01- LangChain入门（基于1.x版本）

whz-emm

432人浏览 · 2025-11-25 23:34:06

whz-emm · 2025-11-25 23:34:06 发布

介绍

LangChain

LangChain 是使用 LLM 构建代理和应用程序的最简单方式。只需不到 10 行代码，您就可以连接到 OpenAI、Anthropic、Google 等。LangChain 提供了预构建的代理架构和模型集成，帮助您快速开始，并轻松将 LLM 集成到您的代理和应用程序中。
如果我们想快速构建代理和自主应用程序，我们推荐使用 LangChain。

LangGraph

当您有更高级的需求，需要结合确定性流程和自主流程、大量定制以及严格控制延迟时，请使用 LangGraph，这是我们低级别的代理编排框架和运行时。
LangChain 代理建立在 LangGraph 之上，以提供持久执行、流式传输、人工参与、持久性等功能。对于基本的 LangChain 代理使用，您不需要了解 LangGraph。

安装LangChain

LangChain是一个Python的包（包的集合），所以可以使用pip来安装

pip install -U langchain

推荐使用uv来安装

uv add langchain

如果没有安装uv，可以使用pip安装，也是个python包

pip install uv

详细步骤参考：创建第一个agent

使用阿里百炼大模型

（也可以本地部署大模型）

创建API-KEY

左侧菜单找到秘钥管理

创建一个秘钥

复制秘钥，请勿泄露

选择模型

可以在模型广场找一个模型，比如qwen3-max

点进入复制模型名即可

获取模型地址

百炼的模型都是这个地址

创建第一个agent

创建工作空间

mkdir langchain-101
cd langchain-101

使用uv初始化工作空间

uv init

安装langchain

uv add langchain
uv add langchain-openai

代码如下

import os

from langchain.agents import create_agent
from langchain_openai import ChatOpenAI

def get_weather(city: str) -> str:
    """Get weather for a given city."""
    return f"It's always sunny in {city}!"

llm = ChatOpenAI(
    model="qwen3-max",
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
    api_key=os.environ.get("CHAT_OPENAI_API_KEY"),
    streaming=True,
)

agent = create_agent(
    model=llm,
    tools=[get_weather],
    system_prompt="You are a helpful assistant",
)

# Run the agent
resp = agent.invoke(
    {"messages": [{"role": "user", "content": "what is the weather in sf"}]}
)

print(resp)

输出如下，自动调用了weather工具，输出永远是晴天

{'messages': [HumanMessage(content='what is the weather in sf', additional_kwargs={}, response_metadata={}, id='679ba2f9-4744-4a7f-a608-9895068ed767'), AIMessage(content='', additional_kwargs={}, respo
nse_metadata={'finish_reason': 'tool_calls', 'model_name': 'qwen3-max', 'model_provider': 'openai'}, id='lc_run--04a23332-90b3-464a-98df-ee4e0e819377', tool_calls=[{'name': 'get_weather', 'args': {'cit
y': 'sf'}, 'id': 'call_f8377606502149ddb21c67f5', 'type': 'tool_call'}]), ToolMessage(content="It's always sunny in sf!", name='get_weather', id='dd49dc87-3c21-42b5-9d2f-78700e41f830', tool_call_id='ca
ll_f8377606502149ddb21c67f5'), AIMessage(content="It's always sunny in San Francisco!", additional_kwargs={}, response_metadata={'finish_reason': 'stop', 'model_name': 'qwen3-max', 'model_provider': 'openai'}, id='lc_run--33f8c265-24b9-40ce-adc4-98599669731b')]}

调用本地模型

可以按照以下local的方式创建，本地模型通常不需要api_key，但是必传，所以给个假的即可

import os

from langchain_openai import ChatOpenAI
from pydantic import SecretStr

bailian = ChatOpenAI(
    model="qwen3-max",
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
    api_key=os.environ.get("CHAT_OPENAI_API_KEY"),
    streaming=True,
)

local = ChatOpenAI(
    model="Qwen/Qwen3-VL-2B-Instruct",
    base_url="http://127.0.0.1:8000/v1",
    api_key=SecretStr("local"),
)

all_models = {
    "bailian": bailian,
    "local": local
}

def get_model(name: str) -> ChatOpenAI | None:
    if name in all_models:
        return all_models[name]

更多配置

定义系统提示词

也可以用中文描述，描述怎样使用工具，包含两个工具：

get_weather_for_location：使用这个工具获取某个特定位置的天气

get_user_location：使用这个工具获取用户所在位置

SYSTEM_PROMPT = """You are an expert weather forecaster, who speaks in puns.

You have access to two tools:

- get_weather_for_location: use this to get the weather for a specific location
- get_user_location: use this to get the user's location

If a user asks you for the weather, make sure you know the location. If you can tell from the question that they mean wherever they are, use the get_user_location tool to find their location."""

创建工具

目前将返回值写死，因为本地工具无法真的去获取到天气，需要使用第三方工具才能实现

注意：工具必须添加注释，注释是个大模型用的，这样大模型才知道什么时候调这个工具

from dataclasses import dataclass
from langchain.tools import tool, ToolRuntime

@tool
def get_weather_for_location(city: str) -> str:
    """Get weather for a given city."""
    return f"It's always sunny in {city}!"

@dataclass
class Context:
    """Custom runtime context schema."""
    user_id: str

@tool
def get_user_location(runtime: ToolRuntime[Context]) -> str:
    """Retrieve user information based on user ID."""
    user_id = runtime.context.user_id
    return "Florida" if user_id == "1" else "SF"

配置模型

bailian = ChatOpenAI(
    model="qwen3-max",
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
    api_key=os.environ.get("CHAT_OPENAI_API_KEY"),
    streaming=True,
    temperature=0.5,  // 越低越稳定，越高越发散
    timeout=10,
    max_tokens=1000
)

定义响应格式

from dataclasses import dataclass

# We use a dataclass here, but Pydantic models are also supported.
@dataclass
class ResponseFormat:
    """Response schema for the agent."""
    # A punny response (always required)
    punny_response: str
    # Any interesting information about the weather if available
    weather_conditions: str | None = None

定义记忆

使用内存记忆，thread_id唯一标识一个会话

from langgraph.checkpoint.memory import InMemorySaver

checkpointer = InMemorySaver()
config = {"configurable": {"thread_id": "1"}}

配置agent

将上述参数配置到agent上

agent = create_agent(
    model=get_model("bailian"),
    tools=[get_user_location, get_weather_for_location],
    context_schema=Context,
    response_format=ToolStrategy(ResponseFormat),
    checkpointer=checkpointer
)

调用大模型

resp = agent.invoke(
    {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
    config=config,
    context=Context(user_id="1"),
)

完整代码

main.py

from langchain.agents import create_agent
from models.models import get_model
from dataclasses import dataclass
from langgraph.checkpoint.memory import InMemorySaver
from langchain.agents.structured_output import ToolStrategy

from tools.local_tools import get_user_location, get_weather_for_location, Context


SYSTEM_PROMPT = """You are an expert weather forecaster, who speaks in puns.

You have access to two tools:

- get_weather_for_location: use this to get the weather for a specific location
- get_user_location: use this to get the user's location

If a user asks you for the weather, make sure you know the location. If you can tell from the question that they mean wherever they are, use the get_user_location tool to find their location."""

# We use a dataclass here, but Pydantic models are also supported.
@dataclass
class ResponseFormat:
    """Response schema for the agent."""
    # A punny response (always required)
    punny_response: str
    # Any interesting information about the weather if available
    weather_conditions: str | None = None


checkpointer = InMemorySaver()
config = {"configurable": {"thread_id": "1"}}


agent = create_agent(
    model=get_model("bailian"),
    tools=[get_user_location, get_weather_for_location],
    context_schema=Context,
    response_format=ToolStrategy(ResponseFormat),
    checkpointer=checkpointer
)

# Run the agent
resp = agent.invoke(
    {"messages": [{"role": "user", "content": "what is the weather in sf"}]},
    config=config,
    context=Context(user_id="1"),
)

print(resp['structured_response'])
print("-------------------------------------")


resp = agent.invoke(
    {"messages": [{"role": "user", "content": "thank you!"}]},
    config=config,
    context=Context(user_id="1")
)

print(resp['structured_response'])

models.py

import os

from langchain_openai import ChatOpenAI
from pydantic import SecretStr

bailian = ChatOpenAI(
    model="qwen3-max",
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
    api_key=os.environ.get("CHAT_OPENAI_API_KEY"),
    streaming=True,
    temperature=0.5,
    timeout=10,
    max_tokens=1000
)

local = ChatOpenAI(
    model="Qwen/Qwen3-VL-2B-Instruct",
    base_url="http://127.0.0.1:8000/v1",
    api_key=SecretStr("local"),
    temperature=0.5,
    timeout=10,
    max_tokens=1000
)

all_models = {
    "bailian": bailian,
    "local": local
}

def get_model(name: str) -> ChatOpenAI | None:
    if name in all_models:
        return all_models[name]

tools.py

from dataclasses import dataclass
from langchain.tools import tool, ToolRuntime

@tool
def get_weather_for_location(city: str) -> str:
    """Get weather for a given city."""
    return f"It's always sunny in {city}!"

@dataclass
class Context:
    """Custom runtime context schema."""
    user_id: str

@tool
def get_user_location(runtime: ToolRuntime[Context]) -> str:
    """Retrieve user information based on user ID."""
    user_id = runtime.context.user_id
    return "Florida" if user_id == "1" else "SF"

AtomGit开源社区

AtomGit 是由开放原子开源基金会联合 CSDN 等生态伙伴共同推出的新一代开源与人工智能协作平台。平台坚持“开放、中立、公益”的理念，把代码托管、模型共享、数据集托管、智能体开发体验和算力服务整合在一起，为开发者提供从开发、训练到部署的一站式体验。

更多推荐

一个入口搞定 Claude/Grok，技术难题全拿捏

AtomGit开源社区

通义千问核心能力与实战表现深度评测

在实际开发和技术选型的过程中，我们常常面临这样一个困境：面对市面上层出不穷的大语言模型，究竟哪一款才能真正融入我们的日常工作流？很多时候，宣传参数是一回事，实际落地又是另一回事。有的模型在简单问答上表现惊艳，一旦进入多轮对话就“失忆”；有的在代码生成上看似流畅，却埋下了难以察觉的逻辑隐患；还有的在处理长文档时，关键信息提取总是差强人意。对于依赖 AI 辅助编程、文档分析或创意构思的开发者而言，盲目