We’re excited that the new API is finally live. Most of the important new stuff is going on behind the scenes with the latest version of our graph database, but here’s a rundown of the new goodies:
- Full read-write support for all items, rather than having to upload full database dumps. This includes methods for adding and removing tags or links as well as rewriting items. Items can also be added or removed from the database.
- Better results for obscure or tag-specific searches. You can see an example of this in our last blog post.
- Faster results. Our new database is several times faster than our already speedy previous version.
- Max results parameter in the search results, configurable from 1 to 100 results.
- “Ignore linked” option to avoid returning recommendations for items a user has already interacted with — i.e. items they’ve already purchased or clicked on.
The bulk of the effort in the last months has been getting our graph database up to snuff as a proper database. We’ve had a lot of questions about our graph database, so the next blog entry will be full of technical information on it. We’ll also now be creating new accounts again, which we’d been holding off on for the last few weeks as they’d need to be moved over to the new database format. All in all the upgrade went as planned and services were down for only around 5 minutes.
We’re already at work merging in the next batch of features which will add special elements for common use cases — e.g. ratings, purchases and click-tracking and will also bring more a more advanced recommendations model that’s been brewing in one of our experimental development trees.