Cachetools
I was looking for a simple, off-the-shelf caching library for Python. Dissatisfied with what I could find, I wrote my own, and called it cachetools.
Cachetools provides a simple set/get interface to cache objects manually, either in-memory or to a filesystem shelve-based store; a caching decorator for functions, that indexes return values based on the sha1 hash of a pickle of the call arguments; and some logging facilities. In the future I plan to add proper expiry logic (right now it supports simple LRU, only for in-memory caches) and use SQLite as a backend, since it can easily store and index metadata for cached objects.
I'd love to hear if you find cachetools useful, or if you have any ideas for improvements.

Reader Comments
lazy q: what's wrong with beaker ( http://pypi.python.org/pypi/Beaker ) ?
I wasn't aware of Beaker. It looks nice, but there are several problems:
* it's centered towards web applications;
* it doesn't handle cache strategies except time-based expiration;
* decorator only accepts positional arguments;
* cache keys are as large as your input data (I'm parsing thousand-line CSS files so it's a non-starter).
Still, it's a really nice piece of work. I'll talk to Ben and see if we can work something out.