Milor123 Debuts Huggingface-model-filter To Clean Up Model Searches
The new release of Huggingface-model-filter provides a powerful userscript to sort through Hugging Face models using positive and negative keywords. This tool displays a floating panel on your screen that lets you hide or dim models you do not want to see. It supports regular expressions for advanced pattern matching and detects infinite scrolling automatically.
Developer Milor123 created this project to solve the frustrating experience of scrolling through hundreds of irrelevant search results. They built the entire script using vanilla JavaScript without relying on any external dependencies or build steps. Users can install it easily through popular userscript managers like ViolentMonkey or Greasemonkey.
Filtering tools for browsing models
Key Features
- Positive and negative keyword model filtering.
- Modern draggable floating user interface design.
- Automatic infinite scroll page content detection.
- Visual green and red color badges.
- Regular expression advanced pattern matching support.
- Local storage configuration and full persistence.
Anyone who frequently searches for machine learning models and struggles with cluttered search results will find this tool helpful. It allows you to quickly narrow down options by typing specific terms to instantly hide unwanted files. You can save your preferred keywords and panel position so your setup remains exactly how you left it for next time.
Development details and project notes
The entire codebase contains around 1148 lines of pure vanilla JavaScript and requires no bundler. If you enter an invalid regular expression, the script will silently ignore it and log a message in the browser console for debugging. The interface natively supports English, Spanish, and Simplified Chinese, with an optional opt-in auto-translate feature for 30 other languages.
"I built a small userscript for HuggingFace's model hub because I got annoyed trying to find specific models." Source: Reddit
Project Links
