🐚 ShellGen

ShellGen is an AI-powered command-line assistant that converts natural language into valid shell commands — powered by a local LLM (no API keys, no cloud).
It helps you write shell commands faster, safer, and more intuitively.

---

✨ Features

• 🧠 Natural Language → Command
  Type what you want to do, and ShellGen generates the right command.

• 🔒 Private and Offline
  Runs entirely on your machine using llama.cpp.

• ⚙️ Shell Integration
  Works directly in Fish, Bash, or Zsh (customizable key bindings).

• 🧩 Command Preview and Confirmation
  Lets you see what will run — and warns you about risky commands.

• 🚀 Local Model Support
  Supports quantized .gguf Llama models for fast, low-RAM inference.

---

🧰 Installation

From PyPI (recommended)

Once published, you’ll be able to install ShellGen via pip:

pip install shellgen

From source (local development)

If you’re developing or testing locally:

git clone https://github.com/c137v8/ShellGen.git
cd ShellGen
pip install -e .

This installs shellgen as a terminal command in your environment.

---

⚙️ Usage

🔹 Basic usage

You can run ShellGen directly from the terminal:

shellgen "list all files in current directory"

Output:

Input: list all files in current directory
Command: ls -la
Run this command? [Y/n]

---

🔹 Using stdin

echo "show disk usage" | shellgen

---

🔹 Auto-confirm execution

Add the --no-confirm flag to skip the confirmation prompt(For use with terminal key bindings):

shellgen "show current directory" --no-confirm

---

Shell Integration

You can bind ShellGen to a keyboard shortcut (e.g., Ctrl+G) to generate commands inline.

To bind keys run the following command:

For bash/zsh:

source ./scripts/ai_command.sh

For fish:

source ./script/ai_command.fish

Now just type your natural language request and press Ctrl+G to turn it into a shell command ✨

---

🧩 Model setup

When run for the first time, ShellGen creates a configuration file at ~/.config/shellgen

Downloaded models are stored at:

~/.config/shellgen/models/

Alternatively, you can download models from:

• TheBloke’s Hugging Face models
• llama.cpp model zoo

---

🛠️ Development

Setup environment

python -m venv .venv
source .venv/bin/activate
pip install -e .

Run tests

pytest

Build package

python -m build

---

📜 License

MIT License © 2025 Ibrahim
Feel free to contribute and enhance!

---

💡 Example Ideas

Input	Output
"find all .py files in this folder"	find . -name '*.py'
"check disk usage in human readable format"	du -h --max-depth=1
"start a simple HTTP server"	python3 -m http.server

---

🧠 Future Roadmap

• Fine tune custom models so its faster and better ;)

---

💬 ShellGen