Show HN: I Built a Tool to Visualize "Bus Factor" and Knowledge Silos in GitHub
1 comments
It seems the results (knowledge silos, review cliques, and rubber stamping) would be inaccurate often since the heuristics can’t actually capture what they say they do? I would be concerned about users not having much faith in it once it returns one too many false signals.
DORA metrics are great for telling you that you are slow, but they rarely tell you why.
I built CodePulse to answer the structural questions that standard dashboards miss. Instead of just counting PRs, it scans your repo's history (metadata only) to build a Knowledge Graph of your team.
It visualizes: Knowledge Silos:* Files that are actively churned but only understood/reviewed by one person (high Bus Factor). Review Cliques:* A network graph showing who actually reviews whom, identifying isolated junior devs. Rubber Stamping:* Detecting reviews that happen in <1 minute on large PRs.
The Tech: It's a FastAPI + React app that connects via a read-only GitHub App. We strictly analyze metadata (timestamps, file paths, graph relationships) and never store your actual source code.
I'm looking for feedback on the "Risk Algorithm"—specifically, does our detection of "Risky Merges" (e.g. no approval + failing tests) match what you see in your teams?
https://codepulsehq.com