Build a Spring Boot 3 Web QA App with Vector Store and AI in Minutes

1. Introduction

This article demonstrates how to implement a web‑page question‑answering service using Spring Boot 3.4.2 and Spring AI. A given URL is loaded, its content stored in a vector database, and queries are answered by retrieving relevant chunks and sending them to a large language model.

2. Practical Example

2.1 Prepare Environment

Add the required Maven dependencies:

<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-advisors-vector-store</artifactId>
</dependency>

<!-- Ollama BGE‑M3 model for embeddings -->
<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-model-ollama</artifactId>
</dependency>

<!-- Milvus vector store -->
<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-vector-store-milvus</artifactId>
</dependency>

<!-- Alibaba DashScope (optional) -->
<dependency>
  <groupId>com.alibaba.cloud.ai</groupId>
  <artifactId>spring-ai-alibaba-autoconfigure-dashscope</artifactId>
  <version>1.0.0.2</version>
</dependency>

<!-- Jsoup for web page parsing -->
<dependency>
  <groupId>org.jsoup</groupId>
  <artifactId>jsoup</artifactId>
  <version>1.18.3</version>
</dependency>

Configure Spring AI properties (YAML style):

spring:
  ai:
    dashscope:
      api-key: sk-xxxooo
      base-url: https://dashscope.aliyuncs.com/compatible-mode/v1
      chat:
        options:
          model: qwen-turbo
      embedding:
        enabled: false
---
spring:
  ai:
    ollama:
      base-url: http://localhost:11111
      embedding:
        model: bge-m3:latest
---
spring:
  ai:
    vectorstore:
      milvus:
        client:
          host: localhost
          port: 19530
          username: xxx
          password: ooo
        initialize-schema: true
        databaseName: "mydb"
        collectionName: "mydocs"
        autoId: true
        id-field-name: id
        embeddingDimension: 1024

Exclude conflicting auto‑configurations:

@SpringBootApplication(
    exclude = {
        OllamaChatAutoConfiguration.class,
        DashScopeEmbeddingAutoConfiguration.class
    }
)
public class SpringBootAiWebQaApplication {}

2.2 Document Reader

Implement DocumentReader using Jsoup to fetch a page and convert it to a Spring AI Document:

public class WebDocumentReader implements DocumentReader {
    private static final Logger logger = LoggerFactory.getLogger(WebDocumentReader.class);
    private final String url;
    public WebDocumentReader(String url) { this.url = url; }
    @Override
    public List<Document> get() {
        try {
            X509ExtendedTrustManager trustManager = createTrustManager();
            SSLContext sslContext = SSLContext.getInstance("TLS");
            sslContext.init(null, new TrustManager[]{trustManager}, new SecureRandom());
            org.jsoup.nodes.Document doc = Jsoup.connect(url)
                .sslSocketFactory(sslContext.getSocketFactory())
                .get();
            Map<String, Object> metadata = Map.of(
                "source", url,
                "date", DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss").format(LocalDateTime.now())
            );
            return List.of(new Document(doc.body().text(), metadata));
        } catch (Exception e) {
            logger.error("Failed to load [{}]: {}", this.url, e);
            return List.of();
        }
    }
    private X509ExtendedTrustManager createTrustManager() {
        return new X509ExtendedTrustManager() {
            public X509Certificate[] getAcceptedIssuers() { return new X509Certificate[0]; }
            // other methods omitted for brevity
        };
    }
}

2.3 Write Documents to Vector Store

@Service
public class WebDocumentService {
    private final VectorStore vectorStore;
    public WebDocumentService(VectorStore vectorStore) { this.vectorStore = vectorStore; }
    public void save(String url) {
        WebDocumentReader reader = new WebDocumentReader(url);
        List<Document> docs = reader.get();
        TokenTextSplitter splitter = new TokenTextSplitter(1000, 400, 10, 5000, true);
        List<Document> chunks = splitter.apply(docs);
        this.vectorStore.add(chunks);
    }
}

Spring AI provides several text splitters, e.g., TokenTextSplitter, ContentFormatTransformer, KeywordMetadataEnricher, and SummaryMetadataEnricher.

2.4 Configure ChatClient

@Configuration
public class ChatConfig {
    @Bean
    ChatClient chatClient(ChatClient.Builder builder, VectorStore vectorStore) {
        PromptTemplate promptTemplate = new PromptTemplate("""
            Context:
            ---------------------
            {context}
            ---------------------
            Answer the query without prior knowledge.
            Rules:
            1. If the answer is not in the context, reply \"I don't know\".
            2. Avoid phrases like \"According to the context...\".
            Query: {query}
            Answer:
            """);
        QueryAugmenter queryAugmenter = ContextualQueryAugmenter.builder()
            .promptTemplate(promptTemplate)
            .build();
        Advisor retrievalAdvisor = RetrievalAugmentationAdvisor.builder()
            .queryAugmenter(queryAugmenter)
            .documentRetriever(VectorStoreDocumentRetriever.builder()
                .similarityThreshold(0.40)
                .topK(1)
                .vectorStore(vectorStore)
                .build())
            .build();
        List<Advisor> advisors = List.of(new SimpleLoggerAdvisor(), retrievalAdvisor);
        return builder.defaultAdvisors(advisors).build();
    }
}

The RetrievalAugmentationAdvisor performs similarity search in the vector store and feeds the retrieved text together with the user query to the LLM.

2.5 Controller Endpoints

@RestController
@RequestMapping("/web")
public class WebController {
    private final WebDocumentService webDocumentService;
    public WebController(WebDocumentService webDocumentService) { this.webDocumentService = webDocumentService; }
    @GetMapping("/load")
    public ResponseEntity<?> load(String url) {
        this.webDocumentService.save(url);
        return ResponseEntity.ok("success");
    }
}

@RestController
@RequestMapping("/chat")
public class ChatController {
    private final ChatClient chatClient;
    public ChatController(ChatClient chatClient) { this.chatClient = chatClient; }
    @GetMapping
    public ResponseEntity<?> chat(String prompt) {
        String answer = this.chatClient.prompt(prompt).call().content();
        return ResponseEntity.ok(answer);
    }
}

2.6 Testing

1. Use the /web/load endpoint to store a web page (e.g., a CCTV news article) into Milvus. 2. Query the /chat endpoint with a question; the system retrieves the most relevant chunk and returns the AI‑generated answer. The console shows successful storage and retrieval.

All steps above constitute the complete tutorial for building a Spring Boot 3 web‑QA application powered by Spring AI and Milvus.