Add vector embedding extensions #307
Replies: 4 comments 2 replies
|
As your AI chatbot recommend, you can use Alex's build is rather old, so this repository demonstrates how you can build and use your own binary. It's really not that difficult. If you are going to use this, I recommend that you build it yourself. If you're happy with the result maybe publish, share and maintain it. I don't really use vector search so I'd like to avoid the maintenance burden. |
|
Okay. Vector search is useful for quickly finding similar rows. For example, if I store news article titles, I can find titles that are conceptually similar, not just in words. Maybe you'll find a reason to use it :) |
|
In contrast, https://github.com/sqliteai/sqlite-vector (not to be confused with |
|
The Vec1 Vector Extension by the SQLite authors as been added to the latest release. It hasn't been formally released upstream, and this port is very much unoptimized. But please experiment, and if you have a good benchmark to share, that's the first step for me to optimize it. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
I request that various good compatible extensions for vector embeddings be added.
For a discussion on vector embedding extensions, which are very useful for AI/ML, please see this preliminary chat: https://chatgpt.com/share/689e680c-9fdc-8012-835d-e2e334bf802a
To start, in the simple case, it is not necessary for the extension itself to be able to compute the embedding. To start, I can compute the embedding vector myself.
All reactions