I created VibeGit after spending too many nights untangling my not-so-clean version control habits. We've all been there: you code for hours, solve multiple problems, and suddenly, you're staring at 30+ changed files with no clear commit strategy.
Instead of the painful git add -p dance or just giving up and doing a massive git commit -a -m "stuff", I wanted something smarter. VibeGit uses AI to analyze your working directory, understand the semantic relationships between your changes (up to hunk-level granularity), and automatically group them into logical, atomic commits.
Just run "vibegit commit" and it:
- Examines your code changes and what they actually do
- Groups related changes across different files
- Generates meaningful commit messages that match your repo's style
- Lets you choose how much control you want (from fully automated to interactive review)
It works with Gemini, GPT-4o, and other LLMs. Gemini 2.5 Flash is used by default because it offers the best speed/cost/quality balance.
I built this tool mostly for myself, but I'd love to hear what other developers think. Python 3.11+ required, MIT licensed.
I created VibeGit after spending too many nights untangling my not-so-clean version control habits. We've all been there: you code for hours, solve multiple problems, and suddenly, you're staring at 30+ changed files with no clear commit strategy.
Instead of the painful git add -p dance or just giving up and doing a massive git commit -a -m "stuff", I wanted something smarter. VibeGit uses AI to analyze your working directory, understand the semantic relationships between your changes (up to hunk-level granularity), and automatically group them into logical, atomic commits.
Just run "vibegit commit" and it: - Examines your code changes and what they actually do - Groups related changes across different files - Generates meaningful commit messages that match your repo's style - Lets you choose how much control you want (from fully automated to interactive review)
It works with Gemini, GPT-4o, and other LLMs. Gemini 2.5 Flash is used by default because it offers the best speed/cost/quality balance.
I built this tool mostly for myself, but I'd love to hear what other developers think. Python 3.11+ required, MIT licensed.
OMG, this is exactly what we all need!
Code stuff, and let AI refactor out the mess of threads of ideas later.
Love it!