PageSourceSearch

https://www.autoinfra.cn/assets/js/eea94652.1ae096fe.js

js autoinfra.cn collected 2026-10-03 21:14:30 UTC 12,452 bytes, 1 lines download raw bytes

1"use strict";(self.webpackChunklangchain_zh=self.webpackChunklangchain_zh||[]).push([[1341],{3905:(e,n,t)=>{t.d(n,{Zo:()=>c,kt:()=>h});var r=t(67294);function a(e,n,t){return n in e?Object.defineProperty(e,n,{value:t,enumerable:!0,configurable:!0,writable:!0}):e[n]=t,e}function o(e,n){var t=Object.keys(e);if(Object.getOwnPropertySymbols){var r=Object.getOwnPropertySymbols(e);n&&(r=r.filter((function(n){return Object.getOwnPropertyDescriptor(e,n).enumerable}))),t.push.apply(t,r)}return t}function l(e){for(var n=1;n<arguments.length;n++){var t=null!=arguments[n]?arguments[n]:{};n%2?o(Object(t),!0).forEach((function(n){a(e,n,t[n])})):Object.getOwnPropertyDescriptors?Object.defineProperties(e,Object.getOwnPropertyDescriptors(t)):o(Object(t)).forEach((function(n){Object.defineProperty(e,n,Object.getOwnPropertyDescriptor(t,n))}))}return e}function s(e,n){if(null==e)return{};var t,r,a=function(e,n){if(null==e)return{};var t,r,a={},o=Object.keys(e);for(r=0;r<o.length;r++)t=o[r],n.indexOf(t)>=0||(a[t]=e[t]);return a}(e,n);if(Object.getOwnPropertySymbols){var o=Object.getOwnPropertySymbols(e);for(r=0;r<o.length;r++)t=o[r],n.indexOf(t)>=0||Object.prototype.propertyIsEnumerable.call(e,t)&&(a[t]=e[t])}return a}var p=r.createContext({}),i=function(e){var n=r.useContext(p),t=n;return e&&(t="function"==typeof e?e(n):l(l({},n),e)),t},c=function(e){var n=i(e.components);return r.createElement(p.Provider,{value:n},e.children)},m="mdxType",u={inlineCode:"code",wrapper:function(e){var n=e.children;return r.createElement(r.Fragment,{},n)}},d=r.forwardRef((function(e,n){var t=e.components,a=e.mdxType,o=e.originalType,p=e.parentName,c=s(e,["components","mdxType","originalType","parentName"]),m=i(t),d=a,h=m["".concat(p,".").concat(d)]||m[d]||u[d]||o;return t?r.createElement(h,l(l({ref:n},c),{},{components:t})):r.createElement(h,l({ref:n},c))}));function h(e,n){var t=arguments,a=n&&n.mdxType;if("string"==typeof e||a){var o=t.length,l=new Array(o);l[0]=d;var s={};for(var p in n)hasOwnProperty.call(n,p)&&(s[p]=n[p]);s.originalType=e,s[m]="string"==typeof e?e:a,l[1]=s;for(var i=2;i<o;i++)l[i]=t[i];return r.createElement.apply(null,l)}return r.createElement.apply(null,t)}d.displayName="MDXCreateElement"},57212:(e,n,t)=>{t.r(n),t.d(n,{assets:()=>p,contentTitle:()=>l,default:()=>h,frontMatter:()=>o,metadata:()=>s,toc:()=>i});var r=t(87462),a=(t(67294),t(3905));const o={sidebar_position:7},l="\u4f7f\u7528HyDE\u6539\u8fdb\u6587\u6863\u7d22\u5f15",s={unversionedId:"use_cases/question_answering/how_to/hyde",id:"use_cases/question_answering/how_to/hyde",title:"\u4f7f\u7528HyDE\u6539\u8fdb\u6587\u6863\u7d22\u5f15",description:"\u672c\u7b14\u8bb0\u672c\u4ecb\u7ecd\u4e86\u5982\u4f55\u4f7f\u7528Hypothetical Document Embeddings (HyDE)\u6765\u6539\u8fdb\u6587\u6863\u7d22\u5f15\uff0c\u5982\u6b64\u8bba\u6587\u6240\u8ff0\u3002",source:"@site/docs/use_cases/question_answering/how_to/hyde.md",sourceDirName:"use_cases/question_answering/how_to",slug:"/use_cases/question_answering/how_to/hyde",permalink:"/docs/use_cases/question_answering/how_to/hyde",draft:!1,tags:[],version:"current",sidebarPosition:7,frontMatter:{sidebar_position:7},sidebar:"use_cases",previous:{title:"\u901a\u8fc7FLARE\u8fdb\u884c\u751f\u6210\u65f6\u68c0\u7d22",permalink:"/docs/use_cases/question_answering/how_to/flare"},next:{title:"\u4f7f\u7528\u672c\u5730LLM",permalink:"/docs/use_cases/question_answering/how_to/local_retrieval_qa"}},p={},i=[{value:"\u591a\u4e2a\u751f\u6210",id:"\u591a\u4e2a\u751f\u6210",level:2},{value:"\u4f7f\u7528\u81ea\u5df1\u7684\u63d0\u793a",id:"\u4f7f\u7528\u81ea\u5df1\u7684\u63d0\u793a",level:2},{value:"\u4f7f\u7528HyDE",id:"\u4f7f\u7528hyde",level:2}],c=(m="CodeOutputBlock",function(e){return console.warn("Component "+m+" was not imported, exported, or provided by MDXProvider as global scope"),(0,a.kt)("div",e)});var m;const u={toc:i},d="wrapper";function h(e){let{components:n,...t}=e;return(0,a.kt)(d,(0,r.Z)({},u,t,{components:n,mdxType:"MDXLayout"}),(0,a.kt)("h1",{id:"\u4f7f\u7528hyde\u6539\u8fdb\u6587\u6863\u7d22\u5f15"},"\u4f7f\u7528HyDE\u6539\u8fdb\u6587\u6863\u7d22\u5f15"),(0,a.kt)("p",null,"\u672c\u7b14\u8bb0\u672c\u4ecb\u7ecd\u4e86\u5982\u4f55\u4f7f\u7528Hypothetical Document Embeddings (HyDE)\u6765\u6539\u8fdb\u6587\u6863\u7d22\u5f15\uff0c\u5982",(0,a.kt)("a",{parentName:"p",href:"https://arxiv.org/abs/2212.10496"},"\u6b64\u8bba\u6587"),"\u6240\u8ff0\u3002"),(0,a.kt)("p",null,"\u5728\u9ad8\u5c42\u6b21\u4e0a\uff0cHyDE\u662f\u4e00\u79cd\u5d4c\u5165\u6280\u672f\uff0c\u5b83\u63a5\u53d7\u67e5\u8be2\uff0c\u751f\u6210\u4e00\u4e2a\u5047\u8bbe\u7684\u7b54\u6848\uff0c\u7136\u540e\u5c06\u751f\u6210\u7684\u6587\u6863\u5d4c\u5165\u5e76\u5c06\u5176\u4f5c\u4e3a\u6700\u7ec8\u793a\u4f8b\u4f7f\u7528\u3002"),(0,a.kt)("p",null,"\u4e3a\u4e86\u4f7f\u7528HyDE\uff0c\u6211\u4eec\u9700\u8981\u63d0\u4f9b\u4e00\u4e2a\u57fa\u672c\u7684\u5d4c\u5165\u6a21\u578b\uff0c\u4ee5\u53ca\u4e00\u4e2a\u53ef\u4ee5\u7528\u4e8e\u751f\u6210\u8fd9\u4e9b\u6587\u6863\u7684LLMChain\u3002\u9ed8\u8ba4\u60c5\u51b5\u4e0b\uff0c
1HyDE\u7c7b\u5e26\u6709\u4e00\u4e9b\u9ed8\u8ba4\u7684\u63d0\u793a\uff08\u6709\u5173\u8be6\u7ec6\u4fe1\u606f\uff0c\u8bf7\u53c2\u9605\u8bba\u6587\uff09\uff0c\u4f46\u6211\u4eec\u4e5f\u53ef\u4ee5\u521b\u5efa\u81ea\u5df1\u7684\u63d0\u793a\u3002"),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},"from langchain.llms import OpenAI\nfrom langchain.embeddings import OpenAIEmbeddings\nfrom langchain.chains import LLMChain, HypotheticalDocumentEmbedder\nfrom langchain.prompts import PromptTemplate\n")),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},"base_embeddings = OpenAIEmbeddings()\nllm = OpenAI()\n")),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'# \u4f7f\u7528`web_search`\u63d0\u793a\u52a0\u8f7d\nembeddings = HypotheticalDocumentEmbedder.from_llm(llm, base_embeddings, "web_search")\n')),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'# \u73b0\u5728\u6211\u4eec\u53ef\u4ee5\u50cf\u4f7f\u7528\u4efb\u4f55\u5d4c\u5165\u7c7b\u4e00\u6837\u4f7f\u7528\u5b83\uff01\nresult = embeddings.embed_query("\u6cf0\u59ec\u9675\u5728\u54ea\u91cc\uff1f")\n')),(0,a.kt)("h2",{id:"\u591a\u4e2a\u751f\u6210"},"\u591a\u4e2a\u751f\u6210"),(0,a.kt)("p",null,"\u6211\u4eec\u8fd8\u53ef\u4ee5\u751f\u6210\u591a\u4e2a\u6587\u6863\uff0c\u7136\u540e\u5c06\u8fd9\u4e9b\u6587\u6863\u7684\u5d4c\u5165\u7ec4\u5408\u8d77\u6765\u3002\u9ed8\u8ba4\u60c5\u51b5\u4e0b\uff0c\u6211\u4eec\u901a\u8fc7\u53d6\u5e73\u5747\u503c\u6765\u7ec4\u5408\u8fd9\u4e9b\u6587\u6863\u7684\u5d4c\u5165\u3002\u6211\u4eec\u53ef\u4ee5\u901a\u8fc7\u6539\u53d8\u7528\u4e8e\u751f\u6210\u6587\u6863\u7684LLM\u6765\u5b9e\u73b0\u8fd9\u4e00\u70b9\u3002"),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},"multi_llm = OpenAI(n=4, best_of=4)\n")),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'embeddings = HypotheticalDocumentEmbedder.from_llm(\n    multi_llm, base_embeddings, "web_search"\n)\n')),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'result = embeddings.embed_query("\u6cf0\u59ec\u9675\u5728\u54ea\u91cc\uff1f")\n')),(0,a.kt)("h2",{id:"\u4f7f\u7528\u81ea\u5df1\u7684\u63d0\u793a"},"\u4f7f\u7528\u81ea\u5df1\u7684\u63d0\u793a"),(0,a.kt)("p",null,"\u9664\u4e86\u4f7f\u7528\u9884\u914d\u7f6e\u7684\u63d0\u793a\u5916\uff0c\u6211\u4eec\u8fd8\u53ef\u4ee5\u8f7b\u677e\u6784\u5efa\u81ea\u5df1\u7684\u63d0\u793a\uff0c\u5e76\u5728\u751f\u6210\u6587\u6863\u7684LLMChain\u4e2d\u4f7f\u7528\u5b83\u4eec\u3002\u5982\u679c\u6211\u4eec\u77e5\u9053\u67e5\u8be2\u6240\u5728\u7684\u9886\u57df\uff0c\u8fd9\u6837\u53ef\u4ee5\u5f88\u6709\u7528\uff0c\u56e0\u4e3a\u6211\u4eec\u53ef\u4ee5\u5c06\u63d0\u793a\u6761\u4ef6\u8bbe\u7f6e\u4e3a\u751f\u6210\u66f4\u7c7b\u4f3c\u4e8e\u8be5\u9886\u57df\u7684\u6587\u672c\u3002"),(0,a.kt)("p",null,"\u5728\u4e0b\u9762\u7684\u793a\u4f8b\u4e2d\uff0c\u8ba9\u6211\u4eec\u5c06\u5176\u8bbe\u7f6e\u4e3a\u751f\u6210\u5173\u4e8e\u56fd\u60c5\u54a8\u6587\u7684\u6587\u672c\uff08\u56e0\u4e3a\u6211\u4eec\u5c06\u5728\u4e0b\u4e00\u4e2a\u793a\u4f8b\u4e2d\u4f7f\u7528\u5b83\uff09\u3002"),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'prompt_template = """\u8bf7\u56de\u7b54\u7528\u6237\u5173\u4e8e\u6700\u8fd1\u4e00\u6b21\u56fd\u60c5\u54a8\u6587\u7684\u95ee\u9898\n\u95ee\u9898\uff1a{question}\n\u7b54\u6848\uff1a"""\nprompt = PromptTemplate(input_variables=["question"], template=prompt_template)\nllm_chain = LLMChain(llm=llm, prompt=prompt)\n')),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},"embeddings = HypotheticalDocumentEmbedder(\n    llm_chain=llm_chain, base_embeddings=base_embeddings\n)\n")),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'result = embeddings.embed_query(\n    "\u603b\u7edf\u5bf9Ketanji Brown Jackson\u6709\u4ec0\u4e48\u8bf4\u6cd5\uff1f"\n)\n')),(0,a.kt)("h2",{id:"\u4f7f\u7528hyde"},"\u4f7f\u7528HyDE"),(0,a.kt)("p",null,"\u73b0\u5728\u6211\u4eec\u6709\u4e86HyDE\uff0c\u6211\u4eec\u53ef\u4ee5\u50cf\u4f7f\u7528\u4efb\u4f55\u5176\u4ed6\u5d4c\u5165\u7c7b\u4e00\u6837\u4f7f\u7528\u5b83\uff01\u4e0b\u9762\u662f\u4f7f\u7528\u5b83\u5728\u56fd\u60c5\u54a8\u6587\u793a\u4f8b\u4e2d\u67e5\u627e\u76f8\u4f3c\u6bb5\u843d\u7684\u793a\u4f8b\u3002"),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'from langchain.text_splitter import CharacterTextSplitter\nfrom langchain.vectorstores import Chroma\n\nwith open("../../state_of_the_union.txt") as f:\n    state_of_the_union = f.read()\ntext_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)\ntexts = text_splitter.split_text(state_of_the_union)\n')),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},'docsearch = Chroma.from_texts(texts, embeddings)\n\nquery = "\u603b\u7edf\u5bf9Ketanji Brown Jackson\u6709\u4ec0\u4e48\u8bf4\u6cd5\uff1f"\ndocs = docsearch.similarity_search(query)\n')),(0,a.kt)(c,{lang:"python",mdxType:"C
1odeOutputBlock"},(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre"},"    \u4f7f\u7528\u76f4\u63a5\u672c\u5730API\u8fd0\u884cChroma\u3002\n    \u4f7f\u7528DuckDB\u5185\u5b58\u6570\u636e\u5e93\u3002\u6570\u636e\u5c06\u662f\u4e34\u65f6\u7684\u3002\n"))),(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre",className:"language-python"},"print(docs[0].page_content)\n")),(0,a.kt)(c,{lang:"python",mdxType:"CodeOutputBlock"},(0,a.kt)("pre",null,(0,a.kt)("code",{parentName:"pre"},"    \u5728\u4e00\u4e2a\u4e2a\u5dde\uff0c\u65b0\u7684\u6cd5\u5f8b\u5df2\u7ecf\u901a\u8fc7\uff0c\u4e0d\u4ec5\u538b\u5236\u9009\u4e3e\u6743\uff0c\u800c\u4e14\u98a0\u8986\u6574\u4e2a\u9009\u4e3e\u8fc7\u7a0b\u3002\n    \n    \u6211\u4eec\u4e0d\u80fd\u8ba9\u8fd9\u79cd\u60c5\u51b5\u53d1\u751f\u3002\n    \n    \u4eca\u665a\uff0c\u6211\u547c\u5401\u53c2\u8bae\u9662\uff1a\u901a\u8fc7\u300a\u81ea\u7531\u6295\u7968\u6cd5\u6848\u300b\u3002\u901a\u8fc7\u300a\u7ea6\u7ff0\xb7\u5218\u6613\u65af\u9009\u4e3e\u6743\u6cd5\u6848\u300b\u3002\u800c\u4e14\uff0c\u987a\u4fbf\u8bf4\u4e00\u53e5\uff0c\u901a\u8fc7\u300a\u62ab\u9732\u6cd5\u6848\u300b\uff0c\u8fd9\u6837\u7f8e\u56fd\u4eba\u5c31\u53ef\u4ee5\u77e5\u9053\u8c01\u5728\u8d44\u52a9\u6211\u4eec\u7684\u9009\u4e3e\u3002\n    \n    \u4eca\u665a\uff0c\u6211\u60f3\u5411\u4e00\u4e2a\u81f4\u529b\u4e8e\u4e3a\u8fd9\u4e2a\u56fd\u5bb6\u670d\u52a1\u7684\u4eba\u8868\u793a\u656c\u610f\uff1a\u53f8\u6cd5\u90e8\u957f\u65af\u8482\u82ac\xb7\u5e03\u96f7\u8036\u2014\u2014\u4e00\u4f4d\u9646\u519b\u9000\u4f0d\u519b\u4eba\u3001\u5baa\u6cd5\u5b66\u8005\u548c\u5373\u5c06\u9000\u4f11\u7684\u7f8e\u56fd\u6700\u9ad8\u6cd5\u9662\u6cd5\u5b98\u3002\u53f8\u6cd5\u90e8\u957f\u5e03\u96f7\u8036\uff0c\u611f\u8c22\u60a8\u7684\u670d\u52a1\u3002\n    \n    \u4f5c\u4e3a\u603b\u7edf\uff0c\u6700\u91cd\u8981\u7684\u5baa\u6cd5\u8d23\u4efb\u4e4b\u4e00\u5c31\u662f\u63d0\u540d\u4eba\u9009\u62c5\u4efb\u7f8e\u56fd\u6700\u9ad8\u6cd5\u9662\u6cd5\u5b98\u3002\n    \n    \u6211\u57284\u5929\u524d\u5c31\u8fd9\u6837\u505a\u4e86\uff0c\u5f53\u65f6\u6211\u63d0\u540d\u4e86\u5de1\u56de\u4e0a\u8bc9\u6cd5\u9662\u6cd5\u5b98Ketanji Brown Jackson\u3002\u5979\u662f\u6211\u4eec\u56fd\u5bb6\u6700\u9876\u5c16\u7684\u6cd5\u5f8b\u4e13\u5bb6\u4e4b\u4e00\uff0c\u5c06\u7ee7\u7eed\u5ef6\u7eed\u5e03\u96f7\u8036\u6cd5\u5b98\u7684\u5353\u8d8a\u4f20\u7edf\u3002\n"))))}h.isMDXComponent=!0}}]);

Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.