// the find
anuraghazra/github-readme-stats
:zap: Dynamically generated stats for your github readmes
GitHub Readme Stats generates SVG stat cards for your GitHub profile README via a hosted API endpoint or self-deployed instance. It's a pure profile decoration tool — commits, stars, top languages, repo pins, WakaTime hours — all rendered as embeddable images. The target audience is every developer who has ever wanted more than six pinned repos.
The self-hosting path via GitHub Actions is genuinely well thought out — static SVGs committed to your profile repo sidestep the rate limit problem entirely rather than papering over it. The language stat algorithm is configurable (byte count vs. repo count weighting) and they document the math, which is more honest than most. Theme support is extensive without being a maintenance nightmare — they've explicitly frozen new theme PRs, which is the right call. Test coverage is solid: unit tests, snapshot tests, and e2e tests all present, with benchmarks too.
The public Vercel instance is openly unreliable and they say so upfront, which means every embed in millions of READMEs is a coin flip. The top languages card only looks at your first 100 repos due to GitHub API pagination limits, so for prolific users it's measuring the wrong thing. There's no way to include language contributions to other people's repos — a significant gap for open-source contributors whose primary work is upstream. The ranking algorithm (S/A+/A/A-...) is based on a weighted percentile of commits, PRs, stars, and followers, which conflates very different kinds of GitHub activity and will mislead anyone who takes it seriously.