As a developer in today’s LLM climate, I’ve been catching myself sending big prompts and accepting huge changes without even bothering to seriously look at the code. I skimmed the changes, glanced at the output, and went ‘oh yea thats fine’. I was turning into a vibe coder by fatigue. It wasn’t because I trusted the model, but because reviewing the output was much harder than just accepting it. That’s why I built Vestige, a Vscode extension to review and understand LLM changes easily. Vestige takes a large change and splits it into smaller, semantically-grouped chunks you can review one at a time. It uses tree-sitter to parse the code, graph analysis to find the dependencies between changes, and an LLM to group and summarize them into features. The result is shown as a repo graph that you can step through and see at a glance how changes propagate. The UI is the part I'm least sure about. I don't know yet what the right way to present this is, which is mainly why I'm posting. I'd rather hear that the graph view is wrong now than after I've sunk more time and energy building on top of it. Thanks for your attention and hope to hear from you!

Supported Languages: TypeScript/JavaScript, Python, Go, Java. Route extraction only supports: Express, Go, Python, Serverless. Note on your code and my API: Please don't point it at a huge repo… I’m paying for the API calls and you'll get throttled. Note on the source: Vestige is closed source. I'm hoping to turn it into a business eventually (free for individual developers, paid for enterprise). If that's a dealbreaker for you, I understand and I’d still like to know what you think. I haven’t completely ruled out open source yet, and your honest feedback would seriously affect my decision.

Thanks for taking the time to look at my work, wishing all of you happy hacking~