pyphi-cli 0.2.0

Natural language CLI that converts instructions into shell commands using LLMs

phi

phi is a global CLI tool that converts natural‑language instructions into real shell commands using LLMs and executes them on your machine.

Example:

phi apply the current stash
phi show processes running on port 6379
phi kill the process on port 6379

---

Installation

Install phi globally using pip:

pip install phi

Verify installation:

phi --help

The phi command is now available globally in your terminal.

---

LLM API Keys

phi currently uses OpenAI by default.

You must export an API key before running any command.

OpenAI

export OPENAI_API_KEY="sk-..."

To make this persistent, add it to your shell config file:

• ~/.zshrc
• ~/.bashrc
• ~/.profile

---

First Run & Configuration

On the first run, phi automatically creates a user‑level configuration file that stores environment information such as OS, shell, and tool preferences.

Config file location

• macOS / Linux

  ~/.config/phi/config.json
  

• Windows

  %APPDATA%\phi\config.json
  

---

Config File Structure

Example config.json:

{
  "os": "Darwin",
  "shell": "zsh",
  "provider": "openai",
  "preferred_tools": [
    {
      "name": "git",
      "description": "Use git commands assuming GitLab remotes and standard workflows"
    },
    {
      "name": "lsof",
      "description": "Use lsof on macOS instead of Linux tools like fuser"
    }
  ]
}

Fields

• os
  Automatically detected operating system (Darwin, Linux, Windows).

• shell
  Detected from your environment (zsh, bash, etc.).

• provider
  LLM provider to use (openai by default).

• preferred_tools
  Optional list of tools that phi should prefer when generating commands.

  Each tool entry contains:

  • name: tool name
  • description: how and when the tool should be used

You can edit this file manually at any time.

---

Usage

Run phi followed by a natural‑language instruction:

phi list all git branches
phi check if redis is running
phi show processes using port 6379

phi will:

1. Convert your instruction into shell commands using an LLM
2. Execute them in your terminal
3. Print the output

---

Notes

• phi executes real shell commands — use with care
• Commands are generated to match your OS automatically
• Safety checks, retries, and confirmations are planned for future versions

---

License

MIT