Closed
Description
Please read this first
- Have you read the custom model provider docs, including the 'Common issues' section? ✅ Yes
- Have you searched for related issues? ✅ Yes
Describe the question
I'm using the OpenAI Agents SDK with a custom model provider (Google Gemini via https://generativelanguage.googleapis.com/v1beta/openai/) and encountering a BadRequestError.
The Gemini API rejects the request with this error:
BadRequestError: Error code: 400 - [{'error': {'code': 400, 'message': 'Invalid JSON payload received. Unknown name "metadata": Cannot find field.\nInvalid JSON payload received. Unknown name "store": Cannot find field.'}}]
It appears that the Agents SDK is sending metadata and store fields automatically, but the Gemini API doesn't support these.
Debug information
- Agents SDK version:
v0.0.3 - Python version:
3.11 - Custom model provider:
Google Gemini (Flash 1.5)/(Flash 2.0) - Request Type:
chat.completions.create
Repro steps
Here’s a minimal script to reproduce the issue:
from openai import AsyncOpenAI
from agents import Agent, Runner, OpenAIChatCompletionsModel
from agents.run import RunConfig
import nest_asyncio
nest_asyncio.apply()
# External Async client for custom provider
client = AsyncOpenAI(
api_key=google_api_key,
base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
)
# Wrap client in SDK model
model = OpenAIChatCompletionsModel(
model="gemini-1.5-flash", # "gemini-2.0-flash"
openai_client=client,
)
config = RunConfig(
model=model,
model_provider=client,
tracing_disabled=True
)
agent = Agent(name="Assistant", instructions="You are a helpful assistant")
# Run agent synchronously
result = Runner.run_sync(agent, "Write a haiku about recursion in programming.")
print(result.final_output)
**Metadata**
Please clarify:
Is there a way to prevent the SDK from sending unsupported fields like metadata or store?
Will future versions of the SDK provide better integration options for non-OpenAI endpoints?