PseudoTest AI

• 

  PseudoTest AI uses NVIDIA Nemotron to generate stronger Python tests, reducing weakly tested functions.

Inspiration

Passing tests don’t always mean the code is properly tested. We wanted to find the bugs that existing tests might miss.

What it does

PseudoTest AI uses mutation testing to find weakly tested Python functions. It then uses NVIDIA Nemotron to generate stronger tests and checks whether they catch the mutations.

How we built it

We built a Python CLI using pytest, AST analysis, mutation testing, and the NVIDIA Nemotron API.

Challenges we ran into

Finding the right tests for each function, safely running mutated code, and making sure AI-generated tests actually catch bugs.

Accomplishments that we're proud of

In our 30-function demo, we reduced pseudo-tested candidates from 6 to 2 by adding four validated AI-generated tests.

What we learned

A passing test isn’t necessarily a useful test. We also learned that AI-generated tests need to be validated, not just accepted.

What's next for PseudoTest AI

Support more Python function types, improve test generation, and integrate PseudoTest AI into GitHub Actions.

Built With

• nemotron
• python
• test