{"id":53730,"date":"2025-02-16T15:29:32","date_gmt":"2025-02-16T07:29:32","guid":{"rendered":"https:\/\/fwq.ai\/blog\/53730\/"},"modified":"2025-02-16T15:29:32","modified_gmt":"2025-02-16T07:29:32","slug":"%e7%94%a8%e7%bd%91%e7%ab%99%e5%86%85%e5%ae%b9%e6%9e%84%e5%bb%barag%e5%ba%94%e7%94%a8","status":"publish","type":"post","link":"https:\/\/fwq.ai\/blog\/53730\/","title":{"rendered":"\u7528\u7f51\u7ad9\u5185\u5bb9\u6784\u5efaRAG\u5e94\u7528"},"content":{"rendered":"

\u6700\u8fd1\uff0c\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08LLMs\uff09\u7684\u8fdb\u6b65\u4e3a\u590d\u6742\u7684\u81ea\u7136\u8bed\u8a00\u5e94\u7528\u89e3\u9501\u4e86\u4ee4\u4eba\u5174\u594b\u7684\u53ef\u80fd\u6027\u3002\u8fd9\u4e9b\u6a21\u578b\uff0c\u5982ChatGPT\u3001LLAMA\u548cMistral\uff0c\u6b63\u5728\u9769\u65b0\u6211\u4eec\u4e0eAI\u7684\u4e92\u52a8\u65b9\u5f0f\uff0c\u4ece\u751f\u6210\u7c7b\u4eba\u6587\u672c\u5230\u9a71\u52a8\u4e2a\u6027\u5316\u804a\u5929\u673a\u5668\u4eba\u3002\u7136\u800c\uff0c\u4e00\u4e2a\u4e3b\u8981\u7684\u9650\u5236\u4ecd\u7136\u5b58\u5728\uff1a\u8fd9\u4e9b\u6a21\u578b\u53d7\u9650\u4e8e\u5b83\u4eec\u8bad\u7ec3\u65f6\u7684\u77e5\u8bc6\uff0c\u5e76\u4e14\u65e0\u6cd5\u66f4\u65b0\u65b0\u7684\u4fe1\u606f\u3002\u8fd9\u79cd\u9650\u5236\u963b\u788d\u4e86\u5b83\u4eec\u5e94\u5bf9\u65f6\u95f4\u654f\u611f\u6216\u9886\u57df\u7279\u5b9a\u67e5\u8be2\u7684\u80fd\u529b\u3002<\/p>\n

\u8fd9\u5c31\u662f\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08RAG\uff09\u53d1\u6325\u4f5c\u7528\u7684\u5730\u65b9\u3002RAG\u4f7f\u6211\u4eec\u80fd\u591f\u5c06\u5b9e\u65f6\u4e0a\u4e0b\u6587\u4fe1\u606f\u8f93\u5165\u5230LLMs\u4e2d\uff0c\u4f7f\u5b83\u4eec\u80fd\u591f\u63d0\u4f9b\u66f4\u76f8\u5173\u548c\u7cbe\u786e\u7684\u7b54\u6848\u3002\u4e00\u4e2a\u6709\u4ef7\u503c\u7684\u4e0a\u4e0b\u6587\u4fe1\u606f\u6765\u6e90\u662f\u7f51\u7ad9\u5185\u5bb9\u3002<\/p>\n

\u5728\u8fd9\u7bc7\u6307\u5357\u4e2d\uff0c\u6211\u4eec\u5c06\u89e3\u91ca\u5982\u4f55\u4ece\u7f51\u7ad9\u63d0\u53d6\u5185\u5bb9\u5e76\u5229\u7528\u5b83\u6765\u6539\u8fdbLLMs\u5728RAG\u5e94\u7528\u7a0b\u5e8f\u4e2d\u7684\u54cd\u5e94\u3002\u6211\u4eec\u5c06\u6db5\u76d6\u4ece\u7f51\u7edc\u6293\u53d6\u7684\u57fa\u7840\u77e5\u8bc6\u5230\u5206\u5757\u7b56\u7565\u4ee5\u53ca\u521b\u5efa\u5411\u91cf\u5d4c\u5165\u4ee5\u5b9e\u73b0\u9ad8\u6548\u68c0\u7d22\u7684\u6240\u6709\u5185\u5bb9\u3002\u8ba9\u6211\u4eec\u5f00\u59cb\u5427\uff01<\/p>\n

1\u3001\u7f51\u7edc\u6293\u53d6\u57fa\u7840<\/h2>\n

\u4e3a\u4e86\u5c06\u7f51\u7ad9\u5185\u5bb9\u96c6\u6210\u5230RAG\u7cfb\u7edf\u4e2d\uff0c\u7b2c\u4e00\u6b65\u662f\u4ece\u7f51\u7ad9\u63d0\u53d6\u5185\u5bb9\u3002\u8fd9\u4e2a\u8fc7\u7a0b\u88ab\u79f0\u4e3a\u7f51\u7edc\u6293\u53d6\u3002\u867d\u7136\u4e00\u4e9b\u7f51\u7ad9\u63d0\u4f9b\u4e86\u8bbf\u95ee\u5176\u6570\u636e\u7684API\uff0c\u4f46\u8bb8\u591a\u6ca1\u6709\u3002\u5728\u8fd9\u79cd\u60c5\u51b5\u4e0b\uff0c\u7f51\u7edc\u6293\u53d6\u53d8\u5f97\u975e\u5e38\u6709\u4ef7\u503c\u3002<\/p>\n

\u6709\u51e0\u4e2a\u6d41\u884c\u7684Python\u5e93\u53ef\u4ee5\u5e2e\u52a9\u63d0\u53d6\u7f51\u9875\u6570\u636e\u3002\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528Beautiful Soup\u89e3\u6790HTML\u5185\u5bb9\u548crequests\u8fdb\u884cHTTP\u8bf7\u6c42\u3002\u8fd8\u53ef\u4ee5\u4f7f\u7528\u66f4\u9ad8\u7ea7\u7684\u5de5\u5177\uff0c\u5982Selenium\uff08\u7528\u4e8e\u52a8\u6001\u5185\u5bb9\uff09\u6216Scrapy\uff08\u7528\u4e8e\u5927\u89c4\u6a21\u6293\u53d6\uff09\u3002<\/p>\n

\n \u793a\u4f8b\uff1a\u6293\u53d6\u7ef4\u57fa\u767e\u79d1\n<\/p><\/blockquote>\n

\u8ba9\u6211\u4eec\u4ece\u4f7f\u7528BeautifulSoup\u6293\u53d6\u7ef4\u57fa\u767e\u79d1\u9875\u9762\u5f00\u59cb\u3002<\/p>\n

import requests  \nfrom bs4 import BeautifulSoup  \n  \n# \u5411\u7ef4\u57fa\u767e\u79d1\u7684\u6570\u636e\u79d1\u5b66\u9875\u9762\u53d1\u9001\u8bf7\u6c42  \nresponse = requests.get(  \n    url=\"https:\/\/en.wikipedia.org\/wiki\/Data_science\",  \n)  \n# \u89e3\u6790HTML\u5185\u5bb9  \nsoup = BeautifulSoup(response.content, 'html.parser')  \n# \u83b7\u53d6\u6587\u7ae0\u4e3b\u4f53\u5185\u7684\u6587\u672c\u5185\u5bb9  \ncontent = soup.find(id=\"bodyContent\")  \nprint(content.text)\n<\/code><\/pre>\n
\u8fd9\u6bb5\u4ee3\u7801\u5411\u7ef4\u57fa\u767e\u79d1\u53d1\u9001\u8bf7\u6c42\uff0c\u83b7\u53d6\u201c\u6570\u636e\u79d1\u5b66\u201d\u9875\u9762\u7684\u5185\u5bb9\uff0c\u5e76\u63d0\u53d6\u4e3b\u4f53\u6587\u672c\u4ee5\u4f9b\u8fdb\u4e00\u6b65\u5904\u7406\u3002<\/p>\n
2\u3001\u5206\u5757\uff1a\u5c06\u5185\u5bb9\u5206\u89e3<\/h2>\n
\u6210\u529f\u6293\u53d6\u4e00\u4e9b\u5185\u5bb9\u540e\uff0c\u4e0b\u4e00\u6b65\u662f\u5c06\u5176\u62c6\u5206\u4e3a\u5757\u3002\u5206\u5757\u6709\u51e0\u4e2a\u91cd\u8981\u539f\u56e0\uff1a<\/p>\n
    \n
  • \u7c92\u5ea6<\/strong>\uff1a\u5c06\u6587\u672c\u62c6\u5206\u4e3a\u8f83\u5c0f\u7684\u90e8\u5206\u4f7f\u5176\u66f4\u5bb9\u6613\u68c0\u7d22\u6700\u76f8\u5173\u7684\u4fe1\u606f\u3002<\/li>\n
  • \u63d0\u9ad8\u8bed\u4e49\u6027<\/strong>\uff1a\u5bf9\u6574\u4e2a\u6587\u6863\u4f7f\u7528\u5355\u4e2a\u5d4c\u5165\u4f1a\u5bfc\u81f4\u6709\u610f\u4e49\u7684\u4fe1\u606f\u4e22\u5931\u3002<\/li>\n
  • \u6548\u7387<\/strong>\uff1a\u8f83\u5c0f\u7684\u6587\u672c\u5757\u5728\u5d4c\u5165\u8fc7\u7a0b\u4e2d\u5bfc\u81f4\u66f4\u9ad8\u6548\u7684\u8ba1\u7b97\u3002<\/li>\n<\/ul>\n
    \n  \u56fa\u5b9a\u5927\u5c0f vs. \u4e0a\u4e0b\u6587\u611f\u77e5\u5206\u5757\n<\/p><\/blockquote>\n
    \u6700\u5e38\u89c1\u7684\u5206\u5757\u65b9\u6cd5\u662f\u56fa\u5b9a\u5927\u5c0f\u5206\u5757\u548c\u4e0a\u4e0b\u6587\u611f\u77e5\u5206\u5757\u3002\u56fa\u5b9a\u5927\u5c0f\u5206\u5757\u5728\u9884\u5b9a\u4e49\u7684\u65f6\u95f4\u95f4\u9694\u5185\u5206\u5272\u6587\u672c\uff0c\u800c\u4e0a\u4e0b\u6587\u611f\u77e5\u5206\u5757\u6839\u636e\u53e5\u5b50\u6216\u6bb5\u843d\u8fb9\u754c\u8c03\u6574\u5206\u5757\u5927\u5c0f\u3002<\/p>\n
    \u5728\u8fd9\u4e2a\u6307\u5357\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528LangChain\u6846\u67b6\u4e2d\u7684RecursiveCharacterTextSplitter<\/code>\u6267\u884c\u5206\u5757\uff0c\u786e\u4fdd\u5728\u6587\u672c\u7684\u903b\u8f91\u70b9\u4e0a\u8fdb\u884c\u62c6\u5206\u3002<\/p>\n
    from langchain.text_splitter import RecursiveCharacterTextSplitter  \n  \ntext_splitter = RecursiveCharacterTextSplitter(  \n    chunk_size=512,   # \u8bbe\u7f6e\u5757\u5927\u5c0f\u4e3a512\u5b57\u7b26  \n    length_function=len  \n)  \nchunked_text = text_splitter.split_text(content.text)\n<\/code><\/pre>\n
    \u8fd9\u6bb5\u4ee3\u7801\u5c06\u6293\u53d6\u7684\u6587\u672c\u62c6\u5206\u4e3a\u5927\u7ea6512\u4e2a\u5b57\u7b26\u7684\u5757\uff0c\u5e76\u6839\u636e\u81ea\u7136\u65ad\u70b9\u8c03\u6574\u62c6\u5206\u3002<\/p>\n
    3\u3001\u4ece\u5757\u5230\u5411\u91cf\u5d4c\u5165<\/h2>\n
    \u4e00\u65e6\u6211\u4eec\u6709\u4e86\u6587\u672c\u5757\uff0c\u4e0b\u4e00\u6b65\u5c31\u662f\u5c06\u5b83\u4eec\u8f6c\u6362\u4e3a\u5411\u91cf\u5d4c\u5165<\/strong>\u3002\u5d4c\u5165\u662f\u6587\u672c\u7684\u6570\u503c\u8868\u793a\uff0c\u6355\u6349\u5176\u8bed\u4e49\u610f\u4e49\uff0c\u5141\u8bb8\u9ad8\u6548\u7684\u76f8\u4f3c\u6027\u6bd4\u8f83\u3002<\/p>\n
    \n  \u5d4c\u5165\u7c7b\u578b\n<\/p><\/blockquote>\n
    \u6709\u4e24\u79cd\u4e3b\u8981\u7c7b\u578b\u7684\u5d4c\u5165\uff1a<\/p>\n
      \n
    • \u5bc6\u96c6\u5d4c\u5165<\/strong>\uff1a\u7531OpenAI\u6216Sentence Transformers\u7b49\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u751f\u6210\u3002\u5b83\u4eec\u5f88\u597d\u5730\u7f16\u7801\u8bed\u4e49\u76f8\u4f3c\u6027\u3002<\/li>\n
    • \u7a00\u758f\u5d4c\u5165<\/strong>\uff1a\u7531TF-IDF\u6216BM25\u7b49\u7ecf\u5178\u65b9\u6cd5\u751f\u6210\u3002\u5b83\u4eec\u5bf9\u4e8e\u57fa\u4e8e\u5173\u952e\u5b57\u7684\u76f8\u4f3c\u6027\u6709\u6548\u3002<\/li>\n<\/ul>\n
      \u5bf9\u4e8e\u6211\u4eec\u7684RAG\u5e94\u7528\u7a0b\u5e8f\uff0c\u6211\u4eec\u5c06\u4f7f\u7528Sentence Transformers\u751f\u6210\u7684\u5bc6\u96c6\u5d4c\u5165<\/strong>\uff0c\u5373all-MiniLM-L6-v2<\/code>\u6a21\u578b\u3002<\/p>\n
      from langchain.embeddings import SentenceTransformerEmbeddings  \n  \n# \u52a0\u8f7d\u7528\u4e8e\u751f\u6210\u5d4c\u5165\u7684\u6a21\u578b  \nembeddings = SentenceTransformerEmbeddings(model_name=\"all-MiniLM-L6-v2\")  \n# \u4e3a\u7b2c\u4e09\u4e2a\u6587\u672c\u5757\u521b\u5efa\u5d4c\u5165  \nchunk_embedding = embeddings.embed_documents([chunked_text[3]])\n<\/code><\/pre>\n
      \u8fd9\u6bb5\u4ee3\u7801\u4f7f\u7528MiniLM-L6-v2\u6a21\u578b\u5c06\u5176\u4e2d\u4e00\u4e2a\u5757\u8f6c\u6362\u4e3a\u5bc6\u96c6\u5d4c\u5165\u3002\u5b9e\u9645\u4e0a\uff0c\u60a8\u9700\u8981\u4e3a\u6240\u6709\u5757\u751f\u6210\u5d4c\u5165\u3002<\/p>\n
      4\u3001\u4f7f\u7528Milvus\u5b58\u50a8\u548c\u68c0\u7d22\u5d4c\u5165<\/h2>\n
      \u751f\u6210\u5d4c\u5165\u540e\uff0c\u6211\u4eec\u9700\u8981\u5c06\u5b83\u4eec\u5b58\u50a8\u5728\u5411\u91cf\u6570\u636e\u5e93<\/strong>\u4e2d\u4ee5\u5b9e\u73b0\u9ad8\u6548\u68c0\u7d22\u3002Milvus\u662f\u4e00\u4e2a\u5f00\u6e90\u5411\u91cf\u6570\u636e\u5e93\uff0c\u4e13\u95e8\u7528\u4e8e\u5b58\u50a8\u548c\u641c\u7d22\u5d4c\u5165\u3002\u5b83\u4e0eLangChain\u96c6\u6210\u826f\u597d\uff0c\u662fRAG\u5e94\u7528\u7a0b\u5e8f\u7684\u7406\u60f3\u9009\u62e9\u3002<\/p>\n
      \u4ee5\u4e0b\u662f\u5c06\u5757\u5d4c\u5165\u5b58\u50a8\u5230Milvus\u7684\u65b9\u6cd5\uff1a<\/p>\n
      from langchain.vectorstores.milvus import Milvus  \n  \n# \u5c06\u5d4c\u5165\u5b58\u50a8\u5230Milvus  \nvector_db = Milvus.from_texts(texts=chunked_text, embedding=embeddings, collection_name=\"rag_milvus\")\n<\/code><\/pre>\n
      \u8fd9\u6bb5\u4ee3\u7801\u5728Milvus\u4e2d\u521b\u5efa\u4e00\u4e2a\u96c6\u5408\u5e76\u5b58\u50a8\u6240\u6709\u5757\u5d4c\u5165\u4ee5\u4fbf\u5c06\u6765\u68c0\u7d22\u3002<\/p>\n
      5\u3001\u6784\u5efaRAG\u7ba1\u9053<\/h2>\n
      \u6709\u4e86\u5757\u5b58\u50a8\u548c\u5d4c\u5165\u51c6\u5907\u5c31\u7eea\uff0c\u73b0\u5728\u662f\u65f6\u5019\u6784\u5efa\u6211\u4eec\u7684RAG\u7ba1\u9053\u4e86\u3002\u8be5\u7ba1\u9053\u5c06\u6839\u636e\u7528\u6237\u67e5\u8be2\u68c0\u7d22\u6700\u76f8\u5173\u7684\u5d4c\u5165\uff0c\u5e76\u5c06\u5176\u4f20\u9012\u7ed9LLM\u4ee5\u751f\u6210\u54cd\u5e94\u3002<\/p>\n
      \u7b2c\u4e00\u6b65\uff1a\u8bbe\u7f6e\u68c0\u7d22\u5668<\/p>\n
      \u9996\u5148\uff0c\u6211\u4eec\u9700\u8981\u8bbe\u7f6e\u4e00\u4e2a\u68c0\u7d22\u5668<\/strong>\uff0c\u8be5\u68c0\u7d22\u5668\u6839\u636e\u7528\u6237\u7684\u67e5\u8be2\u4ece\u5411\u91cf\u6570\u636e\u5e93\u4e2d\u68c0\u7d22\u6700\u76f8\u5173\u7684\u5d4c\u5165\u3002<\/p>\n
      retriever = vector_db.as_retriever()\n<\/code><\/pre>\n
      \u7b2c\u4e8c\u6b65\uff1a\u521d\u59cb\u5316LLM<\/p>\n
      \u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u4f7f\u7528OpenAI\u7684GPT-3.5-turbo\u521d\u59cb\u5316\u6211\u4eec\u7684\u8bed\u8a00\u6a21\u578b\uff1a<\/p>\n
      from langchain_openai import ChatOpenAI  \n  \nllm = ChatOpenAI(model=\"gpt-3.5-turbo-0125\")\n<\/code><\/pre>\n
      \u7b2c\u4e09\u6b65\uff1a\u5b9a\u4e49\u81ea\u5b9a\u4e49\u63d0\u793a<\/p>\n
      \u6211\u4eec\u9700\u8981\u521b\u5efa\u4e00\u4e2a\u63d0\u793a\u6a21\u677f\uff0c\u8be5\u6a21\u677f\u5c06\u6307\u5bfcLLM\u6839\u636e\u68c0\u7d22\u5230\u7684\u5185\u5bb9\u751f\u6210\u9002\u5f53\u7684\u7b54\u6848\u3002<\/p>\n
      from langchain_core.prompts import PromptTemplate  \n  \ntemplate = \"\"\"\u4f7f\u7528\u4ee5\u4e0b\u90e8\u5206\u4e0a\u4e0b\u6587\u56de\u7b54\u95ee\u9898\u3002  \n\u5982\u679c\u4f60\u4e0d\u77e5\u9053\u7b54\u6848\uff0c\u5c31\u8bf4\u4f60\u4e0d\u77e5\u9053\uff0c\u4e0d\u8981\u8bd5\u56fe\u7f16\u9020\u7b54\u6848\u3002  \n\u6700\u591a\u4f7f\u7528\u4e09\u53e5\u8bdd\uff0c\u5e76\u5c3d\u53ef\u80fd\u7b80\u6d01\u5730\u56de\u7b54\u3002  \n\u7b54\u6848\u7ed3\u5c3e\u603b\u662f\u8bf4\u201c\u611f\u8c22\u63d0\u95ee\uff01\u201d  \n{context}  \n\u95ee\u9898\uff1a{question}  \n\u6709\u7528\u7684\u7b54\u6848:\"\"\"  \ncustom_rag_prompt = PromptTemplate.from_template(template)\n<\/code><\/pre>\n
      \u7b2c\u56db\u6b65\uff1a\u6784\u5efaRAG\u94fe<\/p>\n
      \u6700\u540e\uff0c\u6211\u4eec\u5c06\u521b\u5efaRAG\u94fe<\/strong>\uff0c\u8be5\u94fe\u5c06\u68c0\u7d22\u6700\u76f8\u5173\u7684\u5757\uff0c\u5c06\u5176\u4f20\u9012\u7ed9LLM\uff0c\u5e76\u8f93\u51fa\u751f\u6210\u7684\u54cd\u5e94\u3002<\/p>\n
      from langchain_core.runnables import RunnablePassthrough  \nfrom langchain_core.output_parsers import StrOutputParser  \n  \ndef format_docs(docs):  \n    return \"\\n\\n\".join(doc.page_content for doc in docs)  \nrag_chain = (  \n    {\"context\": retriever | format_docs, \"question\": RunnablePassthrough()}  \n    | custom_rag_prompt  \n    | llm  \n    | StrOutputParser()  \n)\n<\/code><\/pre>\n
      \u8bbe\u7f6e\u4e86RAG\u94fe\u540e\uff0c\u4f60\u53ef\u4ee5\u5411\u7ba1\u9053\u53d1\u9001\u67e5\u8be2\u5e76\u63a5\u6536\u57fa\u4e8e\u7f51\u7ad9\u5185\u5bb9\u7684\u7b54\u6848\u3002<\/p>\n
      for chunk in rag_chain.stream(\"\u4ec0\u4e48\u662f\u6570\u636e\u79d1\u5b66\u5bb6\uff1f\"):  \n    print(chunk, end=\"\", flush=True)\n<\/code><\/pre>\n
      6\u3001\u7ed3\u675f\u8bed<\/h2>\n
      \u5728\u8fd9\u7bc7\u6307\u5357\u4e2d\uff0c\u6211\u4eec\u4ecb\u7ecd\u4e86\u5982\u4f55\u4ece\u7f51\u7ad9\u63d0\u53d6\u5185\u5bb9\u5e76\u4f7f\u7528\u5b83\u6765\u6539\u8fdbLLMs\u5728RAG\u5e94\u7528\u7a0b\u5e8f\u4e2d\u7684\u54cd\u5e94\u3002\u6211\u4eec\u8ba8\u8bba\u4e86\u7f51\u7edc\u6293\u53d6\u3001\u6587\u672c\u5206\u5757\u3001\u751f\u6210\u5411\u91cf\u5d4c\u5165\u4ee5\u53ca\u5728\u5411\u91cf\u6570\u636e\u5e93\uff08\u5982Milvus\uff09\u4e2d\u5b58\u50a8\u8fd9\u4e9b\u5d4c\u5165\u7684\u8fc7\u7a0b\u3002<\/p>\n
      \u901a\u8fc7\u4f7f\u7528\u8fd9\u79cd\u65b9\u6cd5\uff0c\u60a8\u53ef\u4ee5\u5f00\u53d1\u51fa\u66f4\u77e5\u60c5\u548c\u4e0a\u4e0b\u6587\u611f\u77e5\u7684AI\u5e94\u7528\u7a0b\u5e8f\u3002\u65e0\u8bba\u662f\u521b\u5efa\u804a\u5929\u673a\u5668\u4eba\u8fd8\u662f\u95ee\u7b54\u7cfb\u7edf\uff0cRAG\u90fd\u80fd\u63d0\u5347\u751f\u6210\u54cd\u5e94\u7684\u76f8\u5173\u6027\u548c\u51c6\u786e\u6027\u3002<\/p>\n
      \u91cd\u8981\u7684\u662f\u8981\u8bb0\u4f4f\uff0c\u60a8\u7684RAG\u7ba1\u9053\u7684\u6210\u529f\u53d6\u51b3\u4e8e\u6570\u636e\u7684\u8d28\u91cf\u4ee5\u53ca\u60a8\u5982\u4f55\u7ec4\u7ec7\u5757\u3001\u5d4c\u5165\u548c\u68c0\u7d22\u8fc7\u7a0b\u3002\u5c1d\u8bd5\u4e0d\u540c\u7684\u6a21\u578b\u3001\u5206\u5757\u5927\u5c0f\u548c\u68c0\u7d22\u65b9\u6cd5\u4ee5\u4f18\u5316\u60a8\u7684\u7cfb\u7edf\u3002<\/p>\n
      \n
      \n","protected":false},"excerpt":{"rendered":"
      \u6700\u8fd1\uff0c\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08LLMs\uff09\u7684\u8fdb\u6b65\u4e3a\u590d\u6742\u7684\u81ea\u7136\u8bed\u8a00\u5e94\u7528\u89e3\u9501\u4e86\u4ee4\u4eba\u5174\u594b\u7684\u53ef\u80fd\u6027\u3002\u8fd9\u4e9b\u6a21\u578b\uff0c\u5982ChatGPT\u3001LLAMA\u548cMistral\uff0c\u6b63\u5728\u9769\u65b0\u6211\u4eec\u4e0eAI\u7684\u4e92\u52a8\u65b9\u5f0f\uff0c\u4ece\u751f\u6210\u7c7b\u4eba\u6587\u672c\u5230\u9a71\u52a8\u4e2a\u6027\u5316\u804a\u5929\u673a\u5668\u4eba\u3002\u7136\u800c\uff0c\u4e00\u4e2a\u4e3b\u8981\u7684\u9650\u5236\u4ecd\u7136\u5b58\u5728\uff1a\u8fd9\u4e9b\u6a21\u578b\u53d7\u9650\u4e8e\u5b83\u4eec\u8bad\u7ec3\u65f6\u7684\u77e5\u8bc6\uff0c\u5e76\u4e14\u65e0\u6cd5\u66f4\u65b0\u65b0\u7684\u4fe1\u606f\u3002\u8fd9\u79cd\u9650\u5236\u963b\u788d\u4e86\u5b83\u4eec\u5e94\u5bf9\u65f6\u95f4\u654f\u611f\u6216\u9886\u57df\u7279\u5b9a\u67e5\u8be2\u7684\u80fd\u529b\u3002 \u8fd9\u5c31\u662f\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08RAG\uff09\u53d1\u6325\u4f5c\u7528\u7684\u5730\u65b9\u3002RAG\u4f7f\u6211\u4eec\u80fd\u591f\u5c06\u5b9e\u65f6\u4e0a\u4e0b\u6587\u4fe1\u606f\u8f93\u5165\u5230LLMs\u4e2d\uff0c\u4f7f\u5b83\u4eec\u80fd\u591f\u63d0\u4f9b\u66f4\u76f8\u5173\u548c\u7cbe\u786e\u7684\u7b54\u6848\u3002\u4e00\u4e2a\u6709\u4ef7\u503c\u7684\u4e0a\u4e0b\u6587\u4fe1\u606f\u6765\u6e90\u662f\u7f51\u7ad9\u5185\u5bb9\u3002 \u5728\u8fd9\u7bc7\u6307\u5357\u4e2d\uff0c\u6211\u4eec\u5c06\u89e3\u91ca\u5982\u4f55\u4ece\u7f51\u7ad9\u63d0\u53d6\u5185\u5bb9\u5e76\u5229\u7528\u5b83\u6765\u6539\u8fdbLLMs\u5728RAG\u5e94\u7528\u7a0b\u5e8f\u4e2d\u7684\u54cd\u5e94\u3002\u6211\u4eec\u5c06\u6db5\u76d6\u4ece\u7f51\u7edc\u6293\u53d6\u7684\u57fa\u7840\u77e5\u8bc6\u5230\u5206\u5757\u7b56\u7565\u4ee5\u53ca\u521b\u5efa\u5411\u91cf\u5d4c\u5165\u4ee5\u5b9e\u73b0\u9ad8\u6548\u68c0\u7d22\u7684\u6240\u6709\u5185\u5bb9\u3002\u8ba9\u6211\u4eec\u5f00\u59cb\u5427\uff01 1\u3001\u7f51\u7edc\u6293\u53d6\u57fa\u7840 \u4e3a\u4e86\u5c06\u7f51\u7ad9\u5185\u5bb9\u96c6\u6210\u5230RAG\u7cfb\u7edf\u4e2d\uff0c\u7b2c\u4e00\u6b65\u662f\u4ece\u7f51\u7ad9\u63d0\u53d6\u5185\u5bb9\u3002\u8fd9\u4e2a\u8fc7\u7a0b\u88ab\u79f0\u4e3a\u7f51\u7edc\u6293\u53d6\u3002\u867d\u7136\u4e00\u4e9b\u7f51\u7ad9\u63d0\u4f9b\u4e86\u8bbf\u95ee\u5176\u6570\u636e\u7684API\uff0c\u4f46\u8bb8\u591a\u6ca1\u6709\u3002\u5728\u8fd9\u79cd\u60c5\u51b5\u4e0b\uff0c\u7f51\u7edc\u6293\u53d6\u53d8\u5f97\u975e\u5e38\u6709\u4ef7\u503c\u3002 \u6709\u51e0\u4e2a\u6d41\u884c\u7684Python\u5e93\u53ef\u4ee5\u5e2e\u52a9\u63d0\u53d6\u7f51\u9875\u6570\u636e\u3002\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528Beautiful Soup\u89e3\u6790HTML\u5185\u5bb9\u548crequests\u8fdb\u884cHTTP\u8bf7\u6c42\u3002\u8fd8\u53ef\u4ee5\u4f7f\u7528\u66f4\u9ad8\u7ea7\u7684\u5de5\u5177\uff0c\u5982Selenium\uff08\u7528\u4e8e\u52a8\u6001\u5185\u5bb9\uff09\u6216Scrapy\uff08\u7528\u4e8e\u5927\u89c4\u6a21\u6293\u53d6\uff09\u3002 \u793a\u4f8b\uff1a\u6293\u53d6\u7ef4\u57fa\u767e\u79d1 \u8ba9\u6211\u4eec\u4ece\u4f7f\u7528BeautifulSoup\u6293\u53d6\u7ef4\u57fa\u767e\u79d1\u9875\u9762\u5f00\u59cb\u3002 import requests from bs4 import BeautifulSoup # \u5411\u7ef4\u57fa\u767e\u79d1\u7684\u6570\u636e\u79d1\u5b66\u9875\u9762\u53d1\u9001\u8bf7\u6c42 response = requests.get( url=”https:\/\/en.wikipedia.org\/wiki\/Data_science”, ) # \u89e3\u6790HTML\u5185\u5bb9 soup = BeautifulSoup(response.content, ‘html.parser’) # \u83b7\u53d6\u6587\u7ae0\u4e3b\u4f53\u5185\u7684\u6587\u672c\u5185\u5bb9 content = soup.find(id=”bodyContent”) print(content.text) \u8fd9\u6bb5\u4ee3\u7801\u5411\u7ef4\u57fa\u767e\u79d1\u53d1\u9001\u8bf7\u6c42\uff0c\u83b7\u53d6\u201c\u6570\u636e\u79d1\u5b66\u201d\u9875\u9762\u7684\u5185\u5bb9\uff0c\u5e76\u63d0\u53d6\u4e3b\u4f53\u6587\u672c\u4ee5\u4f9b\u8fdb\u4e00\u6b65\u5904\u7406\u3002 2\u3001\u5206\u5757\uff1a\u5c06\u5185\u5bb9\u5206\u89e3 \u6210\u529f\u6293\u53d6\u4e00\u4e9b\u5185\u5bb9\u540e\uff0c\u4e0b\u4e00\u6b65\u662f\u5c06\u5176\u62c6\u5206\u4e3a\u5757\u3002\u5206\u5757\u6709\u51e0\u4e2a\u91cd\u8981\u539f\u56e0\uff1a \u7c92\u5ea6\uff1a\u5c06\u6587\u672c\u62c6\u5206\u4e3a\u8f83\u5c0f\u7684\u90e8\u5206\u4f7f\u5176\u66f4\u5bb9\u6613\u68c0\u7d22\u6700\u76f8\u5173\u7684\u4fe1\u606f\u3002 \u63d0\u9ad8\u8bed\u4e49\u6027\uff1a\u5bf9\u6574\u4e2a\u6587\u6863\u4f7f\u7528\u5355\u4e2a\u5d4c\u5165\u4f1a\u5bfc\u81f4\u6709\u610f\u4e49\u7684\u4fe1\u606f\u4e22\u5931\u3002 \u6548\u7387\uff1a\u8f83\u5c0f\u7684\u6587\u672c\u5757\u5728\u5d4c\u5165\u8fc7\u7a0b\u4e2d\u5bfc\u81f4\u66f4\u9ad8\u6548\u7684\u8ba1\u7b97\u3002 \u56fa\u5b9a\u5927\u5c0f vs. \u4e0a\u4e0b\u6587\u611f\u77e5\u5206\u5757 \u6700\u5e38\u89c1\u7684\u5206\u5757\u65b9\u6cd5\u662f\u56fa\u5b9a\u5927\u5c0f\u5206\u5757\u548c\u4e0a\u4e0b\u6587\u611f\u77e5\u5206\u5757\u3002\u56fa\u5b9a\u5927\u5c0f\u5206\u5757\u5728\u9884\u5b9a\u4e49\u7684\u65f6\u95f4\u95f4\u9694\u5185\u5206\u5272\u6587\u672c\uff0c\u800c\u4e0a\u4e0b\u6587\u611f\u77e5\u5206\u5757\u6839\u636e\u53e5\u5b50\u6216\u6bb5\u843d\u8fb9\u754c\u8c03\u6574\u5206\u5757\u5927\u5c0f\u3002 \u5728\u8fd9\u4e2a\u6307\u5357\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528LangChain\u6846\u67b6\u4e2d\u7684RecursiveCharacterTextSplitter\u6267\u884c\u5206\u5757\uff0c\u786e\u4fdd\u5728\u6587\u672c\u7684\u903b\u8f91\u70b9\u4e0a\u8fdb\u884c\u62c6\u5206\u3002 from langchain.text_splitter import RecursiveCharacterTextSplitter text_splitter = RecursiveCharacterTextSplitter( chunk_size=512, # \u8bbe\u7f6e\u5757\u5927\u5c0f\u4e3a512\u5b57\u7b26 […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[],"class_list":["post-53730","post","type-post","status-publish","format-standard","hentry","category-ai"],"_links":{"self":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/53730","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/comments?post=53730"}],"version-history":[{"count":0,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/53730\/revisions"}],"wp:attachment":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/media?parent=53730"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/categories?post=53730"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/tags?post=53730"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}