git-autoc 0.1.0

AI-powered git commit message generator

git-autoc

AI-powered git commit message generator. Reads your staged changes and writes a conventional commit message using any OpenAI-compatible API.

Features

• Generates Conventional Commits from your staged diff
• Uses your recent commit history for style-consistent output
• Interactive confirmation (y / e / N) before committing
• Dry-run mode with no side effects
• Three body styles: long, bullet list, paragraph
• Works with any OpenAI-compatible API (Ollama, LM Studio, OpenAI, …)
• Configuration managed from the CLI — no .env

Usage

Stage the changes you want to commit, then run:

git autoc

The tool generates a commit message, shows it, and asks how to proceed:

• y — commit with the generated message
• e — open the editor to adjust the message before committing
• N (default) — abort, keep changes staged

To generate without committing (or to preview the message first):

git autoc --dry-run

Options

Option	Alias	Description
--dry-run	-n	Generate a message without committing
--long	-l	Allow an optional body paragraph or bullet list
--list		Always add a bullet-list body
--desc		Always add a paragraph body

--long, --list, and --desc are mutually exclusive.

Body styles

Style	Flag	Behavior
Default		One-line message, no body
Long	--long	Body added only if the change is worth explaining
List	--list	Body always present, as a bullet list
Description	--desc	Body always present, as a single paragraph

Configuration

Configuration lives in ~/.config/git-autoc/config.toml.

Show the current configuration (creates a default file on first run):

git autoc config

Set a value interactively:

git autoc config set openai_model

Settings

Key	Default	Description
openai_base_url	http://localhost:11434/v1	Base URL of the OpenAI-compatible API
openai_model	model	Model name to use
openai_api_key	(empty)	API key; leave empty for local providers
openai_temperature	0.2	Sampling temperature (0–1, lower is more deterministic)
openai_max_tokens	512	Maximum tokens in the generated response

openai_base_url and openai_model are required and must be non-empty; the tool refuses to run otherwise.

Development

make run     # run the CLI
make format  # format code (ruff)
make check   # type check (ty) + lint + format check
make test    # run tests (pytest)

Project layout

src/git_autoc/
├── main.py           # CLI (typer), wires all layers
├── core/config.py    # settings, TOML loading, config management
├── git/utils.py      # git operations (diff, log, stage, commit)
└── llm/              # prompt building, provider, commit styles

Stack

• CLI: typer + rich
• Config: pydantic + tomllib
• AI client: openai (OpenAI-compatible API)
• Quality: ruff, ty, pytest