0hq / tinyvector

A tiny nearest-neighbor embedding database built with SQLite and Pytorch. (In development!)
MIT License
772 stars 24 forks source link

Add table metadata and metadata filtering #6

Open 0hq opened 1 year ago

0hq commented 1 year ago

We need to allow users to add metadata to each embedding and allow querying/filtering on each index. This may be a complex issue as integrating filtering into the indexes might need to be properly architectured.

Ideally, we should build with the ability to perform more complex operations than available on other platforms. Read more about this.

finnnnnnnnnnnnnnnnn commented 1 year ago

Hi, Do we know how to actually do this, because the article is quite vague and doesn't really explain how there system works.

Exactly how this works will be covered in a later article. For now, let’s explore some of the new single-stage filtering features and test its performance.

I also cannot seem to find this article.