OpenAI Provider

Drop-in replacement for openai.OpenAI. Wraps chat.completions.create() with automatic cost capture.

Usage

# Before
from openai import OpenAI

# After
from kostrack import OpenAI

client = OpenAI(
    tags={"project": "openmanagr", "feature": "gl-classification"}
)

response = client.chat.completions.create(
    model="gpt-4o",
    max_tokens=512,
    messages=[{"role": "user", "content": "Classify this transaction..."}],
)

Streaming

Kostrack automatically injects stream_options={"include_usage": True} into streaming requests so token counts are captured from the final chunk:

for chunk in client.chat.completions.create(
    model="gpt-4o", stream=True, messages=[...]
):
    print(chunk.choices[0].delta.content, end="")

Supported models

Token breakdown