This IPython Notebook goes over the basics of Collaborative Filtering, specifically covering distance metrics, the nearest neighbor problem, and how to generate recommendations. I used the Pandas Python library for most of the heavy lifting.
The example in this particular notebook is doing User-Based Collaborative Filtering using the Euclidean distance metric, which is pretty much the simplest recommendation engine you could build.
This started as a lecture I gave about recommendation engines at UMBC (75 minutes), which goes over the basics of a user-based recommendation engine.
To run this, you'll need to install pandas, matplotlib, and ipython.
This tutorial uses the MovieLens data set. It's not included in the repo, but you can get it here: http://grouplens.org/datasets/movielens/ I didn't commit it in the repo because it is big.