Prodigy itself will let you annotate whatever you stream in and doesn't really care about the language. The display should also be optimised for LTR languages. I actually know of several users who are working with Prodigy to create training data for Chinese and Arabic.
If you want to use spaCy, there are currently no pre-trained models, but you can start off with a blank model and add your annotations on top of it – for example, to train a new entity recognizer or text classification component. This thread has more details and examples: