Filtrar pesquisas por relevância no nível do documento
Mantenha tudo organizado com as coleções
Salve e categorize o conteúdo com base nas suas preferências.
Ao pesquisar no app da Vertex AI para Pesquisa, é possível aplicar um limite de relevância para que apenas os documentos que atendem a esse limite sejam retornados como resultados. Esta página explica como especificar um
limiar de relevância para reduzir o número de documentos retornados em
consultas.
Sobre a filtragem por relevância no nível do documento
Cada documento retornado por uma consulta de pesquisa recebe um nível de relevância, que
indica a relevância do documento retornado para a consulta. Ao fazer uma consulta por uma chamada de API, é possível definir um limite de relevância. Definir um limite de relevância alto pode reduzir o número de documentos retornados por uma consulta.
Por exemplo, se você perceber que a pesquisa está retornando muitos documentos de relevância insuficiente para seus usuários, defina o limite de relevância como "Alto" para restringir os resultados apenas aos mais relevantes. Se a configuração alta for muito restritiva, tente a média.
Tipos de dados e apps compatíveis com o filtro de relevância no nível do documento
O filtro de relevância no nível do documento pode ser aplicado a repositórios de dados com os seguintes tipos de dados:
Dados de sites com indexação avançada de sites
Dados não estruturados personalizados
Dados estruturados personalizados
O filtro de relevância no nível do documento não funciona para repositórios de dados com indexação básica de sites, dados de mídia ou dados de saúde.
Além disso, o filtro de relevância no nível do documento não pode ser usado com apps de pesquisa combinada. Os apps de pesquisa combinada são conectados a vários repositórios de dados.
Outros tipos de filtros
O filtro de relevância no nível do documento não é a única maneira de filtrar os dados retornados pelas consultas. Também é possível usar expressões de filtro para filtrar resultados com base em metadados (na indexação avançada de sites e repositórios de dados não estruturados com metadados) e valores de campo (em repositórios de dados estruturados).
Se você usar uma expressão de filtro e o filtro de relevância no nível do documento, a expressão será aplicada primeiro aos resultados e, em seguida, o filtro de relevância no nível do documento.
Aqui, o limite de relevância está definido como "alto", então apenas os resultados mais relevantes são retornados. Neste exemplo, apenas um documento foi considerado altamente relevante.
Teste várias consultas com diferentes limites para determinar as melhores configurações de limite para seus dados e aplicativo.
[[["Fácil de entender","easyToUnderstand","thumb-up"],["Meu problema foi resolvido","solvedMyProblem","thumb-up"],["Outro","otherUp","thumb-up"]],[["Difícil de entender","hardToUnderstand","thumb-down"],["Informações incorretas ou exemplo de código","incorrectInformationOrSampleCode","thumb-down"],["Não contém as informações/amostras de que eu preciso","missingTheInformationSamplesINeed","thumb-down"],["Problema na tradução","translationIssue","thumb-down"],["Outro","otherDown","thumb-down"]],["Última atualização 2025-09-03 UTC."],[[["\u003cp\u003eVertex AI Search allows filtering search results by document-level relevance, reducing the number of returned documents based on their relevance to the query.\u003c/p\u003e\n"],["\u003cp\u003eYou can set the relevance threshold to \u003ccode\u003eHIGH\u003c/code\u003e, \u003ccode\u003eMEDIUM\u003c/code\u003e, \u003ccode\u003eLOW\u003c/code\u003e, or \u003ccode\u003eLOWEST\u003c/code\u003e when making an API call to narrow down the search results to only the most relevant ones, using the \u003ccode\u003erelevanceThreshold\u003c/code\u003e field.\u003c/p\u003e\n"],["\u003cp\u003eThis document-level relevance filter is applicable to data stores with website data with advanced indexing, generic unstructured data, and generic structured data, but it is not supported for blended search apps, or data stores with basic website indexing, media data, or healthcare data.\u003c/p\u003e\n"],["\u003cp\u003eDocument-level relevance filtering can be used alongside filter expressions based on metadata or field values, with filter expressions being applied first.\u003c/p\u003e\n"],["\u003cp\u003eTo use this filtering method, search over an app using the \u003ccode\u003eengines.servingConfigs.search\u003c/code\u003e method, and input your app ID and query alongside the relevance threshold.\u003c/p\u003e\n"]]],[],null,["# Filter searches by document-level relevance\n\n| **Note:** This feature is a Preview offering, subject to the \"Pre-GA Offerings Terms\" of the [GCP Service Specific Terms](https://cloud.google.com/terms/service-terms). Pre-GA products and features may have limited support, and changes to pre-GA products and features may not be compatible with other pre-GA versions. For more information, see the [launch stage descriptions](https://cloud.google.com/products#product-launch-stages). Further, by using this feature, you agree to the [Generative AI Preview terms and conditions](https://cloud.google.com/trustedtester/aitos) (\"Preview Terms\"). For this feature, you can process personal data as outlined in the [Cloud Data Processing Addendum](https://cloud.google.com/terms/data-processing-terms), subject to applicable restrictions and obligations in the Agreement (as defined in the Preview Terms).\n|\n| \u003cbr /\u003e\n|\nWhen searching in your Vertex AI Search app, you can apply a\nrelevance threshold so that only the documents that meet this threshold\nare returned as results. This page explains how to specify a\nrelevance threshold in order to reduce the number of documents returned in\nqueries.\n\nAbout filtering by document-level relevance\n-------------------------------------------\n\nEach document returned by a search query is given a relevance level, which\nindicates the relevance of the returned document to the query. When you make a\nquery through an API call, you can set a relevance threshold. Setting a high\nrelevance threshold can reduce the number of documents returned by a query.\n\nFor example, if you find that search is returning too many documents of\ninsufficient relevance to your users, set the relevance threshold to high to\nnarrow the results to only those few that are most relevant. If the high setting\nis too restrictive, try the medium setting.\n| **Note:** This document-level relevance filtering feature is different from and less precise than the [document-relevance score](/generative-ai-app-builder/docs/preview-search-results#relevance-scores) that can be returned for search results.\n\nData types and apps supported for document-level relevance filter\n-----------------------------------------------------------------\n\nThe document-level relevance filter can be applied to data stores with following kinds of data:\n\n- Website data with advanced website indexing\n- Custom unstructured data\n- Custom structured data\n\nThe document-level relevance filter doesn't work for data stores with basic website indexing,\nmedia data, or healthcare data.\n\nFurthermore, the document-level relevance filter can't be used with blended search apps. Blended\nsearch apps are apps that are connected to multiple data stores.\n\nOther kinds of filters\n----------------------\n\nThe document-level relevance filter is not the only way you can filter data returned by queries. You\ncan also use filter expressions to filter results based on metadata (in\nadvanced website indexing and unstructured data with metadata data stores) and field\nvalues (in structured data stores).\n\nFor information, see:\n\n- [Filter expressions with advanced website indexing](/generative-ai-app-builder/docs/filter-website-search#filter-expressions-advanced-indexing)\n\n- [Filter custom search for structured or unstructured data](/generative-ai-app-builder/docs/filter-search-metadata)\n\nIf you use both a filter expression and the document-level relevance filter, the filter expression\nis applied first to the results and then the document-level relevance filter is applied.\n\nBefore you begin\n----------------\n\nMake sure you have created an app and data store and have ingested data\ninto your data store. For more information, see [Create a search\napp](/generative-ai-app-builder/docs/create-engine-es). See also [Data types and apps supported for\ndocument-level relevance filter](#supported).\n\nSearch and filter results by document-level relevance\n-----------------------------------------------------\n\nTo filter by relevance, follow these steps:\n| **Note:** You can search over an app using the [`engines.servingConfigs.search`](/generative-ai-app-builder/docs/reference/rest/v1/projects.locations.collections.engines.servingConfigs/search) method and you can search over a data store using the [`dataStores.servingConfigs.search`](/generative-ai-app-builder/docs/reference/rest/v1/projects.locations.collections.dataStores.servingConfigs/search) method. For the following procedure, Google recommends searching using the `engines.servingConfigs.search` method.\n\n1. Find your app ID. If you already have your app ID, skip to the next step.\n\n 1. In the Google Cloud console, go to the **AI Applications** page.\n\n [Go to Apps](https://console.cloud.google.com/gen-app-builder/engines)\n 2. On the **Apps** page, find the name of your app and get the app's ID from\n the **ID** column.\n\n2. To filter search by document-level relevance, use the `relevanceThreshold`\n field with the [`engines.servingConfigs.search`](/generative-ai-app-builder/docs/reference/rest/v1alpha/projects.locations.collections.engines.servingConfigs/search) method.\n\n **Key Term:** In Vertex AI Search, the term *app* can be used interchangeably with the term *engine* in the context of APIs. \n\n curl -X POST -H \"Authorization: Bearer $(gcloud auth application-default print-access-token)\" \\\n -H \"Content-Type: application/json\" \\\n \"https://discoveryengine.googleapis.com/v1alpha/projects/\u003cvar translate=\"no\"\u003ePROJECT_ID\u003c/var\u003e/locations/global/collections/default_collection/engines/\u003cvar translate=\"no\"\u003eAPP_ID\u003c/var\u003e/servingConfigs/default_search:search\" \\\n -d '{\n \"servingConfig\": \"projects/\u003cvar translate=\"no\"\u003ePROJECT_ID\u003c/var\u003e/locations/global/collections/default_collection/engines/\u003cvar translate=\"no\"\u003eAPP_ID\u003c/var\u003e/servingConfigs/default_search\",\n \"query\": \"\u003cvar translate=\"no\"\u003eQUERY\u003c/var\u003e\",\n \"relevanceThreshold\": \"\u003cvar translate=\"no\"\u003eRELEVANCE_THRESHOLD\u003c/var\u003e\"\n }'\n\n Replace the following:\n - \u003cvar translate=\"no\"\u003ePROJECT_ID\u003c/var\u003e: the ID of your Google Cloud project.\n - \u003cvar translate=\"no\"\u003eAPP_ID\u003c/var\u003e: the ID of the Vertex AI Search app that you want to query.\n - \u003cvar translate=\"no\"\u003eQUERY\u003c/var\u003e: the query text to search.\n - \u003cvar translate=\"no\"\u003eRELEVANCE_THRESHOLD\u003c/var\u003e: one of the following: `HIGH`, `MEDIUM`, `LOW`, `LOWEST`.\n\n #### Example command and result\n\n ```\n curl -X POST -H \"Authorization: Bearer $(gcloud auth print-access-token)\"\n -H \"Content-Type: application/json\" \\\n \"https://discoveryengine.googleapis.com/v1alpha/projects/my-project-123/locations/global/collections/default_collection/engines/my-search-app/servingConfigs/default_search:search\" \\\n -d '{\n \"servingConfig\": \"projects/my-project-123/locations/global/collections/default_collection/engines/my-search-app/servingConfigs/default_search\",\n \"query\": \"What is the check grounding API\",\n \"relevanceThreshold\": \"HIGH\"\n }'\n\n {\n \"results\": [\n {\n \"id\": \"a082e70352c073a4443502477255bd2a\",\n \"document\": {\n \"name\": \"projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/a082e70352c073a4443502477255bd2a\",\n \"id\": \"a082e70352c073a4443502477255bd2a\",\n \"derivedStructData\": {\n \"displayLink\": \"cloud.google.com\",\n \"link\": \"https://cloud.google.com/generative-ai-app-builder/docs/check-grounding\",\n \"htmlTitle\": \"Check grounding\",\n \"title\": \"Check grounding\"\n }\n }\n }\n ],\n \"totalSize\": 1,\n \"attributionToken\": \"f_B-CgwIidzwswYQyue15gESJDY2N2M1NmJkLTAwMDAtMjk3Ni1iMGI4LTg4M2QyNGZmNTZhOCIHR0VORVJJQypAjr6dFavEii3b7Ygt3o-aIoCymiLC8J4Vo4CXIra3jC3Usp0V24-aIt7tiC3n7YgtrsSKLeTtiC2DspoixsvzFw\",\n \"guidedSearchResult\": {},\n \"summary\": {}\n }\n ```\n\n Here, the relevance threshold is set to high, so only the most\n relevant results are returned. In this example, only one document was determined\n to be highly relevant.\n3. Test multiple queries with different thresholds to determine the best\n threshold settings for your data and application."]]
