Inspiration
During my internship, a client came with a requirement to query databases through a simplified interface. I worked in a support role then, and the approach was different. DB Bot is my own take on solving the same problem - cleaner, more intuitive, and AI-native.
What it does
DB Bot lets you talk to your database in plain English. Just type something like "Find the student with the highest marks" or "Show all orders placed last month above ₹5000" - and it figures out the rest. It uses Google Gemini to generate the correct database query from your input, executes it via ORM, and returns the results instantly. Table names are encrypted to ensure schema privacy - the actual structure of your database is never exposed. No SQL knowledge needed.
How I built it
Built in Python with a three-layer architecture - UserEnd handles the interface and user input, LLM_API sends the natural language query to Google Gemini and gets back a structured database query, and DB executes it via ORM and returns the result. Table names are encrypted so even the AI layer never sees your real schema - only mapped references. Clean separation of concerns.
Challenges I ran into
The real challenge was the approach itself - figuring out how to sit an LLM between a user and a live database safely and reliably, and also how to make queries encrypted. Once that mental model clicked -the implementation followed naturally.
Accomplishments that proud of
Got a working end-to-end pipeline where a completely non-technical user can interact with a real database using just natural language - with zero SQL exposure. Adding encryption on top of that means the solution is not just usable, but actually production-safe from a privacy standpoint.
What I learned
- Gen AI in the applications.
- ORM working
What's next for DB Bot
Support for multiple DB types (MySQL, PostgreSQL, MongoDB) Robust Encryption. Command line Desktop application Role-based access so different users only query what they're allowed to. A complete Product.
Built With
- api
- orm
- python
Log in or sign up for Devpost to join the conversation.