Hi HN, im the author of Skylos. We built it because traditional static analyzers (like Vulture) fall apart on modern, dynamic Python frameworks like FastAPI and Pydantic. Don't get me wrong, vulture is really really good. However, in some frameworks, they either miss a ton of dead code or throw you a lot of false positives.
So.. built a hybrid AST + LLM analyzer. We just benchmarked it against 9 popular open-source libraries to see how the data actually compares in the real world.
This is our github: github.com/duriantaco/skylos and https://github.com/duriantaco/skylos-demo this is to our benchmark. the blog post above is essentially the summary
Happy to answer any questions or hear your critiques on the methodology!
Hi HN, im the author of Skylos. We built it because traditional static analyzers (like Vulture) fall apart on modern, dynamic Python frameworks like FastAPI and Pydantic. Don't get me wrong, vulture is really really good. However, in some frameworks, they either miss a ton of dead code or throw you a lot of false positives.
So.. built a hybrid AST + LLM analyzer. We just benchmarked it against 9 popular open-source libraries to see how the data actually compares in the real world.
This is our github: github.com/duriantaco/skylos and https://github.com/duriantaco/skylos-demo this is to our benchmark. the blog post above is essentially the summary
Happy to answer any questions or hear your critiques on the methodology!