• uv tool install rubber-ducky (or uv pip install rubber-ducky inside a project)
• Install ollama
• ollama pull qwen3:4b (first time and then you can just have application in background)
• Run with ducky <path> or ducky <question>

You will need Ollama installed on your machine. The default model used for this project is qwen3:4b.

For the first installation you can run ollama pull qwen3:4b and it should pull the necessary binaries for you.

Ollama is also great because it'll spin up a server which can run in the background and can even do automatic model switching as long as you have it installed.

Install through PyPI with uv:

uv tool install rubber-ducky

If you prefer a project-local installation, use uv pip install rubber-ducky from within your repository.

ducky

or

ducky <question>

or

ducky -f <path>

ducky --file <path> --prompt <prompt> --directory <directory> --chain --model <model> --quick --command

Where:

• --prompt or -p: Custom prompt to be used
• --file or -f: The file to be processed
• --directory or -d: The directory to be processed
• --model or -m: The model to be used (default is "qwen3:4b")
• --quick or -q: Quick mode for concise responses
• --command or -c: Ask the model for a single bash command and print it in place

This repository ships a pyproject.toml and uv.lock, so the recommended workflow is:

uv sync
uv run ducky --help

uv sync will create an isolated environment with the project dependencies, while uv run lets you execute CLI commands without activating the environment manually.

When you just need a command suggestion, add --command (optionally paired with --quick).

ducky --command --prompt "install dependencies and start the server"

The model replies with a single bash command. Press Enter to execute it immediately, or type a new natural-language prompt (history is preserved) to request another suggestion. Prefix a line with ! if you'd like to run a custom command without calling the model again.