OpenAI SDK Adapter

Wraps client.chat.completions.create() with LOCO scheduling and cost tracking.

Setup

import openai
from loco import AsyncLOCOScheduler, SharedResource
from loco.adapters.openai import OpenAIAdapter

scheduler = AsyncLOCOScheduler([], SharedResource("llm_api", capacity=3))
adapter = OpenAIAdapter(scheduler, client=openai.AsyncOpenAI())

Usage

response = await adapter.create_chat(
    "assistant",
    model="gpt-4o",
    messages=[{"role": "user", "content": "Summarize this report..."}],
)

Model Weights

Model	Weight
gpt-4o	3.0
gpt-4o-mini	1.0
o3	5.0
o3-mini	2.0
Unknown	2.0 (default)