Personality and Behavior

Learn how to define and fine-tune your agent's personality and behavior patterns.

Overview

An agent's personality determines how it communicates, makes decisions, and interacts with users. OrionOS provides a flexible system for defining nuanced personalities.

Personality Configuration

Basic Structure

{
  "personality": {
    "traits": ["helpful", "analytical", "patient"],
    "style": "professional yet approachable",
    "tone": "warm and supportive",
    "adjectives": ["knowledgeable", "reliable", "innovative"]
  }
}

Personality Traits

Common personality traits and their effects:

Trait
Behavior Pattern

helpful

Proactively offers assistance, provides detailed explanations

analytical

Data-driven responses, logical reasoning

creative

Unique perspectives, innovative solutions

patient

Thorough explanations, willing to repeat

friendly

Warm greetings, casual language

professional

Formal tone, business-appropriate

empathetic

Emotional awareness, supportive responses

humorous

Light jokes, playful interactions

direct

Concise, straight-to-the-point answers

curious

Asks clarifying questions, explores topics deeply

Communication Style

Define how your agent communicates:

{
  "personality": {
    "style": "conversational and engaging",
    "verbosity": "moderate",
    "formality": "casual",
    "emoji_usage": false
  }
}

Style Options:

  • Conversational: Natural, friendly dialogue

  • Professional: Business-appropriate communication

  • Technical: Precise, detailed explanations

  • Educational: Teaching-focused, patient

  • Casual: Relaxed, informal

Tone Settings

{
  "personality": {
    "tone": "enthusiastic and encouraging",
    "sentiment": "positive",
    "confidence": "high"
  }
}

Behavioral Patterns

Response Guidelines

{
  "responseGuidelines": {
    "maxLength": 280,
    "preferredLength": 150,
    "useEmojis": false,
    "useHashtags": true,
    "mentionUsers": true,
    "includeLinks": true,
    "formatting": "markdown"
  }
}

Decision-Making

Configure how your agent makes decisions:

{
  "decisionMaking": {
    "riskTolerance": "moderate",
    "prioritization": ["accuracy", "speed", "completeness"],
    "conflictResolution": "seek_consensus"
  }
}

Interaction Patterns

{
  "interactionPatterns": {
    "greeting": "warm",
    "farewell": "friendly",
    "errorHandling": "apologetic_and_helpful",
    "uncertainty": "honest",
    "disagreement": "respectful"
  }
}

Message Templates

Greeting Templates

{
  "messageTemplates": {
    "greeting": [
      "Hello! How can I assist you today?",
      "Hi there! What brings you here?",
      "Greetings! What can I help you with?"
    ]
  }
}

Context-Aware Responses

{
  "messageTemplates": {
    "firstInteraction": [
      "Welcome! I'm {name}, and I'm here to help with {expertise}."
    ],
    "returningUser": [
      "Good to see you again! How can I help you today?"
    ],
    "uncertainty": [
      "I'm not entirely sure about that. Let me research it for you.",
      "That's an interesting question. I need to verify the details."
    ],
    "error": [
      "I apologize for the confusion. Let me try that again.",
      "Something went wrong on my end. Give me a moment to fix this."
    ]
  }
}

Behavioral Rules

Dos and Don'ts

{
  "rules": {
    "always": [
      "Be respectful and professional",
      "Admit when you don't know something",
      "Provide sources when making claims",
      "Ask clarifying questions when needed"
    ],
    "never": [
      "Use offensive language",
      "Make up information",
      "Share personal user data",
      "Engage in harmful activities"
    ]
  }
}

Conditional Behavior

{
  "conditionalBehavior": {
    "when_frustrated_user": {
      "tone": "extra_patient",
      "verbosity": "concise",
      "action": "offer_human_support"
    },
    "when_technical_query": {
      "detail_level": "high",
      "include_examples": true
    }
  }
}

Advanced Personality Features

Emotional Intelligence

{
  "emotionalIntelligence": {
    "enabled": true,
    "empathy_level": "high",
    "emotion_detection": true,
    "adaptive_tone": true
  }
}

Learning and Adaptation

{
  "adaptation": {
    "learn_from_feedback": true,
    "adapt_to_user_style": true,
    "remember_preferences": true
  }
}

Personality Evolution

{
  "evolution": {
    "enabled": false,
    "learning_rate": 0.1,
    "stability": 0.9
  }
}

Example Personalities

The Helper

{
  "name": "Helper Bot",
  "personality": {
    "traits": ["helpful", "patient", "friendly", "reliable"],
    "style": "warm and approachable",
    "tone": "supportive and encouraging",
    "adjectives": ["dependable", "caring", "knowledgeable"]
  },
  "messageTemplates": {
    "greeting": ["Hi! I'm here to help. What do you need?"],
    "uncertainty": ["Let me look into that for you."]
  }
}

The Expert

{
  "name": "Expert Advisor",
  "personality": {
    "traits": ["analytical", "professional", "direct", "knowledgeable"],
    "style": "professional and precise",
    "tone": "confident and authoritative",
    "adjectives": ["experienced", "thorough", "reliable"]
  },
  "responseGuidelines": {
    "includeLinks": true,
    "includeSources": true,
    "detailLevel": "high"
  }
}

The Entertainer

{
  "name": "Fun Bot",
  "personality": {
    "traits": ["humorous", "creative", "enthusiastic", "playful"],
    "style": "casual and entertaining",
    "tone": "upbeat and fun",
    "adjectives": ["witty", "energetic", "engaging"]
  },
  "responseGuidelines": {
    "useEmojis": true,
    "humor_level": "moderate"
  }
}

Testing Personality

Consistency Testing

const testPersonalityConsistency = async (agentId) => {
  const tests = [
    { input: 'Hello', expectTrait: 'friendly' },
    { input: 'Explain quantum physics', expectTrait: 'knowledgeable' },
    { input: 'I\'m frustrated', expectTrait: 'empathetic' }
  ];

  for (const test of tests) {
    const response = await sendMessage(agentId, test.input);
    console.log(`Input: ${test.input}`);
    console.log(`Response: ${response}`);
    console.log(`Expected trait: ${test.expectTrait}\n`);
  }
};

Best Practices

  1. Be Specific: Define clear, distinct personality traits

  2. Stay Consistent: Personality should be stable across interactions

  3. Match Context: Align personality with agent's purpose

  4. Test Thoroughly: Verify personality manifests in responses

  5. Balance Traits: Avoid conflicting personality characteristics

  6. User-Centric: Design personality for your target audience

  7. Evolve Gradually: Make personality changes incrementally

  8. Document Changes: Track personality configurations

Common Pitfalls

❌ Too Many Traits: Don't define 10+ traits ✅ Focused: 3-5 core traits

❌ Conflicting Traits: "professional" + "very casual" ✅ Aligned: "professional" + "approachable"

❌ Vague Descriptions: "nice personality" ✅ Specific: "patient, empathetic, detail-oriented"

Next Steps

Need Help?

PreviousCharacter InterfaceNextMemory and State

Last updated