claude-brain 0.30.2 → 0.30.3

package/src/cli/commands/export-training.ts

@@ -1,70 +1,70 @@
-/**
- * Export Training Data Command — Phase 1A (SLM Upgrade)
- * Exports logged classification data as JSONL for model training.
- *
- * Usage:
- *   claude-brain export-training intent
- *   claude-brain export-training entity --verified-only
- *   claude-brain export-training compress --limit 500
- *   claude-brain export-training --stats
- */
-
-import { parseArgs } from 'citty'
-import { exportTrainingData, getTrainingStats, type TrainingTask } from '@/training/data-store'
-
-const VALID_TASKS: TrainingTask[] = ['intent', 'entity', 'query', 'knowledge', 'compress', 'pattern']
-
-export async function runExportTraining() {
-  const args = parseArgs(process.argv.slice(3), {
-    task: { type: 'positional', description: 'Task to export: intent, entity, query, knowledge, compress, pattern' },
-    verifiedOnly: { type: 'boolean', alias: ['verified-only'], description: 'Only export human-verified examples' },
-    limit: { type: 'string', description: 'Max examples to export' },
-    stats: { type: 'boolean', description: 'Show training data statistics' },
-    format: { type: 'string', default: 'jsonl', description: 'Output format: jsonl or csv' },
-  })
-
-  // Stats mode
-  if (args.stats) {
-    const stats = getTrainingStats()
-    console.log('Training Data Statistics:')
-    console.log('========================')
-    let grandTotal = 0
-    let grandVerified = 0
-    for (const [task, counts] of Object.entries(stats)) {
-      console.log(`  ${task.padEnd(12)} ${String(counts.total).padStart(6)} total  ${String(counts.verified).padStart(6)} verified`)
-      grandTotal += counts.total
-      grandVerified += counts.verified
-    }
-    console.log('  ' + '-'.repeat(40))
-    console.log(`  ${'TOTAL'.padEnd(12)} ${String(grandTotal).padStart(6)} total  ${String(grandVerified).padStart(6)} verified`)
-    return
-  }
-
-  const task = args.task as TrainingTask
-  if (!task || !VALID_TASKS.includes(task)) {
-    console.error(`Error: task must be one of: ${VALID_TASKS.join(', ')}`)
-    console.error('Usage: claude-brain export-training <task> [--verified-only] [--limit N]')
-    console.error('       claude-brain export-training --stats')
-    process.exit(1)
-  }
-
-  const limit = args.limit ? parseInt(args.limit, 10) : undefined
-  const lines = exportTrainingData(task, {
-    verifiedOnly: args.verifiedOnly || false,
-    limit,
-  })
-
-  if (lines.length === 0) {
-    console.error(`No training data found for task "${task}".`)
-    console.error('Training data is collected automatically as Claude Brain processes requests.')
-    process.exit(0)
-  }
-
-  // Output to stdout (pipe-friendly)
-  for (const line of lines) {
-    console.log(line)
-  }
-
-  // Summary to stderr so it doesn't pollute piped output
-  console.error(`\nExported ${lines.length} examples for task "${task}"`)
-}
+/**
+ * Export Training Data Command — Phase 1A (SLM Upgrade)
+ * Exports logged classification data as JSONL for model training.
+ *
+ * Usage:
+ *   claude-brain export-training intent
+ *   claude-brain export-training entity --verified-only
+ *   claude-brain export-training compress --limit 500
+ *   claude-brain export-training --stats
+ */
+
+import { parseArgs } from 'citty'
+import { exportTrainingData, getTrainingStats, type TrainingTask } from '@/training/data-store'
+
+const VALID_TASKS: TrainingTask[] = ['intent', 'entity', 'query', 'knowledge', 'compress', 'pattern']
+
+export async function runExportTraining() {
+  const args = parseArgs(process.argv.slice(3), {
+    task: { type: 'positional', description: 'Task to export: intent, entity, query, knowledge, compress, pattern' },
+    verifiedOnly: { type: 'boolean', alias: ['verified-only'], description: 'Only export human-verified examples' },
+    limit: { type: 'string', description: 'Max examples to export' },
+    stats: { type: 'boolean', description: 'Show training data statistics' },
+    format: { type: 'string', default: 'jsonl', description: 'Output format: jsonl or csv' },
+  })
+
+  // Stats mode
+  if (args.stats) {
+    const stats = getTrainingStats()
+    console.log('Training Data Statistics:')
+    console.log('========================')
+    let grandTotal = 0
+    let grandVerified = 0
+    for (const [task, counts] of Object.entries(stats)) {
+      console.log(`  ${task.padEnd(12)} ${String(counts.total).padStart(6)} total  ${String(counts.verified).padStart(6)} verified`)
+      grandTotal += counts.total
+      grandVerified += counts.verified
+    }
+    console.log('  ' + '-'.repeat(40))
+    console.log(`  ${'TOTAL'.padEnd(12)} ${String(grandTotal).padStart(6)} total  ${String(grandVerified).padStart(6)} verified`)
+    return
+  }
+
+  const task = args.task as TrainingTask
+  if (!task || !VALID_TASKS.includes(task)) {
+    console.error(`Error: task must be one of: ${VALID_TASKS.join(', ')}`)
+    console.error('Usage: claude-brain export-training <task> [--verified-only] [--limit N]')
+    console.error('       claude-brain export-training --stats')
+    process.exit(1)
+  }
+
+  const limit = args.limit ? parseInt(args.limit, 10) : undefined
+  const lines = exportTrainingData(task, {
+    verifiedOnly: args.verifiedOnly || false,
+    limit,
+  })
+
+  if (lines.length === 0) {
+    console.error(`No training data found for task "${task}".`)
+    console.error('Training data is collected automatically as Claude Brain processes requests.')
+    process.exit(0)
+  }
+
+  // Output to stdout (pipe-friendly)
+  for (const line of lines) {
+    console.log(line)
+  }
+
+  // Summary to stderr so it doesn't pollute piped output
+  console.error(`\nExported ${lines.length} examples for task "${task}"`)
+}

package/src/cli/commands/export.ts

@@ -1,130 +1,130 @@
-/**
- * Export Command — Phase 30
- * Export observations for a project in markdown or JSON format.
- *
- * Usage:
- *   claude-brain export --project expense-tracker --format md > decisions.md
- *   claude-brain export --project expense-tracker --format json > decisions.json
- *   claude-brain export --project expense-tracker --category decision
- */
-
-import { join } from 'node:path'
-import { existsSync } from 'node:fs'
-import { parseArgs } from 'citty'
-import { getClaudeBrainHome } from '@/config/home'
-
-export async function runExport() {
-  const args = parseArgs(process.argv.slice(3), {
-    project: { type: 'string', description: 'Project name to export' },
-    format: { type: 'string', default: 'md', description: 'Output format: md or json' },
-    category: { type: 'string', description: 'Filter by category: decision, pattern, correction' },
-    limit: { type: 'string', default: '1000', description: 'Max observations to export' },
-  })
-
-  const project = args.project
-  const format = args.format || 'md'
-  const category = args.category
-  const limit = parseInt(args.limit || '1000', 10)
-
-  if (!project) {
-    console.error('Error: --project is required')
-    console.error('Usage: claude-brain export --project <name> [--format md|json] [--category decision|pattern|correction]')
-    process.exit(1)
-  }
-
-  const dbPath = join(getClaudeBrainHome(), 'data', 'memory.db')
-  if (!existsSync(dbPath)) {
-    console.error('Error: No database found. Run claude-brain first to initialize.')
-    process.exit(1)
-  }
-
-  const { Database } = await import('bun:sqlite')
-  const db = new Database(dbPath, { readonly: true })
-
-  let sql = 'SELECT * FROM observations WHERE project = ? AND archived = 0'
-  const params: any[] = [project]
-
-  if (category) {
-    sql += ' AND category = ?'
-    params.push(category)
-  }
-
-  sql += ' ORDER BY created_at DESC LIMIT ?'
-  params.push(limit)
-
-  const rows = db.prepare(sql).all(...params) as any[]
-  db.close()
-
-  if (rows.length === 0) {
-    console.error(`No observations found for project "${project}"${category ? ` with category "${category}"` : ''}.`)
-    process.exit(0)
-  }
-
-  if (format === 'json') {
-    const output = rows.map(row => ({
-      id: row.id,
-      project: row.project,
-      category: row.category,
-      content: row.content,
-      reasoning: row.reasoning || null,
-      context: row.context || null,
-      confidence: row.confidence,
-      source: row.source,
-      tags: parseTags(row.tags),
-      created_at: row.created_at,
-      updated_at: row.updated_at,
-    }))
-    console.log(JSON.stringify(output, null, 2))
-  } else {
-    // Markdown format
-    const lines: string[] = []
-    lines.push(`# ${project} — Observations Export`)
-    lines.push('')
-    lines.push(`> Exported ${rows.length} observation(s) on ${new Date().toISOString().split('T')[0]}`)
-    lines.push('')
-
-    // Group by category
-    const grouped: Record<string, any[]> = {}
-    for (const row of rows) {
-      const cat = row.category || 'other'
-      if (!grouped[cat]) grouped[cat] = []
-      grouped[cat].push(row)
-    }
-
-    for (const [cat, items] of Object.entries(grouped)) {
-      lines.push(`## ${capitalize(cat)}s (${items.length})`)
-      lines.push('')
-
-      for (const item of items) {
-        const date = item.created_at ? item.created_at.split('T')[0] : 'unknown'
-        lines.push(`### ${(item.content as string).slice(0, 80)}`)
-        lines.push('')
-        lines.push(`- **Date:** ${date}`)
-        lines.push(`- **Content:** ${item.content}`)
-        if (item.reasoning) lines.push(`- **Reasoning:** ${item.reasoning}`)
-        if (item.context) lines.push(`- **Context:** ${item.context}`)
-        const tags = parseTags(item.tags)
-        if (tags.length > 0) lines.push(`- **Tags:** ${tags.join(', ')}`)
-        lines.push('')
-        lines.push('---')
-        lines.push('')
-      }
-    }
-
-    console.log(lines.join('\n'))
-  }
-}
-
-function parseTags(value: string | null): string[] {
-  if (!value) return []
-  try {
-    const parsed = JSON.parse(value)
-    return Array.isArray(parsed) ? parsed : []
-  } catch {
-    return value.split(',').map(s => s.trim()).filter(Boolean)
-  }
-}
-
-function capitalize(s: string): string {
-  return s.charAt(0).toUpperCase() + s.slice(1)
-}
+/**
+ * Export Command — Phase 30
+ * Export observations for a project in markdown or JSON format.
+ *
+ * Usage:
+ *   claude-brain export --project expense-tracker --format md > decisions.md
+ *   claude-brain export --project expense-tracker --format json > decisions.json
+ *   claude-brain export --project expense-tracker --category decision
+ */
+
+import { join } from 'node:path'
+import { existsSync } from 'node:fs'
+import { parseArgs } from 'citty'
+import { getClaudeBrainHome } from '@/config/home'
+
+export async function runExport() {
+  const args = parseArgs(process.argv.slice(3), {
+    project: { type: 'string', description: 'Project name to export' },
+    format: { type: 'string', default: 'md', description: 'Output format: md or json' },
+    category: { type: 'string', description: 'Filter by category: decision, pattern, correction' },
+    limit: { type: 'string', default: '1000', description: 'Max observations to export' },
+  })
+
+  const project = args.project
+  const format = args.format || 'md'
+  const category = args.category
+  const limit = parseInt(args.limit || '1000', 10)
+
+  if (!project) {
+    console.error('Error: --project is required')
+    console.error('Usage: claude-brain export --project <name> [--format md|json] [--category decision|pattern|correction]')
+    process.exit(1)
+  }
+
+  const dbPath = join(getClaudeBrainHome(), 'data', 'memory.db')
+  if (!existsSync(dbPath)) {
+    console.error('Error: No database found. Run claude-brain first to initialize.')
+    process.exit(1)
+  }
+
+  const { Database } = await import('bun:sqlite')
+  const db = new Database(dbPath, { readonly: true })
+
+  let sql = 'SELECT * FROM observations WHERE project = ? AND archived = 0'
+  const params: (string | number)[] = [project]
+
+  if (category) {
+    sql += ' AND category = ?'
+    params.push(category)
+  }
+
+  sql += ' ORDER BY created_at DESC LIMIT ?'
+  params.push(limit)
+
+  const rows = db.prepare(sql).all(...params) as Record<string, unknown>[]
+  db.close()
+
+  if (rows.length === 0) {
+    console.error(`No observations found for project "${project}"${category ? ` with category "${category}"` : ''}.`)
+    process.exit(0)
+  }
+
+  if (format === 'json') {
+    const output = rows.map(row => ({
+      id: row.id,
+      project: row.project,
+      category: row.category,
+      content: row.content,
+      reasoning: row.reasoning || null,
+      context: row.context || null,
+      confidence: row.confidence,
+      source: row.source,
+      tags: parseTags(row.tags),
+      created_at: row.created_at,
+      updated_at: row.updated_at,
+    }))
+    console.log(JSON.stringify(output, null, 2))
+  } else {
+    // Markdown format
+    const lines: string[] = []
+    lines.push(`# ${project} — Observations Export`)
+    lines.push('')
+    lines.push(`> Exported ${rows.length} observation(s) on ${new Date().toISOString().split('T')[0]}`)
+    lines.push('')
+
+    // Group by category
+    const grouped: Record<string, Record<string, unknown>[]> = {}
+    for (const row of rows) {
+      const cat = row.category || 'other'
+      if (!grouped[cat]) grouped[cat] = []
+      grouped[cat].push(row)
+    }
+
+    for (const [cat, items] of Object.entries(grouped)) {
+      lines.push(`## ${capitalize(cat)}s (${items.length})`)
+      lines.push('')
+
+      for (const item of items) {
+        const date = item.created_at ? item.created_at.split('T')[0] : 'unknown'
+        lines.push(`### ${(item.content as string).slice(0, 80)}`)
+        lines.push('')
+        lines.push(`- **Date:** ${date}`)
+        lines.push(`- **Content:** ${item.content}`)
+        if (item.reasoning) lines.push(`- **Reasoning:** ${item.reasoning}`)
+        if (item.context) lines.push(`- **Context:** ${item.context}`)
+        const tags = parseTags(item.tags)
+        if (tags.length > 0) lines.push(`- **Tags:** ${tags.join(', ')}`)
+        lines.push('')
+        lines.push('---')
+        lines.push('')
+      }
+    }
+
+    console.log(lines.join('\n'))
+  }
+}
+
+function parseTags(value: string | null): string[] {
+  if (!value) return []
+  try {
+    const parsed = JSON.parse(value)
+    return Array.isArray(parsed) ? parsed : []
+  } catch {
+    return value.split(',').map(s => s.trim()).filter(Boolean)
+  }
+}
+
+function capitalize(s: string): string {
+  return s.charAt(0).toUpperCase() + s.slice(1)
+}