别为了 Agent 而 Agent：2026 年多 Agent 工作流实战指南

1
2
3
4

python -m venv langgraph-env
source langgraph-env/bin/activate
pip install langgraph langchain-openai langchain
export OPENAI_API_KEY="your-key"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55

"""LangGraph 最小示例：Researcher → Writer 协作"""
import os
from typing import TypedDict, Annotated
import operator
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, START, END
from langgraph.checkpoint.memory import MemorySaver
from langchain_core.messages import HumanMessage, AIMessage

os.environ["OPENAI_API_KEY"] = "your-key"
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)

# 定义共享状态
class AgentState(TypedDict):
    messages: Annotated[list, operator.add]
    research_result: Annotated[str, operator.add]
    next: str

# 定义节点
def researcher_node(state: AgentState) -> dict:
    response = llm.invoke([
        HumanMessage(content=f"请针对以下问题提供详细研究结果:\n{state['messages'][-1].content}")
    ])
    return {
        "messages": [AIMessage(content=f"[Researcher] {response.content}")],
        "research_result": response.content,
        "next": "writer"
    }

def writer_node(state: AgentState) -> dict:
    response = llm.invoke([
        HumanMessage(content=f"基于以下研究结果生成最终报告:\n{state['research_result']}")
    ])
    return {
        "messages": [AIMessage(content=f"[Writer] {response.content}")],
        "next": "END"
    }

# 构建图
workflow = StateGraph(AgentState)
workflow.add_node("researcher", researcher_node)
workflow.add_node("writer", writer_node)
workflow.add_edge(START, "researcher")
workflow.add_edge("researcher", "writer")
workflow.add_edge("writer", END)

# 编译并执行
memory = MemorySaver()
app = workflow.compile(checkpointer=memory)

result = app.invoke(
    {"messages": [HumanMessage(content="分析 Python 在 AI 领域的主导地位")],
     "research_result": "", "next": ""},
    config={"configurable": {"thread_id": "demo-1"}}
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

def router_node(state: AgentState) -> dict:
    last_msg = state["messages"][-1].content.lower()
    if any(kw in last_msg for kw in ["code", "编程", "代码"]):
        return {"next": "coder"}
    else:
        return {"next": "generalist"}

router_graph.add_conditional_edges(
    "router",
    lambda state: state["next"],
    {"coder": "coder", "generalist": "generalist"}
)

1
2
3
4

python -m venv crewai-env
source crewai-env/bin/activate
pip install crewai crewai-tools
export OPENAI_API_KEY="your-key"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

"""CrewAI 最小示例：Researcher + Writer 协作"""
import os
os.environ["OPENAI_API_KEY"] = "your-key"

from crewai import Agent, Task, Crew, Process

researcher = Agent(
    role="高级研究员",
    goal="收集和分析相关信息",
    backstory="你是一个经验丰富的研究员，擅长信息搜集和分析。",
    verbose=True
)

writer = Agent(
    role="技术写作者",
    goal="将研究结果转化为清晰的报告",
    backstory="你是一个技术写作者，擅长将复杂信息简化。",
    verbose=True
)

research_task = Task(
    description="研究 Python 在 AI 领域的主导地位及其原因",
    agent=researcher,
    expected_output="结构化的研究分析"
)

write_task = Task(
    description="基于研究结果撰写一份简洁报告",
    agent=writer,
    expected_output="最终报告"
)

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential  # 顺序执行
)

result = crew.kickoff()
print(result)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

manager = Agent(
    role="项目经理",
    goal="协调团队完成项目",
    backstory="你是一个项目经理，擅长任务分配和协调。",
    verbose=True
)

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.hierarchical,
    manager_agent=manager
)

1
2
3
4

python -m venv autogen-env
source autogen-env/bin/activate
pip install autogen-agentchat openai
export OPENAI_API_KEY="your-key"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

"""AutoGen 最小示例：Coder + UserProxy 对话"""
import os
import autogen

os.environ["OPENAI_API_KEY"] = "your-key"

llm_config = {
    "config_list": [{"model": "gpt-4o", "api_key": os.environ.get("OPENAI_API_KEY")}],
    "temperature": 0.7,
}

coder = autogen.AssistantAgent(
    name="coder",
    llm_config=llm_config,
    system_message="你是一个 Python 编程专家。请编写可运行的代码。",
)

user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",  # 完全自动化
    code_execution_config={
        "last_n_messages": 3,
        "work_dir": "coding",
        "use_docker": False,
    },
)

user_proxy.initiate_chat(
    coder,
    message="请写一个 Python 函数，计算斐波那契数列的前 10 项。",
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

pm = autogen.AssistantAgent(name="pm", llm_config=llm_config,
    system_message="你是一个产品经理。请描述 Web 应用的功能需求。")

developer = autogen.AssistantAgent(name="developer", llm_config=llm_config,
    system_message="你是一个全栈开发者。根据需求编写技术实现方案。")

user_proxy = autogen.UserProxyAgent(name="user_proxy",
    human_input_mode="NEVER", max_consecutive_auto_reply=1)

groupchat = autogen.GroupChat(
    agents=[pm, developer, user_proxy],
    messages=[], max_round=6,
    speaker_selection_method="round_robin",
)

manager = autogen.GroupChatManager(groupchat=groupchat, llm_config=llm_config)
user_proxy.initiate_chat(manager, message="请设计并实现一个待办事项清单 Web 应用。")

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# 确保 Docker 已安装并运行
docker --version

# 克隆并启动
git clone https://github.com/All-Hands-AI/OpenHands.git
cd OpenHands
echo "LLM_API_KEY=your-key" >> .env
echo "LLM_MODEL=gpt-4o" >> .env
docker compose up -d

# 访问 http://localhost:3000

1
2
3
4
5

pip install openhands
export LLM_API_KEY="your-key"
export LLM_MODEL="gpt-4o"
openhands
# 输入任务描述即可