Git Whisper | Devpost

These are the commands
It's all local and private
setup – you can configure your own APIs
Project summary
off course demo for hackers!!
last 5 commits tracking got easy
You can check specific commits
lastly changelog – there are more things we can do and add in future

Inspiration 💡

Every hacker has faced this moment: you’ve built something solid, pushed dozens of commits, and when it’s time to explain it—your mind goes blank 😅. Git history is rich, honest, and detailed, but it’s written for machines and future-you, not for demos, READMEs, or judges.

Git Whisper was inspired by that gap. We wanted a tool that could read how a project evolved and translate that history into a clear human story—what problem we were solving, how the solution changed, and what actually matters now.

What it does 🧠

Git Whisper is a local-first CLI tool that turns git commit history into plain-English explanations.

With simple commands, it can:

Summarize the entire project from commit history
Explain what changed since the last commit, tag, or release
Generate a clean changelog
Produce a ready-to-read 60–90 second demo script
Explain your most recent work in context

Instead of asking “what does this code do?”, Git Whisper answers
“what story does this repo tell?”

How we built it 🛠️

Git Whisper is built in Rust, focusing on speed, safety, and a clean CLI experience.

We parse local git history using native git commands
Commit data is structured and stored as project “memory”
We curate this history into meaningful context instead of raw logs
The structured history is sent to an AI model to extract intent, evolution, and impact
Outputs are returned as clean text suitable for READMEs, demos, and changelogs

Everything works locally—no GitHub auth, no CI setup, no cloud ingestion required 🚀.

Challenges we ran into ⚠️

Avoiding overanalysis: Reading full code diffs quickly became noisy, so we focused on intent-rich signals like commit messages and summaries.
Prompt precision: Getting consistent outputs required framing the AI as a technical storyteller, not a code explainer.
Scope control: It was tempting to add dashboards and visuals, but we stayed disciplined and shipped a strong CLI-first MVP.