Proposal:
Now we have the possibility to create a vector field that will run auto-embeddings generation but are missing the ability to do a selection.
The easiest approach is to adapt the current knn(..) method to accept a STRING instead of an exact VECTOR.
It should produce an ERROR if we pass a string but AUTO embeddings are not configured.
So the specification may be like this:
select id, knn_dist() from test where knn ( image_vector, 5, 'smth here', {ef = 2000} )
For the JSON request, instead of query_vector, we should accept query that will represent the text query for auto embeddings generation. It should also return an error in case auto embeddings are not configured at all.
Checklist:
To be completed by the assignee. Check off tasks that have been completed or are not applicable.
Details
Proposal:
Now we have the possibility to create a vector field that will run auto-embeddings generation but are missing the ability to do a selection.
The easiest approach is to adapt the current knn(..) method to accept a STRING instead of an exact VECTOR.
It should produce an ERROR if we pass a string but AUTO embeddings are not configured.
So the specification may be like this:
For the JSON request, instead of
query_vector, we should acceptquerythat will represent the text query for auto embeddings generation. It should also return an error in case auto embeddings are not configured at all.Checklist:
To be completed by the assignee. Check off tasks that have been completed or are not applicable.
Details