Glossary¶

Key terms used throughout Lumiverse and these guides.

A¶

Alternate Fields: Variant versions of a character's description, personality, or scenario that can be selected per-chat.
Alternate Greetings: Multiple first-message options for a character. Chosen when starting a new chat.
API Key: A secret credential from an AI provider that authenticates your requests. Stored encrypted in Lumiverse.
Assembly: The process of building the complete prompt from preset blocks, character data, world info, macros, and chat history.
Author's Note: A hidden instruction injected at a specific depth in the conversation to influence the AI's behavior.

Block (Prompt Block): A section of text in a preset that contributes to the assembled prompt. Blocks have roles, positions, and can be enabled/disabled.
Branch: A fork of a conversation at a specific message, creating a new chat with shared history up to that point.

Character Card: A standardized format (CCSv1/v2/v3) for packaging character data, optionally embedded in a PNG image.
CHARX: A ZIP archive format for character cards that can include the card JSON, avatar, expressions, and Lumiverse modules.
Connection: A saved configuration linking Lumiverse to an AI provider (includes provider, model, API key, and URL).
Constant Entry: A world book entry that is always included in the prompt regardless of keyword matching.
Context Window: The maximum number of tokens a model can process at once (prompt + response combined).
Council: A multi-persona deliberation system where AI personas analyze the scene and provide guidance before the main generation.

Depth: How many messages from the end of the chat to insert content. Lower depth = closer to the end = more influence.
Dry Run: A test that assembles the full prompt without actually calling the AI. Shows exactly what the model would see.

Expression: An emotion-mapped image that changes dynamically based on the conversation mood.

Generation: The process of sending a prompt to the AI and receiving a response. Types: normal, regenerate, continue, swipe, impersonate, quiet.
Group Chat: A conversation with multiple AI characters who take turns responding and interact with each other.

Loom: Content blocks from packs, categorized as narrative styles, utilities, or retrofits.
Lorebook: Another name for World Book — a collection of keyword-triggered contextual entries.
Lumia: An AI persona from a pack, used as a council member or for narrative style selection.

Macro: A template variable (e.g., {{char}}) that gets replaced with dynamic content during prompt assembly.

Pack: A content bundle containing Lumia items, Loom items, and/or Loom tools.
Persona: Your identity in conversations — includes name, description, and avatar.
Preset: A saved configuration defining prompt block order, sampler settings, and completion behavior.
Preset Profile: A snapshot of block enabled/disabled states, bindable to default, character, or chat contexts.
Provider: An AI service (OpenAI, Anthropic, Google, etc.) that Lumiverse connects to for generation.

Recursion (World Info): When activated world book entries contain keywords that trigger additional entries.
Regex Script: A text transformation rule using regular expressions, applied at various stages of the pipeline.

Sampler Settings: Parameters that control how the AI generates text (temperature, top-p, penalties, etc.).
Scan Depth: How many recent messages are checked for world book keywords.
Selective Logic: Secondary keyword conditions (AND, OR, NOT) that refine when a world book entry activates.
Sidecar: A separate, usually lighter AI model used for background tasks like council tools and expression detection.
Spindle: Lumiverse's extension system for installing and running third-party add-ons.
Sticky Entry: A world book entry that stays active for a set number of turns after its keywords stop appearing.
Swipe: An alternate version of a message. Generated by regenerating a response, creating a deck of alternatives.

Token: The basic unit of text for AI models. Roughly 3/4 of a word. Used to measure context size and response length.

World Book: A collection of entries that inject contextual information into the prompt when their keywords appear in the conversation. Also called a lorebook.
World Info: Another name for World Book content and the system that activates it.