So, I see where you are coming from, but one advantage Google has in any vertical is the ability to do PageRank. Google crawls billions of pages, and the results they give you in any given "vertical" benefit from being ranked by the directed graph of all pages, not just the "vertical" pages. You seem to think that limiting the number of pages you look at is an advantage, but I'd suggest it's a big disadvantage. Google is very good at not returning irrelevant results, even while using them for context.

Looking at your points above, in my experience, Google already provides "perfect answers," which often link the docs. I have no problem with how current the results are - they often include recent bug reports. And I think your custom relevancy algorithms are nice, but I think you'll find that Google's results are so good, because they crawl so many pages. Their algorithms aren't generic - they are highly advanced indexing and linguistic analysis, mixed with machine learning to tune the thing.

I think that if you are trying to return the most relevant web pages for a query, it's going to be a long uphill battle to beat Google. I have never thought to myself - "Google isn't doing a good job, I wish there were a Django search engine." I can always answer my Django questions using Google.

When you are looking to make money off of something too, it's important to be a painkiller, and not a vitamin. And right now, I don't have any Django search pain.

Good luck with the thing though. I'll check back later to see how you are doing. I encourage you not to try and compete with Google though, that is to get in the game of "returning lists of webpages." If you want to make a good Django search, you'll need to find a different niche. Your point about metadata sounds interesting.