Sends a prompt or series of prompts to the OpenAI discussions API and outputs the response as it arrives.

gpt -s 'You are a helpful assistant' \
    -p 'Rewrite the second act of Much Ado About Nothing in modern English'

Provides a command line interface to have conversational interactions similar to ChatGPT from the command line, with a searchable conversation history. Simply execute chat and follow the menu prompts to start a conversation.

Generates a command that can be directly executed in your $SHELL.

cmd 'list files in the current directory with time stamps'

If you want commands you execute with cmd to be added to your shell's history, add something like this to your bashrc:

alias cmd="source $(which cmd)"

A little wrapper around gpt focused on producing code without texty explanations.

stack_class=$(code -l python -p 'stack class')
unit_test=$(code -l python -c "$stack_class" -p 'unit test for this class')

Runs your unit tests. If they fail, sends the error messages and relevant file contents to gpt to come up with suggested fixes. Then it will offer to rerun your tests to verify the fix.

Library of small utility functions and aliases:

• image <prompt> [<size>] > image.png - generate an image
• re-image <source-file> <prompt> > new-image.png - modify an image

There are a couple of scripts that work with the new beta APIs.

The analog to the gpt utility above. It provides sub-commands to interact directly with the API.

A much more advanced version of chat that works with the new API. The only thing is lacking is the streaming protocol, so you get to wait on the run to complete before you get your answers.

On the other hand, you also get a lot of new features:

• \f $filepath in your message to send the contents of a file
• \exec cmd --arg2 --arg2 to send the output of a command
• \begin ... \end to send the output of a multi-line command
• \www https://example.com to send the contents of a web page

If the output of any of these is too large for a single message, it will be split into multiple numbered messages.

• curl
• jq
• gum

1. Ensure that /usr/local/bin is in your PATH
2. Ensure that you have write permissions to /usr/local/bin
3. Download to /usr/local/bin

for tool in gpt chat code cmd tester; do
  curl -sSL "https://raw.githubusercontent.com/sysread/bash-gpt/main/$tool" -o "/usr/local/bin/$tool" \
    && chmod +x "/usr/local/bin/$tool"
done

1. Check out repository:

git clone https://github.com/sysread/bash-gpt

2. Add to your PATH

export PATH="$PATH:/path/to/bash-gpt"

Want an easy way to install nix? Checkout the Determinite Systems Installer

1. Check out repository:

git clone https://github.com/sysread/bash-gpt

2. Add Env Vars to your shell environment

• create a .env file with the following keys:

OPENAI_API_KEY=""
OPENAI_API_MODEL="gpt-3.5-turbo-16k"

• OPENAI_API_KEY - your OpenAI API key (required; create one at https://beta.openai.com/account/api-keys)
• OPENAI_MODEL - the model to use; defaults to gpt-3.5-turbo-16k

3. Run the nix shell

nix develop

To use the utilities in utils, first download it:

curl -sSL https://raw.githubusercontent.com/sysread/bash-gpt/main/utils -o /usr/local/lib/bash-gpt-utils

Then add it to your .bashrc or .bash_profile:

echo 'source /usr/local/lib/bash-gpt-utils' >> ~/.bashrc

• OPENAI_API_KEY - your OpenAI API key (required; create one at https://beta.openai.com/account/api-keys)
• OPENAI_MODEL - the model to use; defaults to gpt-3.5-turbo-16k

Tested on macOS. Ought to work on linux as well. Please file a ticket if you find an issue on mac or linux. Better yet, send me a pull request :)