1
from openai import OpenAI
2
client = OpenAI(api_key="<DeepSeek API Key>", base_url="https://api.deepseek.com")
3

4
# Round 1
5
messages = [{"role": "user", "content": "What's the highest mountain in the world?"}]
6
response = client.chat.completions.create(
7
    model="deepseek-v4-pro",
8
    messages=messages
9
)
10

11
messages.append(response.choices[0].message)
12
print(f"Messages Round 1: {messages}")
13

14
# Round 2
15
messages.append({"role": "user", "content": "What is the second?"})
16
response = client.chat.completions.create(
17
    model="deepseek-v4-pro",
18
    messages=messages
19
)
20

21
messages.append(response.choices[0].message)
22
print(f"Messages Round 2: {messages}")

1
[
2
    {"role": "user", "content": "What's the highest mountain in the world?"},
3
    {"role": "assistant", "content": "The highest mountain in the world is Mount Everest."},
4
    {"role": "user", "content": "What is the second?"}
5
]

1
from dotenv import load_dotenv
2
import os
3
from openai import OpenAI
4
load_dotenv()
5
api_key = os.getenv("API_KEY")
6

7
client = OpenAI(
8
            api_key=api_key,
9
            base_url="https://dashscope.aliyuncs.com/compatible-mode/v1"
10
        )
11

12
response = client.chat.completions.create(
13
    model="XXX",
14
    messages=[
15
        {"role": "system", "content": "You are a helpful assistant"},
16
        {"role": "user", "content": "Hello"},
17
    ],
18
    stream=False
19
)
20
print(response.choices[0].message.content)
21
# Hello! How can I assist you today?

1
# uv add langchain
2
# uv add langchain-openai
3
# uv add python-dotenv
4

5
# 在.env文件中添加
6
# OPENAI_API_KEY=XXXXXXXXX
7
# OPENAI_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
8

9
from dotenv import load_dotenv
10
import os
11
load_dotenv()
12
api_key = os.getenv("OPENAI_API_KEY") # 获取API KEY
13

14
# 定义工具
15
from langchain.tools import tool
16
@tool
17
def getWeather(location:str)->str:
18
    """
19
    获取指定位置的天气信息。
20

21
    参数:
22
        location: 城市名称，如"北京"、"上海"
23
    """
24
    # 模拟天气数据（实际使用中可以替换成真实API）
25
    weather_db = {
26
        "北京": "晴天 25°C",
27
        "上海": "多云 28°C",
28
        "深圳": "阵雨 30°C",
29
        "广州": "阴天 27°C",
30
        "杭州": "晴转多云 26°C"
31
    }
32
     # 查找天气
33
    if location in weather_db:
34
        result = f"{location}天气：{weather_db[location]}"
35
    else:
36
        # 默认返回通用天气
37
        result = f"{location}天气：晴 22°C"
38

39
    return result
40

41
# 创建agent
42
from langchain.agents import create_agent
43
agent = create_agent(
44
    model="openai:qwen-plus",
45
    tools=[getWeather]
46
)
47

48
# 发起调用
49
response = agent.invoke({
50
    "messages": [
51
        {"role": "user", "content": "今天深圳天气如何"},
52
    ]
53
})
54

55
print(response)

1
from dotenv import load_dotenv
2
from langchain.chat_models import init_chat_model
3
import os
4
# 在.env文件中添加
5
# OPENAI_API_KEY=XXXXXXXXX
6
# OPENAI_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
7

8
load_dotenv()
9

10
model = init_chat_model(
11
    model="qwen-plus-2025-07-28",
12
    model_provider="openai", # 提供商（qwen不支持自动识别，因此伪装为openai）
13
    temperature=0.7,         # 随机性
14
    max_tokens=1024,         # 输出最大token数
15
    timeout=10,              # 请求超时秒数
16
    max_retries=3,           # 失败重试次数
17
    top_p=0.5,               # 范围 0-1，通常与 temperature 二选一调整
18
    stop=["4.", "4）", "4、"]  # 用于控制回复的结束位置
19
)
20

21
response = model.invoke("你好，请介绍一下自己")
22
print(response.content)
23
print(type(model))
24
# <class 'langchain_openai.chat_models.base.ChatOpenAI'>

1
from langchain.agents import create_agent
2

3
# 方法1:初始化已经定义好的model
4
agent = create_agent(model=model)
5

6
# 方法2:指定model名称
7
agent1 = create_agent("openai:qwen-plus-2025-07-28")

1
# 方法1:invoke
2
response = agent1.invoke({
3
    "messages": [{"role":"user","content":"Hi"}]
4
})
5
print(response["messages"][-1].content)
6
# Hello! ٩(◕‿◕｡)۶ How can I assist you today?
7

8
# 方法2:stream
9
response = agent.stream(
10
    {"messages": [{"role":"user","content":"Hi"}]},
11
    stream_mode="messages"
12
)
13

14
for token, metadata in response:
15
    if token.content:
16
        print(token.content,end="",flush=True)

1
'''省略部分，可参考 3.LangChain使用 - 3.1Demo案例'''
2
# 发起调用
3
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
4
response = agent.invoke({
5
    "messages": [
6
        SystemMessage("可以使用工具获取天气状态"),
7
        HumanMessage("你好"),
8
        AIMessage("你好我是DeepSeek，请问有什么可以帮忙的？"),
9
        HumanMessage("香港的天气怎么样？")
10
    ]
11
})
12
print(response["messages"][-1].content)
13
# 香港目前天气晴朗，气温为22°C，非常舒适。建议穿着轻便衣物，适合外出活动。
14
for message in response["messages"]:
15
    message.pretty_print()
16
'''
17
=== System Message ========================
18
可以使用工具获取天气状态
19
=== Human Message =========================
20
你好
21
=== Ai Message ============================
22
你好我是DeepSeek，请问有什么可以帮忙的？
23
=== Human Message =========================
24
今天香港天气怎么样？
25
===Ai Message =============================
26
Tool Calls:
27
  getWeather (call_XXX)
28
 Call ID: call_XXX
29
  Args:
30
    location: 香港
31
=== Tool Message ==========================
32
Name: getWeather
33
香港天气：晴 22°C
34
==== Ai Message ===========================
35
今天香港天气晴朗，气温为22°C。适合外出活动哦！
36
'''

1
from langchain.agents import create_agent
2
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
3
from langchain.chat_models import init_chat_model
4
import os
5
from dotenv import load_dotenv
6
load_dotenv()
7

8
# 创建model
9
model = init_chat_model(
10
    model="qwen-plus-2025-07-28",
11
    model_provider="openai",
12
    temperature=0.7,
13
    max_tokens=1024
14
)
15

16
# 创建agent
17
agent = create_agent(
18
    model=model,
19
    system_prompt="请你以XXX的语气并进行简短回答",
20
)
21

22
# 调用智能体
23
for token, metadata in agent.stream(
24
        {
25
            "messages": [HumanMessage(content="你是谁？")]
26
        },
27
        stream_mode="messages"
28

29
):
30
    print(token.content,end="",flush=True)

1
from langchain_core.tools import tool
2
@tool("square_tool",description="用于计算一个数的平方根值")
3
def square_tool(x:float)->float:
4
    return x**2

1
from langchain_core.tools import tool
2
@tool
3
def square_tool(x:float)->float:
4
    """用于计算一个数的平方值"""
5
    return x**2

1
@tool
2
def get_weather(location: str, units: str = "celsius", include_forecast: bool = False) -> str:
3
    """
4
    获取当前天气和可选的天气预报。
5

6
    参数:
7
        location: 城市名称或坐标
8
        units: 温度单位
9
        include_forecast: 是否包含天气预报
10
    """
11
    temp = 22 if units == "celsius" else 72
12
    result = f"{location}当前天气：{temp} 度 {units[0].upper()}"
13
    if include_forecast:
14
        result += "\n未来5天：晴朗"
15
    return result
16

17
get_weather.invoke({"location":"HongKong","units":"celsius","include_forecast":False})

1
# 通过自定义模型约束入参
2
from pydantic import BaseModel, Field
3
from typing import Literal # 枚举
4

5
# 天气查询工具的输入参数模型
6
class WeatherInput(BaseModel):
7
    location: str = Field(description="城市名或坐标")
8
    units: Literal["celsius", "fahrenheit"] = Field(
9
        default="celsius",
10
        description="温度单位"
11
    )
12
    include_forecast: bool = Field(
13
        default=False,
14
        description="是否包含5天预报"
15
    )
16

17
@tool(args_schema=WeatherInput)  # 参数约束 args_schema
18
def get_weather(location: str, units: str = "celsius", include_forecast: bool = False) -> str:
19
    """获取当前天气和可选的天气预报。"""
20
    temp = 22 if units == "celsius" else 72
21
    result = f"{location}当前天气：{temp} 度 {units[0].upper()}"
22
    if include_forecast:
23
        result += "\n未来5天：晴朗"
24
    return result

1
from langchain_tavily import TavilySearch
2
from dotenv import load_dotenv
3
import os
4
load_dotenv()
5

6
search_tool = TavilySearch(
7
    max_results=5,                     # 最大返回结果数
8
    topic="general",                   # 搜索主题：general/news/finance
9
    # include_answer=False,            # 是否包含AI生成答案
10
    # include_raw_content=False,       # 是否包含完整页面内容
11
    # include_images=False,            # 是否包含图片列表
12
    # include_image_descriptions=False,# 是否包含图片描述
13
    # search_depth="basic",            # 搜索深度：basic/advanced
14
    # time_range="day",                # 时间过滤：day/week/month/year
15
    # include_domains=None,            # 限定域名列表
16
    # exclude_domains=None             # 排除域名列表
17
)
18
search_tool.invoke({"query": "上周热点新闻有哪些？简短回答即可"})

1
tavily = TavilySearch(
2
    max_results=5,
3
    topic="general",
4
)
5

6
@tool
7
def web_search(query:str):
8
    """用于Web网页信息检索"""
9
    return tavily.invoke(query)
10

11
from pydantic import BaseModel,Field
12
# Agent回答内容引用的网页信息
13
class Reference(BaseModel):
14
    title: str = Field(description="答案中引用的网页标题")
15
    url: str = Field(description="答案中引用的网页链接")
16

17
# Agent的回答内容
18
class AnswerInfo(BaseModel):
19
    answer: str = Field(description="给用户的最终答案")
20
    reference: list[Reference] = Field(description="答案所引用的网页")
21

22
model = init_chat_model(
23
    model="qwen-plus-2025-07-28",
24
    model_provider="openai",
25
    temperature=0.7,
26
    max_tokens=1024
27
)
28

29
agent = create_agent(
30
    model=model,
31
    tools=[web_search],
32
    system_prompt="你是一个智能助手，用于解决用户问题",
33
    response_format=AnswerInfo   # 控制 Agent 返回结构化数据的核心参数
34
)
35

36
response = agent.invoke(
37
    {"messages":[HumanMessage("知乎今日热点都有哪些？")]}
38
)
39

40
structured = response["structured_response"]
41
print("=== 答案 ===")
42
print(structured.answer)
43
print("\n=== 引用来源 ===")
44
for idx, ref in enumerate(structured.reference, 1):
45
    print(f"{idx}. {ref.title}")
46
    print(f"   链接：{ref.url}")
47

48
'''
49
=== 答案 ===
50
知乎的今日热点内容通常包括实时讨论度高的社会事件、科技动态、文化娱乐等话题。您可以通过知乎App或网页端的“今日热点”话题页（如 https://www.zhihu.com/topic/23509571）查看最新热门讨论。此外，聚合类平台如“今日热榜”（tophub.today）也提供知乎实时热点榜单，便于一站式浏览全网热点。
51

52
=== 引用来源 ===
53
1. 今日热点 - 知乎
54
   链接：https://www.zhihu.com/topic/23509571
55
2. 真的能追溯微博、知乎等历史热门榜单吗？
56
   链接：https://zhuanlan.zhihu.com/p/1925271264437733172
57
'''