Wolfgang, the Mozart Data AI Query Assistant

Ask Wolfgang a question colloquially, and he'll write the corresponding query into Mozart Data's SQL editor.

Comment

Inspiration

We are trying to obsolete the analyst language (SQL) as a barrier to doing data work.

What it does

Type questions in english and get the corresponding SQL (in the editor).

How we built it

Added chat window next to the Mozart Data SQL editor. Automated ChatGPT SQL response into the SQL editor. Added new django views.

Challenges we ran into

UX of managing slowness of query returned. Hallucinated table names. Description of many thousands of tables > ChatGPT token limit.

Accomplishments that we're proud of

Deployed to production. Detecting whether the english question is appropriate for an SQL query response.

What we learned

Would be nice to demonstrate progress while attempting to return an answer (customer experience gap awaiting SQL results). Some advantages to ChatGPT3.5 vs. ChatGPT4.

What's next for Wolfgang, the Mozart Data AI Query Assistant

Make it fast and reliable for customer usage (fix bugs and ship it broadly).

Built With

Share this project:

Updates