AutoFNS is a wrapper around OpenAI's Chat Completion API.
It allows you to define functions as you normally would, and have them be called automatically when a message containing tool calls is detected. This allows you to focus on more important things, like the logic of your application.
Note: 🧪 This is a work in progress and may not be ready for production use.
- python >= 3.10
- openai
pip install autofns
from autofns import AutoFNS
FUNCTION_DEFINITIONS = [
{
"type": "function",
"function": {
"name": "get_temp_units",
"description": "Get a list of temperature units",
}
},
{
"type": "function",
"function": {
"name": "get_current_temperature",
"description": "Get the current temperature in a city",
},
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string",
"description": "The city to get the temperature of",
},
"unit": {
"type": "string",
"description": "The unit to get the temperature in",
"enum": ["Fahrenheit", "Celsius", "Kelvin"],
}
},
"required": ["city", "unit"],
}
},
]
def get_temp_units():
return ["Fahrenheit", "Celsius", "Kelvin"]
def get_current_temperature(city: str, unit: str):
return "The current temperature in {} is {} degrees {}".format(
city, 72, unit
)
fns = AutoFNS(
"gpt-4-32k",
fns_mapping={
"get_temp_units": get_temp_units,
"get_current_temperature": get_current_temperature
},
fns_definitions=FUNCTION_DEFINITIONS,
)
result = fns.create_completion(messages=[...])
You can also use the AutoFNSAsync
class to use async functions:
from autofns import AutoFNSAsync
fns = AutoFNSAsync(...)
result = await fns.create_completion(messages=[...])
This project is licensed under the terms of the MIT license.
Contributions are welcome! Please open an issue or a pull request.