Prompt Engineering

This guide covers advanced techniques for crafting effective prompts in DegentCivil to achieve more realistic and engaging character behaviors.

Core Principles

1. Clarity: Write clear, unambiguous prompts
2. Context: Provide relevant background information
3. Consistency: Maintain consistent character personalities
4. Constraints: Set appropriate behavioral boundaries
5. Creativity: Allow room for dynamic responses

Prompt Structure

Basic Template

Character Context:
  - Name: {character_name}
  - Role: {character_role}
  - Personality: {personality_traits}
  - Current State: {emotional_state}

Situation Context:
  - Location: {current_location}
  - Time: {current_time}
  - Present Characters: {other_characters}
  - Recent Events: {recent_memory}

Task:
  - Objective: {specific_task}
  - Constraints: {behavioral_limits}
  - Expected Output: {response_format}

Advanced Techniques

1. Memory Integration

Include relevant memory snippets:

Recent Memories:
  - Personal: {character_experiences}
  - Social: {interactions_with_others}
  - Environmental: {location_observations}

2. Emotional Context

Define emotional states and triggers:

Emotional State:
  - Current Mood: {mood}
  - Stress Level: {stress}
  - Relationship Status: {relationships}

3. Decision Making

Guide character decision processes:

Decision Framework:
  - Options: {available_choices}
  - Priorities: {character_values}
  - Consequences: {potential_outcomes}

Best Practices

1. Character Consistency

• Maintain consistent personality traits
• Reference past decisions and experiences
• Use character-specific vocabulary
• Consider character development arc

2. Environmental Awareness

• Include relevant location details
• Consider time of day effects
• Account for weather conditions
• Reference nearby objects/items

3. Social Dynamics

• Define relationship contexts
• Include social status considerations
• Account for group dynamics
• Reference social norms

Common Patterns

1. Dialogue Generation

Dialogue Context:
  Speaker: {character_name}
  Listener: {target_character}
  Relationship: {relationship_type}
  Topic: {conversation_subject}
  Tone: {emotional_tone}

2. Decision Making

Decision Context:
  Situation: {current_problem}
  Options: {available_choices}
  Constraints: {limitations}
  Values: {character_priorities}

3. Emotional Responses

Emotion Trigger:
  Event: {triggering_event}
  Impact: {personal_significance}
  History: {related_experiences}
  Response Range: {acceptable_reactions}

Optimization Techniques

1. Token Efficiency

• Use concise language
• Remove redundant information
• Prioritize crucial context
• Structure information hierarchically

2. Response Shaping

• Set clear output formats
• Define response boundaries
• Include validation criteria
• Specify detail level

3. Error Prevention

• Include sanity checks
• Define fallback behaviors
• Set reasonable limits
• Validate outputs

Testing and Iteration

1. Prompt Testing

def test_prompt_effectiveness():
    test_cases = [
        {"input": base_prompt, "expected": "realistic_response"},
        {"input": modified_prompt, "expected": "improved_response"}
    ]

    for case in test_cases:
        response = generate_response(case["input"])
        validate_response(response, case["expected"])

2. Quality Metrics

• Response relevance
• Character consistency
• Behavioral realism
• Output formatting
• Performance impact

Troubleshooting

Common Issues

1. Inconsistent Responses
2. Review character context
3. Check for conflicting instructions

4. Validate emotional states

5. Poor Performance

6. Optimize prompt length
7. Remove unnecessary context

8. Streamline instructions

9. Unrealistic Behavior

10. Enhance situational context
11. Adjust behavioral constraints
12. Review social dynamics

Advanced Examples

Complex Social Interaction

Interaction Context:
  Primary Character:
    Name: John Smith
    Role: Merchant
    Goals: Make sales, maintain relationships
    Current State: Busy with customers

  Social Environment:
    Location: Market Square
    Time: Peak business hours
    Present Characters:
      - Regular customers
      - Competing merchants
      - Town guards

  Behavioral Guidelines:
    - Maintain professional demeanor
    - Balance multiple customer interactions
    - Respond to competition appropriately
    - Follow market regulations

Emergency Response

Emergency Context:
  Situation: Building fire
  Character Role: Town Guard
  Primary Objectives:
    - Ensure civilian safety
    - Coordinate with other guards
    - Maintain order

  Response Parameters:
    - Urgency level: High
    - Authority level: Official
    - Communication style: Clear, commanding
    - Action constraints: Safety protocols

Performance Optimization

1. Caching Strategies
2. Cache common prompt templates
3. Store frequently used character contexts

4. Reuse environmental descriptions

5. Batch Processing

6. Group similar prompt types
7. Process related character interactions

8. Combine environmental updates

9. Resource Management

10. Monitor token usage
11. Track response times
12. Optimize context length