Ever noticed how code review bots are great at yelling “Your code sucks!” but terrible at noticing when your team communication does or how to support open source contributors grow? I built a Python GitHub bot that does something different: it watches how developers collaborate during pull requests. Not just “does this code work” but “are we being good teammates about it?”
The bot analyzes three things most tools ignore: - PR descriptions: Did you link the issue? Explain the “why”? Or just write “fixed stuff” and call it a day? - Review responses: When someone suggests changes, do you acknowledge them like a human or just push commits silently into the void? - Change tracking: That reviewer asked you to refactor something two days ago—did you actually do it, or are you hoping they forgot?
The poster walks through the Python architecture—webhook listeners, LLM-powered communication analysis, conversation graph tracking, and diff parsing that connects requested changes to actual commits. I’ll show how combining pattern matching with language processing models can give you a bot that understands both code AND people.
You'll leave with confidence for building tools that make your collaborative contributions better at the soft skills of software development. Plus I'll compare it to existing tools like CodeRabbit and GitHub Copilot to show where the collaboration gap lives.
Bring your war stories about terrible PR etiquette. Let's make code review culture less painful, one bot comment at a time.