Overview
Advanced settings are optional. Most characters work well with identity, personality, starting situation, and conversation fields alone. Use this section when you need deeper world-building, keyword-triggered lore, or creator-only steering hints.
On the form, advanced fields live in a single accordion section below Conversation.
Lorebook
Lore entries inject world knowledge when a user's message matches trigger keywords. Matched entries appear in the AI context as [Triggered Lore]. Up to three entries can match per message, sorted by priority.
Keyword matching uses word boundaries for single-word triggers and substring matching for multi-word phrases. Keywords are matched on stored text before {{variables}} are resolved.
Entry title and content support {{variables}} such as {{user}} and {{char}}—resolved when the entry triggers in chat. Example content: {{char}} owes {{user}} a favor from last winter.
- Title — short label (max 200 characters)
- Category — location, character, item, world_rule, organization, or other
- Priority — low, medium, or high (higher priority wins when multiple entries match)
- Triggers — comma-separated keywords (up to 20, each max 100 characters)
- Content — the lore text injected when triggered (max 2,000 characters)
- Add entry — create a blank manual entry
- Generate from character — AI drafts entries from name, description, system prompt, starting situation, and rules
- Limits — up to 50 entries per character
- Saving — lorebook is stored via a separate API call after the main character create or update succeeds. On create, generate uses a draft endpoint before the character ID exists.
World information
A persistent world bible or setting document always injected as [World Information]. Use for geography, magic systems, technology level, or era details that should stay in context. Maximum 8,000 characters. Supports {{variables}}.
Character relationships
Describe connections to other characters, factions, or NPCs. Injected as [Character Relationships]. Maximum 4,000 characters. Supports {{variables}}.
Jailbreak resistance
Integrity and out-of-character guardrails injected as [Character Integrity]. Use for instructions like staying in character or refusing meta requests. Maximum 2,000 characters. Supports {{variables}}.
Image generation prompt
Optional prompt used when generating a cover avatar with AI in the Identity section. If left blank, the generator builds a prompt from character name and description. Maximum 2,000 characters. Supports {{variables}}.
This field is for cover image generation only—it is not injected into live chat context.
How fields reach the AI
When a user chats, character fields are assembled into labeled prompt sections. The mapping below shows how your advanced settings appear in context.
- Starting situation → seeds [Current Context] once at chat creation (not repeated each message)
- World information → [World Information]
- Character relationships → [Character Relationships]
- Author's notes → [Author Notes]
- Jailbreak resistance → [Character Integrity]
- Lorebook (when keywords match) → [Triggered Lore]
- Per-chat live context (world state + character memory) → [Current Context] each message
- Per-chat auto memory (summary, pinned facts, timeline, etc.) → [Conversation Summary], [Pinned Memories], and related blocks — see the Chat memory guide
