felixangell/otzbe
how it works: scraper service will run periodically
persistent database that stores keywords (search terms) and references.
redis cache containing keyword mappings to references. each redis entry has an expiration date. the more something is searched the expiration length increases slightly - maybe this should be based off traffic? in effect, search terms that aren't looked for much will be forgotten.
set the redis cache to be in LRU mode
$ heroku redis:maxmemory redis-concave-92155 --policy volatile-lru --app otz1
docker pull felixangell/otzbe