Node GPT CLI

A powerful CLI tool for GPT prompts written in JavaScript, supporting both OpenAI and DeepInfra models. Requires Node.js 20+

Install

npm i -g node-gpt-cli

Configuration

The tool supports both OpenAI and DeepInfra models. On install, default parameters are written to ~/.node-gpt-config.json. You can modify them there:

OpenAI Configuration

{
  "url": "https://api.openai.com/v1/chat/completions",
  "model": "gpt-4-turbo-preview",
  "max_tokens": 1024,
  "temperature": 0.3,
  "top_p": 0.5,
  "frequency_penalty": 0,
  "presence_penalty": 0
}

DeepInfra Configuration

{
  "url": "https://api.deepinfra.com/v1/inference/meta-llama/Llama-2-70b-chat-hf",
  "model": "meta-llama/Llama-2-70b-chat-hf",
  "max_tokens": 1024,
  "temperature": 0.3,
  "top_p": 0.5
}

API Keys

OpenAI

Create an OpenAI API key and assign it to OPENAI_API_KEY environment variable:

export OPENAI_API_KEY=<YOUR-API-KEY>

DeepInfra

Create a DeepInfra API key and assign it to DEEPINFRA_API_KEY environment variable:

export DEEPINFRA_API_KEY=<YOUR-API-KEY>

Usage

Basic Usage

gpt4 what is best in life?

Pipe in prompts

echo "what is my name?" | gpt4

Using Different Models

The tool will automatically use the appropriate API based on your configuration. To switch between providers:

1. Edit your config file:

nano ~/.node-gpt-config.json

2. Update the URL and model name:

   • For OpenAI: Use https://api.openai.com/v1/chat/completions
   • For DeepInfra: Use https://api.deepinfra.com/v1/inference/<model-name>

3. Set the appropriate API key environment variable

Available Models

OpenAI Models

• gpt-4-turbo-preview (Fastest, recommended)
• gpt-4
• gpt-3.5-turbo

DeepInfra Models

• meta-llama/Llama-2-70b-chat-hf
• mistralai/Mixtral-8x7B-Instruct-v0.1
• google/gemma-7b-it
• Many more available at DeepInfra Models

Configuration Parameters

• url: API endpoint URL
• model: Model identifier
• max_tokens: Maximum response length (affects speed)
• temperature: Controls randomness (0 = focused, 1 = creative)
• top_p: Controls diversity (lower = faster)
• frequency_penalty: Reduces repetition (0 = faster)
• presence_penalty: Encourages new topics (0 = faster)

Features

• Supports both OpenAI and DeepInfra APIs
• Fast response times with optimized settings
• Easy model switching
• Pipe support for integration with other tools
• Modern Node.js features and best practices

Development

Run with debug logging:

DEBUG=true gpt4 "your prompt"

Test configuration:

npm run test