Spring AI 企业RAG进阶|文档智能分片调优 + ES混合检索 + 接口限流熔断监控

javaimport org.springframework.ai.document.Document;import org.springframework.ai.transformer.splitter.TokenTextSplitter;import java.util.List;/*** RAG 智能文档分片工具* 支持动态块大小、动态重叠度，解决语义断裂、重复冗余问题*/public class RagDocSplitUtil {/*** 自定义分片规则* @param originDocs 原始文档集合* @param chunkSize 单块最大Token数* @param chunkOverlap 分片重叠Token数* @return 分片后文档*/public static List<Document> smartSplit(List<Document> originDocs, int chunkSize, int chunkOverlap) {TokenTextSplitter splitter = new TokenTextSplitter(chunkSize,chunkOverlap,20, // 最小分片长度，过滤无效短片段2000, // 最大分片上限，防止超长文本true);return splitter.apply(originDocs);}}

java// 长文本合同文档分片List<Document> ruleDocs = RagDocSplitUtil.smartSplit(rawDocs, 800, 120);// FAQ短文档分片List<Document> faqDocs = RagDocSplitUtil.smartSplit(rawDocs, 450, 80);// 入库vectorStore.add(ruleDocs);

xml<!– Spring AI ES 向量库 –><dependency><groupId>org.springframework.ai</groupId><artifactId>spring-ai-starter-vector-store-elasticsearch</artifactId><version>1.0.0</version></dependency><!– SpringBoot ES 客户端 –><dependency><groupId>org.springframework.boot</groupId><artifactId>spring-boot-starter-data-elasticsearch</artifactId></dependency>

yamlspring:elasticsearch:uris: http://127.0.0.1:9200ai:openai:api-key: sk-xxxchat:model: gpt-3.5-turbovectorstore:elasticsearch:index-name: enterprise_rag_knowledge# 向量维度适配模型dimensions: 1536similarity: cosine

javaimport org.springframework.ai.chat.client.ChatClient;import org.springframework.ai.document.Document;import org.springframework.ai.vectorstore.ElasticsearchVectorStore;import org.springframework.ai.vectorstore.SearchRequest;import org.springframework.web.bind.annotation.GetMapping;import org.springframework.web.bind.annotation.RequestParam;import org.springframework.web.bind.annotation.RestController;import java.util.List;import java.util.stream.Collectors;/*** Spring AI + Elasticsearch 双引擎混合检索* 全文检索 + 向量语义检索 融合方案*/@RestControllerpublic class EsHybridRagController {private final ChatClient chatClient;private final ElasticsearchVectorStore elasticsearchVectorStore;// 权重配比：语义0.6 关键词0.4private static final int TOP_K = 4;public EsHybridRagController(ChatClient.Builder builder, ElasticsearchVectorStore elasticsearchVectorStore) {this.chatClient = builder.build();this.elasticsearchVectorStore = elasticsearchVectorStore;}@GetMapping(“/rag/es/hybrid/chat”)public String hybridChat(@RequestParam String question) {// 构建检索条件：ES自动支持 向量相似度 + 文本匹配打分SearchRequest searchRequest = SearchRequest.query(question).withTopK(TOP_K).withSimilarityThreshold(0.7d);// 召回高分文档List<Document> documents = elasticsearchVectorStore.similaritySearch(searchRequest);String context = documents.stream().map(Document::getText).collect(Collectors.joining(“\n=====\n”));// 组装Prompt回答return chatClient.prompt().user(“””请严格基于下方知识库内容回答问题，禁止编造幻觉信息，无匹配内容如实告知用户。知识库：{context}用户问题：{question}“””).param(“context”, context).param(“question”, question).call().getContent();}}

xml<!– 阿里Sentinel 限流熔断 –><dependency><groupId>com.alibaba.cloud</groupId><artifactId>spring-cloud-starter-alibaba-sentinel</artifactId><version>2023.0.1.2</version></dependency>

javaimport com.alibaba.csp.sentinel.annotation.SentinelResource;import com.alibaba.csp.sentinel.slots.block.BlockException;import org.springframework.ai.chat.client.ChatClient;import org.springframework.web.bind.annotation.GetMapping;import org.springframework.web.bind.annotation.RequestParam;import org.springframework.web.bind.annotation.RestController;/*** 带限流、熔断、降级的AI问答接口*/@RestControllerpublic class AiSafeRagController {private final ChatClient chatClient;public AiSafeRagController(ChatClient chatClient) {this.chatClient = chatClient;}@GetMapping(“/rag/safe/chat”)@SentinelResource(value = “spring-ai-rag-chat”,blockHandler = “blockHandlerLimit”,fallback = “fallbackHandlerError”)public String safeChat(@RequestParam String question) {// 正常AI问答逻辑return chatClient.prompt().user(question).call().getContent();}/*** 限流兜底：访问频繁触发*/public String blockHandlerLimit(String question, BlockException e) {return “系统访问繁忙，请稍后再试！”;}/*** 熔断降级：接口异常、超时、报错触发*/public String fallbackHandlerError(String question, Throwable e) {return “AI服务暂时异常，已触发熔断保护，请稍后重试。”;}}