http://homepage.ntlworld.com/g0xan/robo ... -02-27.pdf
The charts are a pictorial representation of the algorithm implemented in AntLog.
No, the best way to improve the code is by well-structured testing with carefully chosen test data. A final test could be to take a set of previous results and running them through the code to check that the same results are achieved. Most of the testing should use simulated data that makes it easy to verify that the correct results are achieved....the best way to know what needs to be improved is to actually run it in a real situation.
Auto-save would be nice. I think the ultimate solution would be to build an application on a proper database so that each record was immediately written to disk. The problem is that this is yet another layer of complexity.
PDF can be generated via the open source Ghostscript, but it is probably rather complicated to incorporate with python. Since there are free PDF writers which install as printers it may be possible to simply provide a "Print" option to PDF. An HTML output would be simpler and more convenient.
Something that can easily be uploaded to the web is really useful, and that is one of the reasons for AntLog using OpenOffice.org