How to Create and Playtest a Private SpicyChat Rulebook for One Chat

1. Make a distinctly named private starter copy

Follow the first-Rulebook guide: main menu > Rulebooks > Create Rulebook > Start from a template > Starter: Affection > Use this template.

In Properties, name the copy 'Lantern Library Trial' and leave Visibility on Private. Under Meters, inspect Affection: a Bar spanning 0 to 100, starting at 30. Expected result: your identifiable private copy, not an edited shared book. Leave its cover and tags alone.

2. Change one rule and inspect its influence levels

Under AI evaluation, read the starter's conditions and impacts. Change only the sincere-compliment impact from +4 to +5; keep other rules unchanged for this comparison.

Open Influence rules and read the level instructions and status lines. Expected result: you can distinguish what changes the number from what guides the character at that level. Do not rewrite several instructions while testing one impact.

Use an adult fictional librarian and visitor. Treat Affection as invented story state, never evidence of human consent, emotional dependence or relationship health. Privately record the old and new impact so an unexpected result has a clear comparison.

3. Playtest a clear action, a quiet turn and a recollection

In Playtest, drag the meter to preview low and high levels. Slider previews save nothing. Enter a bot message followed by a user message, then select Send & evaluate.

Try bot: 'The librarian displays a handmade map.' User: 'Your map is beautifully drawn.' Inspect the trace's question, answer, threshold outcome and resulting change. Expected result: an explainable evaluation, not a guaranteed score increase.

Use Reset state before separate cases; Re-run repeats the last evaluation with slider changes retained. Cooldowns and quiet-turn counts persist between evaluations. Check a neutral line and a reference to an earlier event, which should not count as a new action. Allow for the starter's decay on quiet turns.

Keep a tiny comparison note: case, starting value, trace result and ending value. If the compliment also appears to satisfy another condition, simplify the fictional exchange instead of treating it as an isolated test.

4. Resolve editor checks and save without publishing

Open the checks chip beneath Save and inspect field errors and notes about immobile meters or empty level instructions. Keep the largest instruction block under the recommended 1,200 tokens; above 1,800 it cannot be saved.

Fix relevant errors, then select Save with visibility still Private. Expected result: a saved private book ready for the bounded chat trial. Do not switch to Public or Unlisted to complete this exercise.

5. Attach only to the intended conversation

The chat attachment guide documents chat menu > Attach Rulebook. Select your saved trial book in the chosen private-character conversation, without editing character or persona defaults.

A chat attachment overrides those defaults for that conversation. Tap the Rulebook name to inspect its identity in the explanation sheet. Expected result: the intended book and visible meter belong to this chat.

Send one harmless exchange and compare its observed result with your note. Playtest uses checks like chat, but you author both sides there; real replies come from the model and can change evaluations. A passing preview is not proof of identical live behavior. This article is documentation-based; no signed-in save, attachment or generation was tested.

When the meter is absent or moves unexpectedly

Use the troubleshooting guide to separate missing attachment from rule evaluation. Recheck the selected book and scope before changing its logic.

Character/persona defaults do not retrofit older chats; direct chat attachment applies regardless of chat age. Check persona precedence and meter visibility. Interactive features off freezes meters; Hide meters conceals them while state continues.

For no movement, inspect action wording, confidence, cooldown, Once and Only when restrictions, and whether decreases are allowed. A smaller change can reflect bounds or a higher-priority matching rule. Use the trace to change one cause, not every setting.

6. End the trial or choose a simpler alternative

Use chat menu > Remove Rulebook when finished. An underlying persona or character book may resume; otherwise the chat has no meters. Expected result: the trial override is removed and any underlying book identified.

Removal is not a transcript rollback or data-erasure check. Stop if the controls, privacy terms or costs remain unclear. Ask official support with a redacted description, not private transcripts or credentials. For a scene that needs no persistent score, keep an offline fictional outline and use ordinary chat instructions instead.

Conclusion: accept only the bounded result

Keep the trial only if you can identify the private book, explain its test outcomes and confirm the single-chat attachment. Unresolved behavior stays unresolved; a plausible character reply alone does not validate the rules.

Frequently asked questions