Show HN: Scanline – open-source AI PR Reviews
github.comSharing the PR Review CLI tool I built with 2 friends recently. I have found that it's surprisingly capable of finding bugs, logical errors, and typos before pushing.
You can use it on your own branches or to get a second "opinion" on colleagues code - just check out the target branch and run `scan`. Nice job Scott! We're seeing a lot of interest in this sort of thing at Reviewable.io (the best way to review code on GitHub ... manually :D) though most users have been hesitant to say "okay, then send all of our code to GPT-4". Have you tried using Llama 2 or other models that can be run locally? I tried with Llama and llama2 and saw much better results with 2