Git-GPT

A CLI tool to auto-generate git commit messages and issues using OpenAI's GPT model or Ollama.

Use ollama might be a more secure way to generate commit messages and issues. Your data will not be sent to any third-party server.

Installation

Install git-gpt via pip:

pip install git+https://github.com/ShinChven/git-gpt.git

Upgrade:

pip install --upgrade git+https://github.com/ShinChven/git-gpt.git

Development

To set up the development environment:

git clone https://github.com/ShinChven/git-gpt.git
cd git-gpt
python -m venv venv
source venv/bin/activate
pip install -e .

Project Structure

The project is organized as follows:

• git_gpt/main.py: The main entry point of the CLI application.
• git_gpt/__init__.py: Initializes the package and defines the version.
• git_gpt/config_command.py: Handles the configuration command.
• git_gpt/commit_command.py: Implements the commit message generation.
• git_gpt/issue_command.py: Manages the issue creation functionality.
• git_gpt/quality_command.py: Performs quality checks on code changes.
• git_gpt/changelog_command.py: Generates changelogs based on commits.
• git_gpt/ask_command.py: Allows asking custom questions about code diffs.
• git_gpt/request_module.py: Handles API requests to OpenAI or Ollama.

Each command is implemented in its own file for better organization and maintainability.

Configuration

Before using git-gpt, you'll need to configure it with your API settings. Run the following command and follow the prompts:

git-gpt config --api-key <API_KEY> --api-type <API_TYPE>

Options:

• --api-key: The API key to use with OpenAI (not required for Ollama).
• --base: The alternative OpenAI host.
• --model: The model to use for generating messages.
• --lang: Target language for the generated message (default is 'en').
• --max-tokens-quality: Maximum number of tokens for quality module responses.
• --max-tokens-issue: Maximum number of tokens for issue module responses.
• --max-tokens-changelog: Maximum number of tokens for changelog module responses.
• --api-type: The type of API to use ('openai' or 'ollama').
• --ollama-base: The base URL for Ollama API (default is 'http://localhost:11434').

Ollama Support

Git-GPT now supports Ollama, an open-source, locally hosted language model. This allows you to use Git-GPT without relying on external API services.

Setting up Ollama:

1. Install and set up Ollama on your local machine (visit Ollama's website for instructions).
2. Pull the desired model(s) using Ollama's CLI (e.g., ollama pull gemma2).

Configuring Git-GPT for Ollama:

To use Ollama with Git-GPT, configure it as follows:

git-gpt config --api-type ollama --ollama-base http://localhost:11434 --model <MODEL_NAME>

Replace <MODEL_NAME> with the model you've pulled in Ollama (e.g., llama2, codellama, mistral, etc.).

Default Model:

The default model for Ollama in Git-GPT is set to 'gpt-4o-mini'. You can change this by specifying a different model during configuration or when running commands.

Using Ollama:

Once configured, you can use Git-GPT with Ollama just like you would with OpenAI. All commands (commit, issue, quality, changelog, ask) will automatically use your Ollama configuration.

Note: When using Ollama, you don't need to provide an API key.

Usage

Generating Commit Messages

Stage all changes and generate a commit message:

git-gpt commit [--lang <LANGUAGE>] [--model <MODEL>] [--run-dry]

Options:

• --lang: Target language for the generated message (default is 'en').
• --model: The model to use for generating messages (default is set in config).
• --run-dry: Print the generated message without committing.

Creating Issues

To create an issue based on the diffs of the latest commit(s), run:

git-gpt issue [--lang <LANGUAGE>] [--model <MODEL>] [--max-tokens <MAX_TOKENS>] [--commit-range <COMMIT_RANGE>]

Options:

• --lang: Target language for the generated message (default is 'en').
• --model: The model to use for generating messages (default is set in config).
• --max-tokens: The maximum number of tokens to use for the issue prompt (overrides the configured value).
• --commit-range: The range of commits to consider for generating the issue.

Performing a Quality Check

To perform a quality check on the diffs of the latest commit(s), run:

git-gpt quality [--lang <LANGUAGE>] [--model <MODEL>] [--max-tokens <MAX_TOKENS>] [--commit-range <COMMIT_RANGE>]

Options:

• --lang: Target language for the generated message (default is 'en').
• --model: The model to use for generating messages (default is set in config).
• --max-tokens: The maximum number of tokens to use for the quality check prompt (overrides the configured value).
• --commit-range: The range of commits to consider for the quality check.

Generating a Changelog

To generate a changelog based on the diffs of the latest commit(s), run:

git-gpt changelog [--lang <LANGUAGE>] [--model <MODEL>] [--max-tokens <MAX_TOKENS>] [--commit-range <COMMIT_RANGE>]

Options:

• --lang: Target language for the generated changelog (default is 'en').
• --model: The model to use for generating the changelog (default is set in config).
• --max-tokens: The maximum number of tokens to use for the changelog prompt (overrides the configured value).
• --commit-range: The range of commits to consider for generating the changelog.

Asking a Custom Question

To ask a custom question about the code diffs, run:

git-gpt ask --question <YOUR_QUESTION> [--model <MODEL>] [--commit-range <COMMIT_RANGE>]

Options:

• --question: The question to ask about the code diffs.
• --model: The model to use for generating the response (default is set in config).
• --commit-range: The range of commits to consider when forming the response.

Trouble Shooting

aiohttp

If you encounter any issues concerning aiohttp, please try to downgrade python to 3.11, as this issue is reported with python 3.12:

ERROR: Could not build wheels for aiohttp, which is required to install pyproject.toml-based projects

Disclaimer

• Content generated by this tool is not guaranteed to be correct. You should always review and edit the generated content before committing.

Contributing

Feel free to fork the repository, create a feature branch, and open a Pull Request.

License

MIT License