Quick Start

Getting started with gptlint is easy and should only take a few minutes to set up with a new codebase.

Prerequisites

node >= 18

Install

npm i -D gptlint

We recommended installing gptlint as a dev dependency just like eslint.

OpenAI

GPTLint defaults to using the OpenAI API, so you’ll need to sign up for an OpenAI API key. Add your OPENAI_API_KEY as an environment variable or store it in a .env file in your project’s root.

💡

GPTLint supports local LLMs and other providers, so if you don’t want to use OpenAI, see this guide.

Dry Run

npx gptlint --help

By default, gptlint uses **/*.{js,ts,jsx,tsx,cjs,mjs} to match source files. .gitignore will be respected, and optionally .gptlintignore as well if it exists.

Now let’s try running gptlint with the --dry-run flag:

npx gptlint --dry-run

⚠️

The --dry-run flag is very useful for estimating how much it would cost to run the linter for real.

Since LLM costs can add up quickly for large codebases, it’s always a good idea to make sure you understand the projected costs before running gptlint on a new project.

If you want to try out gptlint without worrying about the cost, you can run gptlint on a single source file like this:

npx gptlint --dry-run src/utils.ts

If you’re ever not sure about what will be run, you can use the --print-config flag to view all of the resolved rules, settings, and source files that will be processed:

npx gptlint --print-config

If there are source files being picked up by gptlint that you’d like to ignore, you can add glob patterns to a .gptlintignore file at the root of your project to ignore them.

Real Run

Oooooohhhh now comes the fun part! Let’s run gptlint for real and see what it finds…

npx gptlint

Depending on the size of your codebase, this could take anywhere from a few seconds to a few minutes. It’ll display a summary of the linter tasks as it runs.

Once it completes, you should see something resembling this output:

Example output

This example output comes from running gptlint fresh on it’s own codebase with caching disabled. It cost a total of $0.83 USD and made 351 LLM calls to OpenAI, with the vast majority of them using gpt-3.5-turbo.

More than likely, you’ll see some lint errors that you should consider addressing. Just keep in mind that since gptlint isn’t 100% accurate (like a human code reviewer), it’s possible for gptlint to report false positives and miss false negatives, but we’re working hard to improve the accuracy of the core linting engine and built-in rules over time.

For more details on how to interpret the output, see usage and our faq. Check out the built-in rules to see what gptlint can enforce out of the box.

Overview CLI