[Feature] Preload ML Models on Web Access to Improve Search Performance

I have searched the existing feature requests, both open and closed, to make sure this is not a duplicate request.

• Yes

The feature

Hi,

Currently, ML models (e.g., CLIP) are loaded on-demand, meaning they are only loaded when a search request is made. This can cause noticeable delays for the first search.

I would like to suggest a feature where ML models start loading automatically when a user accesses the web interface or opens the app, so that by the time a search is performed, the models are already in memory. Ideally, this behavior could be configurable via an environment variable or setting, giving users flexibility while improving user experience.

中文 / Chinese:

你好，

目前，机器学习模型（例如 CLIP）是按需加载的，也就是说只有在搜索请求触发时才会加载，这会导致首次搜索出现明显延迟。

我希望可以增加一个功能，使 ML 模型 在用户访问网页或打开 APP 时就开始加载，这样在进行搜索时模型已经在内存中。最好这个功能可以通过 环境变量或配置项 来控制，让用户可以选择是否启用，同时提升用户体验。

Platform

• Server
• Web
• Mobile