Word embeddings enable the use of state of the art NLP DL models.

However, you need those embeddings on inference (not just on training).

If you decide to run the inference client-side, you either have to bundle the embeddings with the client code (too impractical) or use a server API.

With this API service, you get word embeddings to run state of the art NLP DL inference client-side.

The word embeddings were taken from  fastText public repository  where they are licensed under  CC BY-SA 3.0 

The size of the word embeddings was reduced. Only the most common ones were extracted and they were reduced with PCA.