@alanhelmick/memorable/dist/index.cjs.map

An AI memory system enabling personalized, context-aware interactions through advanced memory management and emotional intelligence

{"version":3,"file":"index.cjs","sources":["../src/utils/logger.js","../src/constants/emotions.js","../src/services/humeService.js","../src/services/videoStreamService.js","../src/config/redis.js","../src/services/emotionalContextService.js","../src/config/database.js","../src/services/customModelService.js","../src/services/modelSelectionService.js","../src/utils/healthCheck.js","../src/config/weaviate.js","../src/index.js"],"sourcesContent":["import winston from 'winston';\n\nconst logLevels = {\n  error: 0,\n  warn: 1,\n  info: 2,\n  debug: 3,\n};\n\nconst logColors = {\n  error: 'red',\n  warn: 'yellow',\n  info: 'green',\n  debug: 'blue',\n};\n\nlet logger = null;\n\nexport async function setupLogger() {\n  if (logger) {\n    return logger;\n  }\n  \n  await createLogsDirectory();\n\n  // Add colors to Winston\n  winston.addColors(logColors);\n\n  const format = winston.format.combine(\n    winston.format.timestamp({ format: 'YYYY-MM-DD HH:mm:ss' }),\n    winston.format.errors({ stack: true }),\n    winston.format.colorize({ all: true }),\n    winston.format.printf(({ level, message, timestamp, stack }) => {\n      if (stack) {\n        return `${timestamp} ${level}: ${message}\\n${stack}`;\n      }\n      return `${timestamp} ${level}: ${message}`;\n    })\n  );\n\n  logger = winston.createLogger({\n    level: process.env.LOG_LEVEL || 'info',\n    levels: logLevels,\n    format,\n    transports: [\n      // Console transport\n      new winston.transports.Console(),\n      \n      // File transport for errors\n      new winston.transports.File({\n        filename: 'logs/error.log',\n        level: 'error',\n        format: winston.format.uncolorize(),\n      }),\n      \n      // File transport for all logs\n      new winston.transports.File({\n        filename: 'logs/combined.log',\n        format: winston.format.uncolorize(),\n      }),\n    ],\n    exceptionHandlers: [\n      new winston.transports.File({\n        filename: 'logs/exceptions.log',\n        format: winston.format.uncolorize(),\n      }),\n    ],\n    rejectionHandlers: [\n      new winston.transports.File({\n        filename: 'logs/rejections.log',\n        format: winston.format.uncolorize(),\n      }),\n    ],\n  });\n\n  // Create a stream for Morgan HTTP logging\n  logger.stream = {\n    write: (message) => {\n      logger.info(message.trim());\n    },\n  };\n\n  return logger;\n}\n\nexport async function createLogsDirectory() {\n  const { mkdir } = await import('fs/promises');\n  try {\n    await mkdir('logs', { recursive: true });\n  } catch (error) {\n    if (error.code !== 'EEXIST') {\n      console.error('Failed to create logs directory:', error);\n    }\n  }\n}\n\nexport { logger };","export const expressionColors = {\n  admiration: \"#ffc58f\",\n  adoration: \"#ffc6cc\",\n  aestheticAppreciation: \"#e2cbff\",\n  amusement: \"#febf52\",\n  anger: \"#b21816\",\n  annoyance: \"#ffffff\",\n  anxiety: \"#6e42cc\",\n  awe: \"#7dabd3\",\n  awkwardness: \"#d7d99d\",\n  boredom: \"#a4a4a4\",\n  calmness: \"#a9cce1\",\n  concentration: \"#336cff\",\n  contemplation: \"#b0aeef\",\n  confusion: \"#c66a26\",\n  contempt: \"#76842d\",\n  contentment: \"#e5c6b4\",\n  craving: \"#54591c\",\n  determination: \"#ff5c00\",\n  disappointment: \"#006c7c\",\n  disapproval: \"#ffffff\",\n  disgust: \"#1a7a41\",\n  distress: \"#c5f264\",\n  doubt: \"#998644\",\n  ecstasy: \"#ff48a4\",\n  embarrassment: \"#63c653\",\n  empathicPain: \"#ca5555\",\n  enthusiasm: \"#ffffff\",\n  entrancement: \"#7554d6\",\n  envy: \"#1d4921\",\n  excitement: \"#fff974\",\n  fear: \"#d1c9ef\",\n  gratitude: \"#ffffff\",\n  guilt: \"#879aa1\",\n  horror: \"#772e7a\",\n  interest: \"#a9cce1\",\n  joy: \"#ffd600\",\n  love: \"#f44f4c\",\n  neutral: \"#879aa1\",\n  nostalgia: \"#b087a1\",\n  pain: \"#8c1d1d\",\n  pride: \"#9a4cb6\",\n  realization: \"#217aa8\",\n  relief: \"#fe927a\",\n  romance: \"#f0cc86\",\n  sadness: \"#305575\",\n  sarcasm: \"#ffffff\",\n  satisfaction: \"#a6ddaf\",\n  sexualDesire: \"#aa0d59\",\n  shame: \"#8a6262\",\n  surprise: \"#70e63a\",\n  surpriseNegative: \"#70e63a\",\n  surprisePositive: \"#7affff\",\n  sympathy: \"#7f88e0\",\n  tiredness: \"#757575\",\n  triumph: \"#ec8132\",\n};\n\nexport const isExpressionColor = (color) => {\n  return color in expressionColors;\n};\n\n// Get the number of emotional dimensions from the colors\nexport const EMOTION_DIMENSIONS = Object.keys(expressionColors).length;\n\n// Convert emotion to RGB vector for neural processing\nexport const emotionToVector = (emotion) => {\n  const color = expressionColors[emotion];\n  if (!color) return null;\n\n  // Convert hex to RGB\n  const r = parseInt(color.slice(1, 3), 16) / 255;\n  const g = parseInt(color.slice(3, 5), 16) / 255;\n  const b = parseInt(color.slice(5, 7), 16) / 255;\n\n  return [r, g, b];\n};\n\n// Convert RGB vector back to closest emotion\nexport const vectorToEmotion = (vector) => {\n  if (!vector || vector.length !== 3) return 'neutral';\n\n  // Convert vector back to hex color\n  const toHex = (n) => {\n    const hex = Math.round(n * 255).toString(16);\n    return hex.length === 1 ? '0' + hex : hex;\n  };\n  const color = `#${toHex(vector[0])}${toHex(vector[1])}${toHex(vector[2])}`;\n\n  // Find closest matching emotion\n  let closestEmotion = 'neutral';\n  let minDistance = Infinity;\n\n  for (const [emotion, emotionColor] of Object.entries(expressionColors)) {\n    const distance = colorDistance(color, emotionColor);\n    if (distance < minDistance) {\n      minDistance = distance;\n      closestEmotion = emotion;\n    }\n  }\n\n  return closestEmotion;\n};\n\n// Calculate Euclidean distance between two hex colors\nconst colorDistance = (color1, color2) => {\n  const r1 = parseInt(color1.slice(1, 3), 16);\n  const g1 = parseInt(color1.slice(3, 5), 16);\n  const b1 = parseInt(color1.slice(5, 7), 16);\n\n  const r2 = parseInt(color2.slice(1, 3), 16);\n  const g2 = parseInt(color2.slice(3, 5), 16);\n  const b2 = parseInt(color2.slice(5, 7), 16);\n\n  return Math.sqrt(\n    Math.pow(r1 - r2, 2) +\n    Math.pow(g1 - g2, 2) +\n    Math.pow(b1 - b2, 2)\n  );\n};","import WebSocket from 'ws';\nimport { logger } from '../utils/logger.js';\nimport { expressionColors, emotionToVector } from '../constants/emotions.js';\n\nexport class HumeService {\n  constructor() {\n    this.ws = null;\n    this.apiKey = process.env.HUME_API_KEY;\n    this.endpoint = process.env.HUME_ENDPOINT;\n    this.isConnected = false;\n    this.activeStreams = new Map();\n    this.messageQueue = [];\n    this.processingQueue = false;\n    this.lastActivityTime = Date.now();\n    this.inactivityTimeout = 60000; // 1 minute inactivity timeout\n    this.reconnectAttempts = 0;\n    this.maxReconnectAttempts = 5;\n    this.reconnectDelay = 1000; // Start with 1 second\n  }\n\n  async connect(config = {}) {\n    if (!this.apiKey) {\n      throw new Error('Hume API key not configured');\n    }\n\n    return new Promise((resolve, reject) => {\n      try {\n        const params = new URLSearchParams({\n          apiKey: this.apiKey,\n          ...config\n        });\n\n        const wsUrl = `${this.endpoint}?${params.toString()}`;\n        this.ws = new WebSocket(wsUrl);\n\n        this.ws.on('open', () => {\n          logger.info('Connected to Hume.ai websocket');\n          this.isConnected = true;\n          this.reconnectAttempts = 0;\n          this.reconnectDelay = 1000;\n          this.setupInactivityCheck();\n          resolve();\n        });\n\n        this.ws.on('message', (data) => {\n          this.lastActivityTime = Date.now();\n          this.handleMessage(data);\n        });\n\n        this.ws.on('error', (error) => {\n          logger.error('Hume websocket error:', error);\n          this.handleError(error);\n        });\n\n        this.ws.on('close', () => {\n          logger.info('Hume websocket closed');\n          this.isConnected = false;\n          this.handleDisconnect();\n        });\n      } catch (error) {\n        reject(error);\n      }\n    });\n  }\n\n  setupInactivityCheck() {\n    setInterval(() => {\n      const inactiveTime = Date.now() - this.lastActivityTime;\n      if (inactiveTime >= this.inactivityTimeout) {\n        logger.warn('WebSocket inactive, reconnecting...');\n        this.reconnect();\n      }\n    }, 10000); // Check every 10 seconds\n  }\n\n  async reconnect() {\n    if (this.reconnectAttempts < this.maxReconnectAttempts) {\n      this.reconnectAttempts++;\n      this.reconnectDelay *= 2; // Exponential backoff\n      logger.info(`Attempting to reconnect in ${this.reconnectDelay}ms (attempt ${this.reconnectAttempts})`);\n      \n      setTimeout(async () => {\n        try {\n          await this.connect();\n          // Resubscribe active streams\n          for (const [streamId, config] of this.activeStreams) {\n            await this.startStream(streamId, config);\n          }\n        } catch (error) {\n          logger.error('Reconnection attempt failed:', error);\n        }\n      }, this.reconnectDelay);\n    } else {\n      logger.error('Max reconnection attempts reached');\n    }\n  }\n\n  async startStream(streamId, config) {\n    if (!this.isConnected) {\n      await this.connect();\n    }\n\n    const streamConfig = {\n      models: config.models || { language: {}, face: {}, prosody: {} },\n      raw_text: config.rawText || true,\n      reset_stream: config.resetStream || false\n    };\n\n    this.activeStreams.set(streamId, {\n      config: streamConfig,\n      callbacks: new Map(),\n      buffer: []\n    });\n\n    await this.sendMessage({\n      type: 'stream_start',\n      stream_id: streamId,\n      config: streamConfig\n    });\n\n    logger.info(`Started stream ${streamId}`);\n  }\n\n  async stopStream(streamId) {\n    const stream = this.activeStreams.get(streamId);\n    if (!stream) return;\n\n    await this.sendMessage({\n      type: 'stream_end',\n      stream_id: streamId\n    });\n\n    this.activeStreams.delete(streamId);\n    logger.info(`Stopped stream ${streamId}`);\n  }\n\n  async processText(text, streamId = null) {\n    const id = streamId || `text_${Date.now()}`;\n    if (!streamId) {\n      await this.startStream(id, { models: { language: {} } });\n    }\n\n    return this.sendData(id, {\n      type: 'text',\n      data: text\n    });\n  }\n\n  async processVoice(audioData, streamId = null) {\n    const id = streamId || `voice_${Date.now()}`;\n    if (!streamId) {\n      await this.startStream(id, { models: { prosody: {} } });\n    }\n\n    // Split audio into 5-second chunks\n    const chunks = this.splitAudioIntoChunks(audioData);\n    const results = [];\n\n    for (const chunk of chunks) {\n      const result = await this.sendData(id, {\n        type: 'prosody',\n        data: chunk.toString('base64')\n      });\n      results.push(result);\n    }\n\n    if (!streamId) {\n      await this.stopStream(id);\n    }\n\n    return this.mergeResults(results);\n  }\n\n  async processFacial(imageData, streamId = null) {\n    const id = streamId || `face_${Date.now()}`;\n    if (!streamId) {\n      await this.startStream(id, { models: { face: {} } });\n    }\n\n    const result = await this.sendData(id, {\n      type: 'face',\n      data: imageData.toString('base64')\n    });\n\n    if (!streamId) {\n      await this.stopStream(id);\n    }\n\n    return result;\n  }\n\n  splitAudioIntoChunks(audioData, chunkSize = 5000) {\n    // Split audio data into 5-second chunks\n    const chunks = [];\n    let offset = 0;\n    while (offset < audioData.length) {\n      chunks.push(audioData.slice(offset, offset + chunkSize));\n      offset += chunkSize;\n    }\n    return chunks;\n  }\n\n  mergeResults(results) {\n    // Merge multiple chunk results into a single result\n    const merged = {\n      emotions: new Map()\n    };\n\n    results.forEach(result => {\n      result.emotions.forEach(emotion => {\n        const existing = merged.emotions.get(emotion.name) || { score: 0, count: 0 };\n        existing.score += emotion.score;\n        existing.count += 1;\n        merged.emotions.set(emotion.name, existing);\n      });\n    });\n\n    // Average the scores\n    return Array.from(merged.emotions.entries()).map(([name, data]) => ({\n      name,\n      score: data.score / data.count\n    }));\n  }\n\n  async sendData(streamId, data) {\n    return new Promise((resolve, reject) => {\n      const messageId = Date.now().toString();\n      const stream = this.activeStreams.get(streamId);\n      \n      if (!stream) {\n        reject(new Error(`Stream ${streamId} not found`));\n        return;\n      }\n\n      stream.callbacks.set(messageId, (response) => {\n        if (response.error) {\n          reject(new Error(response.error));\n        } else {\n          resolve(this.processEmotions(response.emotions));\n        }\n      });\n\n      this.sendMessage({\n        id: messageId,\n        stream_id: streamId,\n        ...data\n      });\n    });\n  }\n\n  async sendMessage(message) {\n    if (!this.isConnected) {\n      throw new Error('WebSocket not connected');\n    }\n\n    this.messageQueue.push(message);\n    if (!this.processingQueue) {\n      await this.processMessageQueue();\n    }\n  }\n\n  async processMessageQueue() {\n    this.processingQueue = true;\n    while (this.messageQueue.length > 0) {\n      const message = this.messageQueue.shift();\n      try {\n        this.ws.send(JSON.stringify(message));\n        this.lastActivityTime = Date.now();\n        // Small delay to respect rate limits\n        await new Promise(resolve => setTimeout(resolve, 20));\n      } catch (error) {\n        logger.error('Failed to send message:', error);\n        this.messageQueue.unshift(message);\n        break;\n      }\n    }\n    this.processingQueue = false;\n  }\n\n  handleMessage(data) {\n    try {\n      const message = JSON.parse(data);\n      const stream = this.activeStreams.get(message.stream_id);\n      \n      if (stream && message.id && stream.callbacks.has(message.id)) {\n        const callback = stream.callbacks.get(message.id);\n        callback(message);\n        stream.callbacks.delete(message.id);\n      }\n    } catch (error) {\n      logger.error('Error handling Hume message:', error);\n    }\n  }\n\n  processEmotions(emotions) {\n    return emotions.map(emotion => ({\n      name: emotion.name,\n      score: emotion.score,\n      vector: emotionToVector(emotion.name),\n      color: expressionColors[emotion.name],\n      confidence: emotion.confidence || emotion.score\n    }))\n    .filter(emotion => emotion.confidence >= 0.1)\n    .sort((a, b) => b.score - a.score);\n  }\n\n  async close() {\n    // Stop all active streams\n    for (const streamId of this.activeStreams.keys()) {\n      await this.stopStream(streamId);\n    }\n\n    if (this.ws) {\n      this.ws.close();\n      this.ws = null;\n      this.isConnected = false;\n      logger.info('Hume websocket connection closed');\n    }\n  }\n}\n\n// Create singleton instance\nconst humeService = new HumeService();\n\nexport default humeService;","import { logger } from '../utils/logger.js';\nimport humeService from './humeService.js';\n\nexport class VideoStreamService {\n  constructor() {\n    this.activeStreams = new Map();\n    this.chunkDuration = 5000; // 5 seconds in milliseconds\n    this.maxResolution = { width: 3000, height: 3000 };\n    this.processingInterval = 1000; // Process every second\n  }\n\n  async startStream(streamId, onEmotionUpdate, config = {}) {\n    if (this.activeStreams.has(streamId)) {\n      logger.warn(`Stream ${streamId} is already active`);\n      return;\n    }\n\n    const streamContext = {\n      id: streamId,\n      buffer: [],\n      lastProcessed: Date.now(),\n      onUpdate: onEmotionUpdate,\n      processingInterval: null,\n      config: {\n        resetStream: config.resetStream || false,\n        models: {\n          face: config.faceConfig || {}\n        },\n        ...config\n      }\n    };\n\n    try {\n      // Start Hume stream\n      await humeService.startStream(streamId, streamContext.config);\n\n      // Start processing interval\n      streamContext.processingInterval = setInterval(\n        () => this.processStreamBuffer(streamContext),\n        this.processingInterval\n      );\n\n      this.activeStreams.set(streamId, streamContext);\n      logger.info(`Started video stream ${streamId}`);\n    } catch (error) {\n      logger.error(`Failed to start video stream ${streamId}:`, error);\n      throw error;\n    }\n  }\n\n  async stopStream(streamId) {\n    const streamContext = this.activeStreams.get(streamId);\n    if (!streamContext) {\n      logger.warn(`Stream ${streamId} not found`);\n      return;\n    }\n\n    // Clear processing interval\n    if (streamContext.processingInterval) {\n      clearInterval(streamContext.processingInterval);\n    }\n\n    // Process any remaining frames\n    await this.processStreamBuffer(streamContext);\n\n    // Stop Hume stream\n    await humeService.stopStream(streamId);\n\n    this.activeStreams.delete(streamId);\n    logger.info(`Stopped video stream ${streamId}`);\n  }\n\n  async addFrame(streamId, frameData, timestamp = Date.now()) {\n    const streamContext = this.activeStreams.get(streamId);\n    if (!streamContext) {\n      logger.warn(`Stream ${streamId} not found, frame discarded`);\n      return;\n    }\n\n    try {\n      // Validate frame dimensions\n      const dimensions = await this.getFrameDimensions(frameData);\n      if (!this.validateFrameDimensions(dimensions)) {\n        logger.warn(`Frame dimensions exceed maximum (${dimensions.width}x${dimensions.height})`);\n        return;\n      }\n\n      streamContext.buffer.push({\n        data: frameData,\n        timestamp\n      });\n\n      // Trim buffer if it gets too large\n      this.trimBuffer(streamContext);\n    } catch (error) {\n      logger.error(`Error adding frame to stream ${streamId}:`, error);\n    }\n  }\n\n  async processStreamBuffer(streamContext) {\n    if (streamContext.buffer.length === 0) return;\n\n    try {\n      const now = Date.now();\n      const chunkStartTime = now - this.chunkDuration;\n\n      // Get frames within the current chunk\n      const chunkFrames = streamContext.buffer.filter(\n        frame => frame.timestamp >= chunkStartTime\n      );\n\n      if (chunkFrames.length === 0) return;\n\n      // Select the best frame from the chunk\n      const selectedFrame = this.selectBestFrame(chunkFrames);\n\n      // Process the frame with Hume\n      const emotions = await humeService.processFacial(\n        selectedFrame.data,\n        streamContext.id\n      );\n\n      // Update emotional state\n      if (streamContext.onUpdate && emotions.length > 0) {\n        streamContext.onUpdate({\n          streamId: streamContext.id,\n          timestamp: now,\n          emotions,\n          frameCount: chunkFrames.length,\n          selectedFrameTime: selectedFrame.timestamp\n        });\n      }\n\n      // Remove processed frames\n      streamContext.buffer = streamContext.buffer.filter(\n        frame => frame.timestamp > chunkStartTime\n      );\n\n      streamContext.lastProcessed = now;\n    } catch (error) {\n      logger.error(`Error processing stream ${streamContext.id}:`, error);\n    }\n  }\n\n  selectBestFrame(frames) {\n    // For now, select the middle frame\n    // Could be enhanced with frame quality detection\n    return frames[Math.floor(frames.length / 2)];\n  }\n\n  trimBuffer(streamContext) {\n    const now = Date.now();\n    // Keep only frames from the last chunk duration\n    streamContext.buffer = streamContext.buffer.filter(\n      frame => now - frame.timestamp <= this.chunkDuration\n    );\n  }\n\n  async getFrameDimensions(frameData) {\n    // Implementation would depend on how frames are provided\n    // This is a placeholder that should be implemented based on\n    // the actual frame format (e.g., raw pixels, base64 image, etc.)\n    return {\n      width: 1280,  // placeholder\n      height: 720   // placeholder\n    };\n  }\n\n  validateFrameDimensions(dimensions) {\n    return dimensions.width <= this.maxResolution.width &&\n           dimensions.height <= this.maxResolution.height;\n  }\n\n  getStreamStatus(streamId) {\n    const streamContext = this.activeStreams.get(streamId);\n    if (!streamContext) return null;\n\n    return {\n      id: streamContext.id,\n      isActive: true,\n      bufferSize: streamContext.buffer.length,\n      lastProcessed: streamContext.lastProcessed,\n      timeSinceLastProcess: Date.now() - streamContext.lastProcessed,\n      config: streamContext.config\n    };\n  }\n\n  getAllStreams() {\n    return Array.from(this.activeStreams.keys()).map(id => this.getStreamStatus(id));\n  }\n\n  async cleanup() {\n    // Stop all active streams\n    for (const streamId of this.activeStreams.keys()) {\n      await this.stopStream(streamId);\n    }\n  }\n}\n\n// Create singleton instance\nconst videoStreamService = new VideoStreamService();\n\nexport default videoStreamService;","import { createClient } from 'redis';\nimport { logger } from '../utils/logger.js';\n\nlet client = null;\n\nexport async function setupRedis() {\n  try {\n    const url = process.env.REDIS_URL;\n    if (!url) {\n      throw new Error('REDIS_URL environment variable is not set');\n    }\n\n    client = createClient({\n      url,\n      socket: {\n        reconnectStrategy: (retries) => {\n          const maxRetryTime = 3000; // 3 seconds\n          const retryTime = Math.min(retries * 100, maxRetryTime);\n          logger.info(`Retrying Redis connection in ${retryTime}ms (attempt ${retries + 1})`);\n          return retryTime;\n        },\n        connectTimeout: 10000, // 10 seconds\n        keepAlive: 5000, // 5 seconds\n      },\n      database: 0,\n      commandsQueueMaxLength: 100000,\n      readonly: false,\n      legacyMode: false,\n      isolationPoolOptions: {\n        min: 5,\n        max: 20,\n        acquireTimeoutMillis: 5000,\n        createTimeoutMillis: 5000,\n        idleTimeoutMillis: 5000,\n        createRetryIntervalMillis: 200,\n      }\n    });\n\n    // Event handlers\n    client.on('connect', () => {\n      logger.info('Redis client connecting...');\n    });\n\n    client.on('ready', () => {\n      logger.info('Redis client connected and ready');\n    });\n\n    client.on('error', (err) => {\n      logger.error('Redis client error:', err);\n    });\n\n    client.on('reconnecting', () => {\n      logger.info('Redis client reconnecting...');\n    });\n\n    client.on('end', () => {\n      logger.info('Redis client connection closed');\n    });\n\n    // Connect to Redis\n    await client.connect();\n\n    // Test connection\n    await testConnection();\n\n    // Initialize data structures\n    await initializeDataStructures();\n\n    return client;\n  } catch (error) {\n    logger.error('Failed to setup Redis:', error);\n    throw error;\n  }\n}\n\nasync function testConnection() {\n  try {\n    await client.ping();\n    logger.info('Redis connection test successful');\n  } catch (error) {\n    logger.error('Redis connection test failed:', error);\n    throw error;\n  }\n}\n\nasync function initializeDataStructures() {\n  try {\n    // Initialize emotional state hash if it doesn't exist\n    const emotionalStateExists = await client.exists('emotional_state');\n    if (!emotionalStateExists) {\n      await client.hSet('emotional_state', {\n        emotion: 'neutral',\n        vector: '[]',\n        confidence: '1.0',\n        timestamp: Date.now().toString(),\n        type: 'system'\n      });\n    }\n\n    // Initialize memory windows\n    const windows = ['short_term', 'medium_term', 'long_term'];\n    for (const window of windows) {\n      const key = `memory:window:${window}`;\n      const exists = await client.exists(key);\n      if (!exists) {\n        await client.zAdd(key, {\n          score: Date.now(),\n          value: JSON.stringify({\n            type: 'system',\n            content: 'initialized',\n            timestamp: Date.now()\n          })\n        });\n      }\n    }\n\n    // Initialize attention system structures\n    const attentionExists = await client.exists('attention:4w');\n    if (!attentionExists) {\n      await client.hSet('attention:4w', {\n        who: '[]',\n        what: '[]',\n        when: '[]',\n        where: '[]'\n      });\n    }\n\n    logger.info('Redis data structures initialized');\n  } catch (error) {\n    logger.error('Failed to initialize Redis data structures:', error);\n    throw error;\n  }\n}\n\nexport function getRedisClient() {\n  if (!client) {\n    throw new Error('Redis client not initialized. Call setupRedis first.');\n  }\n  return client;\n}\n\nexport async function closeRedis() {\n  if (client) {\n    try {\n      await client.quit();\n      client = null;\n      logger.info('Redis connection closed gracefully');\n    } catch (error) {\n      logger.error('Error closing Redis connection:', error);\n      // Force close if graceful shutdown fails\n      client.disconnect();\n      client = null;\n    }\n  }\n}\n\n// Cleanup handler for graceful shutdown\nprocess.on('SIGTERM', async () => {\n  logger.info('SIGTERM received, closing Redis connection...');\n  await closeRedis();\n});\n\nprocess.on('SIGINT', async () => {\n  logger.info('SIGINT received, closing Redis connection...');\n  await closeRedis();\n});","import { logger } from '../utils/logger.js';\nimport humeService from './humeService.js';\nimport videoStreamService from './videoStreamService.js';\nimport { getRedisClient } from '../config/redis.js';\nimport { emotionToVector, vectorToEmotion } from '../constants/emotions.js';\n\nexport class EmotionalContextService {\n  constructor() {\n    this.redis = null;\n    this.activeContexts = new Map();\n    this.emotionalBuffer = new Map();\n    this.bufferTimeout = 5000; // 5 seconds\n    this.customModelEnabled = false;\n    this.weights = {\n      evi: 0.5,      // EVI's built-in emotional processing\n      video: 0.3,    // Video facial analysis\n      voice: 0.2     // Voice prosody analysis\n    };\n  }\n\n  async initialize() {\n    this.redis = getRedisClient();\n    await this.setupEmotionalBuffers();\n    await this.loadCustomModel();\n  }\n\n  async loadCustomModel() {\n    try {\n      const hasCustomModel = await this.redis.exists('custom_model_config');\n      if (hasCustomModel) {\n        const config = JSON.parse(await this.redis.get('custom_model_config'));\n        this.customModelEnabled = config.enabled;\n        if (config.weights) {\n          this.weights = config.weights;\n        }\n        logger.info('Custom model configuration loaded');\n      }\n    } catch (error) {\n      logger.error('Failed to load custom model:', error);\n    }\n  }\n\n  async setupEmotionalBuffers() {\n    try {\n      const bufferKey = 'emotional_context_buffers';\n      const exists = await this.redis.exists(bufferKey);\n      \n      if (!exists) {\n        await this.redis.hSet(bufferKey, {\n          active_sessions: '{}',\n          buffer_timeouts: '{}'\n        });\n      }\n    } catch (error) {\n      logger.error('Failed to setup emotional buffers:', error);\n      throw error;\n    }\n  }\n\n  async startContext(contextId, options = {}) {\n    const context = {\n      id: contextId,\n      startTime: Date.now(),\n      options: {\n        useVideo: options.useVideo ?? false,\n        useVoice: options.useVoice ?? true,\n        useEVI: options.useEVI ?? false,\n        customModel: options.customModel ?? this.customModelEnabled,\n        bufferSize: options.bufferSize ?? 5,\n        ...options\n      },\n      emotionalState: {\n        current: 'neutral',\n        confidence: 1.0,\n        vector: emotionToVector('neutral'),\n        history: [],\n        sources: {}\n      }\n    };\n\n    this.activeContexts.set(contextId, context);\n\n    // Start video stream if enabled\n    if (context.options.useVideo) {\n      await videoStreamService.startStream(contextId, (emotionData) => {\n        this.handleVideoEmotion(contextId, emotionData);\n      }, {\n        resetStream: true,\n        faceConfig: {\n          // Configure face detection settings\n          minConfidence: 0.7,\n          returnPoints: true\n        }\n      });\n    }\n\n    // Initialize Hume stream for voice if not using EVI\n    if (context.options.useVoice && !context.options.useEVI) {\n      await humeService.startStream(contextId, {\n        models: { prosody: {} },\n        resetStream: true\n      });\n    }\n\n    logger.info(`Started emotional context ${contextId} with options:`, context.options);\n    return context;\n  }\n\n  async handleEVIEmotion(contextId, eviEmotion) {\n    const context = this.activeContexts.get(contextId);\n    if (!context) {\n      logger.warn(`Context ${contextId} not found for EVI emotion`);\n      return;\n    }\n\n    // Store EVI emotion in context sources\n    context.emotionalState.sources.evi = {\n      emotion: eviEmotion.emotion,\n      confidence: eviEmotion.confidence,\n      vector: eviEmotion.vector,\n      timestamp: Date.now()\n    };\n\n    await this.updateEmotionalState(contextId, {\n      emotion: eviEmotion.emotion,\n      confidence: eviEmotion.confidence,\n      vector: eviEmotion.vector,\n      source: 'evi',\n      timestamp: Date.now()\n    });\n  }\n\n  async handleVideoEmotion(contextId, videoEmotion) {\n    const context = this.activeContexts.get(contextId);\n    if (!context) {\n      logger.warn(`Context ${contextId} not found for video emotion`);\n      return;\n    }\n\n    if (videoEmotion.emotions.length > 0) {\n      const dominant = videoEmotion.emotions[0];\n      \n      // Store video emotion in context sources\n      context.emotionalState.sources.video = {\n        emotion: dominant.name,\n        confidence: dominant.confidence,\n        vector: dominant.vector,\n        timestamp: videoEmotion.timestamp\n      };\n\n      await this.updateEmotionalState(contextId, {\n        emotion: dominant.name,\n        confidence: dominant.confidence,\n        vector: dominant.vector,\n        source: 'video',\n        timestamp: videoEmotion.timestamp\n      });\n    }\n  }\n\n  async handleVoiceEmotion(contextId, voiceData) {\n    const context = this.activeContexts.get(contextId);\n    if (!context) {\n      logger.warn(`Context ${contextId} not found for voice emotion`);\n      return;\n    }\n\n    try {\n      let emotions;\n      if (context.options.useEVI) {\n        // EVI handles voice emotion processing\n        return;\n      } else {\n        // Process voice through Hume\n        emotions = await humeService.processVoice(voiceData, contextId);\n      }\n\n      if (emotions.length > 0) {\n        const dominant = emotions[0];\n        \n        // Store voice emotion in context sources\n        context.emotionalState.sources.voice = {\n          emotion: dominant.name,\n          confidence: dominant.confidence,\n          vector: dominant.vector,\n          timestamp: Date.now()\n        };\n\n        await this.updateEmotionalState(contextId, {\n          emotion: dominant.name,\n          confidence: dominant.confidence,\n          vector: dominant.vector,\n          source: 'voice',\n          timestamp: Date.now()\n        });\n      }\n    } catch (error) {\n      logger.error(`Error processing voice emotion for context ${contextId}:`, error);\n    }\n  }\n\n  async updateEmotionalState(contextId, emotionData) {\n    const context = this.activeContexts.get(contextId);\n    if (!context) return;\n\n    // Add to emotional buffer\n    if (!this.emotionalBuffer.has(contextId)) {\n      this.emotionalBuffer.set(contextId, []);\n    }\n    this.emotionalBuffer.get(contextId).push(emotionData);\n\n    // Process buffer if it reaches the size limit or after timeout\n    if (this.emotionalBuffer.get(contextId).length >= context.options.bufferSize) {\n      await this.processEmotionalBuffer(contextId);\n    } else {\n      // Set timeout to process buffer\n      setTimeout(async () => {\n        await this.processEmotionalBuffer(contextId);\n      }, this.bufferTimeout);\n    }\n  }\n\n  async processEmotionalBuffer(contextId) {\n    const buffer = this.emotionalBuffer.get(contextId);\n    if (!buffer || buffer.length === 0) return;\n\n    const context = this.activeContexts.get(contextId);\n    if (!context) return;\n\n    const combinedVector = new Array(buffer[0].vector.length).fill(0);\n    let totalWeight = 0;\n\n    // Process each source type separately\n    const sourceGroups = this.groupBySource(buffer);\n    \n    for (const [source, emotions] of Object.entries(sourceGroups)) {\n      const weight = this.weights[source] * this.calculateSourceConfidence(emotions);\n      const sourceVector = this.combineSourceEmotions(emotions);\n      \n      sourceVector.forEach((v, i) => {\n        combinedVector[i] += v * weight;\n      });\n      totalWeight += weight;\n    }\n\n    // Normalize vector\n    if (totalWeight > 0) {\n      combinedVector.forEach((_, i) => {\n        combinedVector[i] /= totalWeight;\n      });\n    }\n\n    // Update context state\n    context.emotionalState = {\n      current: vectorToEmotion(combinedVector),\n      confidence: totalWeight,\n      vector: combinedVector,\n      sources: context.emotionalState.sources,\n      history: [\n        ...context.emotionalState.history,\n        {\n          timestamp: Date.now(),\n          emotions: buffer,\n          sources: { ...context.emotionalState.sources }\n        }\n      ].slice(-100) // Keep last 100 emotional states\n    };\n\n    // Store in Redis\n    await this.redis.hSet(`emotional_context:${contextId}`, {\n      state: JSON.stringify(context.emotionalState),\n      lastUpdate: Date.now().toString()\n    });\n\n    // Clear buffer\n    this.emotionalBuffer.set(contextId, []);\n  }\n\n  groupBySource(buffer) {\n    return buffer.reduce((groups, emotion) => {\n      const source = emotion.source;\n      if (!groups[source]) {\n        groups[source] = [];\n      }\n      groups[source].push(emotion);\n      return groups;\n    }, {});\n  }\n\n  calculateSourceConfidence(emotions) {\n    return emotions.reduce((sum, e) => sum + e.confidence, 0) / emotions.length;\n  }\n\n  combineSourceEmotions(emotions) {\n    const vector = new Array(emotions[0].vector.length).fill(0);\n    emotions.forEach(emotion => {\n      emotion.vector.forEach((v, i) => {\n        vector[i] += v * emotion.confidence;\n      });\n    });\n    return vector;\n  }\n\n  async getEmotionalContext(contextId) {\n    const context = this.activeContexts.get(contextId);\n    if (!context) {\n      const storedContext = await this.redis.hGetAll(`emotional_context:${contextId}`);\n      if (storedContext.state) {\n        return JSON.parse(storedContext.state);\n      }\n      return null;\n    }\n    return context.emotionalState;\n  }\n\n  async stopContext(contextId) {\n    const context = this.activeContexts.get(contextId);\n    if (!context) return;\n\n    // Process any remaining emotions in buffer\n    await this.processEmotionalBuffer(contextId);\n\n    // Stop video stream if active\n    if (context.options.useVideo) {\n      await videoStreamService.stopStream(contextId);\n    }\n\n    // Stop Hume stream if using it for voice\n    if (context.options.useVoice && !context.options.useEVI) {\n      await humeService.stopStream(contextId);\n    }\n\n    // Clean up\n    this.activeContexts.delete(contextId);\n    this.emotionalBuffer.delete(contextId);\n\n    logger.info(`Stopped emotional context ${contextId}`);\n  }\n\n  async cleanup() {\n    // Stop all active contexts\n    for (const contextId of this.activeContexts.keys()) {\n      await this.stopContext(contextId);\n    }\n  }\n}\n\n// Create singleton instance\nconst emotionalContextService = new EmotionalContextService();\n\nexport default emotionalContextService;","import { MongoClient } from 'mongodb';\nimport { logger } from '../utils/logger.js';\n\nlet client = null;\n\nexport async function setupDatabase() {\n  try {\n    const uri = process.env.MONGODB_URI;\n    if (!uri) {\n      throw new Error('MONGODB_URI environment variable is not set');\n    }\n\n    client = new MongoClient(uri, {\n      useNewUrlParser: true,\n      useUnifiedTopology: true,\n    });\n\n    await client.connect();\n    logger.info('Successfully connected to MongoDB');\n\n    // Create time series collections\n    const db = client.db('memorable');\n    \n    // Create collections with time series configuration\n    await createTimeSeriesCollection(db, 'emotions', 'timestamp');\n    await createTimeSeriesCollection(db, 'interactions', 'timestamp');\n    await createTimeSeriesCollection(db, 'contextual_data', 'timestamp');\n    \n    return client;\n  } catch (error) {\n    logger.error('MongoDB connection error:', error);\n    throw error;\n  }\n}\n\nasync function createTimeSeriesCollection(db, collectionName, timeField) {\n  try {\n    await db.createCollection(collectionName, {\n      timeseries: {\n        timeField,\n        metaField: \"metadata\",\n        granularity: \"seconds\"\n      }\n    });\n    logger.info(`Created time series collection: ${collectionName}`);\n  } catch (error) {\n    // Collection might already exist\n    if (error.code !== 48) { // 48 is the error code for \"collection already exists\"\n      throw error;\n    }\n  }\n}\n\nexport function getDatabase() {\n  if (!client) {\n    throw new Error('Database not initialized. Call setupDatabase first.');\n  }\n  return client.db('memorable');\n}\n\nexport async function closeDatabase() {\n  if (client) {\n    await client.close();\n    client = null;\n    logger.info('MongoDB connection closed');\n  }\n}","import { logger } from '../utils/logger.js';\nimport { getRedisClient } from '../config/redis.js';\nimport { getDatabase } from '../config/database.js';\n\nexport class CustomModelService {\n  constructor() {\n    this.redis = null;\n    this.db = null;\n    this.apiKey = process.env.HUME_API_KEY;\n    this.apiEndpoint = 'https://api.hume.ai/v0/custom/models';\n    this.activeModels = new Map();\n    this.trainingJobs = new Map();\n  }\n\n  async initialize() {\n    this.redis = getRedisClient();\n    this.db = getDatabase();\n    await this.loadActiveModels();\n  }\n\n  async loadActiveModels() {\n    try {\n      const models = await this.db.collection('custom_models')\n        .find({ status: 'active' })\n        .toArray();\n\n      models.forEach(model => {\n        this.activeModels.set(model.modelId, model);\n      });\n\n      logger.info(`Loaded ${models.length} active custom models`);\n    } catch (error) {\n      logger.error('Failed to load active models:', error);\n    }\n  }\n\n  async createTrainingJob(userId, config) {\n    try {\n      const jobId = `train_${Date.now()}_${userId}`;\n      const job = {\n        id: jobId,\n        userId,\n        status: 'preparing',\n        config: {\n          name: config.name,\n          description: config.description,\n          labelSet: config.labels || [],\n          dataConfig: {\n            includeExpressions: true,\n            includeLanguage: true,\n            includeProsody: true\n          },\n          ...config\n        },\n        created: new Date(),\n        updated: new Date()\n      };\n\n      await this.db.collection('training_jobs').insertOne(job);\n      this.trainingJobs.set(jobId, job);\n\n      // Start collecting training data\n      await this.collectTrainingData(jobId);\n\n      return jobId;\n    } catch (error) {\n      logger.error('Failed to create training job:', error);\n      throw error;\n    }\n  }\n\n  async collectTrainingData(jobId) {\n    const job = this.trainingJobs.get(jobId);\n    if (!job) throw new Error(`Training job ${jobId} not found`);\n\n    try {\n      // Update job status\n      job.status = 'collecting';\n      await this.updateJobStatus(job);\n\n      // Get user's emotional history\n      const emotionalHistory = await this.getEmotionalHistory(job.userId);\n\n      // Process and label the data\n      const labeledData = await this.processTrainingData(emotionalHistory, job.config);\n\n      // Store processed data\n      await this.storeTrainingData(jobId, labeledData);\n\n      // Start training if enough data\n      if (labeledData.length >= 100) {\n        await this.startModelTraining(jobId);\n      } else {\n        job.status = 'insufficient_data';\n        await this.updateJobStatus(job);\n      }\n    } catch (error) {\n      logger.error(`Failed to collect training data for job ${jobId}:`, error);\n      job.status = 'failed';\n      job.error = error.message;\n      await this.updateJobStatus(job);\n    }\n  }\n\n  async getEmotionalHistory(userId) {\n    const history = await this.db.collection('emotional_history')\n      .find({\n        userId,\n        timestamp: { $gte: new Date(Date.now() - 30 * 24 * 60 * 60 * 1000) } // Last 30 days\n      })\n      .sort({ timestamp: 1 })\n      .toArray();\n\n    return history;\n  }\n\n  async processTrainingData(history, config) {\n    const labeledData = [];\n\n    for (const entry of history) {\n      if (!entry.emotionalState || !entry.context) continue;\n\n      const processedEntry = {\n        timestamp: entry.timestamp,\n        labels: this.generateLabels(entry, config.labelSet),\n        data: {\n          expressions: entry.emotionalState.sources.video || null,\n          language: entry.context.text || null,\n          prosody: entry.emotionalState.sources.voice || null\n        }\n      };\n\n      if (this.validateTrainingEntry(processedEntry)) {\n        labeledData.push(processedEntry);\n      }\n    }\n\n    return labeledData;\n  }\n\n  generateLabels(entry, labelSet) {\n    const labels = new Set();\n\n    // Generate labels based on emotional state and context\n    labelSet.forEach(label => {\n      if (this.matchesLabelCriteria(entry, label)) {\n        labels.add(label);\n      }\n    });\n\n    return Array.from(labels);\n  }\n\n  matchesLabelCriteria(entry, label) {\n    // Implement label matching logic based on emotional state and context\n    // This would be customized based on the specific requirements\n    return false;\n  }\n\n  validateTrainingEntry(entry) {\n    return entry.labels.length > 0 &&\n           (entry.data.expressions || entry.data.language || entry.data.prosody);\n  }\n\n  async storeTrainingData(jobId, data) {\n    await this.db.collection('training_data').insertOne({\n      jobId,\n      data,\n      timestamp: new Date()\n    });\n  }\n\n  async startModelTraining(jobId) {\n    const job = this.trainingJobs.get(jobId);\n    if (!job) throw new Error(`Training job ${jobId} not found`);\n\n    try {\n      // Update job status\n      job.status = 'training';\n      await this.updateJobStatus(job);\n\n      // Get training data\n      const trainingData = await this.db.collection('training_data')\n        .findOne({ jobId });\n\n      // Submit training job to Hume API\n      const response = await fetch(this.apiEndpoint, {\n        method: 'POST',\n        headers: {\n          'Content-Type': 'application/json',\n          'X-Hume-Api-Key': this.apiKey\n        },\n        body: JSON.stringify({\n          name: job.config.name,\n          description: job.config.description,\n          data: trainingData.data\n        })\n      });\n\n      if (!response.ok) {\n        throw new Error(`Hume API error: ${response.statusText}`);\n      }\n\n      const result = await response.json();\n      job.modelId = result.model_id;\n      job.status = 'training';\n      await this.updateJobStatus(job);\n\n      // Start monitoring training progress\n      this.monitorTraining(jobId);\n    } catch (error) {\n      logger.error(`Failed to start training for job ${jobId}:`, error);\n      job.status = 'failed';\n      job.error = error.message;\n      await this.updateJobStatus(job);\n    }\n  }\n\n  async monitorTraining(jobId) {\n    const job = this.trainingJobs.get(jobId);\n    if (!job || !job.modelId) return;\n\n    try {\n      const response = await fetch(`${this.apiEndpoint}/${job.modelId}`, {\n        headers: { 'X-Hume-Api-Key': this.apiKey }\n      });\n\n      if (!response.ok) {\n        throw new Error(`Hume API error: ${response.statusText}`);\n      }\n\n      const status = await response.json();\n\n      if (status.status === 'completed') {\n        await this.handleTrainingComplete(job);\n      } else if (status.status === 'failed') {\n        await this.handleTrainingFailed(job, status.error);\n      } else {\n        // Check again in 5 minutes\n        setTimeout(() => this.monitorTraining(jobId), 5 * 60 * 1000);\n      }\n    } catch (error) {\n      logger.error(`Error monitoring training for job ${jobId}:`, error);\n    }\n  }\n\n  async handleTrainingComplete(job) {\n    try {\n      // Update job status\n      job.status = 'completed';\n      await this.updateJobStatus(job);\n\n      // Add model to active models\n      const model = {\n        modelId: job.modelId,\n        name: job.config.name,\n        description: job.config.description,\n        userId: job.userId,\n        status: 'active',\n        created: new Date(),\n        lastUsed: null\n      };\n\n      await this.db.collection('custom_models').insertOne(model);\n      this.activeModels.set(job.modelId, model);\n\n      logger.info(`Training completed for job ${job.id}`);\n    } catch (error) {\n      logger.error(`Error handling training completion for job ${job.id}:`, error);\n    }\n  }\n\n  async handleTrainingFailed(job, error) {\n    job.status = 'failed';\n    job.error = error;\n    await this.updateJobStatus(job);\n    logger.error(`Training failed for job ${job.id}:`, error);\n  }\n\n  async updateJobStatus(job) {\n    await this.db.collection('training_jobs').updateOne(\n      { id: job.id },\n      {\n        $set: {\n          status: job.status,\n          error: job.error,\n          updated: new Date(),\n          modelId: job.modelId\n        }\n      }\n    );\n  }\n\n  async getJobStatus(jobId) {\n    const job = this.trainingJobs.get(jobId);\n    if (!job) {\n      const stored = await this.db.collection('training_jobs')\n        .findOne({ id: jobId });\n      return stored;\n    }\n    return job;\n  }\n\n  async getActiveModels(userId) {\n    return Array.from(this.activeModels.values())\n      .filter(model => model.userId === userId);\n  }\n\n  async deleteModel(modelId) {\n    try {\n      // Delete from Hume API\n      await fetch(`${this.apiEndpoint}/${modelId}`, {\n        method: 'DELETE',\n        headers: { 'X-Hume-Api-Key': this.apiKey }\n      });\n\n      // Update local state\n      this.activeModels.delete(modelId);\n\n      // Update database\n      await this.db.collection('custom_models').updateOne(\n        { modelId },\n        { $set: { status: 'deleted' } }\n      );\n\n      logger.info(`Deleted custom model ${modelId}`);\n    } catch (error) {\n      logger.error(`Failed to delete model ${modelId}:`, error);\n      throw error;\n    }\n  }\n}\n\n// Create singleton instance\nconst customModelService = new CustomModelService();\n\nexport default customModelService;","import { logger } from '../utils/logger.js';\nimport os from 'os';\n\nexport class ModelSelectionService {\n  constructor() {\n    this.isServerEnvironment = process.env.NODE_ENV === 'production';\n    this.hasGPU = process.env.ENABLE_CUDA === '1';\n    this.modelConfigs = {\n      local: {\n        default: 'ollama/mistral:3.2-small',\n        management: 'ollama/mistral:3.2-small',\n        embedding: 'ollama/nomic-embed-text',\n        fallback: 'ollama/tinyllama'\n      },\n      server: {\n        default: 'ollama/mistral:7b-instruct',\n        management: 'ollama/mixtral:8x7b-instruct',\n        embedding: 'ollama/nomic-embed-text:latest',\n        fallback: 'ollama/mistral:3.2-small'\n      }\n    };\n\n    // Performance tracking\n    this.metrics = {\n      requestCount: 0,\n      totalLatency: 0,\n      errors: 0,\n      lastSwitchTime: Date.now(),\n      modelUsage: new Map()\n    };\n\n    // Memory thresholds (percentage)\n    this.memoryThresholds = {\n      warning: 80,\n      critical: 90\n    };\n\n    // Memoization caches\n    this.responseCache = new Map();\n    this.modelStateCache = new Map();\n    this.taskPatternCache = new Map();\n\n    // Cache configuration\n    this.cacheConfig = {\n      maxSize: 1000,\n      ttl: 3600000, // 1 hour\n      criticalityThreshold: 0.8\n    };\n\n    // Initialize performance monitoring\n    this.startPerformanceMonitoring();\n  }\n\n  getModelConfig(type = 'default') {\n    const environment = this.isServerEnvironment ? 'server' : 'local';\n    const config = this.modelConfigs[environment];\n\n    // If GPU is available locally, we can use larger models\n    if (!this.isServerEnvironment && this.hasGPU) {\n      return this.modelConfigs.server[type];\n    }\n\n    return config[type];\n  }\n\n  async validateModel(modelName) {\n    try {\n      // Check if model is available in Ollama\n      const response = await fetch('http://localhost:11434/api/tags');\n      const { models } = await response.json();\n      \n      return models.some(model => model.name === modelName);\n    } catch (error) {\n      logger.error('Error validating model:', error);\n      this.metrics.errors++;\n      return false;\n    }\n  }\n\n  async ensureModel(type = 'default') {\n    const modelName = this.getModelConfig(type);\n    const isAvailable = await this.validateModel(modelName);\n\n    if (!isAvailable) {\n      logger.info(`Model ${modelName} not found, falling back to smaller model`);\n      return this.modelConfigs[this.isServerEnvironment ? 'server' : 'local'].fallback;\n    }\n\n    return modelName;\n  }\n\n  getResourceLimits() {\n    if (this.isServerEnvironment) {\n      return {\n        maxMe