This project provides a web-based tool for building and managing prompts for AI image generation models, specifically targeting Danbooru and NovelAI tags. It offers features like tag classification, definitions, image examples, search, prompt construction with weighting, and support for advanced prompt engineering techniques.

How It Works

The tool leverages a structured data approach, storing tags and presets in YAML files. This allows for easy organization, searching, and modification of prompt elements. It supports importing existing tag combinations and automatically matching them. The system also facilitates the inclusion and management of AI model embeddings and hypernetworks by referencing their metadata and providing links to their storage.

Quick Start & Requirements

• Install/Run: yarn install then yarn dev for development, yarn build for production.
• Prerequisites: Node.js, Yarn. Font Awesome v6 Pro license and access to its private NPM server are required for full development; free icons can be used as a temporary substitute.
• Resources: Requires local development environment setup.

Highlighted Details

• Supports advanced prompt engineering like Prompt Editing and Alternating Words.
• Facilitates importing and managing AI model embeddings and hypernetworks with metadata.
• Includes a system for adding tag descriptions, wiki links, and illustrative images.
• Provides tools for checking duplicate tags and importing images with automatic cropping/compression.

Maintenance & Community

The project is maintained by wfjsw. Contributions are encouraged by editing data/tags/**/*.yaml and data/presets/**/*.yaml. A minimum standard for adding tags includes English and Chinese names.

Licensing & Compatibility

The project's licensing is not explicitly stated in the README. However, the requirement for a Font Awesome Pro license suggests potential licensing considerations for commercial use or redistribution of the development environment.

Limitations & Caveats

The project is noted to have compatibility issues between Masonry layout and certain UI optimizations like "Collapse." The README also mentions that .pt model files for embeddings require conversion via a provided Colab notebook, and hypernetwork models must be uploaded to a HuggingFace repository due to their size.