LLM Tools

This package wraps the OpenAI API and other LLM models that share the completion endpoint feature. It provides a simple interface to generate text from a prompt.

Installation

pip install llm-tools

Usage

from llm_tools.openai import prompt
from pathlib import Path


# Create a prompt object that will be logged to the `my-task` directory under the
# current working directory.
# The PromptZeros class will use arguments that will maximize the determinism of the outputs.
pz = prompt.PromptZeros(Path("./"), "my-task")

# Send a prompt to the API and return the response.
w = pz.send_prompt("What is the WDI code for GDP?", return_data=True)

# Print the response.
print(w)

# Output:

# {'message': [{'role': 'system', 'content': ''}, {'role': 'user', 'content': 'What is the WDI code for GDP?'}], 'response': {'id': 'chatcmpl-8V0IN0VbeYTrsBAohs309DHJQLJL8', 'choices': [{'finish_reason': 'stop', 'index': 0, 'message': {'content': 'The WDI code for GDP (Gross Domestic Product) is "NY.GDP.MKTP.CD".', 'role': 'assistant', 'function_call': None, 'tool_calls': None}}], 'created': 1702399995, 'model': 'gpt-3.5-turbo-0613', 'object': 'chat.completion', 'system_fingerprint': None, 'usage': {'completion_tokens': 23, 'prompt_tokens': 20, 'total_tokens': 43}}, 'content': 'The WDI code for GDP (Gross Domestic Product) is "NY.GDP.MKTP.CD".', 'api_kwargs': {'temperature': 0, 'top_p': 0, 'n': 1, 'stream': False, 'frequency_penalty': 0, 'presence_penalty': 0, 'model': 'gpt-3.5-turbo', 'max_tokens': 4029}, 'metadata': {}, 'message_hash': '1a5816095c232f29428d4cec70498391'}

NOTE: The output is also logged to the my-task directory under the current working directory.