gpterm 0.2.9

A Terminal emulation to interact with OpenAI's GPT models

GPTerm

Interact and chat with OpenAI's GPT-3 models via an REPL terminal emulation on macOS. Including voiced output, image generation and model selection.

Installation

To install GPTerm, simply execute the following command in your terminal:

pip3 install --upgrade gpterm

Note: GPTerm works best when installed and run under a Conda environment. See Known Issues

To install under conda (after installing anaconda or miniconda):

conda create --name gpterm python=3.10.8
conda activate gpterm
pip3 install --upgrade gpterm

Usage

gpterm

Note: GPTerm uses the OpenAI API which requires an OpenAI API Key

• Setting an OpenAI API Key:

  You can get your OpenAI API Key from OpenAI

  Storing it in the file ~/.openai-api-key will allow to run GPTerm without manually providing it with the API key. In the command below replace api_key with your OpenAI API Key.

  echo api_key > ~/.openai-api-key
  

  Alternatively, you can provide your API Key to gpterm by passing it with one of these flags:

  --api_key <api_key>

  --api_key_path <path/to/file/containing/api_key>

  Or by setting a shell environment variable OPENAI_API_KEY with your API Key before running gpterm

  
  

• Setting the Color Theme:

  For the best experience a terminal with a dark background theme is recommended. (Like the one provided by iTerm)

  However, if you are using a light background terminal start gpterm with the following flag:

  --theme light

  Options are light or dark (default)

Features

GPTerm provides the following features as supported by OpenAI's API:

• Streamed output
• Multiline input
• Preserves conversation context including the option to reset it
• Voiced output using text to speech (with a variety of voices to choose from)
• Image Generation using a DALL·E model (note that as the saying goes an image is worth a thousand tokens...)
• Model selection and parameters (including the chatgpt model which is set as default: gpt-3.5-turbo)
• Displays number of remaining tokens for current conversation context before it is reset (Due to max_tokens limited by OpenAI's API)
• Code blocks formatted with Syntax highlighting (experimental)

Known Issues

• Tab completion doesn't seem to work well in the readline module that comes with native python. The Conda environment includes a python binary with a readline module that does support it.