293 lines
13 KiB
Java
293 lines
13 KiB
Java
package com.huangge1199.aiagent.Service.impl;
|
||
|
||
import com.huangge1199.aiagent.Service.RagService;
|
||
import com.huangge1199.aiagent.config.MyLoggerAdvisor;
|
||
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.client.advisor.RetrievalAugmentationAdvisor;
|
||
import org.springframework.ai.chat.client.advisor.api.Advisor;
|
||
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.document.Document;
|
||
import org.springframework.ai.embedding.EmbeddingModel;
|
||
import org.springframework.ai.rag.Query;
|
||
import org.springframework.ai.rag.generation.augmentation.ContextualQueryAugmenter;
|
||
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.rag.retrieval.search.DocumentRetriever;
|
||
import org.springframework.ai.rag.retrieval.search.VectorStoreDocumentRetriever;
|
||
import org.springframework.ai.vectorstore.VectorStore;
|
||
import org.springframework.ai.vectorstore.SimpleVectorStore;
|
||
import org.springframework.ai.vectorstore.filter.FilterExpressionBuilder;
|
||
import org.springframework.stereotype.Service;
|
||
|
||
import java.util.ArrayList;
|
||
import java.util.List;
|
||
import java.util.Map;
|
||
|
||
/**
|
||
* RagServiceImpl
|
||
*
|
||
* @author huangge1199
|
||
* @since 2025/5/24 9:21:38
|
||
*/
|
||
@Service
|
||
public class RagServiceImpl implements RagService {
|
||
|
||
@Resource
|
||
private ChatClient chatClient;
|
||
|
||
@Resource
|
||
private VectorStore vectorStore;
|
||
|
||
@Resource
|
||
private ChatModel ollamaChatModel;
|
||
|
||
@Resource
|
||
private EmbeddingModel ollamaEmbeddingModel;
|
||
|
||
@Override
|
||
public String localDoc(String question) {
|
||
return chatClient.prompt().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.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().chatClientBuilder(builder)
|
||
// 不包含原始查询
|
||
.includeOriginal(false)
|
||
// 生成3个查询变体
|
||
.numberOfQueries(4).build();
|
||
return queryExpander.expand(new Query(question));
|
||
}
|
||
|
||
@Override
|
||
public String queryRewrite(String question) {
|
||
ChatClient.Builder builder = ChatClient.builder(ollamaChatModel);
|
||
// 创建一个模拟用户学习AI的查询场景
|
||
Query query = new Query(question);
|
||
|
||
// 创建查询重写转换器
|
||
QueryTransformer queryTransformer = RewriteQueryTransformer.builder().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();
|
||
|
||
}
|
||
|
||
@Override
|
||
public String baseAdvisor(String question) {
|
||
Advisor advisor = RetrievalAugmentationAdvisor.builder()
|
||
.documentRetriever(VectorStoreDocumentRetriever.builder()
|
||
.vectorStore(vectorStore)
|
||
.build())
|
||
.build();
|
||
return advisor(question, advisor);
|
||
}
|
||
|
||
@Override
|
||
public String advancedAdvisor(String question) {
|
||
Advisor advisor = RetrievalAugmentationAdvisor.builder()
|
||
// 配置查询增强器
|
||
.queryAugmenter(ContextualQueryAugmenter.builder()
|
||
// 允许空上下文查询
|
||
.allowEmptyContext(true)
|
||
.build())
|
||
// 配置文档检索器
|
||
.documentRetriever(VectorStoreDocumentRetriever.builder()
|
||
.vectorStore(vectorStore)
|
||
// 相似度阈值
|
||
.similarityThreshold(0.5)
|
||
// 返回文档数量
|
||
.topK(3)
|
||
.filterExpression(new FilterExpressionBuilder()
|
||
.eq("genre", "fairytale")
|
||
.build()) // 文档过滤表达式
|
||
.build())
|
||
.build();
|
||
return advisor(question, advisor);
|
||
}
|
||
|
||
/**
|
||
* 检索增强顾问
|
||
*
|
||
* @param question 问题
|
||
* @param advisor 检索增强顾问
|
||
* @return 查询结果
|
||
*/
|
||
private String advisor(String question, Advisor advisor) {
|
||
// 1. 初始化向量存储
|
||
SimpleVectorStore vectorStore = SimpleVectorStore.builder(ollamaEmbeddingModel)
|
||
.build();
|
||
|
||
// 2. 添加文档到向量存储
|
||
List<Document> documents = List.of(
|
||
new Document("产品说明书:产品名称：智能机器人\n" +
|
||
"产品描述：智能机器人是一个智能设备，能够自动完成各种任务。\n" +
|
||
"功能：\n" +
|
||
"1. 自动导航：机器人能够自动导航到指定位置。\n" +
|
||
"2. 自动抓取：机器人能够自动抓取物品。\n" +
|
||
"3. 自动放置：机器人能够自动放置物品。\n"));
|
||
vectorStore.add(documents);
|
||
|
||
// 3. 创建检索增强顾问 advisor
|
||
|
||
// 4. 在聊天客户端中使用顾问
|
||
return chatClient.prompt()
|
||
.user(question)
|
||
// 添加检索增强顾问
|
||
.advisors(advisor)
|
||
.call()
|
||
.content();
|
||
}
|
||
|
||
@Override
|
||
public String documentSelection() {
|
||
ChatClient.Builder builder = ChatClient.builder(ollamaChatModel);
|
||
// 生成室内设计案例文档
|
||
List<Document> documents = new ArrayList<>();
|
||
|
||
// 现代简约风格客厅案例
|
||
documents.add(new Document(
|
||
"案例编号：LR-2023-001\n" +
|
||
"项目概述：180平米大平层现代简约风格客厅改造\n" +
|
||
"设计要点：\n" +
|
||
"1. 采用5.2米挑高的落地窗，最大化自然采光\n" +
|
||
"2. 主色调：云雾白(哑光，NCS S0500-N)配合莫兰迪灰\n" +
|
||
"3. 家具选择：意大利B&B品牌真皮沙发，北欧白橡木茶几\n" +
|
||
"4. 照明设计：嵌入式筒灯搭配意大利Flos吊灯\n" +
|
||
"5. 软装配饰：进口黑胡桃木电视墙，几何图案地毯\n" +
|
||
"空间效果：通透大气，适合商务接待和家庭日常起居",
|
||
Map.of(
|
||
// 文档类型
|
||
"type", "interior",
|
||
// 年份
|
||
"year", "2023",
|
||
// 月份
|
||
"month", "06",
|
||
// 位置类型
|
||
"location", "indoor",
|
||
// 装修风格
|
||
"style", "modern",
|
||
// 房间类型
|
||
"room", "living_room"
|
||
)
|
||
));
|
||
// 1. 初始化向量存储
|
||
SimpleVectorStore vectorStore = SimpleVectorStore.builder(ollamaEmbeddingModel)
|
||
.build();
|
||
|
||
// 2. 配置AI助手角色
|
||
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();
|
||
|
||
// 3. 构建复杂的文档过滤条件
|
||
var filterExpression = getOp();
|
||
|
||
// 4. 配置文档检索器
|
||
DocumentRetriever retriever = VectorStoreDocumentRetriever.builder()
|
||
.vectorStore(vectorStore)
|
||
// 设置相似度阈值
|
||
.similarityThreshold(0.5)
|
||
// 返回前3个最相关的文档
|
||
.topK(3)
|
||
.filterExpression(filterExpression.build())
|
||
.build();
|
||
|
||
// 5. 创建上下文感知的查询增强器
|
||
Advisor advisor = RetrievalAugmentationAdvisor.builder()
|
||
.queryAugmenter(ContextualQueryAugmenter.builder()
|
||
.allowEmptyContext(true)
|
||
.build())
|
||
.documentRetriever(retriever)
|
||
.build();
|
||
|
||
// 6. 执行查询并获取响应
|
||
String userQuestion = "根据已经提供的资料，请描述所有相关的场景风格，输出案例编号，尽可能详细地描述其内容。";
|
||
return chatClient.prompt()
|
||
.user(userQuestion)
|
||
.advisors(advisor)
|
||
.call()
|
||
.content();
|
||
}
|
||
|
||
private static FilterExpressionBuilder.Op getOp() {
|
||
var b = new FilterExpressionBuilder();
|
||
// 筛选2023年的案例
|
||
// 仅选择室内案例
|
||
// 类型为室内设计
|
||
// 指定房间类型
|
||
return b.and(
|
||
b.and(
|
||
// 筛选2023年的案例
|
||
b.eq("year", "2023"),
|
||
// 仅选择室内案例
|
||
b.eq("location", "indoor")),
|
||
b.and(
|
||
// 类型为室内设计
|
||
b.eq("type", "interior"),
|
||
// 指定房间类型
|
||
b.in("room", "living_room", "study", "kitchen")
|
||
));
|
||
}
|
||
}
|