ghost-in-the-shell 0.1.5

CLI with OpenAI integration

Ghost in the Shell

AI Shell is a cross-platform command-line assistant that lets you use natural language to run shell commands. Whether you’re on macOS, Linux, or Windows, just type what you want to do — AI Shell will suggest the correct shell command, ask for your confirmation, and optionally execute it.

Supports both:

• One-shot mode: ai "archive.tar" → returns shell command
• Interactive mode: ai -i → opens a terminal chat with the AI

---

✨ Features

• 🧠 Natural Language to Shell Command (via OpenAI API)
• 🖥️ Works on macOS, Linux, and Windows (Python-based)
• ✅ Confirmation before execution
• 💬 Interactive chat mode for iterative CLI tasks
• 🎨 Colored terminal output for better readability
• 🪟 Optional tmux split-pane workflow (chat left, shell right)

---

Installation

Install via pip

pip install ghost-in-the-shell

Clone & Install

git clone https://github.com/codingmoh/ghost_in_the_shell.git
cd ghost_in_the_shell
pip install .

Now you can use the ai command globally.

2. Set OpenAI API Key

You need an OpenAI API key to use this tool.

export OPENAI_API_KEY="your-api-key-here"

---

🚀 Usage

One-shot prompt

ai "untar backup.tar.gz"

✅ AI suggests a shell command ✅ Asks for confirmation ✅ Runs the command if you approve

---

Interactive mode

ai -i

🧠 Open a chat session right in the terminal. 💬 Type natural language instructions like:

You: compress all PNGs in this folder AI suggests: tar -czvf images.tar.gz *.png

Type exit or quit to leave.

---

tmux Split Mode (optional)

Want to split your terminal into chat + shell?

tmux
# Then split horizontally:
Ctrl-B %
# In left pane:
ai -i
# In right pane:
use your normal shell

---

🛡️ Security Notice

Always review AI-generated commands before executing them.

---

🧱 Future Plans

• Voice input via Whisper
• Local LLM support (e.g., llama.cpp)
• Fancy TUI with textual or rich
• Command history and undo
• Plugin system for workflows

---

🧑‍💻 Contributing

PRs welcome! Ideas, issues, improvements — all appreciated.

---

📝 License

MIT

---

❤️ Credits

Built with love and caffeine by codingmoh.