Pylon, a source code analyser for Python powered by AI
Sep 16, 2026
What it is
Pylon is a command line interface (CLI) that combines classic static analysis with a review pass powered by an LLM. The base layer walks the abstract syntax tree (AST) of your python source codebase to catch structural issues. These can include but are not limited to:
- Unused imports
- Bare excepts
- Shadowed variables
- Overly complex function structure (nested loops, etc.)
- Style violations
- Mutable default detection
- And many more!
But something differentiates this tool from others that already do this job really well:
The LLM Layer reads code that has been flagged by the linter and surrounding file context to give the user readable explanations in an interactive manner, suggest potential fixes with human oversight in mind, and catch poor logic that static linting analysis would otherwise miss such as inconsistent style or misleading naming schemes.
Why am I building it?
After starting a Ray Tracer in C++, I wanted something more practical to work on in the mean time, especially something that could actually be used / useful rather than simply cool or just technically impressive. A tool like a smart-linter has everday utilisation in mind as well as industry relevance, being built in Python. Python is an uncompiled language which is versatile and commonly used in scripting, mathematics, and espcially machine learning; fields I'm interested in pursuing and making a difference in. I am bringing AI into the stack not just as a "throw AI into a project" thing but rather a decision where I believe AI naturally improves the quality of output.
Current progress
I have gone ahead and implemented a few rules already, and next I will be implementing a "rule decision machine." More to come on additional rules, the LLM integration, CLI tooling, finalising the stack etc.
P.S: Pylon is short for Python linter & optimization navigator but that might change...