Install: If you already have numpy, there's no installation at all. Just run the script. (pypy compatible)
Job: uQ re-encodes FASTQ files into binary, in such a way that you can compress the output with your favourite compression program and not have to rely on some other sketchier ones that may or may not be around in 10 years from now. It's only dependency is numpy, however I recommend you run it in pypy for speed.
Motivation: I've noticed a lot of people who have made FASTQ compression tools make large assumptions about the data (only 4 bases + N, N must have specific quality scores, doesn't work with variable-length FASTQ entries, etc. uQ is a different sort of philosophy, as it focuses solely on the encoding, removing unnecessary entropy, and finding the best encoding for your compressor - whatever it is. As a result, it often results in the smallest compressed files out of all FASTQ compressors. In an ideal world, someone would write a compressor for uQ files (or more specifically, structured tables in general)
Please check out the README on GitHub for more infos :)
Update: Now uses CFFI and a C memory array instead of Numpy array to speed up encoding by around 25x. This speeds up the encoding of small/large FASTQ files by roughly 5-25x (depends on how much of the work is encoding).