Whitelists: customize how your files are categorized

Why whitelists exist

AI File Sorter is designed to be flexible. By default, it analyzes your files and suggests categories based on their content, names, and context. This works well for general cleanup, especially when you are dealing with mixed or unknown folders.

The app already provides two categorization modes:

• More refined - allows more nuance and variety in categories
• More consistent - encourages more uniform, repeatable results

For many cases, especially everyday folder cleanup, these modes are enough. But sometimes, even More consistent is not strict enough. In some situations, you already know exactly how your files should be organized:

• you have a predefined folder structure
• you are sorting a specific dataset or project
• you need strong consistency across a large batch
• you want to avoid any variation in naming

In those cases, you don't want the model to explore - you want it to stay within clearly defined boundaries.

For example, instead of allowing variations like Work Documents, Docs, Important Files, you may want everything to fall strictly under Documents. That’s where whitelists come in. They let you define exactly which categories are allowed, so the AI works within your structure rather than creating its own.

What a whitelist actually does

A whitelist is a set of allowed categories (and subcategories) that the AI is constrained to use. Instead of asking "What category should this file go into?" you're effectively asking "Which of these categories does this file belong to?". This makes the results more consistent.

In AI File Sorter, whitelists are:

• user-defined
• selectable per run
• injected directly into the model prompt

When whitelists are most useful

Whitelists are especially helpful when:

1. You want consistency across large batches

Sorting 10 files? You can fix things manually. Sorting 10,000 files? You need constraints.

2. You already have a folder structure

Example:

Documents/
Photos/
Music/
Invoices/
Projects/

3. You are organizing repeatedly over time

Without these constraints, categories may drift. With a whitelist, your structure stays stable across runs.

4. You are using smaller local models

Local models (like 3B or 7B) are powerful, but not perfect. Giving them a bounded vocabulary reduces ambiguity and improves accuracy.

How to use whitelists effectively

Keep them small

A good whitelist is not 50 categories covering everything imaginable. It's rather 10-20 clear, distinct categories. This is not just a UX choice - it directly affects model performance.

Use multiple whitelists instead of one giant one

Instead of an Everything list with 50+ categories, create a few lists like

• Work
• Personal
• Media
• Archive

Then pick the one that matches your current task.

Combine with categorization modes

Whitelists work with both modes:

• More Refined → flexible, but constrained to your list
• More Consistent → strongly uniform output

If you want strict structure:

whitelist + more consistent = maximum control

Whitelists are not about limiting the AI. They are about aligning it with your idea of order. The model still does the hard work: understanding your files, and suggesting structure. But instead of sticking to the default system, it now works within one you define yourself.