Here's an example from the pgvector docs
SELECT * FROM items WHERE id != 1 ORDER BY embedding <-> (SELECT embedding FROM items WHERE id = 1) LIMIT 5;
Notice how the right-hand side of the <-> is a select statement that should return only one value.
The pgvector-python sqlalchemy examples and test cases only ever seem to use a hardcoded embedding like [1, 1, 1] on the right-hand side.
Can it be made explicit in the documentation if a select like (SELECT embedding FROM items WHERE id = 1) can be used as an argument to the cosine_distance method, for example? If right-hand side can be a select, an example query would be helpful. If the right-hand side cannot be a select, I think that should be highlighted as a gap in the sqlalchemy support
Here's an example from the pgvector docs
Notice how the right-hand side of the
<->is a select statement that should return only one value.The pgvector-python sqlalchemy examples and test cases only ever seem to use a hardcoded embedding like
[1, 1, 1]on the right-hand side.Can it be made explicit in the documentation if a select like
(SELECT embedding FROM items WHERE id = 1)can be used as an argument to the cosine_distance method, for example? If right-hand side can be a select, an example query would be helpful. If the right-hand side cannot be a select, I think that should be highlighted as a gap in the sqlalchemy support