Custom Patterns

Fabric offers a robust system for augmenting humans with AI, and a key part of this flexibility is the ability to create and manage your own personalized AI prompts, known as Patterns and Prompting. These custom patterns allow you to tailor AI solutions precisely to your unique needs, keep proprietary logic private, and ensure your workflow remains stable even as the core Fabric framework evolves.

This page will guide you through setting up a dedicated directory for your custom patterns, creating new patterns, and using them effectively with the Fabric CLI.

Setting Up Your Custom Patterns Directory

To begin using custom patterns, you first need to configure a dedicated directory where Fabric can find them. This ensures your personal patterns are kept separate from Fabric's built-in patterns↗ and are safe from being overwritten during updates.

Follow these steps to set up your custom patterns directory:

1. Run the Fabric setup command:

2. Navigate the setup menu: Fabric's interactive setup wizard will present a list of configurable components. Select the option for "Custom Patterns" from the "Tools" menu.
3. Specify your directory path: You will be prompted to enter the desired path for your custom patterns directory. For example, you might choose ~/my-custom-patterns. Fabric will automatically create this directory if it doesn't already exist.

Once configured, Fabric will remember this location and automatically include patterns found there.

Creating Your First Custom Pattern

After setting up your custom patterns directory, you can start creating your own. A Fabric pattern is typically a directory containing a system.md file (and optionally other files like user.md for multi-turn conversations or example.md for documentation). For detailed information on the structure and philosophy behind Fabric patterns, please refer to the [[Patterns and Prompting|Patterns and Prompting]] wiki page.

Here’s how to create a simple custom pattern:

1. Create a new pattern directory inside your designated custom patterns folder. For example, if your custom patterns directory is ~/my-custom-patterns and you want to create a pattern named my-analyzer:

2. Create the system.md file within this new directory. This file will contain the core instructions for your AI pattern. You can use any text editor, or simply the echo command for a quick start:
   This command creates a system.md file defining a pattern that acts as a text analyzer.

You now have a functional custom pattern ready for use!

Using Custom Patterns

Once created, your custom patterns can be used just like any of Fabric's built-in patterns↗. You invoke them using the fabric --pattern flag, followed by the name of your pattern.

For instance, to use the my-analyzer pattern created above with some input text:

Fabric will process the input through your my-analyzer pattern, and the AI will respond based on the instructions you provided in ~/my-custom-patterns/my-analyzer/system.md.

For convenience, you can also set up shell aliases to invoke your custom patterns directly as commands (e.g., my-analyzer "input text" instead of fabric --pattern my-analyzer "input text").

Advantages of Custom Patterns

Custom patterns offer several compelling benefits:

• Priority System: If you create a custom pattern with the same name as a built-in pattern↗, your custom pattern will take precedence. This allows you to override or personalize existing functionality without modifying the core distribution.
• Seamless Integration: Custom patterns appear in fabric --listpatterns alongside all built-in patterns, creating a unified and intuitive experience. There's no separate command or interface needed to manage them.
• Update Safe: Your custom patterns reside in their own dedicated directory. This means they will never be affected or overwritten when you update Fabric's built-in patterns↗ using fabric --updatepatterns. This ensures your personalized AI solutions remain stable across software versions.
• Private by Default: Custom patterns are stored locally on your system and are not shared with anyone unless you explicitly choose to do so. This makes them ideal for proprietary tasks, sensitive information, or highly personal applications.

By leveraging custom patterns, you can unlock the full potential of Fabric to create a truly personalized and powerful AI augmentation framework.