jina-ai/clip-as-service ? reverse-engineered prompt

Build me a simple service and client for using CLIP with both images and text.

I want to be able to start a local server, send it a sentence or an image URL or file, and get back embeddings, plus a ranking mode where it can compare an image against a few text options and tell me which one fits best. Make it easy to use from Python, and also make it work over an API so I can call it from other apps. If it makes sense, include support for running it in a container too, and make the setup feel straightforward so someone can try it quickly without a lot of extra work.

Please make the project polished enough that I can run a basic check, see example requests and responses, and understand how to use it from the README. If you need to look up current docs online, go ahead.