Raw openai SDK
If you use the openai package directly (no LangChain / LlamaIndex /
CrewAI on top), install the OpenAI extra:
raw_openai_client.py
import nullrun
from openai import OpenAI
nullrun.init(api_key="nr_live_...")
client = OpenAI()
resp = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": "Hello"}],
)
Patched via the httpx transport hook in nullrun.instrumentation.auto
— the OpenAI SDK uses httpx under the hood, so the SDK reads the
response body, extracts usage.prompt_tokens /
usage.completion_tokens, and emits a track_llm event.
nullrun[openai]is for the rawopenaiSDK — it is not the OpenAI Agents SDK. For agents usenullrun[agents].