构建AI智能体：大模型思维链提示工程：技术原理与行业应用案例分析

import requests
import json
from typing import List, Dict
import os
 
class QwenChainOfThought:
    def __init__(self, api_key: str, base_url: str = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation"):
        self.api_key = api_key
        self.base_url = base_url
        self.headers = {
            "Authorization": f"Bearer {api_key}",
            "Content-Type": "application/json"
        }
    
    def generate_response(self, prompt: str, max_tokens: int = 1000) -> str:
        """调用Qwen API生成响应"""
        data = {
            "model": "qwen-max",
            "input": {
                "messages": [
                    {
                        "role": "user",
                        "content": prompt
                    }
                ]
            },
            "parameters": {
                "max_tokens": max_tokens,
                "temperature": 0.1  # 低温度值保证推理稳定性
            }
        }
        
        try:
            response = requests.post(self.base_url, headers=self.headers, json=data)
            result = response.json()
            # print(result)
            # return result['output']['choices'][0]['message']['content']
            return result['output']['text']
        except Exception as e:
            return f"API调用错误: {str(e)}"
    
    def math_problem_solver(self, problem: str) -> Dict:
        """数学问题思维链求解"""
        prompt = f"""请用思维链方式解决以下数学问题，展示完整的推理过程：
问题：{problem}
请按以下格式回答：
1. 理解问题：[简要说明问题要求]
2. 分解步骤：[将问题分解为多个求解步骤]
3. 逐步计算：[展示每一步的具体计算]
4. 最终答案：[给出最终结果]
推理过程："""
        
        reasoning_process = self.generate_response(prompt)
        return {
            "problem": problem,
            "reasoning_process": reasoning_process,
            "prompt_used": prompt
        }
    
 
# 使用示例
def main():
    # 初始化Qwen客户端
    qwen_client = QwenChainOfThought(api_key=os.environ.get("DASHSCOPE_API_KEY"))
    
    # 数学问题示例
    math_result = qwen_client.math_problem_solver(
        "一个水池有进水管和出水管，进水管单独注满需要6小时，出水管单独排空需要8小时。如果同时打开进水管和出水管，需要多少小时才能注满水池？"
    )
    print("数学问题求解结果：")
    print(json.dumps(math_result, ensure_ascii=False, indent=2))
 
if __name__ == "__main__":
    main()

import requests
import json
from typing import List, Dict
import os
 
class QwenChainOfThought:
    def __init__(self, api_key: str, base_url: str = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation"):
        self.api_key = api_key
        self.base_url = base_url
        self.headers = {
            "Authorization": f"Bearer {api_key}",
            "Content-Type": "application/json"
        }
    
    def generate_response(self, prompt: str, max_tokens: int = 1000) -> str:
        """调用Qwen API生成响应"""
        data = {
            "model": "qwen-max",
            "input": {
                "messages": [
                    {
                        "role": "user",
                        "content": prompt
                    }
                ]
            },
            "parameters": {
                "max_tokens": max_tokens,
                "temperature": 0.1  # 低温度值保证推理稳定性
            }
        }
        
        try:
            response = requests.post(self.base_url, headers=self.headers, json=data)
            result = response.json()
            # print(result)
            # return result['output']['choices'][0]['message']['content']
            return result['output']['text']
        except Exception as e:
            return f"API调用错误: {str(e)}"
    
    def logical_reasoning(self, premise: str, question: str) -> Dict:
        """逻辑推理问题思维链求解"""
        prompt = f"""请基于以下前提进行逻辑推理，展示完整的思维链：
前提：{premise}
问题：{question}
推理步骤：
1. 分析前提条件：[提取关键信息]
2. 建立逻辑关系：[分析元素间的逻辑联系]
3. 逐步推导：[展示推理链条]
4. 得出结论：[回答原始问题]
请开始推理："""
        
        reasoning = self.generate_response(prompt)
        return {
            "premise": premise,
            "question": question,
            "logical_reasoning": reasoning
        }
    
# 使用示例
def main():
    # 初始化Qwen客户端
    qwen_client = QwenChainOfThought(api_key=os.environ.get("DASHSCOPE_API_KEY"))
    
    # 逻辑推理示例
    logic_result = qwen_client.logical_reasoning(
        premise="所有哺乳动物都有脊椎，鲸鱼是哺乳动物，鱼类不是哺乳动物",
        question="鲸鱼有脊椎吗？"
    )
    print("\n逻辑推理结果：")
    print(json.dumps(logic_result, ensure_ascii=False, indent=2))
 
if __name__ == "__main__":
    main()

import requests
import json
from typing import List, Dict
import os
 
class QwenChainOfThought:
    def __init__(self, api_key: str, base_url: str = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation"):
        self.api_key = api_key
        self.base_url = base_url
        self.headers = {
            "Authorization": f"Bearer {api_key}",
            "Content-Type": "application/json"
        }
    
    def generate_response(self, prompt: str, max_tokens: int = 1000) -> str:
        """调用Qwen API生成响应"""
        data = {
            "model": "qwen-max",
            "input": {
                "messages": [
                    {
                        "role": "user",
                        "content": prompt
                    }
                ]
            },
            "parameters": {
                "max_tokens": max_tokens,
                "temperature": 0.1  # 低温度值保证推理稳定性
            }
        }
        
        try:
            response = requests.post(self.base_url, headers=self.headers, json=data)
            result = response.json()
            # print(result)
            # return result['output']['choices'][0]['message']['content']
            return result['output']['text']
        except Exception as e:
            return f"API调用错误: {str(e)}"
    
    def math_problem_solver(self, problem: str) -> Dict:
        """数学问题思维链求解"""
        prompt = f"""请用思维链方式解决以下数学问题，展示完整的推理过程：
问题：{problem}
请按以下格式回答：
1. 理解问题：[简要说明问题要求]
2. 分解步骤：[将问题分解为多个求解步骤]
3. 逐步计算：[展示每一步的具体计算]
4. 最终答案：[给出最终结果]
推理过程："""
        
        reasoning_process = self.generate_response(prompt)
        return {
            "problem": problem,
            "reasoning_process": reasoning_process,
            "prompt_used": prompt
        }
    
    def logical_reasoning(self, premise: str, question: str) -> Dict:
        """逻辑推理问题思维链求解"""
        prompt = f"""请基于以下前提进行逻辑推理，展示完整的思维链：
前提：{premise}
问题：{question}
推理步骤：
1. 分析前提条件：[提取关键信息]
2. 建立逻辑关系：[分析元素间的逻辑联系]
3. 逐步推导：[展示推理链条]
4. 得出结论：[回答原始问题]
请开始推理："""
        
        reasoning = self.generate_response(prompt)
        return {
            "premise": premise,
            "question": question,
            "logical_reasoning": reasoning
        }
    
 
class AdvancedChainOfThought(QwenChainOfThought):
    def multi_step_reasoning(self, complex_problem: str) -> Dict:
        """处理需要多轮推理的复杂问题"""
        # 第一轮：问题分析与规划
        analysis_prompt = f"""请分析以下复杂问题并制定解决计划：
问题：{complex_problem}
请输出：
1. 问题类型识别：[判断问题属于哪类推理任务]
2. 关键信息提取：[列出问题中的关键数据和条件]
3. 解决策略：[规划大致的解决步骤]
4. 潜在难点：[预测可能遇到的困难]
分析结果："""
        
        analysis = self.generate_response(analysis_prompt)
        
        # 第二轮：详细推理
        reasoning_prompt = f"""基于以下分析，请进行详细推理：
问题：{complex_problem}
初步分析：{analysis}
现在请执行详细的逐步推理："""
        
        detailed_reasoning = self.generate_response(reasoning_prompt)
        
        return {
            "complex_problem": complex_problem,
            "problem_analysis": analysis,
            "detailed_reasoning": detailed_reasoning
        }
 
# 使用示例
def main():
    # 初始化Qwen客户端
    qwen_client = QwenChainOfThought(api_key=os.environ.get("DASHSCOPE_API_KEY"))
 
    # 复杂问题示例
    advanced_client = AdvancedChainOfThought(api_key=os.environ.get("DASHSCOPE_API_KEY"))
    complex_result = advanced_client.multi_step_reasoning(
        "某公司有A、B、C三个部门，A部门人数是B部门的2倍，C部门人数比A部门少20人。"
        "如果三个部门总人数是280人，且B部门女性占40%，A部门女性占60%，C部门男女各半，"
        "请问全公司女性总人数是多少？"
    )
    print("复杂问题求解结果：")
    print(json.dumps(complex_result, ensure_ascii=False, indent=2)) 
 
if __name__ == "__main__":
    main()