https://github.com/pycui/universal-api-cohere

https://github.com/pycui/universal-api-cohere

Last synced: 1 day ago
JSON representation

• Host: GitHub
• URL: https://github.com/pycui/universal-api-cohere
• Owner: pycui
• License: mit
• Created: 2024-05-30T22:31:31.000Z (6 months ago)
• Default Branch: main
• Last Pushed: 2024-05-30T22:41:32.000Z (6 months ago)
• Last Synced: 2024-05-31T00:42:35.106Z (6 months ago)
• Language: Python
• Size: 8.79 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Universal API

*Don't know how to implement a function? Let Universal API help you on the fly!*

This is a non-serious library that can implement any functions on the fly using LLMs. To use it, simply call a function with a descriptive name and parameters, and it will be defined, implemented and called at runtime. Generated functions are cached so you don't pay for things that are already implemented.

(Note: this repo is based on my repo at https://github.com/pycui/universal-api and adopted to Cohere API.)

## Usage

```python

api = UniversalAPI()

# Start to call arbitrary functions 

print(api.sort([3, 2, 1])) # returns [1, 2, 3]

print(api.sort([4, 3, 2, 1])) # returns [1, 2, 3, 4] using cached implementation

print(api.sort([1, 2, 3], reverse=True)) # returns [3, 2, 1]

print(api.add(1, 2)) # returns 3

print(api.reverse('hello')) # returns 'olleh'

api.fizzbuzz(15) # prints the fizzbuzz sequence up to 15

api.print_picachu() # prints an ASCII art of Picachu

```

## Warning

This library will execute unverified code in your local machine. It's **NOT** safe to run in production (or really, any serious environment).

## Notes

By default this library uses OpenAI GPT API. You can modify environment variables like `OPENAI_BASE_URL` to use other LLM endpoints (e.g. Anyscale Endpoint), which allows you to run on other models like the LLaMa series. It's also possible to use local LLMs.