zengli
/
coding-agent


			
							12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072
							<!doctype html>
<html lang="zh" class="no-js">
  <head>
    
      <meta charset="utf-8">
      <meta name="viewport" content="width=device-width,initial-scale=1">
      
      
        <link rel="prev" href="4.6%20prompts%E8%AE%BE%E8%AE%A1%E4%B8%8Equery%E6%84%8F%E5%9B%BE%E8%AF%86%E5%88%AB.html">
      
      
        <link rel="next" href="4.8%20RAG%E7%B3%BB%E7%BB%9F%E8%BF%90%E8%A1%8C.html">
      
      
      <link rel="icon" href="../img/AI.jpg">
      <meta name="generator" content="mkdocs-1.6.1, mkdocs-material-9.6.21">
    
    
        <title>4.7 检索策略与RAG系统设计 - 企业知识库RAG智能问答项目V1.0</title>
      
    
      <link rel="stylesheet" href="../assets/stylesheets/main.2a3383ac.min.css">
      
      
        <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
        <link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Roboto:300,300i,400,400i,700,700i%7CRoboto+Mono:400,400i,700,700i&amp;display=fallback">
        <style>:root{--md-text-font:"Roboto";--md-code-font:"Roboto Mono"}</style>
      
    
      <link rel="stylesheet" href="../stylesheets/extra.css">
    
    <script>__md_scope=new URL("..",location),__md_hash=e=>[...e].reduce(((e,_)=>(e<<5)-e+_.charCodeAt(0)),0),__md_get=(e,_=localStorage,t=__md_scope)=>JSON.parse(_.getItem(t.pathname+"."+e)),__md_set=(e,_,t=localStorage,a=__md_scope)=>{try{t.setItem(a.pathname+"."+e,JSON.stringify(_))}catch(e){}}</script>
    
      
  </head>
  
  
    <body dir="ltr">
  
    
    <input class="md-toggle" data-md-toggle="drawer" type="checkbox" id="__drawer" autocomplete="off">
    <input class="md-toggle" data-md-toggle="search" type="checkbox" id="__search" autocomplete="off">
    <label class="md-overlay" for="__drawer"></label>
    <div data-md-component="skip">
      
        
        <a href="#_1" class="md-skip">
          跳转至
        </a>
      
    </div>
    <div data-md-component="announce">
      
    </div>
    
    
<header class="md-header md-header--shadow" data-md-component="header">
  <nav class="md-header__inner md-grid" aria-label="页眉">
    <a href="../index.html" title="企业知识库RAG智能问答项目V1.0" class="md-header__button md-logo" aria-label="企业知识库RAG智能问答项目V1.0" data-md-component="logo">
      
  <img src="../img/AI.jpg" alt="logo">

    </a>
    <label class="md-header__button md-icon" for="__drawer">
      
      <svg xmlns="http://www.w3.org/2000/svg" viewbox="0 0 24 24"><path d="M3 6h18v2H3zm0 5h18v2H3zm0 5h18v2H3z"></path></svg>
    </label>
    <div class="md-header__title" data-md-component="header-title">
      <div class="md-header__ellipsis">
        <div class="md-header__topic">
          <span class="md-ellipsis">
            企业知识库RAG智能问答项目V1.0
          </span>
        </div>
        <div class="md-header__topic" data-md-component="header-topic">
          <span class="md-ellipsis">
            
              4.7 检索策略与RAG系统设计
            
          </span>
        </div>
      </div>
    </div>
    
    
      <script>var palette=__md_get("__palette");if(palette&&palette.color){if("(prefers-color-scheme)"===palette.color.media){var media=matchMedia("(prefers-color-scheme: light)"),input=document.querySelector(media.matches?"[data-md-color-media='(prefers-color-scheme: light)']":"[data-md-color-media='(prefers-color-scheme: dark)']");palette.color.media=input.getAttribute("data-md-color-media"),palette.color.scheme=input.getAttribute("data-md-color-scheme"),palette.color.primary=input.getAttribute("data-md-color-primary"),palette.color.accent=input.getAttribute("data-md-color-accent")}for(var[key,value]of Object.entries(palette.color))document.body.setAttribute("data-md-color-"+key,value)}</script>
    
    
        <label class="md-header__button md-icon" for="__search">
          
          <svg xmlns="http://www.w3.org/2000/svg" viewbox="0 0 24 24"><path d="M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.52 6.52 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5"></path></svg>
        </label>
        <div class="md-search" data-md-component="search" role="dialog">
  <label class="md-search__overlay" for="__search"></label>
  <div class="md-search__inner" role="search">
    <form class="md-search__form" name="search">
      <input type="text" class="md-search__input" name="query" aria-label="搜索" placeholder="搜索" autocapitalize="off" autocorrect="off" autocomplete="off" spellcheck="false" data-md-component="search-query" required>
      <label class="md-search__icon md-icon" for="__search">
        
        <svg xmlns="http://www.w3.org/2000/svg" viewbox="0 0 24 24"><path d="M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.52 6.52 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5"></path></svg>
        
        <svg xmlns="http://www.w3.org/2000/svg" viewbox="0 0 24 24"><path d="M20 11v2H8l5.5 5.5-1.42 1.42L4.16 12l7.92-7.92L13.5 5.5 8 11z"></path></svg>
      </label>
      <nav class="md-search__options" aria-label="查找">
        
        <button type="reset" class="md-search__icon md-icon" title="清空当前内容" aria-label="清空当前内容" tabindex="-1">
          
          <svg xmlns="http://www.w3.org/2000/svg" viewbox="0 0 24 24"><path d="M19 6.41 17.59 5 12 10.59 6.41 5 5 6.41 10.59 12 5 17.59 6.41 19 12 13.41 17.59 19 19 17.59 13.41 12z"></path></svg>
        </button>
      </nav>
      
    </form>
    <div class="md-search__output">
      <div class="md-search__scrollwrap" tabindex="0" data-md-scrollfix>
        <div class="md-search-result" data-md-component="search-result">
          <div class="md-search-result__meta">
            正在初始化搜索引擎
          </div>
          <ol class="md-search-result__list" role="presentation"></ol>
        </div>
      </div>
    </div>
  </div>
</div>
      
    
  </nav>
  
</header>
    
    <div class="md-container" data-md-component="container">
      
      
      <main class="md-main" data-md-component="main">
        <div class="md-main__inner md-grid">
          
            
              <div class="md-sidebar md-sidebar--primary" data-md-component="sidebar" data-md-type="navigation">
                <div class="md-sidebar__scrollwrap">
                  <div class="md-sidebar__inner">
                    

<nav class="md-nav md-nav--primary" aria-label="导航栏" data-md-level="0">
  <label class="md-nav__title" for="__drawer">
    <a href="../index.html" title="企业知识库RAG智能问答项目V1.0" class="md-nav__button md-logo" aria-label="企业知识库RAG智能问答项目V1.0" data-md-component="logo">
      
  <img src="../img/AI.jpg" alt="logo">

    </a>
    企业知识库RAG智能问答项目V1.0
  </label>
  
  <ul class="md-nav__list" data-md-scrollfix>
    
      
    <li class="md-nav__item">
      <a href="../index.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    知识库内容
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_2">
        
          
          <label class="md-nav__link" for="__nav_2" id="__nav_2_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ① 项目概述
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_2_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_2">
            <span class="md-nav__icon md-icon"></span>
            ① 项目概述
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A0%20%E9%A1%B9%E7%9B%AE%E6%A6%82%E8%BF%B0/1.1%20%E9%A1%B9%E7%9B%AE%E6%9E%B6%E6%9E%84.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    1.1 项目架构
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A0%20%E9%A1%B9%E7%9B%AE%E6%A6%82%E8%BF%B0/1.2%20%E9%A1%B9%E7%9B%AE%E7%8E%AF%E5%A2%83%28window%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    1.2 项目环境(window)
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A0%20%E9%A1%B9%E7%9B%AE%E6%A6%82%E8%BF%B0/1.3%20%E9%A1%B9%E7%9B%AE%E7%8E%AF%E5%A2%83%28mac%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    1.3 项目环境(mac)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_3">
        
          
          <label class="md-nav__link" for="__nav_3" id="__nav_3_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ② 项目工具
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_3_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_3">
            <span class="md-nav__icon md-icon"></span>
            ② 项目工具
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
              
                
    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.1%20Ollama%E5%A4%A7%E6%A8%A1%E5%9E%8B%E9%AB%98%E6%95%88%E7%AE%A1%E7%90%86%E5%B7%A5%E5%85%B7.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.1 Ollama大模型高效管理工具
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.2%20LangChain%E5%9F%BA%E7%A1%80%E7%9F%A5%E8%AF%86.md.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.2 LangChain基础知识.md
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.3%20%E5%90%91%E9%87%8F%E7%B4%A2%E5%BC%95%EF%BC%9AHNSW%E3%80%81IVF%E4%B8%8EIVF_PQ.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.3 向量索引：HNSW、IVF与IVF_PQ
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.4%20Milvus%E5%90%91%E9%87%8F%E6%95%B0%E6%8D%AE%E5%BA%93.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.4 Milvus向量数据库
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_4">
        
          
          <label class="md-nav__link" for="__nav_4" id="__nav_4_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ③ 基于Mysql库的问答系统
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_4_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_4">
            <span class="md-nav__icon md-icon"></span>
            ③ 基于Mysql库的问答系统
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.1%20Python%E6%97%A5%E5%BF%97%E4%BB%8B%E7%BB%8D%E4%B8%8E%E5%BA%94%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.1 Python日志介绍与应用
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.2%20Redis%E6%95%B0%E6%8D%AE%E5%BA%93%E7%AE%80%E4%BB%8B%E4%B8%8E%E4%BD%BF%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.2 Redis数据库简介与使用
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.3%20%E5%9F%BA%E4%BA%8EMysql%E6%95%B0%E6%8D%AE%E5%BA%93%E5%AE%9E%E7%8E%B0%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.3 基于Mysql数据库实现问答系统
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.4%20BM25%E7%AE%97%E6%B3%95%E7%AE%80%E4%BB%8B%E4%B8%8E%E5%BA%94%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.4 BM25算法简介与应用
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--active md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_5" checked>
        
          
          <label class="md-nav__link" for="__nav_5" id="__nav_5_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ④ 基于Milvus库的问答系统
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_5_label" aria-expanded="true">
          <label class="md-nav__title" for="__nav_5">
            <span class="md-nav__icon md-icon"></span>
            ④ 基于Milvus库的问答系统
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="4.1%20%E6%95%B4%E4%BD%93%E6%9E%B6%E6%9E%84%E4%B8%8E%E5%B7%A5%E7%A8%8B%E6%B5%81%E7%A8%8B.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.1 整体架构与工程流程
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="4.2%20%E5%9F%BA%E7%A1%80%E6%A8%A1%E5%9D%97.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.2 基础模块
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="4.3%20%E6%96%87%E6%A1%A3%E5%A4%84%E7%90%86%E6%A8%A1%E5%9D%97.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.3 文档处理模块
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="4.4%20%E5%90%91%E9%87%8F%E5%AD%98%E5%82%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.4 向量存储
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item"><a href="4.5%20BERT%E5%9F%BA%E7%A1%80%E4%B8%8E%E4%BB%BB%E5%8A%A1%E5%BE%AE%E8%B0%83.html" class="md-nav__link"><span class="md-ellipsis">4.5 BERT基础与任务微调</span></a></li><li class="md-nav__item">
      <a href="4.6%20prompts%E8%AE%BE%E8%AE%A1%E4%B8%8Equery%E6%84%8F%E5%9B%BE%E8%AF%86%E5%88%AB.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.6 prompts设计与query意图识别
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item md-nav__item--active">
      
      <input class="md-nav__toggle md-toggle" type="checkbox" id="__toc">
      
      
        <label class="md-nav__link md-nav__link--active" for="__toc">
          
  
  <span class="md-ellipsis">
    4.7 检索策略与RAG系统设计
    
  </span>
  

          <span class="md-nav__icon md-icon"></span>
        </label>
      
      <a href="4.7%20%E6%A3%80%E7%B4%A2%E7%AD%96%E7%95%A5%E4%B8%8ERAG%E7%B3%BB%E7%BB%9F%E8%AE%BE%E8%AE%A1.html" class="md-nav__link md-nav__link--active">
        
  
  <span class="md-ellipsis">
    4.7 检索策略与RAG系统设计
    
  </span>
  

      </a>
      
        
<nav class="md-nav md-nav--secondary" aria-label="目录">
  
  
    <label class="md-nav__title" for="__toc">
      <span class="md-nav__icon md-icon"></span>
      目录
    </label>
    <ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
      
        <li class="md-nav__item">
  <a href="#_1" class="md-nav__link">
    <span class="md-ellipsis">
      学习目标
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_2" class="md-nav__link">
    <span class="md-ellipsis">
      一、整体概述
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_3" class="md-nav__link">
    <span class="md-ellipsis">
      二、检索策略选择
    </span>
  </a>
  
    <nav class="md-nav" aria-label="二、检索策略选择">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1" class="md-nav__link">
    <span class="md-ellipsis">
      1 功能概述
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2" class="md-nav__link">
    <span class="md-ellipsis">
      2 代码示例
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3" class="md-nav__link">
    <span class="md-ellipsis">
      3 实现细节
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#4" class="md-nav__link">
    <span class="md-ellipsis">
      4 说明
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#rag" class="md-nav__link">
    <span class="md-ellipsis">
      三、RAG核心逻辑
    </span>
  </a>
  
    <nav class="md-nav" aria-label="三、RAG核心逻辑">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1_1" class="md-nav__link">
    <span class="md-ellipsis">
      1 功能概述
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2_1" class="md-nav__link">
    <span class="md-ellipsis">
      2 代码示例
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3_1" class="md-nav__link">
    <span class="md-ellipsis">
      3 实现细节
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_4" class="md-nav__link">
    <span class="md-ellipsis">
      四、完整流程的整合
    </span>
  </a>
  
    <nav class="md-nav" aria-label="四、完整流程的整合">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1_2" class="md-nav__link">
    <span class="md-ellipsis">
      1 从查询到回答的流程
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2_2" class="md-nav__link">
    <span class="md-ellipsis">
      2 代码示例（整合逻辑）
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3_2" class="md-nav__link">
    <span class="md-ellipsis">
      3 说明
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_5" class="md-nav__link">
    <span class="md-ellipsis">
      章节小结
    </span>
  </a>
  
</li>
      
    </ul>
  
</nav>
      
    </li>
  

    <li class="md-nav__item">
      <a href="4.8%20RAG%E7%B3%BB%E7%BB%9F%E8%BF%90%E8%A1%8C.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.8 RAG系统运行
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="4.9%20%E6%96%87%E6%A1%A3%E8%A7%A3%E6%9E%90%E5%B7%A5%E5%85%B7%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.9 文档解析工具(扩展资料)
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="4.10%20RAG%E4%B8%AD%E7%9A%84Query%E6%94%B9%E5%86%99%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.10 RAG中的Query改写(扩展资料)
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="%E6%95%B0%E6%8D%AE%E9%9B%86%E7%94%9F%E6%88%90%E4%B8%8E%E4%BC%98%E5%8C%96%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    数据集生成与优化(扩展资料)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_6">
        
          
          <label class="md-nav__link" for="__nav_6" id="__nav_6_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ⑤ RAG系统评估
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_6_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_6">
            <span class="md-nav__icon md-icon"></span>
            ⑤ RAG系统评估
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A4%20RAG%E7%B3%BB%E7%BB%9F%E8%AF%84%E4%BC%B0/5.1%20RAG%E7%B3%BB%E7%BB%9F%E8%AF%84%E4%BC%B0%E5%B7%A5%E5%85%B7.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    5.1 RAG系统评估工具
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A4%20RAG%E7%B3%BB%E7%BB%9F%E8%AF%84%E4%BC%B0/5.2%20%E5%AE%9E%E7%8E%B0EnterpriseKB-RAG%E7%B3%BB%E7%BB%9F%E7%9A%84%E8%AF%84%E4%BC%B0.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    5.2 实现企业知识库RAG系统的评估
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A4%20RAG%E7%B3%BB%E7%BB%9F%E8%AF%84%E4%BC%B0/5.3%20RAG%E8%AF%84%E4%BC%B0%E6%95%B0%E6%8D%AE%E9%9B%86%E7%9A%84%E6%9E%84%E9%80%A0%28%E6%89%A9%E5%B1%95%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    5.3 RAG评估数据集的构造(扩展)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_7">
        
          
          <label class="md-nav__link" for="__nav_7" id="__nav_7_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ⑥ 融合Mysql的RAG系统
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_7_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_7">
            <span class="md-nav__icon md-icon"></span>
            ⑥ 融合Mysql的RAG系统
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.1%20%E8%9E%8D%E5%90%88FQA%E5%92%8C%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.1 融合FQA和知识库查询
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.2%20%E8%9E%8D%E5%90%88FAQ%20%E5%92%8C%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2%EF%BC%88%E4%BC%98%E5%8C%96%E7%89%88%EF%BC%89.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.2 融合FAQ 和知识库查询（优化版）
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.3%20%E5%9F%BA%E4%BA%8E%20FastAPI%20%E7%9A%84%20FAQ%20%E4%B8%8E%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2%E6%8E%A5%E5%8F%A3.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.3 基于 FastAPI 的 FAQ 与知识库查询接口
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.4%20FastAPI%E5%92%8CFlask%E7%9A%84%E5%8C%BA%E5%88%AB%E5%92%8C%E8%81%94%E7%B3%BB%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.4 FastAPI和Flask的区别和联系(扩展资料)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_8">
        
          
          <label class="md-nav__link" for="__nav_8" id="__nav_8_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ⑦ 企业级生产环境部署
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_8_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_8">
            <span class="md-nav__icon md-icon"></span>
            ⑦ 企业级生产环境部署
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A6%20%E4%BC%81%E4%B8%9A%E7%BA%A7%E7%94%9F%E4%BA%A7%E7%8E%AF%E5%A2%83%E9%83%A8%E7%BD%B2/7.1%20Docker%E7%9A%84%E5%8E%9F%E7%90%86%E5%92%8C%E5%9F%BA%E6%9C%AC%E4%BD%BF%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    7.1 Docker的原理和基本使用
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A6%20%E4%BC%81%E4%B8%9A%E7%BA%A7%E7%94%9F%E4%BA%A7%E7%8E%AF%E5%A2%83%E9%83%A8%E7%BD%B2/7.2%20%E6%9C%8D%E5%8A%A1%E9%83%A8%E7%BD%B2.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    7.2 服务部署
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

  </ul>
</nav>
                  </div>
                </div>
              </div>
            
            
              <div class="md-sidebar md-sidebar--secondary" data-md-component="sidebar" data-md-type="toc">
                <div class="md-sidebar__scrollwrap">
                  <div class="md-sidebar__inner">
                    

<nav class="md-nav md-nav--secondary" aria-label="目录">
  
  
    <label class="md-nav__title" for="__toc">
      <span class="md-nav__icon md-icon"></span>
      目录
    </label>
    <ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
      
        <li class="md-nav__item">
  <a href="#_1" class="md-nav__link">
    <span class="md-ellipsis">
      学习目标
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_2" class="md-nav__link">
    <span class="md-ellipsis">
      一、整体概述
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_3" class="md-nav__link">
    <span class="md-ellipsis">
      二、检索策略选择
    </span>
  </a>
  
    <nav class="md-nav" aria-label="二、检索策略选择">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1" class="md-nav__link">
    <span class="md-ellipsis">
      1 功能概述
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2" class="md-nav__link">
    <span class="md-ellipsis">
      2 代码示例
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3" class="md-nav__link">
    <span class="md-ellipsis">
      3 实现细节
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#4" class="md-nav__link">
    <span class="md-ellipsis">
      4 说明
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#rag" class="md-nav__link">
    <span class="md-ellipsis">
      三、RAG核心逻辑
    </span>
  </a>
  
    <nav class="md-nav" aria-label="三、RAG核心逻辑">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1_1" class="md-nav__link">
    <span class="md-ellipsis">
      1 功能概述
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2_1" class="md-nav__link">
    <span class="md-ellipsis">
      2 代码示例
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3_1" class="md-nav__link">
    <span class="md-ellipsis">
      3 实现细节
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_4" class="md-nav__link">
    <span class="md-ellipsis">
      四、完整流程的整合
    </span>
  </a>
  
    <nav class="md-nav" aria-label="四、完整流程的整合">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1_2" class="md-nav__link">
    <span class="md-ellipsis">
      1 从查询到回答的流程
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2_2" class="md-nav__link">
    <span class="md-ellipsis">
      2 代码示例（整合逻辑）
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3_2" class="md-nav__link">
    <span class="md-ellipsis">
      3 说明
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_5" class="md-nav__link">
    <span class="md-ellipsis">
      章节小结
    </span>
  </a>
  
</li>
      
    </ul>
  
</nav>
                  </div>
                </div>
              </div>
            
          
            <div class="md-content" data-md-component="content">
              <article class="md-content__inner md-typeset">
                
                  
  <h1>4.7 检索策略与RAG系统设计</h1>

<h2 id="_1">学习目标<a class="headerlink" href="#_1" title="Permanent link">¶</a></h2>
<ul>
<li>掌握如何根据查询特点动态选择最优策略。</li>
<li>理解RAG系统如何整合查询分类、检索和生成阶段。</li>
<li>能够应用策略选择和RAG逻辑，构建一个从输入处理到答案生成的高效系统。</li>
</ul>
<hr>
<h2 id="_2">一、整体概述<a class="headerlink" href="#_2" title="Permanent link">¶</a></h2>
<p><code>strategy_selector.py</code>和<code>rag_system.py</code>是企业知识库RAG系统中<code>core</code>模块的核心组件，分别负责检索策略的选择和RAG系统的整体逻辑整合。<code>strategy_selector.py</code>通过大语言模型动态选择适合用户查询的检索策略，而<code>rag_system.py</code>则将前几章介绍的模块（如向量存储、Prompt管理和查询分类）整合起来，完成从查询输入到答案生成的完整流程。这两个模块共同确保系统能够高效、准确地响应用户的各种查询。</p>
<h2 id="_3">二、检索策略选择<a class="headerlink" href="#_3" title="Permanent link">¶</a></h2>
<h3 id="1">1 功能概述<a class="headerlink" href="#1" title="Permanent link">¶</a></h3>
<p><code>strategy_selector.py</code>定义了<code>StrategySelector</code>类，通过调用大语言模型根据用户查询选择最合适的检索策略。支持的策略包括直接检索、HyDE（假设问题检索）、子查询检索和回溯问题检索，旨在优化检索阶段的输入处理。</p>
<h3 id="2">2 代码示例<a class="headerlink" href="#2" title="Permanent link">¶</a></h3>
<div class="highlight"><pre><span></span><code><span class="c1"># core/strategy_selector.py 源码</span>
<span class="c1"># 导入 LangChain 提示模板</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">langchain_core.prompts</span><span class="w"> </span><span class="kn">import</span> <span class="n">PromptTemplate</span>
<span class="c1"># 导入日志和配置</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">base.config</span><span class="w"> </span><span class="kn">import</span> <span class="n">config</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">base.logger</span><span class="w"> </span><span class="kn">import</span> <span class="n">logger</span>
<span class="c1"># 导入 OpenAI</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">openai</span><span class="w"> </span><span class="kn">import</span> <span class="n">OpenAI</span>


<span class="k">class</span><span class="w"> </span><span class="nc">StrategySelector</span><span class="p">:</span>
    <span class="k">def</span><span class="w"> </span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
        <span class="c1"># 初始化 OpenAI 客户端</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">client</span> <span class="o">=</span> <span class="n">OpenAI</span><span class="p">(</span><span class="n">api_key</span><span class="o">=</span><span class="n">config</span><span class="o">.</span><span class="n">DASHSCOPE_API_KEY</span><span class="p">,</span>
                             <span class="n">base_url</span><span class="o">=</span><span class="n">config</span><span class="o">.</span><span class="n">DASHSCOPE_BASE_URL</span><span class="p">)</span>
        <span class="c1"># 获取策略选择提示模板</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">strategy_prompt_template</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_get_strategy_prompt</span><span class="p">()</span>

    <span class="k">def</span><span class="w"> </span><span class="nf">call_dashscope</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">prompt</span><span class="p">):</span>
        <span class="c1"># 调用 DashScope API</span>
        <span class="k">try</span><span class="p">:</span>
            <span class="c1"># 创建聊天完成请求</span>
            <span class="n">completion</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">client</span><span class="o">.</span><span class="n">chat</span><span class="o">.</span><span class="n">completions</span><span class="o">.</span><span class="n">create</span><span class="p">(</span>
                <span class="n">model</span><span class="o">=</span><span class="n">Config</span><span class="p">()</span><span class="o">.</span><span class="n">LLM_MODEL</span><span class="p">,</span>
                <span class="n">messages</span><span class="o">=</span><span class="p">[</span>
                    <span class="p">{</span><span class="s2">"role"</span><span class="p">:</span> <span class="s2">"system"</span><span class="p">,</span> <span class="s2">"content"</span><span class="p">:</span> <span class="s2">"你是一个有用的助手，能够根据用户输入的Prompt严格执行并返回可靠的结果"</span><span class="p">},</span>
                    <span class="p">{</span><span class="s2">"role"</span><span class="p">:</span> <span class="s2">"user"</span><span class="p">,</span> <span class="s2">"content"</span><span class="p">:</span> <span class="n">prompt</span><span class="p">},</span>
                <span class="p">],</span>
                <span class="n">temperature</span><span class="o">=</span><span class="mf">0.1</span>
            <span class="p">)</span>
            <span class="c1"># 返回完成结果</span>
            <span class="k">return</span> <span class="n">completion</span><span class="o">.</span><span class="n">choices</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">message</span><span class="o">.</span><span class="n">content</span> <span class="k">if</span> <span class="n">completion</span><span class="o">.</span><span class="n">choices</span> <span class="k">else</span> <span class="s2">"直接检索"</span>
        <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
            <span class="c1"># 记录 API 调用失败</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="sa">f</span><span class="s2">"DashScope API 调用失败: </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
            <span class="c1"># 默认返回直接检索</span>
            <span class="k">return</span> <span class="s2">"直接检索"</span>


    <span class="k">def</span><span class="w"> </span><span class="nf">_get_strategy_prompt</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
        <span class="c1">#   定义私有方法，获取策略选择 Prompt 模板</span>
        <span class="k">return</span> <span class="n">PromptTemplate</span><span class="p">(</span>
            <span class="n">template</span><span class="o">=</span><span class="s2">"""</span>
<span class="s2">            你是一个智能助手，负责分析用户查询 </span><span class="si">{query}</span><span class="s2">，并从以下四种检索增强策略中选择一个最适合的策略，直接返回策略名称，不需要解释过程。</span>

<span class="s2">            以下是几种检索增强策略及其适用场景：</span>

<span class="s2">            1.  **直接检索：**</span>
<span class="s2">                * 描述：对用户查询直接进行检索，不进行任何增强处理。</span>
<span class="s2">                * 适用场景：适用于查询意图明确，需要从知识库中检索**特定信息**的问题，例如：</span>
<span class="s2">                    * 示例：</span>
<span class="s2">                        * 查询：AI 知识域报销标准是多少？</span>
<span class="s2">                        * 策略：直接检索</span>
<span class="s2">                    * 查询：JAVA的知识文档目录是什么？</span>
<span class="s2">                        * 策略：直接检索</span>
<span class="s2">            2.  **假设问题检索（HyDE）：**</span>
<span class="s2">                * 描述：使用 LLM 生成一个假设的答案，然后基于假设答案进行检索。</span>
<span class="s2">                * 适用场景：适用于查询较为抽象，直接检索效果不佳的问题，例如：</span>
<span class="s2">                    * 示例：</span>
<span class="s2">                        * 查询：人工智能在企业知识管理场景的应用有哪些？</span>
<span class="s2">                        * 策略：假设问题检索</span>
<span class="s2">            3.  **子查询检索：**</span>
<span class="s2">                * 描述：将复杂的用户查询拆分为多个简单的子查询，分别检索并合并结果。</span>
<span class="s2">                * 适用场景：适用于查询涉及多个实体或方面，需要分别检索不同信息的问题，例如：</span>
<span class="s2">                    * 示例：</span>
<span class="s2">                        * 查询：比较 Milvus 和 Zilliz Cloud 的优缺点。</span>
<span class="s2">                        * 策略：子查询检索</span>
<span class="s2">            4.  **回溯问题检索：**</span>
<span class="s2">                * 描述：将复杂的用户查询转化为更基础、更易于检索的问题，然后进行检索。</span>
<span class="s2">                * 适用场景：适用于查询较为复杂，需要简化后才能有效检索的问题，例如：</span>
<span class="s2">                    * 示例：</span>
<span class="s2">                        * 查询：我有一个包含 100 亿条记录的数据集，想把它存储到 Milvus 中进行查询。可以吗？</span>
<span class="s2">                        * 策略：回溯问题检索</span>

<span class="s2">            根据用户查询 </span><span class="si">{query}</span><span class="s2">，直接返回最适合的策略名称，例如 "直接检索"。不要输出任何分析过程或其他内容。</span>
<span class="s2">            """</span>
            <span class="p">,</span>
            <span class="n">input_variables</span><span class="o">=</span><span class="p">[</span><span class="s2">"query"</span><span class="p">],</span>
        <span class="p">)</span>

    <span class="c1">#   定义方法，选择检索策略</span>
    <span class="k">def</span><span class="w"> </span><span class="nf">select_strategy</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query</span><span class="p">):</span>
        <span class="c1">#   调用 LLM 获取检索策略</span>
        <span class="n">strategy</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">call_dashscope</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">strategy_prompt_template</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">query</span><span class="o">=</span><span class="n">query</span><span class="p">))</span><span class="o">.</span><span class="n">strip</span><span class="p">()</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"为查询 '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">' 选择的检索策略：</span><span class="si">{</span><span class="n">strategy</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
        <span class="k">return</span> <span class="n">strategy</span>

<span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
    <span class="n">ss</span> <span class="o">=</span> <span class="n">StrategySelector</span><span class="p">()</span>
    <span class="n">ss</span><span class="o">.</span><span class="n">select_strategy</span><span class="p">(</span><span class="s1">'你好吗'</span><span class="p">)</span>
</code></pre></div>
<h3 id="3">3 实现细节<a class="headerlink" href="#3" title="Permanent link">¶</a></h3>
<ol>
<li><strong><code>__init__</code></strong>：</li>
<li><strong>作用</strong>：初始化DashScope客户端和策略选择Prompt。</li>
<li><strong>逻辑</strong>：连接大语言模型API，准备Prompt模板。</li>
<li><strong><code>call_dashscope</code></strong>：</li>
<li><strong>作用</strong>：封装DashScope API调用，处理异常并返回模型输出。</li>
<li><strong>逻辑</strong>：确保API调用的鲁棒性，记录错误日志。</li>
<li><strong><code>_get_strategy_prompt</code></strong>：</li>
<li><strong>作用</strong>：定义用于策略选择的Prompt模板。</li>
<li><strong>设计逻辑</strong>：简洁描述四种策略及其适用场景，要求模型直接返回策略名称。</li>
<li><strong><code>select_strategy</code></strong>：</li>
<li><strong>作用</strong>：根据查询调用模型选择策略并返回。</li>
<li><strong>逻辑</strong>：记录选择的策略，便于调试。</li>
</ol>
<h3 id="4">4 说明<a class="headerlink" href="#4" title="Permanent link">¶</a></h3>
<ul>
<li><strong>动态选择</strong>：利用大语言模型的语义理解能力，灵活适应不同查询。</li>
<li><strong>策略多样性</strong>：支持四种策略，覆盖常见查询场景。</li>
<li><strong>效率</strong>：直接返回策略名称，避免冗余输出。</li>
</ul>
<hr>
<h2 id="rag">三、RAG核心逻辑<a class="headerlink" href="#rag" title="Permanent link">¶</a></h2>
<h3 id="1_1">1 功能概述<a class="headerlink" href="#1_1" title="Permanent link">¶</a></h3>
<p><code>rag_system.py</code>定义了<code>RAGSystem</code>类，整合系统的各个模块，完成从查询输入到答案生成的完整流程。它通过查询分类选择处理路径，利用检索策略优化文档检索，并结合上下文生成最终答案。</p>
<h3 id="2_1">2 代码示例<a class="headerlink" href="#2_1" title="Permanent link">¶</a></h3>
<div class="highlight"><pre><span></span><code><span class="c1"># core/rag_system.py 源码</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">rag_qa.core.prompts</span><span class="w"> </span><span class="kn">import</span> <span class="n">RAGPrompts</span>
<span class="c1">#   导入 time 模块，用于计算时间</span>
<span class="kn">import</span><span class="w"> </span><span class="nn">time</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">base.config</span><span class="w"> </span><span class="kn">import</span> <span class="n">config</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">base.logger</span><span class="w"> </span><span class="kn">import</span> <span class="n">logger</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">rag_qa.core.query_classifier</span><span class="w"> </span><span class="kn">import</span> <span class="n">QueryClassifier</span>  <span class="c1"># 导入查询分类器</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">rag_qa.core.strategy_selector</span><span class="w"> </span><span class="kn">import</span> <span class="n">StrategySelector</span>  <span class="c1"># 导入策略选择器</span>


<span class="c1">#   定义 RAGSystem 类，封装 RAG 系统的核心逻辑</span>
<span class="k">class</span><span class="w"> </span><span class="nc">RAGSystem</span><span class="p">:</span>
    <span class="c1">#   初始化方法，设置 RAG 系统的基本参数</span>
    <span class="k">def</span><span class="w"> </span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">vector_store</span><span class="p">,</span> <span class="n">llm</span><span class="p">):</span>
        <span class="c1">#   设置向量数据库对象</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">vector_store</span> <span class="o">=</span> <span class="n">vector_store</span>
        <span class="c1">#   设置大语言模型调用函数</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">llm</span> <span class="o">=</span> <span class="n">llm</span>
        <span class="c1">#   获取 RAG 提示模板</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">rag_prompt</span> <span class="o">=</span> <span class="n">RAGPrompts</span><span class="o">.</span><span class="n">rag_prompt</span><span class="p">()</span>
        <span class="c1">#   初始化查询分类器</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">query_classifier</span> <span class="o">=</span><span class="n">QueryClassifier</span><span class="p">(</span><span class="n">model_path</span><span class="o">=</span><span class="sa">f</span><span class="s1">'</span><span class="si">{</span><span class="n">config</span><span class="o">.</span><span class="n">MODELS_DIR</span><span class="si">}</span><span class="s1">/bert_query_classifier'</span><span class="p">)</span>
        <span class="c1">#   初始化策略选择器</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">strategy_selector</span> <span class="o">=</span> <span class="n">StrategySelector</span><span class="p">()</span>

    <span class="c1">#   定义私有方法，使用假设文档进行检索（HyDE）</span>
    <span class="k">def</span><span class="w"> </span><span class="nf">_retrieve_with_hyde</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query</span><span class="p">):</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"使用 HyDE 策略进行检索 (查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span><span class="p">)</span>
        <span class="c1">#   获取假设问题生成的 Prompt 模板</span>
        <span class="n">hyde_prompt_template</span> <span class="o">=</span> <span class="n">RAGPrompts</span><span class="o">.</span><span class="n">hyde_prompt</span><span class="p">()</span> <span class="c1"># 使用 template 后缀区分</span>
        <span class="c1">#   调用大语言模型生成假设答案</span>
        <span class="k">try</span><span class="p">:</span>
            <span class="n">hypo_answer</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">llm</span><span class="p">(</span><span class="n">hyde_prompt_template</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">query</span><span class="o">=</span><span class="n">query</span><span class="p">))</span><span class="o">.</span><span class="n">strip</span><span class="p">()</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"HyDE 生成的假设答案: '</span><span class="si">{</span><span class="n">hypo_answer</span><span class="si">}</span><span class="s2">'"</span><span class="p">)</span>
            <span class="c1">#   使用假设答案进行检索，并返回检索结果</span>
            <span class="c1">#   注意：HyDE 通常只用于生成检索向量，不一定需要 rerank 这一步，但这里复用了</span>
            <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">vector_store</span><span class="o">.</span><span class="n">hybrid_search_with_rerank</span><span class="p">(</span>
                <span class="n">hypo_answer</span><span class="p">,</span> <span class="n">k</span><span class="o">=</span><span class="n">conf</span><span class="o">.</span><span class="n">RETRIEVAL_K</span> <span class="c1"># 使用 K 而非 M</span>
            <span class="p">)</span>
        <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="sa">f</span><span class="s2">"HyDE 策略执行失败: </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
            <span class="k">return</span> <span class="p">[]</span>


    <span class="c1">#   定义私有方法，使用子查询进行检索</span>
    <span class="k">def</span><span class="w"> </span><span class="nf">_retrieve_with_subqueries</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query</span><span class="p">):</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"使用子查询策略进行检索 (查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span><span class="p">)</span>
        <span class="c1">#   获取子查询生成的 Prompt 模板</span>
        <span class="n">subquery_prompt_template</span> <span class="o">=</span> <span class="n">RAGPrompts</span><span class="o">.</span><span class="n">subquery_prompt</span><span class="p">()</span> <span class="c1"># 使用 template 后缀区分</span>
        <span class="k">try</span><span class="p">:</span>
            <span class="c1">#   调用大语言模型生成子查询列表</span>
            <span class="n">subqueries_text</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">llm</span><span class="p">(</span><span class="n">subquery_prompt_template</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">query</span><span class="o">=</span><span class="n">query</span><span class="p">))</span><span class="o">.</span><span class="n">strip</span><span class="p">()</span>
            <span class="n">subqueries</span> <span class="o">=</span> <span class="p">[</span><span class="n">q</span><span class="o">.</span><span class="n">strip</span><span class="p">()</span> <span class="k">for</span> <span class="n">q</span> <span class="ow">in</span> <span class="n">subqueries_text</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="s2">"</span><span class="se">\n</span><span class="s2">"</span><span class="p">)</span> <span class="k">if</span> <span class="n">q</span><span class="o">.</span><span class="n">strip</span><span class="p">()]</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"生成的子查询: </span><span class="si">{</span><span class="n">subqueries</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
            <span class="k">if</span> <span class="ow">not</span> <span class="n">subqueries</span><span class="p">:</span>
                 <span class="n">logger</span><span class="o">.</span><span class="n">warning</span><span class="p">(</span><span class="s2">"未能生成有效的子查询"</span><span class="p">)</span>
                 <span class="k">return</span> <span class="p">[]</span>

            <span class="c1">#   初始化空列表，用于存储所有子查询的检索结果</span>
            <span class="n">all_docs</span> <span class="o">=</span> <span class="p">[]</span>
            <span class="c1">#   遍历每个子查询</span>
            <span class="k">for</span> <span class="n">sub_q</span> <span class="ow">in</span> <span class="n">subqueries</span><span class="p">:</span>
                <span class="c1">#   使用子查询进行检索，并将结果添加到列表中</span>
                <span class="c1">#   这里对每个子查询都执行了 hybrid search + rerank，开销可能较大</span>
                <span class="n">docs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">vector_store</span><span class="o">.</span><span class="n">hybrid_search_with_rerank</span><span class="p">(</span>
                    <span class="n">sub_q</span><span class="p">,</span> <span class="n">k</span><span class="o">=</span><span class="n">conf</span><span class="o">.</span><span class="n">RETRIEVAL_K</span> <span class="c1"># 使用 K</span>
                <span class="p">)</span>
                <span class="n">all_docs</span><span class="o">.</span><span class="n">extend</span><span class="p">(</span><span class="n">docs</span><span class="p">)</span>
                <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"子查询 '</span><span class="si">{</span><span class="n">sub_q</span><span class="si">}</span><span class="s2">' 检索到 </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">docs</span><span class="p">)</span><span class="si">}</span><span class="s2"> 个文档"</span><span class="p">)</span>

            <span class="c1">#   对所有检索结果进行去重 (基于对象内存地址，如果 Document 内容相同但对象不同则无法去重)</span>
            <span class="c1">#   更可靠的去重方式是基于文档内容或 ID</span>
            <span class="n">unique_docs_dict</span> <span class="o">=</span> <span class="p">{</span><span class="n">doc</span><span class="o">.</span><span class="n">page_content</span><span class="p">:</span> <span class="n">doc</span> <span class="k">for</span> <span class="n">doc</span> <span class="ow">in</span> <span class="n">all_docs</span><span class="p">}</span> <span class="c1"># 基于内容去重</span>
            <span class="n">unique_docs</span> <span class="o">=</span> <span class="nb">list</span><span class="p">(</span><span class="n">unique_docs_dict</span><span class="o">.</span><span class="n">values</span><span class="p">())</span>

            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"所有子查询共检索到 </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">all_docs</span><span class="p">)</span><span class="si">}</span><span class="s2"> 个文档, 去重后剩 </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">unique_docs</span><span class="p">)</span><span class="si">}</span><span class="s2"> 个"</span><span class="p">)</span>
            <span class="c1">#   返回去重后的文档，限制数量 (是否需要在此处限制? retrieve_and_merge 末尾会限制)</span>
            <span class="c1"># return unique_docs[: Config.CANDIDATE_M]</span>
            <span class="k">return</span> <span class="n">unique_docs</span> <span class="c1"># 返回所有唯一文档，让 retrieve_and_merge 处理数量</span>

        <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="sa">f</span><span class="s2">"子查询策略执行失败: </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
            <span class="k">return</span> <span class="p">[]</span>

    <span class="c1">#   定义私有方法，使用回溯问题进行检索</span>
    <span class="k">def</span><span class="w"> </span><span class="nf">_retrieve_with_backtracking</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query</span><span class="p">):</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"使用回溯问题策略进行检索 (查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span><span class="p">)</span>
        <span class="c1">#   获取回溯问题生成的 Prompt 模板</span>
        <span class="n">backtrack_prompt_template</span> <span class="o">=</span> <span class="n">RAGPrompts</span><span class="o">.</span><span class="n">backtracking_prompt</span><span class="p">()</span> <span class="c1"># 使用 template 后缀区分</span>
        <span class="k">try</span><span class="p">:</span>
            <span class="c1">#   调用大语言模型生成回溯问题</span>
            <span class="n">simplified_query</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">llm</span><span class="p">(</span><span class="n">backtrack_prompt_template</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">query</span><span class="o">=</span><span class="n">query</span><span class="p">))</span><span class="o">.</span><span class="n">strip</span><span class="p">()</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"生成的回溯问题: '</span><span class="si">{</span><span class="n">simplified_query</span><span class="si">}</span><span class="s2">'"</span><span class="p">)</span>
            <span class="c1">#   使用回溯问题进行检索，并返回检索结果</span>
            <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">vector_store</span><span class="o">.</span><span class="n">hybrid_search_with_rerank</span><span class="p">(</span>
                <span class="n">simplified_query</span><span class="p">,</span> <span class="n">k</span><span class="o">=</span><span class="n">conf</span><span class="o">.</span><span class="n">RETRIEVAL_K</span> <span class="c1"># 使用 K</span>
            <span class="p">)</span>
        <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="sa">f</span><span class="s2">"回溯问题策略执行失败: </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
            <span class="k">return</span> <span class="p">[]</span>

    <span class="c1">#   定义方法，检索并合并相关文档</span>
    <span class="k">def</span><span class="w"> </span><span class="nf">retrieve_and_merge</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query</span><span class="p">,</span> <span class="n">source_filter</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">strategy</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>  <span class="c1">#   新增 strategy 参数</span>
        <span class="c1">#   如果未指定检索策略，则使用策略选择器选择</span>
        <span class="k">if</span> <span class="ow">not</span> <span class="n">strategy</span><span class="p">:</span>
            <span class="n">strategy</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">strategy_selector</span><span class="o">.</span><span class="n">select_strategy</span><span class="p">(</span><span class="n">query</span><span class="p">)</span>

        <span class="c1">#   根据检索策略选择不同的检索方式</span>
        <span class="n">ranked_sub_chunks</span> <span class="o">=</span> <span class="p">[]</span> <span class="c1"># 初始化</span>
        <span class="k">if</span> <span class="n">strategy</span> <span class="o">==</span> <span class="s2">"回溯问题检索"</span><span class="p">:</span>
            <span class="n">ranked_sub_chunks</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_retrieve_with_backtracking</span><span class="p">(</span><span class="n">query</span><span class="p">)</span>
        <span class="k">elif</span> <span class="n">strategy</span> <span class="o">==</span> <span class="s2">"子查询检索"</span><span class="p">:</span>
            <span class="n">ranked_sub_chunks</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_retrieve_with_subqueries</span><span class="p">(</span><span class="n">query</span><span class="p">)</span> <span class="c1"># 返回的是唯一文档列表</span>
             <span class="c1"># 注意：子查询返回的是已 rerank 过的父文档或子块列表，后续合并逻辑可能需要调整</span>
             <span class="c1"># 当前实现中，子查询返回的是初步检索（可能已rerank）的块，再进行合并</span>
        <span class="k">elif</span> <span class="n">strategy</span> <span class="o">==</span> <span class="s2">"假设问题检索"</span><span class="p">:</span>
            <span class="n">ranked_sub_chunks</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_retrieve_with_hyde</span><span class="p">(</span><span class="n">query</span><span class="p">)</span>
        <span class="k">else</span><span class="p">:</span>  <span class="c1">#   默认或“直接检索”</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"使用直接检索策略 (查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span><span class="p">)</span>
            <span class="n">ranked_sub_chunks</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">vector_store</span><span class="o">.</span><span class="n">hybrid_search_with_rerank</span><span class="p">(</span>
                <span class="n">query</span><span class="p">,</span> <span class="n">k</span><span class="o">=</span><span class="n">conf</span><span class="o">.</span><span class="n">RETRIEVAL_K</span><span class="p">,</span> <span class="n">source_filter</span><span class="o">=</span><span class="n">source_filter</span>
            <span class="p">)</span> <span class="c1"># 注意 hybrid_search_with_rerank 返回的是 rerank 后的父文档</span>

        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"策略 '</span><span class="si">{</span><span class="n">strategy</span><span class="si">}</span><span class="s2">' 检索到 </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">ranked_sub_chunks</span><span class="p">)</span><span class="si">}</span><span class="s2"> 个候选文档 (可能已是父文档)"</span><span class="p">)</span>
        <span class="n">final_context_docs</span> <span class="o">=</span> <span class="n">ranked_sub_chunks</span><span class="p">[:</span><span class="n">conf</span><span class="o">.</span><span class="n">CANDIDATE_M</span><span class="p">]</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"最终选取 </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">final_context_docs</span><span class="p">)</span><span class="si">}</span><span class="s2"> 个文档作为上下文"</span><span class="p">)</span>
        <span class="k">return</span> <span class="n">final_context_docs</span>

    <span class="c1">#   定义方法，生成答案</span>
    <span class="k">def</span><span class="w"> </span><span class="nf">generate_answer</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query</span><span class="p">,</span> <span class="n">source_filter</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
        <span class="c1">#   记录查询开始时间</span>
        <span class="n">start_time</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"开始处理查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">', 知识域过滤: </span><span class="si">{</span><span class="n">source_filter</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>

        <span class="c1">#   判断查询类型</span>
        <span class="n">query_category</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">query_classifier</span><span class="o">.</span><span class="n">predict_category</span><span class="p">(</span><span class="n">query</span><span class="p">)</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"查询分类结果：</span><span class="si">{</span><span class="n">query_category</span><span class="si">}</span><span class="s2"> (查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span><span class="p">)</span>

        <span class="c1">#   如果查询属于“通用知识”类别，则直接使用 LLM 回答</span>
        <span class="k">if</span> <span class="n">query_category</span> <span class="o">==</span> <span class="s2">"通用知识"</span><span class="p">:</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">"查询为通用知识，直接调用 LLM"</span><span class="p">)</span>
             <span class="n">prompt_input</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">rag_prompt</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
                <span class="n">context</span><span class="o">=</span><span class="s2">""</span><span class="p">,</span> <span class="n">question</span><span class="o">=</span><span class="n">query</span><span class="p">,</span> <span class="n">history</span><span class="o">=</span><span class="s2">""</span><span class="p">,</span> <span class="n">phone</span><span class="o">=</span><span class="n">config</span><span class="o">.</span><span class="n">CUSTOMER_SERVICE_PHONE</span>
            <span class="p">)</span>  <span class="c1">#   不使用上下文</span>
            <span class="k">try</span><span class="p">:</span>
                <span class="n">answer</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">llm</span><span class="p">(</span><span class="n">prompt_input</span><span class="p">)</span>
            <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
                <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="sa">f</span><span class="s2">"直接调用 LLM 失败: </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
                <span class="n">answer</span> <span class="o">=</span> <span class="sa">f</span><span class="s2">"抱歉，处理您的通用知识问题时出错。请联系人工客服：</span><span class="si">{</span><span class="n">conf</span><span class="o">.</span><span class="n">CUSTOMER_SERVICE_PHONE</span><span class="si">}</span><span class="s2">"</span>
            <span class="n">processing_time</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span> <span class="o">-</span> <span class="n">start_time</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span>
                <span class="sa">f</span><span class="s2">"通用知识查询处理完成 (耗时: </span><span class="si">{</span><span class="n">processing_time</span><span class="si">:</span><span class="s2">.2f</span><span class="si">}</span><span class="s2">s, 查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span>
            <span class="p">)</span>
            <span class="k">return</span> <span class="n">answer</span>

        <span class="c1">#   否则，进行 RAG 检索并生成答案</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">"查询为企业知识，执行 RAG 流程"</span><span class="p">)</span>
        <span class="c1">#   选择检索策略</span>
        <span class="n">strategy</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">strategy_selector</span><span class="o">.</span><span class="n">select_strategy</span><span class="p">(</span><span class="n">query</span><span class="p">)</span>

        <span class="c1">#   检索相关文档</span>
        <span class="n">context_docs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">retrieve_and_merge</span><span class="p">(</span>
            <span class="n">query</span><span class="p">,</span> <span class="n">source_filter</span><span class="o">=</span><span class="n">source_filter</span><span class="p">,</span> <span class="n">strategy</span><span class="o">=</span><span class="n">strategy</span>
        <span class="p">)</span>  <span class="c1">#   传递 strategy</span>

        <span class="c1">#   准备上下文</span>
        <span class="k">if</span> <span class="n">context_docs</span><span class="p">:</span>
            <span class="n">context</span> <span class="o">=</span> <span class="s2">"</span><span class="se">\n\n</span><span class="s2">"</span><span class="o">.</span><span class="n">join</span><span class="p">([</span><span class="n">doc</span><span class="o">.</span><span class="n">page_content</span> <span class="k">for</span> <span class="n">doc</span> <span class="ow">in</span> <span class="n">context_docs</span><span class="p">])</span> <span class="c1"># 使用换行符分隔文档</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"构建上下文完成，包含 </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">context_docs</span><span class="p">)</span><span class="si">}</span><span class="s2"> 个文档块"</span><span class="p">)</span>
            <span class="c1"># logger.debug(f"上下文内容:\n{context[:500]}...") # Debug 日志可以打印部分上下文</span>
        <span class="k">else</span><span class="p">:</span>
            <span class="n">context</span> <span class="o">=</span> <span class="s2">""</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">"未检索到相关文档，上下文为空"</span><span class="p">)</span>

        <span class="c1">#   构造 Prompt，调用大语言模型生成答案</span>
        <span class="n">prompt_input</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">rag_prompt</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
            <span class="n">context</span><span class="o">=</span><span class="n">context</span><span class="p">,</span> <span class="n">question</span><span class="o">=</span><span class="n">query</span><span class="p">,</span> <span class="n">phone</span><span class="o">=</span><span class="n">conf</span><span class="o">.</span><span class="n">CUSTOMER_SERVICE_PHONE</span>
        <span class="p">)</span>
        <span class="c1"># logger.debug(f"最终生成的 Prompt:\n{prompt_input}") # Debug 日志</span>

        <span class="k">try</span><span class="p">:</span>
            <span class="n">answer</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">llm</span><span class="p">(</span><span class="n">prompt_input</span><span class="p">)</span>
        <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
            <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="sa">f</span><span class="s2">"调用 LLM 生成最终答案失败: </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
            <span class="n">answer</span> <span class="o">=</span> <span class="sa">f</span><span class="s2">"抱歉，处理您的企业知识问题时出错。请联系人工客服：</span><span class="si">{</span><span class="n">conf</span><span class="o">.</span><span class="n">CUSTOMER_SERVICE_PHONE</span><span class="si">}</span><span class="s2">"</span>


        <span class="c1">#   记录查询处理完成的日志</span>
        <span class="n">processing_time</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span> <span class="o">-</span> <span class="n">start_time</span>
        <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">"查询处理完成 (耗时: </span><span class="si">{</span><span class="n">processing_time</span><span class="si">:</span><span class="s2">.2f</span><span class="si">}</span><span class="s2">s, 查询: '</span><span class="si">{</span><span class="n">query</span><span class="si">}</span><span class="s2">')"</span><span class="p">)</span>
        <span class="k">return</span> <span class="n">answer</span>
</code></pre></div>
<h3 id="3_1">3 实现细节<a class="headerlink" href="#3_1" title="Permanent link">¶</a></h3>
<ol>
<li><strong>init</strong>： </li>
<li><strong>作用</strong>：初始化RAG系统，整合向量存储、大语言模型和其他核心组件。</li>
<li><strong>依赖</strong>：依赖VectorStore、RAGPrompts、QueryClassifier和StrategySelector。</li>
<li><strong>_retrieve_with_hyde</strong>： </li>
<li><strong>作用</strong>：生成假设答案并调用混合检索，适合抽象查询。</li>
<li><strong>逻辑</strong>：使用hyde_prompt生成假设答案，传递给hybrid_search_with_rerank。</li>
<li><strong>_retrieve_with_subqueries</strong>： </li>
<li><strong>作用</strong>：分解查询为子查询，分别检索并去重。</li>
<li><strong>逻辑</strong>：使用subquery_prompt分解查询，合并结果并限制数量。</li>
<li><strong>_retrieve_with_backtracking</strong>： </li>
<li><strong>作用</strong>：简化查询后检索，降低复杂度。</li>
<li><strong>逻辑</strong>：使用backtracking_prompt简化查询，调用混合检索。</li>
<li><strong>retrieve_and_merge</strong>： </li>
<li><strong>作用</strong>：根据策略选择执行检索，直接返回结果。</li>
<li><strong>优化</strong>：移除冗余的合并逻辑，直接使用hybrid_search_with_rerank的结果（去重的父文档）。</li>
<li><strong>generate_answer</strong>： </li>
<li><strong>作用</strong>：整合分类、检索和生成，输出最终答案。</li>
<li>流程： <ul>
<li>使用QueryClassifier判断查询类型。</li>
<li>“通用知识”直接生成答案，“企业知识”触发检索。</li>
<li>结合上下文调用rag_prompt生成回答。</li>
</ul>
</li>
</ol>
<hr>
<h2 id="_4">四、完整流程的整合<a class="headerlink" href="#_4" title="Permanent link">¶</a></h2>
<h3 id="1_2">1 从查询到回答的流程<a class="headerlink" href="#1_2" title="Permanent link">¶</a></h3>
<ol>
<li><strong>输入处理</strong>：</li>
<li><code>QueryClassifier</code>分类查询，决定是否需要检索。</li>
<li><strong>策略选择</strong>：</li>
<li><code>StrategySelector</code>根据查询选择最佳检索策略。</li>
<li><strong>文档检索</strong>：</li>
<li>根据策略调用<code>VectorStore</code>的混合检索，获取相关文档。</li>
<li><strong>答案生成</strong>：</li>
<li>使用<code>RAGPrompts</code>的模板，结合上下文调用大语言模型生成答案。</li>
<li><strong>输出</strong>：</li>
<li>返回最终答案，并记录日志。</li>
</ol>
<h3 id="2_2">2 代码示例（整合逻辑）<a class="headerlink" href="#2_2" title="Permanent link">¶</a></h3>
<div class="highlight"><pre><span></span><code><span class="c1"># 示例：完整查询处理</span>
<span class="n">query</span> <span class="o">=</span> <span class="s2">"AI知识域报销标准是多少？"</span>
<span class="n">rag_system</span> <span class="o">=</span> <span class="n">RAGSystem</span><span class="p">(</span><span class="n">vector_store</span><span class="p">,</span> <span class="n">llm</span><span class="p">)</span>
<span class="n">answer</span> <span class="o">=</span> <span class="n">rag_system</span><span class="o">.</span><span class="n">generate_answer</span><span class="p">(</span><span class="n">query</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="n">answer</span><span class="p">)</span>
</code></pre></div>
<h3 id="3_2">3 说明<a class="headerlink" href="#3_2" title="Permanent link">¶</a></h3>
<ul>
<li><strong>端到端设计</strong>：从输入到输出无缝衔接。</li>
<li><strong>智能优化</strong>：通过分类和策略选择减少不必要计算。</li>
</ul>
<hr>
<h2 id="_5">章节小结<a class="headerlink" href="#_5" title="Permanent link">¶</a></h2>
<p>本章节深入探讨了<code>strategy_selector.py</code>和<code>rag_system.py</code>：</p>
<ul>
<li>
<p><strong>strategy_selector.py</strong>：通过大语言模型动态选择检索策略（直接检索、HyDE、子查询、回溯），优化检索输入。</p>
</li>
<li>
<p><strong><code>rag_system.py</code></strong>：整合查询分类、检索策略、向量存储和Prompt管理，完成从查询到回答的完整RAG流程。</p>
</li>
<li><strong>协同作用</strong>：两者结合实现了灵活、高效的查询处理，确保系统能够准确响应用户需求。</li>
</ul>
<p>学习者通过本章节掌握了RAG系统的核心工作机制，为后续的系统运行和扩展奠定了基础。</p>


              </article>
            </div>
          
          
<script>var target=document.getElementById(location.hash.slice(1));target&&target.name&&(target.checked=target.name.startsWith("__tabbed_"))</script>
        </div>
        
      </main>
      
        <footer class="md-footer">
  
  <div class="md-footer-meta md-typeset">
    <div class="md-footer-meta__inner md-grid">
      <div class="md-copyright">
  
    <div class="md-copyright__highlight">
      © EnterpriseKB-RAG 2026. Internal knowledge base demo site.
    </div>
  
  
    Made with
    <a href="https://squidfunk.github.io/mkdocs-material/" target="_blank" rel="noopener">
      Material for MkDocs
    </a>
  
</div>
      
    </div>
  </div>
</footer>
      
    </div>
    <div class="md-dialog" data-md-component="dialog">
      <div class="md-dialog__inner md-typeset"></div>
    </div>
    
    
      <script id="__config" type="application/json">{"base": "..", "features": [], "search": "../assets/javascripts/workers/search.973d3a69.min.js", "tags": null, "translations": {"clipboard.copied": "\u5df2\u590d\u5236", "clipboard.copy": "\u590d\u5236", "search.result.more.one": "\u5728\u8be5\u9875\u4e0a\u8fd8\u6709 1 \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.more.other": "\u5728\u8be5\u9875\u4e0a\u8fd8\u6709 # \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.none": "\u6ca1\u6709\u627e\u5230\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.one": "\u627e\u5230 1 \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.other": "# \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.placeholder": "\u952e\u5165\u4ee5\u5f00\u59cb\u641c\u7d22", "search.result.term.missing": "\u7f3a\u5c11", "select.version": "\u9009\u62e9\u5f53\u524d\u7248\u672c"}, "version": null}</script>
    
    
      <script src="../assets/javascripts/bundle.f55a23d4.min.js"></script>
      
        <script src="../javascripts/extra.js"></script>
      
        <script src="../javascripts/mathjax.js"></script>
      
        <!--        <script src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>-->
      
        <script src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
      
    
  </body>
</html>