Apple Tools MCP

/** * Real data loader for performance tests * Connects to actual LanceDB index and macOS data sources */ import fs from 'fs' import path from 'path' import { connect } from '@lancedb/lancedb' import { pipeline } from '@xenova/transformers' // Real paths const DATA_DIR = path.join(process.env.HOME, '.apple-tools-mcp') const DB_PATH = path.join(DATA_DIR, 'vector-index') let db = null let embedder = null let tables = {} /** * Check if real data index exists */ export function isRealDataAvailable() { return fs.existsSync(DB_PATH) } /** * Get the real embedder (Xenova/all-MiniLM-L6-v2) */ export async function getRealEmbedder() { if (!embedder) { embedder = await pipeline('feature-extraction', 'Xenova/all-MiniLM-L6-v2') } return embedder } /** * Generate embedding using real model */ export async function realEmbed(text) { const emb = await getRealEmbedder() const output = await emb(text, { pooling: 'mean', normalize: true }) return Array.from(output.data) } /** * Batch embed using real model */ export async function realEmbedBatch(texts) { if (texts.length === 0) return [] if (texts.length === 1) return [await realEmbed(texts[0])] const emb = await getRealEmbedder() const result = await emb(texts, { pooling: 'mean', normalize: true }) const EMBEDDING_DIM = 384 const embeddings = [] for (let i = 0; i < texts.length; i++) { const start = i * EMBEDDING_DIM const end = start + EMBEDDING_DIM embeddings.push(Array.from(result.data.slice(start, end))) } return embeddings } /** * Connect to real LanceDB */ export async function connectToRealDB() { if (!db && isRealDataAvailable()) { db = await connect(DB_PATH) } return db } /** * Get a table from real DB */ export async function getRealTable(tableName) { if (tables[tableName]) return tables[tableName] const database = await connectToRealDB() if (!database) return null try { const tableNames = await database.tableNames() if (!tableNames.includes(tableName)) return null tables[tableName] = await database.openTable(tableName) return tables[tableName] } catch (e) { console.error(`Error opening table ${tableName}:`, e.message) return null } } /** * Get real emails from the index */ export async function getRealEmails(limit = 1000) { const table = await getRealTable('emails') if (!table) return [] try { const results = await table.query().limit(limit).toArray() return results.map(r => ({ id: r.id, path: r.filePath, from: r.from, fromEmail: r.fromEmail, to: r.to, subject: r.subject, body: r.body, date: r.date, timestamp: r.dateTimestamp, hasAttachment: r.hasAttachment, mailbox: r.mailbox, isSent: r.isSent, isFlagged: r.isFlagged, vector: r.vector })) } catch (e) { console.error('Error loading real emails:', e.message) return [] } } /** * Get real messages from the index */ export async function getRealMessages(limit = 500) { const table = await getRealTable('messages') if (!table) return [] try { const results = await table.query().limit(limit).toArray() return results.map(r => ({ id: r.id, ROWID: r.id, text: r.text, sender: r.sender, date: r.date, timestamp: r.dateTimestamp, chatId: r.chatId, chatIdentifier: r.chatIdentifier, chatName: r.chatName, participantCount: r.participantCount || 2, isGroup: r.isGroup, attachmentCount: r.attachmentCount || 0, vector: r.vector })) } catch (e) { console.error('Error loading real messages:', e.message) return [] } } /** * Get real calendar events from the index */ export async function getRealCalendarEvents(limit = 200) { const table = await getRealTable('calendar') if (!table) return [] try { const results = await table.query().limit(limit).toArray() return results.map(r => ({ id: r.id, ROWID: r.id, title: r.title, summary: r.title, start: r.start, end: r.end, startTimestamp: r.startTimestamp, endTimestamp: r.endTimestamp, isAllDay: r.isAllDay, all_day: r.isAllDay ? 1 : 0, calendar: r.calendar, calendar_name: r.calendar, location: r.location, notes: r.notes, vector: r.vector })) } catch (e) { console.error('Error loading real calendar events:', e.message) return [] } } /** * Perform real vector search */ export async function realVectorSearch(tableName, query, limit = 20) { const table = await getRealTable(tableName) if (!table) return [] try { const embedding = await realEmbed(query) const results = await table.search(embedding).limit(limit).toArray() return results } catch (e) { console.error(`Error searching ${tableName}:`, e.message) return [] } } /** * Get table statistics */ export async function getTableStats() { const stats = {} for (const tableName of ['emails', 'messages', 'calendar']) { const table = await getRealTable(tableName) if (table) { try { stats[tableName] = await table.countRows() } catch { stats[tableName] = 0 } } else { stats[tableName] = 0 } } return stats } /** * Create performance mocks that use real data */ export function createRealDataMocks() { return { embedder: { embedder: async (texts) => { const vectors = await realEmbedBatch(Array.isArray(texts) ? texts : [texts]) return { data: new Float32Array(vectors.flat()) } }, pipeline: async () => getRealEmbedder() }, lancedb: { connect: connectToRealDB, tables }, // Real data loaders getEmails: getRealEmails, getMessages: getRealMessages, getCalendarEvents: getRealCalendarEvents, search: realVectorSearch } } /** * Cleanup resources */ export async function cleanup() { db = null tables = {} embedder = null }