The headline should probably mention that this has a LevelDB backend. It's not just an exercise in reinventing something that already works; it's a different approach to persistence, with an API very similar to that of Redis.
Instead of reinventing wheels, why not patch Redis ?
Changing every line of code results in a hefty patch.
This is not a minor patch. Modifying Redis to support a LevelDB backend would be difficult, and would probably not find its way into the main version of Redis.
I believe it will still be less work than recreating the whole thing.