xontrib-chatgpt 0.1.2

Gives the ability to use ChatGPT directly from the command line

Gives the ability to use ChatGPT directly from the command line

If you like the idea click ⭐ on the repo and tweet.

Installation

To install use pip:

xpip install xontrib-chatgpt
# or: xpip install -U git+https://github.com/jpal91/xontrib-chatgpt

Then use the following command in your xonsh shell or add it directly to your .xonshrc

xontrib load chatgpt

Environment Variables

❗REQUIRED ❗

# Recommended you put this in your .xonshrc
$OPENAI_API_KEY = <your api key>

If you do not have an api key, you can create an account on OpenAI. You will need to add a credit card on file to pay for any of the api usage charges.

You are 100% responsible for any and all cost you incur by using this xontrib.

For more information on OpenAI usage pricing, please go to OpenAI Pricing.

Optional

$OPENAI_CHAT_MODEL = 'gpt-3.5-turbo'

If this is not set, it will default to gpt-3.5-turbo. Currently the only accepted options are gpt-3.5-turbo or gpt-4.

Usage

xontrib-chatgpt was designed to be versatile. To that end, it has capabilities of a python function, a bash-like function, and a xonsh context block.

Basic

When loaded, an alias chatgpt will automatically be added into your shell. You can use this for simple one-off questions -

chatgpt Can you write me a simple 'hello world' function in python?
# ChatGPT responds here

Context Manager

To carry on a conversation with ChatGPT, first assign the ChatGPT class to a variable

gpt = ChatGPT()

This package extends the built-in Block context manager capabilities from xonsh, allowing for multi-line inputs to be sent to the ChatGPT api

with! gpt:
   My python function below is broken, can you help me fix it?

   def hello_world():
      return None
      print('Hello world!')

# ChatGPT responds here
# Any other inputs added will be included in the messages sent to ChatGPT

with! gpt:
   ...

With Alias

To unlock the full range of conversation options, you can optionally assign an alias to the instance -

# Best practice is to use the same name as your variable assignment
gpt = ChatGPT(alias='gpt')

# The following is automatically ran for you...
# aliases['gpt'] = lambda args, stdin=None: gpt(args, stdin)

# Then use as a context manager...
with! gpt:
   Hello! Tell me more about yourself.

# Or as an alias function...
gpt Hello! Tell me more about yourself.
echo 'Hello! Tell me more about yourself.' | gpt

# Go crazy with it!
cat my_input.txt | gpt
my_next_input = $(echo 'Hello! Tell me more about yourself')
echo @(my_next_input) | gpt

When your variable/conversation is deleted, the registered alias is automatically deleted as well

del gpt
Deleting alias gpt

Get Help

For class/instance related help -

ChatGPT?
gpt = ChatGPT()
gpt?

For cli related help

chatgpt -h
gpt = ChatGPT()
gpt -h

Additional Capabilities

Printing Convo

To print out your conversation to the shell, use one of the following -

gpt = ChatGPT('gpt')

gpt.print_convo()
# or
gpt -p

See gpt -h or gpt.print_convo? for additional help and options

Saving Convo

To save your current conversation -

gpt = ChatGPT('gpt')

gpt.save_convo(path='path/to/my_convo.txt')
# or
gpt -s -P 'path/to/my_convo.txt'

See gpt -h or gpt.save_convo? for additional help and options

Loading Convo

You can load a past conversation to a variable using the class method ChatGPT.fromconvo

gpt = ChatGPT.fromconvo(path='path/to/my_convo.txt', alias='gpt')

Future Plans

• Chat Manager
  • Currently, the package has quite a few options for interaction, but some overlap and aren't very concise. The plan is to create a 'one-stop-shop' for interacting with this package in the form of a 'manager' cli function.
  • Ideally from this, you would be able to:
    • Print/load/save conversations
    • Get consolidated 'help' functions (ie get the same info when calling '-h' or '?')
    • Create instances/aliases
    • Monitor and get stats for all active conversations
  • The idea around this being there's one manager who controls everything and you don't have to go searching around for a specific convo, help, instance, alias, etc.
• Streaming Responses
  • Expand the ability to get streaming responses on the command line as opposed to waiting until the full completion is done

Credits

This package was created with xontrib template.