Auto Embeddings Search

General Discussions - English
1

I have started testing the auto embeddings feature and it seems to be working quite nice!

I was wondering if there is a way to delay the generation of iembeddings in order to preserve the same speed for inserts as without the embeddings column

Or if there is any other way to approach this matter

Also, I was wondering if there is any way to cache the embeddings for search because it seems that it’s significantly slower now when using OpenAI even for repeated searches

2

Or if there is any other way to approach this matter

Try using quantization: Introducing Vector Quantization in Manticore Search
If the delay is caused by generating the HNSW index, this may help.

Also, I was wondering if there is any way to cache the embeddings for search because it seems that it’s significantly slower now when using OpenAI even for repeated searches

Unfortunately, not yet. There’s also no open issue about it. Feel free to create a feature request to encourage the team to work on it sooner.