说起来，

Node 基础笔记

在 LangGraph 中。Node 是执行逻辑的基本单元。一个节点通常是一个函数，它接收当前状态并返回更新后的状态。下面给出一个完整示例，并结合实际开发痛点说明如何配置重试策略、缓存策略还有参数绑定。

from functools import partial
from typing import TypedDict
from langgraph.graph import StateGraph,START,END
from langgraph.types import RetryPolicy,CachePolicy
from requests.exceptions import RequestException,Timeout
# 1️⃣ 定义状态结构
class GraphState:
process_data: dict # 存放业务数据
# 2️⃣ 节点函数
def input_node -> GraphState:
print
return {"process_data": {"input": "input_value"}}
def process_node -> dict:
print
print
print
return {"process_data": {"process": "process_value"}}
# 3️⃣ 重试与缓存策略
retry_policy = RetryPolicy(
max_attempts=3,initial_interval=1.0,jitter=True,backoff_factor=2.0。retry_on=
)
cache_policy = CachePolicy # 缓存 5 分钟
# 4️⃣ 建立图形
state_graph = StateGraph
# 添加节点
state_graph.add_node
# 为 process_node 绑定参数并添加策略
process_with_params = partial
state_graph.add_node(
node="process",action=process_with_params,retry_policy=retry_policy,cache_policy=cache_policy
)
# 定义边
state_graph.add_edge
state_graph.add_edge
state_graph.add_edge
# 编译生成计算图
graph = state_graph.compile
# 打印图结构
print
print
print.print_ascii)
# 执行图
initial_state = {"process_data": {}}
result = graph.invoke
print

⚠️ 常见痛点：

• 节点间参数传递难题：在上例中使用 partial 固定了参数，如果后期需求变更，需要手动修改所有绑定。
• 缺少统一错误处理：虽然设置了重试。但如果出现自定义异常仍会直接抛出，需要进一步封装统一异常捕获。怎么说呢，
• 缓存失效管理：缓存 TTL 设置过大或过小都会影响性能。需要根据业务场景，

Edge 基础笔记

. 普通边

普通边代表无条件跳转，从一个节点直接走到下一个节点。

"""LangGraph 普通边演示"""
from typing_extensions import TypedDict
from langgraph.graph import StateGraph,START,END
class AtguiguState:
说到value，int
至于step，str
def node_a -> dict:
print
return {"value": state + 10,"step": "A执行完毕"}
def node_b -> dict:
print
return {"value": state * 5,"step": "B执行完毕"}
def node_c -> dict:
print
return {"value": state - 7。"step": "C执行完毕"}
def main:
print
builder = StateGraph
builder.add_node
builder.add_node
builder.add_node
# 顺序连线
builder.add_edge
builder.add_edge
builder.add_edge
builder.add_edge
app = builder.compile
result = app.invoke
print
print
print.print_ascii)
print.draw_mermaid)
if __name__ == "__main__":
main

注意： 普通边无法实现分支或循环，若业务需求变化需改动代码； 建议使用条件边或循环边来提高灵活性。其实，

. 条件边

条件边允许根据运行时状态做分支跳转。是实现复杂业务流程的主要。

"""LangGraph 条件边演示"""
from typing import Optional
from langgraph.constants import START。END
from langgraph.graph import StateGraph
from loguru import logger
from pydantic import BaseModel
class MyState:
x:int
result : Optional = None
def check_x-> MyState :
logger.info
return state
def is_even-> bool :
logger.info
return state.x % 2 ==0
def handle_even->MyState :
logger.info
return MyState
def handle_odd ->MyState :
logger.info
return MyState
builder = StateGraph
builder .add_node
builder .add_conditional_edges
builder .add_edge
builder .add_edge
builder .add_edge
graph = builder.compile
print.print_ascii)
logger.info ")
graph.invoke)
logger.info ")
graph.invoke)

常见问题：

• 返回类型必须严格匹配： 如果条件函数返回非布尔值，将导致运行时错误。话说回来，
• 日志混乱： 大量日志会掩盖真正错误信息。建议统一日志级别和格式，
• 可视化难度高： 复杂路由可用 Mermaid 输出。但理解成本仍高，可考虑添加注释说明每条路由含义。

当需要根据多种情况选择不同方法时可以使用多个条件边或路由器实现更细粒度控制。

"""多条件路由示例"""
from langgraph.graph import StateGraph,START,END
from typing_extensions import TypedDict
class AtguiguStatet:
x:int
def addition1 :
"""加法1"""
print
return{"x": state +10 }
def addition2 :
print
return{"x": state +20 }
def addition3 :
print
return{"x": state +30 }
def route_by_sentiment : # 路由函数返回分支名称
flag = statiat
if flag ==10:
return"condition_1"
elif flag==20:
return"condition_2"
从else来看，return"condition_3"
graph = StateGraph
graph .add_node
graph .add_conditional_edges(source=START,path=route_by_sentiment,path_map={"condition_1":"node1","condition_2":"node2"。"condition_3":"node3"})
## 所有处理节点都连接到 END
for name in :
graph .add_edge
app = graph.compile
initial_state={"x":10}
resapp.invoke   # 根据路由选择对应加法
print.draw_mermaid)

⚡️ 调整建议：

• Create a dedicated routing module so that route logic stays isolated from business logic.
• Add descriptive comments for each condition mapping to ease onboarding.
• If conditions become numerous consider using a lookup table instead of many `if/elif` branches.

:  

• No clear separation 娱乐ween nodes and edges leads to tangled graphs.
• Difficulties in parameter binding and retry/caching configuration cause runtime failures.
• Lack of dynamic routing causes frequent code rewrites when business rules change.
• Error handling and logging are inconsistent across nodes.
• Lack of visual debugging tools makes troubleshooting hard.

以上内容为 LangGraph 学习笔记整理与常见痛点分析，仅供参考。如需进一步帮助，请随时提问！🛠️🚀 `