A CLI tool that generates git commit messages and conventional changelogs from staged/unstaged diffs using local LLMs (Ollama, llama.cpp). Features privacy-first operation with no external API calls, supports conventional commits format, includes interactive mode for message refinement, auto-detects change types, and can generate CHANGELOG.md from git history.

Features

Privacy-First: All processing happens locally using Ollama - no external API calls
Conventional Commits: Generates messages following the conventional commits format
Interactive Mode: Refine and edit messages before committing
Auto-Detection: Automatically detects change types (feat, fix, docs, etc.)
Scope Detection: Identifies affected areas from file paths
Changelog Generation: Creates CHANGELOG.md from git history
Flexible Configuration: Supports environment variables and config files

Installation

pip install git-commit-message-generator

Or from source:

pip install -e .

Prerequisites

Python 3.9+
Ollama running locally
A supported model (llama3, codellama, etc.)

Install Ollama and pull a model:

# Install Ollama from https://ollama.ai
ollama serve
ollama pull llama3

Usage

Generate Commit Message

Generate a commit message from staged changes:

git-commit-gen generate

Generate from unstaged changes:

git-commit-gen generate --unstaged

Use interactive mode for message refinement:

git-commit-gen generate --interactive

Specify a different model:

git-commit-gen generate --model codellama

Write message to file:

git-commit-gen generate -o commit-message.txt

Generate Changelog

Generate a changelog from git history:

git-commit-gen changelog

Generate with a limit on commits:

git-commit-gen changelog --limit 100

Generate simple changelog (without LLM):

git-commit-gen changelog --simple

Write to file:

git-commit-gen changelog -o CHANGELOG.md

Check Status

Check system status and configuration:

git-commit-gen status

Configure Settings

Show current configuration:

git-commit-gen config --show

Set default model:

git-commit-gen config --model llama3

Configuration

Environment Variables

Variable	Description	Default
`OLLAMA_HOST`	Ollama server URL	`http://localhost:11434`
`OLLAMA_MODEL`	Default model to use	`llama3`
`GIT_COMMIT_GENERATOR_CONFIG`	Path to config file	-

Config File

Create a config.yaml file:

ollama_host: http://localhost:11434
ollama_model: llama3

Prompt Templates

Customize generation prompts by creating files in the prompts directory:

prompts/commit_message.txt - Template for commit message generation
prompts/changelog.txt - Template for changelog generation

Error Handling

Error	Solution
Ollama connection failed	Check if Ollama is running, verify host URL
No staged changes	Stage with `git add` or use `--unstaged`
Invalid format	Regenerate or manually edit
Git repo not found	Run from within a git repository
Model not found	Pull with `ollama pull <model_name>`

Development

Running Tests

pytest tests/ -v --cov=src --cov-report=term-missing

Project Structure

git-commit-message-generator/
├── src/git_commit_generator/
│   ├── __init__.py
│   ├── __main__.py          # Entry point
│   ├── cli.py              # CLI interface
│   ├── config.py           # Configuration management
│   ├── git_utils.py        # Git operations
│   ├── ollama_client.py    # Ollama integration
│   ├── message_generator.py # Commit message generation
│   ├── changelog_generator.py # Changelog generation
│   └── interactive.py      # Interactive mode
├── prompts/
│   ├── commit_message.txt
│   └── changelog.txt
├── tests/
│   ├── test_cli.py
│   ├── test_git_utils.py
│   ├── test_message_generator.py
│   ├── test_changelog_generator.py
│   └── conftest.py
├── pyproject.toml
├── requirements.txt
└── README.md

License

MIT