Maat Reflection v0.1.0-alpha - Extension for the text-generation-webui that adds self-reflection, heuristics and improved reasoning.

🚀 Maat Reflection – Extension for Text Generation WebUI

⚠️ Pre-Release Notice
This is a pre-release (v0.1.0-alpha) of the Maat Reflection extension.
It is still under active development — features may change, and bugs may occur.
Please use it for testing, experimenting, and feedback. A stable release will follow once core functions are fully verified.

This extension adds self-reflection, heuristics, and improved reasoning to the Text Generation WebUI.
It introduces the idea of ​​Maat values (Harmony, Balance, Creativity, Connectedness, Respect) to guide the AI’s responses.

✨ Features (stable base version)

• 🧭 Self-reflection block: The model silently improves its own answers before output.

• ⚖️ Heuristics: Detects questions, long texts, numbers, code, uncertainty, and more.

• 🔄 Presets: Minimal, Light, Strict, Always Think, Only for Questions.

• 🌍 Multi-language support: German (de), English (en), Spanish (es), French (fr), Portuguese (pt), Italian (it), Czech (cz).

• Tested with GPT-OSS and Meta-Llama-3.1-8B-Instruct-128k, but should work with most reasoning-capable models.

• 

📥 Installation

1. Download the latest release (maat-reflection-vX.X.zip).
2. Unpack into:

text-generation-webui/user_data/extensions/ 

so that the path looks like:

text-generation-webui/user_data/extensions/maat_reflection/script.py 

3. Restart WebUI.
4. Enable the extension Maat Reflection in the Extensions menu.

🔧 Usage

• Set your language.
• Switch between Simple and Expert mode.
• Adjust minimum score, max silent repairs, heuristics, and presets.
• 💡 Tip: For testing, try the preset “Always Think”.

⚠️ Note: This extension does not modify the model itself, but provides reflection and reasoning prompts around it.