Spring AI 高级RAG功能实现：查询翻译、上下文感知

src/main/java/com/huangge1199/aiagent/Service/RagService.java

@@ -16,4 +16,8 @@ public interface RagService {
     List<Query> getMultiQueryExpand(String question);
 
     String queryRewrite(String question);
+
+    String queryTranslation(String question);
+
+    String contextAwareQueries(String question);
 }

src/main/java/com/huangge1199/aiagent/Service/impl/RagServiceImpl.java

@@ -6,11 +6,14 @@ import com.huangge1199.aiagent.rag.MyMultiQueryExpander;
 import jakarta.annotation.Resource;
 import org.springframework.ai.chat.client.ChatClient;
 import org.springframework.ai.chat.client.advisor.QuestionAnswerAdvisor;
+import org.springframework.ai.chat.messages.AssistantMessage;
+import org.springframework.ai.chat.messages.UserMessage;
 import org.springframework.ai.chat.model.ChatModel;
 import org.springframework.ai.rag.Query;
-import org.springframework.ai.rag.preretrieval.query.expansion.MultiQueryExpander;
+import org.springframework.ai.rag.preretrieval.query.transformation.CompressionQueryTransformer;
 import org.springframework.ai.rag.preretrieval.query.transformation.QueryTransformer;
 import org.springframework.ai.rag.preretrieval.query.transformation.RewriteQueryTransformer;
+import org.springframework.ai.rag.preretrieval.query.transformation.TranslationQueryTransformer;
 import org.springframework.ai.vectorstore.VectorStore;
 import org.springframework.stereotype.Service;
 
@@ -36,35 +39,21 @@ public class RagServiceImpl implements RagService {
 
     @Override
     public String localDoc(String question) {
-        return chatClient.prompt()
+        return chatClient.prompt().user(question).advisors(new MyLoggerAdvisor()).advisors(new QuestionAnswerAdvisor(vectorStore)).call().content();
-                .user(question)
-                .advisors(new MyLoggerAdvisor())
-                .advisors(new QuestionAnswerAdvisor(vectorStore))
-                .call()
-                .content();
     }
 
     @Override
     public List<Query> getMultiQueryExpand(String question) {
         ChatClient.Builder builder = ChatClient.builder(ollamaChatModel);
 
-        ChatClient chatClient = builder
+        ChatClient chatClient = builder.defaultSystem("你是一位专业的室内设计顾问，精通各种装修风格、材料选择和空间布局。请基于提供的参考资料，为用户提供专业、详细且实用的建议。在回答时，请注意：\n" + "1. 准确理解用户的具体需求\n" + "2. 结合参考资料中的实际案例\n" + "3. 提供专业的设计理念和原理解释\n" + "4. 考虑实用性、美观性和成本效益\n" + "5. 如有需要，可以提供替代方案").build();
-                .defaultSystem("你是一位专业的室内设计顾问，精通各种装修风格、材料选择和空间布局。请基于提供的参考资料，为用户提供专业、详细且实用的建议。在回答时，请注意：\n" +
-                        "1. 准确理解用户的具体需求\n" +
-                        "2. 结合参考资料中的实际案例\n" +
-                        "3. 提供专业的设计理念和原理解释\n" +
-                        "4. 考虑实用性、美观性和成本效益\n" +
-                        "5. 如有需要，可以提供替代方案")
-                .build();
 
 //        MultiQueryExpander queryExpander = MultiQueryExpander.builder()
-        MyMultiQueryExpander queryExpander = MyMultiQueryExpander.builder()
+        MyMultiQueryExpander queryExpander = MyMultiQueryExpander.builder().chatClientBuilder(builder)
-                .chatClientBuilder(builder)
                 // 不包含原始查询
                 .includeOriginal(false)
                 // 生成3个查询变体
-                .numberOfQueries(4)
+                .numberOfQueries(4).build();
-                .build();
         return queryExpander.expand(new Query(question));
     }
 
@@ -75,12 +64,53 @@ public class RagServiceImpl implements RagService {
         Query query = new Query(question);
 
         // 创建查询重写转换器
-        QueryTransformer queryTransformer = RewriteQueryTransformer.builder()
+        QueryTransformer queryTransformer = RewriteQueryTransformer.builder().chatClientBuilder(builder).build();
-                .chatClientBuilder(builder)
-                .build();
 
         // 执行查询重写
         Query transformedQuery = queryTransformer.transform(query);
         return transformedQuery.text();
     }
+
+    @Override
+    public String queryTranslation(String question) {
+        ChatClient.Builder builder = ChatClient.builder(ollamaChatModel);
+        // 创建一个英文查询
+        Query query = new Query(question);
+
+        // 创建查询翻译转换器，设置目标语言为中文
+        QueryTransformer queryTransformer = TranslationQueryTransformer.builder().chatClientBuilder(builder)
+                // 设置目标语言为中文
+                .targetLanguage("chinese").build();
+
+        // 执行查询翻译
+        Query transformedQuery = queryTransformer.transform(query);
+
+        // 输出翻译后的查询
+        return transformedQuery.text();
+    }
+
+    @Override
+    public String contextAwareQueries(String question) {
+        ChatClient.Builder builder = ChatClient.builder(ollamaChatModel);
+        // 构建带有历史上下文的查询
+        // 这个例子模拟了一个房地产咨询场景，用户先问小区位置，再问房价
+        Query query = Query.builder()
+                // 当前用户的提问
+                .text(question)
+                // 历史对话中用户的问题
+                .history(new UserMessage("深圳市南山区的碧海湾小区在哪里?"),
+                        // AI的回答
+                        new AssistantMessage("碧海湾小区位于深圳市南山区后海中心区，临近后海地铁站。")).build();
+        // 创建查询转换器
+        // QueryTransformer用于将带有上下文的查询转换为完整的独立查询
+        QueryTransformer queryTransformer = CompressionQueryTransformer.builder()
+                .chatClientBuilder(builder)
+                .build();
+
+        // 执行查询转换
+        // 将模糊的代词引用（"这个小区"）转换为明确的实体名称（"碧海湾小区"）
+        Query transformedQuery = queryTransformer.transform(query);
+        return transformedQuery.text();
+
+    }
 }

src/main/java/com/huangge1199/aiagent/controller/RagController.java

@@ -47,4 +47,18 @@ public class RagController {
         String queryList = ragService.queryRewrite(question);
         return R.ok(queryList);
     }
+
+    @PostMapping("/queryTranslation")
+    @Operation(summary = "查询翻译")
+    public R<String> queryTranslation(@RequestBody String question) {
+        String queryList = ragService.queryTranslation(question);
+        return R.ok(queryList);
+    }
+
+    @PostMapping("/contextAwareQueries")
+    @Operation(summary = "上下文感知查询")
+    public R<String> contextAwareQueries(@RequestBody String question) {
+        String queryList = ragService.contextAwareQueries(question);
+        return R.ok(queryList);
+    }
 }