Test Bench

🧪 Test Bench

The Test Bench checks a world before you play it. It lives in the World Editor and works in Simple and Advanced mode.

The bench shows only what the app computes from your world. It never calls the AI, so it works offline, costs nothing and updates as you type. To see what the AI writes, play a turn.

How to Check a World for Issues

1. In the World Editor, select the Test Bench flask button at the right of the header. The World Doctor popover opens.
2. Read the findings. Errors come first, then warnings, then info.
3. Select a name on a finding to go to that item in the editor.
4. Select Fix or Fix All when a finding has one. Otherwise, edit the item yourself.

The list updates as you edit. To see the full bench, select Open Test Bench at the bottom of the popover.

How to Test Which Dictionary Entries Trigger

1. Open the Test Bench, then select the Triggers tab.
2. Paste story text into the Scene text box. If you've played this world, select Paste Last Turn to fill it from your latest save.
3. Read Dictionary. Each entry says why it fired, or why it stayed out.
4. Select a marked word in your text to jump to the entity or entry it matched.

To test scan depth, open History and paste earlier messages, oldest first.

How to Preview the Opening

1. Open the Test Bench, then select the Opening tab.
2. In Testing as, pick who you play.
3. Read the Opening Pool. Select an opening to see what it sends on turn one.
4. Select Reroll to draw new Wildcard values.

If the world has more than one starting location, pick one in the Starting Location list.

What the Bench Shows

The bench runs the same functions a real turn runs, so its results match play. It shows the text and lists the AI receives. It doesn't guess what the AI does with them.

Where a result ends in the AI's choice, the bench says so. For example, AI Context lists every place a player can go. Whether an action counts as travel is the AI's call.

Opening the Bench

The Test Bench flask button sits at the right of the World Editor's header. Its badge counts findings.

• An amber badge counts new findings: ones you haven't seen yet.
• A gray badge counts every finding, after you've seen them all.

The Bench Popover

The flask opens the Bench Popover first. It holds only the World Doctor list, so you can fix a few findings without the full bench. Select Open Test Bench at its foot for the full panel. Select the flask again to close it.

The Full Panel

The full panel has a tab for each Instrument. Each Instrument answers one question about your world.

Tab	Answers
Issues	What is wrong with this world?
Triggers	What does this text make fire?
AI Context	What does the AI get from this location?
Opening	What does a new game look like?

On desktop, the panel can sit inside the editor's list panel or beside it. Select Pop Out or Embed in Editor in the panel's header to move it. The app remembers your pick. On mobile, the panel opens as a sheet over the editor.

Testing As and At

Triggers, AI Context and Opening read the bar under the tabs: Testing as who you play, at a location.

• Testing as lists the traits in groups where the player picks one. Anyone tests with no pick.
• at lists every location. Nowhere tests with no location.

Your pick stays when you switch tabs. A broken pin on your Testing as pick shows in red under the bar.

Issues

The Issues tab is the World Doctor. It checks your world's structure: broken links, unused placeholders, stats that start out of range and more. It never judges your writing.

• Publish Size shows how much of the publish limit the world uses. Images, sounds and 3D models count toward it.
• Each finding is one row with the items it names. A New tag marks a row you haven't seen.
• Fix repairs a row when the fix is safe. Fix All repairs every item in the row.
• The eye button dismisses a row. Open dismissed at the bottom, then select the restore button to bring it back.
• Mark All Seen clears the New tags. Closing the bench also marks the list seen.

Some checks run only when you ask:

• Check Stat Code runs each stat's code and lists what fails. It shows in Advanced mode when a stat has code.
• Check Sources asks the server whether the library items this world's copies follow still exist. It shows when a copy follows a published item. A copy whose source is gone gets a repair list and Apply.

In Simple mode, findings about hidden fields fold into one line. Switch to Advanced mode to see them.

Triggers

The Triggers tab is the Activation Tester. Paste text, and it shows what that text makes fire.

• Entities Present lists each entity the text names, and whether its name, part of its name or an alias matched. A name only inside quotes is a mention, so it doesn't count.
• Dictionary lists every entry, book by book. A fired entry shows the keyword that matched. An entry that stayed out says why, for example a keyword that almost matched.
• Rendered Context shows the lore text the AI would get, with its token cost.
• History holds earlier messages, so you can test an entry's scan depth.

Semantic adds meaning-based matches. It's off each time you open the bench, so a keyword result never looks like a semantic one. It needs embeddings, which the app builds when you play the world with Semantic Lore on.

Matching warnings from Issues show on the rows they name, with the same Fix.

AI Context

The AI Context tab shows what the AI gets from the location in the at list.

• The top line estimates a turn's token cost from here.
• Context Blocks lists each block of text the prompts can use, with its cost. Select a block to read it.
• Destinations lists every place a player can go from here. A place not on the list can't be reached.
• Entities the AI Is Told About shows whether each entity arrives in full, as a summary or by name only.

Opening

The Opening tab shows turn one of a new game for your Testing as pick.

• Persona picks who you play, when the world has persona choices.
• Opening Pool lists every opening a new game here can draw, with its chance.
• Stats at Game Start shows each stat's starting value. Drag a slider to see which descriptor the AI gets at other values. Dragging never changes the world.
• Active Traits lists the traits in force at the start, with their pins and stat changes.
• Placeholder Rolls shows what each Wildcard drew, every value's chance, and the chance of a repeat.
• First Prompt shows the System Prompt and Opening User Turn. An Opening Narration shows Page One instead, since no request goes out on turn one.

The First Prompt uses the default prompts. Custom prompt presets aren't read here.

Related

• 🛠️ World Editor
• 📖 Dictionary
• 🎬 Openings
• 🔗 Linked Content