pr-pilot-cli 1.6.0

CLI for PR Pilot, a text-to-task automation platform for Github.

Install | Documentation | Blog | Website

PR Pilot CLI

PR Pilot gives you a natural language interface for your Github projects. Given a prompt, it uses LLMs (Large Language Models) to autonomously fulfill tasks by interacting with your code base and Github issues.

Using Jinja templates, you can create powerful, reusable commands that can be executed by PR Pilot.

🛠️ Usage

Open a terminal and ls into a repository you have installed PR Pilot.

📋 Examples

You can run a prompt directly on the command line:

pilot "Tell me about this project!"

Ask PR Pilot to edit a file for you:

pilot --edit cli/cli.py "Make sure all functions and classes have docstrings."

Generate code quickly:

pilot -o test_utils.py --code "Write some unit tests for the utils.py file."

Ask about something on your screen:

pilot --snap "What is this code doing?"

Get an organized view of your Github issues:

pilot "Find all open Github issues labeled as 'bug', categorize and prioritize them"

Generate parts of your README with a template:

pilot --direct -f prompts/README.md.jinja2 -o README.md

For more examples, check out the prompts directory in this repository.

⚙️ Options and Parameters

You can change the default settings with parameters and options:

Usage: python -m cli.cli [OPTIONS] [PROMPT]...

  Main function to handle the CLI commands and options.

Options:
  --wait / --no-wait        Wait for the result.
  --repo TEXT               Github repository in the format owner/repo.
  --snap                    Adds a part of your screen as an image to the
                            task.
  -e, --edit PATH           Let PR Pilot edit a file for you.
  --spinner / --no-spinner  Display a loading indicator
  --quiet                   No pretty-print, no status indicator or messages.
  --cheap                   Use the cheapest GPT model (gpt-3.5-turbo)
  --code                    Optimize prompt and settings for generating code
  -f, --file PATH           Load prompt from a template file.
  --direct                  Do not feed the rendered template as a prompt into
                            PR Pilot, but render it directly as output.
  -o, --output PATH         Output file for the result.
  -m, --model TEXT          GPT model to use.
  --debug                   Display debug information.
  --help                    Show this message and exit.

⚙️ Configuration

The configuration file is located at ~/.pr-pilot.yaml.

🤝 Contributing

Contributors are welcome to improve the CLI by submitting pull requests or reporting issues. For more details, check the project's GitHub repository.

📜 License

The PR Pilot CLI is open-source software licensed under the GPL-3 license.