{"id":115255,"date":"2026-10-11T16:40:55","date_gmt":"2026-10-11T08:40:55","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/115255.html"},"modified":"2026-10-11T16:40:55","modified_gmt":"2026-10-11T08:40:55","slug":"%e7%94%a8-java-%e6%89%8b%e5%86%99%e4%b8%80%e4%b8%aa%e6%9c%80%e5%b0%8f-rag%ef%bc%9a%e4%bb%8e%e6%96%87%e6%a1%a3%e5%88%87%e5%88%86%e3%80%81%e5%90%91%e9%87%8f%e6%a3%80%e7%b4%a2%e5%88%b0-prompt-%e6%8b%bc","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/115255.html","title":{"rendered":"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def"},"content":{"rendered":"<h2>\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG&#xff1a;\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def<\/h2>\n<p>\u73af\u5883&#xff1a;Windows 11\u3001JDK 17.0.12\u3001Maven 3.9.14<br \/>\n\u4f9d\u8d56&#xff1a;Spring AI 2.0.1&#xff08;\u5d4c\u5165\u6a21\u578b\u8c03\u7528&#xff09;\u3001H2 \u672c\u5730\u5411\u91cf\u5b58\u50a8&#xff08;\u53ef\u9009&#xff0c;\u672c\u6587\u7528\u5185\u5b58 List \u5b9e\u73b0&#xff09;<br \/>\n\u8bf4\u660e&#xff1a;\u672c\u6587\u91cd\u70b9\u662f\u628a RAG \u7684\u6bcf\u4e2a\u73af\u8282\u62c6\u5f00\u8bb2\u6e05\u695a&#xff0c;\u4ee3\u7801\u53ef\u76f4\u63a5\u8dd1\u901a\u6700\u5c0f\u94fe\u8def&#xff1b;\u751f\u4ea7\u73af\u5883\u4ecd\u9700\u63a5\u5165\u771f\u5b9e\u5411\u91cf\u6570\u636e\u5e93\u548c\u91cd\u6392\u7b56\u7565\u3002<\/p>\n<p>\u76ee\u5f55<\/p>\n<ul>\n<li>RAG \u89e3\u51b3\u7684\u4e0d\u662f&#034;\u6a21\u578b\u4e0d\u591f\u806a\u660e&#034;&#xff0c;\u800c\u662f&#034;\u77e5\u8bc6\u4e0d\u5728\u53c2\u6570\u91cc&#034;<\/li>\n<li>\u6700\u5c0f RAG \u7684\u4e94\u4e2a\u73af\u8282&#xff1a;\u5207\u5206\u3001\u5d4c\u5165\u3001\u5b58\u50a8\u3001\u68c0\u7d22\u3001\u751f\u6210<\/li>\n<li>\u7b2c\u4e00\u6b65&#xff1a;\u6587\u6863\u5207\u5206&#xff0c;\u4e0d\u662f\u8d8a\u7ec6\u8d8a\u597d<\/li>\n<li>\u7b2c\u4e8c\u6b65&#xff1a;\u6587\u672c\u5d4c\u5165&#xff0c;\u628a\u8bed\u4e49\u538b\u7f29\u6210\u5411\u91cf<\/li>\n<li>\u7b2c\u4e09\u6b65&#xff1a;\u5411\u91cf\u5b58\u50a8&#xff0c;\u4e3a\u4ec0\u4e48\u4e0d\u80fd\u53ea\u7528 HashMap<\/li>\n<li>\u7b2c\u56db\u6b65&#xff1a;\u76f8\u4f3c\u5ea6\u68c0\u7d22&#xff0c;Top-K \u53ea\u662f\u5f00\u59cb<\/li>\n<li>\u7b2c\u4e94\u6b65&#xff1a;Prompt \u62fc\u63a5&#xff0c;\u628a\u68c0\u7d22\u7ed3\u679c\u5582\u7ed9\u6a21\u578b<\/li>\n<li>\u8dd1\u901a\u6700\u5c0f\u94fe\u8def&#xff1a;\u4e00\u4e2a\u53ef\u8fd0\u884c\u7684 Java \u793a\u4f8b<\/li>\n<li>\u5e38\u89c1\u7684\u5751\u548c\u4f18\u5316\u65b9\u5411<\/li>\n<li>\u5c0f\u7ed3<\/li>\n<\/ul>\n<h3>RAG \u89e3\u51b3\u7684\u4e0d\u662f&#034;\u6a21\u578b\u4e0d\u591f\u806a\u660e&#034;&#xff0c;\u800c\u662f&#034;\u77e5\u8bc6\u4e0d\u5728\u53c2\u6570\u91cc&#034;<\/h3>\n<p>\u5f88\u591a\u4eba\u7b2c\u4e00\u6b21\u63a5\u89e6 RAG&#xff0c;\u4f1a\u628a\u5b83\u7406\u89e3\u6210&#034;\u8ba9\u6a21\u578b\u66f4\u806a\u660e\u7684\u6280\u672f&#034;\u3002\u8fd9\u4e2a\u65b9\u5411\u504f\u4e86\u3002RAG&#xff08;Retrieval Augmented Generation&#xff0c;\u68c0\u7d22\u589e\u5f3a\u751f\u6210&#xff09;\u89e3\u51b3\u7684\u95ee\u9898\u5f88\u5177\u4f53&#xff1a;\u6a21\u578b\u7684\u53c2\u6570\u91cc\u6ca1\u6709\u4f60\u9700\u8981\u7684\u77e5\u8bc6\u3002<\/p>\n<p>\u8fd9\u4e9b\u77e5\u8bc6\u53ef\u80fd\u662f&#xff1a;<\/p>\n<ul>\n<li>\u516c\u53f8\u5185\u90e8\u7684\u4ea7\u54c1\u6587\u6863\u548c\u64cd\u4f5c\u624b\u518c<\/li>\n<li>\u521a\u66f4\u65b0\u7684\u653f\u7b56\u6587\u4ef6\u548c\u5408\u540c\u6761\u6b3e<\/li>\n<li>\u79c1\u6709\u4ee3\u7801\u5e93\u7684\u63a5\u53e3\u8bf4\u660e<\/li>\n<li>\u7528\u6237\u81ea\u5df1\u4e0a\u4f20\u7684 PDF\u3001Word\u3001Markdown<\/li>\n<\/ul>\n<p>\u6a21\u578b\u4e0d\u53ef\u80fd\u628a\u6240\u6709\u79c1\u6709\u6570\u636e\u90fd\u8bad\u7ec3\u8fdb\u53c2\u6570\u91cc\u2014\u2014\u6210\u672c\u3001\u65f6\u6548\u3001\u5b89\u5168\u90fd\u4e0d\u5141\u8bb8\u3002RAG \u7684\u601d\u8def\u662f&#xff1a;\u751f\u6210\u56de\u7b54\u4e4b\u524d&#xff0c;\u5148\u4ece\u5916\u90e8\u77e5\u8bc6\u5e93\u4e2d\u68c0\u7d22\u51fa\u76f8\u5173\u7247\u6bb5&#xff0c;\u628a\u5b83\u4eec\u548c\u7528\u6237\u95ee\u9898\u4e00\u8d77\u585e\u8fdb Prompt&#xff0c;\u8ba9\u6a21\u578b\u57fa\u4e8e\u68c0\u7d22\u5230\u7684\u5185\u5bb9\u56de\u7b54\u3002<\/p>\n<p>\u6362\u4e2a\u89d2\u5ea6\u8bf4&#xff0c;RAG \u505a\u7684\u662f&#034;\u5f00\u5377\u8003\u8bd5&#034;&#xff1a;\u6a21\u578b\u672c\u8eab\u7684\u63a8\u7406\u80fd\u529b\u6ca1\u53d8&#xff0c;\u4f46\u5b83\u7b54\u9898\u65f6\u53ef\u4ee5\u7ffb\u53c2\u8003\u8d44\u6599\u3002\u53c2\u8003\u8d44\u6599\u627e\u5f97\u51c6\u4e0d\u51c6&#xff0c;\u76f4\u63a5\u51b3\u5b9a\u56de\u7b54\u8d28\u91cf\u3002<\/p>\n<p>\u8fd9\u7bc7\u6587\u7ae0\u4e0d\u6253\u7b97\u4ecb\u7ecd LangChain4j \u6216 Spring AI \u7684 RAG \u7ec4\u4ef6\u600e\u4e48\u4e00\u952e\u914d\u7f6e\u3002\u6211\u60f3\u628a\u94fe\u8def\u62c6\u5f00&#xff0c;\u7528\u7eaf Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f\u53ef\u7528\u7684 RAG&#xff0c;\u8ba9\u6bcf\u4e2a\u73af\u8282\u90fd\u66b4\u9732\u51fa\u6765\u3002\u770b\u5b8c\u4ee5\u540e&#xff0c;\u4f60\u518d\u53bb\u770b\u4efb\u4f55\u6846\u67b6\u7684 RAG \u5b9e\u73b0&#xff0c;\u90fd\u80fd\u5bf9\u5e94\u5230\u5177\u4f53\u6b65\u9aa4\u3002<\/p>\n<h3>\u6700\u5c0f RAG \u7684\u4e94\u4e2a\u73af\u8282&#xff1a;\u5207\u5206\u3001\u5d4c\u5165\u3001\u5b58\u50a8\u3001\u68c0\u7d22\u3001\u751f\u6210<\/h3>\n<p>\u4e00\u6761\u5b8c\u6574\u7684 RAG \u94fe\u8def\u5206\u4e94\u4e2a\u73af\u8282&#xff1a;<\/p>\n<p>\u539f\u59cb\u6587\u6863 \u2192 \u6587\u6863\u5207\u5206 \u2192 \u6587\u672c\u5d4c\u5165 \u2192 \u5411\u91cf\u5b58\u50a8 \u2192 \u76f8\u4f3c\u5ea6\u68c0\u7d22 \u2192 Prompt \u62fc\u63a5 \u2192 \u6a21\u578b\u751f\u6210<\/p>\n<p>\u6bcf\u4e2a\u73af\u8282\u90fd\u6709\u81ea\u5df1\u7684\u8bbe\u8ba1\u7a7a\u95f4\u548c\u5e38\u89c1\u5751&#xff1a;<\/p>\n<table>\n<tr>\u73af\u8282\u6838\u5fc3\u95ee\u9898\u5e38\u89c1\u505a\u6cd5\u76f4\u63a5\u5f71\u54cd<\/tr>\n<tbody>\n<tr>\n<td>\u6587\u6863\u5207\u5206<\/td>\n<td>\u4e00\u5757\u6587\u672c\u591a\u5927\u5408\u9002&#xff1f;\u4ece\u54ea\u5207\u5f00&#xff1f;<\/td>\n<td>\u6309\u5b57\u7b26\u6570\u3001\u6309 token\u3001\u6309\u6bb5\u843d\u3001\u6309\u8bed\u4e49<\/td>\n<td>\u53ec\u56de\u7387&#xff1a;\u76f8\u5173\u4fe1\u606f\u4f1a\u4e0d\u4f1a\u88ab\u5207\u5728\u4e24\u5757\u91cc<\/td>\n<\/tr>\n<tr>\n<td>\u6587\u672c\u5d4c\u5165<\/td>\n<td>\u7528\u4ec0\u4e48\u6a21\u578b&#xff1f;\u591a\u5c11\u7ef4&#xff1f;<\/td>\n<td>OpenAI Embedding\u3001BGE\u3001M3E<\/td>\n<td>\u8bed\u4e49\u76f8\u4f3c\u5ea6\u7684\u51c6\u786e\u6027<\/td>\n<\/tr>\n<tr>\n<td>\u5411\u91cf\u5b58\u50a8<\/td>\n<td>\u600e\u4e48\u5b58&#xff1f;\u600e\u4e48\u67e5&#xff1f;<\/td>\n<td>\u5411\u91cf\u6570\u636e\u5e93&#xff08;Milvus\u3001Pinecone\u3001Chroma&#xff09;<\/td>\n<td>\u68c0\u7d22\u901f\u5ea6\u548c\u89c4\u6a21<\/td>\n<\/tr>\n<tr>\n<td>\u76f8\u4f3c\u5ea6\u68c0\u7d22<\/td>\n<td>\u53d6 Top \u51e0&#xff1f;\u8981\u4e0d\u8981\u91cd\u6392&#xff1f;<\/td>\n<td>\u4f59\u5f26\u76f8\u4f3c\u5ea6\u3001MMR\u3001Reranker<\/td>\n<td>\u4e0a\u4e0b\u6587\u76f8\u5173\u6027\u548c\u5197\u4f59\u5ea6<\/td>\n<\/tr>\n<tr>\n<td>Prompt \u62fc\u63a5<\/td>\n<td>\u68c0\u7d22\u7ed3\u679c\u600e\u4e48\u7ec4\u7ec7&#xff1f;\u653e\u54ea&#xff1f;<\/td>\n<td>System Prompt \u91cc\u3001User Message \u91cc<\/td>\n<td>\u6a21\u578b\u662f\u5426\u4f1a\u6b63\u786e\u5f15\u7528\u4e0a\u4e0b\u6587<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u4e0b\u9762\u9010\u4e2a\u62c6\u5f00\u8bb2&#xff0c;\u6700\u540e\u5408\u6210\u4e00\u4e2a\u53ef\u8fd0\u884c\u7684\u6700\u5c0f\u793a\u4f8b\u3002<\/p>\n<h3>\u7b2c\u4e00\u6b65&#xff1a;\u6587\u6863\u5207\u5206&#xff0c;\u4e0d\u662f\u8d8a\u7ec6\u8d8a\u597d<\/h3>\n<p>\u6587\u6863\u5207\u5206&#xff08;Chunking&#xff09;\u662f RAG \u7684\u7b2c\u4e00\u6b65&#xff0c;\u4e5f\u662f\u6700\u5bb9\u6613\u88ab\u4f4e\u4f30\u7684\u4e00\u6b65\u3002\u5207\u5f97\u592a\u5927&#xff0c;\u5355\u5757\u5185\u5bb9\u566a\u58f0\u591a&#xff0c;\u68c0\u7d22\u4e0d\u51c6&#xff1b;\u5207\u5f97\u592a\u5c0f&#xff0c;\u4e0a\u4e0b\u6587\u65ad\u88c2&#xff0c;\u6a21\u578b\u62fc\u4e0d\u51fa\u5b8c\u6574\u7b54\u6848\u3002<\/p>\n<p>\u4e00\u4e2a\u5e38\u89c1\u7684\u8bef\u89e3\u662f&#034;\u5207\u5f97\u8d8a\u7ec6&#xff0c;\u53ec\u56de\u8d8a\u51c6&#034;\u3002\u5b9e\u9645\u4e0a&#xff0c;\u5207\u5206\u7c92\u5ea6\u9700\u8981\u548c\u7528\u6237\u95ee\u9898\u7684\u7c92\u5ea6\u5339\u914d\u3002\u5982\u679c\u7528\u6237\u95ee\u7684\u662f&#034;\u7b2c\u4e09\u7ae0\u8bb2\u4e86\u4ec0\u4e48&#034;&#xff0c;\u4f60\u628a\u6587\u6863\u5207\u6210 100 \u5b57\u4e00\u5757&#xff0c;\u68c0\u7d22\u51fa\u6765\u7684\u788e\u7247\u6839\u672c\u56de\u7b54\u4e0d\u4e86\u8fd9\u4e2a\u95ee\u9898\u3002<\/p>\n<h4>\u51e0\u79cd\u5207\u5206\u7b56\u7565\u7684\u5bf9\u6bd4<\/h4>\n<p>1. \u56fa\u5b9a\u957f\u5ea6\u5207\u5206&#xff08;Fixed-size Chunking&#xff09;<\/p>\n<p>\u6700\u7b80\u5355\u7684\u505a\u6cd5&#xff1a;\u6bcf N \u4e2a\u5b57\u7b26\u6216 token \u5207\u4e00\u5200&#xff0c;\u4e24\u5757\u4e4b\u95f4\u4fdd\u7559\u4e00\u70b9\u91cd\u53e0&#xff08;overlap&#xff09;&#xff0c;\u907f\u514d\u53e5\u5b50\u88ab\u5207\u65ad\u65f6\u4e22\u5931\u4fe1\u606f\u3002<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">fixedSizeSplit<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> text<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">int<\/span> chunkSize<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">int<\/span> overlap<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> chunks <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">ArrayList<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">int<\/span> start <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">while<\/span> <span class=\"token punctuation\">(<\/span>start <span class=\"token operator\">&lt;<\/span> text<span class=\"token punctuation\">.<\/span><span class=\"token function\">length<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> end <span class=\"token operator\">&#061;<\/span> <span class=\"token class-name\">Math<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">min<\/span><span class=\"token punctuation\">(<\/span>start <span class=\"token operator\">&#043;<\/span> chunkSize<span class=\"token punctuation\">,<\/span> text<span class=\"token punctuation\">.<\/span><span class=\"token function\">length<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        chunks<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span>text<span class=\"token punctuation\">.<\/span><span class=\"token function\">substring<\/span><span class=\"token punctuation\">(<\/span>start<span class=\"token punctuation\">,<\/span> end<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        start <span class=\"token operator\">&#043;&#061;<\/span> chunkSize <span class=\"token operator\">&#8211;<\/span> overlap<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> chunks<span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u4f18\u70b9&#xff1a;\u5b9e\u73b0\u7b80\u5355&#xff0c;\u901f\u5ea6\u5feb\u3002<br \/>\n\u7f3a\u70b9&#xff1a;\u53ef\u80fd\u4ece\u53e5\u5b50\u4e2d\u95f4\u5207\u65ad&#xff0c;\u7834\u574f\u8bed\u4e49\u5b8c\u6574\u6027\u3002<\/p>\n<p>2. \u6309\u8bed\u4e49\u5355\u5143\u5207\u5206&#xff08;Semantic Chunking&#xff09;<\/p>\n<p>\u5148\u6309\u6bb5\u843d\u3001\u53e5\u5b50\u6216 Markdown \u6807\u9898\u5207&#xff0c;\u518d\u6839\u636e\u9700\u8981\u5408\u5e76\u6216\u62c6\u5206\u3002\u6bd4\u5982&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">paragraphSplit<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> text<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token comment\">\/\/ \u6309\u7a7a\u884c\u5207\u5206\u6bb5\u843d<\/span><br \/>\n    <span class=\"token class-name\">String<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> paragraphs <span class=\"token operator\">&#061;<\/span> text<span class=\"token punctuation\">.<\/span><span class=\"token function\">split<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;\\\\\\\\n\\\\\\\\s*\\\\\\\\n&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> <span class=\"token class-name\">Arrays<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">stream<\/span><span class=\"token punctuation\">(<\/span>paragraphs<span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">map<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span><span class=\"token operator\">::<\/span><span class=\"token function\">trim<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">filter<\/span><span class=\"token punctuation\">(<\/span>p <span class=\"token operator\">-&gt;<\/span> <span class=\"token operator\">!<\/span>p<span class=\"token punctuation\">.<\/span><span class=\"token function\">isEmpty<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">toList<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u4f18\u70b9&#xff1a;\u8bed\u4e49\u5b8c\u6574\u6027\u597d\u3002<br \/>\n\u7f3a\u70b9&#xff1a;\u6bb5\u843d\u957f\u5ea6\u5dee\u5f02\u5927&#xff0c;\u6709\u7684\u592a\u77ed\u6709\u7684\u592a\u957f\u3002<\/p>\n<p>3. \u9012\u5f52\u5b57\u7b26\u5207\u5206&#xff08;Recursive Character Splitting&#xff09;<\/p>\n<p>LangChain \u91cc\u5e38\u7528\u7684\u7b56\u7565&#xff1a;\u5148\u6309\u6362\u884c\u5207&#xff0c;\u592a\u5927\u518d\u6309\u53e5\u5b50\u5207&#xff0c;\u8fd8\u5927\u518d\u6309\u5355\u8bcd\u5207&#xff0c;\u6700\u540e\u6309\u5b57\u7b26\u5207\u3002\u5c3d\u91cf\u5728\u81ea\u7136\u8fb9\u754c\u5904\u65ad\u5f00\u3002<\/p>\n<p>\u5728\u6700\u5c0f\u5b9e\u73b0\u91cc&#xff0c;\u6211\u63a8\u8350\u7528&#034;\u6309\u6bb5\u843d\u5207 &#043; \u957f\u5ea6\u8d85\u9650\u518d\u4e8c\u6b21\u62c6\u5206&#034;\u7684\u7ec4\u5408\u3002\u5927\u591a\u6570\u573a\u666f\u591f\u7528&#xff0c;\u5b9e\u73b0\u6210\u672c\u4e5f\u4f4e\u3002<\/p>\n<h4>\u5207\u5206\u5927\u5c0f\u600e\u4e48\u9009<\/h4>\n<p>\u6ca1\u6709\u4e07\u80fd\u516c\u5f0f&#xff0c;\u4f46\u6709\u51e0\u4e2a\u5224\u65ad\u7ef4\u5ea6&#xff1a;<\/p>\n<ul>\n<li>\u5d4c\u5165\u6a21\u578b\u7684\u6700\u5927 token \u9650\u5236&#xff1a;\u5355\u5757\u4e0d\u80fd\u8d85\u8fc7\u6a21\u578b\u7684\u8f93\u5165\u4e0a\u9650&#xff0c;\u6bd4\u5982 text-embedding-3-small \u662f 8191 token\u3002<\/li>\n<li>\u7528\u6237\u95ee\u9898\u7684\u5e73\u5747\u957f\u5ea6&#xff1a;\u95ee\u9898\u8d8a\u5b8f\u89c2&#xff0c;\u5757\u53ef\u4ee5\u8d8a\u5927&#xff1b;\u95ee\u9898\u8d8a\u5177\u4f53&#xff0c;\u5757\u53ef\u4ee5\u8d8a\u5c0f\u3002<\/li>\n<li>Prompt \u7a97\u53e3\u7684\u5269\u4f59\u7a7a\u95f4&#xff1a;\u68c0\u7d22\u7ed3\u679c\u8981\u585e\u8fdb Prompt&#xff0c;\u5757\u8d8a\u5927&#xff0c;\u80fd\u585e\u7684\u5757\u6570\u8d8a\u5c11\u3002<\/li>\n<\/ul>\n<p>\u4e00\u4e2a\u5b9e\u7528\u7684\u8d77\u70b9&#xff1a;\u4e2d\u6587 300-500 \u5b57\u4e00\u5757&#xff0c;\u91cd\u53e0 50-100 \u5b57\u3002\u7136\u540e\u6839\u636e\u5b9e\u9645\u68c0\u7d22\u6548\u679c\u8c03\u6574\u3002<\/p>\n<h3>\u7b2c\u4e8c\u6b65&#xff1a;\u6587\u672c\u5d4c\u5165&#xff0c;\u628a\u8bed\u4e49\u538b\u7f29\u6210\u5411\u91cf<\/h3>\n<p>\u5d4c\u5165&#xff08;Embedding&#xff09;\u5c31\u662f\u628a\u4e00\u6bb5\u6587\u672c\u8f6c\u6362\u6210\u4e00\u4e2a\u56fa\u5b9a\u7ef4\u5ea6\u7684\u6d6e\u70b9\u6570\u5411\u91cf\u3002\u8bed\u4e49\u76f8\u8fd1\u7684\u6587\u672c&#xff0c;\u5411\u91cf\u8ddd\u79bb\u4e5f\u8fd1\u3002\u8fd9\u662f RAG \u80fd&#034;\u6309\u610f\u601d\u627e\u6587\u672c&#034;\u7684\u57fa\u7840\u3002<\/p>\n<p>\u5d4c\u5165\u6a21\u578b\u7684\u9009\u62e9\u76f4\u63a5\u5f71\u54cd\u68c0\u7d22\u8d28\u91cf\u3002\u5e38\u89c1\u7684\u9009\u62e9&#xff1a;<\/p>\n<table>\n<tr>\u6a21\u578b\u7ef4\u5ea6\u7279\u70b9\u9002\u7528\u573a\u666f<\/tr>\n<tbody>\n<tr>\n<td>text-embedding-3-small<\/td>\n<td>1536&#xff08;\u53ef\u538b\u7f29&#xff09;<\/td>\n<td>OpenAI \u5b98\u65b9&#xff0c;\u8d28\u91cf\u7a33\u5b9a&#xff0c;\u6309\u91cf\u4ed8\u8d39<\/td>\n<td>\u82f1\u6587\u4e3a\u4e3b&#xff0c;\u9884\u7b97\u5145\u8db3<\/td>\n<\/tr>\n<tr>\n<td>text-embedding-3-large<\/td>\n<td>3072&#xff08;\u53ef\u538b\u7f29&#xff09;<\/td>\n<td>\u8d28\u91cf\u66f4\u9ad8&#xff0c;\u66f4\u8d35<\/td>\n<td>\u5bf9\u68c0\u7d22\u7cbe\u5ea6\u8981\u6c42\u9ad8<\/td>\n<\/tr>\n<tr>\n<td>BGE-m3<\/td>\n<td>1024<\/td>\n<td>\u5f00\u6e90&#xff0c;\u4e2d\u82f1\u6587\u6548\u679c\u597d<\/td>\n<td>\u79c1\u6709\u5316\u90e8\u7f72&#xff0c;\u4e2d\u6587\u573a\u666f<\/td>\n<\/tr>\n<tr>\n<td>M3E<\/td>\n<td>768<\/td>\n<td>\u5f00\u6e90&#xff0c;\u4e2d\u6587\u4f18\u5316<\/td>\n<td>\u4e2d\u6587\u8f7b\u91cf\u573a\u666f<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5728 Spring AI \u91cc\u8c03\u7528\u5d4c\u5165\u6a21\u578b\u5f88\u7b80\u5355&#xff1a;<\/p>\n<p><span class=\"token annotation punctuation\">&#064;Autowired<\/span><br \/>\n<span class=\"token keyword\">private<\/span> <span class=\"token class-name\">EmbeddingModel<\/span> embeddingModel<span class=\"token punctuation\">;<\/span><\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> <span class=\"token function\">embed<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> text<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token class-name\">EmbeddingResponse<\/span> response <span class=\"token operator\">&#061;<\/span> embeddingModel<span class=\"token punctuation\">.<\/span><span class=\"token function\">call<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">EmbeddingRequest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">List<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">of<\/span><span class=\"token punctuation\">(<\/span>text<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">EmbeddingOptionsBuilder<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">builder<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">build<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> response<span class=\"token punctuation\">.<\/span><span class=\"token function\">getResult<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">getOutput<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u4f46\u7406\u89e3\u5d4c\u5165\u7684\u51e0\u4e2a\u7279\u6027\u66f4\u91cd\u8981&#xff1a;<\/p>\n<p>1. \u5411\u91cf\u7ef4\u5ea6\u4e0d\u662f\u8d8a\u9ad8\u8d8a\u597d<\/p>\n<p>\u9ad8\u7ef4\u5411\u91cf\u80fd\u8868\u8fbe\u66f4\u4e30\u5bcc\u7684\u8bed\u4e49&#xff0c;\u4f46\u5b58\u50a8\u548c\u8ba1\u7b97\u6210\u672c\u4e5f\u66f4\u9ad8\u3002OpenAI \u7684 text-embedding-3-large \u6709 3072 \u7ef4&#xff0c;\u4f46\u53ef\u4ee5\u901a\u8fc7 API \u53c2\u6570\u538b\u7f29\u5230 256 \u7ef4&#xff0c;\u8d28\u91cf\u635f\u5931\u4e0d\u5927\u3002\u9009\u578b\u65f6\u5148\u6d4b\u4f60\u81ea\u5df1\u7684\u6570\u636e\u96c6&#xff0c;\u4e0d\u8981\u9ed8\u8ba4\u9009\u6700\u9ad8\u7ef4\u3002<\/p>\n<p>2. \u5d4c\u5165\u6a21\u578b\u548c\u751f\u6210\u6a21\u578b\u662f\u4e24\u56de\u4e8b<\/p>\n<p>\u5f88\u591a\u4eba\u4ee5\u4e3a\u7528 GPT-4 \u751f\u6210\u5c31\u5fc5\u987b\u7528 OpenAI \u7684\u5d4c\u5165\u6a21\u578b\u3002\u4e0d\u662f\u7684\u3002\u5d4c\u5165\u53ea\u548c\u68c0\u7d22\u8d28\u91cf\u76f8\u5173&#xff0c;\u751f\u6210\u6a21\u578b\u7528\u4ec0\u4e48\u90fd\u884c\u3002\u4f60\u5b8c\u5168\u53ef\u4ee5\u7528 BGE \u505a\u5d4c\u5165&#xff0c;\u7528 DeepSeek \u505a\u751f\u6210\u3002<\/p>\n<p>3. \u5d4c\u5165\u4e0d\u662f\u65e0\u635f\u538b\u7f29<\/p>\n<p>\u4e00\u6bb5 500 \u5b57\u7684\u6587\u672c\u538b\u7f29\u6210 1536 \u4e2a\u6d6e\u70b9\u6570&#xff0c;\u5fc5\u7136\u4e22\u5931\u4fe1\u606f\u3002\u7cbe\u7ec6\u7684\u6570\u5b57\u3001\u7cbe\u786e\u7684\u547d\u540d\u3001\u590d\u6742\u7684\u903b\u8f91\u5173\u7cfb&#xff0c;\u5728\u5411\u91cf\u7a7a\u95f4\u91cc\u53ef\u80fd\u533a\u5206\u4e0d\u5f00\u3002\u8fd9\u4e5f\u662f\u4e3a\u4ec0\u4e48\u7eaf\u5411\u91cf\u68c0\u7d22\u7684\u53ec\u56de\u7387\u6709\u4e0a\u9650&#xff0c;\u901a\u5e38\u9700\u8981\u914d\u5408\u5173\u952e\u8bcd\u68c0\u7d22\u6216\u91cd\u6392\u3002<\/p>\n<h3>\u7b2c\u4e09\u6b65&#xff1a;\u5411\u91cf\u5b58\u50a8&#xff0c;\u4e3a\u4ec0\u4e48\u4e0d\u80fd\u53ea\u7528 HashMap<\/h3>\n<p>\u6709\u4e86\u5411\u91cf&#xff0c;\u63a5\u4e0b\u6765\u8981\u5b58\u8d77\u6765&#xff0c;\u4ee5\u4fbf\u540e\u7eed\u68c0\u7d22\u3002\u6700\u7b80\u5355\u7684\u505a\u6cd5\u2014\u2014\u5168\u90e8\u653e\u8fdb\u5185\u5b58\u7684 List \u91cc&#xff0c;\u67e5\u8be2\u65f6\u9010\u6761\u8ba1\u7b97\u76f8\u4f3c\u5ea6\u3002<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">InMemoryVectorStore<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">VectorDocument<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> documents <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">ArrayList<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token keyword\">void<\/span> <span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> id<span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">String<\/span> content<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> vector<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        documents<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">new<\/span> <span class=\"token class-name\">VectorDocument<\/span><span class=\"token punctuation\">(<\/span>id<span class=\"token punctuation\">,<\/span> content<span class=\"token punctuation\">,<\/span> vector<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">SearchResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">search<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> queryVector<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">int<\/span> topK<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> documents<span class=\"token punctuation\">.<\/span><span class=\"token function\">stream<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">map<\/span><span class=\"token punctuation\">(<\/span>doc <span class=\"token operator\">-&gt;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">SearchResult<\/span><span class=\"token punctuation\">(<\/span>doc<span class=\"token punctuation\">,<\/span> <span class=\"token function\">cosineSimilarity<\/span><span class=\"token punctuation\">(<\/span>queryVector<span class=\"token punctuation\">,<\/span> doc<span class=\"token punctuation\">.<\/span><span class=\"token function\">vector<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">sorted<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">(<\/span>a<span class=\"token punctuation\">,<\/span> b<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">-&gt;<\/span> <span class=\"token class-name\">Double<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">compare<\/span><span class=\"token punctuation\">(<\/span>b<span class=\"token punctuation\">.<\/span><span class=\"token function\">score<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> a<span class=\"token punctuation\">.<\/span><span class=\"token function\">score<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">limit<\/span><span class=\"token punctuation\">(<\/span>topK<span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">toList<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u6570\u636e\u91cf\u5c0f\u7684\u65f6\u5019&#xff08;\u51e0\u5343\u6761\u4ee5\u5185&#xff09;&#xff0c;\u7ebf\u6027\u626b\u63cf\u5b8c\u5168\u591f\u7528\u3002\u4f46\u5f53\u6587\u6863\u91cf\u5230\u4e86\u51e0\u4e07\u3001\u51e0\u5341\u4e07\u6761&#xff0c;\u6bcf\u6b21\u67e5\u8be2\u90fd\u8981\u548c\u6240\u6709\u5411\u91cf\u7b97\u4e00\u904d\u76f8\u4f3c\u5ea6&#xff0c;\u5ef6\u8fdf\u4f1a\u975e\u5e38\u660e\u663e\u3002<\/p>\n<p>\u8fd9\u65f6\u5019\u5c31\u9700\u8981\u5411\u91cf\u6570\u636e\u5e93\u3002\u5411\u91cf\u6570\u636e\u5e93\u505a\u7684\u6838\u5fc3\u4f18\u5316\u662f\u8fd1\u4f3c\u6700\u8fd1\u90bb&#xff08;ANN&#xff09;\u7d22\u5f15&#xff1a;\u4e0d\u7cbe\u786e\u8ba1\u7b97\u6bcf\u4e00\u6761\u7684\u8ddd\u79bb&#xff0c;\u800c\u662f\u7528\u67d0\u79cd\u6570\u636e\u7ed3\u6784&#xff08;HNSW\u3001IVF\u3001PQ \u7b49&#xff09;\u5feb\u901f\u7f29\u5c0f\u8303\u56f4&#xff0c;\u727a\u7272\u4e00\u70b9\u70b9\u7cbe\u5ea6\u6362\u51e0\u5341\u4e0a\u767e\u500d\u7684\u901f\u5ea6\u3002<\/p>\n<p>\u5e38\u89c1\u5411\u91cf\u6570\u636e\u5e93\u5bf9\u6bd4&#xff1a;<\/p>\n<table>\n<tr>\u6570\u636e\u5e93\u90e8\u7f72\u65b9\u5f0f\u7279\u70b9\u9002\u7528\u573a\u666f<\/tr>\n<tbody>\n<tr>\n<td>Milvus<\/td>\n<td>\u72ec\u7acb\u90e8\u7f72<\/td>\n<td>\u529f\u80fd\u6700\u5168&#xff0c;\u5206\u5e03\u5f0f&#xff0c;\u751f\u6001\u6210\u719f<\/td>\n<td>\u5927\u89c4\u6a21\u751f\u4ea7\u73af\u5883<\/td>\n<\/tr>\n<tr>\n<td>Pinecone<\/td>\n<td>SaaS<\/td>\n<td>\u6258\u7ba1\u670d\u52a1&#xff0c;\u96f6\u8fd0\u7ef4<\/td>\n<td>\u4e0d\u60f3\u81ea\u5df1\u8fd0\u7ef4<\/td>\n<\/tr>\n<tr>\n<td>Chroma<\/td>\n<td>\u672c\u5730\/\u670d\u52a1\u7aef<\/td>\n<td>\u8f7b\u91cf&#xff0c;Python \u751f\u6001\u597d<\/td>\n<td>\u539f\u578b\u9a8c\u8bc1\u3001\u5c0f\u89c4\u6a21<\/td>\n<\/tr>\n<tr>\n<td>PostgreSQL &#043; pgvector<\/td>\n<td>\u6269\u5c55<\/td>\n<td>\u590d\u7528\u73b0\u6709 PG&#xff0c;\u652f\u6301 SQL<\/td>\n<td>\u5df2\u6709 PG&#xff0c;\u5411\u91cf\u6570\u636e\u91cf\u4e0d\u5927<\/td>\n<\/tr>\n<tr>\n<td>Redis &#043; RedisVL<\/td>\n<td>\u6269\u5c55<\/td>\n<td>\u590d\u7528 Redis&#xff0c;\u4f4e\u5ef6\u8fdf<\/td>\n<td>\u7f13\u5b58\u7ea7\u5411\u91cf\u68c0\u7d22<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5728\u6700\u5c0f RAG \u5b9e\u73b0\u91cc&#xff0c;\u7528\u5185\u5b58 List \u5c31\u591f\u4e86\u3002\u7b49\u4f60\u9a8c\u8bc1\u5b8c\u6574\u4e2a\u94fe\u8def\u3001\u786e\u8ba4\u68c0\u7d22\u6548\u679c\u6709\u4ef7\u503c&#xff0c;\u518d\u8003\u8651\u8fc1\u79fb\u5230\u5411\u91cf\u6570\u636e\u5e93\u3002<\/p>\n<h4>\u4f59\u5f26\u76f8\u4f3c\u5ea6\u7684\u8ba1\u7b97<\/h4>\n<p>\u5411\u91cf\u76f8\u4f3c\u5ea6\u6700\u5e38\u7528\u7684\u662f\u4f59\u5f26\u76f8\u4f3c\u5ea6&#xff08;Cosine Similarity&#xff09;&#xff0c;\u8861\u91cf\u4e24\u4e2a\u5411\u91cf\u65b9\u5411\u7684\u5939\u89d2&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">double<\/span> <span class=\"token function\">cosineSimilarity<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> a<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> b<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>a<span class=\"token punctuation\">.<\/span>length <span class=\"token operator\">!&#061;<\/span> b<span class=\"token punctuation\">.<\/span>length<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">throw<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">IllegalArgumentException<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;Vector dimensions do not match&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n    <span class=\"token keyword\">double<\/span> dotProduct <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">double<\/span> normA <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">double<\/span> normB <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">for<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token keyword\">int<\/span> i <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span> i <span class=\"token operator\">&lt;<\/span> a<span class=\"token punctuation\">.<\/span>length<span class=\"token punctuation\">;<\/span> i<span class=\"token operator\">&#043;&#043;<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        dotProduct <span class=\"token operator\">&#043;&#061;<\/span> a<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">*<\/span> b<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        normA <span class=\"token operator\">&#043;&#061;<\/span> a<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">*<\/span> a<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        normB <span class=\"token operator\">&#043;&#061;<\/span> b<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">*<\/span> b<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n    <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>normA <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token number\">0.0<\/span> <span class=\"token operator\">||<\/span> normB <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> dotProduct <span class=\"token operator\">\/<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token class-name\">Math<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">sqrt<\/span><span class=\"token punctuation\">(<\/span>normA<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">*<\/span> <span class=\"token class-name\">Math<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">sqrt<\/span><span class=\"token punctuation\">(<\/span>normB<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u53d6\u503c\u8303\u56f4\u662f [-1, 1]&#xff0c;\u8d8a\u5927\u8868\u793a\u8d8a\u76f8\u4f3c\u3002\u6587\u672c\u5d4c\u5165\u7684\u7ed3\u679c\u901a\u5e38\u90fd\u662f\u6b63\u7684&#xff0c;\u6240\u4ee5\u5b9e\u9645\u770b\u5230\u7684\u5206\u6570\u5927\u591a\u5728 0 \u5230 1 \u4e4b\u95f4\u3002<\/p>\n<h3>\u7b2c\u56db\u6b65&#xff1a;\u76f8\u4f3c\u5ea6\u68c0\u7d22&#xff0c;Top-K \u53ea\u662f\u5f00\u59cb<\/h3>\n<p>\u68c0\u7d22\u73af\u8282\u6700\u6734\u7d20\u7684\u505a\u6cd5\u662f&#xff1a;\u628a\u7528\u6237\u95ee\u9898\u4e5f\u8f6c\u6210\u5411\u91cf&#xff0c;\u548c\u5e93\u91cc\u6240\u6709\u5411\u91cf\u7b97\u76f8\u4f3c\u5ea6&#xff0c;\u53d6\u5206\u6570\u6700\u9ad8\u7684\u524d K \u4e2a\u3002<\/p>\n<p>\u4f46\u76f4\u63a5\u53d6 Top-K \u6709\u51e0\u4e2a\u95ee\u9898&#xff1a;<\/p>\n<p>1. \u5197\u4f59\u95ee\u9898<\/p>\n<p>\u524d K \u4e2a\u7ed3\u679c\u53ef\u80fd\u5185\u5bb9\u9ad8\u5ea6\u91cd\u53e0\u3002\u6bd4\u5982\u540c\u4e00\u6bb5\u6587\u5b57\u5728\u6587\u6863\u91cc\u51fa\u73b0\u4e86\u591a\u6b21&#xff0c;\u6216\u8005\u76f8\u90bb\u4e24\u5757\u5185\u5bb9\u5dee\u4e0d\u591a\u3002\u628a\u5b83\u4eec\u90fd\u585e\u8fdb Prompt \u662f\u6d6a\u8d39 token\u3002<\/p>\n<p>\u89e3\u51b3\u65b9\u6cd5&#xff1a;MMR&#xff08;Maximal Marginal Relevance&#xff09;\u2014\u2014\u5148\u53d6\u4e00\u4e2a\u6700\u76f8\u5173\u7684&#xff0c;\u7136\u540e\u4e0b\u4e00\u4e2a\u65e2\u8981\u548c\u67e5\u8be2\u76f8\u5173&#xff0c;\u53c8\u8981\u548c\u5df2\u9009\u7684\u5dee\u5f02\u5927\u3002\u5728\u76f8\u5173\u6027\u548c\u591a\u6837\u6027\u4e4b\u95f4\u505a\u6743\u8861\u3002<\/p>\n<p>2. \u76f8\u5173\u6027\u9608\u503c\u95ee\u9898<\/p>\n<p>Top-K \u4e0d\u7ba1\u76f8\u5173\u4e0d\u76f8\u5173&#xff0c;\u603b\u80fd\u8fd4\u56de K \u4e2a\u3002\u5982\u679c\u5e93\u91cc\u6839\u672c\u6ca1\u6709\u76f8\u5173\u5185\u5bb9&#xff0c;\u8fd4\u56de\u7684\u8fd9 K \u4e2a\u5c31\u662f\u566a\u58f0&#xff0c;\u6a21\u578b\u53ef\u80fd\u4f1a\u88ab\u8bef\u5bfc\u3002<\/p>\n<p>\u89e3\u51b3\u65b9\u6cd5&#xff1a;\u8bbe\u4e00\u4e2a\u76f8\u4f3c\u5ea6\u9608\u503c&#xff0c;\u4f4e\u4e8e\u9608\u503c\u7684\u7ed3\u679c\u76f4\u63a5\u4e22\u5f03&#xff0c;\u5b81\u53ef\u4e0d\u68c0\u7d22\u4e5f\u4e0d\u585e\u65e0\u5173\u5185\u5bb9\u3002<\/p>\n<p>3. \u68c0\u7d22\u7c92\u5ea6\u95ee\u9898<\/p>\n<p>\u6709\u65f6\u5019\u6700\u76f8\u5173\u7684\u662f\u7b2c 5 \u5757&#xff0c;\u4f46\u7b54\u6848\u5176\u5b9e\u5728\u7b2c 4 \u5757\u548c\u7b2c 6 \u5757\u7684\u8fde\u63a5\u5904\u3002\u53ea\u53d6 Top-K \u53ef\u80fd\u6f0f\u6389\u5173\u952e\u4e0a\u4e0b\u6587\u3002<\/p>\n<p>\u89e3\u51b3\u65b9\u6cd5&#xff1a;\u68c0\u7d22\u5230\u76f8\u5173\u5757\u540e&#xff0c;\u628a\u5b83\u524d\u540e\u5404\u4e00\u5757\u4e5f\u5e26\u4e0a&#xff08;\u4e0a\u4e0b\u6587\u7a97\u53e3\u6269\u5c55&#xff09;&#xff0c;\u6216\u8005\u5728\u5207\u5206\u65f6\u5c31\u52a0\u5927 overlap\u3002<\/p>\n<p>\u4e00\u4e2a\u5b9e\u7528\u7684\u68c0\u7d22\u6d41\u7a0b&#xff1a;<\/p>\n<p>1. \u95ee\u9898\u8f6c\u6210\u67e5\u8be2\u5411\u91cf<br \/>\n2. \u521d\u6b65\u68c0\u7d22 Top-2K&#xff08;\u591a\u53d6\u4e00\u4e9b&#xff0c;\u7ed9\u540e\u9762\u8fc7\u6ee4\u7559\u4f59\u91cf&#xff09;<br \/>\n3. \u6309\u76f8\u4f3c\u5ea6\u9608\u503c\u8fc7\u6ee4&#xff0c;\u4f4e\u4e8e 0.7 \u7684\u6254\u6389<br \/>\n4. \u7528 MMR \u505a\u591a\u6837\u6027\u91cd\u6392&#xff0c;\u9009\u51fa\u6700\u7ec8 Top-K<br \/>\n5. \u5bf9\u6bcf\u4e2a\u547d\u4e2d\u5757&#xff0c;\u6269\u5c55\u5176\u524d\u540e\u76f8\u90bb\u5757&#xff08;\u53ef\u9009&#xff09;<\/p>\n<h3>\u7b2c\u4e94\u6b65&#xff1a;Prompt \u62fc\u63a5&#xff0c;\u628a\u68c0\u7d22\u7ed3\u679c\u5582\u7ed9\u6a21\u578b<\/h3>\n<p>\u68c0\u7d22\u5230\u7684\u6587\u672c\u7247\u6bb5\u4e0d\u80fd\u76f4\u63a5\u6254\u7ed9\u6a21\u578b\u3002\u9700\u8981\u7ec4\u7ec7\u6210\u6e05\u6670\u7684\u683c\u5f0f&#xff0c;\u544a\u8bc9\u6a21\u578b&#xff1a;\u8fd9\u4e9b\u662f\u53c2\u8003\u8d44\u6599&#xff0c;\u8bf7\u57fa\u4e8e\u5b83\u4eec\u56de\u7b54\u3002<\/p>\n<p>\u4e00\u4e2a\u57fa\u7840\u7684 RAG Prompt \u6a21\u677f&#xff1a;<\/p>\n<p>\u4f60\u662f\u4e00\u4e2a\u95ee\u7b54\u52a9\u624b\u3002\u8bf7\u57fa\u4e8e\u4ee5\u4e0b\u53c2\u8003\u8d44\u6599\u56de\u7b54\u7528\u6237\u7684\u95ee\u9898\u3002<br \/>\n\u5982\u679c\u53c2\u8003\u8d44\u6599\u4e2d\u6ca1\u6709\u76f8\u5173\u4fe1\u606f&#xff0c;\u8bf7\u76f4\u63a5\u8bf4&#034;\u53c2\u8003\u8d44\u6599\u4e2d\u6ca1\u6709\u627e\u5230\u76f8\u5173\u5185\u5bb9&#034;&#xff0c;\u4e0d\u8981\u7f16\u9020\u7b54\u6848\u3002<\/p>\n<p>\u53c2\u8003\u8d44\u6599&#xff1a;<br \/>\n&#8212;<br \/>\n1. {chunk_1_content}<br \/>\n&#8212;<br \/>\n2. {chunk_2_content}<br \/>\n&#8212;<br \/>\n&#8230;<\/p>\n<p>\u7528\u6237\u95ee\u9898&#xff1a;{user_question}<\/p>\n<p>\u51e0\u4e2a\u8bbe\u8ba1\u7ec6\u8282&#xff1a;<\/p>\n<p>1. \u53c2\u8003\u8d44\u6599\u653e\u5728\u54ea<\/p>\n<p>\u901a\u5e38\u653e\u5728 System Prompt \u6216 User Message \u91cc\u3002\u653e\u5728 System Prompt \u91cc&#xff0c;\u6a21\u578b\u66f4\u5bb9\u6613&#034;\u8bb0\u4f4f&#034;&#xff1b;\u653e\u5728 User Message \u91cc&#xff0c;\u66f4\u7075\u6d3b&#xff0c;\u53ef\u4ee5\u6bcf\u6b21\u66ff\u6362\u3002\u5bf9\u5927\u591a\u6570 RAG \u573a\u666f&#xff0c;\u4e24\u8005\u6548\u679c\u5dee\u5f02\u4e0d\u5927\u3002\u5982\u679c\u68c0\u7d22\u7ed3\u679c\u5f88\u957f&#xff0c;\u6ce8\u610f\u4e0d\u8981\u8d85\u51fa\u6a21\u578b\u7684\u4e0a\u4e0b\u6587\u7a97\u53e3\u3002<\/p>\n<p>2. \u8981\u4e0d\u8981\u7f16\u53f7<\/p>\n<p>\u7f16\u53f7\u6709\u4e24\u4e2a\u597d\u5904&#xff1a;\u6a21\u578b\u56de\u7b54\u65f6\u53ef\u4ee5\u5f15\u7528\u5177\u4f53\u6765\u6e90&#xff08;\u201c\u6839\u636e\u8d44\u6599 2\u2026\u201d&#xff09;&#xff0c;\u65b9\u4fbf\u4e8b\u540e\u6838\u5bf9&#xff1b;\u53e6\u5916&#xff0c;\u7f16\u53f7\u672c\u8eab\u4e5f\u80fd\u5e2e\u52a9\u6a21\u578b\u533a\u5206\u4e0d\u540c\u7684\u7247\u6bb5\u3002<\/p>\n<p>3. &#034;\u4e0d\u77e5\u9053&#034;\u7684\u7ea6\u675f\u5f88\u91cd\u8981<\/p>\n<p>\u4e00\u5b9a\u8981\u660e\u786e\u544a\u8bc9\u6a21\u578b&#xff1a;\u8d44\u6599\u91cc\u6ca1\u6709\u5c31\u8bf4\u4e0d\u77e5\u9053\u3002\u5426\u5219\u6a21\u578b\u4f1a\u57fa\u4e8e\u53c2\u6570\u91cc\u7684\u77e5\u8bc6\u7f16\u9020\u7b54\u6848&#xff0c;\u770b\u8d77\u6765\u5f88\u5408\u7406\u4f46\u53ef\u80fd\u5b8c\u5168\u4e0d\u5bf9\u3002\u8fd9\u662f RAG \u91cc\u6700\u5e38\u89c1\u7684\u5e7b\u89c9\u6765\u6e90\u4e4b\u4e00\u3002<\/p>\n<p>4. \u68c0\u7d22\u7ed3\u679c\u7684\u987a\u5e8f<\/p>\n<p>\u76f8\u5173\u5ea6\u9ad8\u7684\u653e\u524d\u9762\u8fd8\u662f\u540e\u9762&#xff1f;\u6709\u7814\u7a76\u8868\u660e&#xff0c;\u6a21\u578b\u5bf9\u5f00\u5934\u548c\u7ed3\u5c3e\u7684\u4fe1\u606f\u66f4\u654f\u611f&#xff08;\u9996\u56e0\u6548\u5e94\u548c\u8fd1\u56e0\u6548\u5e94&#xff09;\u3002\u6240\u4ee5\u6700\u9ad8\u76f8\u5173\u7684\u653e\u7b2c\u4e00\u4e2a&#xff0c;\u6b21\u9ad8\u7684\u653e\u6700\u540e\u4e00\u4e2a&#xff0c;\u4e2d\u95f4\u653e\u5176\u4ed6\u7684&#xff0c;\u53ef\u80fd\u6bd4\u4e25\u683c\u6309\u5206\u6570\u6392\u5e8f\u6548\u679c\u66f4\u597d\u3002<\/p>\n<h3>\u8dd1\u901a\u6700\u5c0f\u94fe\u8def&#xff1a;\u4e00\u4e2a\u53ef\u8fd0\u884c\u7684 Java \u793a\u4f8b<\/h3>\n<p>\u4e0b\u9762\u628a\u4e94\u4e2a\u73af\u8282\u62fc\u8d77\u6765&#xff0c;\u5199\u4e00\u4e2a\u6700\u5c0f\u53ef\u7528\u7684 RAG \u7c7b\u3002\u4e3a\u4e86\u4e0d\u4f9d\u8d56\u5916\u90e8\u670d\u52a1&#xff0c;\u5d4c\u5165\u90e8\u5206\u6211\u7528\u4e00\u4e2a\u6a21\u62df\u5b9e\u73b0&#xff1b;\u4f60\u63a5\u5165\u771f\u5b9e\u5d4c\u5165\u6a21\u578b\u65f6&#xff0c;\u66ff\u6362 embed() \u65b9\u6cd5\u5373\u53ef\u3002<\/p>\n<h4>\u9879\u76ee\u4f9d\u8d56<\/h4>\n<p><span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>dependencies<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token comment\">&lt;!&#8211; Spring AI \u7528\u4e8e\u8c03\u7528\u771f\u5b9e\u5d4c\u5165\u6a21\u578b\u548c\u751f\u6210\u6a21\u578b &#8211;&gt;<\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>dependency<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>groupId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>org.springframework.ai<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>groupId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>artifactId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>spring-ai-model<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>artifactId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>version<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>2.0.1<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>version<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>dependency<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>dependency<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>groupId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>org.springframework.ai<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>groupId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>artifactId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>spring-ai-openai-spring-boot-starter<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>artifactId<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>version<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>2.0.1<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>version<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>dependency<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>dependencies<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><\/p>\n<h4>\u6587\u6863\u5207\u5206\u5668<\/h4>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">TextSplitter<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token keyword\">int<\/span> chunkSize<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token keyword\">int<\/span> overlap<span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">TextSplitter<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">int<\/span> chunkSize<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">int<\/span> overlap<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">this<\/span><span class=\"token punctuation\">.<\/span>chunkSize <span class=\"token operator\">&#061;<\/span> chunkSize<span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">this<\/span><span class=\"token punctuation\">.<\/span>overlap <span class=\"token operator\">&#061;<\/span> overlap<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token comment\">\/**<br \/>\n     * \u5148\u6309\u6bb5\u843d\u5207&#xff0c;\u6bb5\u843d\u592a\u957f\u7684\u518d\u6309\u56fa\u5b9a\u5927\u5c0f\u4e8c\u6b21\u5207\u5206<br \/>\n     *\/<\/span><br \/>\n    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">split<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> text<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> result <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">ArrayList<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token class-name\">String<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> paragraphs <span class=\"token operator\">&#061;<\/span> text<span class=\"token punctuation\">.<\/span><span class=\"token function\">split<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;\\\\\\\\n\\\\\\\\s*\\\\\\\\n&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token keyword\">for<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> paragraph <span class=\"token operator\">:<\/span> paragraphs<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            <span class=\"token class-name\">String<\/span> trimmed <span class=\"token operator\">&#061;<\/span> paragraph<span class=\"token punctuation\">.<\/span><span class=\"token function\">trim<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>trimmed<span class=\"token punctuation\">.<\/span><span class=\"token function\">isEmpty<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token keyword\">continue<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>trimmed<span class=\"token punctuation\">.<\/span><span class=\"token function\">length<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token operator\">&lt;&#061;<\/span> chunkSize<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n                result<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span>trimmed<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            <span class=\"token punctuation\">}<\/span> <span class=\"token keyword\">else<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n                <span class=\"token comment\">\/\/ \u8d85\u957f\u6bb5\u843d&#xff0c;\u6309\u56fa\u5b9a\u5927\u5c0f &#043; \u91cd\u53e0\u62c6\u5206<\/span><br \/>\n                <span class=\"token keyword\">int<\/span> start <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n                <span class=\"token keyword\">while<\/span> <span class=\"token punctuation\">(<\/span>start <span class=\"token operator\">&lt;<\/span> trimmed<span class=\"token punctuation\">.<\/span><span class=\"token function\">length<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n                    <span class=\"token keyword\">int<\/span> end <span class=\"token operator\">&#061;<\/span> <span class=\"token class-name\">Math<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">min<\/span><span class=\"token punctuation\">(<\/span>start <span class=\"token operator\">&#043;<\/span> chunkSize<span class=\"token punctuation\">,<\/span> trimmed<span class=\"token punctuation\">.<\/span><span class=\"token function\">length<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n                    result<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span>trimmed<span class=\"token punctuation\">.<\/span><span class=\"token function\">substring<\/span><span class=\"token punctuation\">(<\/span>start<span class=\"token punctuation\">,<\/span> end<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n                    start <span class=\"token operator\">&#043;&#061;<\/span> chunkSize <span class=\"token operator\">&#8211;<\/span> overlap<span class=\"token punctuation\">;<\/span><br \/>\n                <span class=\"token punctuation\">}<\/span><br \/>\n            <span class=\"token punctuation\">}<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> result<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h4>\u5185\u5b58\u5411\u91cf\u5b58\u50a8<\/h4>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">InMemoryVectorStore<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">Chunk<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> chunks <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">ArrayList<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">Chunk<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> id<span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">String<\/span> content<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> vector<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">SearchResult<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">Chunk<\/span> chunk<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">double<\/span> score<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token keyword\">void<\/span> <span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">Chunk<\/span> chunk<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        chunks<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span>chunk<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">SearchResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">search<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> queryVector<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">int<\/span> topK<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">double<\/span> threshold<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> chunks<span class=\"token punctuation\">.<\/span><span class=\"token function\">stream<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">map<\/span><span class=\"token punctuation\">(<\/span>chunk <span class=\"token operator\">-&gt;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">SearchResult<\/span><span class=\"token punctuation\">(<\/span>chunk<span class=\"token punctuation\">,<\/span> <span class=\"token function\">cosineSimilarity<\/span><span class=\"token punctuation\">(<\/span>queryVector<span class=\"token punctuation\">,<\/span> chunk<span class=\"token punctuation\">.<\/span><span class=\"token function\">vector<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">filter<\/span><span class=\"token punctuation\">(<\/span>result <span class=\"token operator\">-&gt;<\/span> result<span class=\"token punctuation\">.<\/span><span class=\"token function\">score<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token operator\">&gt;&#061;<\/span> threshold<span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">sorted<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">(<\/span>a<span class=\"token punctuation\">,<\/span> b<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">-&gt;<\/span> <span class=\"token class-name\">Double<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">compare<\/span><span class=\"token punctuation\">(<\/span>b<span class=\"token punctuation\">.<\/span><span class=\"token function\">score<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> a<span class=\"token punctuation\">.<\/span><span class=\"token function\">score<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">limit<\/span><span class=\"token punctuation\">(<\/span>topK<span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">toList<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">double<\/span> <span class=\"token function\">cosineSimilarity<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> a<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> b<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> dot <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">,<\/span> normA <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">,<\/span> normB <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">for<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token keyword\">int<\/span> i <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span> i <span class=\"token operator\">&lt;<\/span> a<span class=\"token punctuation\">.<\/span>length<span class=\"token punctuation\">;<\/span> i<span class=\"token operator\">&#043;&#043;<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            dot <span class=\"token operator\">&#043;&#061;<\/span> a<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">*<\/span> b<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            normA <span class=\"token operator\">&#043;&#061;<\/span> a<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">*<\/span> a<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            normB <span class=\"token operator\">&#043;&#061;<\/span> b<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">*<\/span> b<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><br \/>\n        <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>normA <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token number\">0<\/span> <span class=\"token operator\">||<\/span> normB <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token keyword\">return<\/span> <span class=\"token number\">0.0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> dot <span class=\"token operator\">\/<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token class-name\">Math<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">sqrt<\/span><span class=\"token punctuation\">(<\/span>normA<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">*<\/span> <span class=\"token class-name\">Math<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">sqrt<\/span><span class=\"token punctuation\">(<\/span>normB<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h4>RAG \u670d\u52a1\u4e3b\u7c7b<\/h4>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">MinimalRAG<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">TextSplitter<\/span> splitter<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">InMemoryVectorStore<\/span> store<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">EmbeddingModel<\/span> embeddingModel<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">ChatModel<\/span> chatModel<span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">MinimalRAG<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">EmbeddingModel<\/span> embeddingModel<span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">ChatModel<\/span> chatModel<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">this<\/span><span class=\"token punctuation\">.<\/span>splitter <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">TextSplitter<\/span><span class=\"token punctuation\">(<\/span><span class=\"token number\">400<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">50<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">this<\/span><span class=\"token punctuation\">.<\/span>store <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">InMemoryVectorStore<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">this<\/span><span class=\"token punctuation\">.<\/span>embeddingModel <span class=\"token operator\">&#061;<\/span> embeddingModel<span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">this<\/span><span class=\"token punctuation\">.<\/span>chatModel <span class=\"token operator\">&#061;<\/span> chatModel<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token comment\">\/**<br \/>\n     * \u628a\u6587\u6863\u52a0\u5165\u77e5\u8bc6\u5e93<br \/>\n     *\/<\/span><br \/>\n    <span class=\"token keyword\">public<\/span> <span class=\"token keyword\">void<\/span> <span class=\"token function\">ingestDocument<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> documentId<span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">String<\/span> content<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> chunks <span class=\"token operator\">&#061;<\/span> splitter<span class=\"token punctuation\">.<\/span><span class=\"token function\">split<\/span><span class=\"token punctuation\">(<\/span>content<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">for<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token keyword\">int<\/span> i <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span> i <span class=\"token operator\">&lt;<\/span> chunks<span class=\"token punctuation\">.<\/span><span class=\"token function\">size<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span> i<span class=\"token operator\">&#043;&#043;<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            <span class=\"token class-name\">String<\/span> chunkText <span class=\"token operator\">&#061;<\/span> chunks<span class=\"token punctuation\">.<\/span><span class=\"token function\">get<\/span><span class=\"token punctuation\">(<\/span>i<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> vector <span class=\"token operator\">&#061;<\/span> <span class=\"token function\">embed<\/span><span class=\"token punctuation\">(<\/span>chunkText<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            <span class=\"token class-name\">String<\/span> chunkId <span class=\"token operator\">&#061;<\/span> documentId <span class=\"token operator\">&#043;<\/span> <span class=\"token string\">&#034;-&#034;<\/span> <span class=\"token operator\">&#043;<\/span> i<span class=\"token punctuation\">;<\/span><br \/>\n            store<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span><span class=\"token keyword\">new<\/span> <span class=\"token class-name\">InMemoryVectorStore<span class=\"token punctuation\">.<\/span>Chunk<\/span><span class=\"token punctuation\">(<\/span>chunkId<span class=\"token punctuation\">,<\/span> chunkText<span class=\"token punctuation\">,<\/span> vector<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token comment\">\/**<br \/>\n     * \u57fa\u4e8e\u77e5\u8bc6\u5e93\u56de\u7b54\u95ee\u9898<br \/>\n     *\/<\/span><br \/>\n    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">String<\/span> <span class=\"token function\">answer<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> question<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">int<\/span> topK<span class=\"token punctuation\">,<\/span> <span class=\"token keyword\">double<\/span> threshold<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token comment\">\/\/ 1. \u95ee\u9898\u5411\u91cf\u5316<\/span><br \/>\n        <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> queryVector <span class=\"token operator\">&#061;<\/span> <span class=\"token function\">embed<\/span><span class=\"token punctuation\">(<\/span>question<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token comment\">\/\/ 2. \u68c0\u7d22\u76f8\u5173\u7247\u6bb5<\/span><br \/>\n        <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">InMemoryVectorStore<span class=\"token punctuation\">.<\/span>SearchResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> results <span class=\"token operator\">&#061;<\/span> store<span class=\"token punctuation\">.<\/span><span class=\"token function\">search<\/span><span class=\"token punctuation\">(<\/span>queryVector<span class=\"token punctuation\">,<\/span> topK<span class=\"token punctuation\">,<\/span> threshold<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>results<span class=\"token punctuation\">.<\/span><span class=\"token function\">isEmpty<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            <span class=\"token keyword\">return<\/span> <span class=\"token string\">&#034;\u77e5\u8bc6\u5e93\u4e2d\u6ca1\u6709\u627e\u5230\u76f8\u5173\u5185\u5bb9&#xff0c;\u65e0\u6cd5\u56de\u7b54\u3002&#034;<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><\/p>\n<p>        <span class=\"token comment\">\/\/ 3. \u6784\u5efa Prompt<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> context <span class=\"token operator\">&#061;<\/span> <span class=\"token function\">buildContext<\/span><span class=\"token punctuation\">(<\/span>results<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> prompt <span class=\"token operator\">&#061;<\/span> <span class=\"token function\">buildRAGPrompt<\/span><span class=\"token punctuation\">(<\/span>question<span class=\"token punctuation\">,<\/span> context<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token comment\">\/\/ 4. \u8c03\u7528\u6a21\u578b\u751f\u6210\u56de\u7b54<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> chatModel<span class=\"token punctuation\">.<\/span><span class=\"token function\">call<\/span><span class=\"token punctuation\">(<\/span>prompt<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">float<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> <span class=\"token function\">embed<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> text<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token class-name\">EmbeddingResponse<\/span> response <span class=\"token operator\">&#061;<\/span> embeddingModel<span class=\"token punctuation\">.<\/span><span class=\"token function\">call<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">EmbeddingRequest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">List<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">of<\/span><span class=\"token punctuation\">(<\/span>text<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">EmbeddingOptionsBuilder<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">builder<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">build<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n        <span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> response<span class=\"token punctuation\">.<\/span><span class=\"token function\">getResult<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">getOutput<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">private<\/span> <span class=\"token class-name\">String<\/span> <span class=\"token function\">buildContext<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">InMemoryVectorStore<span class=\"token punctuation\">.<\/span>SearchResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> results<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token class-name\">StringBuilder<\/span> sb <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">StringBuilder<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token keyword\">for<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token keyword\">int<\/span> i <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span> i <span class=\"token operator\">&lt;<\/span> results<span class=\"token punctuation\">.<\/span><span class=\"token function\">size<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span> i<span class=\"token operator\">&#043;&#043;<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            sb<span class=\"token punctuation\">.<\/span><span class=\"token function\">append<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;[&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">append<\/span><span class=\"token punctuation\">(<\/span>i <span class=\"token operator\">&#043;<\/span> <span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">append<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;] &#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n              <span class=\"token punctuation\">.<\/span><span class=\"token function\">append<\/span><span class=\"token punctuation\">(<\/span>results<span class=\"token punctuation\">.<\/span><span class=\"token function\">get<\/span><span class=\"token punctuation\">(<\/span>i<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">chunk<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">content<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n              <span class=\"token punctuation\">.<\/span><span class=\"token function\">append<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;\\\\n\\\\n&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> sb<span class=\"token punctuation\">.<\/span><span class=\"token function\">toString<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token keyword\">private<\/span> <span class=\"token class-name\">String<\/span> <span class=\"token function\">buildRAGPrompt<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span> question<span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">String<\/span> context<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token keyword\">return<\/span> <span class=\"token triple-quoted-string string\">&#034;&#034;&#034;<br \/>\n            \u4f60\u662f\u4e00\u4e2a\u4e13\u4e1a\u7684\u95ee\u7b54\u52a9\u624b\u3002\u8bf7\u4e25\u683c\u57fa\u4e8e\u4ee5\u4e0b\u53c2\u8003\u8d44\u6599\u56de\u7b54\u7528\u6237\u7684\u95ee\u9898\u3002<br \/>\n            \u5982\u679c\u53c2\u8003\u8d44\u6599\u4e2d\u6ca1\u6709\u76f8\u5173\u4fe1\u606f&#xff0c;\u8bf7\u660e\u786e\u8bf4&#034;\u53c2\u8003\u8d44\u6599\u4e2d\u672a\u627e\u5230\u76f8\u5173\u5185\u5bb9&#034;&#xff0c;\u4e0d\u8981\u7f16\u9020\u7b54\u6848\u3002<br \/>\n            \u56de\u7b54\u65f6\u53ef\u4ee5\u5f15\u7528\u8d44\u6599\u7f16\u53f7&#xff0c;\u4f8b\u5982&#034;\u6839\u636e\u8d44\u6599[1]&#034;\u3002<\/p>\n<p>            \u53c2\u8003\u8d44\u6599&#xff1a;<br \/>\n            %s<\/p>\n<p>            \u7528\u6237\u95ee\u9898&#xff1a;%s<br \/>\n            &#034;&#034;&#034;<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">formatted<\/span><span class=\"token punctuation\">(<\/span>context<span class=\"token punctuation\">,<\/span> question<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h4>\u4f7f\u7528\u793a\u4f8b<\/h4>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">RAGDemo<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token keyword\">public<\/span> <span class=\"token keyword\">static<\/span> <span class=\"token keyword\">void<\/span> <span class=\"token function\">main<\/span><span class=\"token punctuation\">(<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">]<\/span> args<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token comment\">\/\/ \u521d\u59cb\u5316\u6a21\u578b&#xff08;\u5b9e\u9645\u4f7f\u7528\u65f6\u4ece Spring \u5bb9\u5668\u83b7\u53d6&#xff09;<\/span><br \/>\n        <span class=\"token class-name\">EmbeddingModel<\/span> embeddingModel <span class=\"token operator\">&#061;<\/span> <span class=\"token punctuation\">.<\/span><span class=\"token punctuation\">.<\/span><span class=\"token punctuation\">.<\/span><span class=\"token punctuation\">;<\/span> <span class=\"token comment\">\/\/ OpenAiEmbeddingModel \u6216\u5176\u4ed6<\/span><br \/>\n        <span class=\"token class-name\">ChatModel<\/span> chatModel <span class=\"token operator\">&#061;<\/span> <span class=\"token punctuation\">.<\/span><span class=\"token punctuation\">.<\/span><span class=\"token punctuation\">.<\/span><span class=\"token punctuation\">;<\/span>           <span class=\"token comment\">\/\/ OpenAiChatModel \u6216\u5176\u4ed6<\/span><\/p>\n<p>        <span class=\"token class-name\">MinimalRAG<\/span> rag <span class=\"token operator\">&#061;<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">MinimalRAG<\/span><span class=\"token punctuation\">(<\/span>embeddingModel<span class=\"token punctuation\">,<\/span> chatModel<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token comment\">\/\/ \u704c\u5165\u6587\u6863<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> doc <span class=\"token operator\">&#061;<\/span> <span class=\"token triple-quoted-string string\">&#034;&#034;&#034;<br \/>\n            Spring AI \u662f Spring \u751f\u6001\u4e2d\u7684 AI \u5e94\u7528\u5f00\u53d1\u6846\u67b6\u3002<br \/>\n            \u5b83\u63d0\u4f9b\u4e86\u5bf9\u591a\u79cd\u5927\u6a21\u578b\u7684\u62bd\u8c61&#xff0c;\u5305\u62ec ChatModel\u3001EmbeddingModel\u3001ImageModel \u7b49\u3002<\/p>\n<p>            \u5de5\u5177\u8c03\u7528&#xff08;Tool Calling&#xff09;\u662f Agent \u7684\u6838\u5fc3\u80fd\u529b\u4e4b\u4e00\u3002<br \/>\n            Spring AI \u901a\u8fc7 &#064;Tool \u6ce8\u89e3\u5c06 Java \u65b9\u6cd5\u66b4\u9732\u7ed9\u6a21\u578b\u8c03\u7528\u3002<br \/>\n            DefaultToolCallingManager \u8d1f\u8d23\u6267\u884c\u5de5\u5177\u8c03\u7528\u7684\u5b8c\u6574\u94fe\u8def&#xff0c;<br \/>\n            \u5305\u62ec\u5de5\u5177\u67e5\u627e\u3001\u53c2\u6570\u89e3\u6790\u3001\u5f02\u5e38\u5904\u7406\u548c\u8c03\u7528\u4e0a\u9650\u63a7\u5236\u3002<\/p>\n<p>            RAG&#xff08;\u68c0\u7d22\u589e\u5f3a\u751f\u6210&#xff09;\u662f\u4e00\u79cd\u5c06\u5916\u90e8\u77e5\u8bc6\u68c0\u7d22\u4e0e\u5927\u6a21\u578b\u751f\u6210\u7ed3\u5408\u7684\u6280\u672f\u3002<br \/>\n            \u5b83\u7684\u6838\u5fc3\u6d41\u7a0b\u5305\u62ec\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u5316\u3001\u5411\u91cf\u5b58\u50a8\u3001\u76f8\u4f3c\u5ea6\u68c0\u7d22\u548c Prompt \u62fc\u63a5\u3002<br \/>\n            RAG \u53ef\u4ee5\u6709\u6548\u89e3\u51b3\u5927\u6a21\u578b\u77e5\u8bc6\u6ede\u540e\u548c\u5e7b\u89c9\u95ee\u9898\u3002<br \/>\n            &#034;&#034;&#034;<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        rag<span class=\"token punctuation\">.<\/span><span class=\"token function\">ingestDocument<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;doc-spring-ai&#034;<\/span><span class=\"token punctuation\">,<\/span> doc<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token comment\">\/\/ \u63d0\u95ee<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> answer <span class=\"token operator\">&#061;<\/span> rag<span class=\"token punctuation\">.<\/span><span class=\"token function\">answer<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;Spring AI \u7684\u5de5\u5177\u8c03\u7528\u662f\u600e\u4e48\u5b9e\u73b0\u7684&#xff1f;&#034;<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">3<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">0.7<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token class-name\">System<\/span><span class=\"token punctuation\">.<\/span>out<span class=\"token punctuation\">.<\/span><span class=\"token function\">println<\/span><span class=\"token punctuation\">(<\/span>answer<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u5e38\u89c1\u7684\u5751\u548c\u4f18\u5316\u65b9\u5411<\/h3>\n<p>\u628a\u6700\u5c0f\u94fe\u8def\u8dd1\u901a\u53ea\u662f\u5f00\u59cb\u3002\u771f\u6b63\u843d\u5730\u65f6&#xff0c;\u8fd8\u6709\u51e0\u4e2a\u95ee\u9898\u5fc5\u987b\u9762\u5bf9&#xff1a;<\/p>\n<h4>1. \u68c0\u7d22\u4e0d\u51c6\u600e\u4e48\u529e<\/h4>\n<p>\u68c0\u7d22\u4e0d\u51c6\u662f RAG \u6700\u5e38\u89c1\u7684\u95ee\u9898\u3002\u53ef\u80fd\u7684\u539f\u56e0\u548c\u5bf9\u7b56&#xff1a;<\/p>\n<ul>\n<li>\u5207\u5206\u7c92\u5ea6\u4e0d\u5bf9&#xff1a;\u8c03\u5927\u6216\u8c03\u5c0f chunk size&#xff0c;\u68c0\u67e5\u6700\u76f8\u5173\u7684\u5185\u5bb9\u662f\u4e0d\u662f\u521a\u597d\u88ab\u5207\u6563\u4e86<\/li>\n<li>\u5d4c\u5165\u6a21\u578b\u4e0d\u5339\u914d&#xff1a;\u4e2d\u6587\u573a\u666f\u8bd5\u8bd5 BGE \u6216 M3E&#xff0c;\u4e0d\u4e00\u5b9a\u6bd4 OpenAI \u7684\u5dee<\/li>\n<li>\u53ea\u9760\u5411\u91cf\u68c0\u7d22\u4e0d\u591f&#xff1a;\u52a0\u4e0a\u5173\u952e\u8bcd\u68c0\u7d22&#xff08;BM25&#xff09;\u505a\u6df7\u5408\u68c0\u7d22&#xff0c;\u901a\u5e38\u80fd\u63d0\u5347 5-10 \u4e2a\u767e\u5206\u70b9<\/li>\n<li>\u52a0\u5165\u91cd\u6392&#xff08;Reranker&#xff09;&#xff1a;\u5148\u7528\u5411\u91cf\u53ec\u56de Top-20&#xff0c;\u518d\u7528\u4ea4\u53c9\u7f16\u7801\u5668\u91cd\u6392\u53d6 Top-3&#xff0c;\u6548\u679c\u63d0\u5347\u660e\u663e<\/li>\n<\/ul>\n<h4>2. \u6a21\u578b\u4e0d\u5f15\u7528\u4e0a\u4e0b\u6587\u600e\u4e48\u529e<\/h4>\n<p>\u6709\u65f6\u5019\u68c0\u7d22\u7ed3\u679c\u662f\u5bf9\u7684&#xff0c;\u4f46\u6a21\u578b\u4e0d\u7167\u7740\u56de\u7b54&#xff0c;\u800c\u662f\u7528\u81ea\u5df1\u53c2\u6570\u91cc\u7684\u77e5\u8bc6\u3002\u5bf9\u7b56&#xff1a;<\/p>\n<ul>\n<li>Prompt \u91cc\u5f3a\u5316&#034;\u53ea\u80fd\u57fa\u4e8e\u53c2\u8003\u8d44\u6599\u56de\u7b54&#034;\u7684\u7ea6\u675f<\/li>\n<li>\u628a\u53c2\u8003\u8d44\u6599\u653e\u5728\u66f4\u9192\u76ee\u7684\u4f4d\u7f6e&#xff0c;\u7528\u660e\u786e\u7684\u5206\u9694\u7b26<\/li>\n<li>\u7528\u66f4\u542c\u8bdd\u7684\u6a21\u578b&#xff08;\u6bd4\u5982 GPT-4o \u901a\u5e38\u6bd4\u5c0f\u6a21\u578b\u66f4\u542c\u8bdd&#xff09;<\/li>\n<li>\u5728 Few-shot \u793a\u4f8b\u91cc\u5c55\u793a\u6b63\u786e\u7684\u5f15\u7528\u65b9\u5f0f<\/li>\n<\/ul>\n<h4>3. \u4e0a\u4e0b\u6587\u592a\u957f\u600e\u4e48\u529e<\/h4>\n<p>\u6587\u6863\u591a\u4e86\u4ee5\u540e&#xff0c;\u68c0\u7d22\u7ed3\u679c\u53ef\u80fd\u5f88\u957f&#xff0c;\u8d85\u51fa Prompt \u7a97\u53e3\u6216\u9020\u6210\u6d6a\u8d39&#xff1a;<\/p>\n<ul>\n<li>\u51cf\u5c0f Top-K<\/li>\n<li>\u7528 MMR \u51cf\u5c11\u5197\u4f59<\/li>\n<li>\u5bf9\u68c0\u7d22\u7ed3\u679c\u505a\u6458\u8981\u538b\u7f29&#xff08;Map-Reduce \u6a21\u5f0f&#xff09;<\/li>\n<li>\u7528\u66f4\u5c0f\u7684 chunk size&#xff0c;\u4f46\u8981\u4fdd\u8bc1\u8bed\u4e49\u5b8c\u6574<\/li>\n<\/ul>\n<h4>4. \u5982\u4f55\u8bc4\u4f30 RAG \u6548\u679c<\/h4>\n<p>\u4e0d\u80fd\u9760&#034;\u611f\u89c9\u56de\u7b54\u5f97\u4e0d\u9519&#034;\u6765\u5224\u65ad\u3002\u81f3\u5c11\u8981\u6709\u51e0\u4e2a\u5b9a\u91cf\u6307\u6807&#xff1a;<\/p>\n<ul>\n<li>\u53ec\u56de\u7387&#xff08;Recall&#xff09;&#xff1a;\u6807\u51c6\u7b54\u6848\u6240\u5728\u7684\u5757\u6709\u6ca1\u6709\u88ab\u68c0\u7d22\u51fa\u6765<\/li>\n<li>\u51c6\u786e\u7387&#xff08;Precision&#xff09;&#xff1a;\u68c0\u7d22\u51fa\u6765\u7684\u5757\u6709\u591a\u5c11\u662f\u771f\u7684\u76f8\u5173<\/li>\n<li>\u56de\u7b54\u6b63\u786e\u7387&#xff1a;\u6700\u7ec8\u56de\u7b54\u5bf9\u4e0d\u5bf9&#xff0c;\u53ef\u4ee5\u7528 LLM-as-judge \u8f85\u52a9\u8bc4\u4f30<\/li>\n<li>\u5f15\u7528\u51c6\u786e\u7387&#xff1a;\u56de\u7b54\u4e2d\u5f15\u7528\u7684\u8d44\u6599\u662f\u4e0d\u662f\u771f\u7684\u652f\u6301\u8fd9\u4e2a\u7ed3\u8bba<\/li>\n<\/ul>\n<h3>\u5c0f\u7ed3<\/h3>\n<p>RAG \u7684\u539f\u7406\u4e0d\u590d\u6742&#xff0c;\u4f46\u6bcf\u4e2a\u73af\u8282\u90fd\u6709\u8bbe\u8ba1\u7a7a\u95f4\u3002\u628a\u6700\u5c0f\u94fe\u8def\u624b\u5199\u4e00\u904d\u7684\u4ef7\u503c\u5728\u4e8e&#xff1a;\u4f60\u77e5\u9053\u6bcf\u4e2a\u53c2\u6570\u5728\u5f71\u54cd\u4ec0\u4e48&#xff0c;\u51fa\u4e86\u95ee\u9898\u77e5\u9053\u4ece\u54ea\u67e5\u3002<\/p>\n<p>\u8fd9\u7bc7\u6587\u7ae0\u7684\u6700\u5c0f\u5b9e\u73b0\u7528\u4e86\u5185\u5b58\u5b58\u50a8\u548c\u57fa\u7840\u76f8\u4f3c\u5ea6\u68c0\u7d22&#xff0c;\u9002\u5408\u5b66\u4e60\u548c\u539f\u578b\u9a8c\u8bc1\u3002\u8d70\u5411\u751f\u4ea7\u7684\u8def\u5f84\u5927\u81f4\u662f&#xff1a;<\/p>\n<li>\u5148\u8dd1\u901a\u6700\u5c0f\u94fe\u8def&#xff0c;\u786e\u8ba4 RAG \u5bf9\u4f60\u7684\u573a\u666f\u6709\u4ef7\u503c<\/li>\n<li>\u63a5\u5165\u771f\u5b9e\u5411\u91cf\u6570\u636e\u5e93&#xff0c;\u89e3\u51b3\u89c4\u6a21\u95ee\u9898<\/li>\n<li>\u52a0\u5165\u6df7\u5408\u68c0\u7d22\u548c\u91cd\u6392&#xff0c;\u63d0\u5347\u53ec\u56de\u8d28\u91cf<\/li>\n<li>\u5efa\u7acb\u8bc4\u6d4b\u96c6&#xff0c;\u7528\u91cf\u5316\u6307\u6807\u9a71\u52a8\u4f18\u5316<\/li>\n<li>\u9010\u6b65\u52a0\u5165\u67e5\u8be2\u6539\u5199\u3001\u591a\u8f6e\u5bf9\u8bdd\u3001\u5143\u6570\u636e\u8fc7\u6ee4\u7b49\u9ad8\u7ea7\u7279\u6027<\/li>\n<p>\u6bcf\u4e00\u6b65\u90fd\u6709\u5bf9\u5e94\u7684\u6846\u67b6\u548c\u5de5\u5177&#xff0c;\u4f46\u6838\u5fc3\u73af\u8282\u6c38\u8fdc\u662f\u90a3\u4e94\u4e2a&#xff1a;\u5207\u5206\u3001\u5d4c\u5165\u3001\u5b58\u50a8\u3001\u68c0\u7d22\u3001\u751f\u6210\u3002\u7406\u89e3\u4e86\u5b83\u4eec&#xff0c;\u6362\u4efb\u4f55\u6846\u67b6\u90fd\u53ea\u662f\u6362 API\u3002<\/p>\n<hr \/>\n<p>\u6807\u7b7e&#xff1a;Java\u3001RAG\u3001Spring AI\u3001\u5411\u91cf\u68c0\u7d22\u3001\u5d4c\u5165\u6a21\u578b\u3001Agent<\/p>\n<p>\u53c2\u8003\u8d44\u6599&#xff1a;<\/p>\n<ul>\n<li>Spring AI \u5b98\u65b9\u6587\u6863&#xff1a;https:\/\/docs.spring.io\/spring-ai\/reference\/<\/li>\n<li>Azure Search OpenAI Demo (Java)&#xff1a;https:\/\/learn.microsoft.com\/en-us\/samples\/azure-samples\/azure-search-openai-demo-java\/<\/li>\n<li>A Survey on Retrieval-Augmented Generation for Large Language Models<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG&#xff1a;\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def\u73af\u5883&#xff1a;Windows 11\u3001JDK 17.0.12\u3001Maven 3.9.14<br \/>\n\u4f9d\u8d56&#xff1a;Spring AI 2.0.1&#xff08;\u5d4c\u5165\u6a21\u578b\u8c03\u7528&#xff09;\u3001H2 \u672c\u5730\u5411\u91cf\u5b58\u50a8&#xff08;\u53ef\u9009&#xff0c;\u672c\u6587\u7528\u5185\u5b58 List \u5b9e\u73b0&#xff09;<br \/>\n\u8bf4\u660e&#xff1a;\u672c\u6587\u91cd\u70b9\u662f\u628a RAG \u7684\u6bcf\u4e2a\u73af\u8282\u62c6\u5f00\u8bb2\u6e05\u695a&#xff0c;\u4ee3\u7801\u53ef\u76f4\u63a5\u8dd1\u901a<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[11820,99,2946,190],"topic":[],"class_list":["post-115255","post","type-post","status-publish","format-standard","hentry","category-server","tag-atomgit","tag-java","tag-prompt","tag-190"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/115255.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG&#xff1a;\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def\u73af\u5883&#xff1a;Windows 11\u3001JDK 17.0.12\u3001Maven 3.9.14 \u4f9d\u8d56&#xff1a;Spring AI 2.0.1&#xff08;\u5d4c\u5165\u6a21\u578b\u8c03\u7528&#xff09;\u3001H2 \u672c\u5730\u5411\u91cf\u5b58\u50a8&#xff08;\u53ef\u9009&#xff0c;\u672c\u6587\u7528\u5185\u5b58 List \u5b9e\u73b0&#xff09; \u8bf4\u660e&#xff1a;\u672c\u6587\u91cd\u70b9\u662f\u628a RAG \u7684\u6bcf\u4e2a\u73af\u8282\u62c6\u5f00\u8bb2\u6e05\u695a&#xff0c;\u4ee3\u7801\u53ef\u76f4\u63a5\u8dd1\u901a\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/115255.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-11T08:40:55+00:00\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/115255.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/115255.html\",\"name\":\"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-10-11T08:40:55+00:00\",\"dateModified\":\"2026-10-11T08:40:55+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/115255.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/115255.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/115255.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/115255.html","og_locale":"zh_CN","og_type":"article","og_title":"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG&#xff1a;\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def\u73af\u5883&#xff1a;Windows 11\u3001JDK 17.0.12\u3001Maven 3.9.14 \u4f9d\u8d56&#xff1a;Spring AI 2.0.1&#xff08;\u5d4c\u5165\u6a21\u578b\u8c03\u7528&#xff09;\u3001H2 \u672c\u5730\u5411\u91cf\u5b58\u50a8&#xff08;\u53ef\u9009&#xff0c;\u672c\u6587\u7528\u5185\u5b58 List \u5b9e\u73b0&#xff09; \u8bf4\u660e&#xff1a;\u672c\u6587\u91cd\u70b9\u662f\u628a RAG \u7684\u6bcf\u4e2a\u73af\u8282\u62c6\u5f00\u8bb2\u6e05\u695a&#xff0c;\u4ee3\u7801\u53ef\u76f4\u63a5\u8dd1\u901a","og_url":"https:\/\/www.wsisp.com\/helps\/115255.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-10-11T08:40:55+00:00","author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"8 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/115255.html","url":"https:\/\/www.wsisp.com\/helps\/115255.html","name":"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-10-11T08:40:55+00:00","dateModified":"2026-10-11T08:40:55+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/115255.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/115255.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/115255.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u7528 Java \u624b\u5199\u4e00\u4e2a\u6700\u5c0f RAG\uff1a\u4ece\u6587\u6863\u5207\u5206\u3001\u5411\u91cf\u68c0\u7d22\u5230 Prompt \u62fc\u63a5\u7684\u5b8c\u6574\u94fe\u8def"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/115255","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=115255"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/115255\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=115255"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=115255"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=115255"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=115255"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}