llm

AI-powered command-line tool for executing shell commands and JavaScript code using natural language.

Table of Contents

• About
• Features
• Prerequisites
• Installation
• Configuration
• Usage
• License

About

LLM CLI is a cross-platform command-line tool that leverages OpenAI-compatible APIs to execute tasks on your computer using natural language prompts. It can run shell commands (PowerShell on Windows, sh on Linux/macOS) and JavaScript/TypeScript code through Bun runtime.

Tech Stack

• Runtime: Bun - Fast JavaScript/TypeScript runtime
• Language: TypeScript
• API: OpenAI-compatible chat completions

Features

• Natural Language Interface - Describe what you want to do in plain English
• Cross-Platform - Works on Windows (PowerShell), Linux, and macOS (sh)
• JavaScript Execution - Run JavaScript/TypeScript code using Bun
• Interactive Processes - Detects when processes wait for input using LLM analysis
• Yolo Mode - Skip confirmations for faster execution
• Configurable - Easy API and behavior configuration

Installation

From NPM (Global)

npm install -g @cypherpotato/llm

From Source

git clone https://github.com/cypherpotato/llm.git
cd llm
bun install
bun run build

Configuration

Before using, configure your API credentials:

# Set API endpoint
llm config --set openai.endpoint "https://api.openai.com/v1"

# Set API key
llm config --set openai.apikey "sk-..."

# Set model (optional, defaults to gpt-4o)
llm config --set openai.model "gpt-4o"

Configuration Commands

| Command | Description | |---------|-------------| | llm config --set <key> <value> | Set a configuration value | | llm config --get <key> | Get a configuration value | | llm config --list | List all configuration values | | llm config --remove <key> | Remove a configuration value | | llm config --clear | Clear all configuration |

Configuration Keys

| Key | Description | Default | |-----|-------------|---------| | openai.endpoint | API endpoint URL | - | | openai.apikey | API key for authentication | - | | openai.model | Model to use | gpt-4o | | yolomode | Skip confirmations globally | false | | idle.timeout | Input detection timeout (ms) | 8000 |

Usage

Basic Usage

# Execute a task with natural language
llm "list all files in the current directory"

# Convert files using ffmpeg
llm "convert all .png files to .jpg using ffmpeg"

# Run with yolo mode (skip confirmations)
llm -y "create a hello.txt file with 'Hello World' content"

Yolo Mode

Yolo mode skips confirmation prompts, allowing for faster automated execution:

# Per-command yolo mode
llm -y "your prompt"
llm --yolo "your prompt"

# Enable globally
llm config --set yolomode true

Interactive Processes

When a command waits for user input, LLM CLI automatically detects it and:

• In normal mode: Shows the analysis and prompts you for input
• In yolo mode: Automatically sends the LLM's suggested input

License

MIT © cypherpotato