John Wins

Software Engineer | Art Enthusiast | Loves Photography

Building QueryWell |

[John Wins @sparrowbrainz]

@DEVIN__SETH https://t.co/rcG1RfOy11

[John Wins @sparrowbrainz]

Rule #1 when building Querywell: never execute raw LLM-generated SQL blindly.

Spent today locking down read-only DB roles and an AST parser that strictly verifies queries are safe SELECTs.

Zero risk of dropping prod tables.

[John Wins @sparrowbrainz]

spent 4 hours debugging an issue just to realize i had a typo in an environment variable i set 6 months ago.

being a solo dev is 10% coding and 90% gaslighting yourself into thinking you know what you’re doing.

[John Wins @sparrowbrainz]

@lightsilver323 https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

"Building in public: We’re finishing up the AI layer for Querywell that translates plain English straight into safe, read-only MySQL queries. No accidental DROP statements, just instant answers and CSV exports. Thoughts?

[John Wins @sparrowbrainz]

@Nonosam2 Yes that's correct. First build and ship then refine.

[John Wins @sparrowbrainz]

@darrxgh Yes that is right

[John Wins @sparrowbrainz]

Building in public: We’re shipping an AI layer that safely turns natural language into MySQL queries. No heavy BI setup, just plain English straight to your database. What safety guardrails would you want to see?
#buildinpublic

[John Wins @sparrowbrainz]

@AnnaBlazickova I am working on a Natural language to sql query generation application where you can ask questions about your data and get the result from your database without writing the query. Please checkout https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

@Erica_Codes_NJ Not yet, I am still being polite. Will test it extensively.

[John Wins @sparrowbrainz]

Building in public: Shipping an AI layer over MySQL that turns 'Show top 10 churned accounts last week' into a read-only query in <2s.

Hardest challenge? Strict safety guardrails so it never runs destructive SQL.

Demo: https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

@darrxgh Thanks. If you get time please test and let me know your feedback.
https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

Hey guys, I am building a natural language to sql conversion application where you can ask questions about your data in plain English and get the output as you would from writing SQL query.

If you guys like what I am building, let's connect.

I would like you guys to try it out

[John Wins @sparrowbrainz]

Non-technical founders: How do you pull metrics from your production database today?

A) Ping an engineer
B) Struggle with raw SQL
C) Heavy BI tool (Looker/Metabase)
D) Fly blind / spreadsheet export

Curious what your workflow looks like!

[John Wins @sparrowbrainz]

@DataAnalyt2539 Yes, that's exactly how we are doing it. We only allow read-only users to connect to the database. Thats the first guardrail. Then parse the sql and only allow select statements. The select statements will be defined based on the request. Please checkout https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

Building in public: Shipping an AI layer over MySQL that turns 'Show top 10 churned accounts last week' into a read-only query in <2s.

Hardest challenge? Strict safety guardrails so it never runs destructive SQL.

https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

@darrxgh He first get the definition of the schemas for around 50 to 100 tables and use it to extract the joins and columns. For anything above around 100 tables we store the definition in vector database so the search for the definition is easier and can be used to find the joins.

[John Wins @sparrowbrainz]

@TanzilaSha9574 https://t.co/5Jn0Acm0rL

[John Wins @sparrowbrainz]

Building in public: spent the week bulletproofing our read-only query engine.

The goal? Let anyone ask MySQL questions in plain English without risking schema leaks or rogue UPDATE commands.

Natural language in, clean data out.
#buildinpublic #Build