For example, if you enter the ingredients "rye whiskey" and "vermouth", following query is generated:
@ingredients (rye SENTENCE whiskey) | vermouth
The SENTENCE operator, basically works like the & operator, just that both operands must occur in the same sentence. That is very helpful, since the index field "ingredients" contains a list of all ingredients of the recipe separated by an "!".
However Sphinx can not only do full text search. You can also filter and sort by attributes and complex expressions, that involve any attribute, the relevance from the fulll-text search, arithmetics and some built-in functions. And Sphinx does that much faster than any RDBMS does. I never managed to generate a sphinx query that took a measurable amount of time, on my 3 years old notebook. ;)
The sphinx index doesn't contain any HTML. However I return HTML, from the WSGI app, that serves the AJAX calls. So I guess that is what you are talking about. It just seemed to me, it would be simpler to generate the HTML for the search results on the server-side with Python than in Javascript.
By the way if you want to use Sphinx for your own project, and you have the data to index already in a MySQL or PostgreSQL database, there is no need to write a script to generate XML. Sphinx can index data from MySQL and PostgreSQL databases directly. However I was just saying that, thanks to xmlpipe support, you don't need an RDBMS just to use Sphinx. And that thanks to index attributes, Sphinx can completely replace an RDBMS in a lot of cases.