Ai-Commit-Gen

Ai-Commit-Gen is a Python binary that leverages various language models to generate meaningful commit messages. There's approximately a thousand such binaries out there, but none that fit into all of my workflows, so here's the 1001st.

ai-commit-gen can be called from scripts to automatically create commits, or integrated into editor workflows to create commit drafts.

Installation

pip install ai-commit-gen

Usage

ai-commit-gen assumes that the current working directory is, or is a subdirectory of, the git root.

• -p or --prompt: Override the default prompt for the language model. Available presets are "default", "refactoring", "documentation", and "mimic".

• -c or --commit: Not only generate a commit message but also commit the changes in git.

• -m or --model: Specify the model to be used. The default is "gpt-4o-mini". Supports a wide range of models from various providers through litellm.

• -d or --debug: Enable debug logging.

• --apibase: Specify a custom API base URL for the model provider.

Supported Models

Ai-Commit-Gen uses litellm, which supports a wide range of language models from various providers. Some of the supported models include:

• OpenAI models (e.g., gpt-3.5-turbo, gpt-4)
• Anthropic models (e.g., claude-2)
• Local models via Ollama (e.g., ollama/llama3.1)

For a full list of supported models, please refer to the litellm documentation.

API Keys

For cloud-based models (like OpenAI or Anthropic), you'll need to provide an API key. The key can be set as an environment variable or stored in the system keyring:

• For OpenAI: Use OPENAI_API_KEY environment variable or store in keyring as "openai_key"
• For Anthropic: Use ANTHROPIC_API_KEY environment variable or store in keyring as "anthropic_key"

If the key is not found when the script is run, you will be prompted to enter it and given the option to store it in the keyring.

Example

Here is an example of how you might use Ai-Commit-Gen:

# Stage your changes
git add .

# Just create the commit message
ai-commit-gen

# Generate a commit message and commit the changes
ai-commit-gen --commit

# Use a specific model
ai-commit-gen --model gpt-4

# Use a local model via Ollama
ai-commit-gen --model ollama/llama3.1

# Use a custom API base URL
ai-commit-gen --apibase "https://custom-api-endpoint.com/v1"