gerg is a command‑line agent that uses your local Ollama model to plan shell commands for a natural‑language task. It can either output the command to help, or execute it for you.

Safety-first: by default gerg asks before running anything. Use --yes to auto‑run, and see --allow-unsafe to permit risky commands.

Install

pip install gerg

Quick start

1. Make sure Ollama is running locally (default http://127.0.0.1:11434).
2. Run gerg:

gerg "list all files in my Downloads directory"

3. Approve the plan or run automatically:

gerg -y "compress the Downloads folder into downloads.zip"

4. Need multi-step planning and commands. Enable the --think flag for Chain of Thought planning and execution.
5. Change default model with:

export GERG_MODEL="DESIRED MODEL"

Examples

# Only print commands (never execute)
gerg --print "find the 5 largest files under ~/Downloads"


# Use a different model just for this run
gerg -m llama3:8b "init a git repo, make first commit"


# Work from another directory (without cd'ing first)
gerg --cwd ~/Projects/website "build the site and serve locally"

gerg --think "create a .txt file in my Documents folder with a simple rhyme"