手把手教你使用Qwen-Agent开发智能体应用实战教程 原创

# 全功能安装（RAG/代码解释器/GUI支持）
pip install -U "qwen-agent[rag,code_interpreter,python_executor,gui]"


# 简约安装版本
pip install -U qwen-agent
1.
2.
3.
4.
5.
6.

export DASHSCOPE_API_KEY='your-api-key'
1.

from vllm import LLM, SamplingParams


llm = LLM(model="Qwen2-7B-Chat")
1.
2.
3.
4.

from qwen_agent.tools.base import BaseTool, register_tool
import json5


@register_tool('calculate')
class Calculator(BaseTool):
    description = '基础运算计算器'
    parameters = [{'name': 'formula', 'type': 'string'}]
    
    def call(self, params: str) -> float:
        return eval(json5.loads(params)['formula'])


# 调用示例
calc = Calculator()
print(calc.call('{"formula": "(3 + 5) * 2"}'))  # 输出 16.0
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.

from qwen_agent.agents import Assistant


class HistoryAssistant(Assistant):
    def _postprocess_messages(self, messages):
        return messages[-10:]  # 保留最近5轮对话


assistant = HistoryAssistant(llm={'model': 'qwen-max'})
1.
2.
3.
4.
5.
6.
7.
8.
9.

from qwen_agent.agents import Assistant
from qwen_agent.tools import BaseTool, register_tool
import requests
import json5


@register_tool('city_info')
class CityInfoTool(BaseTool):
    description = "城市基础信息查询"
    parameters = [{'name': 'name', 'type': 'string'}]
    
    def call(self, params):
        city = json5.loads(params)['name']
        response = requests.get(f"https://api.example.com/cities/{city}")
        return response.json()


# 配置助手
assistant = Assistant(
    llm={'model': 'qwen-max'},
    function_list=['city_info','code_interpreter'],
    system_message="你是一个城市百科助手"
)


# 测试查询
response = assistant.run([{'role': 'user', 'content': '上海有多少个行政区?'}])
print(response[-1]['content'])
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.

import urllib
from qwen_agent.agents import Assistant


assistant = Assistant(
    function_list=['code_interpreter'],
    system_message="图像处理专家"
)


def process_image(prompt):
    messages = [{'role':'user', 'content': prompt}]
    for resp in assistant.run(messages):
        if 'function_call' in resp:
            code = resp['function_call']['arguments']
            exec(code)  # 示例简化执行
    return resp[-1]['content']


process_image('将https://example.com/image.jpg的水平宽度扩大1.5倍')
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.

from qwen_agent import retrieve


# 构建知识库
retrieve.build_index('documents/')


class DocQA(Assistant):
    def _preprocess(self, query):
        contexts = retrieve.search(query)
        return f"根据文档：{contexts}\n回答：{query}"


qa = DocQA(llm={'model': 'qwen-max-longcontext'})
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.

from qwen_agent.gui import WebUI


WebUI(assistant).launch(server_name='0.0.0.0', server_port=7860)
1.
2.
3.
4.

class CustomerService(Assistant):
    def __init__(self):
        super().__init__(
            system_message="你是XX公司客服，回答范围限于产品功能和订单查询",
            function_list=[product_lookup, order_status]
        )
    
    def _validate_query(self, query):
        if '价格' in query:
            return "具体产品价格请访问官网查询"
        return super()._validate_query(query)
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.

@register_tool('data_analysis')
class DataAnalyzer:
    def call(self, params):
        df = pd.read_csv(params['file'])
        return df.describe().to_markdown()


assistant = Assistant(function_list=['data_analysis', 'code_interpreter'])
1.
2.
3.
4.
5.
6.
7.
8.

# 开启调试模式
assistant = Assistant(verbose=True)


# 性能监控
from qwen_agent.monitor import perf_counter


@perf_counter
def critical_function():
    pass
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.