ai-commit

Writing meaningful commit messages is a routine. Most developers write “fix” or “update”.

Problem: Writing messages manually takes time and often leads to an uninformative history.

Solution: A two-phase pipeline:

1. Semantic analysis of git diff to classify changes.
2. Message generation via LLM with strict constraints (banned vocabulary).

Result: Used daily. Saves 30-60 seconds per commit. Consistently high-quality messages in Conventional Commits format.