SillyBunny World Info Lab

See why your lorebook entries did or did not activate.

World Info Lab checks the current chat, or text you paste, and shows what SillyBunny would take from your lorebooks for its next reply.

It can help when:

• An entry should activate, but does not.
• The wrong entries keep activating.
• Your lorebook is using too much prompt space.
• You want to check a large lorebook after making changes.
• You need to update the same setting on many entries.

Running a scan never sends a message, edits a lorebook, or changes timed World Info state. It saves only a summary to recent scan history. That summary does not contain chat, lorebook content, or lorebook names. A lorebook changes only when you save a test or approve a batch edit.

Install

World Info Lab requires SillyBunny 1.7.0 or newer because it uses the public World Info scan modules introduced in that release.

1. Open Extensions in SillyBunny.
2. Choose Install Extension.
3. Paste this URL:

https://github.com/SillyBunnyTeam/SillyBunny-WorldInfo-Lab

4. Finish the installation and reload SillyBunny.

No server plugin or build step is needed.

First scan

1. Open the wand menu and choose World Info Lab.
2. World Info Lab opens as a full-page workspace.
3. Leave Current chat selected.
4. Choose Run scan.
5. Look at Activated entries to see which entries passed their scan rules.
6. Open Trace to see why every other entry was skipped.

You can also open Extensions and expand World Info Lab to use the same workbench inside the Extensions drawer. Choose Open as full page there when you want more room. Moving between the drawer and full page keeps the current workbench session and results.

You can also choose Pasted text to test a sentence without adding it to your chat.

For example, if an entry with the key dragon does not activate, Trace may show that the key was found but a character filter blocked the entry, its probability check failed, or earlier entries used the available token budget.

What the results mean

• Activated means the entry passed its matching and activation rules.
• Skipped means the entry was checked but did not pass one of its rules.
• Token use shows how much prompt space the activated lorebook content uses.
• Budget overflow means there was not enough allowed space for another full entry.
• Insertion result shows where activated content would be inserted, including its depth and role. Author's Note content is conditional on its schedule, and named outlets are inserted only when an outlet macro consumes them.
• Scan rounds show when one activated entry caused another entry to activate.

An activated entry can still produce no prompt content. A regex script may remove its content, or a named outlet may be missing. Open Insertion results in Trace to check what would actually be added.

The seed controls random probability and group choices. Using the same seed with the same chat and lorebooks produces the same random choices, which makes results easier to compare.

The four tabs

Scan

Run a check against the current chat or pasted text. The summary shows activated entries, token use, budget limits, and any important warnings.

Trace

See each lorebook entry in the order it was checked. Expand an entry to see its keys, filters, probability, groups, recursion, and token decision.

Saved Tests

Save a result in a lorebook and run it again later. This is useful for checking whether an edit changed which entries activate, how many tokens they use, or where they are inserted.

Before saving a test: the test travels inside the lorebook. It can contain chat or pasted text; scan-enabled prompt text; character, persona, scenario, creator-note, filename, and tag data; frozen macro results; lorebook names; scan settings; the random-choice seed; timed or forced entry IDs; and expected insertion results with rendered activated content. Anyone you share the lorebook with may receive this information.

Delete saved tests from the Saved Tests tab before sharing a lorebook if they contain private information. Cleaning the extension's settings does not remove tests stored in lorebooks.

The recent history list is separate. It stores only a small summary for the current SillyBunny account. It does not store chat text, lorebook names, entry content, traces, or insertion results.

Batch Edit

Change several entries in one lorebook at the same time. You can replace text or update supported settings such as order, probability, depth, position, group weight, character filters, and enabled state.

Every proposed change is shown before Save these changes to the lorebook becomes available. World Info Lab reloads the lorebook and checks it again before saving. If the reviewed entries changed in the meantime, the edit stops instead of overwriting them.

Limits

• World Info Lab copies SillyBunny's World Info rules in an independent simulator. A future SillyBunny update may change those rules before this extension is updated.
• Current chat scans start from the stored conversation. Some last-minute prompt processing used for an actual reply cannot be recreated exactly.
• Vector similarity is not simulated. The Lab will not invent similarity scores for vectorized entries.
• Batch Edit checks the latest saved lorebook before writing, but SillyBunny does not provide an atomic lorebook update API. A very small race remains if another writer saves at exactly the same time.

Development

npm ci
npm test
npm exec playwright install chromium
npm run test:browser

Development requires Node.js 20.11 or newer.

Set SILLYBUNNY_ROOT to run host contract tests against a different SillyBunny checkout:

SILLYBUNNY_ROOT=/path/to/SillyBunny npm run test:host

In PowerShell:

$env:SILLYBUNNY_ROOT = 'C:\path\to\SillyBunny'
npm run test:host

License

AGPL-3.0.