使用 Python 制作一个属于自己的 AI 搜索引擎

1. 使用到技术

OpenAI KEY
Serper KEY
Bing Search

2. 原理解析

使用Google和Bing的搜搜结果交由OpenAI处理并给出回答。

3. 代码实现

import requests
from lxml import etree
import os
from openai import OpenAI# 从环境变量中加载 API 密钥
os.environ["OPENAI_API_KEY"] = "sk-xxxx-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
os.environ["SERPER_API_KEY"] = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
# 确保在执行代码前已经设置了环境变量
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
SERPER_API_KEY = os.getenv("SERPER_API_KEY")def search_bing(query):headers = {'Referer': 'https://www.bing.com/','User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/127.0.0.0 Safari/537.36',}params = {'q': query,'mkt': 'zh-CN'}response = requests.get('https://www.bing.com/search', params=params, headers=headers)html = etree.HTML(response.text)li_list = html.xpath("//li[@class='b_algo']")result = []for index in range(len(li_list)):title = ";".join(li_list[index].xpath("./h2/a/text()"))link = li_list[index].xpath("./h2/a/@href")[0]snippet = ";".join(li_list[index].xpath("./div/p/text()"))position = indexprint(title, link, snippet, position)result.append({'title': title,'link': link,'snippet': snippet,'position': position,})return resultdef search_serper(query):"""使用Serper API进行搜索并返回结果。"""url = "https://google.serper.dev/search"headers = {"X-API-KEY": SERPER_API_KEY,"Content-Type": "application/json",}params = {'q': query,'gl': "cn",'hl': "zh-cn",}try:response = requests.post(url, headers=headers, json=params)response.raise_for_status()  # 检查HTTP请求状态serper_data = response.json()if not serper_data:return "无法获取搜索结果", []google_context = serper_data.get('organic', [])google_other = serper_data.get('relatedSearches', [])return google_context, google_otherexcept requests.RequestException as e:print(f"请求失败: {e}")return Nonedef search_openai(query, context):"""利用OpenAI API回答问题并引用相关上下文，并使用流的方式输出。"""context_template = ("你是GinLynn构建的大型语言AI助手。给你一个用户问题，请正确、简洁、准确的讲述这个问题的答案。""你会得到一组与问题相关的上下文，其中每个对象都是一个json字符串，""'snippet'字段表示片段，'title'字段表示标题，'link'字段表示链接，'position'字段表示位置。""请使用这些上下文并在每个句子的末尾引用上下文(如果适用)。""你的答案必须是正确、准确的，由专家以公正和专业的语气撰写。请限制为2048token。""不要给出任何与问题无关的信息，也不要重复。如果给定的上下文没有提供足够的信息，""那么在相关主题后面加上“information is missing on”。请以[position]的格式注明出处和参考编号。""以下是一组上下文:")client = OpenAI(api_key=OPENAI_API_KEY)try:completion = client.chat.completions.create(model="gpt-4o",messages=[{"role": "system", "content": context_template + context},{"role": "user", "content": query}],stream=True  # 启用流式响应)# 逐条打印流式输出的结果for chunk in completion:if chunk.choices[0].delta.content is not None:print(chunk.choices[0].delta.content, end="")print()  # 输出换行return "完成输出"except Exception as e:print(f"OpenAI API request failed: {e}")return "无法完成请求", []if __name__ == '__main__':query = input("请输入查询: ")if query.strip() == "":query = "最新俄乌局势信息"print("正在搜索...")serper_context, other_queries = search_serper(query)bing_context = search_bing(query)context = []if bing_context:context.extend(bing_context)if serper_context:# 为Serper上下文的每个条目重置 position 值，以防止重复for index, item in enumerate(serper_context, start=len(bing_context)):item['position'] = index  # 从当前Bing结果的数量开始context.extend(serper_context)print("搜索结果:", context)search_openai(query, str(context))if other_queries:print("相关搜索:", other_queries)