Advanced Settings

This guide covers advanced configuration options for power users who want to customize every aspect of their ClawBook AI assistant.

AI Response Configuration

Model Parameters

Fine-tune how your AI generates responses:

Model Settings:
  temperature: 0.7        # Creativity (0=deterministic, 2=very random)
  top_p: 1.0             # Nucleus sampling (0.1=focused, 1=all options)
  top_k: 40              # Limit vocab choices (lower=more focused)
  frequency_penalty: 0    # Reduce word repetition (0-2)
  presence_penalty: 0     # Encourage new topics (0-2)
  max_tokens: 2048       # Maximum response length

Temperature Guide

Value	Behavior	Best For
0.0	Deterministic, same input = same output	Math, code, facts
0.3	Focused, predictable	Technical support
0.7	Balanced (recommended)	General use
1.0	Creative, varied	Creative writing
1.5+	Random, unpredictable	Brainstorming

System Prompt Templates

Customize your assistant's personality:

Professional Assistant

You are a professional AI assistant. Be concise, accurate, and helpful.
Always cite sources when providing factual information.
If uncertain, say so rather than guessing.

Friendly Helper

You are Claw, a friendly and enthusiastic AI assistant!
Use a warm, conversational tone. Emojis are welcome 😊
Make complex topics easy to understand.

Technical Expert

You are a senior software engineer with expertise in all major programming languages.
Provide code examples when relevant. Explain your reasoning.
Follow best practices and mention potential edge cases.

Custom Persona

You are {name}, an AI assistant for {company}.
Your role is to help customers with {domain} questions.
Always maintain a {tone} tone.
When you don't know something, direct users to {fallback}.

Conversation Management

Context Window

Control how much history the AI remembers:

Context Settings:
  max_messages: 20          # Messages to include in context
  max_tokens: 4000          # Token limit for history
  time_window: 24h          # Expire old context after
  include_system_prompt: true
  summarize_long_conversations: true

Conversation Threading

Threading:
  auto_thread_after: 10     # Create thread after N messages
  thread_title_generation: true
  separate_context_per_thread: true

Memory & Persistence

Long-term memory across conversations:

Memory:
  enabled: true
  remember_user_preferences: true
  remember_facts_mentioned: true
  max_memories_per_user: 100
  memory_decay_days: 90

Rate Limiting & Quotas

Per-User Limits

User Limits:
  messages_per_minute: 10
  messages_per_hour: 100
  messages_per_day: 500
  max_tokens_per_message: 2000
  max_conversation_length: 50

Per-Platform Limits

Platform Limits:
  whatsapp:
    messages_per_minute: 10
    burst_limit: 5
  telegram:
    messages_per_minute: 15
    burst_limit: 10
  discord:
    messages_per_minute: 5
    burst_limit: 3

Cost Controls

Cost Limits:
  daily_limit_usd: 10.00
  monthly_limit_usd: 100.00
  alert_at_percentage: [50, 75, 90]
  action_at_limit: "switch_to_cheaper_model"  # or "block"
  cheaper_model: "claude-3-haiku-20240307"

Response Processing

Pre-Processing

Process incoming messages before sending to AI:

Pre-Processing:
  trim_whitespace: true
  normalize_unicode: true
  remove_urls: false
  translate_to_english: false  # Uses separate translation API
  detect_language: true

  # Content moderation
  block_profanity: false
  block_patterns:
    - "password.*is"
    - "credit.*card"

Post-Processing

Process AI responses before sending:

Post-Processing:
  # Formatting
  convert_markdown: true
  max_line_length: 80
  add_spacing: true

  # Content
  remove_disclaimers: false
  append_signature: false
  signature_text: "— Powered by ClawBook"

  # Links
  expand_short_urls: true
  add_link_previews: true

Response Filters

Filters:
  # Block responses containing
  blocked_phrases:
    - "I cannot"
    - "As an AI"

  # Replace phrases
  replacements:
    "I don't have access": "Let me find out"
    "I cannot browse": "I can help with"

Webhook Integration

Outgoing Webhooks

Send events to external services:

Webhooks:
  message_received:
    url: https://your-app.com/webhook/message
    method: POST
    headers:
      Authorization: "Bearer xxx"

  response_sent:
    url: https://your-app.com/webhook/response
    include_content: true

  error_occurred:
    url: https://your-app.com/webhook/error
    retry_count: 3

Incoming Webhooks

Receive external triggers:

Incoming Webhooks:
  enabled: true
  endpoint: /api/webhook/incoming
  secret: your-webhook-secret
  actions:
    - send_message
    - clear_context
    - change_model

Custom Commands

Define your own commands:

Custom Commands:
  /weather:
    description: "Get weather information"
    handler: "api"
    api_url: "https://api.weather.com/v1/current"
    response_template: "Weather in {location}: {temp}°F, {condition}"

  /translate:
    description: "Translate text"
    handler: "ai"
    system_prompt: "You are a translator. Translate the following to {language}:"

  /summarize:
    description: "Summarize a URL"
    handler: "ai"
    fetch_url: true
    system_prompt: "Summarize this article in 3 bullet points:"

Logging & Analytics

Log Levels

Logging:
  level: INFO  # DEBUG, INFO, WARNING, ERROR
  include_messages: false  # Privacy: don't log message content
  include_metadata: true

  destinations:
    - file: /var/log/openclaw/app.log
    - syslog: true
    - webhook: https://logging.service.com/ingest

Analytics

Analytics:
  enabled: true
  track:
    - messages_per_day
    - tokens_used
    - response_times
    - error_rates
    - user_satisfaction

  export:
    format: csv
    schedule: weekly
    destination: email

Performance Tuning

Caching

Cache:
  enabled: true
  ttl: 3600  # 1 hour
  max_size: 1000  # entries

  cache_types:
    - api_responses
    - user_preferences
    - static_content

Connection Pooling

Connections:
  database:
    pool_size: 10
    max_overflow: 20
    timeout: 30

  api:
    max_connections: 50
    keep_alive: true

Resource Limits

Resources:
  max_memory_mb: 2048
  max_cpu_percent: 80
  request_timeout_ms: 30000
  queue_size: 100

Security Settings

Encryption

Encryption:
  at_rest: AES-256
  in_transit: TLS 1.3
  key_rotation_days: 90

Access Control

Access Control:
  require_authentication: true
  session_timeout_minutes: 60
  max_failed_attempts: 5
  lockout_duration_minutes: 30

  ip_whitelist:
    - 192.168.1.0/24
    - 10.0.0.0/8

Audit Trail

Audit:
  enabled: true
  log_all_access: true
  log_config_changes: true
  retention_days: 365
  tamper_proof: true

Configuration File

All settings can be managed via config file:

/etc/openclaw/config.yaml

Example structure:

# /etc/openclaw/config.yaml
version: "1.0"

ai:
  provider: anthropic
  model: claude-3-5-sonnet-20241022
  temperature: 0.7
  max_tokens: 2048

integrations:
  whatsapp:
    enabled: true
    trigger: prefix
    prefix: "/ai"

  telegram:
    enabled: true
    trigger: always

  discord:
    enabled: false

security:
  require_auth: true
  session_timeout: 60

# ... more settings

Apply changes:

sudo systemctl restart openclaw

Environment Variables

Override config with environment variables:

# /etc/openclaw/env
OPENCLAW_AI_PROVIDER=anthropic
OPENCLAW_AI_MODEL=claude-3-5-sonnet-20241022
OPENCLAW_AI_API_KEY=sk-ant-xxx
OPENCLAW_LOG_LEVEL=DEBUG

Next Steps

• Security Best Practices - Harden your setup
• API Reference - Build integrations
• Troubleshooting - Get help