HN Showcase
Back to browse
GitHub Repository

Find code hotspots before they find you — LLM + static analysis + Git churn

11 starsPython

Nikui – An LLM-Powered "Stench Guard" for Your CI/CD

by amirshk80·Mar 9, 2026·5 points·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemBig Brain

Stench × Churn formula beats generic LLM code analyzers at finding active rot.

Strengths
  • •Churn-weighted scoring surfaces files that are both smelly and actively changing.
  • •Multi-layer analysis: LLM + Semgrep + Simhash duplication + Flake8 metrics.
  • •Local LLM support via Ollama and MLX keeps sensitive code off external APIs.
Weaknesses
  • •LLM code analysis is crowded—Cursor, Cody, and Continue already do semantic review.
  • •No evidence this catches issues that existing static analyzers miss in practice.
Target Audience

Engineering leads and backend developers managing legacy codebases

Similar To

SonarQube · CodeClimate · Sourcegraph Cody

Post Description

I built Nikui because standard linters catch typos but miss architectural rot. It is a forensics tool inspired by Adam Tornhill's "Code as a Crime Scene."

The core idea is the Hotspot Score: Stench (LLM-detected debt) x Churn (Git commit frequency). A messy file that changes daily is a priority; a messy file untouched for years is ignored.

Similar Projects

AI/ML●●Solid

Routiium – self-hosted LLM gateway with a tool-result guard

Guards tool outputs against injection attacks, unlike LiteLLM or Helicone.

Solve My ProblemBig Brain
deadpixel
201mo ago