spring-ai-alibaba第八章使用searxng构建大模型联网搜索应用

1、searxng安装配置 详见

anythingLLM结合searXNG实现联网搜索_anythingllm 配置 searxng-CSDN博客

2、本文介绍如何使用 Spring AI Alibaba 构建大模型联网搜索应用结合模块化 RAG（Module RAG）和信息检索服务（SearXNG）赋能大模型

总体上分为以下几个步骤：

Pre-Retrieval

增强和转换用户输入，使其更有效地执行检索任务，解决格式不正确的查询、query 语义不清晰、或不受支持的语言等。

1. QueryAugmenter 查询增强：使用附加的上下文数据信息增强用户 query，提供大模型回答问题时的必要上下文信息；
2. QueryTransformer 查询改写：因为用户的输入通常是片面的，关键信息较少，不便于大模型理解和回答问题。因此需要使用 prompt 调优手段或者大模型改写用户 query；
3. QueryExpander 查询扩展：将用户 query 扩展为多个语义不同的变体以获得不同视角，有助于检索额外的上下文信息并增加找到相关结果的机会。

Retrieval

负责查询向量存储等数据系统并检索和用户 query 相关性最高的 Document。

1. DocumentRetriever：检索器，根据 QueryExpander 使用不同的数据源进行检索，例如 搜索引擎、向量存储、数据库或知识图等；
2. DocumentJoiner：将从多个 query 和从多个数据源检索到的 Document 合并为一个 Document 集合；

Post-Retrieval

负责处理检索到的 Document 以获得最佳的输出结果，解决模型中的中间丢失和上下文长度限制等。

1. DocumentRanker：根据 Document 和用户 query 的相关性对 Dcoument 进行排序和排名；
2. DocumentSelector：用于从检索到的 Document 列表中删除不相关或冗余文档；
3. DocumentCompressor：用于压缩每个 Document，减少检索到的信息中的噪音和冗余。

生成

生成用户 Query 对应的大模型输出。

Web Search 实践

接下来，将演示如何使用 Spring AI Alibaba 和searXNG搭建联网搜索 RAG 的实现。

资源准备

DashScope apiKey：如何获取API Key_大模型服务平台百炼(Model Studio)-阿里云帮助中心

 

Pre-Retrieval

将用户 Query 使用 qwen-plus 大模型进行增强改写。

CustomContextQueryAugmenter.java 

package com.alibaba.example.chatmemory.rag.prompt;import java.util.List;
import java.util.Map;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.stream.Collectors;import org.jetbrains.annotations.NotNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import org.springframework.ai.chat.prompt.PromptTemplate;
import org.springframework.ai.document.Document;
import org.springframework.ai.rag.Query;
import org.springframework.ai.rag.generation.augmentation.QueryAugmenter;
import org.springframework.ai.util.PromptAssert;
import org.springframework.lang.Nullable;
import org.springframework.util.Assert;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/public class CustomContextQueryAugmenter implements QueryAugmenter {private static final Logger logger = LoggerFactory.getLogger(CustomContextQueryAugmenter.class);private static final PromptTemplate DEFAULT_PROMPT_TEMPLATE = new PromptTemplate("""Context information is below:{{context}}Given the context information and no prior knowledge, answer the query.Follow these rules:1. If the answer is not in the context, just say that you don't know.2. Avoid statements like "Based on the context...." or "The provided information...".Query: {query}Answer:""");private static final PromptTemplate DEFAULT_EMPTY_PROMPT_TEMPLATE = new PromptTemplate("""The user query is outside your knowledge base.Politely inform the user that you cannot answer the query.""");private static final boolean DEFAULT_ALLOW_EMPTY_PROMPT = false;private final PromptTemplate promptTemplate;private final PromptTemplate emptyPromptTemplate;private final boolean allowEmptyContext;public CustomContextQueryAugmenter(@Nullable PromptTemplate promptTemplate,@Nullable PromptTemplate emptyPromptTemplate,@Nullable Boolean allowEmptyContext) {this.promptTemplate = promptTemplate != null ? promptTemplate : DEFAULT_PROMPT_TEMPLATE;this.emptyPromptTemplate = emptyPromptTemplate != null ? emptyPromptTemplate : DEFAULT_EMPTY_PROMPT_TEMPLATE;this.allowEmptyContext = allowEmptyContext != null ? allowEmptyContext : DEFAULT_ALLOW_EMPTY_PROMPT;logger.debug("CustomContextQueryAugmenter promptTemplate: {}", promptTemplate.getTemplate());logger.debug("CustomContextQueryAugmenter emptyPromptTemplate: {}", emptyPromptTemplate);logger.debug("CustomContextQueryAugmenter allowEmptyContext: {}", allowEmptyContext);PromptAssert.templateHasRequiredPlaceholders(this.promptTemplate, "query", "context");}@NotNull@Overridepublic Query augment(@Nullable Query query,@Nullable List<Document> documents) {Assert.notNull(query, "Query must not be null");Assert.notNull(documents, "Documents must not be null");logger.debug("Augmenting query: {}", query);if (documents.isEmpty()) {logger.debug("No documents found. Augmenting query with empty context.");return augmentQueryWhenEmptyContext(query);}logger.debug("Documents found. Augmenting query with context.");// 1. collect content from documents.AtomicInteger idCounter = new AtomicInteger(1);String documentContext = documents.stream().map(document -> {String text = document.getText();return "[[" + (idCounter.getAndIncrement()) + "]]" + text;}).collect(Collectors.joining("\n-----------------------------------------------\n"));// 2. Define prompt parameters.Map<String, Object> promptParameters = Map.of("query", query.text(),"context", documentContext);// 3. Augment user prompt with document context.return new Query(this.promptTemplate.render(promptParameters));}private Query augmentQueryWhenEmptyContext(Query query) {if (this.allowEmptyContext) {logger.debug("Empty context is allowed. Returning the original query.");return query;}logger.debug("Empty context is not allowed. Returning a specific query for empty context.");return new Query(this.emptyPromptTemplate.render());}public static final class Builder {private PromptTemplate promptTemplate;private PromptTemplate emptyPromptTemplate;private Boolean allowEmptyContext;public Builder() {}public Builder withPromptTemplate(PromptTemplate promptTemplate) {this.promptTemplate = promptTemplate;return this;}public Builder withEmptyPromptTemplate(PromptTemplate emptyPromptTemplate) {this.emptyPromptTemplate = emptyPromptTemplate;return this;}public Builder withAllowEmptyContext(Boolean allowEmptyContext) {this.allowEmptyContext = allowEmptyContext;return this;}public CustomContextQueryAugmenter build() {return new CustomContextQueryAugmenter(promptTemplate, emptyPromptTemplate, allowEmptyContext);}}}

QueryTransformer 配置 bean，用于 rewrite 用户 query：

@Beanpublic QueryTransformer queryTransformer(ChatClient.Builder chatClientBuilder,@Qualifier("transformerPromptTemplate") PromptTemplate transformerPromptTemplate) {ChatClient chatClient = chatClientBuilder.defaultOptions(DashScopeChatOptions.builder().withModel("qwen-plus").build()).build();return RewriteQueryTransformer.builder().chatClientBuilder(chatClient.mutate()).promptTemplate(transformerPromptTemplate).targetSearchSystem("Web Search").build();}

PromptTemplate  配置bean  transformerPromptTemplate

	@Beanpublic PromptTemplate transformerPromptTemplate() {return new PromptTemplate("""Given a user query, rewrite the user question to provide better results when querying {target}.You should follow these rules:1. Remove any irrelevant information and make sure the query is concise and specific;2. The output must be consistent with the language of the user's query;3. Ensure better understanding and answers from the perspective of large models.Original query:{query}Query after rewrite:""");}

QueryExpander.java 查询变体  MultiQueryExpander.java   implements QueryExpander.java

package com.alibaba.example.chatmemory.rag.preretrieval.query.expansion;import java.util.Arrays;
import java.util.List;
import java.util.Objects;
import java.util.stream.Collectors;import org.jetbrains.annotations.NotNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import org.springframework.ai.chat.client.ChatClient;
import org.springframework.ai.chat.prompt.PromptTemplate;
import org.springframework.ai.rag.Query;
import org.springframework.ai.rag.preretrieval.query.expansion.QueryExpander;
import org.springframework.ai.util.PromptAssert;
import org.springframework.lang.Nullable;
import org.springframework.util.Assert;
import org.springframework.util.CollectionUtils;
import org.springframework.util.StringUtils;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/public class MultiQueryExpander implements QueryExpander {private static final Logger logger = LoggerFactory.getLogger(MultiQueryExpander.class);private static final PromptTemplate DEFAULT_PROMPT_TEMPLATE = new PromptTemplate("""You are an expert in information retrieval and search optimization.Generate {number} different versions of a given query.Each variation should cover a different perspective or aspect of the topic while maintaining the core intent ofthe original query. The goal is to broaden your search and improve your chances of finding relevant information.Don't interpret the selection or add additional text.Query variants are provided, separated by line breaks.Original query: {query}Query variants:""");private static final Boolean DEFAULT_INCLUDE_ORIGINAL = true;private static final Integer DEFAULT_NUMBER_OF_QUERIES = 3;private final ChatClient chatClient;private final PromptTemplate promptTemplate;private final boolean includeOriginal;private final int numberOfQueries;public MultiQueryExpander(ChatClient.Builder chatClientBuilder,@Nullable PromptTemplate promptTemplate,@Nullable Boolean includeOriginal,@Nullable Integer numberOfQueries) {Assert.notNull(chatClientBuilder, "ChatClient.Builder must not be null");this.chatClient = chatClientBuilder.build();this.promptTemplate = promptTemplate == null ? DEFAULT_PROMPT_TEMPLATE : promptTemplate;this.includeOriginal = includeOriginal == null ? DEFAULT_INCLUDE_ORIGINAL : includeOriginal;this.numberOfQueries = numberOfQueries == null ? DEFAULT_NUMBER_OF_QUERIES : numberOfQueries;PromptAssert.templateHasRequiredPlaceholders(this.promptTemplate, "number", "query");}@NotNull@Overridepublic List<Query> expand(@Nullable Query query) {Assert.notNull(query, "Query must not be null");logger.debug("Generating {} queries for query: {}", this.numberOfQueries, query.text());String resp = this.chatClient.prompt().user(user -> user.text(this.promptTemplate.getTemplate()).param("number", this.numberOfQueries).param("query", query.text())).call().content();logger.debug("MultiQueryExpander#expand() Response from chat client: {}", resp);if (Objects.isNull(resp)) {logger.warn("No response from chat client for query: {}. is return.", query.text());return List.of(query);}List<String> queryVariants = Arrays.stream(resp.split("\n")).filter(StringUtils::hasText).toList();if (CollectionUtils.isEmpty(queryVariants) || this.numberOfQueries != queryVariants.size()) {logger.warn("Query expansion result dose not contain the requested {} variants for query: {}. is return.",this.numberOfQueries, query.text());return List.of(query);}List<Query> queries = queryVariants.stream().filter(StringUtils::hasText).map(queryText -> query.mutate().text(queryText).build()).collect(Collectors.toList());if (this.includeOriginal) {logger.debug("Including original query in the expanded queries for query: {}", query.text());queries.add(0, query);}logger.debug("Rewrite queries: {}", queries);return queries;}public static Builder builder() {return new Builder();}public static final class Builder {private ChatClient.Builder chatClientBuilder;private PromptTemplate promptTemplate;private Boolean includeOriginal;private Integer numberOfQueries;private Builder() {}public Builder chatClientBuilder(ChatClient.Builder chatClientBuilder) {this.chatClientBuilder = chatClientBuilder;return this;}public Builder promptTemplate(PromptTemplate promptTemplate) {this.promptTemplate = promptTemplate;return this;}public Builder includeOriginal(Boolean includeOriginal) {this.includeOriginal = includeOriginal;return this;}public Builder numberOfQueries(Integer numberOfQueries) {this.numberOfQueries = numberOfQueries;return this;}public MultiQueryExpander build() {return new MultiQueryExpander(this.chatClientBuilder, this.promptTemplate, this.includeOriginal, this.numberOfQueries);}}}

Retrieval

从不同数据源查询和用户 query 相似度最高的数据。（这里使用 Web Search）

WebSearchRetriever.java 

package com.alibaba.example.chatmemory.rag;import java.util.List;import com.alibaba.example.chatmemory.entity.websearch.GenericSearchResult;
import com.alibaba.example.chatmemory.rag.core.SearXNGSearchEngine;
import com.alibaba.example.chatmemory.rag.data.DataClean;
import org.jetbrains.annotations.NotNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import org.springframework.ai.document.Document;
import org.springframework.ai.rag.Query;
import org.springframework.ai.rag.postretrieval.ranking.DocumentRanker;
import org.springframework.ai.rag.retrieval.search.DocumentRetriever;
import org.springframework.lang.Nullable;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/public class WebSearchRetriever implements DocumentRetriever {private static final Logger logger = LoggerFactory.getLogger(WebSearchRetriever.class);//使用 searXNGprivate final SearXNGSearchEngine searchEngine;private final int maxResults;private final DataClean dataCleaner;private final DocumentRanker documentRanker;private final boolean enableRanker;private WebSearchRetriever(Builder builder) {this.searchEngine = builder.searchEngine;this.maxResults = builder.maxResults;this.dataCleaner = builder.dataCleaner;this.documentRanker = builder.documentRanker;this.enableRanker = builder.enableRanker;}@NotNull@Overridepublic List<Document> retrieve(@Nullable Query query) {// 搜索GenericSearchResult searchResp = searchEngine.search(query.text());// 清洗数据List<Document> cleanerData = dataCleaner.getData(searchResp);logger.debug("cleaner data: {}", cleanerData);// 返回结果List<Document> documents = dataCleaner.limitResults(cleanerData, maxResults);logger.debug("WebSearchRetriever#retrieve() document size: {}, raw documents: {}",documents.size(),documents.stream().map(Document::getId).toArray());return enableRanker ? ranking(query, documents) : documents;}private List<Document> ranking(Query query, List<Document> documents) {if (documents.size() == 1) {// 只有一个时，不需要 rankreturn documents;}try {List<Document> rankedDocuments = documentRanker.rank(query, documents);logger.debug("WebSearchRetriever#ranking() Ranked documents: {}", rankedDocuments.stream().map(Document::getId).toArray());return rankedDocuments;} catch (Exception e) {// 降级返回原始结果logger.error("ranking error", e);return documents;}}public static Builder builder() {return new Builder();}public static final class Builder {private SearXNGSearchEngine searchEngine;private int maxResults;private DataClean dataCleaner;private DocumentRanker documentRanker;// 默认开启 rankingprivate Boolean enableRanker = true;public Builder searchEngine(SearXNGSearchEngine searchEngine) {this.searchEngine = searchEngine;return this;}public Builder dataCleaner(DataClean dataCleaner) {this.dataCleaner = dataCleaner;return this;}public Builder maxResults(int maxResults) {this.maxResults = maxResults;return this;}public Builder documentRanker(DocumentRanker documentRanker) {this.documentRanker = documentRanker;return this;}public Builder enableRanker(Boolean enableRanker) {this.enableRanker = enableRanker;return this;}public WebSearchRetriever build() {return new WebSearchRetriever(this);}}}

 

ConcatenationDocumentJoiner.java implements  DocumentJoiner.java 合并 Document 

package com.alibaba.example.chatmemory.rag.join;import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Objects;
import java.util.Set;
import java.util.stream.Collectors;import org.jetbrains.annotations.NotNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import org.springframework.ai.document.Document;
import org.springframework.ai.rag.Query;
import org.springframework.ai.rag.retrieval.join.DocumentJoiner;
import org.springframework.lang.Nullable;
import org.springframework.util.Assert;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/public class ConcatenationDocumentJoiner implements DocumentJoiner {private static final Logger logger = LoggerFactory.getLogger(ConcatenationDocumentJoiner.class);@NotNull@Overridepublic List<Document> join(@Nullable Map<Query, List<List<Document>>> documentsForQuery) {Assert.notNull(documentsForQuery, "documentsForQuery cannot be null");Assert.noNullElements(documentsForQuery.keySet(), "documentsForQuery cannot contain null keys");Assert.noNullElements(documentsForQuery.values(), "documentsForQuery cannot contain null values");logger.debug("Joining documents by concatenation");Map<Query, List<List<Document>>> selectDocuments = selectDocuments(documentsForQuery, 10);Set<String> seen = new HashSet<>();return selectDocuments.values().stream()// Flatten List<List<Documents>> to Stream<List<Documents>..flatMap(List::stream)// Flatten Stream<List<Documents> to Stream<Documents>..flatMap(List::stream).filter(doc -> {List<String> keys = extractKeys(doc);for (String key : keys) {if (!seen.add(key)) {logger.info("Duplicate document metadata: {}",doc.getMetadata());// Duplicate keys found.return false;}}// All keys are unique.return true;}).collect(Collectors.toList());}private Map<Query, List<List<Document>>> selectDocuments(Map<Query, List<List<Document>>> documentsForQuery,int totalDocuments) {Map<Query, List<List<Document>>> selectDocumentsForQuery = new HashMap<>();int numberOfQueries = documentsForQuery.size();if (Objects.equals(0, numberOfQueries)) {return selectDocumentsForQuery;}int baseCount = totalDocuments / numberOfQueries;int remainder = totalDocuments % numberOfQueries;// To ensure consistent distribution. sort the keys (optional)List<Query> sortedQueries = new ArrayList<>(documentsForQuery.keySet());// Other sort// sortedQueries.sort(Comparator.comparing(Query::getSomeProperty));Iterator<Query> iterator = sortedQueries.iterator();for (int i = 0; i < numberOfQueries; i ++) {Query query = sortedQueries.get(i);int documentToSelect = baseCount + (i < remainder ? 1 : 0);List<List<Document>> originalDocuments = documentsForQuery.get(query);List<List<Document>> selectedNestLists = new ArrayList<>();int remainingDocuments = documentToSelect;for (List<Document> documentList : originalDocuments) {if (remainingDocuments <= 0) {break;}List<Document> selectSubList = new ArrayList<>();for (Document docs : documentList) {if (remainingDocuments <= 0) {break;}selectSubList.add(docs);remainingDocuments --;}if (!selectSubList.isEmpty()) {selectedNestLists.add(selectSubList);}}selectDocumentsForQuery.put(query, selectedNestLists);}return selectDocumentsForQuery;}private List<String> extractKeys(Document document) {List<String> keys = new ArrayList<>();if (Objects.nonNull(document)) {keys.add(document.getId());}if (Objects.nonNull(document.getMetadata())) {Object src = document.getMetadata().get("source");if (src instanceof String) {keys.add("SOURCE:" + src);}Object fn = document.getMetadata().get("file_name");if (fn instanceof String) {keys.add("FILE_NAME:" + fn);}}return keys;}}

Post-Retrieval

处理从联网搜索种获得的 Document，以获得最佳输出。

DashScopeDocumentRanker.java  implements DocumentRanker.java

package com.alibaba.example.chatmemory.rag.postretrieval;import java.util.ArrayList;
import java.util.List;
import java.util.Objects;
import java.util.Optional;import com.alibaba.cloud.ai.dashscope.rerank.DashScopeRerankOptions;
import com.alibaba.cloud.ai.model.RerankModel;
import com.alibaba.cloud.ai.model.RerankRequest;
import com.alibaba.cloud.ai.model.RerankResponse;
import com.alibaba.example.chatmemory.exception.SAAAppException1;
import org.jetbrains.annotations.NotNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import org.springframework.ai.document.Document;
import org.springframework.ai.rag.Query;
import org.springframework.ai.rag.postretrieval.ranking.DocumentRanker;
import org.springframework.lang.Nullable;
import org.springframework.util.StringUtils;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/public class DashScopeDocumentRanker implements DocumentRanker {private static final Logger logger = LoggerFactory.getLogger(DashScopeDocumentRanker.class);private final RerankModel rerankModel;public DashScopeDocumentRanker(RerankModel rerankModel) {this.rerankModel = rerankModel;}@NotNull@Overridepublic List<Document> rank(@Nullable Query query,@Nullable List<Document> documents) {if (documents == null || documents.isEmpty()) {return new ArrayList<>();}try {List<Document> reorderDocs = new ArrayList<>();// The caller controls the number of documentsDashScopeRerankOptions rerankOptions = DashScopeRerankOptions.builder().withTopN(documents.size()).build();if (Objects.nonNull(query) && StringUtils.hasText(query.text())) {// The assembly parameter calls rerankModelRerankRequest rerankRequest = new RerankRequest(query.text(),documents,rerankOptions);RerankResponse rerankResp = rerankModel.call(rerankRequest);rerankResp.getResults().forEach(res -> {Document outputDocs = res.getOutput();// Find and add to a new listOptional<Document> foundDocsOptional = documents.stream().filter(doc ->{// debug rerank output.logger.debug("DashScopeDocumentRanker#rank() doc id: {}, outputDocs id: {}", doc.getId(), outputDocs.getId());return Objects.equals(doc.getId(), outputDocs.getId());}).findFirst();foundDocsOptional.ifPresent(reorderDocs::add);});}return reorderDocs;}catch (Exception e) {// Further processing is done depending on the type of exceptionthrow new SAAAppException1(e.getMessage());}}
}

 

大模型输出

SAAWebSearchService.java

package com.alibaba.example.chatmemory.service;import com.alibaba.cloud.ai.dashscope.chat.DashScopeChatOptions;
import com.alibaba.example.chatmemory.advisor.ReasoningContentAdvisor1;
import com.alibaba.example.chatmemory.rag.WebSearchRetriever;
import com.alibaba.example.chatmemory.rag.core.SearXNGSearchEngine;
import com.alibaba.example.chatmemory.rag.data.DataClean;
import com.alibaba.example.chatmemory.rag.join.ConcatenationDocumentJoiner;
import com.alibaba.example.chatmemory.rag.prompt.CustomContextQueryAugmenter;
import reactor.core.publisher.Flux;import org.springframework.ai.chat.client.ChatClient;
import org.springframework.ai.chat.client.advisor.RetrievalAugmentationAdvisor;
import org.springframework.ai.chat.client.advisor.SimpleLoggerAdvisor;
import org.springframework.ai.chat.prompt.PromptTemplate;
import org.springframework.ai.rag.postretrieval.ranking.DocumentRanker;
import org.springframework.ai.rag.preretrieval.query.expansion.QueryExpander;
import org.springframework.ai.rag.preretrieval.query.transformation.QueryTransformer;
import org.springframework.beans.factory.annotation.Qualifier;
import org.springframework.stereotype.Service;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/@Service
public class SAAWebSearchService {private final ChatClient chatClient;private final SimpleLoggerAdvisor simpleLoggerAdvisor;private final ReasoningContentAdvisor1 reasoningContentAdvisor;private final QueryTransformer queryTransformer;private final QueryExpander queryExpander;private final PromptTemplate queryArgumentPromptTemplate;private final WebSearchRetriever webSearchRetriever;// It works better here with DeepSeek-R1private static final String DEFAULT_WEB_SEARCH_MODEL = "deepseek-r1";public SAAWebSearchService(ChatClient.Builder chatClientBuilder,QueryTransformer queryTransformer,QueryExpander queryExpander,SearXNGSearchEngine searchEngine,DataClean dataCleaner,DocumentRanker documentRanker,@Qualifier("queryArgumentPromptTemplate") PromptTemplate queryArgumentPromptTemplate) {this.queryTransformer = queryTransformer;this.queryExpander = queryExpander;this.queryArgumentPromptTemplate = queryArgumentPromptTemplate;// reasoning content for DeepSeek-r1 is integrated into the outputthis.reasoningContentAdvisor = new ReasoningContentAdvisor1(1);// Build chatClientthis.chatClient = chatClientBuilder.defaultOptions(DashScopeChatOptions.builder().withModel(DEFAULT_WEB_SEARCH_MODEL)// stream 模式下是否开启增量输出.withIncrementalOutput(true).build()).build();// 日志this.simpleLoggerAdvisor = new SimpleLoggerAdvisor(100);this.webSearchRetriever = WebSearchRetriever.builder().searchEngine(searchEngine).dataCleaner(dataCleaner).maxResults(2).enableRanker(true).documentRanker(documentRanker).build();}//Handle user inputpublic Flux<String> chat(String prompt) {return chatClient.prompt().advisors(createRetrievalAugmentationAdvisor(),reasoningContentAdvisor,simpleLoggerAdvisor).user(prompt).stream().content();}private RetrievalAugmentationAdvisor createRetrievalAugmentationAdvisor() {return RetrievalAugmentationAdvisor.builder().documentRetriever(webSearchRetriever).queryTransformers(queryTransformer).queryAugmenter(new CustomContextQueryAugmenter(queryArgumentPromptTemplate,null,true)).queryExpander(queryExpander).documentJoiner(new ConcatenationDocumentJoiner()).build();}}

PromptTemplate  配置bean  queryArgumentPromptTemplate

@Beanpublic PromptTemplate queryArgumentPromptTemplate() {return new PromptTemplate("""You'll get a set of document contexts that are relevant to the issue.Each document begins with a reference number, such as [[x]], where x is a number that can be repeated.Documents that are not referenced will be marked as [[null]].Use context and refer to it at the end of each sentence, if applicable.The context information is as follows:---------------------{context}---------------------Generate structured responses to user questions given contextual information and without prior knowledge.When you answer user questions, follow these rules:1. If the answer is not in context, say you don't know;2. Don't provide any information that is not relevant to the question, and don't output any duplicate content;3. Avoid using "context-based..." or "The provided information..." said;4. Your answers must be correct, accurate, and written in an expertly unbiased and professional tone;5. The appropriate text structure in the answer is determined according to the characteristics of the content, please include subheadings in the output to improve readability;6. When generating a response, provide a clear conclusion or main idea first, without a title;7. Make sure each section has a clear subtitle so that users can better understand and refer to your output;8. If the information is complex or contains multiple sections, make sure each section has an appropriate heading to create a hierarchical structure;9. Please refer to the sentence or section with the reference number at the end in [[x]] format;10. If a sentence or section comes from more than one context, list all applicable references, e.g. [[x]][[y]];11. Your output answers must be in beautiful and rigorous markdown format.12. Because your output is in markdown format, please include the link in the reference document in the form of a hyperlink when referencing the context, so that users can click to view it;13. If a reference is marked as [[null]], it does not have to be cited;14. Except for Code. Aside from the specific name and citation, your answer must be written in the same language as the question.User Issue:{query}Your answer:""");}

异常类

/** Licensed to the Apache Software Foundation (ASF) under one or more* contributor license agreements.  See the NOTICE file distributed with* this work for additional information regarding copyright ownership.* The ASF licenses this file to You under the Apache License, Version 2.0* (the "License"); you may not use this file except in compliance with* the License.  You may obtain a copy of the License at**     http://www.apache.org/licenses/LICENSE-2.0** Unless required by applicable law or agreed to in writing, software* distributed under the License is distributed on an "AS IS" BASIS,* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.* See the License for the specific language governing permissions and* limitations under the License.*/package com.alibaba.example.chatmemory.exception;import org.slf4j.Logger;
import org.slf4j.LoggerFactory;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>*/public class SAAAppException1 extends RuntimeException {private final static Logger logger = LoggerFactory.getLogger(SAAAppException1.class);public SAAAppException1(String msg) {super(msg);logger.error("Spring AI Playground app exception: {}", msg);}}

searXNG 搜索结果转 java类

package com.alibaba.example.chatmemory.entity.websearch;import java.util.List;
import java.util.Map;import com.aliyun.tea.NameInMap;
import com.aliyun.tea.TeaModel;public class GenericSearchResult extends TeaModel {@NameInMap("results")public List<ScorePageItem> results;@NameInMap("query")public String query;@NameInMap("number_of_results")public Integer numberOfResults;public GenericSearchResult() {}public static GenericSearchResult build(Map<String, ?> map) throws Exception {GenericSearchResult self = new GenericSearchResult();return (GenericSearchResult) TeaModel.build(map, self);}public List<ScorePageItem> getResults() {return results;}public GenericSearchResult setResults(List<ScorePageItem> results) {this.results = results;return this;}public String getQuery() {return query;}public GenericSearchResult setQuery(String query) {this.query = query;return this;}public Integer getNumberOfResults() {return numberOfResults;}public GenericSearchResult setNumberOfResults(Integer numberOfResults) {this.numberOfResults = numberOfResults;return this;}
}

package com.alibaba.example.chatmemory.entity.websearch;import com.aliyun.tea.NameInMap;
import com.aliyun.tea.TeaModel;public class ScorePageItem extends TeaModel {/*** <p>This parameter is required.</p>** <strong>example:</strong>* <p>structure_web_info</p>*/@NameInMap("engine")public String engine;/*** <p>This parameter is required.</p>** <strong>example:</strong>* <p>baijiahao.baidu.com</p>*/@NameInMap("template")public String template;/*** <strong>example:</strong>* <p><a href="https://s2.zimgs.cn/ims?kt=url&at=smstruct&key=aHR0cHM6Ly9ndy5hbGljZG4uY29tL0wxLzcyMy8xNTY1MjU2NjAwLzJhL2YwL2I0LzJhZjBiNDQxMGI5YmVlMDVjOGVlNGJmODk3MTNkNTFjLnBuZw==&sign=yx:CUlNNQVJQjFrk3Kxt2F3KWhTOFU=&tv=400_400">https://s2.zimgs.cn/ims?kt=url&amp;at=smstruct&amp;key=aHR0cHM6Ly9ndy5hbGljZG4uY29tL0wxLzcyMy8xNTY1MjU2NjAwLzJhL2YwL2I0LzJhZjBiNDQxMGI5YmVlMDVjOGVlNGJmODk3MTNkNTFjLnBuZw==&amp;sign=yx:CUlNNQVJQjFrk3Kxt2F3KWhTOFU=&amp;tv=400_400</a></p>*/@NameInMap("category")public String category;/*** <p>This parameter is required.</p>** <strong>example:</strong>* <p><a href="https://baijiahao.baidu.com/s?id=1787881554557805096">https://baijiahao.baidu.com/s?id=1787881554557805096</a></p>*/@NameInMap("url")public String url;/*** <strong>example:</strong>* <p>昨天	，	小米	汽车	没有	发布	，	但	相关	的	信息	透露	的	差	不	多	了	。*         在	发布	会	现场	，	雷军	直接	称	小米	S	U	7	对	标	特斯拉	保时捷	，	有	100	项	行业	领先	，	可见	“	遥遥	领先	”	已经	不再	是	华为	专利	了	？		</p>* <pre><code>    而	在	介绍	技术	时	，	雷军	也	从	电机	、	电池	、	大	压铸	、	自动	驾驶	、	智能	座舱	等	五	大	方面	展开	，	充分	展示	了	小米	汽车	的	技术	以及	技术	储备	，		 		能	堆	的	料	，	全都	堆	上去	了	…	…*     大家	比较	感	兴趣	的	性能	方面	，	2	.	78	s	的	0	-	100	km	/	h	加速	，	265	km	/	h	的	最高	时速* </code></pre>*/@NameInMap("content")public String content;/*** <p>This parameter is required.</p>** <strong>example:</strong>* <p>1704426524000</p>*/@NameInMap("publishedDate")public String  publishedDate;/*** <strong>example:</strong>* <p>0.234325235</p>*/@NameInMap("score")public Double score;/*** <p>This parameter is required.</p>** <strong>example:</strong>* <p>小米SU7售价22.99万元起 高管亲自辟谣：发布前不会有价格</p>*/@NameInMap("title")public String title;public static ScorePageItem build(java.util.Map<String, ?> map) throws Exception {ScorePageItem self = new ScorePageItem();return TeaModel.build(map, self);}public String getEngine() {return engine;}public ScorePageItem setEngine(String engine) {this.engine = engine;return this;}public String getTemplate() {return template;}public ScorePageItem setTemplate(String template) {this.template = template;return this;}public String getCategory() {return category;}public ScorePageItem setCategory(String category) {this.category = category;return this;}public String getUrl() {return url;}public ScorePageItem setUrl(String url) {this.url = url;return this;}public String getContent() {return content;}public ScorePageItem setContent(String content) {this.content = content;return this;}public String getPublishedDate() {return publishedDate;}public ScorePageItem setPublishedDate(String publishedDate) {this.publishedDate = publishedDate;return this;}public Double getScore() {return score;}public ScorePageItem setScore(Double score) {this.score = score;return this;}public String getTitle() {return title;}public ScorePageItem setTitle(String title) {this.title = title;return this;}
}

其他bean配置

	@Beanpublic QueryExpander queryExpander(ChatClient.Builder chatClientBuilder) {ChatClient chatClient = chatClientBuilder.defaultOptions(DashScopeChatOptions.builder().withModel("qwen-plus").build()).build();return MultiQueryExpander.builder().chatClientBuilder(chatClient.mutate()).numberOfQueries(2).build();}@Beanpublic SimpleLoggerAdvisor simpleLoggerAdvisor() {return new SimpleLoggerAdvisor();}@Beanpublic DashScopeDocumentRanker dashScopeDocumentRanker(RerankModel rerankModel) {return new DashScopeDocumentRanker(rerankModel);}

controller

package com.alibaba.example.chatmemory.controller;import com.alibaba.example.chatmemory.service.SAAWebSearchService;
import io.swagger.v3.oas.annotations.tags.Tag;
import jakarta.servlet.http.HttpServletResponse;
import reactor.core.publisher.Flux;import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RequestParam;
import org.springframework.web.bind.annotation.RestController;/*** @author yuluo* @author <a href="mailto:yuluo08290126@gmail.com">yuluo</a>* <p>* The deepseek-r1 model is used by default, which works better.*/@RestController
@Tag(name = "Web Search APIs")
@RequestMapping("/api/v1/")
public class SAAWebSearchController1 {private final SAAWebSearchService webSearch;public SAAWebSearchController1(SAAWebSearchService webSearch) {this.webSearch = webSearch;}@GetMapping("/search")public Flux<String> search(@RequestParam(value = "query") String query,HttpServletResponse response) {response.setCharacterEncoding("UTF-8");return webSearch.chat(query);}}

不联网测试如下

联网测试如下