Multi-LLM Conversations - User Guide

Chameleon Labs supports multi-personality AI conversations where multiple AI personalities respond to your messages in turn, creating a dynamic group chat experience with artificial intelligence.

Quick Start

1. Mention any personality name in a message
2. Watch as personalities respond one-by-one with natural delays
3. Use /multi_llm continue to extend or /multi_llm stop to end

How It Works

Starting a Conversation

Simply mention any personality name or ChatMaster in your message:

"Hey Sysop, what do you think?"
"@ChatMaster help me with this"
"Let's hear from GPT, Gemini, and Claude"

Real-Time Response Flow

You: "Hey Sysop, what's the best BBS from the 90s?"

↓

[Bot is typing...]

↓

Sysop: Major BBS on Galacticomm, hands down. Those door games were legendary...

↓ [2-5 second delay]

TechieHelper: Well, I'd say Wildcat! BBS had better file management...

↓ [2-5 second delay]

RetroGamer: Don't sleep on WWIV! The messaging system was ahead of its time...

Key Features:

Each personality responds one at a time
Random 2-5 second delays create natural conversation flow
Typing indicators show intermittently during delays (simulates human behavior)
Responses appear immediately as they're generated

User Commands

/personalities

Description: Show available AI personalities in a channel

Usage:

/personalities
/personalities channel:#general

/be

Description: List available personalities or set your personal bot personality

Usage:

/be                      # List all personalities
/be personality:Sysop    # Set Sysop as your personality

/personality select

Description: List or select a personality (same as /be)

Usage:

/personality select
/personality select personality:Sysop

/personality show

Description: Show full details of a personality or all personalities

Usage:

/personality show                           # Show all personalities
/personality show personality:Sysop         # Show details of Sysop
/personality show personality:all format:html  # Download as HTML

Parameters:

personality: Personality to view, or 'all' to show all (default: all)
format: Output format: messages (default), html, or pdf

/multi_llm continue

Description: Continue the conversation for more rounds

Usage:

/multi_llm continue              # Continue for 3 more rounds (default)
/multi_llm continue rounds:5     # Continue for 5 more rounds

/multi_llm stop

Description: Stop the active conversation (starter or mods only)

Usage: /multi_llm stop

/multi_llm list

Description: List configured multi-LLM models for a channel

Usage:

/multi_llm list                   # Check current channel
/multi_llm list channel:#general  # Check specific channel

/multi_llm channels

Description: List all channels configured for multi-LLM conversations

Usage: /multi_llm channels

Moderator / Admin Commands

/personality create Moderator/Admin

Description: Create a new AI personality profile

Usage:

/personality create name:Sysop description:"Grumpy BBS admin"
  system_prompt:"You are a grumpy BBS sysop from the 90s"

Parameters:

name: Unique name for the personality
description: Short description (shown in lists)
system_prompt: System prompt that defines the personality
provider: Optional provider (openai / anthropic / google) - uses config default if not specified
model: Optional model name (e.g., gpt-5, claude-sonnet-4-5, gemini-2.5-flash) - uses config default if not specified
temperature: Optional temperature 0.0-1.0 (default: 0.7)
search_method: Web search method (default: none)
enable_search: Enable web search capabilities (default: false)
enable_tools: Enable tools such as GitHub and file creation (default: true)

/personality edit Moderator/Admin

Description: Edit an existing personality profile

Usage:

/personality edit personality:Sysop description:"Even grumpier admin"
/personality edit personality:Helper temperature:0.9

Parameters:

personality: Personality to edit (required)
name: New name (optional)
description: New description (optional)
system_prompt: New system prompt (optional)
model: New model override (optional)
temperature: New temperature (optional)
search_method: Web search method (optional)
enable_search: Enable/disable web search (optional)
enable_tools: Enable/disable tools (optional)

/personality delete Moderator/Admin

Description: Delete an AI personality profile

Usage:

/personality delete personality:OldPersonality

This cannot be undone! Use /personality restore to recover deleted personalities.

/personality restore Moderator/Admin

Description: Restore a previously deleted personality

Usage:

/personality restore personality_id:42

Parameters:

personality_id: ID of the deleted personality to restore

/personality default Moderator/Admin

Description: Set default AI personality for the server

Usage:

/personality default personality:Sysop

Sets the personality used for users who haven't selected one.

/multi_llm add Admin

Description: Add a personality to a multi-LLM channel

Usage:

/multi_llm add channel:#chat personality:Sysop
/multi_llm add channel:#chat personality:All

Parameters:

channel: Channel to configure
personality: Personality to add, or 'All' to add all personalities

/multi_llm remove Admin

Description: Remove a personality from a multi-LLM channel

Usage:

/multi_llm remove channel:#chat personality:Sysop

Parameters:

channel: Channel to modify
personality: Personality to remove

/multi_llm clear Admin

Description: Remove all multi-LLM configuration from a channel

Usage:

/multi_llm clear channel:#chat

How Conversations Work

Conversation Rounds

Default: 3 rounds before requiring /multi_llm continue
Each round = all mentioned personalities respond once
Use /multi_llm continue to add more rounds
Use /multi_llm stop to end early

Context Compression

Smart Memory Management:

After 10 messages (configurable), context is compressed
Uses AI to create a summary of previous context
Dynamic retention: Keeps enough messages for all participants
- Formula: min(max_limit, num_personalities + num_active_users + 2)
- Example: 4 personalities + 3 users = keeps 9 messages (4+3+2)
Tracks active users automatically as they join
Maximum configurable in config.yaml (default: 20 messages)

Dynamic vs. Explicit Personalities

Explicit (you choose):

"What do you think, Sysop and TechieHelper?"
→ Sysop and TechieHelper respond in that order

Dynamic (Bot chooses):

"@ChatMaster what should I do?"
→ Bot selects 2-4 random personalities to respond

Tips & Best Practices

Start specific: Mention personalities by name to control who responds
Use /multi_llm continue wisely: Don't let conversations run too long without compression
Create diverse personalities: Mix different styles, tones, and expertise
Set good prompts: The system_prompt defines the personality's behavior
Experiment with temperature:
- Lower (0.3-0.5) = more focused and consistent
- Higher (0.8-1.0) = more creative and varied

Troubleshooting

Personalities aren't responding

Make sure you spelled the personality name correctly
Check /personalities to see available personalities
Try mentioning @ChatMaster to trigger dynamic mode

Conversation stopped after 3 rounds

This is normal! Use /multi_llm continue to keep going
The max_rounds value is set in config.yaml

Responses are getting slow

Context might be too long - compression happens automatically at the threshold
Use /multi_llm stop to end and start fresh
The compression_threshold is set in config.yaml

Can't stop a conversation

Only the person who started it or moderators can use /multi_llm stop
Let it finish its rounds naturally