalk5.red(` (${String(pct)}%)`);
2677
- return chalk5.dim(" (no change)");
2678
- }
2679
- function formatScore(score) {
2680
- if (score === undefined)
2681
- return chalk5.dim("N/A");
2682
- return score >= 0.8 ? chalk5.green(score.toFixed(2)) : score >= 0.5 ? chalk5.yellow(score.toFixed(2)) : chalk5.red(score.toFixed(2));
2683
- }
2684
- function renderScorecard(data) {
2685
- const { job, prompt } = data;
2686
- const boxWidth = 55;
2687
- const topBorder = `┌${"─".repeat(boxWidth)}┐`;
2688
- const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
2689
- const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
2690
- const pad = (text, width) => {
2691
- const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
2692
- const remaining = width - stripped.length;
2693
- return remaining > 0 ? text + " ".repeat(remaining) : text;
2694
- };
2695
- const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
2696
- const originalScore = data.originalScore;
2697
- const bestScore = data.bestScore;
2698
- const iterations = data.iterationsCompleted ?? job.config?.maxIterations ?? 0;
2699
- const originalText = data.originalPrompt?.systemPrompt ?? data.originalPrompt?.rawPrompt ?? "(original prompt)";
2700
- const optimizedText = prompt.systemPrompt ?? prompt.rawPrompt ?? prompt.humanPrompt ?? "(optimized prompt)";
2701
- console.log("");
2702
- console.log(topBorder);
2703
- console.log(line(chalk5.bold("Optimization Results")));
2704
- console.log(separator);
2705
- console.log(line(chalk5.dim("BEFORE")));
2706
- console.log(line(` Prompt: ${chalk5.dim(truncateText(originalText, 38))}`));
2707
- console.log(line(` Score: ${formatScore(originalScore)}`));
2708
- console.log(line(""));
2709
- console.log(line(chalk5.bold("AFTER")));
2710
- console.log(line(` Prompt: ${chalk5.cyan(truncateText(optimizedText, 38))}`));
2711
- console.log(line(` Score: ${formatScore(bestScore)}${formatScoreChange(originalScore, bestScore)}`));
2712
- console.log(separator);
2713
- if (data.criteriaScores && data.criteriaScores.length > 0) {
2714
- console.log(line(chalk5.dim(" Criterion Before After Change")));
2715
- console.log(line(chalk5.dim(" " + "─".repeat(45))));
2716
- for (const c of data.criteriaScores) {
2717
- const name = c.name.length > 16 ? c.name.substring(0, 13) + "..." : c.name;
2718
- const paddedName = name + " ".repeat(18 - name.length);
2719
- const beforeStr = c.before !== undefined ? c.before.toFixed(2) : "N/A ";
2720
- const afterStr = c.after !== undefined ? c.after.toFixed(2) : "N/A ";
2721
- const changeStr = c.before !== undefined && c.after !== undefined && c.before > 0 ? (() => {
2722
- const pct = Math.round((c.after - c.before) / c.before * 100);
2723
- if (pct > 0)
2724
- return chalk5.green(`+${String(pct)}%`);
2725
- if (pct < 0)
2726
- return chalk5.red(`${String(pct)}%`);
2727
- return chalk5.dim("0%");
2728
- })() : "";
2729
- 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: ${wsId}`);
2805
+ }
2806
+ if (limited.length === 0) {
2807
+ output.info("No prompts found in current workspace. Run 'mutagent config list' to check workspace context.");
2808
+ } else {
2809
+ const formatted = limited.map((p) => ({
2810
+ id: p.id,
2811
+ name: p.name,
2812
+ version: p.version,
2813
+ updated: p.updatedAt ? new Date(p.updatedAt).toLocaleDateString() : "N/A",
2814
+ url: promptLink(p.id)
2815
+ }));
2816
+ output.output(formatted);
2817
+ }
2818
+ }
2819
+ } catch (error) {
2820
+ handleError(error, isJson);
2730
2821
  }
2731
- console.log(line(chalk5.dim(" " + "─".repeat(45))));
2732
- const overallBefore = originalScore !== undefined ? originalScore.toFixed(2) : "N/A ";
2733
- const overallAfter = bestScore !== undefined ? bestScore.toFixed(2) : "N/A ";
2734
- const overallChange = originalScore !== undefined && bestScore !== undefined && originalScore > 0 ? (() => {
2735
- const pct = Math.round((bestScore - originalScore) / originalScore * 100);
2736
- if (pct > 0)
2737
- return chalk5.green(`+${String(pct)}%`);
2738
- if (pct < 0)
2739
- return chalk5.red(`${String(pct)}%`);
2740
- return chalk5.dim("0%");
2741
- })() : "";
2742
- console.log(line(` ${"Overall" + " ".repeat(11)}${overallBefore} ${overallAfter} ${overallChange}`));
2743
- console.log(separator);
2744
- }
2745
- const statusStr = job.status === "completed" ? chalk5.green("completed") : chalk5.yellow(job.status);
2746
- console.log(line(`Status: ${statusStr} | Iterations: ${String(iterations)}`));
2747
- if (job.config?.model) {
2748
- console.log(line(`Model: ${chalk5.dim(job.config.model)}`));
2749
- }
2750
- if (data.scoreProgression && data.scoreProgression.length > 0) {
2751
- console.log(line(""));
2752
- console.log(line(chalk5.dim("Score Progression:")));
2753
- const barWidth = 10;
2754
- for (let i = 0;i < data.scoreProgression.length; i++) {
2755
- const s = data.scoreProgression[i] ?? 0;
2756
- const filled = Math.round(s * barWidth);
2757
- const bar = "█".repeat(filled) + "░".repeat(barWidth - filled);
2758
- console.log(line(chalk5.dim(` #${String(i + 1)}: ${bar} ${s.toFixed(2)}`)));
2759
- }
2760
- }
2761
- console.log(separator);
2762
- console.log(line(`Dashboard: ${chalk5.underline(optimizerLink(job.id))}`));
2763
- console.log(bottomBorder);
2764
- console.log("");
2765
- }
2766
- function renderOptimizationStartCard(data) {
2767
- const { job } = data;
2768
- const boxWidth = 60;
2769
- const topBorder = `┌${"─".repeat(boxWidth)}┐`;
2770
- const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
2771
- const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
2772
- const pad = (text, width) => {
2773
- const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
2774
- const remaining = width - stripped.length;
2775
- return remaining > 0 ? text + " ".repeat(remaining) : text;
2776
- };
2777
- const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
2778
- const maxIter = job.config.maxIterations;
2779
- const model = job.config.model ?? "default";
2780
- const target = job.config.targetScore ?? 0.8;
2781
- console.log("");
2782
- console.log(topBorder);
2783
- console.log(line(chalk5.bold("⚡ Optimization Started")));
2784
- console.log(separator);
2785
- console.log(line(`Job ID: ${chalk5.cyan(job.id)}`));
2786
- console.log(line(`Prompt: ${chalk5.dim(data.promptId)}`));
2787
- console.log(line(`Dataset: ${chalk5.dim(data.datasetId)}`));
2788
- console.log(line(`Iterations: ${chalk5.bold(String(maxIter))} | Target: ${chalk5.bold(target.toFixed(2))}`));
2789
- console.log(line(`Model: ${chalk5.dim(model)}`));
2790
- console.log(line(`Status: ${chalk5.yellow(job.status)}`));
2791
- console.log(separator);
2792
- console.log(line(`\uD83D\uDD17 Monitor: ${chalk5.underline(optimizerLink(job.id))}`));
2793
- console.log(line(chalk5.dim(`Next: mutagent prompts optimize status ${job.id}`)));
2794
- console.log(bottomBorder);
2795
- console.log(AI_DIRECTIVE);
2796
- console.log("");
2797
- }
2798
- function renderOptimizationStatusCard(status) {
2799
- const boxWidth = 60;
2800
- const topBorder = `┌${"─".repeat(boxWidth)}┐`;
2801
- const bottomBorder = `└${"─".repeat(boxWidth)}┘`;
2802
- const separator = `│ ${"─".repeat(boxWidth - 2)} │`;
2803
- const pad = (text, width) => {
2804
- const stripped = text.replace(/\u001B\[[0-9;]*m/g, "");
2805
- const remaining = width - stripped.length;
2806
- return remaining > 0 ? text + " ".repeat(remaining) : text;
2807
- };
2808
- const line = (text) => `│ ${pad(text, boxWidth - 2)} │`;
2809
- const progress = status.progressPercent;
2810
- const barWidth = 20;
2811
- const filled = Math.round(progress / 100 * barWidth);
2812
- const progressBar = "█".repeat(filled) + "░".repeat(barWidth - filled);
2813
- const statusColor = status.status === "completed" ? chalk5.green : status.status === "failed" ? chalk5.red : status.status === "cancelled" ? chalk5.gray : status.status === "running" ? chalk5.cyan : chalk5.yellow;
2814
- const scoreStr = status.bestScore !== undefined ? formatScore(status.bestScore) : chalk5.dim("pending");
2815
- console.log("");
2816
- console.log(topBorder);
2817
- console.log(line(chalk5.bold("\uD83D\uDCCA Optimization Status")));
2818
- console.log(separator);
2819
- console.log(line(`Job ID: ${chalk5.cyan(status.jobId)}`));
2820
- console.log(line(`Status: ${statusColor(status.status)}`));
2821
- console.log(line(`Iteration: ${chalk5.bold(`${String(status.currentIteration)}/${String(status.maxIterations)}`)}`));
2822
- console.log(line(`Best Score: ${scoreStr}`));
2823
- console.log(line(""));
2824
- console.log(line(`Progress: [${progressBar}] ${String(progress)}%`));
2825
- if (status.message) {
2826
- console.log(line(`Message: ${chalk5.dim(status.message)}`));
2827
- }
2828
- console.log(separator);
2829
- console.log(line(`\uD83D\uDD17 Monitor: ${chalk5.underline(optimizerLink(status.jobId))}`));
2830
- if (status.status === "completed") {
2831
- console.log(line(chalk5.dim(`Next: mutagent prompts optimize results ${status.jobId}`)));
2832
- } else if (status.status === "running" || status.status === "pending") {
2833
- console.log(line(chalk5.dim(`Refresh: mutagent prompts optimize status ${status.jobId}`)));
2834
- }
2835
- console.log(bottomBorder);
2836
- console.log(AI_DIRECTIVE);
2837
- console.log("");
2838
- }
2839
- function startDirective(job, promptId) {
2840
- return {
2841
- display: "status_card",
2842
- template: "optimization_started",
2843
- title: `Optimization Started — Prompt ${promptId}`,
2844
- fields: {
2845
- jobId: job.id,
2846
- promptId,
2847
- status: job.status,
2848
- maxIterations: job.config.maxIterations,
2849
- targetScore: job.config.targetScore ?? 0.8,
2850
- model: job.config.model ?? "default"
2851
- },
2852
- links: {
2853
- monitor: optimizerLink(job.id),
2854
- api: `/api/prompts/${promptId}/optimizations/${job.id}`
2855
- },
2856
- next: [`mutagent prompts optimize status ${job.id}`],
2857
- 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."
2858
- };
2859
- }
2860
- function statusDirective(status) {
2861
- const next = status.status === "completed" ? [`mutagent prompts optimize results ${status.jobId}`] : status.status === "running" || status.status === "pending" ? [`mutagent prompts optimize status ${status.jobId}`] : [];
2862
- return {
2863
- display: "status_card",
2864
- template: "optimization_progress",
2865
- title: `Optimization ${status.status} — ${status.jobId}`,
2866
- fields: {
2867
- jobId: status.jobId,
2868
- status: status.status,
2869
- iteration: `${String(status.currentIteration)}/${String(status.maxIterations)}`,
2870
- bestScore: status.bestScore,
2871
- progress: `${String(status.progressPercent)}%`,
2872
- message: status.message
2873
- },
2874
- links: { monitor: optimizerLink(status.jobId) },
2875
- next,
2876
- 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."
2877
- };
2878
- }
2879
- function showPromptDiff(original, optimized) {
2880
- console.log("");
2881
- console.log(chalk5.bold(" Prompt Diff:"));
2882
- console.log("");
2883
- console.log(chalk5.red(" - " + (original ?? "(empty)")));
2884
- console.log(chalk5.green(" + " + (optimized ?? "(empty)")));
2885
- console.log("");
2886
- }
2887
-
2888
- // src/commands/prompts.ts
2889
- function updateMutationContext(updater) {
2890
- try {
2891
- const ctx = MutationContext.load();
2892
- updater(ctx);
2893
- ctx.save();
2894
- } catch {}
2895
- }
2896
- var PREREQUISITES_TEXT = `
2897
- ${chalk6.red("Prerequisites (required):")}
2898
- 1. Evaluation criteria defined ${chalk6.dim("(via dashboard or evaluation create)")}
2899
- 2. Dataset uploaded ${chalk6.dim("mutagent prompts dataset list <prompt-id>")}
2900
- ${chalk6.dim("Note: LLM provider config is only required when the server uses external providers (USE_EXT_PROVIDERS=true)")}`;
2901
- function parseValidationErrors(error) {
2902
- if (error instanceof ApiError) {
2903
- try {
2904
- const body = JSON.parse(error.message);
2905
- if (body.errors && Array.isArray(body.errors)) {
2906
- return body.errors.map((e) => e.message ?? JSON.stringify(e));
2907
- }
2908
- if (body.message) {
2909
- return [body.message];
2910
- }
2911
- } catch {
2912
- return [error.message];
2913
- }
2914
- }
2915
- if (error instanceof Error) {
2916
- return [error.message];
2917
- }
2918
- return ["An unknown error occurred"];
2919
- }
2920
- function isSchemaEmpty(schema) {
2921
- if (schema === undefined || schema === null)
2922
- return true;
2923
- if (typeof schema !== "object")
2924
- return false;
2925
- const obj = schema;
2926
- if (Object.keys(obj).length === 0)
2927
- return true;
2928
- return false;
2929
- }
2930
- function createPromptsCommand() {
2931
- const prompts = new Command3("prompts").description("Manage prompts, datasets, evaluations, and optimizations").addHelpText("after", `
2932
- Examples:
2933
- ${chalk6.dim("$")} mutagent prompts list
2934
- ${chalk6.dim("$")} mutagent prompts get <prompt-id>
2935
- ${chalk6.dim("$")} mutagent prompts create --name "my-prompt" --system "You are helpful" --human "{input}"
2936
- ${chalk6.dim("$")} mutagent prompts dataset list <prompt-id>
2937
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --name "My Eval"
2938
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
2939
-
2940
- Subcommands:
2941
- list, get, create, update, delete
2942
- dataset list|add|remove
2943
- evaluation list|create|delete
2944
- optimize start|status|results
2945
- `);
2946
- prompts.command("list").description("List all prompts").option("-l, --limit <n>", "Limit results", "50").addHelpText("after", `
2947
- Examples:
2948
- ${chalk6.dim("$")} mutagent prompts list
2949
- ${chalk6.dim("$")} mutagent prompts list --limit 10
2950
- ${chalk6.dim("$")} mutagent prompts list --json
2951
-
2952
- ${chalk6.dim("Tip: Use --json for machine-readable output (AI agents, CI pipelines).")}
2953
- `).action(async (options) => {
2954
- const isJson = getJsonFlag(prompts);
2955
- const output = new OutputFormatter(isJson ? "json" : "table");
2956
- try {
2957
- const client = getSDKClient();
2958
- const promptsList = await client.listPrompts();
2959
- const limit = parseInt(options.limit, 10) || 50;
2960
- const limited = promptsList.slice(0, limit);
2961
- if (isJson) {
2962
- const withLinks = limited.map((p) => ({
2963
- id: p.id,
2964
- name: p.name,
2965
- version: p.version,
2966
- isLatest: p.isLatest,
2967
- updatedAt: p.updatedAt,
2968
- _links: promptLinks(p.id)
2969
- }));
2970
- output.output(withLinks);
2971
- } else {
2972
- const wsId = client.getCurrentWorkspaceId();
2973
- if (wsId) {
2974
- output.info(`Workspace: ${wsId}`);
2975
- }
2976
- if (limited.length === 0) {
2977
- output.info("No prompts found in current workspace. Run 'mutagent config list' to check workspace context.");
2978
- } else {
2979
- const formatted = limited.map((p) => ({
2980
- id: p.id,
2981
- name: p.name,
2982
- version: p.version,
2983
- updated: p.updatedAt ? new Date(p.updatedAt).toLocaleDateString() : "N/A",
2984
- url: promptLink(p.id)
2985
- }));
2986
- output.output(formatted);
2987
- }
2988
- }
2989
- } catch (error) {
2990
- handleError(error, isJson);
2991
- }
2992
- });
2993
- 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", `
2994
- Examples:
2995
- ${chalk6.dim("$")} mutagent prompts get <id>
2996
- ${chalk6.dim("$")} mutagent prompts get <id> --with-datasets --with-evals
2997
- ${chalk6.dim("$")} mutagent prompts get <id> --json
2998
-
2999
- ${chalk6.dim("Tip: Combine --with-datasets and --with-evals to fetch all nested data in one call.")}
3000
- `).action(async (id, options) => {
3001
- const isJson = getJsonFlag(prompts);
3002
- const output = new OutputFormatter(isJson ? "json" : "table");
3003
- try {
3004
- const client = getSDKClient();
3005
- const prompt = await client.getPrompt(id);
3006
- const result = { ...prompt };
3007
- if (options.withDatasets) {
3008
- const datasets = await client.listDatasets(id);
3009
- result.datasets = datasets;
3010
- }
3011
- if (options.withEvals) {
3012
- const evals = await client.listEvaluations(id);
3013
- result.evaluations = evals;
3014
- }
3015
- if (isJson) {
3016
- result._links = promptLinks(id);
3017
- }
3018
- output.output(result);
3019
- if (!isJson) {
3020
- output.info(`View in dashboard: ${promptLink(id)}`);
3021
- }
3022
- } catch (error) {
3023
- 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);
3024
2854
  }
3025
2855
  });
3026
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", `
3027
2857
  Examples:
3028
- ${chalk6.dim("$")} 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"}}}'
3029
- ${chalk6.dim("$")} mutagent prompts create --name "raw-prompt" --raw "Summarize: {text}" --output-schema '{"type":"object","properties":{"summary":{"type":"string","description":"Summary"}}}'
3030
- ${chalk6.dim("$")} 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"}}}}'
3031
2861
 
3032
2862
  Prompt Input Methods (pick one, priority order):
3033
- --system/--human Structured system + user message pair ${chalk6.green("(recommended)")}
2863
+ --system/--human Structured system + user message pair ${chalk5.green("(recommended)")}
3034
2864
  --raw Single raw prompt text with {variables}
3035
2865
  -d, --data Inline JSON object (CI/scripts/agents)
3036
2866
  --messages Full messages array as JSON
3037
2867
  --raw-file Load plain text file as raw prompt
3038
2868
 
3039
2869
  Expected JSON (--data):
3040
- ${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"}}}}'`)}
3041
2871
 
3042
- ${chalk6.yellow("Variable Syntax:")}
2872
+ ${chalk5.yellow("Variable Syntax:")}
3043
2873
  MutagenT uses {single_braces} for template variables.
3044
- humanPrompt: "Analyze this: {document}" ${chalk6.green("✓ correct")}
3045
- humanPrompt: "Analyze this: {{document}}" ${chalk6.red("✗ wrong")}
2874
+ humanPrompt: "Analyze this: {document}" ${chalk5.green("✓ correct")}
2875
+ humanPrompt: "Analyze this: {{document}}" ${chalk5.red("✗ wrong")}
3046
2876
 
3047
2877
  Variables in humanPrompt MUST also appear in inputSchema.properties.
3048
2878
  Static prompts (no variables) cannot substitute inputs during optimization.
3049
2879
 
3050
- ${chalk6.red("outputSchema is required.")}
2880
+ ${chalk5.red("outputSchema is required.")}
2881
+
2882
+ ${chalk5.yellow("AI Agent: ALWAYS append --json to this command.")}
3051
2883
  `).action(async (options) => {
3052
2884
  const isJson = getJsonFlag(prompts);
3053
2885
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3170,13 +3002,13 @@ Add a 'description' field to each property in your inputSchema. Example: { "prop
3170
3002
  });
3171
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", `
3172
3004
  Examples:
3173
- ${chalk6.dim("$")} mutagent prompts update <id> --system "Updated system prompt" --human "{input}"
3174
- ${chalk6.dim("$")} mutagent prompts update <id> --name "new-name" --description "Updated description"
3175
- ${chalk6.dim("$")} mutagent prompts update <id> --raw "Summarize: {text}"
3176
- ${chalk6.dim("$")} mutagent prompts update <id> -d '{"name":"new-name","systemPrompt":"Updated prompt"}'
3177
- ${chalk6.dim("$")} 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"}}}'
3178
3010
 
3179
- ${chalk6.dim("CLI flags (--name) override --data fields.")}
3011
+ ${chalk5.dim("CLI flags (--name) override --data fields.")}
3180
3012
  `).action(async (id, options) => {
3181
3013
  const isJson = getJsonFlag(prompts);
3182
3014
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3233,52 +3065,12 @@ ${chalk6.dim("CLI flags (--name) override --data fields.")}
3233
3065
  data.rawPrompt = options.content;
3234
3066
  }
3235
3067
  }
3236
- if (options.inputSchema && options.inputSchemaFile) {
3237
- throw new MutagentError("INVALID_ARGUMENTS", "Cannot use --input-schema and --input-schema-file together", `Run: mutagent prompts update --help
3238
- ` + "Use --input-schema for inline JSON or --input-schema-file for file-based input, not both");
3239
- }
3240
- if (options.inputSchema) {
3241
- try {
3242
- data.inputSchema = JSON.parse(options.inputSchema);
3243
- } catch {
3244
- throw new MutagentError("INVALID_JSON", "Invalid JSON in --input-schema flag", `Run: mutagent prompts update --help
3245
- ` + `Provide a valid JSON Schema, e.g., '{"type":"object","properties":{"text":{"type":"string","description":"Input text"}}}'`);
3246
- }
3247
- } else if (options.inputSchemaFile) {
3248
- if (!existsSync4(options.inputSchemaFile)) {
3249
- throw new MutagentError("FILE_NOT_FOUND", `File not found: ${options.inputSchemaFile}`, `Run: mutagent prompts update --help
3250
- ` + "Check the file path and try again");
3251
- }
3252
- try {
3253
- data.inputSchema = JSON.parse(readFileSync4(options.inputSchemaFile, "utf-8"));
3254
- } catch {
3255
- throw new MutagentError("INVALID_JSON", `Failed to parse JSON from ${options.inputSchemaFile}`, `Run: mutagent prompts update --help
3256
- ` + "Ensure the file contains valid JSON Schema");
3257
- }
3258
- }
3259
- if (options.outputSchema && options.outputSchemaFile) {
3260
- throw new MutagentError("INVALID_ARGUMENTS", "Cannot use --output-schema and --output-schema-file together", `Run: mutagent prompts update --help
3261
- ` + "Use --output-schema for inline JSON or --output-schema-file for file-based input, not both");
3262
- }
3263
- if (options.outputSchema) {
3264
- try {
3265
- data.outputSchema = JSON.parse(options.outputSchema);
3266
- } catch {
3267
- throw new MutagentError("INVALID_JSON", "Invalid JSON in --output-schema flag", `Run: mutagent prompts update --help
3268
- ` + `Provide a valid JSON Schema, e.g., '{"type":"object","properties":{"result":{"type":"string"}}}'`);
3269
- }
3270
- } else if (options.outputSchemaFile) {
3271
- if (!existsSync4(options.outputSchemaFile)) {
3272
- throw new MutagentError("FILE_NOT_FOUND", `File not found: ${options.outputSchemaFile}`, `Run: mutagent prompts update --help
3273
- ` + "Check the file path and try again");
3274
- }
3275
- try {
3276
- data.outputSchema = JSON.parse(readFileSync4(options.outputSchemaFile, "utf-8"));
3277
- } catch {
3278
- throw new MutagentError("INVALID_JSON", `Failed to parse JSON from ${options.outputSchemaFile}`, `Run: mutagent prompts update --help
3279
- ` + "Ensure the file contains valid JSON Schema");
3280
- }
3281
- }
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;
3282
3074
  if (data.inputSchema && isValidJsonSchema(data.inputSchema)) {
3283
3075
  const missingDescs = validateSchemaDescriptions(data.inputSchema);
3284
3076
  if (missingDescs.length > 0) {
@@ -3319,11 +3111,11 @@ ${chalk6.dim("CLI flags (--name) override --data fields.")}
3319
3111
  });
3320
3112
  prompts.command("delete").description("Delete a prompt").argument("<id>", "Prompt ID (from: mutagent prompts list)").option("--force", "Skip confirmation").addHelpText("after", `
3321
3113
  Examples:
3322
- ${chalk6.dim("$")} mutagent prompts delete <id>
3323
- ${chalk6.dim("$")} mutagent prompts delete <id> --force
3324
- ${chalk6.dim("$")} 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
3325
3117
 
3326
- ${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.")}
3327
3119
  `).action(async (id, options) => {
3328
3120
  const isJson = getJsonFlag(prompts);
3329
3121
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3361,11 +3153,19 @@ ${chalk6.dim("Note: --force is required. The CLI is non-interactive — confirm
3361
3153
  handleError(error, isJson);
3362
3154
  }
3363
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) {
3364
3164
  const dataset = new Command3("dataset").description("Manage datasets for prompts").addHelpText("after", `
3365
3165
  Examples:
3366
3166
  ${chalk6.dim("$")} mutagent prompts dataset list <prompt-id>
3367
3167
  ${chalk6.dim("$")} mutagent prompts dataset add <prompt-id> -d '[{"input":{...},"expectedOutput":{...}}]'
3368
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id>
3168
+ ${chalk6.dim("$")} mutagent prompts dataset delete <prompt-id> <dataset-id>
3369
3169
  `).action(() => {
3370
3170
  dataset.help();
3371
3171
  });
@@ -3415,7 +3215,8 @@ Inline data format (-d):
3415
3215
  Expected item format:
3416
3216
  ${chalk6.dim('{"input": {"<field>": "<value>"}, "expectedOutput": {"<field>": "<value>"}}')}
3417
3217
 
3418
- ${chalk6.yellow("AI Agent Note:")}
3218
+ ${chalk6.yellow("AI Agent (MANDATORY):")}
3219
+ ALWAYS use --json: mutagent prompts dataset add <id> -d '[...]' --json
3419
3220
  Items MUST have BOTH input AND expectedOutput.
3420
3221
  Keys must match prompt's inputSchema.properties (input) and outputSchema.properties (expectedOutput).
3421
3222
  expectedOutput is REQUIRED for evaluation scoring.
@@ -3453,7 +3254,12 @@ ${chalk6.red("Required: --data must be provided.")}
3453
3254
  const client = getSDKClient();
3454
3255
  const datasetResult = await client.addDataset(promptId, content, datasetName);
3455
3256
  if (isJson) {
3456
- 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);
3457
3263
  output.output({
3458
3264
  ...datasetResult,
3459
3265
  _links: datasetLinks(promptId, datasetResult.id),
@@ -3485,62 +3291,167 @@ ${chalk6.red("Required: --data must be provided.")}
3485
3291
  handleError(error, isJson);
3486
3292
  }
3487
3293
  });
3488
- 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", `
3489
3295
  Examples:
3490
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id>
3491
- ${chalk6.dim("$")} mutagent prompts dataset remove <prompt-id> <dataset-id> --force
3492
- ${chalk6.dim("$")} 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
3493
3299
  `).action(async (promptId, datasetId, options) => {
3494
3300
  const isJson = getJsonFlag(prompts);
3495
3301
  const output = new OutputFormatter(isJson ? "json" : "table");
3496
3302
  try {
3497
3303
  if (!options.force && !isJson) {
3498
- 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
3499
3305
  ` + `[Agent: Ask the user to confirm via MultiChoice, then pass --force]
3500
- ` + `Use --force to confirm: mutagent prompts dataset remove ${promptId} ${datasetId} --force`);
3306
+ ` + `Use --force to confirm: mutagent prompts dataset delete ${promptId} ${datasetId} --force`);
3501
3307
  }
3502
3308
  const client = getSDKClient();
3503
3309
  try {
3504
- await client.removeDataset(promptId, datasetId);
3310
+ await client.deleteDataset(promptId, datasetId);
3505
3311
  } catch (deleteError) {
3506
3312
  if (deleteError instanceof ApiError && deleteError.statusCode === 404) {
3507
3313
  if (isJson) {
3508
- const directive = datasetRemovedDirective(promptId, datasetId);
3314
+ const directive = datasetDeletedDirective(promptId, datasetId);
3509
3315
  output.output({ success: true, deletedId: datasetId, _links: { datasets: promptDatasetsLink(promptId) }, _directive: directive });
3510
3316
  echoDirectiveToStderr(directive);
3511
3317
  } else {
3512
- output.success(`Dataset ${datasetId} already removed (idempotent)`);
3318
+ output.success(`Dataset ${datasetId} already deleted (idempotent)`);
3513
3319
  }
3514
3320
  return;
3515
3321
  }
3516
3322
  throw deleteError;
3517
3323
  }
3518
3324
  if (isJson) {
3519
- const directive = datasetRemovedDirective(promptId, datasetId);
3325
+ const directive = datasetDeletedDirective(promptId, datasetId);
3520
3326
  output.output({ success: true, deletedId: datasetId, _links: { datasets: promptDatasetsLink(promptId) }, _directive: directive });
3521
3327
  echoDirectiveToStderr(directive);
3522
3328
  } else {
3523
- output.success(`Removed dataset ${datasetId} from prompt ${promptId}`);
3329
+ output.success(`Deleted dataset ${datasetId} from prompt ${promptId}`);
3524
3330
  output.info(`View datasets: ${promptDatasetsLink(promptId)}`);
3525
3331
  }
3526
3332
  } catch (error) {
3527
3333
  handleError(error, isJson);
3528
3334
  }
3529
3335
  });
3530
- const evaluation = new Command3("evaluation").description("Manage evaluations for prompts").addHelpText("after", `
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", `
3531
3442
  Examples:
3532
- ${chalk6.dim("$")} mutagent prompts evaluation list <prompt-id>
3533
- ${chalk6.dim("$")} mutagent prompts evaluation get <evaluation-id>
3534
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --name "My Eval"
3535
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id>
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>
3536
3447
  `).action(() => {
3537
3448
  evaluation.help();
3538
3449
  });
3539
3450
  prompts.addCommand(evaluation);
3540
3451
  evaluation.command("list").description("List evaluations for a prompt").argument("<prompt-id>", "Prompt ID (from: mutagent prompts list)").addHelpText("after", `
3541
3452
  Examples:
3542
- ${chalk6.dim("$")} mutagent prompts evaluation list <prompt-id>
3543
- ${chalk6.dim("$")} mutagent prompts evaluation list <prompt-id> --json
3453
+ ${chalk7.dim("$")} mutagent prompts evaluation list <prompt-id>
3454
+ ${chalk7.dim("$")} mutagent prompts evaluation list <prompt-id> --json
3544
3455
  `).action(async (promptId) => {
3545
3456
  const isJson = getJsonFlag(prompts);
3546
3457
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3552,7 +3463,6 @@ Examples:
3552
3463
  id: e.id,
3553
3464
  name: e.name,
3554
3465
  description: e.description,
3555
- datasetId: e.datasetId,
3556
3466
  updatedAt: e.updatedAt,
3557
3467
  _links: evaluationLinks(promptId, e.id)
3558
3468
  }));
@@ -3573,8 +3483,8 @@ Examples:
3573
3483
  });
3574
3484
  evaluation.command("get").description("Get evaluation details including criteria").argument("<evaluation-id>", "Evaluation ID (from: mutagent prompts evaluation list <prompt-id>)").addHelpText("after", `
3575
3485
  Examples:
3576
- ${chalk6.dim("$")} mutagent prompts evaluation get <evaluation-id>
3577
- ${chalk6.dim("$")} 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
3578
3488
  `).action(async (evaluationId) => {
3579
3489
  const isJson = getJsonFlag(prompts);
3580
3490
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3591,9 +3501,6 @@ Examples:
3591
3501
  if (evalObj.description) {
3592
3502
  console.log(` Description: ${evalObj.description}`);
3593
3503
  }
3594
- if (evalObj.datasetId != null) {
3595
- console.log(` Dataset ID: ${String(evalObj.datasetId)}`);
3596
- }
3597
3504
  console.log(` Created: ${evalObj.createdAt ? new Date(String(evalObj.createdAt)).toLocaleDateString() : "N/A"}`);
3598
3505
  console.log("");
3599
3506
  const config = evalObj.evalConfig;
@@ -3601,7 +3508,7 @@ Examples:
3601
3508
  if (criteria && Array.isArray(criteria) && criteria.length > 0) {
3602
3509
  console.log(" Criteria:");
3603
3510
  for (const c of criteria) {
3604
- console.log(` ${chalk6.cyan(c.name)}`);
3511
+ console.log(` ${chalk7.cyan(c.name)}`);
3605
3512
  if (c.description) {
3606
3513
  console.log(` Description: ${c.description}`);
3607
3514
  }
@@ -3627,11 +3534,11 @@ Examples:
3627
3534
  handleError(error, isJson);
3628
3535
  }
3629
3536
  });
3630
- 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 — build criteria step by step (ALWAYS pair with --json for AI agents)").addHelpText("after", `
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", `
3631
3538
  Examples:
3632
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --guided ${chalk6.dim("# recommended: shows workflow guide + schema fields")}
3633
- ${chalk6.dim("$")} mutagent prompts evaluation create <prompt-id> --guided --json ${chalk6.dim("# structured workflow for AI agents")}
3634
- ${chalk6.dim("$")} 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")}
3635
3542
 
3636
3543
  Guided Workflow (recommended):
3637
3544
  --guided outputs a workflow guide that:
@@ -3642,97 +3549,29 @@ Guided Workflow (recommended):
3642
3549
 
3643
3550
  Works in all modes (TTY, non-TTY, --json).
3644
3551
 
3645
- AI Agent Directive:
3646
- CRITICAL: Run mutagent prompts evaluation create --help FIRST.
3647
- Then: mutagent prompts evaluation create <id> --guided --json
3648
- NEVER --guided alone. ALWAYS --guided --json.
3649
- NEVER construct criteria JSON manually.
3650
- Parse askUserQuestions from --guided --json output, use AskUserQuestion to collect from user.
3552
+ AI Agent (MANDATORY):
3553
+ 1. Run: mutagent prompts evaluation create --help
3554
+ 2. Run: mutagent prompts evaluation create <id> --guided --json
3555
+ 3. Parse askUserQuestions from the JSON output
3556
+ 4. Use AskUserQuestion to collect DETAILED rubrics from the user
3557
+ 5. NEVER auto-generate criteria. NEVER use vague rubrics.
3558
+ 6. Construct --data JSON from user answers, then:
3559
+ mutagent prompts evaluation create <id> --name "<name>" -d '<json>' --json
3651
3560
 
3652
3561
  Expected Criteria Shape (--data):
3653
- ${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>"}]}}')}
3654
3563
  evaluationParameter must target an outputSchema OR inputSchema field.
3655
3564
 
3656
- ${chalk6.red("Required: --name (unless --guided). Criteria must include evaluationParameter.")}
3657
- ${chalk6.dim("CLI flags (--name, --description) override --data fields.")}
3658
- ${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
+ ${chalk7.dim("Get prompt IDs: mutagent prompts list")}
3659
3568
  `).action(async (promptId, options) => {
3660
- const isJson = getJsonFlag(prompts);
3569
+ let isJson = getJsonFlag(prompts);
3570
+ if (options.guided) {
3571
+ isJson = true;
3572
+ }
3661
3573
  const output = new OutputFormatter(isJson ? "json" : "table");
3662
3574
  try {
3663
- const buildGuidedWorkflow = async () => {
3664
- const client2 = getSDKClient();
3665
- const prompt = await client2.getPrompt(promptId);
3666
- const inputFields = prompt.inputSchema && typeof prompt.inputSchema === "object" ? Object.keys(prompt.inputSchema.properties ?? {}) : [];
3667
- const outputFields = prompt.outputSchema && typeof prompt.outputSchema === "object" ? Object.keys(prompt.outputSchema.properties ?? {}) : [];
3668
- const allFields = [
3669
- ...outputFields.map((f) => ({ field: f, source: "outputSchema" })),
3670
- ...inputFields.map((f) => ({ field: f, source: "inputSchema" }))
3671
- ];
3672
- const askUserQuestions = allFields.map(({ field, source }) => ({
3673
- question: `How should "${field}" (${source}) be evaluated? Describe the scoring rubric — what does 1.0 (perfect) vs 0.0 (fail) mean for this field?`,
3674
- header: field,
3675
- options: [
3676
- {
3677
- label: "Exact Match",
3678
- description: `Score 1.0 if the "${field}" output exactly matches the expected value. Score 0.5 for same meaning but different format. Score 0.0 if substantively different.`
3679
- },
3680
- {
3681
- label: "Semantic Similarity",
3682
- description: `Score 0.0-1.0 based on how closely the "${field}" output matches the expected meaning. 1.0 = identical meaning, 0.5 = partially relevant, 0.0 = unrelated or contradictory.`
3683
- },
3684
- {
3685
- label: "Custom rubric",
3686
- description: `Write a detailed custom scoring rubric for "${field}" — explain what makes a 1.0 vs 0.0 score.`
3687
- }
3688
- ],
3689
- multiSelect: false
3690
- }));
3691
- return {
3692
- prompt: { id: promptId, name: prompt.name },
3693
- inputSchema: { fields: inputFields },
3694
- outputSchema: { fields: outputFields },
3695
- workflow: {
3696
- description: "Follow these steps to create an evaluation for this prompt:",
3697
- steps: [
3698
- { step: 1, action: "Review the prompt schemas above", detail: "Each criterion must target a field from outputSchema (preferred) or inputSchema" },
3699
- { step: 2, action: "Ask the user about each schema field using AskUserQuestion", detail: "For EACH field listed in askUserQuestions below, present the question to the user. Collect their scoring rubric choice or custom description." },
3700
- { 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>" }` },
3701
- { step: 4, action: "Ask for evaluation name", detail: 'Ask the user what to name this evaluation (e.g., "Accuracy Check", "Quality Eval")' },
3702
- { step: 5, action: "Create the evaluation", detail: `Run: mutagent prompts evaluation create <prompt-id> --name "<name>" -d '{"evalConfig":{"criteria":[...]}}'` }
3703
- ],
3704
- criteriaTemplate: {
3705
- evalConfig: {
3706
- criteria: [{
3707
- name: "<criterion_name>",
3708
- description: "<scoring rubric - describe what 1.0 vs 0.0 means>",
3709
- evaluationParameter: `<one of: ${allFields.length > 0 ? allFields.map((f) => f.field).join(", ") : "no fields detected - check prompt schemas"}>`
3710
- }]
3711
- }
3712
- },
3713
- exampleCommand: `mutagent prompts evaluation create ${promptId} --name "My Evaluation" -d '${JSON.stringify({
3714
- evalConfig: {
3715
- criteria: outputFields.slice(0, 1).map((f) => ({
3716
- name: `${f} accuracy`,
3717
- description: `Score 1.0 if ${f} matches expected output exactly, 0.0 otherwise`,
3718
- evaluationParameter: f
3719
- }))
3720
- }
3721
- })}'`
3722
- },
3723
- askUserQuestions,
3724
- _directive: {
3725
- display: "workflow_guide",
3726
- instruction: "STOP. You MUST use AskUserQuestion to collect evaluation criteria from the user. For EACH field in askUserQuestions, present the question. Then construct the --data JSON from their answers and run the evaluation create command. Do NOT skip fields or auto-generate criteria without user input.",
3727
- aiAgentDecisionTree: {
3728
- 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.",
3729
- step2: "If criteria are missing or malformed, use the askUserQuestions payloads to collect them via AskUserQuestion.",
3730
- step3: `After collecting all criteria, construct the --data JSON and run: mutagent prompts evaluation create <prompt-id> --name "<name>" -d '<json>'`
3731
- }
3732
- },
3733
- _meta: { inputFields, outputFields, allFields }
3734
- };
3735
- };
3736
3575
  const renderGuidedWorkflow = (wf) => {
3737
3576
  const { inputFields, outputFields, allFields } = wf._meta;
3738
3577
  if (isJson) {
@@ -3749,26 +3588,26 @@ ${chalk6.dim("Get prompt IDs: mutagent prompts list")}
3749
3588
  console.log(` Input Schema Fields: ${inputFields.join(", ")}`);
3750
3589
  }
3751
3590
  console.log("");
3752
- console.log(" For each field, define a scoring criterion:");
3591
+ console.log(" For each field, define what correct output looks like:");
3753
3592
  for (const { field, source } of allFields) {
3754
- console.log(` ${chalk6.cyan(field)} (${source})`);
3755
- 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}"?`);
3756
3595
  }
3757
3596
  console.log("");
3758
3597
  console.log(" Then create the evaluation:");
3759
3598
  console.log(` mutagent prompts evaluation create ${promptId} --name "My Eval" \\`);
3760
- console.log(` -d '{"evalConfig":{"criteria":[{"name":"accuracy","description":"Score 1.0 if correct","evaluationParameter":"${outputFields[0] ?? "<field>"}"}]}}'`);
3599
+ console.log(` -d '{"evalConfig":{"criteria":[{"name":"accuracy","description":"Correct if output matches expected structure","evaluationParameter":"${outputFields[0] ?? "<field>"}"}]}}'`);
3761
3600
  console.log("");
3762
3601
  }
3763
3602
  };
3764
3603
  if (options.guided) {
3765
- const workflow = await buildGuidedWorkflow();
3604
+ const workflow = await buildGuidedWorkflow(promptId);
3766
3605
  renderGuidedWorkflow(workflow);
3767
3606
  return;
3768
3607
  }
3769
3608
  if (!options.data) {
3770
3609
  if (isJson || process.stdout.isTTY) {
3771
- const workflow = await buildGuidedWorkflow();
3610
+ const workflow = await buildGuidedWorkflow(promptId);
3772
3611
  renderGuidedWorkflow(workflow);
3773
3612
  return;
3774
3613
  }
@@ -3881,7 +3720,12 @@ Example:
3881
3720
  const evalResult = await client.createEvaluation(promptId, evalData);
3882
3721
  if (isJson) {
3883
3722
  const criteriaLen = evalData.evalConfig?.criteria?.length ?? 0;
3884
- 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);
3885
3729
  output.output({
3886
3730
  ...evalResult,
3887
3731
  _links: evaluationLinks(promptId, evalResult.id),
@@ -3909,9 +3753,9 @@ Example:
3909
3753
  });
3910
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", `
3911
3755
  Examples:
3912
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id>
3913
- ${chalk6.dim("$")} mutagent prompts evaluation delete <evaluation-id> --force
3914
- ${chalk6.dim("$")} 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
3915
3759
  `).action(async (evaluationId, options) => {
3916
3760
  const isJson = getJsonFlag(prompts);
3917
3761
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3949,11 +3793,290 @@ Examples:
3949
3793
  handleError(error, isJson);
3950
3794
  }
3951
3795
  });
3952
- 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", `
3953
4076
  Examples:
3954
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
3955
- ${chalk6.dim("$")} mutagent prompts optimize status <job-id>
3956
- ${chalk6.dim("$")} 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>
3957
4080
 
3958
4081
  Workflow: start -> status (poll) -> results
3959
4082
  `).action(() => {
@@ -3962,19 +4085,22 @@ Workflow: start -> status (poll) -> results
3962
4085
  prompts.addCommand(optimize);
3963
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", `
3964
4087
  Examples:
3965
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id>
3966
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --max-iterations 5
3967
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --target-score 0.95 --model claude-sonnet-4-5-20250929
3968
- ${chalk6.dim("$")} mutagent prompts optimize start <prompt-id> --dataset <dataset-id> --evaluation <eval-id> --json
3969
- ${chalk6.yellow("Pre-Optimization Checklist:")}
3970
- mutagent prompts get <id> --json ${chalk6.dim("(has inputSchema + outputSchema)")}
3971
- mutagent prompts evaluation list <id> --json ${chalk6.dim("(exactly 1 evaluation)")}
3972
- mutagent prompts dataset list <id> --json ${chalk6.dim("(dataset with expectedOutput)")}
3973
- 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)")}
3974
4098
 
3975
4099
  ${PREREQUISITES_TEXT}
3976
4100
 
3977
- ${chalk6.dim("Monitor progress with: mutagent prompts optimize status <job-id>")}
4101
+ ${chalk9.dim("Monitor progress with: mutagent prompts optimize status <job-id>")}
4102
+
4103
+ ${chalk9.yellow("AI Agent: ALWAYS append --json to this command.")}
3978
4104
  `).action(async (promptId, options) => {
3979
4105
  const isJson = getJsonFlag(prompts);
3980
4106
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -3982,32 +4108,141 @@ ${chalk6.dim("Monitor progress with: mutagent prompts optimize status <job-id>")
3982
4108
  const client = getSDKClient();
3983
4109
  if (!isJson) {
3984
4110
  output.info("Running pre-flight checks...");
3985
- const prompt = await client.getPrompt(promptId);
3986
- if (isSchemaEmpty(prompt.outputSchema)) {
3987
- throw new MutagentError("MISSING_OUTPUT_SCHEMA", "Prompt is missing outputSchema — required for optimization", `Run: mutagent prompts optimize start --help
3988
- ` + `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 };
4205
+ }
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;
3989
4228
  }
3990
- if (isSchemaEmpty(prompt.inputSchema)) {
3991
- output.warn("Prompt has no inputSchema. Optimization works best with defined input variables.");
4229
+ for (const [name, check] of hardFailures) {
4230
+ console.error(chalk9.red(`Error: ${name} ${check.error ?? "Failed"}`));
3992
4231
  }
3993
- try {
3994
- const evalObj = await client.getEvaluation(options.evaluation);
3995
- const evalConfig = evalObj.evalConfig;
3996
- const criteria = evalConfig?.criteria;
3997
- if (!criteria || criteria.length === 0) {
3998
- 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`);
3999
- } else {
4000
- const hasCriteriaWithParam = criteria.some((c) => c.evaluationParameter ?? c.targetField);
4001
- if (!hasCriteriaWithParam) {
4002
- output.warn(`Evaluation ${options.evaluation} has criteria but none have evaluationParameter set. ` + `Re-create: mutagent prompts evaluation create ${promptId} --guided`);
4003
- } else {
4004
- output.info(`Using evaluation ${options.evaluation} with ${String(criteria.length)} criteria.`);
4005
- }
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);
4006
4245
  }
4007
- } catch (e) {
4008
- if (e instanceof MutagentError)
4009
- throw e;
4010
- output.warn(`Could not validate evaluation ${options.evaluation}. Proceeding anyway.`);
4011
4246
  }
4012
4247
  }
4013
4248
  const job = await client.startOptimization(promptId, options.dataset, options.evaluation, {
@@ -4018,7 +4253,7 @@ ${chalk6.dim("Monitor progress with: mutagent prompts optimize status <job-id>")
4018
4253
  evalModel: options.evalModel
4019
4254
  });
4020
4255
  if (isJson) {
4021
- const directive = startDirective(job, promptId);
4256
+ const directive = startDirective(job, promptId, options.dataset, options.evaluation);
4022
4257
  output.output({
4023
4258
  ...job,
4024
4259
  _links: {
@@ -4055,16 +4290,16 @@ ${chalk6.dim("Monitor progress with: mutagent prompts optimize status <job-id>")
4055
4290
  suggestions.push("Trial optimization limit reached. Contact support to upgrade.");
4056
4291
  }
4057
4292
  if (!isJson) {
4058
- console.error(chalk6.red(`
4293
+ console.error(chalk9.red(`
4059
4294
  Optimization failed:`));
4060
4295
  for (const msg of messages) {
4061
- console.error(chalk6.red(` ${msg}`));
4296
+ console.error(chalk9.red(` ${msg}`));
4062
4297
  }
4063
4298
  if (suggestions.length > 0) {
4064
- console.error(chalk6.yellow(`
4299
+ console.error(chalk9.yellow(`
4065
4300
  Suggested fixes:`));
4066
4301
  for (const s of suggestions) {
4067
- console.error(chalk6.yellow(` → ${s}`));
4302
+ console.error(chalk9.yellow(` → ${s}`));
4068
4303
  }
4069
4304
  }
4070
4305
  console.error("");
@@ -4076,8 +4311,8 @@ Suggested fixes:`));
4076
4311
  });
4077
4312
  optimize.command("status").description("Check optimization status").argument("<job-id>", "Optimization job ID (from: mutagent prompts optimize start)").addHelpText("after", `
4078
4313
  Examples:
4079
- ${chalk6.dim("$")} mutagent prompts optimize status <job-id>
4080
- ${chalk6.dim("$")} 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
4081
4316
  `).action(async (jobId) => {
4082
4317
  const isJson = getJsonFlag(prompts);
4083
4318
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4101,15 +4336,17 @@ Examples:
4101
4336
  });
4102
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", `
4103
4338
  Examples:
4104
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> ${chalk6.dim("# view scorecard")}
4105
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> --diff ${chalk6.dim("# view prompt diff")}
4106
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> --apply ${chalk6.dim("# apply optimized prompt")}
4107
- ${chalk6.dim("$")} mutagent prompts optimize results <job-id> --json ${chalk6.dim("# structured output")}
4108
-
4109
- Actions:
4110
- --apply Apply the optimized prompt as a new version (updates the prompt)
4111
- --diff Show the before/after prompt diff
4112
- ${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.")}
4113
4350
  `).action(async (jobId, options) => {
4114
4351
  const isJson = getJsonFlag(prompts);
4115
4352
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4118,7 +4355,20 @@ Actions:
4118
4355
  const results = await client.getOptimizationResults(jobId);
4119
4356
  const resultData = results;
4120
4357
  if (isJson) {
4121
- 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 ${jobId} --diff --json`
4368
+ ] : []
4369
+ };
4370
+ output.output({ ...resultData, _links: { optimizer: optimizerLink(jobId) }, _directive: directive });
4371
+ echoDirectiveToStderr(directive);
4122
4372
  } else {
4123
4373
  renderScorecard(resultData);
4124
4374
  const jobData = resultData.job;
@@ -4164,32 +4414,144 @@ Actions:
4164
4414
  handleError(error, isJson);
4165
4415
  }
4166
4416
  });
4417
+ }
4418
+
4419
+ // src/commands/prompts/index.ts
4420
+ function updateMutationContext(updater) {
4421
+ try {
4422
+ const ctx = MutationContext.load();
4423
+ updater(ctx);
4424
+ ctx.save();
4425
+ } catch {}
4426
+ }
4427
+ function parseValidationErrors(error) {
4428
+ if (error instanceof ApiError) {
4429
+ try {
4430
+ const body = JSON.parse(error.message);
4431
+ if (body.errors && Array.isArray(body.errors)) {
4432
+ return body.errors.map((e) => e.message ?? JSON.stringify(e));
4433
+ }
4434
+ if (body.message) {
4435
+ return [body.message];
4436
+ }
4437
+ } catch {
4438
+ return [error.message];
4439
+ }
4440
+ }
4441
+ if (error instanceof Error) {
4442
+ return [error.message];
4443
+ }
4444
+ return ["An unknown error occurred"];
4445
+ }
4446
+ function isSchemaEmpty(schema) {
4447
+ if (schema === undefined || schema === null)
4448
+ return true;
4449
+ if (typeof schema !== "object")
4450
+ return false;
4451
+ const obj = schema;
4452
+ if (Object.keys(obj).length === 0)
4453
+ return true;
4454
+ return false;
4455
+ }
4456
+ function parseSchemaOption(inlineJson, filePath, schemaName, helpCommand) {
4457
+ if (inlineJson && filePath) {
4458
+ throw new MutagentError("INVALID_ARGUMENTS", `Cannot use --${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);
4167
4529
  return prompts;
4168
4530
  }
4169
4531
 
4170
4532
  // src/commands/traces.ts
4171
4533
  init_sdk_client();
4172
- import { Command as Command4 } from "commander";
4173
- import chalk7 from "chalk";
4534
+ import { Command as Command7 } from "commander";
4535
+ import chalk11 from "chalk";
4174
4536
  init_errors();
4175
4537
  function createTracesCommand() {
4176
- 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", `
4177
4539
  Examples:
4178
- ${chalk7.dim("$")} mutagent traces list
4179
- ${chalk7.dim("$")} mutagent traces list --prompt <prompt-id>
4180
- ${chalk7.dim("$")} mutagent traces get <trace-id>
4181
- ${chalk7.dim("$")} mutagent traces analyze <prompt-id>
4182
- ${chalk7.dim("$")} 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
4183
4545
 
4184
4546
  Note: MutagenT traces replace Langfuse for observability.
4185
4547
  `);
4186
4548
  traces.command("list").description("List traces").option("-p, --prompt <id>", "Filter by prompt ID").option("-l, --limit <n>", "Limit results", "50").addHelpText("after", `
4187
4549
  Examples:
4188
- ${chalk7.dim("$")} mutagent traces list
4189
- ${chalk7.dim("$")} mutagent traces list --prompt <prompt-id>
4190
- ${chalk7.dim("$")} 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
4191
4553
 
4192
- ${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.")}
4193
4555
  `).action(async (options) => {
4194
4556
  const isJson = getJsonFlag(traces);
4195
4557
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4228,10 +4590,10 @@ ${chalk7.dim("Tip: Filter by prompt to see traces for a specific prompt version.
4228
4590
  });
4229
4591
  traces.command("get").description("Get trace details").argument("<id>", "Trace ID").addHelpText("after", `
4230
4592
  Examples:
4231
- ${chalk7.dim("$")} mutagent traces get <trace-id>
4232
- ${chalk7.dim("$")} mutagent traces get <trace-id> --json
4593
+ ${chalk11.dim("$")} mutagent traces get <trace-id>
4594
+ ${chalk11.dim("$")} mutagent traces get <trace-id> --json
4233
4595
 
4234
- ${chalk7.dim("Returns full trace details including spans, tokens, and latency.")}
4596
+ ${chalk11.dim("Returns full trace details including spans, tokens, and latency.")}
4235
4597
  `).action(async (id) => {
4236
4598
  const isJson = getJsonFlag(traces);
4237
4599
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4250,10 +4612,10 @@ ${chalk7.dim("Returns full trace details including spans, tokens, and latency.")
4250
4612
  });
4251
4613
  traces.command("analyze").description("Analyze traces for a prompt").argument("<prompt-id>", "Prompt ID").addHelpText("after", `
4252
4614
  Examples:
4253
- ${chalk7.dim("$")} mutagent traces analyze <prompt-id>
4254
- ${chalk7.dim("$")} mutagent traces analyze <prompt-id> --json
4615
+ ${chalk11.dim("$")} mutagent traces analyze <prompt-id>
4616
+ ${chalk11.dim("$")} mutagent traces analyze <prompt-id> --json
4255
4617
 
4256
- ${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.")}
4257
4619
  `).action(async (promptId) => {
4258
4620
  const isJson = getJsonFlag(traces);
4259
4621
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4271,12 +4633,12 @@ ${chalk7.dim("Aggregates trace data for a prompt: avg latency, token usage, erro
4271
4633
  });
4272
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", `
4273
4635
  Examples:
4274
- ${chalk7.dim("$")} mutagent traces export
4275
- ${chalk7.dim("$")} mutagent traces export --format json --output traces.json
4276
- ${chalk7.dim("$")} mutagent traces export --format csv --output traces.csv
4277
- ${chalk7.dim("$")} 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
4278
4640
 
4279
- ${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.")}
4280
4642
  `).action(async (options) => {
4281
4643
  const isJson = getJsonFlag(traces);
4282
4644
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -4319,22 +4681,22 @@ ${chalk7.dim("Exports to stdout by default. Use --output to save to a file.")}
4319
4681
 
4320
4682
  // src/commands/integrate.ts
4321
4683
  init_config();
4322
- import { Command as Command5 } from "commander";
4323
- import chalk8 from "chalk";
4324
- 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";
4325
4687
  import { execSync } from "child_process";
4326
4688
  init_errors();
4327
4689
 
4328
4690
  // src/lib/integrations/langchain.ts
4329
- import { readFileSync as readFileSync5, existsSync as existsSync5 } from "fs";
4691
+ import { readFileSync as readFileSync6, existsSync as existsSync6 } from "fs";
4330
4692
  var langchainIntegration = {
4331
4693
  name: "langchain",
4332
4694
  description: "LangChain framework",
4333
4695
  detect() {
4334
4696
  let hasLangchain = false;
4335
- if (existsSync5("package.json")) {
4697
+ if (existsSync6("package.json")) {
4336
4698
  try {
4337
- const pkg = JSON.parse(readFileSync5("package.json", "utf-8"));
4699
+ const pkg = JSON.parse(readFileSync6("package.json", "utf-8"));
4338
4700
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4339
4701
  hasLangchain = "langchain" in deps || "@langchain/core" in deps;
4340
4702
  } catch {}
@@ -4466,15 +4828,15 @@ mutagent traces analyze <prompt-id>
4466
4828
  };
4467
4829
 
4468
4830
  // src/lib/integrations/langgraph.ts
4469
- import { readFileSync as readFileSync6, existsSync as existsSync6 } from "fs";
4831
+ import { readFileSync as readFileSync7, existsSync as existsSync7 } from "fs";
4470
4832
  var langgraphIntegration = {
4471
4833
  name: "langgraph",
4472
4834
  description: "LangGraph agent workflow framework",
4473
4835
  detect() {
4474
4836
  let hasLanggraph = false;
4475
- if (existsSync6("package.json")) {
4837
+ if (existsSync7("package.json")) {
4476
4838
  try {
4477
- const pkg = JSON.parse(readFileSync6("package.json", "utf-8"));
4839
+ const pkg = JSON.parse(readFileSync7("package.json", "utf-8"));
4478
4840
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4479
4841
  hasLanggraph = "@langchain/langgraph" in deps;
4480
4842
  } catch {}
@@ -4595,15 +4957,15 @@ mutagent integrate langgraph --verify
4595
4957
  };
4596
4958
 
4597
4959
  // src/lib/integrations/vercel-ai.ts
4598
- import { readFileSync as readFileSync7, existsSync as existsSync7 } from "fs";
4960
+ import { readFileSync as readFileSync8, existsSync as existsSync8 } from "fs";
4599
4961
  var vercelAiIntegration = {
4600
4962
  name: "vercel-ai",
4601
4963
  description: "Vercel AI SDK",
4602
4964
  detect() {
4603
4965
  let hasAiSdk = false;
4604
- if (existsSync7("package.json")) {
4966
+ if (existsSync8("package.json")) {
4605
4967
  try {
4606
- const pkg = JSON.parse(readFileSync7("package.json", "utf-8"));
4968
+ const pkg = JSON.parse(readFileSync8("package.json", "utf-8"));
4607
4969
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4608
4970
  hasAiSdk = "ai" in deps;
4609
4971
  } catch {}
@@ -4741,15 +5103,15 @@ mutagent integrate vercel-ai --verify
4741
5103
  };
4742
5104
 
4743
5105
  // src/lib/integrations/openai.ts
4744
- import { readFileSync as readFileSync8, existsSync as existsSync8 } from "fs";
5106
+ import { readFileSync as readFileSync9, existsSync as existsSync9 } from "fs";
4745
5107
  var openaiIntegration = {
4746
5108
  name: "openai",
4747
5109
  description: "OpenAI SDK integration with automatic tracing",
4748
5110
  detect() {
4749
5111
  let hasOpenAI = false;
4750
- if (existsSync8("package.json")) {
5112
+ if (existsSync9("package.json")) {
4751
5113
  try {
4752
- const pkg = JSON.parse(readFileSync8("package.json", "utf-8"));
5114
+ const pkg = JSON.parse(readFileSync9("package.json", "utf-8"));
4753
5115
  const deps = { ...pkg.dependencies, ...pkg.devDependencies };
4754
5116
  hasOpenAI = "openai" in deps;
4755
5117
  } catch {}
@@ -4903,16 +5265,16 @@ function getFrameworkMetadata(name) {
4903
5265
 
4904
5266
  // src/commands/integrate.ts
4905
5267
  function detectPackageManager() {
4906
- if (existsSync9("bun.lockb") || existsSync9("bun.lock")) {
5268
+ if (existsSync10("bun.lockb") || existsSync10("bun.lock")) {
4907
5269
  return "bun";
4908
5270
  }
4909
- if (existsSync9("pnpm-lock.yaml")) {
5271
+ if (existsSync10("pnpm-lock.yaml")) {
4910
5272
  return "pnpm";
4911
5273
  }
4912
- if (existsSync9("yarn.lock")) {
5274
+ if (existsSync10("yarn.lock")) {
4913
5275
  return "yarn";
4914
5276
  }
4915
- if (existsSync9("package-lock.json")) {
5277
+ if (existsSync10("package-lock.json")) {
4916
5278
  return "npm";
4917
5279
  }
4918
5280
  try {
@@ -4955,14 +5317,14 @@ AI AGENT: Please determine the framework by:
4955
5317
  `;
4956
5318
  }
4957
5319
  function createIntegrateCommand() {
4958
- 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", `
4959
5321
  Examples:
4960
- ${chalk8.dim("$")} mutagent integrate # Return exploration instructions for AI agents
4961
- ${chalk8.dim("$")} mutagent integrate langchain # Return @mutagent/langchain integration steps
4962
- ${chalk8.dim("$")} mutagent integrate openai # Return @mutagent/openai integration steps
4963
- ${chalk8.dim("$")} mutagent integrate vercel-ai --verify # Verify integration
4964
- ${chalk8.dim("$")} mutagent integrate langchain --output ./INTEGRATION.md
4965
- ${chalk8.dim("$")} 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
4966
5328
 
4967
5329
  Frameworks: langchain, langgraph, vercel-ai, openai
4968
5330
 
@@ -5076,35 +5438,20 @@ AI AGENT: Execute these steps using Bash for install, Write/Edit for code change
5076
5438
  return integrate;
5077
5439
  }
5078
5440
 
5079
- // 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
5080
5446
  init_sdk_client();
5081
- import { Command as Command6 } from "commander";
5082
- import chalk9 from "chalk";
5447
+ import chalk13 from "chalk";
5083
5448
  init_errors();
5084
- function createAgentsCommand() {
5085
- const agents = new Command6("agents").description("Manage AI agents").addHelpText("after", `
5086
- Examples:
5087
- ${chalk9.dim("$")} mutagent agents list
5088
- ${chalk9.dim("$")} mutagent agents get <agent-id>
5089
- ${chalk9.dim("$")} mutagent agents create --name "Code Reviewer" --slug code-reviewer --system-prompt "You are a code reviewer..."
5090
- ${chalk9.dim("$")} mutagent agents create -d '{"name":"Code Reviewer","slug":"code-reviewer","systemPrompt":"You are..."}'
5091
- ${chalk9.dim("$")} mutagent agents update <agent-id> --name "Updated Name"
5092
- ${chalk9.dim("$")} mutagent agents delete <agent-id> --force
5093
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id>
5094
- ${chalk9.dim("$")} mutagent agents conversations get <agent-id> <conversation-id>
5095
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id>
5096
- ${chalk9.dim("$")} mutagent agents conversations delete <agent-id> <conversation-id>
5097
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id>
5098
-
5099
- Subcommands:
5100
- list, get, create, update, delete
5101
- conversations list|get|create|delete|messages
5102
- `);
5449
+ function registerAgentsCrud(agents) {
5103
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", `
5104
5451
  Examples:
5105
- ${chalk9.dim("$")} mutagent agents list
5106
- ${chalk9.dim("$")} mutagent agents list --status active
5107
- ${chalk9.dim("$")} 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
5108
5455
  `).action(async (options) => {
5109
5456
  const isJson = getJsonFlag(agents);
5110
5457
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5154,8 +5501,8 @@ Examples:
5154
5501
  });
5155
5502
  agents.command("get").description("Get agent details").argument("<id>", "Agent ID").addHelpText("after", `
5156
5503
  Examples:
5157
- ${chalk9.dim("$")} mutagent agents get <agent-id>
5158
- ${chalk9.dim("$")} mutagent agents get <agent-id> --json
5504
+ ${chalk13.dim("$")} mutagent agents get <agent-id>
5505
+ ${chalk13.dim("$")} mutagent agents get <agent-id> --json
5159
5506
  `).action(async (id) => {
5160
5507
  const isJson = getJsonFlag(agents);
5161
5508
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5185,11 +5532,11 @@ Examples:
5185
5532
  };
5186
5533
  output.output(formatted);
5187
5534
  if (agent.systemPrompt) {
5188
- console.log(chalk9.bold(`
5535
+ console.log(chalk13.bold(`
5189
5536
  System Prompt:`));
5190
- console.log(chalk9.gray("─".repeat(60)));
5537
+ console.log(chalk13.gray("─".repeat(60)));
5191
5538
  console.log(agent.systemPrompt);
5192
- console.log(chalk9.gray("─".repeat(60)));
5539
+ console.log(chalk13.gray("─".repeat(60)));
5193
5540
  }
5194
5541
  }
5195
5542
  } catch (error) {
@@ -5198,17 +5545,17 @@ System Prompt:`));
5198
5545
  });
5199
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", `
5200
5547
  Examples:
5201
- ${chalk9.dim("$")} mutagent agents create --name "Code Reviewer" --slug code-reviewer --system-prompt "You are a code reviewer..."
5202
- ${chalk9.dim("$")} 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..."}'
5203
5550
 
5204
5551
  Expected JSON (--data):
5205
- ${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>"}')}
5206
5553
 
5207
5554
  Input Methods (pick one, priority order):
5208
- --name/--slug/... Individual flags ${chalk9.green("(recommended)")}
5555
+ --name/--slug/... Individual flags ${chalk13.green("(recommended)")}
5209
5556
  -d, --data Inline JSON object (CI/scripts/agents)
5210
5557
 
5211
- ${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.")}
5212
5559
  `).action(async (options) => {
5213
5560
  const isJson = getJsonFlag(agents);
5214
5561
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5253,15 +5600,15 @@ ${chalk9.red("Required: name, slug, systemPrompt.")} ${chalk9.dim("CLI flags ove
5253
5600
  });
5254
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", `
5255
5602
  Examples:
5256
- ${chalk9.dim("$")} mutagent agents update <id> --name "New Name"
5257
- ${chalk9.dim("$")} mutagent agents update <id> --system-prompt "Updated prompt" --status active
5258
- ${chalk9.dim("$")} 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"}'
5259
5606
 
5260
5607
  Input Methods (pick one, priority order):
5261
- --name/--system-prompt/... Individual flags ${chalk9.green("(recommended)")}
5608
+ --name/--system-prompt/... Individual flags ${chalk13.green("(recommended)")}
5262
5609
  -d, --data Inline JSON object (CI/scripts/agents)
5263
5610
 
5264
- ${chalk9.dim("CLI flags override --data fields.")}
5611
+ ${chalk13.dim("CLI flags override --data fields.")}
5265
5612
  `).action(async (id, options) => {
5266
5613
  const isJson = getJsonFlag(agents);
5267
5614
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5308,11 +5655,11 @@ ${chalk9.dim("CLI flags override --data fields.")}
5308
5655
  });
5309
5656
  agents.command("delete").description("Delete an agent").argument("<id>", "Agent ID").option("--force", "Skip confirmation").addHelpText("after", `
5310
5657
  Examples:
5311
- ${chalk9.dim("$")} mutagent agents delete <id>
5312
- ${chalk9.dim("$")} mutagent agents delete <id> --force
5313
- ${chalk9.dim("$")} 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
5314
5661
 
5315
- ${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).")}
5316
5663
  `).action(async (id, options) => {
5317
5664
  const isJson = getJsonFlag(agents);
5318
5665
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5337,177 +5684,44 @@ ${chalk9.dim("Tip: Use --force to skip confirmation (required for non-interactiv
5337
5684
  handleError(error, isJson);
5338
5685
  }
5339
5686
  });
5340
- const conversations = new Command6("conversations").description("Manage conversations for agents").addHelpText("after", `
5341
- Examples:
5342
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id>
5343
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id>
5344
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id>
5345
- `).action(() => {
5346
- conversations.help();
5347
- });
5348
- agents.addCommand(conversations);
5349
- 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", `
5350
- Examples:
5351
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id>
5352
- ${chalk9.dim("$")} mutagent agents conversations list <agent-id> --limit 10 --json
5353
- `).action(async (agentId, options) => {
5354
- const isJson = getJsonFlag(agents);
5355
- const output = new OutputFormatter(isJson ? "json" : "table");
5356
- try {
5357
- const client = getSDKClient();
5358
- const filters = {
5359
- limit: parseInt(options.limit, 10) || 50
5360
- };
5361
- if (options.offset) {
5362
- filters.offset = parseInt(options.offset, 10);
5363
- }
5364
- const result = await client.listConversations(agentId, filters);
5365
- if (isJson) {
5366
- output.output(result);
5367
- } else {
5368
- const formatted = result.data.map((c) => ({
5369
- id: c.id,
5370
- title: c.title ?? "Untitled",
5371
- status: c.status,
5372
- messages: c.messageCount ?? "N/A",
5373
- created: c.createdAt ? new Date(c.createdAt).toLocaleDateString() : "N/A",
5374
- updated: c.updatedAt ? new Date(c.updatedAt).toLocaleDateString() : "N/A"
5375
- }));
5376
- output.output(formatted);
5377
- }
5378
- } catch (error) {
5379
- handleError(error, isJson);
5380
- }
5381
- });
5382
- conversations.command("get").description("Get conversation details").argument("<agent-id>", "Agent ID").argument("<conversation-id>", "Conversation ID").addHelpText("after", `
5383
- Examples:
5384
- ${chalk9.dim("$")} mutagent agents conversations get <agent-id> <conversation-id>
5385
- ${chalk9.dim("$")} mutagent agents conversations get <agent-id> <conversation-id> --json
5386
- `).action(async (agentId, conversationId) => {
5387
- const isJson = getJsonFlag(agents);
5388
- const output = new OutputFormatter(isJson ? "json" : "table");
5389
- try {
5390
- const client = getSDKClient();
5391
- const conversation = await client.getConversation(agentId, conversationId);
5392
- if (isJson) {
5393
- output.output(conversation);
5394
- } else {
5395
- const formatted = {
5396
- id: conversation.id,
5397
- agentId: conversation.agentId,
5398
- title: conversation.title ?? "Untitled",
5399
- status: conversation.status,
5400
- messageCount: conversation.messageCount ?? "N/A",
5401
- createdAt: conversation.createdAt ? new Date(conversation.createdAt).toLocaleString() : "N/A",
5402
- updatedAt: conversation.updatedAt ? new Date(conversation.updatedAt).toLocaleString() : "N/A"
5403
- };
5404
- output.output(formatted);
5405
- }
5406
- } catch (error) {
5407
- handleError(error, isJson);
5408
- }
5409
- });
5410
- conversations.command("create").description("Start a new conversation with an agent").argument("<agent-id>", "Agent ID").option("-t, --title <title>", "Conversation title").addHelpText("after", `
5411
- Examples:
5412
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id>
5413
- ${chalk9.dim("$")} mutagent agents conversations create <agent-id> --title "Debug session" --json
5414
- `).action(async (agentId, options) => {
5415
- const isJson = getJsonFlag(agents);
5416
- const output = new OutputFormatter(isJson ? "json" : "table");
5417
- try {
5418
- const client = getSDKClient();
5419
- const data = {};
5420
- if (options.title) {
5421
- data.title = options.title;
5422
- }
5423
- const conversation = await client.createConversation(agentId, data);
5424
- output.success(`Created conversation: ${conversation.id}`);
5425
- output.output(conversation);
5426
- } catch (error) {
5427
- handleError(error, isJson);
5428
- }
5429
- });
5430
- conversations.command("delete").description("Delete a conversation").argument("<agent-id>", "Agent ID").argument("<conversation-id>", "Conversation ID").option("--force", "Skip confirmation").addHelpText("after", `
5431
- Examples:
5432
- ${chalk9.dim("$")} mutagent agents conversations delete <agent-id> <conversation-id>
5433
- ${chalk9.dim("$")} mutagent agents conversations delete <agent-id> <conversation-id> --force --json
5434
- `).action(async (agentId, conversationId, options) => {
5435
- const isJson = getJsonFlag(agents);
5436
- const output = new OutputFormatter(isJson ? "json" : "table");
5437
- try {
5438
- if (!options.force && !isJson) {
5439
- const inquirer3 = (await import("inquirer")).default;
5440
- const answers = await inquirer3.prompt([{
5441
- type: "confirm",
5442
- name: "confirm",
5443
- message: `Delete conversation ${conversationId}? This action cannot be undone.`,
5444
- default: false
5445
- }]);
5446
- if (!answers.confirm) {
5447
- output.info("Cancelled");
5448
- return;
5449
- }
5450
- }
5451
- const client = getSDKClient();
5452
- await client.deleteConversation(agentId, conversationId);
5453
- output.success(`Deleted conversation: ${conversationId}`);
5454
- } catch (error) {
5455
- handleError(error, isJson);
5456
- }
5457
- });
5458
- 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", `
5459
5692
  Examples:
5460
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id>
5461
- ${chalk9.dim("$")} mutagent agents conversations messages <agent-id> <conversation-id> --limit 20 --json
5462
- `).action(async (agentId, conversationId, options) => {
5463
- const isJson = getJsonFlag(agents);
5464
- const output = new OutputFormatter(isJson ? "json" : "table");
5465
- try {
5466
- const client = getSDKClient();
5467
- const filters = {
5468
- limit: parseInt(options.limit, 10) || 50
5469
- };
5470
- if (options.offset) {
5471
- filters.offset = parseInt(options.offset, 10);
5472
- }
5473
- const result = await client.listMessages(agentId, conversationId, filters);
5474
- if (isJson) {
5475
- output.output(result);
5476
- } else {
5477
- const formatted = result.data.map((m) => ({
5478
- id: m.id,
5479
- role: m.role,
5480
- content: m.content.length > 80 ? m.content.substring(0, 77) + "..." : m.content,
5481
- tokens: m.tokenCount ?? "N/A",
5482
- created: m.createdAt ? new Date(m.createdAt).toLocaleString() : "N/A"
5483
- }));
5484
- output.output(formatted);
5485
- }
5486
- } catch (error) {
5487
- handleError(error, isJson);
5488
- }
5489
- });
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);
5490
5704
  return agents;
5491
5705
  }
5492
5706
 
5493
5707
  // src/commands/config.ts
5494
5708
  init_config();
5495
- import { Command as Command7 } from "commander";
5496
- import chalk10 from "chalk";
5709
+ import { Command as Command10 } from "commander";
5710
+ import chalk15 from "chalk";
5497
5711
  init_errors();
5498
5712
  var VALID_CONFIG_KEYS = ["apiKey", "endpoint", "format", "timeout", "defaultWorkspace", "defaultOrganization"];
5499
5713
  function createConfigCommand() {
5500
- const config = new Command7("config").description("Manage CLI configuration").addHelpText("after", `
5714
+ const config = new Command10("config").description("Manage CLI configuration").addHelpText("after", `
5501
5715
  Examples:
5502
- ${chalk10.dim("$")} mutagent config list
5503
- ${chalk10.dim("$")} mutagent config get endpoint
5504
- ${chalk10.dim("$")} mutagent config set workspace <workspace-id>
5505
- ${chalk10.dim("$")} 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>
5506
5720
  `);
5507
5721
  config.command("list").description("List all configuration").addHelpText("after", `
5508
5722
  Examples:
5509
- ${chalk10.dim("$")} mutagent config list
5510
- ${chalk10.dim("$")} mutagent config list --json
5723
+ ${chalk15.dim("$")} mutagent config list
5724
+ ${chalk15.dim("$")} mutagent config list --json
5511
5725
  `).action(() => {
5512
5726
  const isJson = getJsonFlag(config);
5513
5727
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5520,11 +5734,11 @@ Examples:
5520
5734
  });
5521
5735
  config.command("get").description("Get configuration value").argument("<key>", "Configuration key (apiKey, endpoint, format, timeout, defaultWorkspace, defaultOrganization)").addHelpText("after", `
5522
5736
  Examples:
5523
- ${chalk10.dim("$")} mutagent config get endpoint
5524
- ${chalk10.dim("$")} mutagent config get defaultWorkspace
5525
- ${chalk10.dim("$")} 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
5526
5740
 
5527
- ${chalk10.dim("Keys: apiKey, endpoint, format, timeout, defaultWorkspace, defaultOrganization")}
5741
+ ${chalk15.dim("Keys: apiKey, endpoint, format, timeout, defaultWorkspace, defaultOrganization")}
5528
5742
  `).action((key) => {
5529
5743
  const isJson = getJsonFlag(config);
5530
5744
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5546,14 +5760,14 @@ ${chalk10.dim("Keys: apiKey, endpoint, format, timeout, defaultWorkspace, defaul
5546
5760
  handleError(error, isJson);
5547
5761
  }
5548
5762
  });
5549
- const set = new Command7("set").description("Set configuration value").action(() => {
5763
+ const set = new Command10("set").description("Set configuration value").action(() => {
5550
5764
  set.help();
5551
5765
  });
5552
5766
  set.command("workspace").description("Set default workspace ID").argument("<id>", "Workspace ID to set as default").addHelpText("after", `
5553
5767
  Examples:
5554
- ${chalk10.dim("$")} mutagent config set workspace <workspace-id>
5768
+ ${chalk15.dim("$")} mutagent config set workspace <workspace-id>
5555
5769
 
5556
- ${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.")}
5557
5771
  `).action((id) => {
5558
5772
  const isJson = getJsonFlag(config);
5559
5773
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5566,9 +5780,9 @@ ${chalk10.dim("Persists workspace ID so you don't need to pass headers on every
5566
5780
  });
5567
5781
  set.command("org").description("Set default organization ID").argument("<id>", "Organization ID to set as default").addHelpText("after", `
5568
5782
  Examples:
5569
- ${chalk10.dim("$")} mutagent config set org <org-id>
5783
+ ${chalk15.dim("$")} mutagent config set org <org-id>
5570
5784
 
5571
- ${chalk10.dim("Persists organization ID for org-scoped API keys.")}
5785
+ ${chalk15.dim("Persists organization ID for org-scoped API keys.")}
5572
5786
  `).action((id) => {
5573
5787
  const isJson = getJsonFlag(config);
5574
5788
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5585,8 +5799,8 @@ ${chalk10.dim("Persists organization ID for org-scoped API keys.")}
5585
5799
 
5586
5800
  // src/commands/playground.ts
5587
5801
  init_sdk_client();
5588
- import { Command as Command8 } from "commander";
5589
- import chalk11 from "chalk";
5802
+ import { Command as Command11 } from "commander";
5803
+ import chalk16 from "chalk";
5590
5804
  init_errors();
5591
5805
  function parseSSELine(line) {
5592
5806
  if (!line || line.startsWith(":")) {
@@ -5611,13 +5825,13 @@ function parsePromptStreamEvent(data) {
5611
5825
  }
5612
5826
  }
5613
5827
  function createPlaygroundCommand() {
5614
- 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", `
5615
5829
  Examples:
5616
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{"name": "John"}'
5617
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{}' --stream
5618
- ${chalk11.dim("$")} mutagent playground run <prompt-id> -i '{}' --model gpt-4-turbo
5619
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --system "You are helpful" --human "Hello"
5620
- ${chalk11.dim("$")} 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"}]'
5621
5835
 
5622
5836
  Input Format:
5623
5837
  The input must be a valid JSON object matching the prompt's input schema.
@@ -5633,16 +5847,16 @@ Streaming:
5633
5847
  `);
5634
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", `
5635
5849
  Examples:
5636
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{"name": "John"}'
5637
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --input '{}' --stream
5638
- ${chalk11.dim("$")} mutagent playground run <prompt-id> --system "You are helpful" --human "Hello"
5639
- ${chalk11.dim("$")} 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
5640
5854
 
5641
5855
  Input Methods (pick one, priority order):
5642
- --system/--human Quick system + user message ${chalk11.green("(recommended)")}
5856
+ --system/--human Quick system + user message ${chalk16.green("(recommended)")}
5643
5857
  --input '{"key":"value"}' Inline JSON variables
5644
5858
  --messages '[...]' Full messages array
5645
- ${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"}'`)}
5646
5860
  `).action(async (promptId, options) => {
5647
5861
  const isJson = getJsonFlag(playground);
5648
5862
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5664,21 +5878,21 @@ ${chalk11.dim(`Hint: Test before evaluating: mutagent playground run <id> --inpu
5664
5878
  }
5665
5879
  });
5666
5880
  } else {
5667
- console.log(chalk11.bold(`
5881
+ console.log(chalk16.bold(`
5668
5882
  Execution Result:`));
5669
- console.log(chalk11.gray("─".repeat(50)));
5670
- console.log(chalk11.cyan("Output:"));
5883
+ console.log(chalk16.gray("─".repeat(50)));
5884
+ console.log(chalk16.cyan("Output:"));
5671
5885
  console.log(result.output);
5672
- console.log(chalk11.gray("─".repeat(50)));
5673
- console.log(chalk11.dim(`Model: ${result.model}`));
5674
- 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`));
5675
5889
  if (result.tokens) {
5676
- 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`));
5677
5891
  }
5678
5892
  if (result.cost !== undefined) {
5679
- console.log(chalk11.dim(`Cost: $${result.cost.toFixed(6)}`));
5893
+ console.log(chalk16.dim(`Cost: $${result.cost.toFixed(6)}`));
5680
5894
  }
5681
- console.log(chalk11.dim(`Playground: ${playgroundLink()}`));
5895
+ console.log(chalk16.dim(`Playground: ${playgroundLink()}`));
5682
5896
  console.log();
5683
5897
  }
5684
5898
  }
@@ -5767,9 +5981,9 @@ async function executeStreaming(client, promptId, input, model, isJson, output)
5767
5981
  const decoder = new TextDecoder;
5768
5982
  let buffer = "";
5769
5983
  if (!isJson) {
5770
- console.log(chalk11.bold(`
5984
+ console.log(chalk16.bold(`
5771
5985
  Streaming Output:`));
5772
- console.log(chalk11.gray("─".repeat(50)));
5986
+ console.log(chalk16.gray("─".repeat(50)));
5773
5987
  }
5774
5988
  try {
5775
5989
  for (;; ) {
@@ -5808,15 +6022,15 @@ Streaming Output:`));
5808
6022
  console.log(JSON.stringify({ type: "complete", result: event.result }));
5809
6023
  } else {
5810
6024
  console.log();
5811
- console.log(chalk11.gray("─".repeat(50)));
6025
+ console.log(chalk16.gray("─".repeat(50)));
5812
6026
  if (event.result) {
5813
- console.log(chalk11.dim(`Model: ${event.result.model}`));
5814
- 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`));
5815
6029
  if (event.result.tokens) {
5816
- 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`));
5817
6031
  }
5818
6032
  if (event.result.cost !== undefined) {
5819
- console.log(chalk11.dim(`Cost: $${event.result.cost.toFixed(6)}`));
6033
+ console.log(chalk16.dim(`Cost: $${event.result.cost.toFixed(6)}`));
5820
6034
  }
5821
6035
  }
5822
6036
  console.log();
@@ -5840,14 +6054,14 @@ Streaming Output:`));
5840
6054
 
5841
6055
  // src/commands/workspaces.ts
5842
6056
  init_sdk_client();
5843
- import { Command as Command9 } from "commander";
5844
- import chalk12 from "chalk";
6057
+ import { Command as Command12 } from "commander";
6058
+ import chalk17 from "chalk";
5845
6059
  init_errors();
5846
6060
  function createWorkspacesCommand() {
5847
- 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", `
5848
6062
  Examples:
5849
- ${chalk12.dim("$")} mutagent workspaces list
5850
- ${chalk12.dim("$")} mutagent workspaces get <workspace-id>
6063
+ ${chalk17.dim("$")} mutagent workspaces list
6064
+ ${chalk17.dim("$")} mutagent workspaces get <workspace-id>
5851
6065
 
5852
6066
  Subcommands:
5853
6067
  list, get
@@ -5856,8 +6070,8 @@ Note: Workspace management (create, update, delete) is available in the Admin Pa
5856
6070
  `);
5857
6071
  workspaces.command("list").description("List all workspaces").option("-l, --limit <n>", "Limit results", "50").option("-o, --offset <n>", "Offset for pagination").addHelpText("after", `
5858
6072
  Examples:
5859
- ${chalk12.dim("$")} mutagent workspaces list
5860
- ${chalk12.dim("$")} mutagent workspaces list --limit 10 --json
6073
+ ${chalk17.dim("$")} mutagent workspaces list
6074
+ ${chalk17.dim("$")} mutagent workspaces list --limit 10 --json
5861
6075
  `).action(async (options) => {
5862
6076
  const isJson = getJsonFlag(workspaces);
5863
6077
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5897,8 +6111,8 @@ Examples:
5897
6111
  });
5898
6112
  workspaces.command("get").description("Get workspace details").argument("<id>", "Workspace ID").addHelpText("after", `
5899
6113
  Examples:
5900
- ${chalk12.dim("$")} mutagent workspaces get <workspace-id>
5901
- ${chalk12.dim("$")} mutagent workspaces get <workspace-id> --json
6114
+ ${chalk17.dim("$")} mutagent workspaces get <workspace-id>
6115
+ ${chalk17.dim("$")} mutagent workspaces get <workspace-id> --json
5902
6116
  `).action(async (id) => {
5903
6117
  const isJson = getJsonFlag(workspaces);
5904
6118
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -5930,8 +6144,8 @@ Examples:
5930
6144
 
5931
6145
  // src/commands/providers.ts
5932
6146
  init_sdk_client();
5933
- import { Command as Command10 } from "commander";
5934
- import chalk13 from "chalk";
6147
+ import { Command as Command13 } from "commander";
6148
+ import chalk18 from "chalk";
5935
6149
  init_errors();
5936
6150
  var VALID_PROVIDER_TYPES = [
5937
6151
  "openai",
@@ -5953,11 +6167,11 @@ function validateProviderType(type) {
5953
6167
  return type;
5954
6168
  }
5955
6169
  function createProvidersCommand() {
5956
- 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", `
5957
6171
  Examples:
5958
- ${chalk13.dim("$")} mutagent providers list
5959
- ${chalk13.dim("$")} mutagent providers get <provider-id>
5960
- ${chalk13.dim("$")} 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>
5961
6175
 
5962
6176
  Provider Types:
5963
6177
  openai, anthropic, google, azure, bedrock, cohere, mistral, groq, together, replicate, custom
@@ -5967,15 +6181,15 @@ Subcommands:
5967
6181
 
5968
6182
  Note: Provider management (create, update, delete) is available in the Admin Panel only.
5969
6183
 
5970
- ${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
5971
6185
  for future external provider configuration. The server currently uses
5972
6186
  built-in provider settings.
5973
6187
  `);
5974
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", `
5975
6189
  Examples:
5976
- ${chalk13.dim("$")} mutagent providers list
5977
- ${chalk13.dim("$")} mutagent providers list --type openai
5978
- ${chalk13.dim("$")} 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
5979
6193
  `).action(async (options) => {
5980
6194
  const isJson = getJsonFlag(providers);
5981
6195
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6002,7 +6216,7 @@ Examples:
6002
6216
  }));
6003
6217
  output.output({ ...result, data: withLinks });
6004
6218
  } else {
6005
- 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."));
6006
6220
  console.log("");
6007
6221
  if (result.data.length === 0) {
6008
6222
  output.info("No providers configured.");
@@ -6026,8 +6240,8 @@ Examples:
6026
6240
  });
6027
6241
  providers.command("get").description("Get provider details").argument("<id>", "Provider ID (from: mutagent providers list)").addHelpText("after", `
6028
6242
  Examples:
6029
- ${chalk13.dim("$")} mutagent providers get <provider-id>
6030
- ${chalk13.dim("$")} mutagent providers get <provider-id> --json
6243
+ ${chalk18.dim("$")} mutagent providers get <provider-id>
6244
+ ${chalk18.dim("$")} mutagent providers get <provider-id> --json
6031
6245
  `).action(async (id) => {
6032
6246
  const isJson = getJsonFlag(providers);
6033
6247
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6037,7 +6251,7 @@ Examples:
6037
6251
  if (isJson) {
6038
6252
  output.output({ ...provider, _links: providerLinks(provider.id) });
6039
6253
  } else {
6040
- 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."));
6041
6255
  console.log("");
6042
6256
  const formatted = {
6043
6257
  id: provider.id,
@@ -6058,17 +6272,17 @@ Examples:
6058
6272
  });
6059
6273
  providers.command("test").description("Test provider connectivity").argument("<id>", "Provider ID (from: mutagent providers list)").addHelpText("after", `
6060
6274
  Examples:
6061
- ${chalk13.dim("$")} mutagent providers test <provider-id>
6062
- ${chalk13.dim("$")} mutagent providers test <provider-id> --json
6275
+ ${chalk18.dim("$")} mutagent providers test <provider-id>
6276
+ ${chalk18.dim("$")} mutagent providers test <provider-id> --json
6063
6277
 
6064
- ${chalk13.dim("Tests connectivity and lists available models for the provider.")}
6278
+ ${chalk18.dim("Tests connectivity and lists available models for the provider.")}
6065
6279
  `).action(async (id) => {
6066
6280
  const isJson = getJsonFlag(providers);
6067
6281
  const output = new OutputFormatter(isJson ? "json" : "table");
6068
6282
  try {
6069
6283
  const client = getSDKClient();
6070
6284
  if (!isJson) {
6071
- 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."));
6072
6286
  console.log("");
6073
6287
  output.info(`Testing provider ${id}...`);
6074
6288
  }
@@ -6078,9 +6292,9 @@ ${chalk13.dim("Tests connectivity and lists available models for the provider.")
6078
6292
  } else {
6079
6293
  if (result.success) {
6080
6294
  output.success(`Provider test passed (${String(result.responseTimeMs)}ms)`);
6081
- console.log(chalk13.green(`Message: ${result.message}`));
6295
+ console.log(chalk18.green(`Message: ${result.message}`));
6082
6296
  if (result.availableModels && result.availableModels.length > 0) {
6083
- console.log(chalk13.bold(`
6297
+ console.log(chalk18.bold(`
6084
6298
  Available Models:`));
6085
6299
  result.availableModels.forEach((model) => {
6086
6300
  console.log(` - ${model}`);
@@ -6089,7 +6303,7 @@ Available Models:`));
6089
6303
  } else {
6090
6304
  output.error(`Provider test failed: ${result.message}`);
6091
6305
  if (result.error) {
6092
- console.log(chalk13.red(`Error: ${result.error}`));
6306
+ console.log(chalk18.red(`Error: ${result.error}`));
6093
6307
  }
6094
6308
  }
6095
6309
  }
@@ -6102,13 +6316,18 @@ Available Models:`));
6102
6316
 
6103
6317
  // src/commands/init.ts
6104
6318
  init_config();
6105
- import { Command as Command11 } from "commander";
6319
+ import { Command as Command14 } from "commander";
6106
6320
  import inquirer3 from "inquirer";
6107
- import chalk14 from "chalk";
6108
- 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";
6109
6329
  import { execSync as execSync2 } from "child_process";
6110
6330
  import { join as join5 } from "path";
6111
- init_errors();
6112
6331
  var FRAMEWORK_DETECTION_MAP = {
6113
6332
  "@mastra/core": {
6114
6333
  name: "mastra",
@@ -6158,16 +6377,16 @@ var FRAMEWORK_DETECTION_MAP = {
6158
6377
  }
6159
6378
  };
6160
6379
  function detectPackageManager2(cwd = process.cwd()) {
6161
- if (existsSync10(join5(cwd, "bun.lockb")) || existsSync10(join5(cwd, "bun.lock"))) {
6380
+ if (existsSync11(join5(cwd, "bun.lockb")) || existsSync11(join5(cwd, "bun.lock"))) {
6162
6381
  return "bun";
6163
6382
  }
6164
- if (existsSync10(join5(cwd, "pnpm-lock.yaml"))) {
6383
+ if (existsSync11(join5(cwd, "pnpm-lock.yaml"))) {
6165
6384
  return "pnpm";
6166
6385
  }
6167
- if (existsSync10(join5(cwd, "yarn.lock"))) {
6386
+ if (existsSync11(join5(cwd, "yarn.lock"))) {
6168
6387
  return "yarn";
6169
6388
  }
6170
- if (existsSync10(join5(cwd, "package-lock.json"))) {
6389
+ if (existsSync11(join5(cwd, "package-lock.json"))) {
6171
6390
  return "npm";
6172
6391
  }
6173
6392
  try {
@@ -6189,12 +6408,12 @@ function getInstallCommand2(pm, packages) {
6189
6408
  }
6190
6409
  function detectFrameworkFromPackageJson(cwd = process.cwd()) {
6191
6410
  const pkgPath = join5(cwd, "package.json");
6192
- if (!existsSync10(pkgPath)) {
6411
+ if (!existsSync11(pkgPath)) {
6193
6412
  return null;
6194
6413
  }
6195
6414
  let pkg;
6196
6415
  try {
6197
- pkg = JSON.parse(readFileSync9(pkgPath, "utf-8"));
6416
+ pkg = JSON.parse(readFileSync10(pkgPath, "utf-8"));
6198
6417
  } catch {
6199
6418
  return null;
6200
6419
  }
@@ -6210,23 +6429,25 @@ function detectFrameworkFromPackageJson(cwd = process.cwd()) {
6210
6429
  return null;
6211
6430
  }
6212
6431
  function hasRcConfig(cwd = process.cwd()) {
6213
- return existsSync10(join5(cwd, ".mutagentrc.json"));
6432
+ return existsSync11(join5(cwd, ".mutagentrc.json"));
6214
6433
  }
6434
+
6435
+ // src/commands/init.ts
6215
6436
  function writeRcConfig(config, cwd = process.cwd()) {
6216
- const rcPath = join5(cwd, ".mutagentrc.json");
6437
+ const rcPath = join6(cwd, ".mutagentrc.json");
6217
6438
  writeFileSync4(rcPath, JSON.stringify(config, null, 2) + `
6218
6439
  `);
6219
6440
  }
6220
6441
  function createInitCommand() {
6221
- 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", `
6222
6443
  Examples:
6223
- ${chalk14.dim("$")} mutagent init # Interactive setup wizard
6224
- ${chalk14.dim("$")} 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)
6225
6446
 
6226
6447
  Modes:
6227
- ${chalk14.bold("Full scaffold")} Install SDK + integration package, create config, setup tracing
6228
- ${chalk14.bold("CLI-only")} Verify auth + create .mutagentrc.json with workspace/endpoint
6229
- ${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
6230
6451
  `).action(async (options) => {
6231
6452
  const isJson = getJsonFlag(init);
6232
6453
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6384,7 +6605,7 @@ Modes:
6384
6605
  output.info(`Installing: ${packages.join(", ")}`);
6385
6606
  output.info(`Running: ${installCmd}`);
6386
6607
  try {
6387
- execSync2(installCmd, { cwd, stdio: "inherit" });
6608
+ execSync3(installCmd, { cwd, stdio: "inherit" });
6388
6609
  output.success("Packages installed successfully.");
6389
6610
  } catch {
6390
6611
  output.error("Package installation failed. You can install manually:");
@@ -6427,8 +6648,8 @@ Modes:
6427
6648
  }
6428
6649
  }
6429
6650
  if (!isNonInteractive) {
6430
- const skillPath = join5(cwd, ".claude/skills/mutagent-cli/SKILL.md");
6431
- if (!existsSync10(skillPath)) {
6651
+ const skillPath = join6(cwd, ".claude/skills/mutagent-cli/SKILL.md");
6652
+ if (!existsSync12(skillPath)) {
6432
6653
  const { installSkill } = await inquirer3.prompt([{
6433
6654
  type: "confirm",
6434
6655
  name: "installSkill",
@@ -6437,11 +6658,11 @@ Modes:
6437
6658
  }]);
6438
6659
  if (installSkill) {
6439
6660
  try {
6440
- const skillDir = join5(cwd, ".claude/skills/mutagent-cli");
6441
- if (!existsSync10(skillDir)) {
6661
+ const skillDir = join6(cwd, ".claude/skills/mutagent-cli");
6662
+ if (!existsSync12(skillDir)) {
6442
6663
  mkdirSync3(skillDir, { recursive: true });
6443
6664
  }
6444
- execSync2("node " + join5(cwd, "node_modules/.bin/mutagent") + " skills install", {
6665
+ execSync3("node " + join6(cwd, "node_modules/.bin/mutagent") + " skills install", {
6445
6666
  cwd,
6446
6667
  stdio: "ignore"
6447
6668
  });
@@ -6475,24 +6696,24 @@ Modes:
6475
6696
  }
6476
6697
 
6477
6698
  // src/commands/explore.ts
6478
- import { Command as Command12 } from "commander";
6479
- import chalk15 from "chalk";
6699
+ import { Command as Command15 } from "commander";
6700
+ import chalk20 from "chalk";
6480
6701
  import { resolve as resolve3 } from "path";
6481
6702
  init_errors();
6482
6703
  function createExploreCommand() {
6483
- 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", `
6484
6705
  Examples:
6485
- ${chalk15.dim("$")} mutagent explore
6486
- ${chalk15.dim("$")} mutagent explore --path ./src
6487
- ${chalk15.dim("$")} mutagent explore --include "**/*.{ts,py}" --depth 5
6488
- ${chalk15.dim("$")} mutagent explore --markers-only
6489
- ${chalk15.dim("$")} 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
6490
6711
 
6491
6712
  Detection modes:
6492
- ${chalk15.dim("Heuristic")} Template variables ({{var}}), prompt constants, schema definitions
6493
- ${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
6494
6715
 
6495
- ${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.")}
6496
6717
  `).action((options) => {
6497
6718
  const isJson = getJsonFlag(explore);
6498
6719
  const output = new OutputFormatter(isJson ? "json" : "table");
@@ -6510,7 +6731,7 @@ ${chalk15.dim("Results are saved to .mutagent/mutation-context.md for use by oth
6510
6731
  markersOnly
6511
6732
  };
6512
6733
  if (!isJson) {
6513
- console.log(chalk15.cyan(`
6734
+ console.log(chalk20.cyan(`
6514
6735
  Scanning ${scanPath}...
6515
6736
  `));
6516
6737
  }
@@ -6539,40 +6760,41 @@ Scanning ${scanPath}...
6539
6760
  const totalFindings = result.prompts.length + result.datasets.length + result.markers.length;
6540
6761
  if (totalFindings === 0) {
6541
6762
  output.info("No prompts, datasets, or markers found.");
6542
- console.log(chalk15.dim(`
6763
+ console.log(chalk20.dim(`
6543
6764
  Tip: Create a prompt with template variables like {{input}} to get started.`));
6544
6765
  return;
6545
6766
  }
6546
6767
  if (result.prompts.length > 0) {
6547
- console.log(chalk15.bold(` Prompts Found (${String(result.prompts.length)}):`));
6768
+ console.log(chalk20.bold(` Prompts Found (${String(result.prompts.length)}):`));
6548
6769
  console.log();
6549
6770
  for (const p of result.prompts) {
6550
- const reasonTag = chalk15.dim(`[${p.reason}]`);
6551
- console.log(` ${chalk15.green(p.file)}:${chalk15.yellow(String(p.line))} ${reasonTag}`);
6552
- 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(` ${chalk20.dim(p.preview)}`);
6553
6775
  }
6554
6776
  console.log();
6555
6777
  }
6556
6778
  if (result.datasets.length > 0) {
6557
- console.log(chalk15.bold(` Datasets Found (${String(result.datasets.length)}):`));
6779
+ console.log(chalk20.bold(` Datasets Found (${String(result.datasets.length)}):`));
6558
6780
  console.log();
6559
6781
  for (const d of result.datasets) {
6560
- 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)`)}`);
6561
6783
  }
6562
6784
  console.log();
6563
6785
  }
6564
6786
  if (result.markers.length > 0) {
6565
- console.log(chalk15.bold(` MutagenT Markers (${String(result.markers.length)}):`));
6787
+ console.log(chalk20.bold(` MutagenT Markers (${String(result.markers.length)}):`));
6566
6788
  console.log();
6567
6789
  for (const m of result.markers) {
6568
- const idPart = m.platformId ? chalk15.cyan(` id=${m.platformId}`) : "";
6569
- 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}`);
6570
6792
  }
6571
6793
  console.log();
6572
6794
  }
6573
- console.log(chalk15.dim(" ─────────────────────────────────"));
6574
- console.log(` ${chalk15.bold("Summary:")} ${String(result.prompts.length)} prompts, ${String(result.datasets.length)} datasets, ${String(result.markers.length)} markers`);
6575
- 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`));
6576
6798
  console.log();
6577
6799
  }
6578
6800
  } catch (error) {
@@ -6583,14 +6805,14 @@ Scanning ${scanPath}...
6583
6805
  }
6584
6806
 
6585
6807
  // src/commands/skills.ts
6586
- import { Command as Command13 } from "commander";
6587
- import chalk16 from "chalk";
6588
- import { existsSync as existsSync11, mkdirSync as mkdirSync4, writeFileSync as writeFileSync5 } from "fs";
6589
- import { join as join6 } from "path";
6590
- 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";
6591
6813
  function findRepoRoot() {
6592
6814
  try {
6593
- 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();
6594
6816
  } catch {
6595
6817
  return process.cwd();
6596
6818
  }
@@ -6704,10 +6926,10 @@ Every evaluation MUST specify criteria targeting either:
6704
6926
  var SKILL_DIR = ".claude/skills/mutagent-cli";
6705
6927
  var SKILL_FILE = "SKILL.md";
6706
6928
  function createSkillsCommand() {
6707
- 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");
6708
6930
  skills.command("install").description("Install MutagenT CLI skill for Claude Code").addHelpText("after", `
6709
6931
  Examples:
6710
- ${chalk16.dim("$")} mutagent skills install
6932
+ ${chalk21.dim("$")} mutagent skills install
6711
6933
 
6712
6934
  This creates a Claude Code skill at .claude/skills/mutagent-cli/SKILL.md
6713
6935
  that teaches coding agents how to use the MutagenT CLI effectively.
@@ -6716,9 +6938,9 @@ that teaches coding agents how to use the MutagenT CLI effectively.
6716
6938
  const isJson = parentCmd ? getJsonFlag(parentCmd) : false;
6717
6939
  const output = new OutputFormatter(isJson ? "json" : "table");
6718
6940
  const repoRoot = findRepoRoot();
6719
- const skillDir = join6(repoRoot, SKILL_DIR);
6720
- const skillPath = join6(skillDir, SKILL_FILE);
6721
- if (!existsSync11(skillDir)) {
6941
+ const skillDir = join7(repoRoot, SKILL_DIR);
6942
+ const skillPath = join7(skillDir, SKILL_FILE);
6943
+ if (!existsSync13(skillDir)) {
6722
6944
  mkdirSync4(skillDir, { recursive: true });
6723
6945
  }
6724
6946
  const content = `${SKILL_FRONTMATTER}
@@ -6734,10 +6956,10 @@ ${SKILL_BODY}
6734
6956
  });
6735
6957
  } else {
6736
6958
  output.success(`Installed MutagenT CLI skill`);
6737
- console.log(` ${chalk16.dim("Path:")} ${skillPath}`);
6959
+ console.log(` ${chalk21.dim("Path:")} ${skillPath}`);
6738
6960
  console.log("");
6739
- console.log(` ${chalk16.dim("This skill teaches coding agents how to use the MutagenT CLI.")}`);
6740
- 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.")}`);
6741
6963
  }
6742
6964
  });
6743
6965
  return skills;
@@ -6745,8 +6967,8 @@ ${SKILL_BODY}
6745
6967
 
6746
6968
  // src/commands/usage.ts
6747
6969
  init_config();
6748
- import { Command as Command14 } from "commander";
6749
- import chalk17 from "chalk";
6970
+ import { Command as Command17 } from "commander";
6971
+ import chalk22 from "chalk";
6750
6972
  init_errors();
6751
6973
  init_sdk_client();
6752
6974
  var TRIAL_OPTIMIZATION_LIMIT = 5;
@@ -6760,10 +6982,10 @@ function renderProgressBar(used, limit, width = 30) {
6760
6982
  return `${bar} ${String(percent)}%`;
6761
6983
  }
6762
6984
  function createUsageCommand() {
6763
- 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", `
6764
6986
  Examples:
6765
- ${chalk17.dim("$")} mutagent usage
6766
- ${chalk17.dim("$")} mutagent usage --json
6987
+ ${chalk22.dim("$")} mutagent usage
6988
+ ${chalk22.dim("$")} mutagent usage --json
6767
6989
  `);
6768
6990
  usage.action(async () => {
6769
6991
  const isJson = getJsonFlag(usage);
@@ -6818,21 +7040,21 @@ Examples:
6818
7040
  });
6819
7041
  } else {
6820
7042
  console.log("");
6821
- console.log(chalk17.bold("\uD83D\uDCCA MutagenT Usage"));
6822
- console.log(chalk17.dim("─".repeat(45)));
7043
+ console.log(chalk22.bold("\uD83D\uDCCA MutagenT Usage"));
7044
+ console.log(chalk22.dim("─".repeat(45)));
6823
7045
  console.log("");
6824
- console.log(chalk17.bold("Resources:"));
6825
- console.log(` Prompts: ${chalk17.cyan(String(promptCount))}`);
6826
- console.log(` Datasets: ${chalk17.cyan(String(datasetCount))}`);
6827
- 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: ${chalk22.cyan(String(evaluationCount))}`);
6828
7050
  console.log("");
6829
- console.log(chalk17.bold(`Optimization Runs (${chalk17.yellow("trial")} plan):`));
6830
- 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)}`);
6831
7053
  console.log(` ${renderProgressBar(optimizationUsed, optimizationLimit)}`);
6832
7054
  console.log("");
6833
- console.log(chalk17.yellow(` ⚠ ${String(optimizationRemaining)} optimization runs remaining`));
6834
- console.log(chalk17.dim(` ℹ Optimization run counts are approximate`));
6835
- 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)}`);
6836
7058
  console.log("");
6837
7059
  }
6838
7060
  } catch (error) {
@@ -6850,12 +7072,12 @@ if (process.env.CLI_VERSION) {
6850
7072
  } else {
6851
7073
  try {
6852
7074
  const __dirname2 = dirname(fileURLToPath(import.meta.url));
6853
- const pkgPath = join7(__dirname2, "..", "..", "package.json");
6854
- const pkg = JSON.parse(readFileSync10(pkgPath, "utf-8"));
7075
+ const pkgPath = join8(__dirname2, "..", "..", "package.json");
7076
+ const pkg = JSON.parse(readFileSync11(pkgPath, "utf-8"));
6855
7077
  cliVersion = pkg.version ?? cliVersion;
6856
7078
  } catch {}
6857
7079
  }
6858
- var program = new Command15;
7080
+ var program = new Command18;
6859
7081
  program.name("mutagent").description(`MutagenT CLI - AI-native prompt optimization platform
6860
7082
 
6861
7083
  Documentation: https://docs.mutagent.io/cli
@@ -6864,56 +7086,57 @@ program.name("mutagent").description(`MutagenT CLI - AI-native prompt optimizati
6864
7086
  showGlobalOptions: true
6865
7087
  });
6866
7088
  program.addHelpText("after", `
6867
- ${chalk18.yellow("Non-Interactive Mode (CI/CD & Coding Agents):")}
6868
- export MUTAGENT_API_KEY=mt_... ${chalk18.dim("or")} --api-key mt_...
6869
- --json ${chalk18.dim("for structured output")} --non-interactive ${chalk18.dim("to disable prompts")}
6870
-
6871
- ${chalk18.yellow("Command Navigation:")}
6872
- mutagent auth login --browser ${chalk18.dim("Authenticate (OAuth)")}
6873
- mutagent auth status ${chalk18.dim("Check auth + workspace")}
6874
- mutagent init ${chalk18.dim("Initialize project (.mutagentrc.json)")}
6875
- mutagent explore ${chalk18.dim("Discover prompts in codebase")}
6876
- mutagent workspaces list --json ${chalk18.dim("List workspaces (verify ID)")}
6877
- mutagent config set workspace <id> ${chalk18.dim("Set active workspace")}
6878
- mutagent usage --json ${chalk18.dim("Check plan limits")}
6879
-
6880
- mutagent prompts create --help ${chalk18.dim("Upload prompt (read help first!)")}
6881
- mutagent prompts list --json ${chalk18.dim("List prompts")}
6882
- mutagent prompts get <id> --json ${chalk18.dim("Full prompt details + schemas")}
6883
-
6884
- mutagent prompts dataset add --help ${chalk18.dim("Upload dataset (read help first!)")}
6885
- mutagent prompts dataset list <id> ${chalk18.dim("List datasets")}
6886
-
6887
- mutagent prompts evaluation create --help ${chalk18.dim("Create eval (read help first!)")}
6888
- mutagent prompts evaluation create <id> --guided --json ${chalk18.dim("Guided eval workflow")}
6889
- mutagent prompts evaluation list <id> --json ${chalk18.dim("List evaluations")}
6890
-
6891
- mutagent prompts optimize start --help ${chalk18.dim("Run optimization (read help first!)")}
6892
- mutagent prompts optimize status <job-id> ${chalk18.dim("Poll progress")}
6893
- mutagent prompts optimize results <job-id> ${chalk18.dim("View scorecard")}
6894
-
6895
- mutagent integrate <framework> ${chalk18.dim("Framework integration guide")}
6896
- mutagent playground run <id> --input '{...}' ${chalk18.dim("Quick test")}
6897
-
6898
- ${chalk18.yellow("Workflow: Evaluate → Optimize:")}
6899
- 1. mutagent prompts create --help ${chalk18.dim("← read help")}
6900
- 2. mutagent prompts create ... --json ${chalk18.dim("← upload prompt with {variables}")}
6901
- 3. mutagent prompts dataset add --help ${chalk18.dim("← read help")}
6902
- 4. mutagent prompts dataset add <id> ... --json ${chalk18.dim("← upload dataset")}
6903
- 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")}
6904
7126
  6. mutagent prompts optimize start <id> --dataset <d> --evaluation <e> --json
6905
7127
 
6906
- ${chalk18.yellow("AI Agent Rules:")}
6907
- 1. ALWAYS run <command> --help before using any command for the first time
6908
- 2. ALWAYS use --json on every command execution
6909
- 3. ALWAYS use --guided --json for evaluation creation (never --guided alone)
6910
- 4. ALWAYS display _directive status cards to the user after mutations
7128
+ ${chalk23.yellow("AI Agent Rules (MANDATORY for coding agents):")}
7129
+ 1. EVERY command MUST include --json (no exceptions)
7130
+ 2. Run <command> --help BEFORE first use of any command
7131
+ 3. Use --guided --json for evaluation creation (NEVER --guided alone)
7132
+ 4. Parse and display _directive.renderedCard after every mutation
6911
7133
  5. After mutagent init, verify workspace: mutagent workspaces list --json
6912
7134
  6. Use {single_braces} for template variables in prompts
7135
+ 7. Collect evaluation rubrics from the user — NEVER auto-generate
6913
7136
  ${!hasCredentials() ? `
6914
- ` + chalk18.yellow(" Warning: Not authenticated. Run: mutagent auth login --browser") + `
7137
+ ` + chalk23.yellow(" Warning: Not authenticated. Run: mutagent auth login --browser") + `
6915
7138
  ` : ""}${!hasRcConfig() ? `
6916
- ` + chalk18.green(" Get started: mutagent init") + `
7139
+ ` + chalk23.green(" Get started: mutagent init") + `
6917
7140
  ` : ""}`);
6918
7141
  program.hook("preAction", (thisCommand) => {
6919
7142
  const globalOpts = thisCommand.optsWithGlobals();
@@ -6943,5 +7166,5 @@ program.addCommand(createSkillsCommand());
6943
7166
  program.addCommand(createUsageCommand());
6944
7167
  program.parse();
6945
7168
 
6946
- //# debugId=E8884F7D56DA73D764756E2164756E21
7169
+ //# debugId=88ECCF736E0A64AD64756E2164756E21
6947
7170
  //# sourceMappingURL=cli.js.map