openai_usage_logger

This library provides a simple way to log OpenAI API usage and costs. It maintains a record of the number of prompt and completion tokens used for different OpenAI models, as well as the associated costs.

Install

pip install openai-usage-logger

Features

• Tracks usage and costs for different tags and models.
• Allows logging of token usage and costs in a structured format.
• Provides flexibility to use a custom logger.

Usage

Basic Usage

Use the library to log responses from OpenAI's completion API, including the model used and the associated costs.

import openai
from openai_usage_logger import usageLogger

# Example OpenAI completion API call
response = openai.ChatCompletion.create(
    model='gpt-3.5-turbo-0125',
    messages=messages,
)
usageLogger().log(response)

Logging the response with a usage tag

usageLogger().log(response, tag='usage_tag')

Advanced Usage with Custom Logger

You can also pass a custom logger to the log method if you want to use a specific logging configuration.

import logging
from openai_usage_logger import usageLogger

Set up your custom logger

custom_logger = logging.getLogger('my_custom_logger')
logging.basicConfig(level=logging.INFO)

Using the custom logger

usageLogger().log(response, tag='usage_tag', logger=custom_logger)

Model Costs The package currently supports cost calculation for the following models:

GPT-4 models (including various versions) GPT-3.5-Turbo models (including various versions) Costs are based on the number of tokens used for prompts and completions, with different rates for different models.