Interesting, a lot of this is consistent with our experience!
We've been adding linters, including semantic analysis, in various ai talk2db connectors in loiue.ai that help steer better than normal thin MCPs enable. We've been having to hand roll for making it easy for folks to use splunk, kusto, etc better, so cool to see something like this for sql - we were hoping for precisely that! Our semantic analysis is driven by what's in the DB (schema, ..) x policy controls configurable by team admins, which this does out of the box for SQL.
It's cool workaround for the problem that preserves the problem. An alternative is to write the query in a more reasonable language like https://prql-lang.org/ which has a representation closer to the semantic meaning and mostly avoid the big diff in the first place.
I appreciate that! It's a culmination of years of my own pain in the engineering space and a solution anticipating the flood of AI-generated SQL coming for our databases.
8 comments
[ 2.9 ms ] story [ 25.2 ms ] threadSeems like it's doing something similar to sqlfluff lint, even supporting the same dialects.
Also the GitHub link in the docs section leads to a 404.
We've been adding linters, including semantic analysis, in various ai talk2db connectors in loiue.ai that help steer better than normal thin MCPs enable. We've been having to hand roll for making it easy for folks to use splunk, kusto, etc better, so cool to see something like this for sql - we were hoping for precisely that! Our semantic analysis is driven by what's in the DB (schema, ..) x policy controls configurable by team admins, which this does out of the box for SQL.