gmuse 0.3.0

AI generated commit messages.

gmuse

AI generated git commit messages in the shell using LLMs.

Highlights

• AI-powered shell completions (zsh, experimental) — context-aware suggestions for git commit -m that help you generate commit messages faster.
• Fast, configurable message generation — generate high‑quality commit messages via the CLI with customizable prompts, models, and provider settings.

Quickstart

1. Install gmuse (see Installation below).
2. Set credentials for your provider:

• Interactive setup: gmuse auth set OPENAI_API_KEY
• Non-interactive environments (CI, containers, automation): export environment variables (for example OPENAI_API_KEY)

3. Load completions: eval "$(gmuse git-completions zsh)"
4. Stage changes: git add .
5. Test: git commit -m <TAB> — gmuse will suggest a message; confirm to use it.
6. Alternatively, generate a commit message directly: gmuse msg
7. Preview the prompt without calling LLM: gmuse msg --dry-run

See Completions docs for configuration and how to persist the completion across sessions.

Installation

# Basic installation
pip install gmuse

# With clipboard support
pip install gmuse[clipboard]

Provider Setup

gmuse supports 100+ LLM providers via LiteLLM.

Interactive setup (stores credentials in your OS keyring):

gmuse auth set OPENAI_API_KEY
gmuse auth status

Auth command details: https://gmuse.readthedocs.io/en/latest/reference/cli.html#gmuse-auth

Non-interactive environments (CI, containers, automation):

export OPENAI_API_KEY="sk-..."          # For OpenAI
export ANTHROPIC_API_KEY="sk-ant-..."  # For Anthropic
export GOOGLE_API_KEY="..."             # For Google
# Or configure in ~/.config/gmuse/config.toml

Provider selection is auto-detected from credentials (environment variables first, then keyring when env values are unset or blank).

Default model choices (cost-efficient variants):

• openai: gpt-4o-mini
• anthropic: claude-haiku-4-5
• cohere: command-light
• azure: gpt-4o-mini
• gemini: gemini/gemini-flash-lite-latest

These defaults are chosen to prioritize lower cost and latency for short, high-throughput tasks like commit-message generation. You can override via GMUSE_MODEL env var or the --model CLI option.

Zsh completions (experimental)

Generate a Zsh completion script that provides AI-powered commit message suggestions for git commit -m.

Quick install:

# Add to your ~/.zshrc so the completion is loaded on shell startup
eval "$(gmuse git-completions zsh)"

Configuration:

• GMUSE_COMPLETIONS_ENABLED (default true) — enable/disable completions
• GMUSE_COMPLETIONS_TIMEOUT (default 3.0) — generation timeout in seconds
• GMUSE_COMPLETIONS_CACHE_TTL (default 30) — cache TTL in seconds

See the documentation for details: https://gmuse.readthedocs.io/en/latest/how_to/completions.html