Text to SQL has been a problem many people have tried to address. I'm highly skeptical of a solution that is devoid of a human explaining the schema of a database, or intent, or even metadata.
Don't get me wrong, I'm sure Picard is making good progress, but I'd be surprised if "State-of-the-Art" is actually "useful"