This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine Every codebase has a story, but not every developer remembers how it began.
LAST COMMIT is an AI-powered codebase investigator that digs through Git history, commit messages, and code changes to help developers understand why their projects evolved the way they did.
Ever wondered why a function was rewritten, when a bug first appeared, or what happened to an abandoned feature? LAST COMMIT helps you investigate those questions using actual repository evidence.
Think of it as digital archaeology, bringing forgotten parts of software history back to life.
I'm building LAST COMMIT with Python and GitPython to analyze repository history, and Ollama to run an open-weight AI model locally.
The AI retrieves relevant commits and code changes to answer questions, with references to the evidence behind its conclusions. If the repository doesn't provide enough information, it should acknowledge that instead of making things up.
The frontend uses React and Tailwind CSS to present investigations and their supporting evidence.
Codebases can contain sensitive information, and I don't want developers to upload their entire project history just to understand it.
Running AI locally makes private repository analysis possible without relying on a hosted LLM API. Open-weight models also let me experiment with different models without locking the project into a single provider.
The AI isn't the source of truth; the code and commit history are. Its job is to connect the dots.
Old repositories are full of forgotten decisions, abandoned ideas, and mysterious fixes.
Every commit leaves a clue about someone who worked on that project. Sometimes the reason is documented; sometimes it's lost.
LAST COMMIT is about bringing those stories back to the surface.
Maybe the ghost in the machine is simply the memory of every developer who came before us.