jameswylde / openai-chatgpt-terminal

Python implementation of OpenAI's ChatGPT integrated into your shell, interactively or passing questions to it from the terminal - with support for custom prompts for honed responses, switching of API model and changing of model temperature.

ChatGPT in your terminal

OpenAI's ChatGPT integrated into your shell.

Description

Python implementation of OpenAI's ChatGPT intergrated into your shell - interaction can be either calling genie from shell for one off questions, or entering into an interactive chat with genie. There are some chat prompts to use for more honed responses - these can be added or amended in prompts.py

Dependencies

• Install module dependencies using pip: pip install -r requirements.txt

• OpenAI API key - you can get one here - Dashboard - Settings - View API Keys - Generate

Installing

• Clone the repo and copy the folder to a permanent location (including the /src/ folder).

• Open the script and amend openai.api_key = "API_KEY" with your aforementioned API key and save.

• Create an alias pointing at the script's location, either in your bash profile ~/.bash_profile or ~/.bashrc or ~/.zshrc - i.e: alias genie='python3 /path/to/genie.py'

Usage

Using your chosen alias you can call it from shell and pass your question as an argument for one off questions:

$ genie "Recommend a python module for LP and link documentation?"

Or by calling the alias without argument to enter an interactive chat with ChatGPT:

$ genie

The default model used is gpt-3.5-turbo for a more fluent experience as its replies are much faster and API pricing is significantly cheaper.

You can switch API model using the --model argument (run without --model to use the default model) if you'd like to to use gpt-4:

$ genie --model gpt-4

--model will also accept code-davinci-002 & text-davinci-003 - other API models can be seen here OpenAI ChatGPT API Models and added as required to the script.

The temperature used is 0.7 - which appears to be a good balance between creativity and focused responses.

You can switch temperature using the --temp argument (run without --temp to use the default value).

To end the interactive chat, use either bye,quit,q or ctrl+c.

openai-chatgpt-terminal's Issues

You got forked!

Hi James,

Sorry to abuse the issues system like this, but I figured it was less "pushy" than a PR. I recently forked your project as I was looking for an openAI client at the command line that had some form of baked in prompts to choose from. What I've ended up with is a bit more complicated than Genie, but it serves as a jumping off point for some context management experiments I want to do.

Just wanted to say thanks! At this point there isn't much of Genie left in my project, but it was a helpful place to start.

Take a look if you're interested, https://github.com/grobertson/openai-terminal-client

Cheers,
-Grant