{"id":91588,"date":"2026-08-08T01:48:44","date_gmt":"2026-08-07T17:48:44","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/91588.html"},"modified":"2026-08-08T01:48:44","modified_gmt":"2026-08-07T17:48:44","slug":"spring-ai%e4%bc%81%e4%b8%9a%e7%ba%a7%e5%ba%94%e7%94%a8%e5%ae%9e%e6%88%98%ef%bc%8813%ef%bc%89%ef%bc%9a%e6%b5%8b%e8%af%95%e3%80%81%e9%bb%84%e9%87%91%e6%95%b0%e6%8d%ae%e9%9b%86%e4%b8%8e%e5%8f%91%e5%b8%83","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/91588.html","title":{"rendered":"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981"},"content":{"rendered":"<h3>\u6587\u7ae0\u6458\u8981<\/h3>\n<p>\u524d\u9762\u7684\u7ae0\u8282\u5df2\u7ecf\u5b8c\u6210\u591a\u6a21\u578b\u8def\u7531\u3001\u9884\u7b97\u63a7\u5236\u3001\u6743\u9650\u611f\u77e5RAG\u548c\u5f02\u6b65\u957f\u4efb\u52a1\u3002\u7cfb\u7edf\u53ef\u4ee5\u8c03\u7528\u6a21\u578b\u3001\u68c0\u7d22\u4f01\u4e1a\u77e5\u8bc6\u3001\u6267\u884c\u5de5\u5177\u5e76\u5728\u6545\u969c\u540e\u6062\u590d&#xff0c;\u4f46\u4ecd\u7f3a\u5c11\u4e00\u4e2a\u51b3\u5b9a\u80fd\u5426\u7a33\u5b9a\u8fed\u4ee3\u7684\u6838\u5fc3\u80fd\u529b&#xff1a;\u5982\u4f55\u8bc1\u660e\u4e00\u6b21Prompt\u3001\u6a21\u578b\u3001RAG\u7d22\u5f15\u3001Advisor\u6216Tool\u76ee\u5f55\u53d8\u66f4\u6ca1\u6709\u7834\u574f\u73b0\u6709\u4e1a\u52a1\u3002<\/p>\n<p>\u4f20\u7edf\u5355\u5143\u6d4b\u8bd5\u80fd\u591f\u9a8c\u8bc1\u72b6\u6001\u7801\u3001Schema\u548c\u4e1a\u52a1\u89c4\u5219&#xff0c;\u5374\u65e0\u6cd5\u5b8c\u6574\u5224\u65ad\u5f00\u653e\u5f0f\u56de\u7b54\u662f\u5426\u66f4\u597d&#xff1b;\u53ea\u4f9d\u8d56LLM-as-Judge\u53c8\u4f1a\u53d7\u5230\u968f\u673a\u6027\u3001\u4f4d\u7f6e\u504f\u5dee\u3001\u957f\u5ea6\u504f\u5dee\u548cJudge\u7248\u672c\u6f02\u79fb\u5f71\u54cd\u3002\u751f\u4ea7\u7ea7\u65b9\u6848\u5fc5\u987b\u5efa\u7acb\u5206\u5c42\u6d4b\u8bd5\u4f53\u7cfb&#xff1a;\u786e\u5b9a\u6027\u89c4\u5219\u8d1f\u8d23\u6743\u9650\u3001Schema\u3001\u6570\u5b57\u3001\u72b6\u6001\u673a\u548c\u5de5\u5177\u526f\u4f5c\u7528&#xff1b;Spring AI Evaluator\u8d1f\u8d23\u76f8\u5173\u6027\u548c\u4e8b\u5b9e\u652f\u6301&#xff1b;\u7ecf\u8fc7\u6821\u51c6\u7684Judge\u8d1f\u8d23\u5f00\u653e\u8d28\u91cf&#xff1b;\u9ec4\u91d1\u6570\u636e\u96c6\u548c\u751f\u4ea7Trace\u8d1f\u8d23\u771f\u5b9e\u8986\u76d6&#xff1b;Quality Gate\u8d1f\u8d23\u5728CI\/CD\u4e2d\u963b\u65ad\u9ad8\u98ce\u9669\u9000\u5316\u3002<\/p>\n<p>\u672c\u7ae0\u5c06\u5728\u73b0\u6709Spring AI\u5de5\u7a0b\u4e2d\u589e\u52a0ai-evaluation\u6a21\u5757&#xff0c;\u6784\u5efa\u7248\u672c\u5316\u6570\u636e\u96c6\u3001Release Manifest\u3001Baseline\/Candidate\u53cc\u8dd1\u3001\u91cd\u590d\u91c7\u6837\u3001RAG\u4e0eAgent\u8f68\u8ff9\u8bc4\u6d4b\u3001Slice\u6307\u6807\u3001GitHub Actions\u72b6\u6001\u68c0\u67e5\u3001\u53d1\u5e03\u540eCanary\u548c\u7ebf\u4e0a\u5931\u8d25\u56de\u6d41\u3002<\/p>\n<p>\u5b8c\u6210\u672c\u7ae0\u540e&#xff0c;\u6bcf\u6b21\u6a21\u578b\u6216Prompt\u5347\u7ea7\u90fd\u5fc5\u987b\u56de\u7b54&#xff1a;<\/p>\n<p>\u6539\u4e86\u4ec0\u4e48&#xff1f;<br \/>\n\u5728\u54ea\u4e9b\u6837\u672c\u4e0a\u53d8\u597d&#xff1f;<br \/>\n\u5728\u54ea\u4e9bSlice\u4e0a\u9000\u5316&#xff1f;<br \/>\n\u662f\u5426\u51fa\u73b0\u786c\u5931\u8d25&#xff1f;<br \/>\n\u6210\u672c\u548c\u5ef6\u8fdf\u53d8\u5316\u591a\u5c11&#xff1f;<br \/>\n\u80fd\u5426\u53d1\u5e03&#xff1f;<\/p>\n<h3>\u4e00\u3001\u672c\u7ae0\u4e0e\u524d\u9762\u80fd\u529b\u7684\u5173\u7cfb<\/h3>\n<p>\u73b0\u6709\u8c03\u7528\u94fe&#xff1a;<\/p>\n<p>\u4e1a\u52a1\u8bf7\u6c42<br \/>\n\u2193<br \/>\n\u6a21\u578b\u8def\u7531\u4e0e\u9884\u7b97<br \/>\n\u2193<br \/>\n\u6743\u9650\u611f\u77e5RAG<br \/>\n\u2193<br \/>\nChatClient \/ Tool Calling<br \/>\n\u2193<br \/>\n\u5f02\u6b65\u4efb\u52a1\u4e0eCheckpoint<br \/>\n\u2193<br \/>\n\u7ed3\u679c<\/p>\n<p>\u672c\u7ae0\u589e\u52a0\u4e00\u6761\u5e73\u884c\u8d28\u91cf\u94fe&#xff1a;<\/p>\n<p>Release Manifest<br \/>\n&#xff0b;<br \/>\n\u7248\u672c\u5316\u6570\u636e\u96c6<br \/>\n\u2193<br \/>\nBaseline\/Candidate\u6267\u884c<br \/>\n\u2193<br \/>\n\u786e\u5b9a\u6027Evaluator<br \/>\n\u2193<br \/>\nSpring AI Evaluator<br \/>\n\u2193<br \/>\nLLM Judge<br \/>\n\u2193<br \/>\nSlice\u805a\u5408<br \/>\n\u2193<br \/>\nQuality Gate<br \/>\n\u2193<br \/>\n\u5408\u5e76\u3001\u53d1\u5e03\u6216\u963b\u65ad<\/p>\n<p>\u8bc4\u6d4b\u4e0d\u662f\u4e1a\u52a1\u8c03\u7528\u4e2d\u7684\u4e00\u4e2aAdvisor&#xff0c;\u800c\u662f\u72ec\u7acb\u63a7\u5236\u9762\u3002<\/p>\n<h3>\u4e8c\u3001\u672c\u7ae0\u76ee\u6807<\/h3>\n<p>1. \u56fa\u5b9a\u6a21\u578b\u3001Prompt\u3001\u7d22\u5f15\u4e0e\u5de5\u5177\u8fd0\u884c\u6e05\u5355<br \/>\n2. \u5efa\u7acb\u5f00\u53d1\u96c6\u3001\u56de\u5f52\u96c6\u3001\u5b89\u5168\u96c6\u548c\u9690\u85cf\u9a8c\u6536\u96c6<br \/>\n3. \u4f7f\u7528Fake\u6a21\u578b\u5b8c\u6210\u786e\u5b9a\u6027\u5355\u5143\u6d4b\u8bd5<br \/>\n4. \u4f7f\u7528\u771f\u5b9eProvider\u6267\u884c\u79bb\u7ebf\u8d28\u91cf\u8bc4\u6d4b<br \/>\n5. \u91cd\u590d\u8fd0\u884c\u4f30\u8ba1\u6a21\u578b\u968f\u673a\u6027<br \/>\n6. \u5206\u522b\u8bc4\u6d4bRAG\u68c0\u7d22\u3001Groundedness\u548c\u5f15\u7528<br \/>\n7. \u8bc4\u6d4bTool Calling\u8f68\u8ff9\u4e0e\u526f\u4f5c\u7528<br \/>\n8. \u4f7f\u7528Spring AI RelevancyEvaluator\u548cFactCheckingEvaluator<br \/>\n9. \u4f7f\u7528\u7ecf\u8fc7\u4f4d\u7f6e\u5e73\u8861\u548c\u4eba\u5de5\u6821\u51c6\u7684LLM Judge<br \/>\n10. \u6309\u4e1a\u52a1Slice\u805a\u5408\u6307\u6807<br \/>\n11. \u9ad8\u98ce\u9669\u9519\u8bef\u96f6\u5bb9\u5fcd<br \/>\n12. CI\/CD\u751f\u6210Required Status Check<br \/>\n13. \u53d1\u5e03\u540e\u8fd0\u884cCanary\u4e0e\u7ebf\u4e0a\u62bd\u6837<br \/>\n14. \u5c06\u751f\u4ea7\u4e8b\u6545\u56de\u6d41\u4e3a\u6c38\u4e45\u56de\u5f52Case<\/p>\n<h3>\u4e09\u3001\u6d4b\u8bd5\u91d1\u5b57\u5854<\/h3>\n<p>                    \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510<br \/>\n                    \u2502 Human Acceptance    \u2502<br \/>\n                    \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<br \/>\n                 \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510<br \/>\n                 \u2502 Shadow \/ Online Canary   \u2502<br \/>\n                 \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<br \/>\n              \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510<br \/>\n              \u2502 Offline Quality Evaluation    \u2502<br \/>\n              \u2502 Dataset \/ Judge \/ RAG \/ Agent \u2502<br \/>\n              \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<br \/>\n           \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510<br \/>\n           \u2502 Integration &amp; Contract Tests       \u2502<br \/>\n           \u2502 Provider \/ VectorStore \/ Tools     \u2502<br \/>\n           \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<br \/>\n        \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510<br \/>\n        \u2502 Deterministic Unit Tests                \u2502<br \/>\n        \u2502 Rules \/ Schema \/ State \/ Cost \/ ACL     \u2502<br \/>\n        \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<\/p>\n<p>\u5e95\u5c42\u6570\u91cf\u591a\u3001\u901f\u5ea6\u5feb\u3001\u6210\u672c\u4f4e\u3002<\/p>\n<p>\u8d8a\u5411\u4e0a&#xff1a;<\/p>\n<ul>\n<li>\u6837\u672c\u66f4\u771f\u5b9e&#xff1b;<\/li>\n<li>\u6210\u672c\u66f4\u9ad8&#xff1b;<\/li>\n<li>\u8fd0\u884c\u66f4\u6162&#xff1b;<\/li>\n<li>\u7ed3\u8bba\u66f4\u63a5\u8fd1\u751f\u4ea7\u4f53\u9a8c\u3002<\/li>\n<\/ul>\n<h3>\u56db\u3001\u4e3a\u4ec0\u4e48\u4e0d\u80fd\u53ea\u4f9d\u8d56JUnit<\/h3>\n<p>JUnit\u975e\u5e38\u9002\u5408&#xff1a;<\/p>\n<p><span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span>result<span class=\"token punctuation\">.<\/span><span class=\"token function\">status<\/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\">isEqualTo<\/span><span class=\"token punctuation\">(<\/span><span class=\"token constant\">SUCCESS<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>\u4f46\u5f00\u653e\u95ee\u7b54\u901a\u5e38\u6ca1\u6709\u552f\u4e00\u5b57\u7b26\u4e32\u3002<\/p>\n<p>\u4ee5\u4e0b\u56de\u7b54\u53ef\u80fd\u90fd\u6b63\u786e&#xff1a;<\/p>\n<p>\u6e20\u9053\u52a8\u9500\u662f\u5546\u54c1\u4ece\u6e20\u9053\u5e93\u5b58\u8f6c\u5316\u4e3a\u7ec8\u7aef\u9500\u552e\u7684\u8fc7\u7a0b\u3002<\/p>\n<p>\u6e20\u9053\u52a8\u9500\u5173\u6ce8\u8d27\u7269\u662f\u5426\u771f\u6b63\u4ece\u7ecf\u9500\u548c\u7ec8\u7aef\u73af\u8282\u88ab\u6d88\u8d39\u8005\u8d2d\u4e70\u3002<\/p>\n<p>\u56e0\u6b64\u9700\u8981&#xff1a;<\/p>\n<p>\u786e\u5b9a\u6027\u5951\u7ea6<br \/>\n&#xff0b;\u8bed\u4e49\u8d28\u91cf\u8bc4\u6d4b<\/p>\n<h3>\u4e94\u3001\u4e3a\u4ec0\u4e48\u4e0d\u80fd\u53ea\u4f9d\u8d56LLM Judge<\/h3>\n<p>Judge\u4e0d\u9002\u5408\u66ff\u4ee3&#xff1a;<\/p>\n<ul>\n<li>JSON Schema&#xff1b;<\/li>\n<li>\u7cbe\u786e\u91d1\u989d&#xff1b;<\/li>\n<li>\u89d2\u8272\u6743\u9650&#xff1b;<\/li>\n<li>\u79df\u6237\u9694\u79bb&#xff1b;<\/li>\n<li>\u5de5\u5177\u6b21\u6570&#xff1b;<\/li>\n<li>\u72b6\u6001\u8fc1\u79fb&#xff1b;<\/li>\n<li>\u5e42\u7b49\u952e&#xff1b;<\/li>\n<li>SLA&#xff1b;<\/li>\n<li>\u6210\u672c\u9884\u7b97\u3002<\/li>\n<\/ul>\n<p>\u8fd9\u4e9b\u90fd\u6709\u660e\u786e\u771f\u503c&#xff0c;\u4f7f\u7528\u4ee3\u7801\u66f4\u53ef\u9760\u3002<\/p>\n<p>\u6b63\u786e\u987a\u5e8f&#xff1a;<\/p>\n<p>\u786e\u5b9a\u6027\u89c4\u5219\u5931\u8d25<br \/>\n\u2192 \u76f4\u63a5\u5931\u8d25<\/p>\n<p>\u786e\u5b9a\u6027\u89c4\u5219\u901a\u8fc7<br \/>\n\u2192 \u518d\u505a\u8bed\u4e49Judge<\/p>\n<h3>\u516d\u3001\u6a21\u5757\u7ed3\u6784<\/h3>\n<p>enterprise-ai-platform<br \/>\n\u251c\u2500\u2500 ai-core<br \/>\n\u251c\u2500\u2500 ai-routing<br \/>\n\u251c\u2500\u2500 ai-rag<br \/>\n\u251c\u2500\u2500 ai-task-center<br \/>\n\u251c\u2500\u2500 ai-evaluation<br \/>\n\u2502   \u251c\u2500\u2500 dataset<br \/>\n\u2502   \u251c\u2500\u2500 fixture<br \/>\n\u2502   \u251c\u2500\u2500 target<br \/>\n\u2502   \u251c\u2500\u2500 evaluator<br \/>\n\u2502   \u251c\u2500\u2500 runner<br \/>\n\u2502   \u251c\u2500\u2500 metric<br \/>\n\u2502   \u251c\u2500\u2500 gate<br \/>\n\u2502   \u251c\u2500\u2500 report<br \/>\n\u2502   \u2514\u2500\u2500 integration<br \/>\n\u2514\u2500\u2500 ai-application<\/p>\n<p>\u8bc4\u6d4b\u6a21\u5757\u4f9d\u8d56\u4e1a\u52a1\u63a5\u53e3&#xff0c;\u4f46\u751f\u4ea7\u4e1a\u52a1\u6a21\u5757\u4e0d\u4f9d\u8d56\u8bc4\u6d4b\u5b9e\u73b0\u3002<\/p>\n<h3>\u4e03\u3001Maven Profile<\/h3>\n<p><span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>profiles<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>profile<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>id<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>ai-eval-fast<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>id<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>properties<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n            <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>ai.eval.dataset<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n                pr-fast&#064;2026.08.06-v1<br \/>\n            <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>ai.eval.dataset<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n            <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>ai.eval.repetitions<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>1<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>ai.eval.repetitions<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>properties<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>profile<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><\/p>\n<p>    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>profile<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>id<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>ai-eval-full<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>id<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>properties<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n            <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>ai.eval.dataset<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n                full-regression&#064;2026.08.06-v3<br \/>\n            <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>ai.eval.dataset<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n            <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;<\/span>ai.eval.repetitions<\/span><span class=\"token punctuation\">&gt;<\/span><\/span>3<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>ai.eval.repetitions<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n        <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>properties<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n    <span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>profile<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><br \/>\n<span class=\"token tag\"><span class=\"token tag\"><span class=\"token punctuation\">&lt;\/<\/span>profiles<\/span><span class=\"token punctuation\">&gt;<\/span><\/span><\/p>\n<h3>\u516b\u3001Release Manifest<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">AiReleaseManifest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> releaseId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> gitCommit<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> modelRoutingPolicyVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Map<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">ModelProfileManifest<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> models<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">PromptManifest<\/span> prompt<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> knowledgeIndexVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> embeddingProfileVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> rerankerProfileVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> toolCatalogVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> memoryPolicyVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> outputSchemaVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> applicationConfigHash<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>Prompt Manifest&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">PromptManifest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> systemPromptVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> templateVersion<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> advisorVersions<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> securityPolicyVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> hash<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u6a21\u578bManifest&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">ModelProfileManifest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> profileId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> provider<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> requestedModel<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> resolvedModel<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> endpointRegion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Map<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">Object<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> options<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u4e5d\u3001\u4e3a\u4ec0\u4e48Release Manifest\u5fc5\u987b\u7531\u8fd0\u884c\u65f6\u56de\u663e<\/h3>\n<p>\u914d\u7f6e\u6587\u4ef6\u5199&#xff1a;<\/p>\n<p>knowledgeIndexVersion&#061;v42<\/p>\n<p>\u4e0d\u4ee3\u8868VectorStore\u8fde\u63a5\u7684\u4e00\u5b9a\u662fV42\u3002<\/p>\n<p>\u5019\u9009\u670d\u52a1\u63d0\u4f9b\u5185\u90e8\u7aef\u70b9&#xff1a;<\/p>\n<p>GET \/internal\/ai-runtime-manifest<\/p>\n<p>\u8fd4\u56de\u5b9e\u9645\u52a0\u8f7d\u503c\u3002<\/p>\n<p>\u8bc4\u6d4bRunner\u5728\u6bcf\u4e2aRun\u5f00\u59cb\u524d\u6bd4\u8f83&#xff1a;<\/p>\n<p>\u58f0\u660eManifest<br \/>\n\u4e0e<br \/>\n\u8fd0\u884c\u65f6Manifest<\/p>\n<p>\u4e0d\u4e00\u81f4\u76f4\u63a5\u505c\u6b62\u3002<\/p>\n<h3>\u5341\u3001\u6570\u636e\u96c6\u5206\u5c42<\/h3>\n<p>dev<br \/>\nregression<br \/>\nsecurity<br \/>\nincident<br \/>\nhidden-acceptance<br \/>\nproduction-replay<br \/>\ncanary<\/p>\n<h4>Dev<\/h4>\n<p>\u5f00\u53d1\u4eba\u5458\u53ef\u89c1&#xff0c;\u8c03\u8bd5Prompt\u3002<\/p>\n<h4>Regression<\/h4>\n<p>\u5386\u53f2\u7a33\u5b9a\u80fd\u529b\u3002<\/p>\n<h4>Security<\/h4>\n<p>\u8d8a\u6743\u3001\u6ce8\u5165\u548c\u526f\u4f5c\u7528&#xff0c;\u96f6\u5bb9\u5fcd\u3002<\/p>\n<h4>Incident<\/h4>\n<p>\u751f\u4ea7\u4e8b\u6545\u6c38\u4e45\u56de\u5f52\u3002<\/p>\n<h4>Hidden Acceptance<\/h4>\n<p>\u53d1\u5e03\u524d\u9690\u85cf\u9a8c\u6536\u3002<\/p>\n<h4>Production Replay<\/h4>\n<p>\u771f\u5b9eTrace\u8131\u654f\u56de\u653e\u3002<\/p>\n<h4>Canary<\/h4>\n<p>\u90e8\u7f72\u540e\u5feb\u901f\u9a8c\u8bc1\u3002<\/p>\n<h3>\u5341\u4e00\u3001Evaluation Case<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">AiEvaluationCase<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> caseId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> caseVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> datasetId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> datasetVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> taskType<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">RiskLevel<\/span> riskLevel<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">UserFixture<\/span> user<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\">ConversationTurn<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> history<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">JsonNode<\/span> input<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">ExpectedBehavior<\/span> expected<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> tags<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">CaseProvenance<\/span> provenance<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u5341\u4e8c\u3001Expected Behavior<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">ExpectedBehavior<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">JsonNode<\/span> exactOutput<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\">ExpectedFact<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> requiredFacts<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> forbiddenFacts<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> requiredDocumentIds<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> forbiddenDocumentIds<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">ExpectedToolTrace<\/span> toolTrace<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">RefusalExpectation<\/span> refusal<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">OutputContract<\/span> outputContract<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">QualityRubric<\/span> rubric<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">OperationalBudget<\/span> budget<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u5341\u4e09\u3001Case\u6765\u6e90<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">enum<\/span> <span class=\"token class-name\">EvaluationSource<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token constant\">HUMAN_GOLDEN<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">SYNTHETIC<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">PRODUCTION_TRACE<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">INCIDENT<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">SECURITY_RESEARCH<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u751f\u4ea7Trace\u8fdb\u5165\u6570\u636e\u96c6\u524d\u5fc5\u987b&#xff1a;<\/p>\n<ul>\n<li>\u6388\u6743&#xff1b;<\/li>\n<li>\u8131\u654f&#xff1b;<\/li>\n<li>\u6700\u5c0f\u5316&#xff1b;<\/li>\n<li>\u4e13\u5bb6\u6807\u6ce8&#xff1b;<\/li>\n<li>\u7248\u672c\u51bb\u7ed3\u3002<\/li>\n<\/ul>\n<h3>\u5341\u56db\u3001\u786e\u5b9a\u6027\u6d4b\u8bd5\u4e00&#xff1a;\u8def\u7531<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Test<\/span><br \/>\n<span class=\"token keyword\">void<\/span> <span class=\"token function\">highRiskContractMustUsePowerfulModel<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">RouteDecision<\/span> decision <span class=\"token operator\">&#061;<\/span><br \/>\n            routingPolicy<span class=\"token punctuation\">.<\/span><span class=\"token function\">route<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    fixture<span class=\"token punctuation\">.<\/span><span class=\"token function\">highRiskContract<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            decision<span class=\"token punctuation\">.<\/span><span class=\"token function\">primaryModel<\/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\">isEqualTo<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token class-name\">ModelId<\/span><span class=\"token punctuation\">.<\/span><span class=\"token constant\">POWERFUL_PRIMARY<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u5341\u4e94\u3001\u786e\u5b9a\u6027\u6d4b\u8bd5\u4e8c&#xff1a;\u6743\u9650\u8fc7\u6ee4<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Test<\/span><br \/>\n<span class=\"token keyword\">void<\/span> <span class=\"token function\">tenantFilterMustBePresent<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">SearchRequest<\/span> request <span class=\"token operator\">&#061;<\/span><br \/>\n            searchRequestFactory<span class=\"token punctuation\">.<\/span><span class=\"token function\">create<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token string\">&#034;\u5408\u540c\u98ce\u9669&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    fixture<span class=\"token punctuation\">.<\/span><span class=\"token function\">access<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;T001&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token string\">&#034;KB1&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token number\">20<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token number\">0.6<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            request<span class=\"token punctuation\">.<\/span><span class=\"token function\">getFilterExpression<\/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\">contains<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token string\">&#034;tenant_id &#061;&#061; &#039;T001&#039;&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u66f4\u91cd\u8981\u7684\u662f\u7528Fake VectorStore\u9a8c\u8bc1\u7ed3\u679c\u4e2d\u7edd\u4e0d\u51fa\u73b0T002\u3002<\/p>\n<h3>\u5341\u516d\u3001\u786e\u5b9a\u6027\u6d4b\u8bd5\u4e09&#xff1a;\u7ed3\u6784\u5316\u8f93\u51fa<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Test<\/span><br \/>\n<span class=\"token keyword\">void<\/span> <span class=\"token function\">extractionSchemaMustRejectMissingAmount<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">JsonNode<\/span> output <span class=\"token operator\">&#061;<\/span><br \/>\n            objectMapper<span class=\"token punctuation\">.<\/span><span class=\"token function\">readTree<\/span><span class=\"token punctuation\">(<\/span><span class=\"token triple-quoted-string string\">&#034;&#034;&#034;<br \/>\n                    {<br \/>\n                      &#034;currency&#034;: &#034;CNY&#034;<br \/>\n                    }<br \/>\n                    &#034;&#034;&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token class-name\">ValidationReport<\/span> report <span class=\"token operator\">&#061;<\/span><br \/>\n            schemaValidator<span class=\"token punctuation\">.<\/span><span class=\"token function\">validate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    contractExtractionSchema<span class=\"token punctuation\">,<\/span><br \/>\n                    output<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span>report<span class=\"token punctuation\">.<\/span><span class=\"token function\">valid<\/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\">isFalse<\/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<h3>\u5341\u4e03\u3001\u786e\u5b9a\u6027\u6d4b\u8bd5\u56db&#xff1a;Tool\u8f68\u8ff9<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Test<\/span><br \/>\n<span class=\"token keyword\">void<\/span> <span class=\"token function\">refundToolMustNotBeCalledTwice<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">ToolTrace<\/span> trace <span class=\"token operator\">&#061;<\/span><br \/>\n            fixture<span class=\"token punctuation\">.<\/span><span class=\"token function\">trace<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token string\">&#034;queryOrder&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token string\">&#034;createRefund&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token string\">&#034;createRefund&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token class-name\">ToolTraceValidation<\/span> result <span class=\"token operator\">&#061;<\/span><br \/>\n            toolTraceEvaluator<span class=\"token punctuation\">.<\/span><span class=\"token function\">validate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    expectedRefundTrace<span class=\"token punctuation\">,<\/span><br \/>\n                    trace<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span>result<span class=\"token punctuation\">.<\/span><span class=\"token function\">valid<\/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\">isFalse<\/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<h3>\u5341\u516b\u3001Fake ChatModel<\/h3>\n<p>\u7ec4\u4ef6\u6d4b\u8bd5\u4e0d\u5e94\u6bcf\u6b21\u8c03\u7528\u771f\u5b9eProvider\u3002<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">ScriptedChatModel<\/span><br \/>\n        <span class=\"token keyword\">implements<\/span> <span class=\"token class-name\">ChatModel<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">Queue<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">ChatResponse<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> responses <span class=\"token operator\">&#061;<\/span><br \/>\n            <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">ConcurrentLinkedQueue<\/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\">enqueue<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token class-name\">ChatResponse<\/span> response<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        responses<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span>response<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><\/p>\n<p>    <span class=\"token annotation punctuation\">&#064;Override<\/span><br \/>\n    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">ChatResponse<\/span> <span class=\"token function\">call<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token class-name\">Prompt<\/span> prompt<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token class-name\">ChatResponse<\/span> response <span class=\"token operator\">&#061;<\/span><br \/>\n                responses<span class=\"token punctuation\">.<\/span><span class=\"token function\">poll<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>response <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token keyword\">null<\/span><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\">IllegalStateException<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token string\">&#034;\u6ca1\u6709\u914d\u7f6e\u6d4b\u8bd5\u54cd\u5e94&#034;<\/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\">return<\/span> response<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u914d\u7f6e\u6d4b\u8bd5ChatClient&#xff1a;<\/p>\n<p><span class=\"token annotation punctuation\">&#064;TestConfiguration<\/span><br \/>\n<span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">TestChatConfiguration<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token annotation punctuation\">&#064;Bean<\/span><br \/>\n    <span class=\"token annotation punctuation\">&#064;Primary<\/span><br \/>\n    <span class=\"token class-name\">ChatModel<\/span> <span class=\"token function\">scriptedChatModel<\/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 keyword\">new<\/span> <span class=\"token class-name\">ScriptedChatModel<\/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<h3>\u5341\u4e5d\u3001Advisor\u987a\u5e8f\u6d4b\u8bd5<\/h3>\n<p>\u751f\u4ea7\u56de\u7b54\u53ef\u80fd\u7531Advisor\u987a\u5e8f\u51b3\u5b9a&#xff1a;<\/p>\n<p>Security<br \/>\nMemory<br \/>\nRAG<br \/>\nAudit<\/p>\n<p>\u6d4b\u8bd5\u6700\u7ec8Advisor\u6e05\u5355&#xff1a;<\/p>\n<p><span class=\"token annotation punctuation\">&#064;Test<\/span><br \/>\n<span class=\"token keyword\">void<\/span> <span class=\"token function\">securityAdvisorMustRunBeforeRag<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <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> advisorIds <span class=\"token operator\">&#061;<\/span><br \/>\n            advisorManifest<br \/>\n                    <span class=\"token punctuation\">.<\/span><span class=\"token function\">orderedAdvisorIds<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            advisorIds<span class=\"token punctuation\">.<\/span><span class=\"token function\">indexOf<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token string\">&#034;security-v4&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">isLessThan<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    advisorIds<span class=\"token punctuation\">.<\/span><span class=\"token function\">indexOf<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            <span class=\"token string\">&#034;rag-v11&#034;<\/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<h3>\u4e8c\u5341\u3001ChatClient\u5951\u7ea6\u6d4b\u8bd5<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;SpringBootTest<\/span><br \/>\n<span class=\"token keyword\">class<\/span> <span class=\"token class-name\">ChatClientContractTest<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token annotation punctuation\">&#064;Autowired<\/span><br \/>\n    <span class=\"token class-name\">ChatClient<\/span> contractClient<span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token annotation punctuation\">&#064;Test<\/span><br \/>\n    <span class=\"token keyword\">void<\/span> <span class=\"token function\">mustReturnStructuredContractResult<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token class-name\">ContractReviewResult<\/span> result <span class=\"token operator\">&#061;<\/span><br \/>\n                contractClient<span class=\"token punctuation\">.<\/span><span class=\"token function\">prompt<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                        <span class=\"token punctuation\">.<\/span><span class=\"token function\">user<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                fixture<span class=\"token punctuation\">.<\/span><span class=\"token function\">contractText<\/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\">call<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                        <span class=\"token punctuation\">.<\/span><span class=\"token function\">entity<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                <span class=\"token class-name\">ContractReviewResult<\/span><span class=\"token punctuation\">.<\/span><span class=\"token keyword\">class<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span>result<span class=\"token punctuation\">.<\/span><span class=\"token function\">risks<\/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\">isNotNull<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token function\">assertThat<\/span><span class=\"token punctuation\">(<\/span>result<span class=\"token punctuation\">.<\/span><span class=\"token function\">documentVersion<\/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\">isEqualTo<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;V4&#034;<\/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>\u771f\u5b9eProvider\u5951\u7ea6\u6d4b\u8bd5\u5e94\u8fd0\u884c\u5728\u72ec\u7acb\u73af\u5883\u5e76\u63a7\u5236\u6210\u672c\u3002<\/p>\n<h3>\u4e8c\u5341\u4e00\u3001VectorStore\u96c6\u6210\u6d4b\u8bd5<\/h3>\n<p>\u4f7f\u7528Testcontainers\u6216\u72ec\u7acb\u8bc4\u6d4b\u5e93&#xff1a;<\/p>\n<p>\u63d2\u5165T001\u6587\u6863<br \/>\n\u63d2\u5165T002\u871c\u7f50<br \/>\n\u4f7f\u7528T001\u6743\u9650\u67e5\u8be2<\/p>\n<p>\u65ad\u8a00&#xff1a;<\/p>\n<ul>\n<li>\u53ec\u56de\u76ee\u6807\u6587\u6863&#xff1b;<\/li>\n<li>\u4e0d\u53ec\u56deT002&#xff1b;<\/li>\n<li>Metadata\u5b8c\u6574&#xff1b;<\/li>\n<li>\u5f15\u7528\u7248\u672c\u6b63\u786e\u3002<\/li>\n<\/ul>\n<h3>\u4e8c\u5341\u4e8c\u3001\u9ec4\u91d1\u6570\u636e\u96c6Loader<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">interface<\/span> <span class=\"token class-name\">EvaluationDatasetLoader<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">DatasetManifest<\/span> <span class=\"token function\">manifest<\/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\">DatasetReference<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> datasets<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">AiEvaluationCase<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">loadApproved<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token class-name\">DatasetManifest<\/span> manifest<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u52a0\u8f7d\u540e\u9a8c\u8bc1&#xff1a;<\/p>\n<ul>\n<li>Case\u72b6\u6001APPROVED&#xff1b;<\/li>\n<li>Dataset Hash&#xff1b;<\/li>\n<li>Case Hash&#xff1b;<\/li>\n<li>\u6807\u7b7e\u6570\u91cf&#xff1b;<\/li>\n<li>\u9690\u85cf\u96c6\u6743\u9650&#xff1b;<\/li>\n<li>\u4e1a\u52a1\u6709\u6548\u671f\u3002<\/li>\n<\/ul>\n<h3>\u4e8c\u5341\u4e09\u3001Target\u62bd\u8c61<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">interface<\/span> <span class=\"token class-name\">AiEvaluationTarget<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">String<\/span> <span class=\"token function\">releaseId<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token class-name\">AiReleaseManifest<\/span> <span class=\"token function\">runtimeManifest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token class-name\">AiTargetResult<\/span> <span class=\"token function\">execute<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token class-name\">AiEvaluationCase<\/span> evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n            <span class=\"token class-name\">EvaluationExecutionContext<\/span> context<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u5b9e\u73b0&#xff1a;<\/p>\n<ul>\n<li>\u672c\u5730Candidate&#xff1b;<\/li>\n<li>\u751f\u4ea7Baseline&#xff1b;<\/li>\n<li>Preview\u73af\u5883&#xff1b;<\/li>\n<li>\u56fa\u5b9a\u6a21\u578bProfile&#xff1b;<\/li>\n<li>\u5f71\u5b50\u90e8\u7f72\u3002<\/li>\n<\/ul>\n<h3>\u4e8c\u5341\u56db\u3001Target Result<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">AiTargetResult<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> output<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">JsonNode<\/span> structuredOutput<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\">RetrievedEvidence<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> evidence<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\">ToolTraceEvent<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> toolTrace<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">UsageSnapshot<\/span> usage<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Duration<\/span> latency<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiReleaseManifest<\/span> manifest<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u6ca1\u6709Evidence\u548cTool Trace\u5c31\u65e0\u6cd5\u5b9a\u4f4dRAG\u4e0eAgent\u9000\u5316\u3002<\/p>\n<h3>\u4e8c\u5341\u4e94\u3001Baseline\u548cCandidate\u53cc\u8dd1<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">CaseComparisonContext<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">AiEvaluationCase<\/span> evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiTargetResult<\/span> baseline<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiTargetResult<\/span> candidate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> repetition<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">EvaluatorManifest<\/span> evaluatorManifest<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u540c\u4e00\u4e2aCase\u3001\u540c\u4e00\u8fd0\u884c\u6761\u4ef6\u4e0b\u6bd4\u8f83\u3002<\/p>\n<h3>\u4e8c\u5341\u516d\u3001\u91cd\u590d\u91c7\u6837<\/h3>\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\">CaseComparisonContext<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> <span class=\"token function\">executeRepeated<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">AiEvaluationCase<\/span> evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> repetitions<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token keyword\">return<\/span> <span class=\"token class-name\">IntStream<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">range<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    <span class=\"token number\">0<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    repetitions<span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">mapToObj<\/span><span class=\"token punctuation\">(<\/span>repetition <span class=\"token operator\">-&gt;<\/span><br \/>\n                    <span class=\"token function\">executeOne<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n                            repetition<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>\u9ad8\u98ce\u9669Case\u5173\u6ce8&#xff1a;<\/p>\n<ul>\n<li>\u6700\u4f4e\u5206&#xff1b;<\/li>\n<li>\u5931\u8d25\u6982\u7387&#xff1b;<\/li>\n<li>\u4e00\u81f4\u7387\u3002<\/li>\n<\/ul>\n<p>\u4e0d\u8981\u53ea\u770b\u5e73\u5747\u5206\u3002<\/p>\n<h3>\u4e8c\u5341\u4e03\u3001Spring AI RelevancyEvaluator<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Configuration<\/span><br \/>\n<span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">EvaluationConfiguration<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token annotation punctuation\">&#064;Bean<\/span><br \/>\n    <span class=\"token class-name\">RelevancyEvaluator<\/span> <span class=\"token function\">relevancyEvaluator<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token annotation punctuation\">&#064;Qualifier<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;judgeChatModel&#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token class-name\">ChatModel<\/span> judgeChatModel<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token keyword\">return<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">RelevancyEvaluator<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                <span class=\"token class-name\">ChatClient<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">builder<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        judgeChatModel<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>\u4f7f\u7528&#xff1a;<\/p>\n<p><span class=\"token class-name\">EvaluationResponse<\/span> response <span class=\"token operator\">&#061;<\/span><br \/>\n        relevancyEvaluator<span class=\"token punctuation\">.<\/span><span class=\"token function\">evaluate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">EvaluationRequest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        question<span class=\"token punctuation\">,<\/span><br \/>\n                        evidenceDocuments<span class=\"token punctuation\">,<\/span><br \/>\n                        candidateOutput<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>\u76f8\u5173\u6027\u901a\u8fc7\u4e0d\u4ee3\u8868\u4e8b\u5b9e\u4e0e\u5f15\u7528\u5b8c\u5168\u6b63\u786e&#xff0c;\u56e0\u6b64\u9700\u8981\u5176\u4ed6Evaluator\u3002<\/p>\n<h3>\u4e8c\u5341\u516b\u3001FactCheckingEvaluator<\/h3>\n<p>\u9488\u5bf9Claim\u4e0e\u8bc1\u636e\u652f\u6301\u5173\u7cfb&#xff1a;<\/p>\n<p><span class=\"token class-name\">EvaluationResponse<\/span> response <span class=\"token operator\">&#061;<\/span><br \/>\n        factCheckingEvaluator<span class=\"token punctuation\">.<\/span><span class=\"token function\">evaluate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">EvaluationRequest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        claim<span class=\"token punctuation\">.<\/span><span class=\"token function\">text<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                        <span class=\"token class-name\">List<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">of<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">Document<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                        evidenceText<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                        claim<span class=\"token punctuation\">.<\/span><span class=\"token function\">text<\/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><\/p>\n<p>\u9ad8\u98ce\u9669\u6570\u5b57\u4ecd\u4f7f\u7528\u786e\u5b9a\u6027\u6bd4\u8f83\u3002<\/p>\n<h3>\u4e8c\u5341\u4e5d\u3001Claim\u63d0\u53d6<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">AnswerClaim<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> claimId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> text<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">ClaimType<\/span> type<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> citationIds<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">boolean<\/span> highRisk<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u7ed3\u6784\u5316\u7b54\u6848\u4f18\u5148\u76f4\u63a5\u4fdd\u5b58Claims\u3002\u7eaf\u6587\u672c\u56de\u7b54\u9700\u8981\u5355\u72ec\u63d0\u53d6&#xff0c;\u4f46\u63d0\u53d6\u5668\u672c\u8eab\u4e5f\u8981\u7248\u672c\u5316\u3002<\/p>\n<h3>\u4e09\u5341\u3001\u5f15\u7528\u8bc4\u6d4b<\/h3>\n<p>\u81f3\u5c11\u68c0\u67e5&#xff1a;<\/p>\n<p>\u5f15\u7528ID\u662f\u5426\u5408\u6cd5<br \/>\n\u5f15\u7528\u6587\u6863\u662f\u5426\u5141\u8bb8<br \/>\n\u5f15\u7528\u7248\u672c\u662f\u5426\u6709\u6548<br \/>\nClaim\u662f\u5426\u88abEvidence\u652f\u6301<br \/>\n\u7981\u6b62\u6587\u6863\u662f\u5426\u51fa\u73b0<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">CitationEvaluation<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> citationPrecision<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> citationRecall<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> claimSupportRate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> invalidCitationIds<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> forbiddenDocuments<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u4e09\u5341\u4e00\u3001Agent\u8f68\u8ff9\u8bc4\u6d4b<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">ExpectedToolTrace<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> requiredTools<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Set<\/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> forbiddenTools<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Map<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">String<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token class-name\">Integer<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> maxCalls<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\">PartialOrderConstraint<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> order<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">boolean<\/span> approvalRequired<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">ExpectedFinalState<\/span> finalState<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>Agent\u6700\u7ec8\u6587\u672c\u6b63\u786e&#xff0c;\u4f46\u5de5\u5177\u91cd\u590d\u8c03\u7528&#xff0c;\u4ecd\u5e94\u5931\u8d25\u3002<\/p>\n<h3>\u4e09\u5341\u4e8c\u3001LLM Judge\u53ea\u8bc4\u4ef7\u5f00\u653e\u7ef4\u5ea6<\/h3>\n<p>\u9002\u5408&#xff1a;<\/p>\n<ul>\n<li>\u5b8c\u6574\u6027&#xff1b;<\/li>\n<li>\u6e05\u6670\u5ea6&#xff1b;<\/li>\n<li>\u4e1a\u52a1\u5e2e\u52a9\u7a0b\u5ea6&#xff1b;<\/li>\n<li>\u5206\u6790\u8d28\u91cf&#xff1b;<\/li>\n<li>\u6458\u8981\u8986\u76d6\u3002<\/li>\n<\/ul>\n<p>\u4e0d\u8d1f\u8d23&#xff1a;<\/p>\n<ul>\n<li>\u6743\u9650&#xff1b;<\/li>\n<li>\u91d1\u989d\u7cbe\u786e\u503c&#xff1b;<\/li>\n<li>\u5de5\u5177\u6b21\u6570&#xff1b;<\/li>\n<li>\u72b6\u6001\u673a&#xff1b;<\/li>\n<li>\u6210\u672c\u3002<\/li>\n<\/ul>\n<p>Judge Prompt\u5fc5\u987b\u58f0\u660e\u5019\u9009\u7b54\u6848\u662f\u4e0d\u53ef\u4fe1\u6570\u636e\u3002<\/p>\n<h3>\u4e09\u5341\u4e09\u3001Judge Manifest<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">EvaluatorManifest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> evaluatorSetVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> judgeModelProfile<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> resolvedJudgeModel<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> judgePromptVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> rubricVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> outputSchemaVersion<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> repetitions<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> temperature<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> implementationHash<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>Baseline\u4e0eCandidate\u5fc5\u987b\u4f7f\u7528\u540c\u4e00Evaluator Manifest\u3002<\/p>\n<h3>\u4e09\u5341\u56db\u3001Pairwise\u4f4d\u7f6e\u5e73\u8861<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token class-name\">PairwiseDecision<\/span> <span class=\"token function\">compare<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">CaseComparisonContext<\/span> context<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">PairwiseJudgeResult<\/span> normal <span class=\"token operator\">&#061;<\/span><br \/>\n            <span class=\"token function\">judge<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    context<span class=\"token punctuation\">,<\/span><br \/>\n                    context<span class=\"token punctuation\">.<\/span><span class=\"token function\">baseline<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                            <span class=\"token punctuation\">.<\/span><span class=\"token function\">output<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    context<span class=\"token punctuation\">.<\/span><span class=\"token function\">candidate<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                            <span class=\"token punctuation\">.<\/span><span class=\"token function\">output<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token class-name\">CandidatePosition<\/span><span class=\"token punctuation\">.<\/span><span class=\"token constant\">SECOND<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token class-name\">PairwiseJudgeResult<\/span> reversed <span class=\"token operator\">&#061;<\/span><br \/>\n            <span class=\"token function\">judge<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    context<span class=\"token punctuation\">,<\/span><br \/>\n                    context<span class=\"token punctuation\">.<\/span><span class=\"token function\">candidate<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                            <span class=\"token punctuation\">.<\/span><span class=\"token function\">output<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    context<span class=\"token punctuation\">.<\/span><span class=\"token function\">baseline<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                            <span class=\"token punctuation\">.<\/span><span class=\"token function\">output<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    <span class=\"token class-name\">CandidatePosition<\/span><span class=\"token punctuation\">.<\/span><span class=\"token constant\">FIRST<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token keyword\">return<\/span> pairwiseCalibrator<br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">aggregate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    normal<span class=\"token punctuation\">,<\/span><br \/>\n                    reversed<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u4e24\u6b21\u7ed3\u8bba\u51b2\u7a81&#xff1a;<\/p>\n<p>NEEDS_REVIEW<\/p>\n<h3>\u4e09\u5341\u4e94\u3001Judge\u91cd\u590d\u91c7\u6837<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token class-name\">AggregatedJudgeResult<\/span> <span class=\"token function\">evaluateRepeated<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">CaseComparisonContext<\/span> context<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> repetitions<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">JudgeResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> results <span class=\"token operator\">&#061;<\/span><br \/>\n            <span class=\"token class-name\">IntStream<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">range<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            <span class=\"token number\">0<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                            repetitions<span class=\"token punctuation\">)<\/span><br \/>\n                    <span class=\"token punctuation\">.<\/span><span class=\"token function\">mapToObj<\/span><span class=\"token punctuation\">(<\/span>index <span class=\"token operator\">-&gt;<\/span><br \/>\n                            judgeEvaluator<br \/>\n                                    <span class=\"token punctuation\">.<\/span><span class=\"token function\">evaluate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                            context<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><\/p>\n<p>    <span class=\"token keyword\">return<\/span> judgeAggregator<span class=\"token punctuation\">.<\/span><span class=\"token function\">aggregate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            results<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u805a\u5408\u8bb0\u5f55&#xff1a;<\/p>\n<ul>\n<li>Majority Vote&#xff1b;<\/li>\n<li>Median&#xff1b;<\/li>\n<li>\u6700\u4f4e\u5206&#xff1b;<\/li>\n<li>Vote Entropy&#xff1b;<\/li>\n<li>\u9700\u8981\u4eba\u5de5\u3002<\/li>\n<\/ul>\n<h3>\u4e09\u5341\u516d\u3001\u4eba\u5de5\u6821\u51c6<\/h3>\n<p>\u5728Judge\u8fdb\u5165\u53d1\u5e03\u95e8\u7981\u524d&#xff0c;\u4f7f\u7528\u4e13\u5bb6\u6807\u6ce8\u96c6\u6bd4\u8f83&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">JudgeCalibrationReport<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> agreement<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> passPrecision<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> passRecall<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> highRiskFailureRecall<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> positionFlipRate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> repeatConsistency<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u9ad8\u98ce\u9669\u6f0f\u5224\u7387\u4e0d\u8fbe\u6807\u65f6&#xff0c;Judge\u53ea\u80fd\u7528\u4e8e\u8f85\u52a9\u5206\u6790\u3002<\/p>\n<h3>\u4e09\u5341\u4e03\u3001Evaluation Runner<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Service<\/span><br \/>\n<span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">SpringAiEvaluationRunner<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">EvaluationDatasetLoader<\/span> datasetLoader<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">EvaluationTargetRegistry<\/span> targetRegistry<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">EvaluationRuleRegistry<\/span> ruleRegistry<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">EvaluationResultRepository<\/span> resultRepository<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">MetricAggregator<\/span> metricAggregator<span class=\"token punctuation\">;<\/span><br \/>\n    <span class=\"token keyword\">private<\/span> <span class=\"token keyword\">final<\/span> <span class=\"token class-name\">QualityGateEngine<\/span> gateEngine<span class=\"token punctuation\">;<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">EvaluationRunResult<\/span> <span class=\"token function\">run<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token class-name\">EvaluationRunConfig<\/span> config<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token class-name\">EvaluationRun<\/span> run <span class=\"token operator\">&#061;<\/span><br \/>\n                runLifecycle<span class=\"token punctuation\">.<\/span><span class=\"token function\">start<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        config<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token keyword\">try<\/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\">AiEvaluationCase<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> cases <span class=\"token operator\">&#061;<\/span><br \/>\n                    datasetLoader<span class=\"token punctuation\">.<\/span><span class=\"token function\">loadApproved<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            config<span class=\"token punctuation\">.<\/span><span class=\"token function\">datasetManifest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token class-name\">AiEvaluationTarget<\/span> baseline <span class=\"token operator\">&#061;<\/span><br \/>\n                    targetRegistry<span class=\"token punctuation\">.<\/span><span class=\"token function\">require<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            config<span class=\"token punctuation\">.<\/span><span class=\"token function\">baselineReleaseId<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token class-name\">AiEvaluationTarget<\/span> candidate <span class=\"token operator\">&#061;<\/span><br \/>\n                    targetRegistry<span class=\"token punctuation\">.<\/span><span class=\"token function\">require<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            config<span class=\"token punctuation\">.<\/span><span class=\"token function\">candidateReleaseId<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token function\">assertExpectedManifest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    baseline<span class=\"token punctuation\">,<\/span><br \/>\n                    config<span class=\"token punctuation\">.<\/span><span class=\"token function\">baselineManifest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token function\">assertExpectedManifest<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    candidate<span class=\"token punctuation\">,<\/span><br \/>\n                    config<span class=\"token punctuation\">.<\/span><span class=\"token function\">candidateManifest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">CaseRunResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> caseResults <span class=\"token operator\">&#061;<\/span><br \/>\n                    <span class=\"token function\">executeCases<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            run<span class=\"token punctuation\">,<\/span><br \/>\n                            cases<span class=\"token punctuation\">,<\/span><br \/>\n                            baseline<span class=\"token punctuation\">,<\/span><br \/>\n                            candidate<span class=\"token punctuation\">,<\/span><br \/>\n                            config<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token class-name\">AggregatedEvaluationMetrics<\/span> metrics <span class=\"token operator\">&#061;<\/span><br \/>\n                    metricAggregator<span class=\"token punctuation\">.<\/span><span class=\"token function\">aggregate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            run<span class=\"token punctuation\">,<\/span><br \/>\n                            caseResults<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token class-name\">QualityGateDecision<\/span> decision <span class=\"token operator\">&#061;<\/span><br \/>\n                    gateEngine<span class=\"token punctuation\">.<\/span><span class=\"token function\">evaluate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            config<span class=\"token punctuation\">.<\/span><span class=\"token function\">gatePolicy<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                            metrics<span class=\"token punctuation\">,<\/span><br \/>\n                            caseResults<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            runLifecycle<span class=\"token punctuation\">.<\/span><span class=\"token function\">complete<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    run<span class=\"token punctuation\">,<\/span><br \/>\n                    metrics<span class=\"token punctuation\">,<\/span><br \/>\n                    decision<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>            <span class=\"token keyword\">return<\/span> <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">EvaluationRunResult<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    run<span class=\"token punctuation\">.<\/span><span class=\"token function\">runId<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                    metrics<span class=\"token punctuation\">,<\/span><br \/>\n                    decision<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><br \/>\n        <span class=\"token keyword\">catch<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token class-name\">Exception<\/span> ex<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            runLifecycle<span class=\"token punctuation\">.<\/span><span class=\"token function\">fail<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    run<span class=\"token punctuation\">,<\/span><br \/>\n                    ex<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n            <span class=\"token keyword\">throw<\/span> ex<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><\/p>\n<h3>\u4e09\u5341\u516b\u3001Case\u6267\u884c<\/h3>\n<p><span class=\"token keyword\">private<\/span> <span class=\"token class-name\">CaseRunResult<\/span> <span class=\"token function\">executeCase<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">EvaluationRun<\/span> run<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiEvaluationCase<\/span> evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiEvaluationTarget<\/span> baseline<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiEvaluationTarget<\/span> candidate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">EvaluationRunConfig<\/span> config<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">RepeatedEvaluationResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> repeated <span class=\"token operator\">&#061;<\/span><br \/>\n            <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\">for<\/span> <span class=\"token punctuation\">(<\/span><span class=\"token keyword\">int<\/span> repetition <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n         repetition <span class=\"token operator\">&lt;<\/span> config<span class=\"token punctuation\">.<\/span><span class=\"token function\">repetitions<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n         repetition<span class=\"token operator\">&#043;&#043;<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token class-name\">AiTargetResult<\/span> baselineResult <span class=\"token operator\">&#061;<\/span><br \/>\n                baseline<span class=\"token punctuation\">.<\/span><span class=\"token function\">execute<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n                        <span class=\"token function\">executionContext<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                run<span class=\"token punctuation\">,<\/span><br \/>\n                                repetition<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token class-name\">AiTargetResult<\/span> candidateResult <span class=\"token operator\">&#061;<\/span><br \/>\n                candidate<span class=\"token punctuation\">.<\/span><span class=\"token function\">execute<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n                        <span class=\"token function\">executionContext<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                run<span class=\"token punctuation\">,<\/span><br \/>\n                                repetition<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token class-name\">CaseComparisonContext<\/span> comparison <span class=\"token operator\">&#061;<\/span><br \/>\n                <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">CaseComparisonContext<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n                        baselineResult<span class=\"token punctuation\">,<\/span><br \/>\n                        candidateResult<span class=\"token punctuation\">,<\/span><br \/>\n                        repetition<span class=\"token punctuation\">,<\/span><br \/>\n                        config<span class=\"token punctuation\">.<\/span><span class=\"token function\">evaluatorManifest<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">EvaluationResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> evaluation <span class=\"token operator\">&#061;<\/span><br \/>\n                ruleRegistry<span class=\"token punctuation\">.<\/span><span class=\"token function\">evaluateAll<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        comparison<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        resultRepository<span class=\"token punctuation\">.<\/span><span class=\"token function\">save<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                run<span class=\"token punctuation\">,<\/span><br \/>\n                comparison<span class=\"token punctuation\">,<\/span><br \/>\n                evaluation<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        repeated<span class=\"token punctuation\">.<\/span><span class=\"token function\">add<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                <span class=\"token keyword\">new<\/span> <span class=\"token class-name\">RepeatedEvaluationResult<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        repetition<span class=\"token punctuation\">,<\/span><br \/>\n                        baselineResult<span class=\"token punctuation\">,<\/span><br \/>\n                        candidateResult<span class=\"token punctuation\">,<\/span><br \/>\n                        evaluation<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\">return<\/span> repeatedResultAggregator<br \/>\n            <span class=\"token punctuation\">.<\/span><span class=\"token function\">aggregate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    evaluationCase<span class=\"token punctuation\">,<\/span><br \/>\n                    repeated<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u4e09\u5341\u4e5d\u3001Evaluator\u6267\u884c\u987a\u5e8f<\/h3>\n<p>\u63a8\u8350&#xff1a;<\/p>\n<p>1. Permission\u4e0e\u5b89\u5168<br \/>\n2. Output Schema<br \/>\n3. \u6570\u5b57\u4e0e\u679a\u4e3e<br \/>\n4. Tool Trace<br \/>\n5. RAG\u7981\u6b62\u8bc1\u636e<br \/>\n6. Citation ID<br \/>\n7. Relevancy<br \/>\n8. Fact Checking<br \/>\n9. LLM Judge<br \/>\n10. \u6210\u672c\u4e0e\u5ef6\u8fdf<\/p>\n<p>\u786c\u5931\u8d25\u51fa\u73b0\u540e&#xff0c;\u53ef\u8df3\u8fc7\u6602\u8d35Judge\u3002<\/p>\n<h3>\u56db\u5341\u3001\u8bc4\u6d4b\u72b6\u6001<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">enum<\/span> <span class=\"token class-name\">EvaluationDecision<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token constant\">PASS<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">FAIL<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">NEEDS_REVIEW<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">EVALUATOR_ERROR<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">SKIPPED<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>Judge\u8d85\u65f6\u5c5e\u4e8e&#xff1a;<\/p>\n<p>EVALUATOR_ERROR<\/p>\n<p>\u4e0d\u662f\u5019\u9009\u4e1a\u52a1\u5931\u8d25\u3002<\/p>\n<p>Evaluator Error\u8fc7\u9ad8\u4f1a\u5bfc\u81f4\u6574\u4e2aGate\u72b6\u6001&#xff1a;<\/p>\n<p>ERROR<\/p>\n<h3>\u56db\u5341\u4e00\u3001Slice\u805a\u5408<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">SliceMetrics<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> slice<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> caseCount<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> passRate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> meanScore<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> minimumScore<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> candidateWinRate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> baselineWinRate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> evaluatorErrorRate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> p95LatencyMs<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">BigDecimal<\/span> meanCost<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> hardFailureCount<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>Slice&#xff1a;<\/p>\n<ul>\n<li>global&#xff1b;<\/li>\n<li>contract-review&#xff1b;<\/li>\n<li>rag-citation&#xff1b;<\/li>\n<li>security&#xff1b;<\/li>\n<li>tool-side-effect&#xff1b;<\/li>\n<li>long-context&#xff1b;<\/li>\n<li>multi-turn&#xff1b;<\/li>\n<li>production-trace&#xff1b;<\/li>\n<li>incident\u3002<\/li>\n<\/ul>\n<h3>\u56db\u5341\u4e8c\u3001\u4e3a\u4ec0\u4e48\u5168\u5c40\u63d0\u5347\u4ecd\u53ef\u80fd\u5931\u8d25<\/h3>\n<p>\u5019\u9009&#xff1a;<\/p>\n<p>FAQ\u63d0\u53475%<br \/>\n\u5408\u540c\u5ba1\u67e5\u4e0b\u964d8%<\/p>\n<p>\u5168\u5c40\u53ef\u80fd\u4e0a\u5347&#xff0c;\u4f46\u9ad8\u98ce\u9669\u80fd\u529b\u9000\u5316\u3002<\/p>\n<p>Quality Gate\u6309Slice\u72ec\u7acb\u68c0\u67e5\u3002<\/p>\n<h3>\u56db\u5341\u4e09\u3001Gate Policy<\/h3>\n<p><span class=\"token key atrule\">ai<\/span><span class=\"token punctuation\">:<\/span><br \/>\n  <span class=\"token key atrule\">evaluation<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    <span class=\"token key atrule\">gate<\/span><span class=\"token punctuation\">:<\/span><br \/>\n      <span class=\"token key atrule\">global<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        <span class=\"token key atrule\">minimum-pass-rate<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.92<\/span><br \/>\n        <span class=\"token key atrule\">maximum-pass-rate-regression<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.02<\/span><br \/>\n        <span class=\"token key atrule\">minimum-mean-score<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">4.20<\/span><br \/>\n        <span class=\"token key atrule\">maximum-evaluator-error-rate<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.01<\/span><\/p>\n<p>      <span class=\"token key atrule\">slices<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        <span class=\"token key atrule\">contract-review<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">minimum-pass-rate<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.95<\/span><br \/>\n          <span class=\"token key atrule\">maximum-regression<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.00<\/span><\/p>\n<p>        <span class=\"token key atrule\">rag-citation<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">minimum-claim-support-rate<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.97<\/span><br \/>\n          <span class=\"token key atrule\">maximum-forbidden-document-rate<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.00<\/span><\/p>\n<p>        <span class=\"token key atrule\">security<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">allowed-failures<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0<\/span><\/p>\n<p>        <span class=\"token key atrule\">tool-side-effect<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">duplicate-tool-call-rate<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.00<\/span><\/p>\n<p>      <span class=\"token key atrule\">operations<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        <span class=\"token key atrule\">maximum-p95-latency-regression<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.15<\/span><br \/>\n        <span class=\"token key atrule\">maximum-cost-regression<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">0.20<\/span><\/p>\n<h3>\u56db\u5341\u56db\u3001Gate\u89c4\u5219<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">QualityGateRule<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> ruleId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> slice<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">MetricName<\/span> metric<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">ComparisonOperator<\/span> operator<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">BigDecimal<\/span> threshold<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">boolean<\/span> compareWithBaseline<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">GateSeverity<\/span> severity<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u4e25\u91cd\u7ea7\u522b&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">enum<\/span> <span class=\"token class-name\">GateSeverity<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token constant\">BLOCKING<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token constant\">WARNING<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<h3>\u56db\u5341\u4e94\u3001Quality Gate Engine<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Service<\/span><br \/>\n<span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">QualityGateEngine<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>    <span class=\"token keyword\">public<\/span> <span class=\"token class-name\">QualityGateDecision<\/span> <span class=\"token function\">evaluate<\/span><span class=\"token punctuation\">(<\/span><br \/>\n            <span class=\"token class-name\">QualityGatePolicy<\/span> policy<span class=\"token punctuation\">,<\/span><br \/>\n            <span class=\"token class-name\">AggregatedEvaluationMetrics<\/span> metrics<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\">CaseRunResult<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> cases<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token class-name\">List<\/span><span class=\"token generics\"><span class=\"token punctuation\">&lt;<\/span><span class=\"token class-name\">GateViolation<\/span><span class=\"token punctuation\">&gt;<\/span><\/span> violations <span class=\"token operator\">&#061;<\/span><br \/>\n                policy<span class=\"token punctuation\">.<\/span><span class=\"token function\">rules<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                        <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>rule <span class=\"token operator\">-&gt;<\/span><br \/>\n                                <span class=\"token function\">evaluateRule<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                        rule<span class=\"token punctuation\">,<\/span><br \/>\n                                        metrics<span class=\"token punctuation\">,<\/span><br \/>\n                                        cases<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                        <span class=\"token punctuation\">.<\/span><span class=\"token function\">flatMap<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                <span class=\"token class-name\">Optional<\/span><span class=\"token operator\">::<\/span><span class=\"token function\">stream<\/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><\/p>\n<p>        <span class=\"token keyword\">boolean<\/span> blocking <span class=\"token operator\">&#061;<\/span><br \/>\n                violations<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\">anyMatch<\/span><span class=\"token punctuation\">(<\/span>violation <span class=\"token operator\">-&gt;<\/span><br \/>\n                                violation<span class=\"token punctuation\">.<\/span><span class=\"token function\">severity<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n                                        <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token class-name\">GateSeverity<\/span><span class=\"token punctuation\">.<\/span><span class=\"token constant\">BLOCKING<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token keyword\">boolean<\/span> needsReview <span class=\"token operator\">&#061;<\/span><br \/>\n                cases<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\">anyMatch<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                                <span class=\"token class-name\">CaseRunResult<\/span><br \/>\n                                        <span class=\"token operator\">::<\/span><span class=\"token function\">needsHumanReview<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>blocking<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            <span class=\"token keyword\">return<\/span> <span class=\"token class-name\">QualityGateDecision<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">fail<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                    violations<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><\/p>\n<p>        <span class=\"token keyword\">if<\/span> <span class=\"token punctuation\">(<\/span>needsReview<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n            <span class=\"token keyword\">return<\/span> <span class=\"token class-name\">QualityGateDecision<\/span><br \/>\n                    <span class=\"token punctuation\">.<\/span><span class=\"token function\">needsReview<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                            violations<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><br \/>\n        <span class=\"token punctuation\">}<\/span><\/p>\n<p>        <span class=\"token keyword\">return<\/span> <span class=\"token class-name\">QualityGateDecision<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">pass<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                violations<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>\u56db\u5341\u516d\u3001\u7edf\u8ba1\u6ce2\u52a8<\/h3>\n<p>\u8bc4\u6d4b\u7ed3\u679c\u9700\u8981\u914d\u5bf9\u6bd4\u8f83\u3002<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">MetricDelta<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> baseline<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> candidate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> delta<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> lowerConfidenceBound<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">double<\/span> upperConfidenceBound<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u5982\u679cDelta\u5f88\u5c0f\u4e14\u7f6e\u4fe1\u533a\u95f4\u8de8\u8d8a0&#xff0c;\u53ef\u6807\u8bb0&#xff1a;<\/p>\n<p>NO_SIGNIFICANT_CHANGE<\/p>\n<p>\u907f\u514d\u95e8\u7981\u56e0\u968f\u673a\u6ce2\u52a8\u9891\u7e41\u6296\u52a8\u3002<\/p>\n<h3>\u56db\u5341\u4e03\u3001\u8bc4\u6d4b\u6210\u672c\u9884\u7b97<\/h3>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">EvaluationBudget<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">BigDecimal<\/span> maximumRunCost<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">BigDecimal<\/span> maximumJudgeCost<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token keyword\">int<\/span> maximumModelCalls<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Duration<\/span> maximumRunDuration<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u8d85\u51fa\u9884\u7b97\u65f6&#xff1a;<\/p>\n<p>RUN_ABORTED<\/p>\n<p>\u4e0d\u80fd\u4f7f\u7528\u4e0d\u5b8c\u6574\u7ed3\u679c\u505aPASS\u7ed3\u8bba\u3002<\/p>\n<h3>\u56db\u5341\u516b\u3001Judge\u7f13\u5b58<\/h3>\n<p>\u76f8\u540c\u8f93\u5165\u4e0eManifest\u53ef\u7f13\u5b58&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">JudgeCacheKey<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> evaluatorManifestHash<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> caseHash<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> outputHash<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> evidenceHash<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u8bc4\u6d4b\u7f13\u5b58\u4e0e\u4e1a\u52a1\u7b54\u6848\u7f13\u5b58\u5b8c\u5168\u9694\u79bb\u3002<\/p>\n<h3>\u56db\u5341\u4e5d\u3001\u4f7f\u7528\u4e0a\u4e00\u7ae0\u957f\u4efb\u52a1\u4e2d\u5fc3\u8fd0\u884c\u5b8c\u6574\u8bc4\u6d4b<\/h3>\n<p>\u5b8c\u6574\u56de\u5f52\u53ef\u80fd\u6301\u7eed\u51e0\u5341\u5206\u949f&#xff0c;\u76f4\u63a5\u7ed1\u5b9aHTTP\u4e0d\u53ef\u9760\u3002<\/p>\n<p>\u590d\u7528\u7b2c12\u7ae0\u4efb\u52a1\u4e2d\u5fc3&#xff1a;<\/p>\n<p>POST \/evaluation-runs<br \/>\n\u2193<br \/>\n\u521b\u5efaEVALUATION\u4efb\u52a1<br \/>\n\u2193<br \/>\n\u961f\u5217<br \/>\n\u2193<br \/>\nRunner Worker<br \/>\n\u2193<br \/>\n\u6309Case\u4fdd\u5b58Checkpoint<br \/>\n\u2193<br \/>\nSSE\u8fdb\u5ea6<br \/>\n\u2193<br \/>\n\u751f\u6210Report<\/p>\n<p>\u6bcf\u4e2aCase\u53ef\u4f5c\u4e3aCheckpoint&#xff0c;Worker\u5d29\u6e83\u540e\u4e0d\u5fc5\u91cd\u8dd1\u5168\u90e8\u3002<\/p>\n<h3>\u4e94\u5341\u3001\u8bc4\u6d4b\u4efb\u52a1\u6b65\u9aa4<\/h3>\n<p>validate-manifest<br \/>\nload-dataset<br \/>\nexecute-baseline<br \/>\nexecute-candidate<br \/>\nrun-deterministic-evaluators<br \/>\nrun-llm-judges<br \/>\naggregate-metrics<br \/>\nevaluate-gate<br \/>\nrender-report<br \/>\npublish-check<\/p>\n<h3>\u4e94\u5341\u4e00\u3001CLI\u63a5\u5165CI<\/h3>\n<p><span class=\"token annotation punctuation\">&#064;Component<\/span><br \/>\n<span class=\"token annotation punctuation\">&#064;Profile<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;ai-eval-cli&#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token keyword\">public<\/span> <span class=\"token keyword\">class<\/span> <span class=\"token class-name\">EvaluationCli<\/span><br \/>\n        <span class=\"token keyword\">implements<\/span> <span class=\"token class-name\">CommandLineRunner<\/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\">run<\/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><span class=\"token punctuation\">.<\/span> args<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><\/p>\n<p>        <span class=\"token class-name\">EvaluationRunResult<\/span> result <span class=\"token operator\">&#061;<\/span><br \/>\n                runner<span class=\"token punctuation\">.<\/span><span class=\"token function\">run<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        configLoader<br \/>\n                                <span class=\"token punctuation\">.<\/span><span class=\"token function\">fromArgs<\/span><span class=\"token punctuation\">(<\/span>args<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        reportService<span class=\"token punctuation\">.<\/span><span class=\"token function\">write<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                result<span class=\"token punctuation\">,<\/span><br \/>\n                <span class=\"token class-name\">Path<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">of<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                        <span class=\"token string\">&#034;build\/ai-evaluation&#034;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token keyword\">int<\/span> code <span class=\"token operator\">&#061;<\/span><br \/>\n                <span class=\"token keyword\">switch<\/span> <span class=\"token punctuation\">(<\/span>result<span class=\"token punctuation\">.<\/span><span class=\"token function\">decision<\/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\">case<\/span> <span class=\"token constant\">PASS<\/span> <span class=\"token operator\">-&gt;<\/span> <span class=\"token number\">0<\/span><span class=\"token punctuation\">;<\/span><br \/>\n                    <span class=\"token keyword\">case<\/span> <span class=\"token constant\">NEEDS_REVIEW<\/span> <span class=\"token operator\">-&gt;<\/span> <span class=\"token number\">2<\/span><span class=\"token punctuation\">;<\/span><br \/>\n                    <span class=\"token keyword\">case<\/span> <span class=\"token constant\">FAIL<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token constant\">ERROR<\/span> <span class=\"token operator\">-&gt;<\/span> <span class=\"token number\">1<\/span><span class=\"token punctuation\">;<\/span><br \/>\n                <span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">;<\/span><\/p>\n<p>        <span class=\"token class-name\">SpringApplication<\/span><span class=\"token punctuation\">.<\/span><span class=\"token function\">exit<\/span><span class=\"token punctuation\">(<\/span><br \/>\n                context<span class=\"token punctuation\">,<\/span><br \/>\n                <span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token operator\">-&gt;<\/span> code<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>\u4e94\u5341\u4e8c\u3001GitHub Actions<\/h3>\n<p><span class=\"token key atrule\">name<\/span><span class=\"token punctuation\">:<\/span> Spring AI Quality Gate<\/p>\n<p><span class=\"token key atrule\">on<\/span><span class=\"token punctuation\">:<\/span><br \/>\n  <span class=\"token key atrule\">pull_request<\/span><span class=\"token punctuation\">:<\/span><br \/>\n  <span class=\"token key atrule\">workflow_dispatch<\/span><span class=\"token punctuation\">:<\/span><\/p>\n<p><span class=\"token key atrule\">jobs<\/span><span class=\"token punctuation\">:<\/span><br \/>\n  <span class=\"token key atrule\">ai-quality-gate<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    <span class=\"token key atrule\">runs-on<\/span><span class=\"token punctuation\">:<\/span> ubuntu<span class=\"token punctuation\">&#8211;<\/span>latest<br \/>\n    <span class=\"token key atrule\">timeout-minutes<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token number\">45<\/span><\/p>\n<p>    <span class=\"token key atrule\">permissions<\/span><span class=\"token punctuation\">:<\/span><br \/>\n      <span class=\"token key atrule\">contents<\/span><span class=\"token punctuation\">:<\/span> read<br \/>\n      <span class=\"token key atrule\">checks<\/span><span class=\"token punctuation\">:<\/span> write<\/p>\n<p>    <span class=\"token key atrule\">steps<\/span><span class=\"token punctuation\">:<\/span><br \/>\n      <span class=\"token punctuation\">&#8211;<\/span> <span class=\"token key atrule\">uses<\/span><span class=\"token punctuation\">:<\/span> actions\/checkout&#064;v4<\/p>\n<p>      <span class=\"token punctuation\">&#8211;<\/span> <span class=\"token key atrule\">uses<\/span><span class=\"token punctuation\">:<\/span> actions\/setup<span class=\"token punctuation\">&#8211;<\/span>java&#064;v4<br \/>\n        <span class=\"token key atrule\">with<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">distribution<\/span><span class=\"token punctuation\">:<\/span> temurin<br \/>\n          <span class=\"token key atrule\">java-version<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;21&#034;<\/span><br \/>\n          <span class=\"token key atrule\">cache<\/span><span class=\"token punctuation\">:<\/span> maven<\/p>\n<p>      <span class=\"token punctuation\">&#8211;<\/span> <span class=\"token key atrule\">name<\/span><span class=\"token punctuation\">:<\/span> Unit and contract tests<br \/>\n        <span class=\"token key atrule\">run<\/span><span class=\"token punctuation\">:<\/span> .\/mvnw <span class=\"token punctuation\">&#8211;<\/span>B test<\/p>\n<p>      <span class=\"token punctuation\">&#8211;<\/span> <span class=\"token key atrule\">name<\/span><span class=\"token punctuation\">:<\/span> Package candidate<br \/>\n        <span class=\"token key atrule\">run<\/span><span class=\"token punctuation\">:<\/span> .\/mvnw <span class=\"token punctuation\">&#8211;<\/span>B <span class=\"token punctuation\">&#8211;<\/span>DskipTests package<\/p>\n<p>      <span class=\"token punctuation\">&#8211;<\/span> <span class=\"token key atrule\">name<\/span><span class=\"token punctuation\">:<\/span> Run PR evaluation<br \/>\n        <span class=\"token key atrule\">env<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">EVAL_PROVIDER_KEY<\/span><span class=\"token punctuation\">:<\/span> $<span class=\"token punctuation\">{<\/span><span class=\"token punctuation\">{<\/span> secrets.EVAL_PROVIDER_KEY <span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">}<\/span><br \/>\n          <span class=\"token key atrule\">BASELINE_RELEASE<\/span><span class=\"token punctuation\">:<\/span> production<br \/>\n          <span class=\"token key atrule\">CANDIDATE_RELEASE<\/span><span class=\"token punctuation\">:<\/span> $<span class=\"token punctuation\">{<\/span><span class=\"token punctuation\">{<\/span> github.sha <span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">}<\/span><br \/>\n        <span class=\"token key atrule\">run<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token punctuation\">|<\/span><span class=\"token scalar string\"><br \/>\n          .\/mvnw -B spring-boot:run \\\\<br \/>\n            -Pai-eval-fast \\\\<br \/>\n            -Dspring-boot.run.profiles&#061;ai-eval-cli \\\\<br \/>\n            -Dspring-boot.run.arguments&#061;&#034;\\\\<br \/>\n            &#8211;baseline&#061;${BASELINE_RELEASE} \\\\<br \/>\n            &#8211;candidate&#061;${CANDIDATE_RELEASE} \\\\<br \/>\n            &#8211;report-dir&#061;build\/ai-evaluation&#034;<\/span><\/p>\n<p>      <span class=\"token punctuation\">&#8211;<\/span> <span class=\"token key atrule\">name<\/span><span class=\"token punctuation\">:<\/span> Upload evaluation report<br \/>\n        <span class=\"token key atrule\">if<\/span><span class=\"token punctuation\">:<\/span> always()<br \/>\n        <span class=\"token key atrule\">uses<\/span><span class=\"token punctuation\">:<\/span> actions\/upload<span class=\"token punctuation\">&#8211;<\/span>artifact&#064;v4<br \/>\n        <span class=\"token key atrule\">with<\/span><span class=\"token punctuation\">:<\/span><br \/>\n          <span class=\"token key atrule\">name<\/span><span class=\"token punctuation\">:<\/span> spring<span class=\"token punctuation\">&#8211;<\/span>ai<span class=\"token punctuation\">&#8211;<\/span>evaluation<br \/>\n          <span class=\"token key atrule\">path<\/span><span class=\"token punctuation\">:<\/span> build\/ai<span class=\"token punctuation\">&#8211;<\/span>evaluation<\/p>\n<p>\u5c06ai-quality-gate\u8bbe\u4e3aRequired Status Check\u540e&#xff0c;\u5931\u8d25\u65f6\u4e0d\u80fd\u5408\u5e76\u53d7\u4fdd\u62a4\u5206\u652f\u3002<\/p>\n<h3>\u4e94\u5341\u4e09\u3001PR\u3001Nightly\u548cRelease\u4e09\u5957\u6d41\u7a0b<\/h3>\n<h4>PR Fast<\/h4>\n<ul>\n<li>\u5173\u952e\u786c\u5931\u8d25&#xff1b;<\/li>\n<li>50\u2014200\u6761&#xff1b;<\/li>\n<li>\u4f4e\u6210\u672cJudge&#xff1b;<\/li>\n<li>1\u20143\u6b21\u91cd\u590d\u3002<\/li>\n<\/ul>\n<h4>Nightly Full<\/h4>\n<ul>\n<li>\u5b8c\u6574\u56de\u5f52&#xff1b;<\/li>\n<li>\u751f\u4ea7Trace&#xff1b;<\/li>\n<li>\u591aJudge&#xff1b;<\/li>\n<li>\u7edf\u8ba1\u5206\u6790&#xff1b;<\/li>\n<li>\u8d8b\u52bf\u3002<\/li>\n<\/ul>\n<h4>Release Acceptance<\/h4>\n<ul>\n<li>\u9690\u85cf\u6570\u636e\u96c6&#xff1b;<\/li>\n<li>\u4eba\u5de5\u62bd\u6837&#xff1b;<\/li>\n<li>Preview\u73af\u5883&#xff1b;<\/li>\n<li>\u5f71\u5b50\u6d41\u91cf&#xff1b;<\/li>\n<li>\u53d1\u5e03\u6279\u51c6\u3002<\/li>\n<\/ul>\n<h3>\u4e94\u5341\u56db\u3001\u62a5\u544a<\/h3>\n<p>Decision: FAIL<\/p>\n<p>Global:<br \/>\n&#8211; Pass rate: 93.4% \u2192 92.8%<br \/>\n&#8211; Mean score: 4.32 \u2192 4.28<br \/>\n&#8211; P95 latency: &#043;9%<br \/>\n&#8211; Cost: &#043;14%<\/p>\n<p>Blocking:<br \/>\n&#8211; security: 1 failure<br \/>\n&#8211; contract-review: 93.8% &lt; 95%<br \/>\n&#8211; citation support: 95.9% &lt; 97%<\/p>\n<p>Top regressions:<br \/>\n&#8211; INCIDENT-2026-031<br \/>\n&#8211; CONTRACT-108<br \/>\n&#8211; RAG-CITATION-022<\/p>\n<h3>\u4e94\u5341\u4e94\u3001\u5931\u8d25\u8bca\u65ad<\/h3>\n<p>\u6309\u7b2c\u4e00\u6b21\u5206\u53c9\u5206\u7c7b&#xff1a;<\/p>\n<p>Routing<br \/>\nRetrieval<br \/>\nEvidence<br \/>\nGeneration<br \/>\nTool<br \/>\nOutput Validation<br \/>\nJudge Infrastructure<br \/>\nOperational<\/p>\n<p>\u4f8b\u5982\u7b54\u6848\u9519\u8bef\u4f46\u68c0\u7d22\u6b63\u786e&#xff1a;<\/p>\n<p>Generation Regression<\/p>\n<p>\u6b63\u786e\u6587\u6863\u672a\u8fdb\u5165Top K&#xff1a;<\/p>\n<p>Retrieval Regression<\/p>\n<h3>\u4e94\u5341\u516d\u3001\u53d1\u5e03\u540eCanary<\/h3>\n<p>\u90e8\u7f72\u540e\u8fd0\u884c\u5c11\u91cf\u56fa\u5b9aCase&#xff1a;<\/p>\n<p>security-canary<br \/>\ncross-tenant-canary<br \/>\ncitation-canary<br \/>\ntool-side-effect-canary<br \/>\nstructured-output-canary<\/p>\n<p>Canary\u4f7f\u7528\u4e13\u7528\u6d4b\u8bd5\u79df\u6237&#xff0c;\u4e0d\u5e94\u6c61\u67d3\u771f\u5b9e\u6570\u636e\u3002<\/p>\n<h3>\u4e94\u5341\u4e03\u3001\u5f71\u5b50\u6d41\u91cf<\/h3>\n<p>\u751f\u4ea7\u8bf7\u6c42\u590d\u5236\u7ed9Candidate&#xff1a;<\/p>\n<p>\u7528\u6237\u4ecd\u6536\u5230Baseline<br \/>\nCandidate\u53ea\u8bb0\u5f55\u7ed3\u679c<\/p>\n<p>\u5f71\u5b50\u6267\u884c\u5fc5\u987b&#xff1a;<\/p>\n<ul>\n<li>\u7981\u6b62\u771f\u5b9e\u526f\u4f5c\u7528&#xff1b;<\/li>\n<li>\u4f7f\u7528\u53ea\u8bfb\u5de5\u5177&#xff1b;<\/li>\n<li>\u8131\u654f&#xff1b;<\/li>\n<li>\u63a7\u5236\u6210\u672c&#xff1b;<\/li>\n<li>\u9075\u5b88\u6570\u636e\u533a\u57df&#xff1b;<\/li>\n<li>\u4e0d\u5f71\u54cd\u4e3b\u94fe\u8def\u5ef6\u8fdf\u3002<\/li>\n<\/ul>\n<h3>\u4e94\u5341\u516b\u3001\u7ebf\u4e0a\u8bc4\u6d4b<\/h3>\n<p>\u7ebf\u4e0a\u53ef\u4ee5\u62bd\u6837\u8fd0\u884c&#xff1a;<\/p>\n<ul>\n<li>\u683c\u5f0f\u68c0\u67e5&#xff1b;<\/li>\n<li>\u5b89\u5168&#xff1b;<\/li>\n<li>Groundedness&#xff1b;<\/li>\n<li>\u5f15\u7528\u5408\u6cd5\u6027&#xff1b;<\/li>\n<li>\u4f4e\u6210\u672cJudge\u3002<\/li>\n<\/ul>\n<p>\u4e0d\u8981\u5728\u4e3b\u8bf7\u6c42\u540c\u6b65\u7b49\u5f85Judge\u3002<\/p>\n<p>\u4e3b\u8bf7\u6c42\u5b8c\u6210<br \/>\n\u2193<br \/>\nTrace\u4e8b\u4ef6<br \/>\n\u2193<br \/>\n\u5f02\u6b65\u7ebf\u4e0aEvaluator<\/p>\n<h3>\u4e94\u5341\u4e5d\u3001\u7ebf\u4e0a\u8d28\u91cf\u4fe1\u53f7<\/h3>\n<p>\u7528\u6237\u70b9\u8e29<br \/>\n\u91cd\u590d\u63d0\u95ee<br \/>\n\u4eba\u5de5\u6539\u5199<br \/>\n\u5f15\u7528\u70b9\u51fb\u5931\u8d25<br \/>\n\u5de5\u5177\u5931\u8d25<br \/>\n\u56de\u9000\u6a21\u578b<br \/>\n\u62d2\u7b54<br \/>\n\u8d85\u65f6<br \/>\n\u9ad8\u6210\u672c<br \/>\n\u8f6c\u4eba\u5de5<br \/>\n\u4e1a\u52a1\u7ed3\u679c\u5931\u8d25<\/p>\n<p>\u7ebf\u4e0a\u4fe1\u53f7\u7528\u4e8e\u53d1\u73b0\u95ee\u9898&#xff0c;\u4e0d\u76f4\u63a5\u7b49\u540c\u4e8e\u6807\u51c6\u7b54\u6848\u3002<\/p>\n<h3>\u516d\u5341\u3001\u5931\u8d25\u56de\u6d41<\/h3>\n<p>\u7ebf\u4e0a\u5931\u8d25<br \/>\n\u2193<br \/>\n\u5019\u9009\u6c60<br \/>\n\u2193<br \/>\n\u8131\u654f<br \/>\n\u2193<br \/>\n\u4e13\u5bb6\u6807\u6ce8<br \/>\n\u2193<br \/>\n\u65b0Incident Case<br \/>\n\u2193<br \/>\n\u56de\u5f52\u6570\u636e\u96c6\u65b0\u7248\u672c<br \/>\n\u2193<br \/>\n\u4fee\u590d\u540e\u91cd\u65b0\u8bc4\u6d4b<\/p>\n<p>\u6bcf\u6b21\u4e25\u91cd\u4e8b\u6545\u81f3\u5c11\u65b0\u589e\u4e00\u4e2a\u786c\u5931\u8d25Case\u3002<\/p>\n<h3>\u516d\u5341\u4e00\u3001Judge\u81ea\u8eabCanary<\/h3>\n<p>Judge\u6a21\u578b\u4e5f\u53ef\u80fd\u5347\u7ea7\u3002<\/p>\n<p>\u56fa\u5b9aJudge Canary&#xff1a;<\/p>\n<ul>\n<li>\u660e\u786ePASS&#xff1b;<\/li>\n<li>\u660e\u786eFAIL&#xff1b;<\/li>\n<li>\u4f4d\u7f6e\u4ea4\u6362&#xff1b;<\/li>\n<li>\u5197\u957f\u53d8\u4f53&#xff1b;<\/li>\n<li>\u591a\u8bed\u8a00&#xff1b;<\/li>\n<li>Prompt Injection\u3002<\/li>\n<\/ul>\n<p>\u76d1\u63a7&#xff1a;<\/p>\n<p>judge_repeat_consistency<br \/>\njudge_position_flip_rate<br \/>\njudge_human_agreement<br \/>\njudge_canary_score<\/p>\n<p>Judge\u6f02\u79fb\u65f6\u51bb\u7ed3\u81ea\u52a8\u95e8\u7981\u3002<\/p>\n<h3>\u516d\u5341\u4e8c\u3001\u6a21\u578b\u5347\u7ea7\u6d41\u7a0b<\/h3>\n<p>\u65b0\u6a21\u578bProfile<br \/>\n\u2193<br \/>\n\u5951\u7ea6\u6d4b\u8bd5<br \/>\n\u2193<br \/>\n\u5f00\u53d1\u6570\u636e\u96c6<br \/>\n\u2193<br \/>\n\u5b8c\u6574\u79bb\u7ebf\u56de\u5f52<br \/>\n\u2193<br \/>\n\u9690\u85cf\u9a8c\u6536<br \/>\n\u2193<br \/>\n\u5f71\u5b50\u6d41\u91cf<br \/>\n\u2193<br \/>\n\u5c0f\u6bd4\u4f8bCanary<br \/>\n\u2193<br \/>\n\u9010\u6b65\u6269\u5927<\/p>\n<p>\u4e0d\u8981\u76f4\u63a5\u628alatest\u522b\u540d\u6307\u5411\u65b0\u6a21\u578b\u5e76\u5168\u91cf\u4e0a\u7ebf\u3002<\/p>\n<h3>\u516d\u5341\u4e09\u3001Prompt\u5347\u7ea7\u6d41\u7a0b<\/h3>\n<p>Prompt\u672c\u8eab\u4f5c\u4e3aArtifact&#xff1a;<\/p>\n<p>promptId<br \/>\nversion<br \/>\nhash<br \/>\nowner<br \/>\nchangeReason<br \/>\napprovedBy<\/p>\n<p>\u6bcf\u6b21\u53d8\u66f4\u5fc5\u987b\u5173\u8054&#xff1a;<\/p>\n<ul>\n<li>Evaluation Run&#xff1b;<\/li>\n<li>Gate Result&#xff1b;<\/li>\n<li>Release ID\u3002<\/li>\n<\/ul>\n<h3>\u516d\u5341\u56db\u3001\u77e5\u8bc6\u7d22\u5f15\u5347\u7ea7\u6d41\u7a0b<\/h3>\n<p>\u65b0Chunk\u7b56\u7565\u6216Embedding\u6a21\u578b\u9700\u8981&#xff1a;<\/p>\n<p>\u7d22\u5f15\u6784\u5efa<br \/>\n\u2193<br \/>\n\u68c0\u7d22Recall\u8bc4\u6d4b<br \/>\n\u2193<br \/>\n\u5f15\u7528\u8bc4\u6d4b<br \/>\n\u2193<br \/>\n\u7b54\u6848\u8bc4\u6d4b<br \/>\n\u2193<br \/>\n\u5f71\u5b50\u53cc\u67e5<br \/>\n\u2193<br \/>\n\u5207\u6362\u522b\u540d<\/p>\n<p>\u53ea\u8bc4\u6d4b\u6700\u7ec8\u7b54\u6848\u5f88\u96be\u5b9a\u4f4d\u7d22\u5f15\u95ee\u9898\u3002<\/p>\n<h3>\u516d\u5341\u4e94\u3001Tool\u76ee\u5f55\u5347\u7ea7\u6d41\u7a0b<\/h3>\n<p>Tool\u63cf\u8ff0\u53d8\u5316\u53ef\u80fd\u6539\u53d8\u6a21\u578b\u9009\u62e9\u3002<\/p>\n<p>\u9700\u8981\u6bd4\u8f83&#xff1a;<\/p>\n<ul>\n<li>\u9009Tool\u51c6\u786e\u7387&#xff1b;<\/li>\n<li>\u53c2\u6570&#xff1b;<\/li>\n<li>\u8c03\u7528\u6b21\u6570&#xff1b;<\/li>\n<li>\u6743\u9650&#xff1b;<\/li>\n<li>\u5e42\u7b49&#xff1b;<\/li>\n<li>\u526f\u4f5c\u7528&#xff1b;<\/li>\n<li>\u5ef6\u8fdf\u3002<\/li>\n<\/ul>\n<p>\u65b0Tool\u9ed8\u8ba4\u4e0d\u5e94\u5bf9\u6240\u6709Agent\u81ea\u52a8\u5f00\u653e\u3002<\/p>\n<h3>\u516d\u5341\u516d\u3001\u76d1\u63a7\u6307\u6807<\/h3>\n<p>ai_eval_run_total{<br \/>\n  status,<br \/>\n  gate_policy<br \/>\n}<\/p>\n<p>ai_eval_pass_rate{<br \/>\n  slice<br \/>\n}<\/p>\n<p>ai_eval_score{<br \/>\n  evaluator,<br \/>\n  slice<br \/>\n}<\/p>\n<p>ai_eval_candidate_win_rate{<br \/>\n  slice<br \/>\n}<\/p>\n<p>ai_eval_hard_failure_total{<br \/>\n  category<br \/>\n}<\/p>\n<p>ai_eval_evaluator_error_rate{<br \/>\n  evaluator<br \/>\n}<\/p>\n<p>ai_eval_latency_seconds{<br \/>\n  target<br \/>\n}<\/p>\n<p>ai_eval_cost_total{<br \/>\n  target,<br \/>\n  evaluator<br \/>\n}<\/p>\n<p>ai_eval_gate_violation_total{<br \/>\n  rule,<br \/>\n  severity<br \/>\n}<\/p>\n<p>ai_judge_position_flip_rate<\/p>\n<p>ai_judge_repeat_consistency<\/p>\n<p>ai_canary_failure_total{<br \/>\n  canary_type<br \/>\n}<\/p>\n<h3>\u516d\u5341\u4e03\u3001\u6d4b\u8bd5\u573a\u666f<\/h3>\n<h4>1. Manifest\u4e0d\u4e00\u81f4<\/h4>\n<p>\u8bc4\u6d4b\u58f0\u660eV42&#xff0c;\u8fd0\u884c\u65f6V41&#xff0c;Run\u5fc5\u987bERROR\u3002<\/p>\n<h4>2. \u5168\u5c40\u63d0\u5347\u4f46\u5b89\u5168\u5931\u8d25<\/h4>\n<p>Gate\u5fc5\u987bFAIL\u3002<\/p>\n<h4>3. Judge 5xx<\/h4>\n<p>Case\u4e3aEvaluator Error&#xff0c;\u4e0d\u76f4\u63a5\u5224Candidate\u5931\u8d25\u3002<\/p>\n<h4>4. Pairwise\u4f4d\u7f6e\u53cd\u8f6c<\/h4>\n<p>\u8fdb\u5165Needs Review\u3002<\/p>\n<h4>5. \u6a21\u578b\u968f\u673a\u5931\u8d25<\/h4>\n<p>\u91cd\u590d\u91c7\u6837\u901a\u8fc7\u7387\u4f4e\u4e8e\u9608\u503c\u65f6FAIL\u3002<\/p>\n<h4>6. RAG\u7981\u6b62\u6587\u6863<\/h4>\n<p>\u5373\u4f7f\u6700\u7ec8\u7b54\u6848\u6ca1\u6cc4\u9732&#xff0c;\u53ea\u8981\u5019\u9009Evidence\u51fa\u73b0\u7981\u6b62\u6587\u6863&#xff0c;\u4e5fFAIL\u3002<\/p>\n<h4>7. \u5de5\u5177\u91cd\u590d<\/h4>\n<p>\u6700\u7ec8\u7ed3\u679c\u6b63\u786e\u4f46\u91cd\u590d\u6267\u884c\u526f\u4f5c\u7528&#xff0c;FAIL\u3002<\/p>\n<h4>8. \u6210\u672c\u56de\u5f52<\/h4>\n<p>\u8d28\u91cf\u63d0\u5347\u4f46\u6210\u672c\u8d85\u51fa\u7b56\u7565&#xff0c;\u963b\u65ad\u6216\u4eba\u5de5\u5ba1\u6279\u3002<\/p>\n<h3>\u516d\u5341\u516b\u3001\u53d1\u5e03\u56de\u6eda<\/h3>\n<p>\u53d1\u5e03\u8bb0\u5f55&#xff1a;<\/p>\n<p><span class=\"token keyword\">public<\/span> <span class=\"token keyword\">record<\/span> <span class=\"token class-name\">AiReleaseRecord<\/span><span class=\"token punctuation\">(<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> releaseId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">AiReleaseManifest<\/span> manifest<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> evaluationRunId<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">QualityGateDecision<\/span> gate<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">Instant<\/span> deployedAt<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token class-name\">String<\/span> previousReleaseId<span class=\"token punctuation\">)<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>\u7ebf\u4e0aCanary\u5931\u8d25\u65f6&#xff0c;\u56de\u6eda&#xff1a;<\/p>\n<ul>\n<li>\u6a21\u578b\u8def\u7531Profile&#xff1b;<\/li>\n<li>Prompt\u522b\u540d&#xff1b;<\/li>\n<li>\u77e5\u8bc6\u7d22\u5f15\u522b\u540d&#xff1b;<\/li>\n<li>Tool\u76ee\u5f55&#xff1b;<\/li>\n<li>\u5e94\u7528\u7248\u672c\u3002<\/li>\n<\/ul>\n<p>AI\u53d1\u5e03\u4e0d\u662f\u53ea\u56de\u6edaDocker\u955c\u50cf\u3002<\/p>\n<h3>\u516d\u5341\u4e5d\u3001\u4e0a\u7ebf\u68c0\u67e5\u6e05\u5355<\/h3>\n<p>\u25a1 \u5df2\u5efa\u7acb\u4e94\u5c42\u6d4b\u8bd5\u4f53\u7cfb<br \/>\n\u25a1 Release Manifest\u8986\u76d6\u6a21\u578b\u3001Prompt\u3001\u7d22\u5f15\u3001Tool\u548cMemory<br \/>\n\u25a1 \u8fd0\u884c\u65f6\u80fd\u591f\u56de\u663e\u771f\u5b9eManifest<br \/>\n\u25a1 \u6570\u636e\u96c6\u5305\u542b\u9ec4\u91d1\u3001\u751f\u4ea7Trace\u3001\u5b89\u5168\u548c\u4e8b\u6545\u6837\u672c<br \/>\n\u25a1 \u6570\u636e\u96c6\u4e0eCase\u5747\u7248\u672c\u5316<br \/>\n\u25a1 \u786e\u5b9a\u6027\u89c4\u5219\u4f18\u5148\u4e8eJudge<br \/>\n\u25a1 RAG\u5206\u522b\u8bc4\u6d4b\u68c0\u7d22\u3001\u8bc1\u636e\u3001Claim\u548c\u5f15\u7528<br \/>\n\u25a1 Agent\u8bc4\u6d4b\u5b8c\u6574Tool\u8f68\u8ff9<br \/>\n\u25a1 Judge\u4f7f\u7528\u56fa\u5b9aManifest\u5e76\u7ecf\u8fc7\u4eba\u5de5\u6821\u51c6<br \/>\n\u25a1 Pairwise\u6267\u884c\u4f4d\u7f6e\u4e92\u6362<br \/>\n\u25a1 \u9ad8\u98ce\u9669Case\u91cd\u590d\u8fd0\u884c<br \/>\n\u25a1 \u6307\u6807\u6309\u4e1a\u52a1Slice\u805a\u5408<br \/>\n\u25a1 \u5b89\u5168\u548c\u526f\u4f5c\u7528\u786c\u5931\u8d25\u96f6\u5bb9\u5fcd<br \/>\n\u25a1 Gate\u540c\u65f6\u68c0\u67e5\u8d28\u91cf\u3001\u6210\u672c\u548c\u5ef6\u8fdf<br \/>\n\u25a1 PR\u3001Nightly\u548cRelease\u5206\u5c42\u8fd0\u884c<br \/>\n\u25a1 Required Status Check\u5df2\u7ecf\u542f\u7528<br \/>\n\u25a1 \u53d1\u5e03\u540e\u8fd0\u884cCanary\u548c\u5f71\u5b50\u6d41\u91cf<br \/>\n\u25a1 Judge\u672c\u8eab\u6709Canary<br \/>\n\u25a1 \u7ebf\u4e0a\u5931\u8d25\u80fd\u591f\u56de\u6d41\u6570\u636e\u96c6<br \/>\n\u25a1 \u53d1\u5e03\u8bb0\u5f55\u652f\u6301\u6a21\u578b\u3001Prompt\u548c\u7d22\u5f15\u56de\u6eda<\/p>\n<h3>\u4e03\u5341\u3001\u672c\u7ae0\u5b8c\u6574\u94fe\u8def<\/h3>\n<p>\u4ee3\u7801\/Prompt\/\u6a21\u578b\/\u7d22\u5f15\u53d8\u66f4<br \/>\n\u2193<br \/>\n\u751f\u6210Candidate Manifest<br \/>\n\u2193<br \/>\n\u786e\u5b9a\u6027\u5355\u5143\u4e0e\u5951\u7ea6\u6d4b\u8bd5<br \/>\n\u2193<br \/>\n\u52a0\u8f7d\u51bb\u7ed3\u6570\u636e\u96c6<br \/>\n\u2193<br \/>\nBaseline\/Candidate\u91cd\u590d\u6267\u884c<br \/>\n\u2193<br \/>\n\u89c4\u5219Evaluator<br \/>\n\u2193<br \/>\nSpring AI Evaluator<br \/>\n\u2193<br \/>\nLLM Judge<br \/>\n\u2193<br \/>\nSlice\u4e0e\u7edf\u8ba1\u805a\u5408<br \/>\n\u2193<br \/>\nQuality Gate<br \/>\n\u2193<br \/>\nGitHub Required Check<br \/>\n\u2193<br \/>\nPreview\u4e0e\u5f71\u5b50<br \/>\n\u2193<br \/>\nCanary<br \/>\n\u2193<br \/>\n\u53d1\u5e03<br \/>\n\u2193<br \/>\n\u7ebf\u4e0a\u8bc4\u6d4b\u4e0e\u5931\u8d25\u56de\u6d41<\/p>\n<h3>\u603b\u7ed3<\/h3>\n<p>\u4f01\u4e1a\u7ea7Spring AI\u5e94\u7528\u4e0d\u80fd\u4f9d\u8d56\u201c\u4eba\u5de5\u8bd5\u51e0\u4e2a\u95ee\u9898\u611f\u89c9\u8fd8\u884c\u201d\u53d1\u5e03&#xff0c;\u4e5f\u4e0d\u80fd\u628a\u5355\u6b21LLM Judge\u5206\u6570\u5f53\u4f5c\u7edd\u5bf9\u6807\u51c6\u3002<\/p>\n<p>\u53ef\u9760\u8d28\u91cf\u4f53\u7cfb\u9700\u8981&#xff1a;<\/p>\n<p>Release Manifest<br \/>\n&#xff0b;\u7248\u672c\u5316\u9ec4\u91d1\u6570\u636e\u96c6<br \/>\n&#xff0b;\u786e\u5b9a\u6027\u6d4b\u8bd5<br \/>\n&#xff0b;Spring AI Evaluator<br \/>\n&#xff0b;\u6821\u51c6Judge<br \/>\n&#xff0b;\u91cd\u590d\u91c7\u6837<br \/>\n&#xff0b;\u4e1a\u52a1Slice<br \/>\n&#xff0b;\u786c\u5931\u8d25<br \/>\n&#xff0b;CI\u8d28\u91cf\u95e8\u7981<br \/>\n&#xff0b;\u7ebf\u4e0a\u53cd\u9988\u95ed\u73af<\/p>\n<p>\u5f53\u6bcf\u4e00\u6b21\u6a21\u578b\u3001Prompt\u3001RAG\u548cTool\u53d8\u66f4\u90fd\u5fc5\u987b\u901a\u8fc7\u76f8\u540c\u7684\u53ef\u590d\u73b0\u57fa\u51c6&#xff0c;\u5e76\u4e14\u751f\u4ea7\u4e8b\u6545\u4f1a\u6c38\u4e45\u8fdb\u5165\u56de\u5f52\u96c6&#xff0c;AI\u7cfb\u7edf\u624d\u5177\u5907\u6301\u7eed\u5347\u7ea7\u800c\u4e0d\u6301\u7eed\u5931\u63a7\u7684\u5de5\u7a0b\u57fa\u7840\u3002<\/p>\n<p>\u4e0b\u4e00\u7bc7\u5c06\u7ee7\u7eed\u5b9e\u73b0&#xff1a;<\/p>\n<p>Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218&#xff08;14&#xff09;&#xff1a;\u5bb9\u5668\u5316\u90e8\u7f72\u3001\u7070\u5ea6\u53d1\u5e03\u4e0e\u81ea\u52a8\u56de\u6eda\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6458\u8981<br \/>\n\u524d\u9762\u7684\u7ae0\u8282\u5df2\u7ecf\u5b8c\u6210\u591a\u6a21\u578b\u8def\u7531\u3001\u9884\u7b97\u63a7\u5236\u3001\u6743\u9650\u611f\u77e5RAG\u548c\u5f02\u6b65\u957f\u4efb\u52a1\u3002\u7cfb\u7edf\u53ef\u4ee5\u8c03\u7528\u6a21\u578b\u3001\u68c0\u7d22\u4f01\u4e1a\u77e5\u8bc6\u3001\u6267\u884c\u5de5\u5177\u5e76\u5728\u6545\u969c\u540e\u6062\u590d&#xff0c;\u4f46\u4ecd\u7f3a\u5c11\u4e00\u4e2a\u51b3\u5b9a\u80fd\u5426\u7a33\u5b9a\u8fed\u4ee3\u7684\u6838\u5fc3\u80fd\u529b&#xff1a;\u5982\u4f55\u8bc1\u660e\u4e00\u6b21Prompt\u3001\u6a21\u578b\u3001RAG\u7d22\u5f15\u3001Advisor\u6216Tool\u76ee\u5f55\u53d8\u66f4\u6ca1\u6709\u7834\u574f\u73b0\u6709\u4e1a\u52a1\u3002<br \/>\n\u4f20\u7edf\u5355\u5143\u6d4b\u8bd5\u80fd\u591f\u9a8c\u8bc1\u72b6\u6001\u7801\u3001Schema\u548c\u4e1a\u52a1\u89c4\u5219&#xff0c;\u5374\u65e0\u6cd5\u5b8c\u6574\u5224\u65ad\u5f00\u653e\u5f0f\u56de\u7b54\u662f\u5426\u66f4\u597d&#xff1b;\u53ea\u4f9d\u8d56LLM-as-Judge\u53c8\u4f1a\u53d7\u5230\u968f\u673a\u6027\u3001\u4f4d\u7f6e\u504f\u5dee\u3001\u957f\u5ea6\u504f\u5dee\u548cJ<\/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":[99,101,50],"topic":[],"class_list":["post-91588","post","type-post","status-publish","format-standard","hentry","category-server","tag-java","tag-spring","tag-50"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981 - \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\/91588.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u6587\u7ae0\u6458\u8981 \u524d\u9762\u7684\u7ae0\u8282\u5df2\u7ecf\u5b8c\u6210\u591a\u6a21\u578b\u8def\u7531\u3001\u9884\u7b97\u63a7\u5236\u3001\u6743\u9650\u611f\u77e5RAG\u548c\u5f02\u6b65\u957f\u4efb\u52a1\u3002\u7cfb\u7edf\u53ef\u4ee5\u8c03\u7528\u6a21\u578b\u3001\u68c0\u7d22\u4f01\u4e1a\u77e5\u8bc6\u3001\u6267\u884c\u5de5\u5177\u5e76\u5728\u6545\u969c\u540e\u6062\u590d&#xff0c;\u4f46\u4ecd\u7f3a\u5c11\u4e00\u4e2a\u51b3\u5b9a\u80fd\u5426\u7a33\u5b9a\u8fed\u4ee3\u7684\u6838\u5fc3\u80fd\u529b&#xff1a;\u5982\u4f55\u8bc1\u660e\u4e00\u6b21Prompt\u3001\u6a21\u578b\u3001RAG\u7d22\u5f15\u3001Advisor\u6216Tool\u76ee\u5f55\u53d8\u66f4\u6ca1\u6709\u7834\u574f\u73b0\u6709\u4e1a\u52a1\u3002 \u4f20\u7edf\u5355\u5143\u6d4b\u8bd5\u80fd\u591f\u9a8c\u8bc1\u72b6\u6001\u7801\u3001Schema\u548c\u4e1a\u52a1\u89c4\u5219&#xff0c;\u5374\u65e0\u6cd5\u5b8c\u6574\u5224\u65ad\u5f00\u653e\u5f0f\u56de\u7b54\u662f\u5426\u66f4\u597d&#xff1b;\u53ea\u4f9d\u8d56LLM-as-Judge\u53c8\u4f1a\u53d7\u5230\u968f\u673a\u6027\u3001\u4f4d\u7f6e\u504f\u5dee\u3001\u957f\u5ea6\u504f\u5dee\u548cJ\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/91588.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-08-07T17:48:44+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=\"12 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/91588.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/91588.html\",\"name\":\"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-08-07T17:48:44+00:00\",\"dateModified\":\"2026-08-07T17:48:44+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/91588.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/91588.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/91588.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981\"}]},{\"@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":"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981 - \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\/91588.html","og_locale":"zh_CN","og_type":"article","og_title":"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u6587\u7ae0\u6458\u8981 \u524d\u9762\u7684\u7ae0\u8282\u5df2\u7ecf\u5b8c\u6210\u591a\u6a21\u578b\u8def\u7531\u3001\u9884\u7b97\u63a7\u5236\u3001\u6743\u9650\u611f\u77e5RAG\u548c\u5f02\u6b65\u957f\u4efb\u52a1\u3002\u7cfb\u7edf\u53ef\u4ee5\u8c03\u7528\u6a21\u578b\u3001\u68c0\u7d22\u4f01\u4e1a\u77e5\u8bc6\u3001\u6267\u884c\u5de5\u5177\u5e76\u5728\u6545\u969c\u540e\u6062\u590d&#xff0c;\u4f46\u4ecd\u7f3a\u5c11\u4e00\u4e2a\u51b3\u5b9a\u80fd\u5426\u7a33\u5b9a\u8fed\u4ee3\u7684\u6838\u5fc3\u80fd\u529b&#xff1a;\u5982\u4f55\u8bc1\u660e\u4e00\u6b21Prompt\u3001\u6a21\u578b\u3001RAG\u7d22\u5f15\u3001Advisor\u6216Tool\u76ee\u5f55\u53d8\u66f4\u6ca1\u6709\u7834\u574f\u73b0\u6709\u4e1a\u52a1\u3002 \u4f20\u7edf\u5355\u5143\u6d4b\u8bd5\u80fd\u591f\u9a8c\u8bc1\u72b6\u6001\u7801\u3001Schema\u548c\u4e1a\u52a1\u89c4\u5219&#xff0c;\u5374\u65e0\u6cd5\u5b8c\u6574\u5224\u65ad\u5f00\u653e\u5f0f\u56de\u7b54\u662f\u5426\u66f4\u597d&#xff1b;\u53ea\u4f9d\u8d56LLM-as-Judge\u53c8\u4f1a\u53d7\u5230\u968f\u673a\u6027\u3001\u4f4d\u7f6e\u504f\u5dee\u3001\u957f\u5ea6\u504f\u5dee\u548cJ","og_url":"https:\/\/www.wsisp.com\/helps\/91588.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-08-07T17:48:44+00:00","author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"12 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/91588.html","url":"https:\/\/www.wsisp.com\/helps\/91588.html","name":"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-08-07T17:48:44+00:00","dateModified":"2026-08-07T17:48:44+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/91588.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/91588.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/91588.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"Spring AI\u4f01\u4e1a\u7ea7\u5e94\u7528\u5b9e\u6218\uff0813\uff09\uff1a\u6d4b\u8bd5\u3001\u9ec4\u91d1\u6570\u636e\u96c6\u4e0e\u53d1\u5e03\u8d28\u91cf\u95e8\u7981"}]},{"@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\/91588","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=91588"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/91588\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=91588"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=91588"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=91588"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=91588"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}