openai-request-tracker 0.1.0

OpenAI cost tracker that logs OpenAI requests using opentelemetry

openai-cost-tracker

Cost tracker that logs OpenAI requests using opentelemetry. Once enabled, it will log every request that is being made via the openai python client to a file in the execution directory. The log file is called "traces_<\datetime>.log".

Installation

Install with poetry:

poetry install

or install with pip:

pip install .

Usage

Record requests

CLI wrapper

The command track-costs is a wrapper that initializes the tracker and then runs the given CLI command.

track-costs <your usual command>

For example, if your are running usually an indexing run with microsoft/graphRAG with

python -m graphrag.index --root foo

you only need to change it to

track-costs graphrag.index --root foo

to record all openai requests in a log file.

Similarly, for calling the graphrag query module you can do

track-costs graphrag.query --root foo --method local "My query"

In code usage

You can just add a call to cost_tracker.init_tracker() at the very beginning of your script:

from cost_tracker import init_tracker

init_tracker()

... # your script

For example, to make it work for the indexing phase of microsoft/graphRAG, modify the file main.py such that it looks like:

# Copyright (c) 2024 Microsoft Corporation.
# Licensed under the MIT License

"""The Indexing Engine package root."""
from cost_tracker import init_tracker

init_tracker()

import argparse

from graphrag.logging import ReporterType
from graphrag.utils.cli import dir_exist, file_exist

from .cli import index_cli
from .emit.types import TableEmitterType


if __name__ == "__main__":
    parser = argparse.ArgumentParser(
        prog="python -m graphrag.index",
        description="The graphrag indexing engine",
    )

... # rest of the unmodified script

Run your script as before.

Show costs

The script "display-costs" can be used to show the openai costs for all requests recorded in a given log file:

display-costs --file <YOUR_LOG_FILE>