OpenAI's ChatGPT integrated into your shell.
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
-
Install module dependencies using pip:
pip install -r requirements.txt -
OpenAI API key - you can get one here - Dashboard - Settings - View API Keys - Generate
-
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'
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.