Buckets:

hf-doc-build
/

doc

Files

xet

hf-doc-build/doc / cookbook /main /ko /vector_search_agent.html

rtrm

5 months ago

download

raw

56.1 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"🤖 벡터 검색 에이전트: 허깅페이스 허브를 백엔드로 하는 지능형 검색 엔진","local":"-벡터-검색-에이전트-허깅페이스-허브를-백엔드로-하는-지능형-검색-엔진","sections":[{"title":"Agentic 접근 방식의 차별점","local":"agentic-접근-방식의-차별점","sections":[],"depth":3},{"title":"DuckDB?","local":"duckdb","sections":[],"depth":3}],"depth":1}">
	<link href="/docs/cookbook/main/ko/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/entry/start.4b65b044.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/scheduler.e0212c12.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/singletons.a4069010.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/paths.d649b3de.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/entry/app.b90033ef.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/preload-helper.1e412078.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/index.a585a37c.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/nodes/0.64193f5f.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/each.e59479a4.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/nodes/12.a5159399.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.2ba1ebea.js">
	<link rel="modulepreload" href="/docs/cookbook/main/ko/_app/immutable/chunks/DocNotebookDropdown.26e1fda8.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"🤖 벡터 검색 에이전트: 허깅페이스 허브를 백엔드로 하는 지능형 검색 엔진","local":"-벡터-검색-에이전트-허깅페이스-허브를-백엔드로-하는-지능형-검색-엔진","sections":[{"title":"Agentic 접근 방식의 차별점","local":"agentic-접근-방식의-차별점","sections":[],"depth":3},{"title":"DuckDB?","local":"duckdb","sections":[],"depth":3}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="flex space-x-1 absolute z-10 right-0 top-0" style=""> <a href="https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/ko/vector_search_agent.ipynb" target="_blank"><img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"></a> </div> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg></span> <span>Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-7 max-sm:h-7 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible sm:size-3.5 size-3 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg></button></div> </div> <h1 class="relative group"><a id="-벡터-검색-에이전트-허깅페이스-허브를-백엔드로-하는-지능형-검색-엔진" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#-벡터-검색-에이전트-허깅페이스-허브를-백엔드로-하는-지능형-검색-엔진"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>🤖 벡터 검색 에이전트: 허깅페이스 허브를 백엔드로 하는 지능형 검색 엔진</span></h1> <p data-svelte-h="svelte-owjfcg"><em>참조: <a href="https://github.com/MartinEls" rel="nofollow">Martin Elstner</a> </em>작성자: <a href="https://github.com/ahnjj" rel="nofollow">안정</a></p> <p data-svelte-h="svelte-1c92z2c">검색 엔진은 크게 키워드 검색과 벡터 검색으로 나눌 수 있습니다.
	키워드 검색과 달리, 벡터 검색을 사용할 때는 두 가지를 고려해야 합니다.</p> <ol data-svelte-h="svelte-368w9u"><li>적합한 임베딩 모델로 데이터셋과 쿼리를 임베딩하는 작업</li> <li>임베딩 데이터를 처리할 수 있는 DB</li></ol> <p data-svelte-h="svelte-1x27ct1">하지만 임베딩 값을 기반으로 한 벡터 검색만으로는 ‘사용자가 원하는 답변’을 보장하기 어렵습니다.</p> <p data-svelte-h="svelte-1az0hy6">그래서 검색 과정에서 에이전트가 자율적으로 판단하고 최적화한다면, 사용자 의도에 더 가까운 검색 결과를 얻을 수 있습니다.</p> <h3 class="relative group"><a id="agentic-접근-방식의-차별점" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#agentic-접근-방식의-차별점"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Agentic 접근 방식의 차별점</span></h3> <p data-svelte-h="svelte-1qp8008">기존 벡터 검색 워크플로우</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->데이터 ➡ 데이터 임베딩<span class="hljs-comment">(고정 모델)</span> ➡ 인덱스 생성 ➡ 사용자 질의 ➡ 유사도 검색 ➡ 답변<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1qd4ash">똑똑한 Agentic 방식</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->사용자 질의 분석<span class="hljs-comment">(검색 전략 세움)</span> ➡ 최적 임베딩 모델 선택 ➡ 데이터 임베딩 ➡ 인덱스 생성 ➡ 유사도 검색 ➡ 검색결과 바탕으로 답변 정제<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="duckdb" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#duckdb"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>DuckDB?</span></h3> <p data-svelte-h="svelte-1b38992">허깅페이스의 데이터셋은 파켓(parquet) 파일을 기반으로 동작하는데, 이때 빠른 인메모리 데이터베이스 시스템인 <a href="https://huggingface.co/docs/hub/en/datasets-duckdb" rel="nofollow">DuckDB를 사용하면 이 파일들과 상호작용</a>할 수 있습니다. 또한 DuckDB의 기능 중 하나는 <a href="https://duckdb.org/docs/extensions/vss.html" rel="nofollow">벡터 유사도 검색</a>으로, 인덱스 유무에 관계없이 사용할 수 있습니다.</p> <p data-svelte-h="svelte-1cn3q17">이번 노트북에서는 단일 Agent에 여러가지 도구를 주어 수행하는 간단한 Agentic 벡터 검색 엔진을 만들어 보겠습니다.</p> <p data-svelte-h="svelte-1af45jk">필요한 의존성을 설치합니다 :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># 본 예제 파일은 Python 3.10 이상 버전에서만 실행할 수 있습니다.</span>
	!pip install -U smolagents datasets sentence-transformers duckdb openai<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-3y32x0">HuggingFace의 추론 API를 사용하기 위해 로그인합니다 :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> notebook_login

	notebook_login()<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> smolagents <span class="hljs-keyword">import</span> tool
	<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> Dataset
	<span class="hljs-keyword">import</span> os

	<span class="hljs-comment"># 도구 사용을 위해 OPENAI 키를 발급 받아야합니다.</span>
	os.environ[<span class="hljs-string">"OPENAI_API_KEY"</span>] = <span class="hljs-string">"YOUR KEY"</span><!-- HTML_TAG_END --></pre></div> <h1 class="relative group"><a id="도구-정의" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#도구-정의"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>도구 정의</span></h1> <p data-svelte-h="svelte-1n2ss7b">정의할 도구는 아래와 같습니다.</p> <ul data-svelte-h="svelte-c88pvg"><li>임베딩 생성 도구</li> <li>인덱스 생성 도구</li> <li>유사도 검색 도구</li> <li>답변 생성 도구</li></ul> <h3 class="relative group"><a id="도구1--임베딩-생성" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#도구1--임베딩-생성"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>도구1 : 임베딩 생성</span></h3> <p data-svelte-h="svelte-1xn2q0w">일반적으로, 임베딩 작업에서는 작은 배치 사이즈로 데이터를 청킹하지만 여기서는 단순히 데이터셋을 임베딩으로 변환하는 과정만 수행하겠습니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">@tool</span>
	<span class="hljs-keyword">def</span> <span class="hljs-title function_">create_embeddings</span>(<span class="hljs-params">
	dataset: Dataset,
	model_id: <span class="hljs-built_in">str</span>,
	column_name: <span class="hljs-built_in">str</span>,
	</span>) -> Dataset:
	<span class="hljs-string">"""
	주어진 데이터셋에 대해 임베딩을 생성합니다.

	Args:
	dataset: 임베딩을 생성할 대상 데이터셋
	model_id: 임베딩에 사용할 모델
	column_name: 임베딩할 열 이름

	Returns:
	임베딩이 추가된 데이터셋
	"""</span>
	<span class="hljs-keyword">from</span> sentence_transformers <span class="hljs-keyword">import</span> SentenceTransformer

	model = SentenceTransformer(model_id)

	<span class="hljs-keyword">def</span> <span class="hljs-title function_">embed_batch</span>(<span class="hljs-params">batch</span>):
	embeddings = model.encode(batch[column_name], convert_to_numpy=<span class="hljs-literal">True</span>)
	batch[<span class="hljs-string">"embeddings"</span>] = embeddings.tolist()
	<span class="hljs-keyword">return</span> batch

	dataset = dataset.<span class="hljs-built_in">map</span>(embed_batch, batched=<span class="hljs-literal">True</span>)

	<span class="hljs-keyword">return</span> dataset<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="duckdb로-벡터-검색-수행하기" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#duckdb로-벡터-검색-수행하기"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>DuckDB로 벡터 검색 수행하기</span></h2> <p data-svelte-h="svelte-18k1ji7"><code>duckdb</code>를 사용하여 데이터셋에서 벡터 검색을 수행할 수 있습니다.
	인덱스는 사용하거나 사용하지 않을 수 있습니다. 인덱스를 <strong>활용하지 않고</strong> 검색하는 것은 더 느리지만 더 정확하고, 인덱스를 <strong>활용하여</strong> 검색하는 것은 더 빠르지만 덜 정확합니다.</p> <h3 class="relative group"><a id="인덱스를-활용하지-않고-검색하기" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#인덱스를-활용하지-않고-검색하기"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>인덱스를 활용하지 않고 검색하기</span></h3> <p data-svelte-h="svelte-of1kmm">인덱스를 활용하지 않고 검색하는 것은 느린 작업지만 일반적으로 약 10만 행까지의 작은 데이터셋에서는 충분히 빠르게 동작합니다. 하지만 이번 노트북에서는 DuckDB의 인덱스를 활용해서 검색해보겠습니다.</p> <h3 class="relative group"><a id="인덱스를-활용하여-검색하기" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#인덱스를-활용하여-검색하기"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>인덱스를 활용하여 검색하기</span></h3> <p data-svelte-h="svelte-16chymz">이 접근법은 데이터셋의 로컬 복사본을 생성하고 이를 사용하여 인덱스를 생성합니다. 약간의 오버헤드가 있지만 한번 인덱스를 생성한 후에는 검색 속도가 개선될 것입니다.</p> <h3 class="relative group"><a id="도구2--duckdb-인덱스-만들기" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#도구2--duckdb-인덱스-만들기"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>도구2 : DuckDB 인덱스 만들기</span></h3> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">@tool</span>
	<span class="hljs-keyword">def</span> <span class="hljs-title function_">create_db_index</span>(<span class="hljs-params">
	dataset_with_embeddings: Dataset, <span class="hljs-comment"># 임베딩이 포함된 데이터셋</span>
	table_name: <span class="hljs-built_in">str</span>,
	embedding_column: <span class="hljs-built_in">str</span> = <span class="hljs-string">"embeddings"</span>
	</span>) -> <span class="hljs-literal">None</span>:
	<span class="hljs-string">"""
	임베딩이 포함된 데이터셋에 대해 DuckDB 인덱스를 생성합니다.

	Args:
	dataset_with_embeddings: 이미 임베딩이 포함된 데이터셋
	table_name: 생성할 테이블 이름
	embedding_column: 임베딩 열 이름

	Returns:
	None
	"""</span>
	<span class="hljs-keyword">import</span> duckdb

	<span class="hljs-comment"># VSS 확장 설치 및 로드</span>
	duckdb.sql(<span class="hljs-string">"INSTALL vss; LOAD vss;"</span>)
	duckdb.sql(<span class="hljs-string">f"DROP TABLE IF EXISTS <span class="hljs-subst">{table_name}</span>;"</span>)

	<span class="hljs-comment"># 데이터셋을 pandas DataFrame으로 변환</span>
	df = dataset_with_embeddings.to_pandas()

	<span class="hljs-comment"># DuckDB에 DataFrame 등록</span>
	duckdb.register(<span class="hljs-string">f"<span class="hljs-subst">{table_name}</span>_temp"</span>, df)

	<span class="hljs-comment"># 모델에서 임베딩 차원 가져오기</span>
	embedding_dim = <span class="hljs-built_in">len</span>(df[embedding_column].iloc[<span class="hljs-number">0</span>])
	<span class="hljs-comment"># embedding_dim = model.get_sentence_embedding_dimension()</span>

	<span class="hljs-comment"># 테이블 생성 (임베딩을 FLOAT 배열로 변환)</span>
	duckdb.sql(<span class="hljs-string">f"""
	CREATE TABLE <span class="hljs-subst">{table_name}</span> AS
	SELECT *, <span class="hljs-subst">{embedding_column}</span>::FLOAT[<span class="hljs-subst">{embedding_dim}</span>] AS <span class="hljs-subst">{embedding_column}</span>_float
	FROM <span class="hljs-subst">{table_name}</span>_temp;
	"""</span>)

	<span class="hljs-comment"># HNSW 인덱스 생성</span>
	duckdb.sql(<span class="hljs-string">f"""
	CREATE INDEX idx_<span class="hljs-subst">{embedding_column}</span> ON <span class="hljs-subst">{table_name}</span>
	USING HNSW (<span class="hljs-subst">{embedding_column}</span>_float) WITH (metric = 'cosine');
	"""</span>)

	<span class="hljs-comment"># 임시 테이블 정리</span>
	duckdb.sql(<span class="hljs-string">f"DROP VIEW IF EXISTS <span class="hljs-subst">{table_name}</span>_temp;"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-b4gp9c">이 도구를 통해 인덱스를 사용하여 벡터 검색을 수행할 수 있으며, 결과는 즉시 반환됩니다.</p> <h3 class="relative group"><a id="도구3--벡터-검색-수행하기" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#도구3--벡터-검색-수행하기"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>도구3 : 벡터 검색 수행하기</span></h3> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">@tool</span>
	<span class="hljs-keyword">def</span> <span class="hljs-title function_">similarity_search_with_duckdb_index</span>(<span class="hljs-params">
	query: <span class="hljs-built_in">str</span>,
	table_name: <span class="hljs-built_in">str</span>,
	model_id: <span class="hljs-built_in">str</span>,
	k: <span class="hljs-built_in">int</span> = <span class="hljs-number">5</span>,
	embedding_column: <span class="hljs-built_in">str</span> = <span class="hljs-string">"embeddings"</span>
	</span>)-> <span class="hljs-built_in">dict</span>:
	<span class="hljs-string">"""
	DuckDB 인덱스를 이용해 벡터 검색을 수행합니다.

	Args:
	query: 검색할 쿼리 문자열
	model_id: 임베딩에 사용할 모델
	k: 반환할 결과 수
	table_name: 검색할 테이블 이름
	embedding_column: 임베딩 컬럼 이름

	Returns:
	dict: 검색 결과
	"""</span>
	<span class="hljs-keyword">from</span> sentence_transformers <span class="hljs-keyword">import</span> SentenceTransformer
	<span class="hljs-keyword">import</span> duckdb

	model = SentenceTransformer(model_id)
	embedding = model.encode(query).tolist()
	<span class="hljs-keyword">return</span> duckdb.sql(
	query=<span class="hljs-string">f"""
	SELECT *, array_cosine_distance(<span class="hljs-subst">{embedding_column}</span>_float, <span class="hljs-subst">{embedding}</span>::FLOAT[<span class="hljs-subst">{model.get_sentence_embedding_dimension()}</span>]) as distance
	FROM <span class="hljs-subst">{table_name}</span>
	ORDER BY distance
	LIMIT <span class="hljs-subst">{k}</span>;
	"""</span>
	).to_df()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1yfgbs2">무거운 벡터 검색 엔진을 따로 배포할 필요 없고, 저장소는 허브에서 처리됩니다.</p> <h3 class="relative group"><a id="도구4--답변-생성-도구" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#도구4--답변-생성-도구"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>도구4 : 답변 생성 도구</span></h3> <p data-svelte-h="svelte-1k09lse">유사도 검색 결과 청크를 기반으로, LLM이 사용자가 원할 만한 답변을 생성합니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">@tool</span>
	<span class="hljs-keyword">def</span> <span class="hljs-title function_">generate_answer</span>(<span class="hljs-params">chunks: <span class="hljs-built_in">list</span>, query: <span class="hljs-built_in">str</span></span>) -> <span class="hljs-built_in">str</span>:
	<span class="hljs-string">"""
	쿼리를 기반으로 주어진 텍스트 청크 목록에서 답변을 생성합니다.

	Args:
	chunks: 답변 생성에 사용할 텍스트 청크 목록
	query: 답변할 쿼리 문자열

	Returns:
	str: 생성된 답변
	"""</span>
	<span class="hljs-keyword">import</span> openai <span class="hljs-comment"># OPENAI 키 발급이 필요합니다.</span>

	context = <span class="hljs-string">"\n\n"</span>.join(chunks)
	prompt = <span class="hljs-string">f"Context:\n<span class="hljs-subst">{context}</span>\n\nQuestion: <span class="hljs-subst">{query}</span>\nAnswer:"</span>
	response = openai.ChatCompletion.create(
	model=<span class="hljs-string">"gpt-4"</span>,
	messages=[{<span class="hljs-string">"role"</span>: <span class="hljs-string">"user"</span>, <span class="hljs-string">"content"</span>: prompt}]
	)
	<span class="hljs-keyword">return</span> response[<span class="hljs-string">"choices"</span>][<span class="hljs-number">0</span>][<span class="hljs-string">"message"</span>][<span class="hljs-string">"content"</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-en7cjr">도구를 정의하였으니, 데이터셋을 로드하고 에이전트를 동작해보겠습니다.</p> <p data-svelte-h="svelte-53u7m">사용한 데이터셋은 <a href="https://huggingface.co/datasets/huggingface-KREW/KoCultre-Descriptions" rel="nofollow">huggingface-KREW/KoCultre-Descriptions</a>으로, 한국어 밈(meme)에 대한 데이터입니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset

	dataset = load_dataset(<span class="hljs-string">"huggingface-KREW/KoCultre-Descriptions"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-gsyogr">에이전트를 정의합니다. 사용할 모델은 Qwen/Qwen2.5-Coder-32B-Instruct 입니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> smolagents <span class="hljs-keyword">import</span> CodeAgent, InferenceClientModel

	model = InferenceClientModel(
	<span class="hljs-string">"Qwen/Qwen2.5-Coder-32B-Instruct"</span>,
	provider=<span class="hljs-string">"together"</span>,
	max_tokens=<span class="hljs-number">2048</span>
	)

	tools = [
	create_embeddings, <span class="hljs-comment"># 임베딩 생성 도구</span>
	create_db_index, <span class="hljs-comment"># 인덱스 생성 도구</span>
	similarity_search_with_duckdb_index, <span class="hljs-comment"># 유사도 검색 도구</span>
	generate_answer <span class="hljs-comment"># 답변 생성 도구</span>
	]

	agent = CodeAgent(
	model=model,
	tools=tools,
	verbosity_level=<span class="hljs-number">2</span>,
	max_steps=<span class="hljs-number">15</span>
	)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-25stim">에이전트가 자율적으로 임베딩 모델을 택할 수 있도록 프롬프트를 구성하겠습니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">agentic_prompt</span>(<span class="hljs-params">query: <span class="hljs-built_in">str</span></span>):
	<span class="hljs-keyword">return</span> <span class="hljs-string">f"""
	당신은 지능형 검색 전문가입니다. <span class="hljs-subst">{query}</span>:

	검색을 위해 먼저 쿼리와 데이터셋을 분석하고, 분석 결과에 따라 가장 적합한 임베딩 모델을 선택하세요
	- sentence-transformers/all-MiniLM-L6-v2 (fast, lightweight, good for simple queries)
	- sentence-transformers/all-mpnet-base-v2 (balanced performance)
	- intfloat/e5-large-v2 (high quality for complex tasks)
	- minishlab/potion-base-8M (very efficient)
	"""</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-q81us6">데이터셋에 대한 설명과, 검색어를 입력하고 검색엔진 에이전트를 실행합니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->result = agent.run(
	agentic_prompt(query=<span class="hljs-string">"아이돌 관련 밈 알려주세요"</span>),
	additional_args={<span class="hljs-string">"dataset"</span>: dataset,
	<span class="hljs-string">"dataset_description"</span>: <span class="hljs-string">"한국어 밈(meme)에 대한 데이터셋으로, meme과(title), meme의 뜻(content)을 알려준다."</span>,
	<span class="hljs-string">"column_name"</span>: <span class="hljs-string">"content"</span>}
	)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span><span class="hljs-built_in">print</span>(<span class="hljs-string">f"Final result: <span class="hljs-subst">{result}</span>"</span>)<!-- HTML_TAG_END --></pre></div> <pre data-svelte-h="svelte-dwlxai">Final result: 엄마가되는 밈은 최근 SNS와 온라인 커뮤니티에서 유행하는 표현으로, 주로 누군가가 매우 귀엽거나 사랑스러워서 모성애적 감정을 느낄 때 사용됩니다. 이 표현은 아이돌 팬덤 문화에서 자주 사용되며, '엄마가 돼'가 올바른 표기이지만 의도적인 맞춤법 파괴를 통해 밈으로서의 독특한 성격을 갖게 되었습니다.
	</pre> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">Final</span> <span class="hljs-keyword">result</span>: 엄마가되는 밈은 최근 SNS와 온라인 커뮤니티에서 유행하는 표현으로, 주로 누군가가 매우 귀엽거나 사랑스러워서 모성애적 감정을 느낄 때 사용됩니다. 이 표현은 아이돌 팬덤 문화에서 자주 사용되며, <span class="hljs-string">'엄마가 돼'</span>가 올바른 표기이지만 의도적인 맞춤법 파괴를 통해 밈으로서의 독특한 성격을 갖게 되었습니다.<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="conclusion" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#conclusion"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Conclusion</span></h3> <p data-svelte-h="svelte-1xsmqoq">단순히 검색 결과만을 가져오는 것이 아니라, 검색결과를 바탕으로 쿼리에 따른 답변을 정제해서 보여주었습니다.✌🏻</p> <p data-svelte-h="svelte-16cnm9y">지금까지 간단한 에이전트 시스템을 만들어보았습니다. 여기에 품질 평가, 분석 등을 추가한다면 진정한 Agentic 검색 엔진을 구현할 수 있습니다.</p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/cookbook/blob/main/notebooks/ko/vector_search_agent.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p>

	<script>
	{
	__sveltekit_1d6oska = {
	assets: "/docs/cookbook/main/ko",
	base: "/docs/cookbook/main/ko",
	env: {}
	};

	const element = document.currentScript.parentElement;

	const data = [null,null];

	Promise.all([
	import("/docs/cookbook/main/ko/_app/immutable/entry/start.4b65b044.js"),
	import("/docs/cookbook/main/ko/_app/immutable/entry/app.b90033ef.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 12],
	data,
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 56.1 kB
Xet hash:: b66d9e52cd52aeb5d382122f520f836d63793c93e29af67673029cca5111a2d0

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.