AI Commit

Inspiration

Every developer knows the pain of writing commit messages. We either spend too much time crafting the perfect message, or we give up and write "fix bug" or "update". Looking at my own git history, I realized most commits were poorly documented - making code review and debugging a nightmare.

I wanted a tool that could analyze code changes and generate meaningful, standardized commit messages automatically.

What it does

AI Commit is a developer tool that:

Analyzes git diffs and generates Conventional Commits formatted messages
Scores commit message quality (0-100) with actionable feedback
Analyzes entire GitHub repositories to visualize commit quality trends
Supports multiple AI providers: OpenAI, Anthropic Claude, and local Ollama
Works as both a CLI tool and a web application

How we built it

CLI Tool: Node.js with Commander.js for argument parsing, Inquirer for interactive prompts
Web App: Next.js 14 with App Router, deployed on Vercel
Visualization: Recharts for interactive charts and graphs
AI Integration: OpenAI API, Anthropic API, and Ollama for local inference
Styling: Tailwind CSS with custom glassmorphism design

Challenges we ran into

Designing a scoring algorithm that accurately reflects commit message quality
Handling large diffs without exceeding API token limits
Making the tool work offline with Ollama while maintaining feature parity
GitHub API rate limiting when analyzing repositories

Accomplishments that we're proud of

The scoring system actually works - it catches vague messages and rewards specific ones
Beautiful visualization that makes commit quality tangible
Multi-provider support means anyone can use it, even without API keys (via Ollama)
The tool dogfoods itself - we used it to write commits for this project