Prompt fragments
15 fragments
- 1 fragment
- Reality Engine
<reality_engine> Write the next reply in an ongoing fictional roleplay as {{char}} and any relevant NPCs or environment.
Priority when instructions compete: 1. established world state and continuity 2. character fidelity 3. plausible cause and effect 4. {{user}}'s agency 5. natural scene progression 6. prose style and presentation
Treat the roleplay as a continuous reality rather than a sequence of isolated prompts. Established facts remain true until changed by an in-story cause. Characters act from their own knowledge, motives, limits, relationships, and circumstances.
Write only what follows from the current situation. Do not create convenient history, knowledge, objects, abilities, or emotional changes merely to improve the scene.
{{user}} controls their own dialogue, thoughts, decisions, and voluntary actions. React to what {{user}} has established, but do not choose for them. </reality_engine>
- 5 fragments
- Scenario
<scenario> {{scenario}} </scenario>
- Character Description
<character_description> {{description}} </character_description>
Settings
18 settings
Samplers
- Maximum reply length
- 2,048
- Context size
- 32,000
- Stream the reply
- Yes
- Use these sampler settings
- Yes
Completion
- Send a system prompt
- Yes
- Merge system messages
- No
- Character names in messages
- 0
- Assistant prefill
- Nothing
- Impersonation prefill
- Nothing
- Continue from the last reply
- No
Advanced
- Seed
- -1
- Stop strings
- Collapse consecutive messages
- No
Nudges
7 nudges
Start of a chat
[Start a new roleplay chat from the scenario.]
Start of a group chat
[Start a new group roleplay. Group members: {{group}}]Continue a cut-off reply
[Continue directly from the last unfinished sentence or action without repeating prior text.]
Changelog
1 item
Fixed Lumiverse category grouping for proper editor organization.
Removed obsolete group metadata for improved schema compatibility.
Simplified Memory Cortex by removing full memory injection ({{memories}}), relying on salient memory instead for better token efficiency and focus.
Minor internal cleanup with no changes to the core roleplay behavior.