You can load in whatever document you want as a source, and it'll randomly select chunks of whatever size you want (i.e. about 200 words-worth). It then gives you analysis of your performance, as well as breaking down what trigrams and words cause you the most trouble.
Combine this with ngram training to get the muscle memory of the most common chunks of english and you can really quickly improve your fluency [^2].
1: https://gitlab.com/franksh/amphetype