How do you implement multi-agent systems using the OpenAI Agents SDK?

from agents import Agent, Runner, handoff
import asyncio

# Define specialised agents:
code_agent = Agent(
    name="CodeWriter",
    instructions="Write clean, well-documented Python code. Focus on correctness.",
    model="gpt-5.5",
)

review_agent = Agent(
    name="CodeReviewer",
    instructions="Review code for bugs, security issues, and performance problems.",
    model="gpt-5.5",
)

doc_agent = Agent(
    name="DocWriter",
    instructions="Write clear docstrings and README documentation.",
    model="gpt-5.4-mini",  # cheaper model for docs
)

# Orchestrator agent with handoffs:
orchestrator = Agent(
    name="Orchestrator",
    instructions="""Manage the coding workflow:
    1. Use CodeWriter to implement features
    2. Use CodeReviewer to review and fix issues
    3. Use DocWriter to document the final code
    """,
    handoffs=[
        handoff(code_agent),
        handoff(review_agent),
        handoff(doc_agent),
    ],
    model="gpt-5.5",
)

async def run_pipeline(task: str):
    result = await Runner.run(
        orchestrator,
        task,
        max_turns=20,  # prevent infinite loops
    )
    return result.final_output

# Run:
output = asyncio.run(run_pipeline(
    "Implement a thread-safe LRU cache class with comprehensive tests"
))
print(output)