大模型应用之工具调用与流程编排

 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
56
57
58
59
60
61
62
63
64
65
66

import random
from langchain.agents import Tool
from langchain.agents import AgentType
from langchain.memory import ConversationBufferMemory
from langchain.agents import initialize_agent
from langchain.llms import Tongyi


# 定义WeatherTool
class WeatherTool:
    def get_weather(self, location: str):
        return random.choice(["Sunny", "Rainy", "Snowy"])


# 定义MusicTool
class MusicTool:
    def get_music(self, song: str):
        return [
            {"name": "HappySong", "type": "happy"},
            {"name": "SadSong", "type": "Sad"},
            {"name": "AngrySong", "type": "Angry"}
        ]


# 创建工具实例
weather_tool = WeatherTool()
music_tool = MusicTool()

# 定义LangChain工具
tools = [
    Tool(
        name="Weather",
        func=weather_tool.get_weather,
        description="Useful for getting the current weather"
    ),
    Tool(
        name="Music",
        func=music_tool.get_music,
        description="Useful for getting music information"
    )
]

# 初始化语言模型
# llm = ChatOpenAI(temperature=0)
import os

os.environ["DASHSCOPE_API_KEY"] = "xxxx"
llm = Tongyi(model="qwen-turbo")
# 初始化内存
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

# 初始化agent
agent = initialize_agent(
    tools,
    llm,
    agent=AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
    memory=memory,
    verbose=True
)

# 使用agent
response = agent.run("What's the weather like today?")
print(response)

response = agent.run("Can you recommend a song based on the weather?")
print(response)

1
2
3
4
5
6
7
8

TOOL RESPONSE: 
---------------------
[{'name': 'HappySong', 'type': 'happy'}, {'name': 'SadSong', 'type': 'Sad'}, {'name': 'AngrySong', 'type': 'Angry'}]

USER'S INPUT
--------------------

Okay, so what is the response to my last comment? If using information obtained from the tools you must mention it explicitly without mentioning the tool names - I have forgotten all TOOL RESPONSES! Remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else.

1
2
3
4

{
    "action": "Final Answer",
    "action_input": "Since it's rainy, you might enjoy listening to SadSong to match the weather."
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

你是一个从自然语言中提取参数的助手，请根据用户的输入，提取指定的参数
根据用户的输入，提取城市：city，以及日期：date两个参数，并用json格式输出。
输出样例如下：
{
    "city":"hangzhou",
    "date":"2024-01-01"
}

用户的输入如下：
${sys.query}

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

def main():
    import json
    import random
    city=json.loads(params["input"])["city"] 
    date=json.loads(params["input"])["date"] 
    ret = {
        "result": {
           "city": city, 
           "date": date, 
            "ret":random.choice(["Sunny", "Rainy", "Snowy"])
            }
    }
    return ret

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

def main(): 
  import json  
  ret = {        
    "result":  [
            {"name": "HappySong", "type": "happy"},
            {"name": "SadSong", "type": "Sad"},
            {"name": "AngrySong", "type": "Angry"}
        ]    
  }    
  return ret