LLM Wrapper

This is an unofficial wrapper for LLM APIs.

• OpenAI models
• Google Palm
• Cohere Commands

Setup

pip install git+https://github.com/ryokamoi/llm_wrapper
pip install --upgrade chardet  # for llama2

Example

from llm_wrapper import llm_api
print(llm_api("gpt-3.5-turbo", prompt="Translate this sentence into Japanese: I am GPT-3.5."))
# {'prompt': 'Translate this sentence into Japanese: I am GPT-3.5.', 'response': '私はGPT-3.5です。 (Watashi wa GPT-3.5 desu.)'}

# When temperature=0. (default), this wrapper will store and read cached results.
llm_api("gpt-3.5-turbo", prompt="Translate this sentence into Japanese: I am GPT-3.5.")
# read cache from llm_cache/554a571588d632d02bd8cc330fee66fea40adf7d242763b462a7691ee858afd4d3a852caaa02058db5e9810edac0fee5021754c0f4479842d374e915a7cb21c0.json
# {'prompt': 'Translate this sentence into Japanese: I am GPT-3.5.', 'response': '私はGPT-3.5です。 (Watashi wa GPT-3.5 desu.)'}

• You can overwrite the cache by setting cache_overwrite=True.
• You can overwrite the temperature by setting updated_parameters={"temperature": 0.5}. You can update any parameters in the same way.