Metadata-Version: 2.4
Name: git-commit-message
Version: 0.9.0
Summary: Generate Git commit messages from staged changes using LLM
Maintainer-email: Mina Her <minacle@live.com>
License: This is free and unencumbered software released into the public domain.
        
        Anyone is free to copy, modify, publish, use, compile, sell, or
        distribute this software, either in source code form or as a compiled
        binary, for any purpose, commercial or non-commercial, and by any
        means.
        
        In jurisdictions that recognize copyright laws, the author or authors
        of this software dedicate any and all copyright interest in the
        software to the public domain. We make this dedication for the benefit
        of the public at large and to the detriment of our heirs and
        successors. We intend this dedication to be an overt act of
        relinquishment in perpetuity of all present and future rights to this
        software under copyright law.
        
        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
        EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
        MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
        IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
        OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
        ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
        OTHER DEALINGS IN THE SOFTWARE.
        
        For more information, please refer to <https://unlicense.org/>
        
Project-URL: Homepage, https://github.com/minacle/git-commit-message
Project-URL: Repository, https://github.com/minacle/git-commit-message
Project-URL: Issues, https://github.com/minacle/git-commit-message/issues
Classifier: Development Status :: 3 - Alpha
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: The Unlicense (Unlicense)
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3 :: Only
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Software Development :: Version Control :: Git
Requires-Python: >=3.13
Description-Content-Type: text/markdown
Requires-Dist: babel>=2.17.0
Requires-Dist: google-genai>=1.56.0
Requires-Dist: ollama>=0.4.0
Requires-Dist: openai>=2.6.1

# git-commit-message

Generate a commit message from your staged changes using OpenAI, Google Gemini, Ollama, or llama.cpp.

[![asciicast](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN.svg)](https://asciinema.org/a/jk0phFqNnc5vaCiIZEYBwZOyN)

## Requirements

- Python 3.13+
- A Git repo with staged changes (`git add ...`) (or use `--amend` even if nothing is staged)

## Install

Install the latest released version from PyPI:

```sh
# User environment (recommended)
python -m pip install --user git-commit-message

# Or system/virtualenv as appropriate
python -m pip install git-commit-message

# Or with pipx for isolated CLI installs
pipx install git-commit-message

# Upgrade to the newest version
python -m pip install --upgrade git-commit-message
```

Quick check:

```sh
git-commit-message --help
```

## Setup

### OpenAI

```sh
export OPENAI_API_KEY="sk-..."
```

### Google Gemini

```sh
export GOOGLE_API_KEY="..."
```

### Ollama (local models)

1. Install Ollama: https://ollama.ai
2. Start the server:

```sh
ollama serve
```

3. Pull a model:

```sh
ollama pull mistral
```

Optional: set defaults:

```sh
export GIT_COMMIT_MESSAGE_PROVIDER=ollama
export OLLAMA_MODEL=mistral
```

### llama.cpp (local models)

1. Build and run llama.cpp server with your model:

```sh
llama-server -hf ggml-org/gpt-oss-20b-GGUF --host 0.0.0.0 --port 8080
```

2. The server runs on `http://localhost:8080` by default.

Optional: set defaults:

```sh
export GIT_COMMIT_MESSAGE_PROVIDER=llamacpp
export LLAMACPP_HOST=http://localhost:8080
```

Note (fish):

```fish
set -x OPENAI_API_KEY "sk-..."
```

## Install (editable)

```sh
python -m pip install -e .
```

## Usage

Generate and print a commit message:

```sh
git add -A
git-commit-message "optional extra context about the change"
```

Generate a single-line subject only (when no trailers are appended):

```sh
git-commit-message --one-line "optional context"

# with trailers, output is subject plus trailer lines
git-commit-message --one-line --co-author 'John Doe <john.doe@example.com>'
```

Use Conventional Commits constraints for the subject/footer only (body format is preserved):

```sh
git-commit-message --conventional

# can be combined with one-line mode
git-commit-message --conventional --one-line

# co-author trailers are appended after any existing footers
git-commit-message --conventional --co-author copilot
```

Select provider:

```sh
# OpenAI (default)
git-commit-message --provider openai

# Google Gemini (via google-genai)
git-commit-message --provider google

# Ollama
git-commit-message --provider ollama

# llama.cpp
git-commit-message --provider llamacpp
```

Commit immediately (optionally open editor):

```sh
git-commit-message --commit "refactor parser for speed"
git-commit-message --commit --edit "refactor parser for speed"

# add co-author trailers
git-commit-message --commit --co-author 'John Doe <john.doe@example.com>'
git-commit-message --commit --co-author 'John Doe <john.doe@example.com>' --co-author 'Jane Doe <jane.doe@example.com>'
git-commit-message --commit --co-author copilot
```

Amend the previous commit:

```sh
# print only (useful for pasting into a GUI editor)
git-commit-message --amend "optional context"

# amend immediately
git-commit-message --commit --amend "optional context"

# amend immediately, but open editor for final tweaks
git-commit-message --commit --amend --edit "optional context"
```

Limit subject length:

```sh
git-commit-message --one-line --max-length 50
```

Chunk/summarise long diffs by token budget:

```sh
# force a single summary pass over the whole diff (default)
git-commit-message --chunk-tokens 0

# chunk the diff into ~4000-token pieces before summarising
git-commit-message --chunk-tokens 4000

# note: for provider 'ollama', values >= 1 are not supported
# use 0 (single summary pass) or -1 (legacy one-shot)
git-commit-message --provider ollama --chunk-tokens 0

# disable summarisation and use the legacy one-shot prompt
git-commit-message --chunk-tokens -1
```

Adjust unified diff context lines:

```sh
# use 5 context lines around each change hunk
git-commit-message --diff-context 5

# include only changed lines (no surrounding context)
git-commit-message --diff-context 0
```

Select output language/locale (IETF language tag):

```sh
git-commit-message --language en-US
git-commit-message --language ko-KR
git-commit-message --language ja-JP
```

Print debug info:

```sh
git-commit-message --debug
```

Configure Ollama host (if running on a different machine):

```sh
git-commit-message --provider ollama --host http://192.168.1.100:11434
```

Configure llama.cpp host:

```sh
git-commit-message --provider llamacpp --host http://192.168.1.100:8080
```

## Options

- `--provider {openai,google,ollama,llamacpp}`: provider to use (default: `openai`)
- `--model MODEL`: model override (provider-specific; ignored for llama.cpp)
- `--language TAG`: output language/locale (default: `en-GB`)
- `--conventional`: apply Conventional Commits constraints to the subject and footer behavior. The body format is unchanged and still includes the translated `Rationale:` line. Breaking changes are expressed with `!` in the subject line, and `BREAKING CHANGE` footer lines are not generated.
- `--one-line`: output subject only when no trailers are appended; with `--co-author`, output is a single-line subject plus `Co-authored-by:` trailer lines
- `--max-length N`: max subject length (default: 72)
- `--chunk-tokens N`: token budget per diff chunk (`0` = single summary pass, `-1` disables summarisation). For `ollama`, values `>= 1` are not supported.
- `--diff-context N`: context lines in unified diff (`N >= 0`). If omitted, uses `GIT_COMMIT_MESSAGE_DIFF_CONTEXT` when set; otherwise uses Git default (usually `3`).
- `--debug`: print request/response details
- `--commit`: run `git commit -m <message>`
- `--amend`: generate a message suitable for amending the previous commit (diff is from the amended commit's parent to the staged index; if nothing is staged, this effectively becomes the diff introduced by `HEAD`)
- `--edit`: with `--commit`, open editor for final message
- `--host URL`: host URL for providers like Ollama or llama.cpp (default: `http://localhost:11434` for Ollama, `http://localhost:8080` for llama.cpp)
- `--co-author VALUE`: append `Co-authored-by:` trailer(s). Repeat to add multiple values. Accepted forms: `Name <email@example.com>` or `copilot` (alias, case-insensitive).

## Environment variables

Required:

- `OPENAI_API_KEY`: when provider is `openai`
- `GOOGLE_API_KEY`: when provider is `google`

Optional:

- `GIT_COMMIT_MESSAGE_PROVIDER`: default provider (`openai` by default). `--provider` overrides this.
- `GIT_COMMIT_MESSAGE_MODEL`: model override for any provider. `--model` overrides this.
- `OPENAI_MODEL`: OpenAI-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
- `OLLAMA_MODEL`: Ollama-only model override (used if `--model`/`GIT_COMMIT_MESSAGE_MODEL` are not set)
- `OLLAMA_HOST`: Ollama server URL (default: `http://localhost:11434`)
- `LLAMACPP_HOST`: llama.cpp server URL (default: `http://localhost:8080`)
- `GIT_COMMIT_MESSAGE_LANGUAGE`: default language/locale (default: `en-GB`)
- `GIT_COMMIT_MESSAGE_CHUNK_TOKENS`: default chunk token budget (default: `0`; for `ollama`, values `>= 1` are not supported)
- `GIT_COMMIT_MESSAGE_DIFF_CONTEXT`: default unified diff context lines (`0` or greater). If unset, Git default is used (usually `3`).

Default models (if not overridden):

- OpenAI: `gpt-5-mini`
- Google: `gemini-2.5-flash`
- Ollama: `gpt-oss:20b`
- llama.cpp: uses pre-loaded model (model parameter is ignored)

## AI-generated code notice

Parts of this project were created with assistance from AI tools (e.g. large language models).
All AI-assisted contributions were reviewed and adapted by maintainers before inclusion.
If you need provenance for specific changes, please refer to the Git history and commit messages.
