AgentScope 멀티 에이전트 파이프라인 — MsgHub + FanoutPipeline

패턴 1: SequentialPipeline (직렬)
A → B → C → 결과

패턴 2: FanoutPipeline (병렬)
    ┌→ B ─┐
A → ├→ C ─┤ → 결과 취합
    └→ D ─┘

패턴 3: MsgHub (그룹 대화)
A ←→ B ←→ C (모두가 모든 메시지를 봄)

from agentscope.agent import ReActAgent
from agentscope.model import OpenAIChatModel
from agentscope.memory import InMemoryMemory
from agentscope.formatter import OpenAIChatFormatter
from agentscope.pipeline import SequentialPipeline
import os, asyncio

model = OpenAIChatModel(
    model_name="gpt-4o",
    api_key=os.environ["OPENAI_API_KEY"],
    stream=True,
)

# 리서처 에이전트
researcher = ReActAgent(
    name="Researcher",
    sys_prompt="You are a research specialist. Given a topic, produce detailed research notes with key facts, statistics, and insights.",
    model=model,
    memory=InMemoryMemory(),
    formatter=OpenAIChatFormatter(),
)

# 분석가 에이전트
analyst = ReActAgent(
    name="Analyst",
    sys_prompt="You are a data analyst. Take research notes and extract actionable insights, patterns, and recommendations.",
    model=model,
    memory=InMemoryMemory(),
    formatter=OpenAIChatFormatter(),
)

# 라이터 에이전트
writer = ReActAgent(
    name="Writer",
    sys_prompt="You are a technical writer. Take analysis results and write a clear, well-structured report for a non-technical audience.",
    model=model,
    memory=InMemoryMemory(),
    formatter=OpenAIChatFormatter(),
)

# 파이프라인 구성
pipeline = SequentialPipeline([researcher, analyst, writer])