SpringAI ONNX Embedding PGVector RAG
·
SpringAI ONNX Embedding PGVector RAG
1 依赖
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 https://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<parent>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-parent</artifactId>
<version>3.5.9</version>
<relativePath/> <!-- lookup parent from repository -->
</parent>
<groupId>com.xu</groupId>
<artifactId>spring-openai-onnx</artifactId>
<version>0.0.1-SNAPSHOT</version>
<name>spring-openai-onnx</name>
<description>Demo project for Spring Boot</description>
<properties>
<java.version>25</java.version>
<spring-ai.version>1.1.2</spring-ai.version>
</properties>
<dependencies>
<!-- SpringBoot 前端请求 -->
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-web</artifactId>
</dependency>
<!-- SpringAi openai -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-starter-model-openai</artifactId>
</dependency>
<!-- Rag 文档解析 -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-tika-document-reader</artifactId>
</dependency>
<!-- transformers 模型 -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-starter-model-transformers</artifactId>
</dependency>
<!-- pgvector -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-starter-vector-store-pgvector</artifactId>
</dependency>
<!-- hutool -->
<dependency>
<groupId>cn.hutool</groupId>
<artifactId>hutool-all</artifactId>
<version>5.8.42</version>
</dependency>
<!-- devtools -->
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-devtools</artifactId>
<scope>runtime</scope>
<optional>true</optional>
</dependency>
<!-- lombok -->
<dependency>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
<optional>true</optional>
</dependency>
<!-- test -->
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-test</artifactId>
<scope>test</scope>
</dependency>
</dependencies>
<dependencyManagement>
<dependencies>
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-bom</artifactId>
<version>${spring-ai.version}</version>
<type>pom</type>
<scope>import</scope>
</dependency>
</dependencies>
</dependencyManagement>
<build>
<plugins>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-compiler-plugin</artifactId>
<configuration>
<annotationProcessorPaths>
<path>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
</path>
</annotationProcessorPaths>
</configuration>
</plugin>
<plugin>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-maven-plugin</artifactId>
<configuration>
<excludes>
<exclude>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
</exclude>
</excludes>
</configuration>
</plugin>
</plugins>
</build>
</project>
2 配置
spring:
datasource:
username: postgres
password: 123456
url: jdbc:postgresql://localhost:5432/postgres
application:
name: spring-openai-onnx
ai:
openai:
api-key: 你的DeepSeek的api-key
base-url: https://api.deepseek.com
chat:
options:
model: deepseek-chat
embedding:
transformer:
onnx:
model-Uri: classpath:onnx/bge-small-zh-v1.5.onnx
model-output-name: last_hidden_state
tokenizer:
uri: classpath:onnx/bge-small-zh-v1.5-tokenizer.json
vectorstore:
pgvector:
dimensions: 512
table-name: vector_store
schema-name: public
distance-type: COSINE_DISTANCE
index-type: HNSW
initialize-schema: false
3 ONNX Embedding 下载
4 PGVector
-- 启用vector
create extension vector;
create extension if not exists vector;
select version();
-- 查看vector
select * from pg_extension where extname = 'vector';
-- 创建vector表
drop table if exists vector_store;
create table if not exists vector_store (
id uuid primary key,
content text,
metadata jsonb,
embedding vector(512)
);
create index on vector_store
using ivfflat (embedding vector_cosine_ops)
with (lists = 100);
select * from vector_store;
5 Java
1 RAG 存储
package com.xu;
import org.springframework.ai.document.Document;
import org.springframework.ai.reader.TextReader;
import org.springframework.ai.reader.tika.TikaDocumentReader;
import org.springframework.ai.transformer.splitter.TextSplitter;
import org.springframework.ai.vectorstore.VectorStore;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.test.context.SpringBootTest;
import org.springframework.core.io.FileSystemResource;
import org.springframework.core.io.Resource;
import org.junit.jupiter.api.Test;
import com.xu.conf.ZhaoXinTextSplitter;
import java.util.List;
@SpringBootTest
class RagTest {
@Autowired
private VectorStore store;
@Test
void text() {
Resource resource = new FileSystemResource("D:\\SourceCode\\简历\\a.txt");
TextReader reader = new TextReader(resource);
List<Document> origin = reader.get();
// 中文拆分
TextSplitter splitter = new ZhaoXinTextSplitter();
List<Document> documents = splitter.split(origin);
for (Document doc : documents) {
doc.getMetadata().put("tenant", "11");
}
store.add(documents);
}
@Test
void file() {
Resource pdf = new FileSystemResource("D:\\SourceCode\\简历\\个人简历.pdf");
Resource doc = new FileSystemResource("D:\\SourceCode\\简历\\个人简历.docx");
Resource excel = new FileSystemResource("D:\\SourceCode\\简历\\个人简历.xlsx");
TikaDocumentReader reader = new TikaDocumentReader(pdf);
List<Document> origin = reader.get();
// 中文拆分
TextSplitter splitter = new ZhaoXinTextSplitter();
List<Document> documents = splitter.split(origin);
store.add(documents);
}
}
2 RAG 搜索
package com.xu;
import org.springframework.ai.document.Document;
import org.springframework.ai.embedding.EmbeddingModel;
import org.springframework.ai.embedding.EmbeddingResponse;
import org.springframework.ai.vectorstore.SearchRequest;
import org.springframework.ai.vectorstore.VectorStore;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.test.context.SpringBootTest;
import org.junit.jupiter.api.Test;
import cn.hutool.json.JSONUtil;
import java.util.List;
@SpringBootTest
class RagTest {
@Autowired
private VectorStore store;
@Autowired
private EmbeddingModel embeddingModel;
@Test
void length() {
EmbeddingResponse response = embeddingModel.embedForResponse(List.of("测试文本"));
System.out.println("向量维度: " + response.getResult().getOutput().length);
}
@Test
void search() {
SearchRequest request = SearchRequest.builder()
.query("艾拉·万斯") // 查询
.topK(2) // 返回的相似文档数量
//.similarityThreshold(0.7) // 相似度阈值
.filterExpression("tenant == '11'") // 过滤条件
.build();
List<Document> docs = store.similaritySearch(request);
System.out.println(JSONUtil.toJsonPrettyStr(docs));
}
}
3 Agent + RAG
package com.xu;
import org.springframework.ai.chat.client.ChatClient;
import org.springframework.ai.chat.model.ChatResponse;
import org.springframework.ai.chat.prompt.Prompt;
import org.springframework.ai.chat.prompt.PromptTemplate;
import org.springframework.ai.document.Document;
import org.springframework.ai.vectorstore.SearchRequest;
import org.springframework.ai.vectorstore.VectorStore;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.test.context.SpringBootTest;
import org.junit.jupiter.api.Test;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.stream.Collectors;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@SpringBootTest
public class ChatTest {
@Autowired
private VectorStore redisVectorStore;
@Autowired
private ChatClient chatClient;
@Test
void common() {
ChatResponse chatResponse = chatClient.prompt("艾拉·万斯是谁, 知道回答知道, 不知道回答不知道").call().chatResponse();
String answer = chatResponse.getResult().getOutput().getText();
System.out.println(answer);
System.out.println("豆包生成回答：\n" + answer);
}
@Test
void spring() {
String userQuery = "艾拉·万斯是谁, 知道回答知道, 不知道回答不知道";
SearchRequest request = SearchRequest.builder()
.query("艾拉·万斯")
.topK(2)
.similarityThreshold(0.7)
.build();
List<Document> docs = redisVectorStore.similaritySearch(request);
String context = docs.stream()
.map(Document::getText)
.collect(Collectors.joining("\n\n"));
String promptTemplate = """
请你基于以下提供的上下文信息，回答用户的问题。
要求：
1. 严格按照上下文内容回答，不要添加你自己的额外知识；
2. 如果上下文没有相关信息，直接回答"无法从现有资料中获取该信息"；
3. 回答语言简洁、准确，符合中文表达习惯；
4. 保留上下文的关键信息，不遗漏重要内容。
上下文信息：
{context}
用户问题：
{userQuery}
""";
PromptTemplate promptTemp = new PromptTemplate(promptTemplate);
Map<String, Object> promptParams = new HashMap<>();
promptParams.put("context", context);
promptParams.put("userQuery", userQuery);
Prompt finalPrompt = promptTemp.create(promptParams);
ChatResponse chatResponse = chatClient.prompt(finalPrompt).call().chatResponse();
String answer = chatResponse.getResult().getOutput().getText();
System.out.println("豆包生成回答：\n" + answer);
}
}
4 ChatClient
package com.xu.conf;
import org.springframework.ai.chat.client.ChatClient;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import org.springframework.stereotype.Component;
/**
* @author hyacinth
*/
@Component
@Configuration
public class ChatClientConf {
@Bean
public ChatClient chatClient(ChatClient.Builder builder) {
return builder.build();
}
}
5 RAG文字拆分
package com.xu.conf;
import org.springframework.ai.transformer.splitter.TextSplitter;
import java.util.ArrayList;
import java.util.List;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
import java.util.stream.Collectors;
/**
* 优化后的中文专属文本拆分器
* 适配中文语义边界，避免拆分后语义断裂
*
* @author hyacinth
*/
public class ZhaoXinTextSplitter extends TextSplitter {
// 中文核心分隔符正则（按「强语义边界→弱语义边界」排序，优先拆分大的语义单元）
private static final Pattern CHINESE_SEPARATOR_PATTERN = Pattern.compile(
"(。|？|！|；|：|\\n\\n|\\n|，)",
Pattern.UNICODE_CASE | Pattern.MULTILINE
);
// 空白字符过滤正则（去除全角/半角空白、制表符等）
private static final Pattern BLANK_PATTERN = Pattern.compile("\\s+");
@Override
protected List<String> splitText(String text) {
// 步骤1：预处理文本，去除多余空白，避免无效拆分
String processedText = preprocessText(text);
// 步骤2：按中文语义边界拆分文本，得到初步拆分结果
List<String> initialSplits = splitByChineseSemantic(processedText);
// 步骤3：过滤无效分块，合并过短分块，得到最终有效分块
return filterAndOptimizeChunks(initialSplits);
}
/**
* 文本预处理：去除多余空白、制表符，统一换行格式
*/
private String preprocessText(String text) {
if (text == null || text.isBlank()) {
return "";
}
// 1. 替换全角空白为半角
String result = text.replace(" ", " ");
// 2. 去除连续多个空白（空格、制表符等），保留单个空格
result = BLANK_PATTERN.matcher(result).replaceAll(" ");
// 3. 去除文本首尾空白
return result.trim();
}
/**
* 核心：按中文语义边界拆分文本，保留分隔符（保证语义完整）
*/
private List<String> splitByChineseSemantic(String text) {
List<String> chunks = new ArrayList<>();
if (text.isBlank()) {
return chunks;
}
// 利用正则拆分，同时保留分隔符（中文标点是语义的重要组成，不可丢弃）
String[] splitParts = CHINESE_SEPARATOR_PATTERN.split(text);
Matcher matcher = CHINESE_SEPARATOR_PATTERN.matcher(text);
List<String> separators = new ArrayList<>();
// 提取所有匹配到的中文分隔符（用于后续拼接，保留语义）
while (matcher.find()) {
separators.add(matcher.group());
}
// 拼接拆分后的内容和对应的分隔符，形成完整语义块
for (int i = 0; i < splitParts.length; i++) {
String part = splitParts[i].trim();
if (part.isBlank()) {
continue;
}
// 拼接当前内容和对应的分隔符（避免分隔符丢失）
StringBuilder chunk = new StringBuilder(part);
if (i < separators.size()) {
chunk.append(separators.get(i));
}
chunks.add(chunk.toString().trim());
}
return chunks;
}
/**
* 过滤无效分块，优化分块质量（去除空值、合并过短分块）
*/
private List<String> filterAndOptimizeChunks(List<String> initialSplits) {
// 步骤1：过滤空字符串、仅含标点的无效分块
List<String> validChunks = initialSplits.stream()
.filter(chunk -> chunk != null && !chunk.isBlank() && !isOnlyPunctuation(chunk))
.collect(Collectors.toList());
// 步骤2：合并过短分块（避免出现单个词语、几个汉字的无效小分块）
return mergeShortChunks(validChunks, 100);
}
/**
* 判断是否仅含标点符号（过滤无效分块）
*/
private boolean isOnlyPunctuation(String chunk) {
return chunk.matches("[。|？|！|；|：|，|、|\\s]+");
}
/**
* 合并过短分块，保证分块语义完整性
*
* @param validChunks 有效分块列表
* @param minChunkLength 最小分块字符数
* @return 优化后的分块列表
*/
private List<String> mergeShortChunks(List<String> validChunks, int minChunkLength) {
List<String> mergedChunks = new ArrayList<>();
StringBuilder currentChunk = new StringBuilder();
for (String chunk : validChunks) {
// 若当前拼接的分块长度不足，继续拼接下一个分块
if (currentChunk.length() + chunk.length() < minChunkLength) {
currentChunk.append(chunk);
} else {
// 长度达标，存入结果列表，重置当前拼接器
if (!currentChunk.isEmpty()) {
mergedChunks.add(currentChunk.toString().trim());
currentChunk.setLength(0);
}
mergedChunks.add(chunk.trim());
}
}
// 处理最后一个未完成拼接的分块
if (!currentChunk.isEmpty()) {
mergedChunks.add(currentChunk.toString().trim());
}
return mergedChunks;
}
// 可选：重写拆分入口方法，适配Spring AI的Document拆分
@Override
public List<org.springframework.ai.document.Document> split(List<org.springframework.ai.document.Document> documents) {
return super.split(documents);
}
}
汇聚全球AI编程工具，助力开发者即刻编程。
更多推荐
3
0- 0
已为社区贡献3条内容
所有评论(0)