contentbase 0.0.2 → 0.0.4

package/src/cli/commands/create.ts

@@ -1,59 +1,185 @@
-import { defineCommand } from "citty";
-import fs from "fs/promises";
-import path from "path";
-import { loadCollection } from "../load-collection";
-import { kebabCase } from "../../utils/inflect";
-
-export default defineCommand({
-  meta: {
-    name: "create",
-    description: "Create a new document for a model type",
-  },
-  args: {
-    model: {
-      type: "positional",
-      description: "Model name",
-      required: true,
-    },
-    title: {
-      type: "string",
-      description: "Document title",
-      required: true,
-    },
-    rootPath: {
-      type: "string",
-      description: "Root path for the collection",
-      alias: "r",
-    },
-  },
-  async run({ args }) {
-    const collection = await loadCollection({
-      rootPath: args.rootPath as string | undefined,
-    });
-
-    const modelName = args.model as string;
-    const title = args.title as string;
-    const def = collection.getModelDefinition(modelName);
-
-    if (!def) {
-      console.error(
-        `Model "${modelName}" not found. Available: ${collection.modelDefinitions.map((d) => d.name).join(", ")}`
-      );
-      process.exit(1);
+import { z } from 'zod'
+import fs from 'fs/promises'
+import path from 'path'
+import matter from 'gray-matter'
+import { commands } from '../registry.js'
+import { loadCollection } from '../load-collection.js'
+import { kebabCase } from '../../utils/inflect.js'
+import { introspectMetaSchema } from '../../collection.js'
+
+const argsSchema = z.object({
+  title: z.string().optional(),
+  contentFolder: z.string().optional(),
+})
+
+async function handler(options: z.infer<typeof argsSchema>, context: { container: any }) {
+  const collection = await loadCollection({
+    contentFolder: options.contentFolder,
+  })
+
+  const modelName = context.container.argv._[1] as string | undefined
+  if (!modelName) {
+    console.error('Usage: cnotes create <model> --title "Document Title"')
+    process.exit(1)
+  }
+
+  const title = options.title
+  if (!title) {
+    console.error('--title is required')
+    process.exit(1)
+  }
+
+  const def =
+    collection.getModelDefinition(modelName) ??
+    collection.modelDefinitions.find(
+      (d) => d.name.toLowerCase() === modelName.toLowerCase()
+    )
+
+  if (!def) {
+    console.error(
+      `Model "${modelName}" not found. Available: ${collection.modelDefinitions.map((d) => d.name).join(', ')}`
+    )
+    process.exit(1)
+  }
+
+  // Parse --meta.* flags from raw argv
+  const metaOverrides: Record<string, unknown> = {}
+  const rawArgs = context.container.argv
+  for (const key of Object.keys(rawArgs)) {
+    if (key.startsWith('meta.')) {
+      metaOverrides[key.slice(5)] = rawArgs[key]
     }
+  }
+
+  // Build meta from priority layers: zod defaults < definition.defaults < template frontmatter < CLI overrides
+  const zodDefaults: Record<string, unknown> = {}
+  for (const field of introspectMetaSchema(def.meta)) {
+    if (field.defaultValue !== undefined) {
+      zodDefaults[field.name] = field.defaultValue
+    }
+  }
+
+  const definitionDefaults: Record<string, unknown> = def.defaults ?? {}
 
-    const slug = kebabCase(title.toLowerCase());
-    const pathId = `${def.prefix}/${slug}`;
-    const filePath = path.resolve(
+  // Template lookup
+  const templateExtensions = ['md', 'mdx']
+  let templateContent: string | null = null
+
+  for (const ext of templateExtensions) {
+    const templatePath = path.resolve(
       collection.rootPath,
-      `${pathId}.mdx`
-    );
+      'templates',
+      `${modelName.toLowerCase()}.${ext}`
+    )
+    try {
+      templateContent = await fs.readFile(templatePath, 'utf8')
+      break
+    } catch {
+      // template not found, try next extension
+    }
+  }
+
+  let content: string
+
+  if (templateContent) {
+    const parsed = matter(templateContent)
+
+    const mergedMeta = {
+      ...zodDefaults,
+      ...definitionDefaults,
+      ...parsed.data,
+      ...metaOverrides,
+    }
+
+    const body = parsed.content.replace(/^# .+$/m, `# ${title}`)
+    content = matter.stringify(body, mergedMeta)
+  } else {
+    const mergedMeta = {
+      ...zodDefaults,
+      ...definitionDefaults,
+      ...metaOverrides,
+    }
+
+    const lines: string[] = []
+    lines.push(`# ${title}`)
+    lines.push('')
+
+    const sections = def.sections ?? {}
+    for (const [, sec] of Object.entries(sections)) {
+      const s = sec as any
+      lines.push(`## ${s.heading}`)
+      lines.push('')
+      if (s.schema?.description) {
+        lines.push(s.schema.description)
+        lines.push('')
+      }
+    }
+
+    content = matter.stringify(lines.join('\n'), mergedMeta)
+  }
+
+  const slug = kebabCase(title.toLowerCase())
+  const pathId = `${def.prefix}/${slug}`
+  const filePath = path.resolve(collection.rootPath, `${pathId}.md`)
+
+  await fs.mkdir(path.dirname(filePath), { recursive: true })
+  await fs.writeFile(filePath, content, 'utf8')
+
+  console.log(`Created ${filePath}`)
+}
+
+commands.register('create', {
+  description: 'Create a new document for a model type',
+  help: `# cnotes create
+
+Create a new document for a registered model, with proper frontmatter defaults and section scaffolding.
+
+## Usage
+
+\`\`\`
+cnotes create <model> --title "Document Title" [options]
+\`\`\`
+
+## Arguments
+
+| Argument | Description |
+|----------|-------------|
+| \`model\` | Model name (e.g. \`Post\`, \`Epic\`, \`Task\`) |
+
+## Options
+
+| Option | Description |
+|--------|-------------|
+| \`--title\` | **Required.** Title for the new document (used as H1 and slug) |
+| \`--meta.*\` | Set frontmatter fields (e.g. \`--meta.status active\`) |
+| \`--contentFolder\` | Path to content folder |
+
+## Template Lookup
+
+If \`templates/<model>.md\` (or \`.mdx\`) exists in the content root, it will be used as the document scaffold. Template frontmatter is merged with schema defaults and CLI overrides.
+
+## Meta Priority
+
+Frontmatter values are merged in this order (last wins):
+
+1. Zod schema defaults
+2. Model definition defaults
+3. Template frontmatter
+4. CLI \`--meta.*\` overrides
+
+## Examples
 
-    const content = `---\n---\n\n# ${title}\n`;
+\`\`\`bash
+# Create a new post
+cnotes create Post --title "My First Post"
 
-    await fs.mkdir(path.dirname(filePath), { recursive: true });
-    await fs.writeFile(filePath, content, "utf8");
+# Create with meta overrides
+cnotes create Task --title "Fix login bug" --meta.status active --meta.priority high
 
-    console.log(`Created ${filePath}`);
-  },
-});
+# Create from a different content folder
+cnotes create Epic --title "Auth System" --contentFolder ./docs
+\`\`\`
+`,
+  argsSchema,
+  handler,
+})

package/src/cli/commands/embed.ts

@@ -0,0 +1,323 @@
+import { z } from 'zod'
+import { existsSync, readdirSync, accessSync, constants as fsConstants } from 'node:fs'
+import path from 'node:path'
+import { commands } from '../registry.js'
+import { loadCollection } from '../load-collection.js'
+
+const argsSchema = z.object({
+  force: z.boolean().default(false),
+  provider: z.enum(['local', 'openai']).default('openai'),
+  status: z.boolean().default(false),
+  local: z.boolean().default(false),
+  installLocal: z.boolean().default(false),
+  contentFolder: z.string().optional(),
+})
+
+function hasSearchIndex(rootPath: string): boolean {
+  const dbDir = path.join(rootPath, '.contentbase')
+  if (!existsSync(dbDir)) return false
+  try {
+    const files = readdirSync(dbDir) as string[]
+    return files.some((f: string) => f.startsWith('search.') && f.endsWith('.sqlite'))
+  } catch {
+    return false
+  }
+}
+
+function collectDocumentInputs(collection: any) {
+  const inputs: any[] = []
+  for (const pathId of collection.available) {
+    const doc = collection.document(pathId)
+    const modelDef = collection.findModelDefinition(pathId)
+
+    const sections: any[] = []
+    const lines = (doc.content as string).split('\n')
+    let currentHeading: string | null = null
+    let currentContent: string[] = []
+
+    for (const line of lines) {
+      const h2Match = line.match(/^## (.+)/)
+      if (h2Match) {
+        if (currentHeading) {
+          sections.push({
+            heading: currentHeading,
+            headingPath: currentHeading,
+            content: currentContent.join('\n').trim(),
+            level: 2,
+          })
+        }
+        currentHeading = h2Match[1].trim()
+        currentContent = []
+      } else if (currentHeading) {
+        currentContent.push(line)
+      }
+    }
+    if (currentHeading) {
+      sections.push({
+        heading: currentHeading,
+        headingPath: currentHeading,
+        content: currentContent.join('\n').trim(),
+        level: 2,
+      })
+    }
+
+    inputs.push({
+      pathId,
+      model: modelDef?.name ?? undefined,
+      title: doc.title,
+      slug: doc.slug,
+      meta: doc.meta,
+      content: doc.content,
+      sections: sections.length > 0 ? sections : undefined,
+    })
+  }
+  return inputs
+}
+
+function isLocalInstalled(): boolean {
+  const modulePath = path.join(process.cwd(), 'node_modules', 'node-llama-cpp')
+  try {
+    return existsSync(modulePath)
+  } catch {
+    return false
+  }
+}
+
+async function installLocal(SemanticSearch: any, container: any): Promise<void> {
+  const cwd = process.cwd()
+  const nodeModules = path.join(cwd, 'node_modules')
+
+  // Check permissions
+  try {
+    if (existsSync(nodeModules)) {
+      accessSync(nodeModules, fsConstants.W_OK)
+    }
+  } catch {
+    const version = SemanticSearch.PINNED_LLAMA_VERSION
+    const cmd = detectInstallCommand(cwd, version)
+    console.error(`Cannot write to ${nodeModules}.`)
+    console.error(`Run manually with elevated permissions:\n  ${cmd}`)
+    process.exit(1)
+  }
+
+  console.error(`Installing node-llama-cpp@${SemanticSearch.PINNED_LLAMA_VERSION}...`)
+
+  // Create a temporary instance to use the install method
+  if (!container.features.available.includes('semanticSearch')) {
+    SemanticSearch.attach(container)
+  }
+  const ss = container.feature('semanticSearch', {
+    dbPath: path.join(cwd, '.contentbase/search.sqlite'),
+    embeddingProvider: 'local',
+  })
+
+  try {
+    await ss.installLocalEmbeddings(cwd)
+    console.log(`node-llama-cpp@${SemanticSearch.PINNED_LLAMA_VERSION} installed successfully.`)
+  } catch (err: any) {
+    const msg = err?.message ?? String(err)
+    if (msg.includes('network') || msg.includes('ETIMEDOUT') || msg.includes('ENOTFOUND')) {
+      console.error(`Install failed (network unreachable). To use local embeddings offline, manually install node-llama-cpp@${SemanticSearch.PINNED_LLAMA_VERSION} into this project.`)
+    } else {
+      console.error(msg)
+    }
+    process.exit(1)
+  }
+}
+
+function detectInstallCommand(cwd: string, version: string): string {
+  const pkg = `node-llama-cpp@${version}`
+  if (existsSync(path.join(cwd, 'bun.lockb')) || existsSync(path.join(cwd, 'bun.lock'))) {
+    return `bun add --optional ${pkg}`
+  } else if (existsSync(path.join(cwd, 'pnpm-lock.yaml'))) {
+    return `pnpm add --save-optional ${pkg}`
+  } else if (existsSync(path.join(cwd, 'yarn.lock'))) {
+    return `yarn add --optional ${pkg}`
+  }
+  return `npm install --save-optional ${pkg}`
+}
+
+async function handler(options: z.infer<typeof argsSchema>, { container }: { container: any }) {
+  const { SemanticSearch } = await import('@soederpop/luca/agi')
+
+  // --install-local: install node-llama-cpp only, then exit
+  if (options.installLocal) {
+    if (isLocalInstalled()) {
+      console.log('node-llama-cpp is already installed. Skipping.')
+      return
+    }
+    await installLocal(SemanticSearch, container)
+    return
+  }
+
+  // Resolve effective provider: --local flag overrides --provider
+  const provider = options.local ? 'local' : options.provider
+
+  // --local: auto-install if needed
+  if (options.local && !isLocalInstalled()) {
+    await installLocal(SemanticSearch, container)
+  }
+
+  const collection = await loadCollection({
+    contentFolder: options.contentFolder,
+  })
+
+  // --status: show index health without modifying anything
+  if (options.status) {
+    if (!hasSearchIndex(collection.rootPath)) {
+      console.log('No search index found.')
+      console.log(`Documents in collection: ${collection.available.length}`)
+      console.log(`\nRun: cnotes embed`)
+      return
+    }
+
+    if (!container.features.available.includes('semanticSearch')) {
+      SemanticSearch.attach(container)
+    }
+    const dbPath = path.join(collection.rootPath, '.contentbase/search.sqlite')
+    const ss = container.feature('semanticSearch', {
+      dbPath,
+      embeddingProvider: provider,
+    })
+    await ss.initDb()
+
+    const stats = ss.getStats()
+    console.log('Search Index Status')
+    console.log('-------------------')
+    console.log(`Documents indexed:  ${stats.documentCount}`)
+    console.log(`Chunks created:     ${stats.chunkCount}`)
+    console.log(`Embeddings stored:  ${stats.embeddingCount}`)
+    console.log(`Last indexed:       ${stats.lastIndexedAt ?? 'never'}`)
+    console.log(`Provider:           ${stats.provider}`)
+    console.log(`Model:              ${stats.model}`)
+    console.log(`Dimensions:         ${stats.dimensions}`)
+    console.log(`Database size:      ${formatBytes(stats.dbSizeBytes)}`)
+    console.log(`Collection docs:    ${collection.available.length}`)
+
+    // Check for staleness
+    const docs = collectDocumentInputs(collection)
+    const stale = docs.filter((d: any) => ss.needsReindex(d)).length
+    if (stale > 0) {
+      console.log(`Stale documents:    ${stale}`)
+    } else {
+      console.log(`Status:             up to date`)
+    }
+    return
+  }
+
+  // Embed / re-embed
+  if (!container.features.available.includes('semanticSearch')) {
+    SemanticSearch.attach(container)
+  }
+  const dbPath = path.join(collection.rootPath, '.contentbase/search.sqlite')
+  const ss = container.feature('semanticSearch', {
+    dbPath,
+    embeddingProvider: provider,
+  })
+  await ss.initDb()
+
+  const docs = collectDocumentInputs(collection)
+  const startTime = Date.now()
+
+  if (options.force) {
+    console.error('Clearing existing index...')
+    await ss.reindex()
+  }
+
+  const toIndex = options.force ? docs : docs.filter((d: any) => ss.needsReindex(d))
+
+  // Remove stale
+  ss.removeStale(docs.map((d: any) => d.pathId))
+
+  if (toIndex.length === 0) {
+    console.log('All documents are up to date. Nothing to embed.')
+    const stats = ss.getStats()
+    printStats(stats, 0, Date.now() - startTime)
+    return
+  }
+
+  console.error(`Embedding ${toIndex.length} of ${docs.length} document(s)...`)
+  console.error(`Provider: ${provider}`)
+
+  const batchSize = 5
+
+  for (let i = 0; i < toIndex.length; i += batchSize) {
+    const batch = toIndex.slice(i, i + batchSize)
+    await ss.indexDocuments(batch)
+    const done = Math.min(i + batchSize, toIndex.length)
+    const pct = Math.round((done / toIndex.length) * 100)
+    const elapsed = Date.now() - startTime
+    const eta = toIndex.length > done
+      ? Math.round((elapsed / done) * (toIndex.length - done) / 1000)
+      : 0
+    const etaStr = eta > 0 ? ` ETA: ${eta}s` : ''
+    console.error(`  [${pct}%] ${done}/${toIndex.length} documents${etaStr}`)
+  }
+
+  const elapsed = Date.now() - startTime
+  const stats = ss.getStats()
+  printStats(stats, toIndex.length, elapsed)
+}
+
+function printStats(stats: any, indexed: number, elapsedMs: number) {
+  console.log()
+  console.log('Embedding complete.')
+  if (indexed > 0) console.log(`  Documents indexed: ${indexed}`)
+  console.log(`  Total documents:   ${stats.documentCount}`)
+  console.log(`  Total chunks:      ${stats.chunkCount}`)
+  console.log(`  Embeddings:        ${stats.embeddingCount}`)
+  console.log(`  Time elapsed:      ${(elapsedMs / 1000).toFixed(1)}s`)
+  console.log(`  Database size:     ${formatBytes(stats.dbSizeBytes)}`)
+}
+
+function formatBytes(bytes: number): string {
+  if (bytes < 1024) return `${bytes} B`
+  if (bytes < 1024 * 1024) return `${(bytes / 1024).toFixed(1)} KB`
+  return `${(bytes / (1024 * 1024)).toFixed(1)} MB`
+}
+
+commands.register('embed', {
+  description: 'Generate embeddings for collection documents',
+  help: `# cnotes embed
+
+Generate or update vector embeddings for all documents in the collection. Required before using \`cnotes search\`.
+
+## Usage
+
+\`\`\`
+cnotes embed [options]
+\`\`\`
+
+## Options
+
+| Option | Default | Description |
+|--------|---------|-------------|
+| \`--force\` | \`false\` | Re-embed everything (ignore content hashes) |
+| \`--provider\` | \`openai\` | Embedding provider: \`openai\` or \`local\` |
+| \`--status\` | \`false\` | Show index health without embedding |
+| \`--local\` | \`false\` | Use local embeddings; auto-installs node-llama-cpp if not found |
+| \`--install-local\` | \`false\` | Only install node-llama-cpp (no embedding), then exit |
+| \`--contentFolder\` | | Path to content folder |
+
+## Examples
+
+\`\`\`bash
+# Generate/update embeddings (uses OpenAI by default)
+cnotes embed
+
+# Re-embed everything from scratch
+cnotes embed --force
+
+# Use local embeddings (auto-installs node-llama-cpp if needed)
+cnotes embed --local
+
+# Only install node-llama-cpp without embedding
+cnotes embed --install-local
+
+# Show index health
+cnotes embed --status
+\`\`\`
+`,
+  argsSchema,
+  handler,
+})

package/src/cli/commands/export.ts

@@ -1,24 +1,58 @@
-import { defineCommand } from "citty";
-import { loadCollection } from "../load-collection";
-
-export default defineCommand({
-  meta: {
-    name: "export",
-    description: "Export collection as JSON",
-  },
-  args: {
-    rootPath: {
-      type: "string",
-      description: "Root path for the collection",
-      alias: "r",
-    },
-  },
-  async run({ args }) {
-    const collection = await loadCollection({
-      rootPath: args.rootPath as string | undefined,
-    });
-
-    const data = await collection.export();
-    console.log(JSON.stringify(data, null, 2));
-  },
-});
+import { z } from 'zod'
+import { commands } from '../registry.js'
+import { loadCollection } from '../load-collection.js'
+
+const argsSchema = z.object({
+  contentFolder: z.string().optional(),
+})
+
+async function handler(options: z.infer<typeof argsSchema>) {
+  const collection = await loadCollection({
+    contentFolder: options.contentFolder,
+  })
+
+  const data = await collection.export()
+  console.log(JSON.stringify(data, null, 2))
+}
+
+commands.register('export', {
+  description: 'Export collection as JSON',
+  help: `# cnotes export
+
+Export the entire collection as a JSON object. Each document includes its path ID, title, frontmatter, content, and model info.
+
+## Usage
+
+\`\`\`
+cnotes export [options]
+\`\`\`
+
+## Options
+
+| Option | Description |
+|--------|-------------|
+| \`--contentFolder\` | Path to content folder |
+
+## Output
+
+Writes JSON to stdout. Pipe to a file or another tool.
+
+## Examples
+
+\`\`\`bash
+# Export to stdout
+cnotes export
+
+# Save to a file
+cnotes export > backup.json
+
+# Export a specific content folder
+cnotes export --contentFolder ./docs > docs.json
+
+# Pipe to jq for filtering
+cnotes export | jq '.[] | select(.meta.status == "published")'
+\`\`\`
+`,
+  argsSchema,
+  handler,
+})