Skip to content

README.md

Latest commit

 

History

History
75 lines (46 loc) · 2.33 KB

README.md

File metadata and controls

75 lines (46 loc) · 2.33 KB

AI CodeLint

Boost code quality as you type.

AI CodeLint uses LLMs to identify logic bugs that survive compilation and traditional linting.

Install (Recommended)

macOS/Linux:

curl -fsSL https://raw.githubusercontent.com/RussellGN/AI-CodeLint/main/scripts/install.sh | sh

Windows (PowerShell):

irm https://raw.githubusercontent.com/RussellGN/AI-CodeLint/main/scripts/install.ps1 | iex

After install:

ai-codelint --version

Development Status

AI CodeLint is in heavy development and is not production-ready yet.

  • Features and APIs can change quickly.
  • Breaking changes are expected while core behavior is refined.
  • Current releases should be treated as pre-release testing builds.

Interface

AI CodeLint consists of one standalone binary: ai-codelint.

That binary currently exposes two modes of operation via the command line:

  • CLI mode: run lint checks from the terminal, like conventional linters.
  • LSP server mode: run as a language server so IDE clients can request diagnostics.

Right now, the only IDE client available is the VS Code extension in this repository (unpublished). More editor integrations are planned later.

Inference

AI CodeLint intentionally uses OpenRouter so inference-provider and model selection is quick and easy to change and experiment with.

  • You will need an OpenRouter api key to use AI CodeLint.
  • At the time of writing, signing up is free, with free models available for use with limits.
  • Run ai-codelint --configure to set up api key and other optional runtime defaults.

Model quality note:

  • Best results so far come from frontier models.
  • Lesser models can produce highly variable lint quality and consistency.
  • Recommended model: anthropic/claude-sonnet-4.6

Examples

This repository includes intentionally buggy code samples under lint samples/.

These samples cover:

  • syntax/semantics mistakes (should be ignored by linter),
  • easy rule-based linter catches,
  • subtle logic bugs,
  • edge-case behavior and review-evasive patterns.

They are useful for evaluation, and comparing model performance.

Resources: