Implement a SQL, Qdrant SQL? about qdrant HOT 2 OPEN

Hey @raulcarlomagno, thanks for bringing that up!

I recognize the problem of JSON filters being complicated to write sometimes, especially with explicit pydantic types in python.

I, however, don't think sql-like language is a right answer to this problem, for the following reasons:

Original SQL was designed for a very different purposes

• it expects data to be normalized, meaning each table is flatten. This is not the case for most search engines.
• vector dbs will support only a subset of original SQL functionality, which will make it basically a syntactic "sugar"
• Along with defining what to retrieve, search engines frequently require instructions of how to retrieve something. E.g. by applying re-ranking step of some kind. This is barely compatible with SQL-like languages
• It is very hard to programmatically edit and modify sql-like language, compared to json

Instead of doing this, I would rather:

• simplify the json language, like allow not to use redundant arrays e.t.c.
• introduce a DSL-type of helpers in client libraries
• improve autocomplete in web-ui

Please feel free to share any additional arguments why would you think the sql is a better alternative

from qdrant.