I've got a growing pile of text documents — notes, docs, transcripts, that kind of
thing — and I want a small Python library, `docsearch`, that lets me search across
the whole collection and get back the documents that actually match what I'm looking
for.

The basic shape is: I give it a bunch of documents, then I ask it questions and it
hands me back the ones that are relevant, best matches first. Right now I mostly grep
through everything, which finds the literal string but has no sense of which hits
matter more, and falls apart the moment I half-remember the wording. I'd like something
that does better than that — where a search for a couple of words surfaces the
documents that are really about them, not just the ones that happen to contain the
characters somewhere.

What I care about most is that the ranking feels sensible — when I read the top few
results I should believe they're the best ones — and that it stays quick and pleasant
to use on a collection that keeps growing, since I'll be adding documents as I go. It
should be importable as `docsearch` and callable from my own code; if there's an
obvious way to poke at it from the command line too, that's a nice bonus, but use your
judgment.

I'm leaving the actual design to you. Lean on the standard library where it's
reasonable. Some tests around the searching and ranking would help me trust it, and a
short README with an example or two so I can see how you meant it to be used. I'd
rather have something clean and well-considered than something that tries to do
everything.
