@pellux/goodvibes-agent 1.1.7 → 1.2.0

package/src/input/agent-workspace-local-library-snapshot.ts

@@ -0,0 +1,167 @@
+import type { MemoryRecord } from '@pellux/goodvibes-sdk/platform/state';
+import type { AgentNoteRecord } from '../agent/note-registry.ts';
+import type { AgentPersonaRecord } from '../agent/persona-registry.ts';
+import { formatAgentRecordOrigin } from '../agent/record-labels.ts';
+import { evaluateAgentRoutineReadiness, type AgentRoutineRecord } from '../agent/routine-registry.ts';
+import { AgentResearchRunRegistry, researchRunLogTail } from '../agent/research-run-registry.ts';
+import { RoutineScheduleReceiptStore } from '../agent/routine-schedule-receipts.ts';
+import { evaluateAgentSkillBundleReadiness, evaluateAgentSkillReadiness, formatAgentSkillRequirement, type AgentSkillBundleRecord, type AgentSkillRecord } from '../agent/skill-registry.ts';
+import type { listAgentRuntimeProfiles, listAgentRuntimeProfileTemplates } from '../agent/runtime-profile.ts';
+import type { AgentWorkspaceLocalLibraryItem, AgentWorkspaceResearchRunSummary, AgentWorkspaceRoutineScheduleReceiptSummary, AgentWorkspaceRuntimeProfileItem, AgentWorkspaceRuntimeStarterTemplateItem } from './agent-workspace-types.ts';
+
+export function summarizePersonaItem(persona: AgentPersonaRecord, activePersonaId: string | null): AgentWorkspaceLocalLibraryItem {
+  return {
+    id: persona.id,
+    name: persona.name,
+    description: persona.description,
+    reviewState: persona.reviewState,
+    source: formatAgentRecordOrigin(persona.source, persona.provenance),
+    tags: persona.tags,
+    triggers: persona.triggers,
+    active: persona.id === activePersonaId,
+  };
+}
+
+export function summarizeSkillItem(skill: AgentSkillRecord): AgentWorkspaceLocalLibraryItem {
+  const readiness = evaluateAgentSkillReadiness(skill);
+  return {
+    id: skill.id,
+    name: skill.name,
+    description: skill.description,
+    reviewState: skill.reviewState,
+    source: formatAgentRecordOrigin(skill.source, skill.provenance),
+    tags: skill.tags,
+    triggers: skill.triggers,
+    enabled: skill.enabled,
+    requirementCount: skill.requirements.length,
+    missingRequirementCount: readiness.missing.length,
+    missingRequirements: readiness.missing.map(formatAgentSkillRequirement),
+  };
+}
+
+export function summarizeSkillBundleItem(bundle: AgentSkillBundleRecord, skills: readonly AgentSkillRecord[]): AgentWorkspaceLocalLibraryItem {
+  const readiness = evaluateAgentSkillBundleReadiness(bundle, skills);
+  const missing = [
+    ...readiness.missingRequirements.map(formatAgentSkillRequirement),
+    ...readiness.missingSkillIds.map((skillId) => `skill:${skillId}`),
+  ];
+  return {
+    id: bundle.id,
+    name: bundle.name,
+    description: `${bundle.description} Skills: ${bundle.skillIds.join(', ')}`,
+    reviewState: bundle.reviewState,
+    source: formatAgentRecordOrigin(bundle.source, bundle.provenance),
+    tags: bundle.skillIds,
+    triggers: [],
+    enabled: bundle.enabled,
+    requirementCount: readiness.includedSkills.reduce((total, skill) => total + skill.requirements.length, 0) + readiness.missingSkillIds.length,
+    missingRequirementCount: missing.length,
+    missingRequirements: missing,
+  };
+}
+
+export function summarizeRoutineItem(routine: AgentRoutineRecord): AgentWorkspaceLocalLibraryItem {
+  const readiness = evaluateAgentRoutineReadiness(routine);
+  return {
+    id: routine.id,
+    name: routine.name,
+    description: routine.description,
+    reviewState: routine.reviewState,
+    source: formatAgentRecordOrigin(routine.source, routine.provenance),
+    tags: routine.tags,
+    triggers: routine.triggers,
+    enabled: routine.enabled,
+    requirementCount: routine.requirements.length,
+    missingRequirementCount: readiness.missing.length,
+    missingRequirements: readiness.missing.map(formatAgentSkillRequirement),
+    startCount: routine.startCount,
+  };
+}
+
+export function summarizeRoutineScheduleReceipt(
+  receipt: ReturnType<RoutineScheduleReceiptStore['snapshot']>['receipts'][number],
+): AgentWorkspaceRoutineScheduleReceiptSummary {
+  return {
+    id: receipt.id,
+    status: receipt.status,
+    routineId: receipt.routineId,
+    routineName: receipt.routineName,
+    scheduleName: receipt.scheduleName,
+    scheduleKind: receipt.scheduleKind,
+    scheduleValue: receipt.scheduleValue,
+    createdAt: receipt.createdAt,
+  };
+}
+
+export function summarizeResearchRunItem(
+  run: ReturnType<AgentResearchRunRegistry['snapshot']>['runs'][number],
+): AgentWorkspaceResearchRunSummary {
+  return {
+    id: run.id,
+    title: run.title,
+    status: run.status,
+    phase: run.phase,
+    progress: run.progress,
+    sourceIds: run.sourceIds,
+    nextSteps: run.nextSteps,
+    checkpointCount: run.checkpoints.length,
+    logTail: researchRunLogTail(run, 4),
+    updatedAt: run.updatedAt,
+    ...(run.note ? { note: run.note } : {}),
+    ...(run.reportArtifactId ? { reportArtifactId: run.reportArtifactId } : {}),
+  };
+}
+
+export function summarizeMemoryItem(record: MemoryRecord): AgentWorkspaceLocalLibraryItem {
+  const detail = record.detail?.trim();
+  return {
+    id: record.id,
+    name: record.summary,
+    description: detail && detail.length > 0 ? detail : `${record.scope}/${record.cls}`,
+    reviewState: record.reviewState,
+    source: 'Agent memory',
+    tags: record.tags,
+    triggers: [],
+    scope: record.scope,
+    cls: record.cls,
+    confidence: record.confidence,
+  };
+}
+
+export function summarizeNoteItem(note: AgentNoteRecord): AgentWorkspaceLocalLibraryItem {
+  const preview = note.body.replace(/\s+/g, ' ').trim();
+  const description = note.sourceUrl
+    ? `${preview.slice(0, 160)}${preview.length > 160 ? '...' : ''} Origin URL ${note.sourceUrl}`
+    : preview;
+  return {
+    id: note.id,
+    name: note.title,
+    description,
+    reviewState: note.reviewState,
+    source: formatAgentRecordOrigin(note.source, note.provenance),
+    tags: note.tags,
+    triggers: [],
+  };
+}
+
+export function summarizeRuntimeProfile(profile: ReturnType<typeof listAgentRuntimeProfiles>[number]): AgentWorkspaceRuntimeProfileItem {
+  return {
+    id: profile.id,
+    homeDirectory: profile.homeDirectory,
+    createdAt: profile.createdAt,
+    starterTemplateId: profile.starterTemplateId,
+    starterTemplateName: profile.starterTemplateName,
+  };
+}
+
+export function summarizeStarterTemplate(template: ReturnType<typeof listAgentRuntimeProfileTemplates>[number]): AgentWorkspaceRuntimeStarterTemplateItem {
+  return {
+    id: template.id,
+    name: template.name,
+    description: template.description,
+    personaName: template.personaName,
+    skillNames: template.skillNames,
+    routineNames: template.routineNames,
+    source: template.source,
+  };
+}

package/src/input/agent-workspace-model-compare-editor.ts

@@ -0,0 +1,376 @@
+import type { AgentWorkspaceLocalEditor, AgentWorkspaceRecentReviewerHandoffArtifact, AgentWorkspaceReviewPacketDefaults } from './agent-workspace-types.ts';
+import type { AgentModelCompareAnalyticsWorkspaceToolArgs, AgentModelCompareApplyWorkspaceToolArgs, AgentModelCompareExportWorkspaceToolArgs, AgentModelCompareHandoffDiffWorkspaceToolArgs, AgentModelCompareJudgmentWorkspaceToolArgs, AgentModelCompareReviewWorkspaceToolArgs, AgentModelCompareRouteDecisionWorkspaceToolArgs, AgentModelCompareWorkspaceToolArgs, AgentWorkspaceFieldReader } from './agent-workspace-model-compare-types.ts';
+import { compareExportMode, compareReviewMode, isAffirmative, readList, readPositiveInteger } from './agent-workspace-model-compare-utils.ts';
+
+export type { AgentModelCompareAnalyticsWorkspaceToolArgs, AgentModelCompareApplyWorkspaceToolArgs, AgentModelCompareExportWorkspaceToolArgs, AgentModelCompareHandoffDiffWorkspaceToolArgs, AgentModelCompareJudgmentWorkspaceToolArgs, AgentModelCompareReviewWorkspaceToolArgs, AgentModelCompareRouteDecisionWorkspaceToolArgs, AgentModelCompareWorkspaceToolArgs, AgentWorkspaceFieldReader } from './agent-workspace-model-compare-types.ts';
+export { buildAgentModelCompareAnalyticsPromptSubmission, buildAgentModelCompareApplyPromptSubmission, buildAgentModelCompareExportPromptSubmission, buildAgentModelCompareHandoffDiffPromptSubmission, buildAgentModelCompareJudgmentPromptSubmission, buildAgentModelComparePromptSubmission, buildAgentModelCompareReviewPromptSubmission, buildAgentModelCompareRouteDecisionPromptSubmission } from './agent-workspace-model-compare-prompt-submission.ts';
+
+export function createAgentModelCompareEditor(): AgentWorkspaceLocalEditor {
+  return {
+    kind: 'model-compare',
+    mode: 'create',
+    title: 'Run Blind Model Compare',
+    selectedFieldIndex: 0,
+    message: 'Run one prompt against two to four selectable models and save a local JSON review artifact. Leave models blank to auto-select candidates. Type yes on the final field to confirm.',
+    fields: [
+      { id: 'prompt', label: 'Prompt', value: '', required: true, multiline: true, hint: 'Exact prompt sent identically to every candidate model. Ctrl-J inserts a new line.' },
+      { id: 'artifactId', label: 'Source artifact', value: '', required: false, multiline: false, hint: 'Optional saved text artifact id to include identically in the candidate prompt.' },
+      { id: 'modelRefs', label: 'Models', value: '', required: false, multiline: true, hint: 'Optional registry keys or model ids, separated by commas or new lines. Blank auto-selects.' },
+      { id: 'candidateCount', label: 'Auto count', value: '2', required: false, multiline: false, hint: 'Used only when Models is blank. 2 to 4.' },
+      { id: 'rubric', label: 'Rubric', value: '', required: false, multiline: true, hint: 'Optional judging rubric shown with blinded results.' },
+      { id: 'systemPrompt', label: 'System prompt', value: '', required: false, multiline: true, hint: 'Optional system prompt sent identically to every candidate.' },
+      { id: 'maxTokens', label: 'Max tokens', value: '2048', required: false, multiline: false, hint: 'Per-candidate output cap. Defaults to 2048.' },
+      { id: 'benchmarkKind', label: 'Benchmark tag', value: '', required: false, multiline: false, hint: 'Optional benchmark tag for filtered analytics.' },
+      { id: 'taskType', label: 'Task type', value: '', required: false, multiline: false, hint: 'Optional tag such as writing, research, code-review, or benchmark.' },
+      { id: 'documentId', label: 'Document id', value: '', required: false, multiline: false, hint: 'Optional document id tag. Source document exports can fill this automatically.' },
+      { id: 'reveal', label: 'Reveal now', value: 'no', required: false, multiline: false, hint: 'yes/no. Blank or no keeps model identities hidden until reveal.' },
+      { id: 'confirm', label: 'Confirm', value: '', required: true, multiline: false, hint: 'Type yes to run a token-spending model comparison and save the local review artifact.' },
+    ],
+  };
+}
+
+export function createAgentLocalModelBenchmarkEditor(): AgentWorkspaceLocalEditor {
+  return {
+    kind: 'local-model-benchmark',
+    mode: 'create',
+    title: 'Run Local Model Benchmark',
+    selectedFieldIndex: 0,
+    message: 'Run the local-route benchmark with the first-class blind comparison tool and save a tagged local review artifact. Enter a local model route plus a baseline route, then type yes on the final field.',
+    fields: [
+      {
+        id: 'prompt',
+        label: 'Prompt',
+        value: [
+          'Benchmark this local route on one practical task:',
+          '1. summarize the current project goal in five bullets,',
+          '2. identify one likely setup risk,',
+          '3. propose one next action with a command or route.',
+        ].join(' '),
+        required: true,
+        multiline: true,
+        hint: 'Exact benchmark prompt sent identically to every candidate model. Ctrl-J inserts a new line.',
+      },
+      { id: 'artifactId', label: 'Source artifact', value: '', required: false, multiline: false, hint: 'Optional saved text artifact id to include identically in the benchmark prompt.' },
+      { id: 'modelRefs', label: 'Models', value: '', required: false, multiline: true, hint: 'Recommended: local route first, then current/baseline route. Registry keys or model ids, separated by commas or new lines.' },
+      { id: 'candidateCount', label: 'Auto count', value: '2', required: false, multiline: false, hint: 'Used only when Models is blank. 2 to 4.' },
+      { id: 'rubric', label: 'Rubric', value: 'Prefer fast first useful output, correct project context, concrete setup-risk identification, and one actionable next route or command.', required: false, multiline: true, hint: 'Judging rubric shown with blinded benchmark results.' },
+      { id: 'systemPrompt', label: 'System prompt', value: '', required: false, multiline: true, hint: 'Optional system prompt sent identically to every candidate.' },
+      { id: 'maxTokens', label: 'Max tokens', value: '1024', required: false, multiline: false, hint: 'Per-candidate output cap. Defaults to 1024 for this benchmark.' },
+      { id: 'benchmarkKind', label: 'Benchmark tag', value: 'local-model-route', required: false, multiline: false, hint: 'Keeps the saved comparison visible in local model benchmark history.' },
+      { id: 'taskType', label: 'Task type', value: 'local-model-route', required: false, multiline: false, hint: 'Tags the saved judgment trend for filtered analytics.' },
+      { id: 'reveal', label: 'Reveal now', value: 'no', required: false, multiline: false, hint: 'yes/no. Blank or no keeps model identities hidden until reveal.' },
+      { id: 'confirm', label: 'Confirm', value: '', required: true, multiline: false, hint: 'Type yes to spend model tokens and save the local benchmark artifact.' },
+    ],
+  };
+}
+
+export function createAgentModelCompareReviewEditor(): AgentWorkspaceLocalEditor {
+  return {
+    kind: 'model-compare-review',
+    mode: 'create',
+    title: 'Review Saved Compare',
+    selectedFieldIndex: 0,
+    message: 'Review saved blind comparison artifacts, render related evidence side by side, or compare two reviewer handoffs. Leave ids blank to list recent saved comparisons or handoffs.',
+    fields: [
+      { id: 'view', label: 'View', value: 'review', required: false, multiline: false, hint: 'review, sideBySide, or handoffDiff. Diff compares two saved reviewer handoff artifacts.' },
+      { id: 'artifactId', label: 'Artifact id', value: '', required: false, multiline: false, hint: 'Saved artifact id such as artifact-123. Blank lists recent saved comparisons.' },
+      { id: 'comparisonId', label: 'Comparison id', value: '', required: false, multiline: false, hint: 'Comparison id such as cmp_... if known. Artifact id is usually easier.' },
+      { id: 'leftArtifactId', label: 'Left handoff', value: '', required: false, multiline: false, hint: 'For handoffDiff: first saved reviewer handoff artifact id. Artifact id can also fill this.' },
+      { id: 'rightArtifactId', label: 'Right handoff', value: '', required: false, multiline: false, hint: 'For handoffDiff: second saved reviewer handoff artifact id.' },
+      { id: 'relatedArtifactIds', label: 'Related artifacts', value: '', required: false, multiline: true, hint: 'For sideBySide: document export, archive, or artifact ids, separated by commas or new lines.' },
+      { id: 'previewBytes', label: 'Preview bytes', value: '2000', required: false, multiline: false, hint: 'For sideBySide: max bytes per related artifact preview, 200 to 10000.' },
+      { id: 'reveal', label: 'Reveal', value: 'no', required: false, multiline: false, hint: 'yes/no. No keeps identities hidden and renders the blind review board.' },
+    ],
+  };
+}
+
+export function createAgentModelCompareHandoffDiffEditor(
+  recentHandoffs: readonly AgentWorkspaceRecentReviewerHandoffArtifact[] = [],
+): AgentWorkspaceLocalEditor {
+  const right = recentHandoffs.length >= 2 ? recentHandoffs[0] : undefined;
+  const left = recentHandoffs.length >= 2 ? recentHandoffs[1] : undefined;
+  const recentHint = recentHandoffs.length >= 2
+    ? ` Prefilled with the two newest handoffs: ${left?.id ?? ''} -> ${right?.id ?? ''}.`
+    : ' Leave both ids blank to list recent saved handoffs.';
+  return {
+    kind: 'model-compare-handoff-diff',
+    mode: 'create',
+    title: 'Diff Reviewer Handoffs',
+    selectedFieldIndex: 0,
+    message: `Compare two saved reviewer handoff artifacts in the workspace split view. Use Section jump to focus the diff on all, metadata, policy, related, or comparison evidence.${recentHint}`,
+    fields: [
+      { id: 'leftArtifactId', label: 'Left handoff', value: left?.id ?? '', required: false, multiline: false, hint: left ? `Older handoff ${left.handoffId}; source ${left.sourceArtifactId}; related ${left.relatedArtifactCount}. Clear both ids to list saved handoffs.` : 'First saved reviewer handoff artifact id. Leave both handoff fields blank to list recent saved handoffs.' },
+      { id: 'rightArtifactId', label: 'Right handoff', value: right?.id ?? '', required: false, multiline: false, hint: right ? `Newer handoff ${right.handoffId}; source ${right.sourceArtifactId}; related ${right.relatedArtifactCount}. Clear both ids to list saved handoffs.` : 'Second saved reviewer handoff artifact id. Leave both handoff fields blank to list recent saved handoffs.' },
+      { id: 'sectionId', label: 'Section jump', value: 'all', required: false, multiline: false, hint: 'all, metadata, policy, related, or comparison. Related focuses changed document/artifact evidence.' },
+    ],
+  };
+}
+
+export function createAgentModelCompareJudgmentEditor(): AgentWorkspaceLocalEditor {
+  return {
+    kind: 'model-compare-judge',
+    mode: 'create',
+    title: 'Save Compare Judgment',
+    selectedFieldIndex: 0,
+    message: 'Save the winning candidate and reasons as a local judgment artifact. This does not change the selected model.',
+    fields: [
+      { id: 'artifactId', label: 'Artifact id', value: '', required: false, multiline: false, hint: 'Saved comparison artifact id such as artifact-123.' },
+      { id: 'comparisonId', label: 'Comparison id', value: '', required: false, multiline: false, hint: 'Comparison id such as cmp_... if known.' },
+      { id: 'winnerBlindId', label: 'Winner', value: '', required: true, multiline: false, hint: 'Candidate label, such as A or Candidate B.' },
+      { id: 'reasons', label: 'Reasons', value: '', required: true, multiline: true, hint: 'Why this candidate won. Ctrl-J inserts a new line.' },
+      { id: 'notes', label: 'Notes', value: '', required: false, multiline: true, hint: 'Optional risks, follow-ups, or rubric notes.' },
+      { id: 'reveal', label: 'Reveal in judgment', value: 'no', required: false, multiline: false, hint: 'yes/no. Yes includes model identity for later route update.' },
+      { id: 'confirm', label: 'Confirm', value: '', required: true, multiline: false, hint: 'Type yes to save this judgment artifact.' },
+    ],
+  };
+}
+
+function packetSourceArtifactId(defaults: AgentWorkspaceReviewPacketDefaults | null): string {
+  return defaults?.revealedJudgmentArtifactId
+    ?? defaults?.judgmentArtifactId
+    ?? defaults?.comparisonArtifactId
+    ?? '';
+}
+
+export function createAgentModelCompareApplyEditor(defaults: AgentWorkspaceReviewPacketDefaults | null = null): AgentWorkspaceLocalEditor {
+  const artifactId = defaults?.revealedJudgmentArtifactId ?? '';
+  const artifactHint = defaults?.summary
+    ? `Default from latest revealed packet judgment. Packet default: ${defaults.summary}. Clear it to choose another saved judgment.`
+    : 'Saved revealed judgment artifact id such as artifact-123.';
+  return {
+    kind: 'model-compare-apply',
+    mode: 'create',
+    title: 'Apply Compare Winner',
+    selectedFieldIndex: 0,
+    message: artifactId
+      ? 'Apply a revealed saved judgment artifact to the main Agent model route.'
+      : 'Apply a revealed saved judgment artifact to the main Agent model route. This changes provider.model after confirmation.',
+    fields: [
+      { id: 'artifactId', label: 'Judgment artifact id', value: artifactId, required: true, multiline: false, hint: artifactId ? artifactHint : 'Saved revealed judgment artifact id such as artifact-123.' },
+      { id: 'confirm', label: 'Confirm', value: '', required: true, multiline: false, hint: 'Type yes to change the selected Agent model.' },
+    ],
+  };
+}
+
+export function createAgentModelCompareRouteDecisionEditor(defaults: AgentWorkspaceReviewPacketDefaults | null = null): AgentWorkspaceLocalEditor {
+  const artifactId = defaults?.revealedJudgmentArtifactId ?? '';
+  const artifactHint = defaults?.summary
+    ? `Default from latest revealed packet judgment. Packet default: ${defaults.summary}.`
+    : 'Saved revealed judgment artifact id such as artifact-123.';
+  return {
+    kind: 'model-compare-route-decision',
+    mode: 'create',
+    title: 'Record Route Decision',
+    selectedFieldIndex: 0,
+    message: artifactId
+      ? 'Record that the current Agent model route should stay unchanged for this revealed judgment.'
+      : 'Record that the current Agent model route should stay unchanged for a revealed saved judgment.',
+    fields: [
+      { id: 'artifactId', label: 'Judgment artifact id', value: artifactId, required: true, multiline: false, hint: artifactHint },
+      { id: 'decision', label: 'Decision', value: 'left-unchanged', required: true, multiline: false, hint: 'left-unchanged. Use Apply Compare Winner for route updates.' },
+      { id: 'confirm', label: 'Confirm', value: '', required: true, multiline: false, hint: 'Type yes to save the route-decision receipt without changing the selected model.' },
+    ],
+  };
+}
+
+export function createAgentModelCompareExportEditor(defaults: AgentWorkspaceReviewPacketDefaults | null = null): AgentWorkspaceLocalEditor {
+  const defaultKind = defaults?.handoffArtifactId
+    ? 'archive'
+    : defaults?.relatedArtifactIds.length
+      ? 'handoff'
+      : 'report';
+  const artifactId = defaultKind === 'archive'
+    ? defaults?.handoffArtifactId ?? ''
+    : packetSourceArtifactId(defaults);
+  const relatedArtifactIds = defaultKind === 'handoff' ? (defaults?.relatedArtifactIds ?? []).join('\n') : '';
+  return {
+    kind: 'model-compare-export',
+    mode: 'create',
+    title: 'Export Compare Report/Handoff',
+    selectedFieldIndex: 0,
+    message: artifactId
+      ? 'Export a saved comparison or judgment as a markdown report, create a reviewer handoff with related document/artifact ids, or archive a saved handoff as one ZIP artifact.'
+      : 'Export a saved comparison or judgment as a markdown report, create a reviewer handoff with related document/artifact ids, or archive a saved handoff as one ZIP artifact. This does not change the selected model.',
+    fields: [
+      { id: 'reportKind', label: 'Kind', value: defaultKind, required: false, multiline: false, hint: defaults?.handoffArtifactId ? `Default archive uses the latest reviewer handoff. Packet default: ${defaults.summary}. Change to report or handoff to use comparison/judgment artifacts.` : defaults?.relatedArtifactIds.length ? `Default handoff carries latest related packet evidence. Packet default: ${defaults.summary}. Change to report for markdown only.` : 'report, handoff, or archive. Archive expects a saved handoff artifact id and creates one ZIP artifact.' },
+      { id: 'artifactId', label: 'Artifact id', value: artifactId, required: true, multiline: false, hint: artifactId ? `Default from latest packet comparison, judgment, or reviewer handoff. Packet default: ${defaults?.summary ?? artifactId}. Clear it to choose another saved artifact.` : 'Saved comparison/judgment artifact for report or handoff; saved handoff artifact for archive.' },
+      { id: 'relatedArtifactIds', label: 'Related artifacts', value: relatedArtifactIds, required: false, multiline: true, hint: relatedArtifactIds ? `Default related evidence from latest document export, attachments, source artifacts, or handoff metadata. Packet default: ${defaults?.summary ?? relatedArtifactIds}.` : 'For handoff: document export, archive, or artifact ids, separated by commas or new lines.' },
+      { id: 'reveal', label: 'Reveal in export', value: defaultKind === 'handoff' ? 'yes' : 'no', required: false, multiline: false, hint: 'yes/no. For comparison artifacts, yes includes model identities. Judgment artifacts use their saved reveal state.' },
+      { id: 'confirm', label: 'Confirm', value: '', required: true, multiline: false, hint: 'Type yes to create the local report, handoff, or ZIP archive artifact.' },
+    ],
+  };
+}
+
+export function createAgentModelCompareAnalyticsEditor(): AgentWorkspaceLocalEditor {
+  return {
+    kind: 'model-compare-analytics',
+    mode: 'create',
+    title: 'Compare Analytics/Synthesis',
+    selectedFieldIndex: 0,
+    message: 'Summarize or synthesize saved blind comparison judgments by winner, model, blind slot, themes, task type, document id, benchmark tag, and recent reasons. This is read-only.',
+    fields: [
+      { id: 'view', label: 'View', value: 'analytics', required: false, multiline: false, hint: 'analytics or synthesis. Synthesis groups saved judgment themes across sessions.' },
+      { id: 'limit', label: 'Judgment limit', value: '20', required: false, multiline: false, hint: 'Maximum saved judgment artifacts to inspect. Defaults to 20.' },
+      { id: 'benchmarkKind', label: 'Benchmark tag', value: '', required: false, multiline: false, hint: 'Optional filter, such as local-model-route.' },
+      { id: 'taskType', label: 'Task type', value: '', required: false, multiline: false, hint: 'Optional filter, such as writing, research, code-review, or local-model-route.' },
+      { id: 'documentId', label: 'Document id', value: '', required: false, multiline: false, hint: 'Optional filter for judgments tied to one Agent document id.' },
+      { id: 'includeReasons', label: 'Include reasons', value: 'yes', required: false, multiline: false, hint: 'yes/no. Yes includes short reason and note excerpts in the recent-judgments list.' },
+    ],
+  };
+}
+
+export function buildAgentModelCompareToolArgs(
+  readField: AgentWorkspaceFieldReader,
+  explicitUserRequest: string,
+): AgentModelCompareWorkspaceToolArgs {
+  const prompt = readField('prompt').trim();
+  const artifactId = readField('artifactId').trim();
+  const modelRefs = readList(readField('modelRefs'));
+  const candidateCount = readPositiveInteger(readField('candidateCount'));
+  const rubric = readField('rubric').trim();
+  const systemPrompt = readField('systemPrompt').trim();
+  const maxTokens = readPositiveInteger(readField('maxTokens'));
+  const benchmarkKind = readField('benchmarkKind').trim();
+  const taskType = readField('taskType').trim();
+  const documentId = readField('documentId').trim();
+  const reveal = isAffirmative(readField('reveal'));
+
+  return {
+    mode: 'run',
+    prompt,
+    ...(artifactId ? { artifactId } : {}),
+    ...(modelRefs.length > 0 ? { modelRefs } : {}),
+    ...(candidateCount !== null ? { candidateCount } : {}),
+    ...(rubric ? { rubric } : {}),
+    ...(systemPrompt ? { systemPrompt } : {}),
+    ...(maxTokens !== null ? { maxTokens } : {}),
+    ...(benchmarkKind ? { benchmarkKind } : {}),
+    ...(taskType ? { taskType } : {}),
+    ...(documentId ? { documentId } : {}),
+    reveal,
+    confirm: true,
+    explicitUserRequest,
+  };
+}
+
+export function buildAgentModelCompareReviewToolArgs(
+  readField: AgentWorkspaceFieldReader,
+): AgentModelCompareReviewWorkspaceToolArgs {
+  const mode = compareReviewMode(readField('view'));
+  const artifactId = readField('artifactId').trim();
+  const comparisonId = readField('comparisonId').trim();
+  const leftArtifactId = readField('leftArtifactId').trim() || artifactId;
+  const rightArtifactId = readField('rightArtifactId').trim();
+  const relatedArtifactIds = readList(readField('relatedArtifactIds'));
+  const previewBytes = readPositiveInteger(readField('previewBytes'));
+  const reveal = isAffirmative(readField('reveal'));
+  if (mode === 'handoffDiff') {
+    const sectionId = readField('sectionId').trim();
+    return {
+      mode,
+      ...(leftArtifactId ? { leftArtifactId } : {}),
+      ...(rightArtifactId ? { rightArtifactId } : {}),
+      ...(sectionId && sectionId.toLowerCase() !== 'all' ? { sectionId } : {}),
+      ...(previewBytes !== null ? { previewBytes } : {}),
+    };
+  }
+  return {
+    mode,
+    ...(artifactId ? { artifactId } : {}),
+    ...(comparisonId ? { comparisonId } : {}),
+    ...(relatedArtifactIds.length > 0 ? { relatedArtifactIds } : {}),
+    ...(previewBytes !== null ? { previewBytes } : {}),
+    reveal,
+  };
+}
+
+export function buildAgentModelCompareHandoffDiffToolArgs(
+  readField: AgentWorkspaceFieldReader,
+): AgentModelCompareHandoffDiffWorkspaceToolArgs {
+  const sectionId = readField('sectionId').trim();
+  return {
+    mode: 'handoffDiff',
+    ...(readField('leftArtifactId').trim() ? { leftArtifactId: readField('leftArtifactId').trim() } : {}),
+    ...(readField('rightArtifactId').trim() ? { rightArtifactId: readField('rightArtifactId').trim() } : {}),
+    ...(sectionId && sectionId.toLowerCase() !== 'all' ? { sectionId } : {}),
+  };
+}
+
+export function buildAgentModelCompareJudgmentToolArgs(
+  readField: AgentWorkspaceFieldReader,
+  explicitUserRequest: string,
+): AgentModelCompareJudgmentWorkspaceToolArgs {
+  const artifactId = readField('artifactId').trim();
+  const comparisonId = readField('comparisonId').trim();
+  const notes = readField('notes').trim();
+  return {
+    mode: 'judge',
+    ...(artifactId ? { artifactId } : {}),
+    ...(comparisonId ? { comparisonId } : {}),
+    winnerBlindId: readField('winnerBlindId').trim(),
+    reasons: readField('reasons').trim(),
+    ...(notes ? { notes } : {}),
+    reveal: isAffirmative(readField('reveal')),
+    confirm: true,
+    explicitUserRequest,
+  };
+}
+
+export function buildAgentModelCompareApplyToolArgs(
+  readField: AgentWorkspaceFieldReader,
+  explicitUserRequest: string,
+): AgentModelCompareApplyWorkspaceToolArgs {
+  return {
+    mode: 'apply',
+    artifactId: readField('artifactId').trim(),
+    confirm: true,
+    explicitUserRequest,
+  };
+}
+
+export function buildAgentModelCompareRouteDecisionToolArgs(
+  readField: AgentWorkspaceFieldReader,
+  explicitUserRequest: string,
+): AgentModelCompareRouteDecisionWorkspaceToolArgs {
+  return {
+    mode: 'routeDecision',
+    artifactId: readField('artifactId').trim(),
+    decision: 'left-unchanged',
+    confirm: true,
+    explicitUserRequest,
+  };
+}
+
+export function buildAgentModelCompareExportToolArgs(
+  readField: AgentWorkspaceFieldReader,
+  explicitUserRequest: string,
+): AgentModelCompareExportWorkspaceToolArgs {
+  const mode = compareExportMode(readField('reportKind'));
+  const relatedArtifactIds = readList(readField('relatedArtifactIds'));
+  return {
+    mode,
+    artifactId: readField('artifactId').trim(),
+    ...(relatedArtifactIds.length > 0 ? { relatedArtifactIds } : {}),
+    reveal: isAffirmative(readField('reveal')),
+    confirm: true,
+    explicitUserRequest,
+  };
+}
+
+export function buildAgentModelCompareAnalyticsToolArgs(
+  readField: AgentWorkspaceFieldReader,
+): AgentModelCompareAnalyticsWorkspaceToolArgs {
+  const view = readField('view').trim().toLowerCase();
+  const limit = readPositiveInteger(readField('limit'));
+  const benchmarkKind = readField('benchmarkKind').trim();
+  const taskType = readField('taskType').trim();
+  const documentId = readField('documentId').trim();
+  return {
+    mode: view === 'synthesis' ? 'synthesis' : 'analytics',
+    ...(limit !== null ? { limit } : {}),
+    ...(benchmarkKind ? { benchmarkKind } : {}),
+    ...(taskType ? { taskType } : {}),
+    ...(documentId ? { documentId } : {}),
+    includeReasons: isAffirmative(readField('includeReasons')),
+  };
+}