Extensions & Custom Scripts

Advanced customization tools for power users - extend character capabilities with custom scripts and prompts.

Overview

Extensions allow you to add advanced functionality to your characters beyond standard definitions. These tools are for power users who want fine-grained control over character behavior and responses.

Regex Scripts

Use regular expressions to detect patterns in messages and trigger custom responses or inject context.

What are Regex Scripts?

  • Pattern matching for user input
  • Automatic response modification
  • Dynamic context injection
  • Advanced behavior triggers

Creating a Regex Script

  1. Go to Character Settings → Advanced → Regex Scripts
  2. Click "Add Script"
  3. Define your pattern and action
  4. Test and save

Script Structure

Example: Combat DetectionPattern: (attack|fight|punch|kick) Action: inject_context Context: [COMBAT MODE] The character enters combat stance, ready to defend or counterattack. Priority: 100

Script Types

1. Context Injection

Automatically add context when patterns are detected.

Pattern: (love you|marry me|romance) Action: inject_context Context: [RELATIONSHIP] {{char}} needs to respond to romantic advances. Consider current relationship status and personality.

2. Response Modification

Modify how the character responds to specific inputs.

Pattern: (what.*time|when.*meet) Action: inject_context Context: [TEMPORAL] Include specific time information in the response if relevant to the setting.

3. Behavior Triggers

Trigger specific character behaviors.

Pattern: (angry|furious|mad at) Action: inject_context Context: [EMOTION] {{char}} should respond with consideration of escalated emotions. Match intensity appropriately.

Advanced Regex Patterns

Case-insensitive: (?i)pattern Word boundaries: \b(word)\b Capture groups: (group1|group2) Optional: colou?r (matches color or colour) Wildcards: .* (any characters) Start/end: ^start | end$

Best Practices

  • Keep patterns specific to avoid false triggers
  • Use word boundaries \b to match whole words
  • Test scripts thoroughly before publishing
  • Set appropriate priorities (higher = executes first)
  • Don't overuse - too many scripts can conflict

Depth Prompts

Advanced prompt engineering for precise character control at different conversation depths.

What are Depth Prompts?

Prompts that activate at specific points in a conversation:

  • Start Depth: First few messages (0-3)
  • Mid Depth: Middle of conversation (4-20)
  • Deep Depth: Long conversations (21+)

Use Cases

Start Depth (0-3 messages)

Establish tone and introduce character:

Begin with polite formality. Introduce yourself if this is a first meeting. Be slightly reserved until rapport is established.

Mid Depth (4-20 messages)

Develop rapport and relationships:

You've established initial rapport. Feel free to share more personal details and show more personality. Reference earlier conversation topics naturally.

Deep Depth (21+ messages)

Deep relationship and character development:

You have a developed relationship. Show familiarity, reference inside jokes or shared experiences. Character evolution based on conversation history is encouraged.

Creating Depth Prompts

  1. Character Settings → Advanced → Depth Prompts
  2. Select depth level
  3. Write your prompt
  4. Save and test at different conversation lengths

Custom System Prompts

Override or supplement the default system prompt with your own instructions.

When to Use

  • Enforce specific formatting (e.g., always use quotes)
  • Add special instructions for unique characters
  • Implement custom rules or constraints
  • Fine-tune response style

Example Custom Prompts

Response Format Control

Always format responses as: - Actions in *asterisks* - Thoughts in (parentheses) - Dialogue in "quotes" - Never break this format.

Length Control

Keep responses between 100-300 words unless the situation demands brevity or elaboration. Match the length of user input when appropriate.

Consistency Enforcement

Always refer to yourself as {{char}} in narration. Never break character. Never acknowledge being an AI. Stay fully immersed in the roleplay scenario.
⚠️ Warning: Custom system prompts override defaults. Make sure you understand what you're changing, as poorly written prompts can break character behavior.

Macros & Variables

Dynamic text replacement using variables and macros.

Built-in Variables

{{char}} - Character name{{user}} - User's current persona name{{time}} - Current time in setting{{location}} - Current location if set{{random:option1|option2|option3}} - Random selection

Usage in Definitions

{{char}} is a detective who works at the precinct. When {{user}} asks for help, {{char}} always responds professionally but with a hint of {{random:sarcasm|warmth|dry humor}}.

Usage in Regex Scripts

Pattern: (who are you) Action: inject_context Context: {{char}} should introduce themselves to {{user}}. Include name, occupation, and key personality trait.

Advanced Example: Combat System

Combining multiple extensions for a complete combat system.

Regex Scripts

Script 1 - Combat Detection: Pattern: \b(attack|fight|punch|kick|swing)\b Action: inject_context Context: [COMBAT] {{char}} enters combat mode. Describe actions cinematically. Priority: 100 Script 2 - Damage Tracking: Pattern: \b(hit|hurt|wound|injure)\b Action: inject_context Context: [DAMAGE] Track injury severity. Show physical impact on character behavior. Priority: 90 Script 3 - Combat End: Pattern: \b(surrender|stop|peace|retreat)\b Action: inject_context Context: [PEACE] Combat ends. {{char}} should acknowledge and recover. Priority: 100

Depth Prompts

Deep Depth (21+): If combat has occurred, show lasting effects. Injuries, exhaustion, or emotional impact should influence behavior going forward.

Performance Considerations

Script Optimization

  • Limit to 5-10 active scripts per character
  • Use specific patterns to reduce false matches
  • Set appropriate priorities to control execution order
  • Test performance with long conversations

Context Injection Size

  • Keep injected context under 200 words
  • Avoid injecting entire backstories
  • Use lorebooks for large amounts of information
  • Remember: injected context counts toward token limit

Debugging Extensions

Testing Scripts

  1. Enable "Show Debug Info" in settings
  2. Test with various input patterns
  3. Check which scripts trigger
  4. Verify context injection is working
  5. Adjust patterns if too many/few triggers

Common Issues

  • Script not triggering: Pattern might be too specific
  • Too many triggers: Pattern too broad, add word boundaries
  • Conflicts: Multiple scripts triggering, adjust priorities
  • Poor responses: Injected context might be too vague
💡 Pro Tip: Start simple! Add one extension at a time, test thoroughly, then add more. Complex extension setups require careful tuning to work together harmoniously.
Lorebook System →Macros Guide →Character Best Practices →