千问AI agent qwan_agent使用

代码，一个简单的例子：

?
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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
# Reference: https://platform.openai.com/docs/guides/function-calling
import json
import os
 
# DASHSCOPE_API_KEY
 
from qwen_agent.llm import get_chat_model
 
 
# Example dummy function hard coded to return the same weather
# In production, this could be your backend API or an external API
def get_current_weather(location, unit='fahrenheit'):
    """Get the current weather in a given location"""
    if 'tokyo' in location.lower():
        return json.dumps({'location': 'Tokyo', 'temperature': '10', 'unit': 'celsius'})
    elif 'san francisco' in location.lower():
        return json.dumps({'location': 'San Francisco', 'temperature': '72', 'unit': 'fahrenheit'})
    elif 'paris' in location.lower():
        return json.dumps({'location': 'Paris', 'temperature': '22', 'unit': 'celsius'})
    else:
        return json.dumps({'location': location, 'temperature': 'unknown'})
 
 
def test():
    llm = get_chat_model({
        # Use the model service provided by DashScope:
        # 'model': 'qwen-max',
        'model': 'qwen-plus',
        'model_server': 'dashscope',
        'api_key':  'sk-c78替换下c8',#os.getenv('DASHSCOPE_API_KEY'),
 
        # Use the model service provided by Together.AI:
        # 'model': 'Qwen/Qwen1.5-14B-Chat',
        # 'model_server': 'https://api.together.xyz',  # api_base
        # 'api_key': os.getenv('TOGETHER_API_KEY'),
 
        # Use your own model service compatible with OpenAI API:
        # 'model': 'Qwen/Qwen1.5-72B-Chat',
        # 'model_server': 'http://localhost:8000/v1',  # api_base
        # 'api_key': 'EMPTY',
    })
 
    # Step 1: send the conversation and available functions to the model
    messages = [{'role': 'user', 'content': "What's the weather like in San Francisco?"}]
    functions = [{
        'name': 'get_current_weather',
        'description': 'Get the current weather in a given location',
        'parameters': {
            'type': 'object',
            'properties': {
                'location': {
                    'type': 'string',
                    'description': 'The city and state, e.g. San Francisco, CA',
                },
                'unit': {
                    'type': 'string',
                    'enum': ['celsius', 'fahrenheit']
                },
            },
            'required': ['location'],
        },
    }]
 
    print('# Assistant Response 1:')
    responses = []
    for responses in llm.chat(messages=messages, functions=functions, stream=True):
        print(responses)
 
    messages.extend(responses)  # extend conversation with assistant's reply
 
    # Step 2: check if the model wanted to call a function
    last_response = messages[-1]
    print("*"*88)
    print(last_response)
    print("*"*88)
    if last_response.get('function_call', None):
 
        # Step 3: call the function
        # Note: the JSON response may not always be valid; be sure to handle errors
        available_functions = {
            'get_current_weather': get_current_weather,
        }  # only one function in this example, but you can have multiple
        function_name = last_response['function_call']['name']
        function_to_call = available_functions[function_name]
        function_args = json.loads(last_response['function_call']['arguments'])
        function_response = function_to_call(
            location=function_args.get('location'),
            unit=function_args.get('unit'),
        )
        print('# Function Response:')
        print(function_response)
 
        # Step 4: send the info for each function call and function response to the model
        messages.append({
            'role': 'function',
            'name': function_name,
            'content': function_response,
        })  # extend conversation with function response
 
        print('# Assistant Response 2:')
        for responses in llm.chat(
                messages=messages,
                functions=functions,
                stream=True,
        ):  # get a new response from the model where it can see the function response
            print(responses)
 
 
if __name__ == '__main__':
    test()

　　

运行效果：

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
# Assistant Response 1:
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': ''}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': ''}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location": "San Francisco'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location": "San Francisco, CA",'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location": "San Francisco, CA",\n  "unit": "'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location": "San Francisco, CA",\n  "unit": "celsius"\n}'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location": "San Francisco, CA",\n  "unit": "celsius"\n}'}}]
****************************************************************************************
{'role': 'assistant', 'content': '', 'function_call': {'name': 'get_current_weather', 'arguments': '{\n  "location": "San Francisco, CA",\n  "unit": "celsius"\n}'}}
****************************************************************************************
# Function Response:
{"location": "San Francisco", "temperature": "72", "unit": "fahrenheit"}
# Assistant Response 2:
[{'role': 'assistant', 'content': 'The'}]
[{'role': 'assistant', 'content': 'The current'}]
[{'role': 'assistant', 'content': 'The current weather'}]
[{'role': 'assistant', 'content': 'The current weather in San Francisco,'}]
[{'role': 'assistant', 'content': 'The current weather in San Francisco, California is 7'}]
[{'role': 'assistant', 'content': 'The current weather in San Francisco, California is 72 degrees Fahrenheit ('}]
[{'role': 'assistant', 'content': 'The current weather in San Francisco, California is 72 degrees Fahrenheit (approximately 22'}]
[{'role': 'assistant', 'content': 'The current weather in San Francisco, California is 72 degrees Fahrenheit (approximately 22.2 degrees Celsius'}]
[{'role': 'assistant', 'content': 'The current weather in San Francisco, California is 72 degrees Fahrenheit (approximately 22.2 degrees Celsius).'}]

　　

再来一个复杂的工具调用例子：

?
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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
import json
import os
from qwen_agent.llm import get_chat_model
 
 
# Example dummy function hard coded to return the same weather
# In production, this could be your backend API or an external API
def get_current_weather(location, unit='fahrenheit'):
    """Get the current weather in a given location"""
    if 'tokyo' in location.lower():
        return json.dumps({'location': 'Tokyo', 'temperature': '10', 'unit': 'celsius'})
    elif 'san francisco' in location.lower():
        return json.dumps({'location': 'San Francisco', 'temperature': '72', 'unit': 'fahrenheit'})
    elif 'paris' in location.lower():
        return json.dumps({'location': 'Paris', 'temperature': '22', 'unit': 'celsius'})
    else:
        return json.dumps({'location': location, 'temperature': 'unknown'})
 
def check_inventory(flower_type: str = "玫瑰花") -> int:
    """
    查询特定类型花的库存数量，默认为玫瑰花。
    参数:
    - flower_type: 花的类型
    返回:
    - 库存数量
    """
    # 实际应用中这里应该是数据库查询或其他形式的库存检查
    return '130'  # 假设每种花都有130个单位
 
def get_extra_asset_price() -> float:
    """
    获取额外的固定资产价格。get extra asset price。
    参数: 无
    返回:
    - 额外的固定资产价格
    """
    return '500'
 
 
def process_responses(llm, messages, functions):
    print('# Assistant Response with:', messages)
    responses = []
    for responses in llm.chat(messages=messages, functions=functions, stream=True):
        print(responses)
        # responses.append(response)
    print("llm.chat over")
    # 注意，这里添加了回复，方便递归调用工具
    messages.extend(responses)  # Extend conversation with assistant's replies
 
 
def handle_function_calls(llm, messages, functions):
    process_responses(llm, messages, functions)
    last_response = messages[-1]
    print("last_res:", last_response)
    while last_response.get('function_call', None):
        available_functions = {
            'get_current_weather': get_current_weather,
            'check_inventory': check_inventory,
            'get_extra_asset_price': get_extra_asset_price
        }
        function_name = last_response['function_call']['name']
        function_to_call = available_functions.get(function_name)
        if function_to_call:
            function_args = json.loads(last_response['function_call']['arguments'])
            function_response = function_to_call(**function_args)
        else:
            function_response = "Function not found or not available in available_functions"
        print('# Function Response:', function_response)
 
        messages.append({
            'role': 'function',
            'name': function_name,
            'content': function_response,
        })  # Extend conversation with function response
 
        print("check messages:", messages)
        process_responses(llm, messages, functions)  # Get a new response from the model
        print("*" * 88)
        last_response = messages[-1]  # Update last response for next iteration
 
    # process_responses(llm, messages, functions)
 
def test():
    llm = get_chat_model({
        # Use the model service provided by DashScope:
        # 'model': 'qwen-max',
        'model': 'qwen-plus',
        'model_server': 'dashscope',
        'api_key':  'sk-c78896504df24xxxx7bf67529c8',#os.getenv('DASHSCOPE_API_KEY'),
 
        # Use the model service provided by Together.AI:
        # 'model': 'Qwen/Qwen1.5-14B-Chat',
        # 'model_server': 'https://api.together.xyz',  # api_base
        # 'api_key': os.getenv('TOGETHER_API_KEY'),
 
        # Use your own model service compatible with OpenAI API:
        # 'model': 'Qwen/Qwen1.5-72B-Chat',
        # 'model_server': 'http://localhost:8000/v1',  # api_base
        # 'api_key': 'EMPTY',
    })
 
    # Step 1: send the conversation and available functions to the model
    # messages = [{'role': 'user', 'content': "What's the weather like in San Francisco?"}]
    messages = [{'role': 'user', 'content': "通过使用工具(函数)，查询玫瑰花的库存数量。已知玫瑰花的单价是10元，帮我计算玫瑰花总价加上额外固定资产的总价是多少?"}]
    functions = [{
        'name': 'get_current_weather',
        'description': 'Get the current weather in a given location',
        'parameters': {
            'type': 'object',
            'properties': {
                'location': {
                    'type': 'string',
                    'description': 'The city and state, e.g. San Francisco, CA',
                },
                'unit': {
                    'type': 'string',
                    'enum': ['celsius', 'fahrenheit']
                },
            },
            'required': ['location']}
        },
        {
            "name": "check_inventory",
            "description": "Check the inventory for a specific type of flower, default is roses",
            "parameters": {
                "type": "object",
                "properties": {
                    "flower_type": {
                        "type": "string",
                        "description": "Type of flower to check inventory for"
                    }
                },
                "required": []
            }
        },
        {
            "name": "get_extra_asset_price",
            "description": "Get the price of an extra fixed asset",
            "parameters": {
                "type": "object",
                "properties": {},
                "required": []
            }
        }
    ]
    handle_function_calls(llm, messages, functions)
 
if __name__ == '__main__':
    test()

　　

发现千问agent效果还不错，基本上每次都能得到正确结果！

?
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
D:\Python\Python312\python.exe D:\source\pythonProject\qianwen_x.py
# Assistant Response with: [{'role': 'user', 'content': '通过使用工具(函数)，查询玫瑰花的库存数量。已知玫瑰花的单价是10元，帮我计算玫瑰花总价加上额外固定资产的总价是多少?'}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': ''}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}]
[{'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}]
llm.chat over
last_res: {'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}
# Function Response: 130
check messages: [{'role': 'user', 'content': '通过使用工具(函数)，查询玫瑰花的库存数量。已知玫瑰花的单价是10元，帮我计算玫瑰花总价加上额外固定资产的总价是多少?'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}, {'role': 'function', 'name': 'check_inventory', 'content': '130'}]
# Assistant Response with: [{'role': 'user', 'content': '通过使用工具(函数)，查询玫瑰花的库存数量。已知玫瑰花的单价是10元，帮我计算玫瑰花总价加上额外固定资产的总价是多少?'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}, {'role': 'function', 'name': 'check_inventory', 'content': '130'}]
[{'role': 'assistant', 'content': '当'}]
[{'role': 'assistant', 'content': '当前'}]
[{'role': 'assistant', 'content': '当前我们'}]
[{'role': 'assistant', 'content': '当前我们有130'}]
[{'role': 'assistant', 'content': '当前我们有130朵玫瑰花在'}]
[{'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}]
[{'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'get_extra', 'arguments': ''}}]
[{'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'get_extra_asset_price', 'arguments': ''}}]
[{'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'get_extra_asset_price', 'arguments': '{}'}}]
[{'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'get_extra_asset_price', 'arguments': '{}'}}]
llm.chat over
****************************************************************************************
# Function Response: 500
check messages: [{'role': 'user', 'content': '通过使用工具(函数)，查询玫瑰花的库存数量。已知玫瑰花的单价是10元，帮我计算玫瑰花总价加上额外固定资产的总价是多少?'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}, {'role': 'function', 'name': 'check_inventory', 'content': '130'}, {'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'get_extra_asset_price', 'arguments': '{}'}}, {'role': 'function', 'name': 'get_extra_asset_price', 'content': '500'}]
# Assistant Response with: [{'role': 'user', 'content': '通过使用工具(函数)，查询玫瑰花的库存数量。已知玫瑰花的单价是10元，帮我计算玫瑰花总价加上额外固定资产的总价是多少?'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'check_inventory', 'arguments': '{"flower_type": "roses"}'}}, {'role': 'function', 'name': 'check_inventory', 'content': '130'}, {'role': 'assistant', 'content': '当前我们有130朵玫瑰花在库存。'}, {'role': 'assistant', 'content': '', 'function_call': {'name': 'get_extra_asset_price', 'arguments': '{}'}}, {'role': 'function', 'name': 'get_extra_asset_price', 'content': '500'}]
[{'role': 'assistant', 'content': '额'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 ='}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 130'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的总价是13'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的总价是1300元 +'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的总价是1300元 + 500'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的总价是1300元 + 500元 = 1'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的总价是1300元 + 500元 = 1800元'}]
[{'role': 'assistant', 'content': '额外固定资产的价格是500元。\n\n现在，玫瑰花的总价是130朵 * 10元/朵 = 1300元。\n\n所以，玫瑰花总价加上额外固定资产的总价是1300元 + 500元 = 1800元。'}]
llm.chat over
****************************************************************************************
 
Process finished with exit code 0

　　

不同模型的表现情况：

'model': 'qwen-max',
# 'model': 'qwen-turbo',
# 'model': 'qwen-long', # 无法正常执行
# 'model': 'qwen-7b-chat', #无法得到正确结果
# 'model': 'qwen-1.8b-chat', # 无法得到正确结果
# 'model': 'qwen-max-0428',
# 'model': 'qwen-plus',

整体来说，比langchain的react框架好用不少！
posted @   bonelee  阅读(111)  评论(0)  编辑  收藏  举报
相关博文：
· AI agent开发——使用react框架
· Langchain Plan-And-Execute框架使用
· 如何实现LLM的通用function-calling能力？
· OpenAi FunctionCalling 案例详解
· 简单剖析qwen-agent回答是怎么获取tool的
阅读排行：
· 全程不用写代码，我用AI程序员写了一个飞机大战
· MongoDB 8.0这个新功能碉堡了，比商业数据库还牛
· 记一次.NET内存居高不下排查解决与启示
· 白话解读 Dapr 1.15：你的「微服务管家」又秀新绝活了
· DeepSeek 开源周回顾「GitHub 热点速览」
历史上的今天：
2022-06-07 红蓝对抗之邮件钓鱼攻击——用到了很多典型无文件攻击手段
2021-06-07 python目录扫描工具——dirsearch使用，可以使用御剑的字典 支持慢速扫描，一般使用-s 60绕过waf，-s DELAY, --delay=DELAY 设置请求之间的延时
2021-06-07 一条Fofa搜索语法，实现批量挖洞——很好的文章，自己实践下
2018-06-07 leetcode 459. Repeated Substring Pattern
2018-06-07 gym强化学习入门demo——随机选取动作 其实有了这些动作和反馈值以后就可以用来训练DNN网络了
2017-06-07 suse的安装命令zypper，类似apt
2017-06-07 ES transport client批量导入
点击右上角即可分享
微信分享提示