CallingGPT
GPT's Function Calling Demo, a experiment of self-hosted ChatGPT-Plugins-like platform.
Recommend reading: function-calling
Abstract
OpenAI's GPT models provide a function calling feature, so we can easily create ChatGPT-Plugins-like
tools. This repository is a proof-of-concept of the function calling feature.
In this experiment, we defined the Plugin
as Namespace
which contains a serial of functions. While user performing a conversation, the functions in Namespace
will be called by the API and return the result to the user.
Usage
-
Clone this repository and install the dependencies.
git clone https://github.com/RockChinQ/CallingGPT cd CallingGPT pip install -r requirements.txt
-
Run the
main.py
to generateconfig.yaml
python main.py
-
Edit the
config.yaml
to set your API key and other settings. -
Run the
main.py
and pass your modules.python main.py <module0> <module1> ...
Example
Use the example/greet.py
, provides a greet
function called when user ask GPT to greet someone.
python main.py example/greet.py
Then you can talk to the bot.
$ python main.py examples/greet.py
Using module: examples.greet
>>> Hello and who are you?
<<< Hello! I am an AI assistant. How can I assist you today?
>>> say hello to Rock
call<examples-greet-greet>: {
"user": "Rock"
}
<<< Hello, Rock! How can I assist you today?
>>> and to Alice
call<examples-greet-greet>: {
"user": "Alice"
}
<<< Hello, Alice! How can I assist you today?
>>>
Type help
to get help.
See wiki for the function format.
Other Examples
examples/draw_and_wrapper_md.py
Provides a dalle_draw
function to use DALL·E model when user ask GPT to draw something.
python main.py examples/draw_and_wrapper_md.py
$ python main.py examples/draw_and_wrapper_md.py
Using module: examples.draw_and_wrapper_md
>>> hello!
<<< Hi there! How can I assist you today?
>>> draw a sunset for me please
call<examples-draw_and_wrapper_md-draw>: {
"prompt": "sunset"
}
<<< Sure! Here's a beautiful sunset for you:
![Sunset](https://oaidalleapiprodscus.blob.core.windows.net/private/org-VS9HEpJba78GXVfOcmVo7qaM/user-OHa7Jo3kL4XJDg9lo7AzdWNT/img-QmDUiwp1IGFcu8pDGZh0i7r8.png)
I hope you like it! Let me know if there's anything else I can help you with.
>>>
For Code
-
Install the package
pip install --upgrade CallingGPT
-
Create your own functions in modules(these modules can also be used in the CLI mode)
# your_module_a.py def func_a(prompt: str) -> str: # Type hint of EACH argument and return value is REQUIRED. """ The description of this func a, will be provided to the api. Args: prompt(str): The prompt of the function. Returns: The result of the function. """ # Google style docstring is REQUIRED, it will be split into # `description` and `params`(required if there are args) and # `returns`(optional), `\n\n` between each part. return "func_a: " + prompt
# your_module_b.py def adder(a: int, b: int) -> int: """ Add two numbers. Args: a: The first number. b: The second number. Returns: The sum of a and b. """ # Type hints of args in docstring is optional. return a + b
-
Call the wrapper
from CallingGPT.session.session import Session import your_module_a, your_module_b import openai openai.api_key = 'your_openai_api_key' session = Session([your_module_a, your_module_b]) for reply in session.ask("your prompt"): # session.ask will yield each time the api returns a result, # before calling function, it will print the function name and args. # e.g. here's a function call: # { # "role": "assistant", # "content": null, # "function_call": { # "name": "examples-draw_and_wrapper_md-draw", # "arguments": "{\n \"prompt\": \"cat\"\n}" # } # } # # while here's a normal reply: # { # "role": "assistant", # "content": "Hello, I am an AI assistant. How can I assist you today?" # } print(reply)
Session
will automatically manage context for you.
See wiki for the function format.