OpenAI Agents SDK - RedPill AI

The OpenAI Agents SDK accepts a custom AsyncOpenAI client. Select its Chat Completions model explicitly because the SDK uses the Responses API by default.

Install

pip install openai-agents
export REDPILL_AI_API_KEY="YOUR_API_KEY"

Build an agent

import os

from agents import (
    Agent,
    OpenAIChatCompletionsModel,
    Runner,
    function_tool,
    set_tracing_disabled,
)
from openai import AsyncOpenAI


set_tracing_disabled(True)

client = AsyncOpenAI(
    base_url="https://api.redpill.ai/v1",
    api_key=os.environ["REDPILL_AI_API_KEY"],
)
model = OpenAIChatCompletionsModel(
    model="z-ai/glm-5.2",
    openai_client=client,
)


@function_tool
def deployment_status(service: str) -> str:
    """Return a sample deployment status."""
    return f"{service}: healthy"


agent = Agent(
    name="Operations assistant",
    instructions="Use the tool to check deployment status.",
    model=model,
    tools=[deployment_status],
)

result = Runner.run_sync(
    agent,
    "Check the gateway deployment status.",
    max_turns=3,
)
print(result.final_output)

Tracing is disabled because the default trace exporter requires an OpenAI platform key. Configure a different trace processor when your application needs tracing.

Production checks

Keep max_turns bounded, add network timeouts, and validate tool input. For confidential-only calls, change the client base URL to https://tee.redpill.ai/v1 and select an is_tee: true model.

Troubleshooting

Symptom	Check
SDK calls the wrong API shape	Wrap the client with `OpenAIChatCompletionsModel`.
Default tracing reports errors	Disable tracing or configure a compatible trace processor.
Agent runs too many turns	Lower `max_turns` and add request limits around the runner.

References