Skip to content

256thFission/coderabbitgiveusmoney

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Git Gud Bud

A "Wall of Shame" (or employment opportunity) that scrapes GitHub profiles, analyzes them, and uses CodeRabbit to generate brutal AI code reviews to create a leaderboard of hackers.

Built by Phillip, Zackery, and Hummam.

Process

Given a list of GitHub usernames, the pipeline:

  1. Scrapes profile data, commit history, and READMEs via the GitHub GraphQL API.
  2. Scores toxicity of commit messages locally using Detoxify, surfacing each user's worst commit.
  3. Computes a "sus score" from emoji density across commits and READMEs (percentile-ranked).
  4. Assigns heuristic badges from README analysis (e.g. "Empty README Enthusiast", "No Tests, No Problem").
  5. Judges code quality with CodeRabbit -- forks each user's top repo, opens a full-codebase PR, and prompts CodeRabbit (as Linus Torvalds) to grade it, roast it, and assign a badge.
  6. Exports everything to a static JSON file consumed by a React frontend.

How We Used CodeRabbit

CodeRabbit is the core of the judging pipeline (judge.py). For each target user's most-starred repository, the pipeline:

  1. Forks the repo into our account.
  2. Creates a branch pointing at the repo's oldest commit, then opens a PR from main into that branch. This forces GitHub to diff the entire codebase as new additions.
  3. Comments @coderabbitai review to trigger a full review, then posts a structured judging prompt asking CodeRabbit to act as Linus Torvalds and return a JSON block containing a letter grade (F- to A+), a one-liner roast referencing real code, and a humorous badge.
  4. Polls PR comments until CodeRabbit responds, then parses the grade/verdict/badge from the reply.
  5. Calls the CodeRabbit Custom Reports API (/api/v1/report.generate) to produce an aggregate report across all reviewed repos.

Grades are bell-curved in export.py so the distribution follows a normal curve regardless of CodeRabbit's raw grading tendencies.

The entire pipeline is resumable -- each phase (fork, PR, comment, poll, report) saves state to judge_state.json and skips already-completed work on re-run.

Stack

Layer Tool
Scraper Python 3.12, GitHub GraphQL API, multi-token rotation
Toxicity Detoxify (local, offline)
AI Judge CodeRabbit GitHub App + Custom Reports API
Frontend React 19, Vite, DoodleCSS
Data Precomputed JSON (no live backend)

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors