I did this for the lulz and I'm not sure if this would ever be useful, but here is a locally-hosted imitation of YQL that can be used from the shell or imported into Python programs. It's an abstraction layer for interacting with various API's in a uniform way.

As of now, it only supports SELECT and does not do any caching. It's also slightly different from YQL in that it's actually SQL, and it uses a different definition format (it would be nice to support the Open Data Tables format, though).

What do you think?