Merge pull request #82 from chenwei1015/fix-20250514-knowledge_info

fix- 20250514.sql, alter单词错误

README.md

@@ -34,6 +34,23 @@
     <a href="https://github.com/ageerle/ruoyi-ai/issues">提出新特性</a>
 </p>
 
+## 快速启动
+
+1. **克隆项目**
+   ```bash
+   git clone https://github.com/alanpeng/ruoyi-ai-docker-deploy
+   cd ruoyi-ai-docker-deploy
+   ```
+
+2. **启动全套应用**
+   ```bash
+   docker-compose up -d
+   ```
+
+3. **访问应用界面**
+   - 用户界面：`http://your-server-ip:8081`
+   - 管理员界面：`http://your-server-ip:8082`
+
 ## 目录
 
 - [系统体验](#系统体验)

ruoyi-admin/src/main/resources/application-dev.yml

@@ -94,3 +94,11 @@ sms:
   # 腾讯专用
   sdkAppId:
 
+pdf:
+  extract:
+    service:
+      url: http://localhost:8080
+    ai-api:
+      url: https://api.pandarobot.chat/v1/chat/completions
+      key: sk-xxxx
+

ruoyi-admin/src/main/resources/application-prod.yml

@@ -172,3 +172,11 @@ sms:
   signName: 测试
   # 腾讯专用
   sdkAppId:
+
+pdf:
+  extract:
+    service:
+      url: http://localhost:8080
+    ai-api:
+      url: https://api.pandarobot.chat/v1/chat/completions
+      key: sk-XXXXXX

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/domain/KnowledgeInfo.java

@@ -1,5 +1,6 @@
 package org.ruoyi.domain;
 
+import com.alibaba.excel.annotation.ExcelProperty;
 import com.baomidou.mybatisplus.annotation.*;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
@@ -78,14 +79,19 @@ public class KnowledgeInfo extends BaseEntity {
     private Long textBlockSize;
 
     /**
-     * 向量库
+     * 向量库模型名称
      */
-    private String vector;
+    private String vectorModelName;
 
     /**
-     * 向量模型
+     * 向量化模型名称
      */
-    private String vectorModel;
+    private String embeddingModelName;
+
+    /**
+     * 系统提示词
+     */
+    private String systemPrompt;
 
     /**
      * 备注

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/domain/PdfFileContentResult.java

@@ -0,0 +1,30 @@
+package org.ruoyi.domain;
+
+/**
+ * 文件内容结果封装类
+ */
+public class PdfFileContentResult {
+    private String filename;
+    private String content;
+
+    public PdfFileContentResult(String filename, String content) {
+        this.filename = filename;
+        this.content = content;
+    }
+
+    public String getFilename() {
+        return filename;
+    }
+
+    public void setFilename(String filename) {
+        this.filename = filename;
+    }
+
+    public String getContent() {
+        return content;
+    }
+
+    public void setContent(String content) {
+        this.content = content;
+    }
+}

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/domain/bo/KnowledgeInfoBo.java

@@ -83,16 +83,22 @@ public class KnowledgeInfoBo extends BaseEntity {
     private Long textBlockSize;
 
     /**
-     * 向量库
+     * 向量库模型名称
      */
     @NotBlank(message = "向量库不能为空", groups = { AddGroup.class, EditGroup.class })
-    private String vector;
+    private String vectorModelName;
 
     /**
-     * 向量模型
+     * 向量化模型名称
      */
     @NotBlank(message = "向量模型不能为空", groups = { AddGroup.class, EditGroup.class })
-    private String vectorModel;
+    private String embeddingModelName;
+
+
+    /**
+     * 系统提示词
+     */
+    private String systemPrompt;
 
     /**
      * 备注

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/domain/bo/QueryVectorBo.java

@@ -26,9 +26,14 @@ public class QueryVectorBo {
     private Integer maxResults;
 
     /**
-     * 模型名称
+     * 向量库模型名称
      */
-    private String modelName;
+    private String vectorModelName;
+
+    /**
+     * 向量化模型名称
+     */
+    private String embeddingModelName;
 
     /**
      * 请求key

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/domain/bo/StoreEmbeddingBo.java

@@ -32,9 +32,14 @@ public class StoreEmbeddingBo {
     private List<String> fids;
 
     /**
-     * 模型名称
+     * 向量库模型名称
      */
-    private String modelName;
+    private String vectorModelName;
+
+    /**
+     * 向量化模型名称
+     */
+    private String embeddingModelName;
 
     /**
      * 请求key

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/domain/vo/KnowledgeInfoVo.java

@@ -98,16 +98,20 @@ public class KnowledgeInfoVo implements Serializable {
     private Integer textBlockSize;
 
     /**
-     * 向量库
+     * 向量库模型名称
      */
-    @ExcelProperty(value = "向量库")
+    private String vectorModelName;
-    private String vector;
 
     /**
-     * 向量模型
+     * 向量化模型名称
      */
-    @ExcelProperty(value = "向量模型")
+    private String embeddingModelName;
-    private String vectorModel;
+
+
+    /**
+     * 系统提示词
+     */
+    private String systemPrompt;
 
     /**
      * 备注

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/service/PdfImageExtractService.java

@@ -0,0 +1,41 @@
+package org.ruoyi.service;
+
+import java.io.IOException;
+import java.util.List;
+import org.ruoyi.domain.PdfFileContentResult;
+import org.springframework.web.multipart.MultipartFile;
+
+/**
+ * PDF图片提取服务接口
+ */
+public interface PdfImageExtractService {
+
+  /**
+   * 从PDF文件中提取图片
+   *
+   * @param pdfFile PDF文件
+   * @param imageFormat 输出图片格式 (png, jpeg, gif)
+   * @param allowDuplicates 是否允许重复图片
+   * @return 包含提取图片的ZIP文件的字节数组
+   * @throws IOException 如果文件处理过程中发生错误
+   */
+  byte[] extractImages(MultipartFile pdfFile, String imageFormat, boolean allowDuplicates)
+      throws IOException;
+
+  /**
+   * 处理文件内容
+   *
+   * @param unzip Base64编码的图片数组
+   * @return 文件内容结果列表
+   * @throws IOException 如果API调用过程中发生错误
+   */
+  List<PdfFileContentResult> dealFileContent(String[] unzip) throws IOException;
+
+  /**
+   * 提取PDF中的图片并调用gpt-4o-mini,识别图片内容并返回
+   * @param file
+   * @return
+   * @throws IOException
+   */
+  List<PdfFileContentResult> extractImages(MultipartFile file) throws IOException;
+}

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/service/VectorStoreService.java

@@ -13,14 +13,14 @@ public interface VectorStoreService {
 
     void storeEmbeddings(StoreEmbeddingBo storeEmbeddingBo);
 
-    void removeByDocId(String kid,String docId);
-
-    void removeByKid(String kid);
-
     List<String> getQueryVector(QueryVectorBo queryVectorBo);
 
     void createSchema(String kid,String modelName);
 
-    void removeByKidAndFid(String kid, String fid);
+    void removeByKid(String kid,String modelName);
+
+    void removeByDocId(String kid,String docId,String modelName);
+
+    void removeByKidAndFid(String kid, String fid,String modelName);
 
 }

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/service/impl/PdfImageExtractServiceImpl.java

@@ -0,0 +1,144 @@
+package org.ruoyi.service.impl;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.TimeUnit;
+import lombok.extern.slf4j.Slf4j;
+import okhttp3.MediaType;
+import okhttp3.MultipartBody;
+import okhttp3.OkHttpClient;
+import okhttp3.OkHttpClient.Builder;
+import okhttp3.Request;
+import okhttp3.RequestBody;
+import okhttp3.Response;
+import org.ruoyi.common.core.domain.R;
+import org.ruoyi.domain.PdfFileContentResult;
+import org.ruoyi.service.PdfImageExtractService;
+import org.ruoyi.utils.ZipUtils;
+import org.springframework.beans.factory.annotation.Value;
+import org.springframework.stereotype.Service;
+import org.springframework.web.multipart.MultipartFile;
+
+/**
+ * PDF图片提取服务实现类
+ */
+@Service
+@Slf4j
+public class PdfImageExtractServiceImpl implements PdfImageExtractService {
+
+  @Value("${pdf.extract.service.url}")
+  private String serviceUrl;
+  @Value("${pdf.extract.ai-api.url}")
+  private String aiApiUrl;
+  @Value("${pdf.extract.ai-api.key}")
+  private String aiApiKey ;
+
+  private final OkHttpClient client = new Builder()
+      .connectTimeout(100, TimeUnit.SECONDS)
+      .readTimeout(150, TimeUnit.SECONDS)
+      .writeTimeout(150, TimeUnit.SECONDS)
+      .callTimeout(300, TimeUnit.SECONDS)
+      .build();
+
+  private static final MediaType JSON = MediaType.parse("application/json; charset=utf-8");
+
+  @Override
+  public byte[] extractImages(MultipartFile pdfFile, String imageFormat, boolean allowDuplicates)
+      throws IOException {
+    // 构建multipart请求
+    RequestBody requestBody = new MultipartBody.Builder()
+        .setType(MultipartBody.FORM)
+        .addFormDataPart("fileInput", pdfFile.getOriginalFilename(),
+            RequestBody.create(MediaType.parse("application/pdf"), pdfFile.getBytes()))
+        .addFormDataPart("format", imageFormat)
+        .addFormDataPart("allowDuplicates", String.valueOf(allowDuplicates))
+        .build();
+
+    // 创建请求
+    Request request = new Request.Builder()
+        .url(serviceUrl + "/api/v1/misc/extract-images")
+        .post(requestBody)
+        .build();
+
+    // 执行请求
+    try (Response response = client.newCall(request).execute()) {
+      if (!response.isSuccessful()) {
+        throw new IOException("请求失败: " + response.code());
+      }
+      return response.body().bytes();
+    }
+  }
+
+  /**
+   * 处理文件内容
+   *
+   * @param unzip Base64编码的图片数组
+   * @return 文件内容结果列表
+   * @throws IOException 如果API调用过程中发生错误
+   */
+  @Override
+  public List<PdfFileContentResult> dealFileContent(String[] unzip) throws IOException {
+    List<PdfFileContentResult> results = new ArrayList<>();
+    int i = 0;
+    for (String base64Image : unzip) {
+      // 构建请求JSON
+      String requestJson = String.format("{"
+          + "\"model\": \"gpt-4o\","
+          + "\"stream\": false,"
+          + "\"messages\": [{"
+          + "\"role\": \"user\","
+          + "\"content\": [{"
+          + "\"type\": \"text\","
+          + "\"text\": \"这张图片有什么\""
+          + "}, {"
+          + "\"type\": \"image_url\","
+          + "\"image_url\": {"
+          + "\"url\": \"%s\""
+          + "}}"
+          + "]}],"
+          + "\"max_tokens\": 400"
+          + "}", base64Image);
+
+      // 创建请求
+      Request request = new Request.Builder()
+          .url(aiApiUrl)
+          .addHeader("Authorization", "Bearer " + aiApiKey)
+          .post(RequestBody.create(JSON, requestJson))
+          .build();
+
+      // 执行请求
+      try {
+        log.info("=============call=" + ++i);
+        Response response = client.newCall(request).execute();
+        log.info("=============response=" + response);
+        if (!response.isSuccessful()) {
+          throw new IOException("API请求失败: " + response.code() + response.toString());
+        }
+
+        String responseBody = response.body().string();
+        log.info("=============responseBody=" + responseBody);
+        // 使用文件名（这里使用base64的前10个字符作为标识）和API返回内容创建结果对象
+        String filename = base64Image.substring(0, Math.min(base64Image.length(), 10));
+        results.add(new PdfFileContentResult(filename, responseBody));
+      } catch (Exception e) {
+        log.error(e.getMessage());
+        throw new RuntimeException(e);
+      }
+    }
+
+    return results;
+  }
+
+  @Override
+  public List<PdfFileContentResult> extractImages(MultipartFile file) throws IOException {
+    String format = "png";
+    boolean allowDuplicates = true;
+    // 获取ZIP数据
+    byte[] zipData = this.extractImages(file, format, allowDuplicates);
+    // 解压文件并识别图片内容并返回
+    String[] unzip = ZipUtils.unzipForBase64(zipData);
+    //解析图片内容
+    return this.dealFileContent(unzip);
+  }
+}

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/service/impl/VectorStoreServiceImpl.java

@@ -1,5 +1,7 @@
 package org.ruoyi.service.impl;
 
+import cn.hutool.core.util.RandomUtil;
+import com.google.protobuf.ServiceException;
 import dev.langchain4j.data.embedding.Embedding;
 import dev.langchain4j.data.segment.TextSegment;
 import dev.langchain4j.model.embedding.EmbeddingModel;
@@ -16,6 +18,7 @@ import dev.langchain4j.store.embedding.milvus.MilvusEmbeddingStore;
 import dev.langchain4j.store.embedding.qdrant.QdrantEmbeddingStore;
 import dev.langchain4j.store.embedding.weaviate.WeaviateEmbeddingStore;
 import lombok.RequiredArgsConstructor;
+import lombok.SneakyThrows;
 import lombok.extern.slf4j.Slf4j;
 import org.ruoyi.common.core.service.ConfigService;
 import org.ruoyi.domain.bo.QueryVectorBo;
@@ -40,11 +43,10 @@ public class VectorStoreServiceImpl implements VectorStoreService {
 
     private final ConfigService configService;
 
-    Map<String,EmbeddingStore<TextSegment>> storeMap = new HashMap<>();
+    private EmbeddingStore<TextSegment> embeddingStore;
 
     @Override
     public void createSchema(String kid,String modelName) {
-        EmbeddingStore<TextSegment> embeddingStore;
         switch (modelName) {
             case "weaviate" -> {
                 String protocol = configService.getConfigValue("weaviate", "protocol");
@@ -84,88 +86,83 @@ public class VectorStoreServiceImpl implements VectorStoreService {
                 embeddingStore = new InMemoryEmbeddingStore<>();
             }
         }
-        storeMap.put(kid,embeddingStore);
     }
 
     @Override
     public void storeEmbeddings(StoreEmbeddingBo storeEmbeddingBo) {
-        EmbeddingStore<TextSegment> store = storeMap.get(storeEmbeddingBo.getKid());
+        createSchema(storeEmbeddingBo.getKid(),storeEmbeddingBo.getVectorModelName());
-        EmbeddingModel embeddingModel = getEmbeddingModel(storeEmbeddingBo.getModelName(),
+        EmbeddingModel embeddingModel = getEmbeddingModel(storeEmbeddingBo.getEmbeddingModelName(),
                 storeEmbeddingBo.getApiKey(), storeEmbeddingBo.getBaseUrl());
-        for (int i = 0; i < storeEmbeddingBo.getChunkList().size(); i++) {
+        List<String> chunkList = storeEmbeddingBo.getChunkList();
+        for (int i = 0; i < chunkList.size(); i++) {
             Map<String, Object> dataSchema = new HashMap<>();
             dataSchema.put("kid", storeEmbeddingBo.getKid());
             dataSchema.put("docId", storeEmbeddingBo.getKid());
             dataSchema.put("fid", storeEmbeddingBo.getFids().get(i));
-            Response<Embedding> response = embeddingModel.embed(storeEmbeddingBo.getChunkList().get(i));
+            Embedding embedding = embeddingModel.embed(chunkList.get(i)).content();
-            Embedding embedding = response.content();
+            TextSegment segment = TextSegment.from(chunkList.get(i));
-            TextSegment segment = TextSegment.from(storeEmbeddingBo.getChunkList().get(i));
             segment.metadata().putAll(dataSchema);
-
+            embeddingStore.add(embedding,segment);
-            store.add(embedding,segment);
         }
     }
 
     @Override
     public List<String> getQueryVector(QueryVectorBo queryVectorBo) {
-        EmbeddingStore<TextSegment> store = storeMap.get(queryVectorBo.getKid());
+        createSchema(queryVectorBo.getKid(),queryVectorBo.getVectorModelName());
-
+        EmbeddingModel embeddingModel = getEmbeddingModel(queryVectorBo.getEmbeddingModelName(),
-        EmbeddingModel embeddingModel = getEmbeddingModel(queryVectorBo.getModelName(),
                 queryVectorBo.getApiKey(), queryVectorBo.getBaseUrl());
-        Filter simpleFilter = new IsEqualTo("kid", queryVectorBo.getKid());
+       // Filter simpleFilter = new IsEqualTo("kid", queryVectorBo.getKid());
         Embedding queryEmbedding = embeddingModel.embed(queryVectorBo.getQuery()).content();
         EmbeddingSearchRequest embeddingSearchRequest = EmbeddingSearchRequest.builder()
                 .queryEmbedding(queryEmbedding)
                 .maxResults(queryVectorBo.getMaxResults())
                 // 添加过滤条件
-                .filter(simpleFilter)
+         //       .filter(simpleFilter)
                 .build();
-        List<EmbeddingMatch<TextSegment>> matches = store.search(embeddingSearchRequest).matches();
+        List<EmbeddingMatch<TextSegment>> matches = embeddingStore.search(embeddingSearchRequest).matches();
-
         List<String> results = new ArrayList<>();
-
         matches.forEach(embeddingMatch -> results.add(embeddingMatch.embedded().text()));
         return results;
     }
 
 
     @Override
-    public void removeByKid(String kid) {
+    public void removeByKid(String kid,String modelName) {
-        EmbeddingStore<TextSegment> store = storeMap.get(kid);
+        createSchema(kid,modelName);
-
         // 根据条件删除向量数据
         Filter simpleFilter = new IsEqualTo("kid", kid);
-        store.removeAll(simpleFilter);
+        embeddingStore.removeAll(simpleFilter);
     }
 
     @Override
-    public void removeByDocId(String kid, String docId) {
+    public void removeByDocId(String kid, String docId,String modelName) {
-        EmbeddingStore<TextSegment> store = storeMap.get(kid);
+        createSchema(kid,modelName);
         // 根据条件删除向量数据
         Filter simpleFilterByDocId = new IsEqualTo("docId", docId);
-        store.removeAll(simpleFilterByDocId);
+        embeddingStore.removeAll(simpleFilterByDocId);
     }
 
     @Override
-    public void removeByKidAndFid(String kid, String fid) {
+    public void removeByKidAndFid(String kid, String fid,String modelName) {
-        EmbeddingStore<TextSegment> store = storeMap.get(kid);
+        createSchema(kid,modelName);
         // 根据条件删除向量数据
         Filter simpleFilterByKid = new IsEqualTo("kid", kid);
         Filter simpleFilterFid = new IsEqualTo("fid", fid);
         Filter simpleFilterByAnd = Filter.and(simpleFilterFid, simpleFilterByKid);
-        store.removeAll(simpleFilterByAnd);
+        embeddingStore.removeAll(simpleFilterByAnd);
     }
 
     /**
      * 获取向量模型
      */
-    public EmbeddingModel getEmbeddingModel(String modelName,String apiKey,String baseUrl) {
+    @SneakyThrows
-        EmbeddingModel embeddingModel = OpenAiEmbeddingModel.builder().build();
+    public EmbeddingModel getEmbeddingModel(String modelName, String apiKey, String baseUrl) {
+        EmbeddingModel embeddingModel;
         if(TEXT_EMBEDDING_3_SMALL.toString().equals(modelName)) {
              embeddingModel = OpenAiEmbeddingModel.builder()
                     .apiKey(apiKey)
                     .baseUrl(baseUrl)
-                    .modelName(TEXT_EMBEDDING_3_SMALL)
+                    .modelName(modelName)
                     .build();
         // TODO 添加枚举
         }else if("quentinz/bge-large-zh-v1.5".equals(modelName)) {
@@ -173,6 +170,14 @@ public class VectorStoreServiceImpl implements VectorStoreService {
                     .baseUrl(baseUrl)
                     .modelName(modelName)
                     .build();
+        }else if("baai/bge-m3".equals(modelName)) {
+            embeddingModel = OpenAiEmbeddingModel.builder()
+                    .apiKey(apiKey)
+                    .baseUrl(baseUrl)
+                    .modelName(modelName)
+                    .build();
+        }else {
+            throw new ServiceException("未找到对应向量化模型!");
         }
         return embeddingModel;
     }

ruoyi-modules-api/ruoyi-knowledge-api/src/main/java/org/ruoyi/utils/ZipUtils.java

@@ -0,0 +1,95 @@
+package org.ruoyi.utils;
+
+import java.io.BufferedOutputStream;
+import java.io.ByteArrayInputStream;
+import java.io.ByteArrayOutputStream;
+import java.io.File;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Base64;
+import java.util.List;
+import java.util.zip.ZipEntry;
+import java.util.zip.ZipInputStream;
+
+/**
+ * ZIP文件处理工具类
+ */
+public class ZipUtils {
+
+    /**
+     * 解压ZIP文件到指定目录
+     *
+     * @param zipData ZIP文件的字节数组
+     * @param destDir 目标目录
+     * @return 解压后的文件路径列表
+     * @throws IOException 如果解压过程中发生错误
+     */
+    public static String[] unzip(byte[] zipData, String destDir) throws IOException {
+        File destDirFile = new File(destDir);
+        if (!destDirFile.exists()) {
+            destDirFile.mkdirs();
+        }
+
+        List<String> extractedPaths = new ArrayList<>();
+        try (ByteArrayInputStream bis = new ByteArrayInputStream(zipData);
+             ZipInputStream zis = new ZipInputStream(bis)) {
+
+            ZipEntry zipEntry;
+            while ((zipEntry = zis.getNextEntry()) != null) {
+                String filePath = destDir + File.separator + zipEntry.getName();
+                if (!zipEntry.isDirectory()) {
+                    extractFile(zis, filePath);
+                    extractedPaths.add(filePath);
+                } else {
+                    new File(filePath).mkdirs();
+                }
+                zis.closeEntry();
+            }
+        }
+        return extractedPaths.toArray(new String[0]);
+    }
+
+    private static void extractFile(ZipInputStream zis, String filePath) throws IOException {
+        try (BufferedOutputStream bos = new BufferedOutputStream(new FileOutputStream(filePath))) {
+            byte[] buffer = new byte[4096];
+            int read;
+            while ((read = zis.read(buffer)) != -1) {
+                bos.write(buffer, 0, read);
+            }
+        }
+    }
+
+    /**
+     * 解压ZIP文件并返回文件内容的Base64编码字符串数组
+     *
+     * @param zipData ZIP文件的字节数组
+     * @return Base64编码的文件内容数组
+     * @throws IOException 如果解压过程中发生错误
+     */
+    public static String[] unzipForBase64(byte[] zipData) throws IOException {
+        List<String> base64Contents = new ArrayList<>();
+        try (ByteArrayInputStream bis = new ByteArrayInputStream(zipData);
+             ZipInputStream zis = new ZipInputStream(bis)) {
+
+            ZipEntry zipEntry;
+            while ((zipEntry = zis.getNextEntry()) != null) {
+                if (!zipEntry.isDirectory()) {
+                    // 读取文件内容到内存
+                    ByteArrayOutputStream baos = new ByteArrayOutputStream();
+                    byte[] buffer = new byte[4096];
+                    int read;
+                    while ((read = zis.read(buffer)) != -1) {
+                        baos.write(buffer, 0, read);
+                    }
+                    
+                    // 将文件内容转换为Base64字符串
+                    String base64Content = Base64.getEncoder().encodeToString(baos.toByteArray());
+                    base64Contents.add(base64Content);
+                }
+                zis.closeEntry();
+            }
+        }
+        return base64Contents.toArray(new String[0]);
+    }
+}

ruoyi-modules/ruoyi-chat/src/main/java/org/ruoyi/chat/controller/knowledge/KnowledgeController.java

@@ -1,9 +1,12 @@
 package org.ruoyi.chat.controller.knowledge;
 
 import cn.dev33.satoken.stp.StpUtil;
+import io.swagger.v3.oas.annotations.Operation;
+import io.swagger.v3.oas.annotations.Parameter;
 import jakarta.servlet.http.HttpServletResponse;
 import jakarta.validation.constraints.NotEmpty;
 import jakarta.validation.constraints.NotNull;
+import java.io.IOException;
 import lombok.RequiredArgsConstructor;
 import org.ruoyi.common.core.domain.R;
 import org.ruoyi.common.core.validate.AddGroup;
@@ -14,6 +17,7 @@ import org.ruoyi.common.satoken.utils.LoginHelper;
 import org.ruoyi.common.web.core.BaseController;
 import org.ruoyi.core.page.PageQuery;
 import org.ruoyi.core.page.TableDataInfo;
+import org.ruoyi.domain.PdfFileContentResult;
 import org.ruoyi.domain.bo.KnowledgeAttachBo;
 import org.ruoyi.domain.bo.KnowledgeFragmentBo;
 import org.ruoyi.domain.bo.KnowledgeInfoBo;
@@ -24,6 +28,7 @@ import org.ruoyi.domain.vo.KnowledgeInfoVo;
 import org.ruoyi.service.IKnowledgeAttachService;
 import org.ruoyi.service.IKnowledgeFragmentService;
 import org.ruoyi.service.IKnowledgeInfoService;
+import org.ruoyi.service.PdfImageExtractService;
 import org.springframework.validation.annotation.Validated;
 import org.springframework.web.bind.annotation.*;
 import org.springframework.web.multipart.MultipartFile;
@@ -47,6 +52,8 @@ public class KnowledgeController extends BaseController {
 
   private final IKnowledgeFragmentService fragmentService;
 
+  private final PdfImageExtractService pdfImageExtractService;
+
   /**
    * 根据用户信息查询本地知识库
    */
@@ -101,7 +108,8 @@ public class KnowledgeController extends BaseController {
    * 查询知识附件信息
    */
   @GetMapping("/detail/{kid}")
-    public TableDataInfo<KnowledgeAttachVo> attach(KnowledgeAttachBo bo, PageQuery pageQuery, @PathVariable String kid) {
+  public TableDataInfo<KnowledgeAttachVo> attach(KnowledgeAttachBo bo, PageQuery pageQuery,
+      @PathVariable String kid) {
     bo.setKid(kid);
     return attachService.queryPageList(bo, pageQuery);
   }
@@ -141,7 +149,8 @@ public class KnowledgeController extends BaseController {
    * 查询知识片段
    */
   @GetMapping("/fragment/list/{docId}")
-    public TableDataInfo<KnowledgeFragmentVo> fragmentList(KnowledgeFragmentBo bo, PageQuery pageQuery, @PathVariable String docId) {
+  public TableDataInfo<KnowledgeFragmentVo> fragmentList(KnowledgeFragmentBo bo,
+      PageQuery pageQuery, @PathVariable String docId) {
     bo.setDocId(docId);
     return fragmentService.queryPageList(bo, pageQuery);
   }
@@ -154,4 +163,18 @@ public class KnowledgeController extends BaseController {
   public String translationByFile(@RequestParam("file") MultipartFile file, String targetLanguage) {
     return attachService.translationByFile(file, targetLanguage);
   }
+
+  /**
+   * 提取PDF中的图片并调用gpt-4o-mini,识别图片内容并返回
+   *
+   * @param file PDF文件
+   * @return 保存的文件路径信息
+   */
+  @PostMapping("/extract-images")
+  @Operation(summary = "提取PDF中的图片并调用大模型,识别图片内容并返回", description = "提取PDF中的图片并调用gpt-4o-mini,识别图片内容并返回")
+  public R<List<PdfFileContentResult>> extractImages(
+      @RequestPart("file") MultipartFile file
+  ) throws IOException {
+    return R.ok(pdfImageExtractService.extractImages(file));
+  }
 }

ruoyi-modules/ruoyi-chat/src/main/java/org/ruoyi/chat/service/chat/impl/SseServiceImpl.java

@@ -2,6 +2,7 @@ package org.ruoyi.chat.service.chat.impl;
 
 import cn.dev33.satoken.stp.StpUtil;
 import cn.hutool.core.collection.CollectionUtil;
+import com.baomidou.mybatisplus.core.toolkit.Wrappers;
 import com.google.protobuf.ServiceException;
 import jakarta.servlet.http.HttpServletRequest;
 import lombok.RequiredArgsConstructor;
@@ -29,6 +30,8 @@ import org.ruoyi.common.redis.utils.RedisUtils;
 import org.ruoyi.domain.bo.ChatSessionBo;
 import org.ruoyi.domain.bo.QueryVectorBo;
 import org.ruoyi.domain.vo.ChatModelVo;
+import org.ruoyi.domain.vo.KnowledgeInfoVo;
+import org.ruoyi.service.IKnowledgeInfoService;
 import org.ruoyi.service.VectorStoreService;
 import org.ruoyi.service.IChatModelService;
 import org.ruoyi.service.IChatSessionService;
@@ -67,6 +70,8 @@ public class SseServiceImpl implements ISseService {
 
     private final IChatSessionService chatSessionService;
 
+    private final IKnowledgeInfoService knowledgeInfoService;
+
     private ChatModelVo chatModelVo;
 
 
@@ -148,50 +153,61 @@ public class SseServiceImpl implements ISseService {
             }
     }
 
-
     /**
      *  构建消息列表
      */
     private void buildChatMessageList(ChatRequest chatRequest){
+        String sysPrompt;
         chatModelVo = chatModelService.selectModelByName(chatRequest.getModel());
         // 获取对话消息列表
         List<Message> messages = chatRequest.getMessages();
-        String sysPrompt = chatModelVo.getSystemPrompt();
+        // 查询向量库相关信息加入到上下文
-
+        if(StringUtils.isNotEmpty(chatRequest.getKid())){
+            List<Message> knMessages = new ArrayList<>();
+            String content = messages.get(messages.size() - 1).getContent().toString();
+            // 通过kid查询知识库信息
+            KnowledgeInfoVo knowledgeInfoVo = knowledgeInfoService.queryById(Long.valueOf(chatRequest.getKid()));
+            // 查询向量模型配置信息
+            ChatModelVo chatModel = chatModelService.selectModelByName(knowledgeInfoVo.getEmbeddingModelName());
 
+            QueryVectorBo queryVectorBo = new QueryVectorBo();
+            queryVectorBo.setQuery(content);
+            queryVectorBo.setKid(chatRequest.getKid());
+            queryVectorBo.setApiKey(chatModel.getApiKey());
+            queryVectorBo.setBaseUrl(chatModel.getApiHost());
+            queryVectorBo.setVectorModelName(knowledgeInfoVo.getVectorModelName());
+            queryVectorBo.setEmbeddingModelName(knowledgeInfoVo.getEmbeddingModelName());
+            queryVectorBo.setMaxResults(knowledgeInfoVo.getRetrieveLimit());
+            List<String> nearestList = vectorStoreService.getQueryVector(queryVectorBo);
+            for (String prompt : nearestList) {
+                Message userMessage = Message.builder().content(prompt).role(Message.Role.USER).build();
+                knMessages.add(userMessage);
+            }
+            messages.addAll(knMessages);
+            // 设置知识库系统提示词
+            sysPrompt = knowledgeInfoVo.getSystemPrompt();
+            if(StringUtils.isEmpty(sysPrompt)){
+                sysPrompt ="###角色设定\n" +
+                        "你是一个智能知识助手，专注于利用上下文中的信息来提供准确和相关的回答。\n" +
+                        "###指令\n" +
+                        "当用户的问题与上下文知识匹配时，利用上下文信息进行回答。如果问题与上下文不匹配，运用自身的推理能力生成合适的回答。\n" +
+                        "###限制\n" +
+                        "确保回答清晰简洁，避免提供不必要的细节。始终保持语气友好" +
+                        "当前时间："+ DateUtils.getDate();
+            }
+        }else {
+            sysPrompt = chatModelVo.getSystemPrompt();
             if(StringUtils.isEmpty(sysPrompt)){
-            // TODO 系统默认提示词,后续会增加提示词管理
                 sysPrompt ="你是一个由RuoYI-AI开发的人工智能助手，名字叫熊猫助手。你擅长中英文对话，能够理解并处理各种问题，提供安全、有帮助、准确的回答。" +
                         "当前时间："+ DateUtils.getDate()+
                         "#注意：回复之前注意结合上下文和工具返回内容进行回复。";
             }
+        }
         // 设置系统默认提示词
         Message sysMessage = Message.builder().content(sysPrompt).role(Message.Role.SYSTEM).build();
         messages.add(0,sysMessage);
 
         chatRequest.setSysPrompt(sysPrompt);
-        // 查询向量库相关信息加入到上下文
-        if(StringUtils.isNotEmpty(chatRequest.getKid())){
-            List<Message> knMessages = new ArrayList<>();
-            String content = messages.get(messages.size() - 1).getContent().toString();
-            QueryVectorBo queryVectorBo = new QueryVectorBo();
-            queryVectorBo.setQuery(content);
-            queryVectorBo.setKid(chatRequest.getKid());
-            queryVectorBo.setApiKey(chatModelVo.getApiKey());
-            queryVectorBo.setBaseUrl(chatModelVo.getApiHost());
-            queryVectorBo.setModelName(chatModelVo.getModelName());
-            // TODO 查询向量返回条数,这里应该查询知识库配置
-            queryVectorBo.setMaxResults(3);
-            List<String> nearestList = vectorStoreService.getQueryVector(queryVectorBo);
-            for (String prompt : nearestList) {
-                Message userMessage = Message.builder().content(prompt).role(Message.Role.USER).build();
-                knMessages.add(userMessage);
-            }
-            // TODO 提示词,这里应该查询知识库配置
-            Message userMessage = Message.builder().content(content + (!nearestList.isEmpty() ? "\n\n注意：回答问题时，须严格根据我给你的系统上下文内容原文进行回答，请不要自己发挥,回答时保持原来文本的段落层级" : "")).role(Message.Role.USER).build();
-            knMessages.add(userMessage);
-            messages.addAll(knMessages);
-        }
         // 用户对话内容
         String chatString = null;
         // 获取用户对话信息

ruoyi-modules/ruoyi-chat/src/main/java/org/ruoyi/chat/service/knowledge/KnowledgeInfoServiceImpl.java

@@ -102,8 +102,6 @@ public class KnowledgeInfoServiceImpl implements IKnowledgeInfoService {
         lqw.eq(bo.getOverlapChar() != null, KnowledgeInfo::getOverlapChar, bo.getOverlapChar());
         lqw.eq(bo.getRetrieveLimit() != null, KnowledgeInfo::getRetrieveLimit, bo.getRetrieveLimit());
         lqw.eq(bo.getTextBlockSize() != null, KnowledgeInfo::getTextBlockSize, bo.getTextBlockSize());
-        lqw.eq(StringUtils.isNotBlank(bo.getVector()), KnowledgeInfo::getVector, bo.getVector());
-        lqw.eq(StringUtils.isNotBlank(bo.getVectorModel()), KnowledgeInfo::getVectorModel, bo.getVectorModel());
         return lqw;
     }
 
@@ -161,7 +159,7 @@ public class KnowledgeInfoServiceImpl implements IKnowledgeInfoService {
             }
             baseMapper.insert(knowledgeInfo);
             if (knowledgeInfo != null) {
-                vectorStoreService.createSchema(String.valueOf(knowledgeInfo.getId()),bo.getVector());
+                vectorStoreService.createSchema(String.valueOf(knowledgeInfo.getId()),bo.getVectorModelName());
             }
         }else {
             baseMapper.updateById(knowledgeInfo);
@@ -177,7 +175,7 @@ public class KnowledgeInfoServiceImpl implements IKnowledgeInfoService {
         check(knowledgeInfoList);
         // 删除向量库信息
         knowledgeInfoList.forEach(knowledgeInfoVo -> {
-            vectorStoreService.removeByKid(String.valueOf(knowledgeInfoVo.getId()));
+            vectorStoreService.removeByKid(String.valueOf(knowledgeInfoVo.getId()),knowledgeInfoVo.getVectorModelName());
         });
         // 删除附件和知识片段
         fragmentMapper.deleteByMap(map);
@@ -231,17 +229,18 @@ public class KnowledgeInfoServiceImpl implements IKnowledgeInfoService {
 
         // 通过kid查询知识库信息
         KnowledgeInfoVo knowledgeInfoVo = baseMapper.selectVoOne(Wrappers.<KnowledgeInfo>lambdaQuery()
-                .eq(KnowledgeInfo::getKid, kid));
+                .eq(KnowledgeInfo::getId, kid));
 
         // 通过向量模型查询模型信息
-        ChatModelVo chatModelVo = chatModelService.selectModelByName(knowledgeInfoVo.getVectorModel());
+        ChatModelVo chatModelVo = chatModelService.selectModelByName(knowledgeInfoVo.getEmbeddingModelName());
 
         StoreEmbeddingBo storeEmbeddingBo = new StoreEmbeddingBo();
         storeEmbeddingBo.setKid(kid);
         storeEmbeddingBo.setDocId(docId);
         storeEmbeddingBo.setFids(fids);
         storeEmbeddingBo.setChunkList(chunkList);
-        storeEmbeddingBo.setModelName(knowledgeInfoVo.getVectorModel());
+        storeEmbeddingBo.setVectorModelName(knowledgeInfoVo.getVectorModelName());
+        storeEmbeddingBo.setEmbeddingModelName(knowledgeInfoVo.getEmbeddingModelName());
         storeEmbeddingBo.setApiKey(chatModelVo.getApiKey());
         storeEmbeddingBo.setBaseUrl(chatModelVo.getApiHost());
         vectorStoreService.storeEmbeddings(storeEmbeddingBo);

script/sql/update/20250514.sql

@@ -0,0 +1,6 @@
+ALTER TABLE `knowledge_info`
+ADD COLUMN `system_prompt` varchar(255) NULL COMMENT '系统提示词' AFTER `vector_model`;
+
+ALTER TABLE `knowledge_info`
+    CHANGE COLUMN `vector` `vector_model_name` varchar(50) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci NULL DEFAULT NULL COMMENT '向量库' AFTER `text_block_size`,
+    CHANGE COLUMN `vector_model` `embedding_model_name` varchar(50) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci NULL DEFAULT NULL COMMENT '向量模型' AFTER `vector_model_name`;