Advanced Configuration Advanced

This document covers advanced configuration options and techniques for users who need deep customization.

💡 Reading Tips

Read Basic Config and Channels before this document.

Configuration Hierarchy

ChatAI Plugin uses a multi-layer configuration system:

Configs are merged by priority; later layers override earlier ones.

Environment Variables

Configuration overrides via environment variables are supported:

# API Keys
export CHATAI_OPENAI_KEY=sk-xxx
export CHATAI_CLAUDE_KEY=sk-ant-xxx

# Proxy Config
export CHATAI_PROXY=http://127.0.0.1:7890
export HTTPS_PROXY=http://127.0.0.1:7890

# Debug Mode
export CHATAI_DEBUG=true
export CHATAI_LOG_LEVEL=debug

Environment Variable Mapping

Environment Variable	Config Path	Description
CHATAI_OPENAI_KEY	channels[0].key	OpenAI API Key
CHATAI_PROXY	proxy.url	Proxy address
CHATAI_DEBUG	debug	Debug mode
CHATAI_PORT	server.port	Web server port

Multi-Environment Config

config.dev.yaml Development

debug: true
logLevel: debug

server:
  port: 3001

channels:
  - name: dev-channel
    baseUrl: http://localhost:8080/v1
    key: test-key

config.prod.yaml Production

debug: false
logLevel: info

server:
  port: 3000
  cors:
    enabled: true
    origins:
      - https://your-domain.com

rateLimit:
  enabled: true
  maxRequests: 60

📦 Loading a specific environment config

# Specify via environment variable
export CHATAI_ENV=prod

Advanced Channel Config

⚙️ Advanced Configuration

The following config suits users with multiple API channels, enabling high availability and load balancing.

Load Balancing

channels:
  - name: openai-1
    priority: 1        # Priority; smaller number means higher priority
    weight: 3          # Weight, used for weighted round-robin
    maxConcurrent: 10  # Max concurrency

  - name: openai-2
    priority: 1
    weight: 2
    maxConcurrent: 10

  - name: backup
    priority: 2        # Backup channel
    weight: 1

loadBalance:
  strategy: weighted   # round-robin | weighted | priority | random
  healthCheck:
    enabled: true
    interval: 60       # Health check interval (seconds)
    timeout: 5         # Timeout (seconds)

Failover

channels:
  - name: primary
    priority: 1
    failover:
      enabled: true
      maxRetries: 3
      retryDelay: 1000     # Retry delay (milliseconds)
      fallbackChannel: backup

  - name: backup
    priority: 2

Request Limits

channels:
  - name: limited-channel
    rateLimit:
      requestsPerMinute: 60
      tokensPerMinute: 100000
      requestsPerDay: 1000

    quotas:
      daily: 100000      # Daily token quota
      monthly: 3000000   # Monthly token quota
      alertThreshold: 0.8  # Alert at 80%

Model Aliases

Configure model aliases to simplify switching:

models:
  aliases:
    default: gpt-4o
    fast: gpt-4o-mini
    smart: claude-3-5-sonnet-20241022
    cheap: deepseek-chat

  # Per-purpose mapping
  usage:
    chat: default
    memory: fast
    summary: fast
    embedding: text-embedding-3-small

Usage:

#switch model smart

Advanced Trigger Config

Regex Triggers

triggers:
  regex:
    enabled: true
    patterns:
      - pattern: "^(help me|excuse me|can you)"
        flags: "i"
        priority: 1
      - pattern: "(AI|artificial intelligence|robot)"
        flags: "gi"
        priority: 2

Contextual Triggers

triggers:
  contextual:
    enabled: true
    # Keep responding if replied within the last N messages
    replyWindow: 5
    # Timeout (seconds)
    timeout: 300

Scheduled Triggers

triggers:
  schedule:
    enabled: true
    rules:
      - cron: "0 9 * * *"    # 9:00 every day
        action: morning_greeting
        groups: ["123456"]
      - cron: "0 22 * * *"   # 22:00 every day
        action: summary_push

Content Filtering

Input Filter

filter:
  input:
    enabled: true
    # Sensitive word filter
    keywords:
      - sensitive word 1
      - sensitive word 2
    # Regex filter
    patterns:
      - "\\d{11}"           # Phone numbers
      - "\\d{18}"           # ID card numbers
    # Replacement rules
    replacements:
      "profanity": "[blocked]"

Output Filter

filter:
  output:
    enabled: true
    # Remove specific content
    remove:
      - " thinking"
      - " response"
    # Length limit
    maxLength: 2000
    truncateMessage: "...(content too long, truncated)"

Cache Config

cache:
  # Response cache
  response:
    enabled: true
    ttl: 3600           # Cache time (seconds)
    maxSize: 1000       # Max cached entries

  # Model list cache
  models:
    ttl: 86400          # 24 hours

  # Tool result cache
  tools:
    enabled: true
    ttl: 300
    # Cacheable tools
    cacheable:
      - get_current_time
      - get_weather

Logging Config

logging:
  level: info           # debug | info | warn | error

  # File logging
  file:
    enabled: true
    path: ./logs
    maxSize: 10M
    maxFiles: 7

  # Console logging
  console:
    enabled: true
    colorize: true

  # Request logging
  request:
    enabled: true
    includeBody: false   # Disable in production

  # Sensitive data redaction
  redact:
    - key
    - password
    - token

Performance Tuning

Concurrency Control

performance:
  # Global concurrency limit
  maxConcurrentRequests: 50

  # Queue config
  queue:
    enabled: true
    maxSize: 100
    timeout: 30000

  # Connection pool
  connectionPool:
    maxConnections: 20
    keepAlive: true
    timeout: 30000

Memory Optimization

performance:
  memory:
    # Conversation history limit
    maxHistoryLength: 50
    # Tool log retention
    toolLogRetention: 7  # days
    # Periodic GC
    gcInterval: 3600

Security Config

CORS

server:
  cors:
    enabled: true
    origins:
      - https://your-domain.com
      - https://admin.your-domain.com
    methods:
      - GET
      - POST
      - PUT
      - DELETE
    credentials: true

Authentication Hardening

auth:
  # JWT config
  jwt:
    secret: your-secret-key
    expiresIn: 7d

  # Login limits
  login:
    maxAttempts: 5
    lockoutDuration: 300  # seconds

  # IP whitelist
  ipWhitelist:
    enabled: false
    ips:
      - 127.0.0.1
      - 192.168.1.0/24

Encrypted Storage

security:
  encryption:
    enabled: true
    algorithm: aes-256-gcm
    # Key read from an environment variable
    keyEnv: CHATAI_ENCRYPTION_KEY

Plugin System

Custom Middleware

middleware:
  custom:
    - path: ./plugins/my-middleware.js
      enabled: true
      config:
        option1: value1

// plugins/my-middleware.js
export default {
  name: 'my-middleware',

  // Pre-request hook
  async onRequest(ctx, next) {
    console.log('Request:', ctx.path)
    await next()
  },

  // Post-response hook
  async onResponse(ctx, response) {
    return response
  }
}

Event Hooks

hooks:
  onStart:
    - ./plugins/on-start.js
  onMessage:
    - ./plugins/on-message.js
  onToolCall:
    - ./plugins/on-tool-call.js

Config Validation

Validate config at startup:

validation:
  enabled: true
  strict: false      # Strict mode: error on unknown fields

  # Required config checks
  required:
    - channels
    - server.port

Config Import & Export

Export Config

#ai导出配置

Import Config

#ai导入配置 [config file path]

Config Sync

sync:
  enabled: false
  provider: git      # git | s3 | webdav
  remote: https://github.com/your/config.git
  branch: main
  interval: 3600     # Sync interval (seconds)

Next Steps

• Basic Config - Basic configuration options
• Channels - Detailed channel configuration
• Triggers - Trigger method configuration