At one point my Atomgen script used Amara, because I liked the bindery interface. However, I was having problems with wide Unicode characters, so I've updated Atomgen to use more widely supported XML libraries (lxml's objectify API is similar to Amara's bindery). For future reference, I'll recount my history with Amara.
I started off using Sylvain Hellegouarch's Atomixlib, which is
based on Amara2 or ElementTree. However, Sylvain seems to be
falling behind as ElementTree and Amara continue to evolve, so I
rewrote Atomgen to run off Amara2 directly, since Amara's
bindery interface is wonderful. Things worked fine when I was
installing Amara via
sudo apt-get install python-setuptools easy_install --prefix=~ amara