refactor(ai): 合并ai_document表到vector_store表，切换文档ID类型为字符串

- 删除旧的ai_document表及相关索引，新增vector_store表兼容Spring AI PgVectorStore结构 - 调整实体类AiDocument映射到vector_store表，使用字符串ID代替Long类型 - 修改Mapper接口及XML中所有ID相关SQL使用字符串类型，并替换表名为vector_store - 修改服务接口与实现类，文档ID参数类型统一为字符串 - 处理文档分块时改用UUID生成chunk ID，确保唯一且格式正确 - 禁用Spring Ai PgVectorStore的自动schema初始化，使用手动创建的表结构 - 更新配置文件OpenAI模型API key及基础URL配置，支持多模型与聊天功能 - 优化日志输出，增加分块文档ID和父文档ID显示，方便调试追踪
2026-03-30 17:43:29 +08:00
parent 9f972f5e30
commit 37da5da044
11 changed files with 202 additions and 188 deletions
--- a/docs/dev-ops/pgsql/sql/weform_run.sql
+++ b/docs/dev-ops/pgsql/sql/weform_run.sql
@@ -8,6 +8,122 @@
 CREATE EXTENSION IF NOT EXISTS vector;
 CREATE EXTENSION IF NOT EXISTS "uuid-ossp";

+-- =====================================================
+-- Spring AI PgVectorStore 向量存储表
+-- 用于RAG文档向量存储和相似度搜索
+-- 兼容PgVectorStore默认结构，同时支持完整文档管理
+-- =====================================================
+DROP TABLE IF EXISTS vector_store CASCADE;
+CREATE TABLE vector_store (
+    -- PgVectorStore 核心字段
+    id VARCHAR(255) PRIMARY KEY,
+    content TEXT,
+    metadata JSONB,
+    embedding vector(1536),
+
+    -- 关联关系
+    project_id BIGINT,
+    timeline_node_id BIGINT,
+    kb_id BIGINT,
+
+    -- 文档来源
+    source_type VARCHAR(50),
+    source_id BIGINT,
+
+    -- 文档扩展信息
+    title VARCHAR(500),
+    content_raw TEXT,
+    summary TEXT,
+
+    -- 文档元数据
+    doc_type VARCHAR(50),
+    language VARCHAR(10) DEFAULT 'zh',
+    file_type VARCHAR(50),
+    file_size BIGINT,
+    file_path VARCHAR(500),
+
+    -- 时间信息 (用于时间维度检索)
+    doc_date DATE,
+    doc_datetime TIMESTAMP,
+
+    -- 分块信息(大文档分块存储)
+    chunk_index INT DEFAULT 0,
+    chunk_total INT DEFAULT 1,
+    chunk_parent_id VARCHAR(255),
+
+    -- 标签和分类
+    tags JSONB,
+    category VARCHAR(100),
+
+    -- 使用统计
+    view_count INT DEFAULT 0,
+    query_count INT DEFAULT 0,
+    last_queried_at TIMESTAMP,
+
+    -- 状态
+    status VARCHAR(20) DEFAULT 'active',
+    error_message TEXT,
+
+    -- 创建信息
+    create_by BIGINT,
+    create_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    update_by BIGINT,
+    update_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    deleted SMALLINT DEFAULT 0,
+
+    -- 外键约束
+    CONSTRAINT fk_vs_project FOREIGN KEY (project_id) REFERENCES project(id) ON DELETE SET NULL,
+    CONSTRAINT fk_vs_timeline FOREIGN KEY (timeline_node_id) REFERENCES project_timeline(id) ON DELETE SET NULL,
+    CONSTRAINT fk_vs_kb FOREIGN KEY (kb_id) REFERENCES ai_knowledge_base(id) ON DELETE SET NULL,
+    CONSTRAINT fk_vs_create_by FOREIGN KEY (create_by) REFERENCES sys_user(id) ON DELETE SET NULL
+);
+
+-- 创建向量索引（使用IVFFlat算法，适合中等数据量）
+CREATE INDEX idx_vector_store_embedding ON vector_store
+    USING ivfflat (embedding vector_cosine_ops)
+    WITH (lists = 100);
+
+-- 创建其他常用索引
+CREATE INDEX idx_vs_project ON vector_store(project_id) WHERE deleted = 0;
+CREATE INDEX idx_vs_timeline ON vector_store(timeline_node_id) WHERE deleted = 0;
+CREATE INDEX idx_vs_kb ON vector_store(kb_id) WHERE deleted = 0;
+CREATE INDEX idx_vs_source ON vector_store(source_type, source_id) WHERE deleted = 0;
+CREATE INDEX idx_vs_status ON vector_store(status);
+CREATE INDEX idx_vs_type ON vector_store(doc_type);
+CREATE INDEX idx_vs_tags ON vector_store USING GIN(tags);
+CREATE INDEX idx_vs_chunk_parent ON vector_store(chunk_parent_id) WHERE chunk_parent_id IS NOT NULL;
+
+COMMENT ON TABLE vector_store IS '向量存储表 - 用于RAG文档检索和知识库管理';
+COMMENT ON COLUMN vector_store.id IS '文档ID（字符串类型，兼容PgVectorStore）';
+COMMENT ON COLUMN vector_store.content IS '文档内容文本';
+COMMENT ON COLUMN vector_store.metadata IS '文档元数据（JSONB格式，兼容PgVectorStore）';
+COMMENT ON COLUMN vector_store.embedding IS '向量嵌入（1536维）';
+COMMENT ON COLUMN vector_store.project_id IS '关联项目ID';
+COMMENT ON COLUMN vector_store.timeline_node_id IS '关联时间节点ID';
+COMMENT ON COLUMN vector_store.kb_id IS '关联知识库ID';
+COMMENT ON COLUMN vector_store.source_type IS '来源类型: project-项目文档, risk-风险文档, ticket-工单, report-日报, upload-上传文件, knowledge-知识库, chat-对话记录';
+COMMENT ON COLUMN vector_store.source_id IS '来源记录ID';
+COMMENT ON COLUMN vector_store.title IS '文档标题';
+COMMENT ON COLUMN vector_store.content_raw IS '原始内容(带格式)';
+COMMENT ON COLUMN vector_store.summary IS 'AI生成的摘要';
+COMMENT ON COLUMN vector_store.doc_type IS '文档类型: requirement-需求, design-设计, plan-计划, report-报告, contract-合同, photo-照片, other-其他';
+COMMENT ON COLUMN vector_store.language IS '语言: zh-中文, en-英文';
+COMMENT ON COLUMN vector_store.file_type IS '文件类型: pdf, doc, txt, md, jpg, png等';
+COMMENT ON COLUMN vector_store.file_size IS '文件大小(字节)';
+COMMENT ON COLUMN vector_store.file_path IS '文件存储路径';
+COMMENT ON COLUMN vector_store.doc_date IS '文档日期(如日报日期、照片拍摄日期)';
+COMMENT ON COLUMN vector_store.doc_datetime IS '文档时间戳';
+COMMENT ON COLUMN vector_store.chunk_index IS '分块序号';
+COMMENT ON COLUMN vector_store.chunk_total IS '总分块数';
+COMMENT ON COLUMN vector_store.chunk_parent_id IS '父文档ID(分块时使用)';
+COMMENT ON COLUMN vector_store.tags IS '标签数组';
+COMMENT ON COLUMN vector_store.category IS '分类';
+COMMENT ON COLUMN vector_store.view_count IS '查看次数';
+COMMENT ON COLUMN vector_store.query_count IS '被检索次数';
+COMMENT ON COLUMN vector_store.last_queried_at IS '最后被检索时间';
+COMMENT ON COLUMN vector_store.status IS '状态: active-可用, processing-处理中, error-错误, archived-归档';
+COMMENT ON COLUMN vector_store.error_message IS '错误信息';
+
 -- 设置时区
 SET timezone = 'Asia/Shanghai';

@@ -969,118 +1085,8 @@ COMMENT ON COLUMN ai_knowledge_base.status IS '状态: active-可用, archived-
 -- 11. AI服务相关表
 -- =====================================================

-- AI文档向量表 (用于RAG知识库)
-DROP TABLE IF EXISTS ai_document CASCADE;
-CREATE TABLE ai_document (
-                             id BIGSERIAL PRIMARY KEY,
-                             doc_id UUID DEFAULT uuid_generate_v4(),
-
-    -- 关联关系
-                             project_id BIGINT,
-                             timeline_node_id BIGINT,
-                             kb_id BIGINT,
-    -- 文档来源
-                             source_type VARCHAR(50) NOT NULL,
-                             source_id BIGINT,
-
-    -- 文档内容
-                             title VARCHAR(500),
-                             content TEXT NOT NULL,
-                             content_raw TEXT,
-                             summary TEXT,
-
-    -- 向量嵌入 (1536维适配OpenAI, 可调整为其他维度)
-                             embedding vector(1536),
-
-    -- 文档元数据
-                             doc_type VARCHAR(50),
-                             language VARCHAR(10) DEFAULT 'zh',
-                             file_type VARCHAR(50),
-                             file_size BIGINT,
-                             file_path VARCHAR(500),
-
-    -- 时间信息 (用于时间维度检索)
-                             doc_date DATE,
-                             doc_datetime TIMESTAMP,
-
-    -- 分块信息(大文档分块存储)
-                             chunk_index INT DEFAULT 0,
-                             chunk_total INT DEFAULT 1,
-                             chunk_parent_id BIGINT,
-
-    -- 标签和分类
-                             tags JSONB,
-                             category VARCHAR(100),
-
-    -- 使用统计
-                             view_count INT DEFAULT 0,
-                             query_count INT DEFAULT 0,
-                             last_queried_at TIMESTAMP,
-
-    -- 状态
-                             status VARCHAR(20) DEFAULT 'active',
-                             error_message TEXT,
-
-    -- 创建信息
-                             create_by BIGINT,
-                             create_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
-                             update_by BIGINT,
-                             update_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
-                             deleted SMALLINT DEFAULT 0,
-
-    -- 外键约束
-                             CONSTRAINT fk_ai_doc_project FOREIGN KEY (project_id) REFERENCES project(id) ON DELETE SET NULL,
-                             CONSTRAINT fk_ai_doc_timeline FOREIGN KEY (timeline_node_id) REFERENCES project_timeline(id) ON DELETE SET NULL,
-                             CONSTRAINT fk_ai_doc_kb FOREIGN KEY (kb_id) REFERENCES ai_knowledge_base(id) ON DELETE SET NULL
-);
-
-- 创建向量索引 (使用IVFFlat或HNSW)
-- IVFFlat: 适合中等数据量, 内存占用小
-- HNSW: 适合大数据量, 查询更快但内存占用大
-CREATE INDEX idx_ai_document_embedding ON ai_document
-    USING ivfflat (embedding vector_cosine_ops)
-    WITH (lists = 100);
-
-- 创建其他常用索引
-CREATE INDEX idx_ai_doc_project ON ai_document(project_id) WHERE deleted = 0;
-CREATE INDEX idx_ai_doc_timeline ON ai_document(timeline_node_id) WHERE deleted = 0;
-CREATE INDEX idx_ai_doc_kb ON ai_document(kb_id) WHERE deleted = 0;
-CREATE INDEX idx_ai_doc_source ON ai_document(source_type, source_id) WHERE deleted = 0;
-CREATE INDEX idx_ai_doc_status ON ai_document(status);
-CREATE INDEX idx_ai_doc_type ON ai_document(doc_type);
-CREATE INDEX idx_ai_doc_tags ON ai_document USING GIN(tags);
-
-COMMENT ON TABLE ai_document IS 'AI文档向量表 - 存储所有用于RAG的文档向量';
-COMMENT ON COLUMN ai_document.doc_id IS '文档唯一标识';
-COMMENT ON COLUMN ai_document.project_id IS '关联项目ID';
-COMMENT ON COLUMN ai_document.timeline_node_id IS '关联时间节点ID';
-COMMENT ON COLUMN ai_document.kb_id IS '关联知识库ID';
-COMMENT ON COLUMN ai_document.source_type IS '来源类型: project-项目文档, risk-风险文档, ticket-工单, report-日报, upload-上传文件, knowledge-知识库, chat-对话记录';
-COMMENT ON COLUMN ai_document.source_id IS '来源记录ID';
-COMMENT ON COLUMN ai_document.title IS '文档标题';
-COMMENT ON COLUMN ai_document.content IS '文档内容(纯文本)';
-COMMENT ON COLUMN ai_document.content_raw IS '原始内容(带格式)';
-COMMENT ON COLUMN ai_document.summary IS 'AI生成的摘要';
-COMMENT ON COLUMN ai_document.embedding IS '向量嵌入';
-COMMENT ON COLUMN ai_document.doc_type IS '文档类型: requirement-需求, design-设计, plan-计划, report-报告, contract-合同, photo-照片, other-其他';
-COMMENT ON COLUMN ai_document.language IS '语言: zh-中文, en-英文';
-COMMENT ON COLUMN ai_document.file_type IS '文件类型: pdf, doc, txt, md, jpg, png等';
-COMMENT ON COLUMN ai_document.file_size IS '文件大小(字节)';
-COMMENT ON COLUMN ai_document.file_path IS '文件存储路径';
-COMMENT ON COLUMN ai_document.doc_date IS '文档日期(如日报日期、照片拍摄日期)';
-COMMENT ON COLUMN ai_document.doc_datetime IS '文档时间戳';
-COMMENT ON COLUMN ai_document.chunk_index IS '分块序号';
-COMMENT ON COLUMN ai_document.chunk_total IS '总分块数';
-COMMENT ON COLUMN ai_document.chunk_parent_id IS '父文档ID(分块时使用)';
-COMMENT ON COLUMN ai_document.tags IS '标签数组';
-COMMENT ON COLUMN ai_document.category IS '分类';
-COMMENT ON COLUMN ai_document.view_count IS '查看次数';
-COMMENT ON COLUMN ai_document.query_count IS '被检索次数';
-COMMENT ON COLUMN ai_document.last_queried_at IS '最后被检索时间';
-COMMENT ON COLUMN ai_document.status IS '状态: active-可用, processing-处理中, error-错误, archived-归档';
-COMMENT ON COLUMN ai_document.error_message IS '错误信息';
-
 -- AI对话记录表 (合并会话管理功能，无需单独的session表)
+-- 注：ai_document表已合并到vector_store表中
 DROP TABLE IF EXISTS ai_chat_history CASCADE;
 CREATE TABLE ai_chat_history (
                                 id BIGSERIAL PRIMARY KEY,
--- a/src/main/java/cn/yinlihupo/common/config/SpringAiConfig.java
+++ b/src/main/java/cn/yinlihupo/common/config/SpringAiConfig.java
@@ -41,7 +41,7 @@ public class SpringAiConfig {
        return PgVectorStore.builder(jdbcTemplate, embeddingModel)
                .dimensions(1536)  // 向量维度，与配置一致
                .distanceType(PgVectorStore.PgDistanceType.COSINE_DISTANCE)
-                .initializeSchema(true)  // 自动初始化schema
+                .initializeSchema(false)  // 禁用自动初始化，使用SQL文件中已创建的表
                .build();
    }
 }
--- a/src/main/java/cn/yinlihupo/domain/entity/AiDocument.java
+++ b/src/main/java/cn/yinlihupo/domain/entity/AiDocument.java
@@ -9,20 +9,20 @@ import java.time.LocalDate;
 import java.time.LocalDateTime;

 /**
- * AI文档向量实体
- * 对应 ai_document 表
+ * 向量存储实体
+ * 对应 vector_store 表
+ * 兼容 Spring AI PgVectorStore 默认结构
 */
@Data
-@TableName("ai_document")
+@TableName("vector_store")
 public class AiDocument {

-    @TableId(type = IdType.AUTO)
-    private Long id;
-
    /**
-     * 文档唯一标识(UUID)
+     * 文档ID（字符串类型，兼容PgVectorStore）
+     * 使用标准UUID格式（带连字符）
     */
-    private String docId;
+    @TableId(type = IdType.INPUT)
+    private String id;

    /**
     * 关联项目ID
@@ -124,7 +124,7 @@ public class AiDocument {
    /**
     * 父文档ID(分块时使用)
     */
-    private Long chunkParentId;
+    private String chunkParentId;

    /**
     * 标签数组(JSON)
--- a/src/main/java/cn/yinlihupo/domain/vo/KbDocumentVO.java
+++ b/src/main/java/cn/yinlihupo/domain/vo/KbDocumentVO.java
@@ -13,10 +13,10 @@ public class KbDocumentVO {
    /**
     * 文档ID
     */
-    private Long id;
+    private String id;

    /**
-     * 文档UUID
+     * 文档UUID（与id相同）
     */
    private String docId;

--- a/src/main/java/cn/yinlihupo/domain/vo/ReferencedDocVO.java
+++ b/src/main/java/cn/yinlihupo/domain/vo/ReferencedDocVO.java
@@ -11,10 +11,10 @@ public class ReferencedDocVO {
    /**
     * 文档ID
     */
-    private Long id;
+    private String id;

    /**
-     * 文档UUID
+     * 文档UUID（与id相同）
     */
    private String docId;

--- a/src/main/java/cn/yinlihupo/mapper/AiDocumentMapper.java
+++ b/src/main/java/cn/yinlihupo/mapper/AiDocumentMapper.java
@@ -10,7 +10,7 @@ import org.apache.ibatis.annotations.Param;
 import java.util.List;

 /**
- * AI文档向量Mapper
+ * 向量存储Mapper
 */
@Mapper
 public interface AiDocumentMapper extends BaseMapper<AiDocument> {
@@ -24,17 +24,17 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
    List<KbDocumentVO> selectProjectDocuments(@Param("projectId") Long projectId);

    /**
-     * 根据docId查询文档
+     * 根据id查询文档
     *
-     * @param docId 文档UUID
+     * @param docId 文档ID
     * @return 文档实体
     */
    AiDocument selectByDocId(@Param("docId") String docId);

    /**
-     * 根据docId删除文档
+     * 根据id删除文档
     *
-     * @param docId 文档UUID
+     * @param docId 文档ID
     * @return 影响行数
     */
    int deleteByDocId(@Param("docId") String docId);
@@ -45,7 +45,7 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
     * @param docIds 文档ID列表
     * @return 文档信息列表
     */
-    List<ReferencedDocVO> selectReferencedDocs(@Param("docIds") List<Long> docIds);
+    List<ReferencedDocVO> selectReferencedDocs(@Param("docIds") List<String> docIds);

    /**
     * 获取父文档的分块数量
@@ -53,12 +53,12 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
     * @param docId 父文档ID
     * @return 分块数量
     */
-    Integer selectChunkCount(@Param("docId") Long docId);
+    Integer selectChunkCount(@Param("docId") String docId);

    /**
     * 更新文档状态
     *
-     * @param docId  文档UUID
+     * @param docId  文档ID
     * @param status 状态
     * @return 影响行数
     */
@@ -67,7 +67,7 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
    /**
     * 更新文档错误信息
     *
-     * @param docId        文档UUID
+     * @param docId        文档ID
     * @param errorMessage 错误信息
     * @return 影响行数
     */
@@ -79,7 +79,7 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
     * @param id 文档ID
     * @return 影响行数
     */
-    int incrementViewCount(@Param("id") Long id);
+    int incrementViewCount(@Param("id") String id);

    /**
     * 增加文档查询次数
@@ -87,5 +87,5 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
     * @param id 文档ID
     * @return 影响行数
     */
-    int incrementQueryCount(@Param("id") Long id);
+    int incrementQueryCount(@Param("id") String id);
 }
--- a/src/main/java/cn/yinlihupo/service/ai/AiKnowledgeBaseService.java
+++ b/src/main/java/cn/yinlihupo/service/ai/AiKnowledgeBaseService.java
@@ -49,12 +49,12 @@ public interface AiKnowledgeBaseService {
     *
     * @param docId 文档ID
     */
-    void processDocument(Long docId);
+    void processDocument(String docId);

    /**
     * 异步处理文档
     *
     * @param docId 文档ID
     */
-    void processDocumentAsync(Long docId);
+    void processDocumentAsync(String docId);
 }
--- a/src/main/java/cn/yinlihupo/service/ai/impl/AiKnowledgeBaseServiceImpl.java
+++ b/src/main/java/cn/yinlihupo/service/ai/impl/AiKnowledgeBaseServiceImpl.java
@@ -55,7 +55,7 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {

        // 4. 保存文档元数据
        AiDocument doc = new AiDocument();
-        doc.setDocId(docId);
+        doc.setId(docId);  // 设置标准UUID格式的ID
        doc.setProjectId(projectId);
        doc.setSourceType("upload");
        doc.setTitle(originalFilename);
@@ -75,7 +75,7 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {
        // 5. 异步处理文档（解析、切片、向量化）
        documentProcessor.processDocumentAsync(doc.getId());

-        log.info("文件上传成功: {}, docId: {}", originalFilename, docId);
+        log.info("文件上传成功: {}, docId: {}", originalFilename, doc.getId());

        // 6. 返回VO
        return convertToVO(doc);
@@ -133,13 +133,13 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {
    }

    @Override
-    public void processDocument(Long docId) {
+    public void processDocument(String docId) {
        documentProcessor.processDocument(docId);
    }

    @Override
    @Async
-    public void processDocumentAsync(Long docId) {
+    public void processDocumentAsync(String docId) {
        documentProcessor.processDocument(docId);
    }

@@ -197,7 +197,7 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {
    private KbDocumentVO convertToVO(AiDocument doc) {
        KbDocumentVO vo = new KbDocumentVO();
        vo.setId(doc.getId());
-        vo.setDocId(doc.getDocId());
+        vo.setDocId(doc.getId());
        vo.setTitle(doc.getTitle());
        vo.setDocType(doc.getDocType());
        vo.setFileType(doc.getFileType());
--- a/src/main/java/cn/yinlihupo/service/ai/rag/DocumentProcessor.java
+++ b/src/main/java/cn/yinlihupo/service/ai/rag/DocumentProcessor.java
@@ -16,6 +16,7 @@ import java.io.InputStream;
 import java.util.Collections;
 import java.util.List;
 import java.util.Map;
+import java.util.UUID;
 import java.util.stream.Collectors;

 /**
@@ -41,7 +42,7 @@ public class DocumentProcessor {
     *
     * @param docId 文档ID
     */
-    public void processDocument(Long docId) {
+    public void processDocument(String docId) {
        AiDocument doc = documentMapper.selectById(docId);
        if (doc == null) {
            log.error("文档不存在: {}", docId);
@@ -91,7 +92,7 @@ public class DocumentProcessor {
     * @param docId 文档ID
     */
    @Async("documentTaskExecutor")
-    public void processDocumentAsync(Long docId) {
+    public void processDocumentAsync(String docId) {
        processDocument(docId);
    }

@@ -169,23 +170,25 @@ public class DocumentProcessor {
     * @param chunks    切片列表
     */
    private void storeChunks(AiDocument parentDoc, List<String> chunks) {
-        String docId = parentDoc.getDocId();
-        Long parentId = parentDoc.getId();
+        String parentId = parentDoc.getId();

        for (int i = 0; i < chunks.size(); i++) {
            String chunkContent = chunks.get(i);
+            // 使用UUID生成唯一的chunk ID，确保格式正确
+            String chunkId = UUID.randomUUID().toString();

            // 创建向量文档
            Document vectorDoc = new Document(
+                    chunkId,
                    chunkContent,
                    Map.of(
-                            "doc_id", docId.toString(),
-                            "project_id", parentDoc.getProjectId(),
-                            "timeline_node_id", parentDoc.getTimelineNodeId() != null ? parentDoc.getTimelineNodeId() : "",
+                            "project_id", parentDoc.getProjectId() != null ? parentDoc.getProjectId().toString() : "",
+                            "timeline_node_id", parentDoc.getTimelineNodeId() != null ? parentDoc.getTimelineNodeId().toString() : "",
                            "chunk_index", i,
                            "chunk_total", chunks.size(),
-                            "title", parentDoc.getTitle(),
-                            "source_type", parentDoc.getSourceType(),
+                            "chunk_parent_id", parentId,
+                            "title", parentDoc.getTitle() != null ? parentDoc.getTitle() : "",
+                            "source_type", parentDoc.getSourceType() != null ? parentDoc.getSourceType() : "",
                            "status", "active"
                    )
            );
@@ -199,7 +202,7 @@ public class DocumentProcessor {
                documentMapper.updateById(parentDoc);
            }

-            log.debug("存储切片: {}/{}, docId: {}", i + 1, chunks.size(), docId);
+            log.debug("存储切片: {}/{}, parentId: {}, chunkId: {}", i + 1, chunks.size(), parentId, chunkId);
        }
    }

--- a/src/main/resources/application-dev.yaml
+++ b/src/main/resources/application-dev.yaml
@@ -46,14 +46,21 @@ spring:

  ai:
    openai:
-      api-key: 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
-      base-url: https://sg1.proxy.yinlihupo.cc/proxy/https://openrouter.ai/api
+      ##嵌入式模型
      embedding:
        options:
-          model: qwen/qwen3-embedding-8b
+          model: text-embedding-v4
+        base-url: https://dashscope.aliyuncs.com/compatible-mode
+        api-key: sk-85c3ccc7c63747c485f9699c90f1972f
+      ##聊天模型
      chat:
        options:
          model: google/gemini-3.1-pro-preview
+        api-key: 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
+        base-url: https://sg1.proxy.yinlihupo.cc/proxy/https://openrouter.ai/api
+      ##其他模型的apikey
+      base-url: https://sg1.proxy.yinlihupo.cc/proxy/https://openrouter.ai/api
+      api-key: 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

 # MinIO 对象存储配置
 minio:
--- a/src/main/resources/mapper/AiDocumentMapper.xml
+++ b/src/main/resources/mapper/AiDocumentMapper.xml
@@ -5,54 +5,52 @@
    <!-- 获取项目文档列表 -->
    <select id="selectProjectDocuments" resultType="cn.yinlihupo.domain.vo.KbDocumentVO">
        SELECT
-            ad.id,
-            ad.doc_id as docId,
-            ad.title,
-            ad.doc_type as docType,
-            ad.file_type as fileType,
-            ad.file_size as fileSize,
-            ad.file_path as filePath,
-            ad.source_type as sourceType,
-            ad.chunk_total as chunkCount,
-            ad.status,
+            vs.id,
+            vs.title,
+            vs.doc_type as docType,
+            vs.file_type as fileType,
+            vs.file_size as fileSize,
+            vs.file_path as filePath,
+            vs.source_type as sourceType,
+            vs.chunk_total as chunkCount,
+            vs.status,
            su.real_name as createByName,
-            ad.create_time as createTime
-        FROM ai_document ad
-        LEFT JOIN sys_user su ON ad.create_by = su.id
-        WHERE ad.project_id = #{projectId}
-          AND ad.deleted = 0
-          AND ad.chunk_parent_id IS NULL
-        ORDER BY ad.create_time DESC
+            vs.create_time as createTime
+        FROM vector_store vs
+        LEFT JOIN sys_user su ON vs.create_by = su.id
+        WHERE vs.project_id = #{projectId}
+          AND vs.deleted = 0
+          AND vs.chunk_parent_id IS NULL
+        ORDER BY vs.create_time DESC
    </select>

-    <!-- 根据docId查询文档 -->
+    <!-- 根据id查询文档 -->
    <select id="selectByDocId" resultType="cn.yinlihupo.domain.entity.AiDocument">
        SELECT *
-        FROM ai_document
-        WHERE doc_id = #{docId}
+        FROM vector_store
+        WHERE id = #{docId}
          AND deleted = 0
        LIMIT 1
    </select>

-    <!-- 根据docId删除文档 -->
+    <!-- 根据id删除文档 -->
    <delete id="deleteByDocId">
-        UPDATE ai_document
+        UPDATE vector_store
        SET deleted = 1,
            update_time = NOW()
-        WHERE doc_id = #{docId}
+        WHERE id = #{docId}
    </delete>

    <!-- 批量查询引用文档信息 -->
    <select id="selectReferencedDocs" resultType="cn.yinlihupo.domain.vo.ReferencedDocVO">
        SELECT
            id,
-            doc_id as docId,
            title,
            doc_type as docType,
            file_type as fileType,
            source_type as sourceType,
            LEFT(content, 200) as content
-        FROM ai_document
+        FROM vector_store
        WHERE id IN
        <foreach collection="docIds" item="id" open="(" separator="," close=")">
            #{id}
@@ -63,38 +61,38 @@
    <!-- 获取父文档的分块数量 -->
    <select id="selectChunkCount" resultType="java.lang.Integer">
        SELECT COUNT(*)
-        FROM ai_document
+        FROM vector_store
        WHERE chunk_parent_id = #{docId}
          AND deleted = 0
    </select>

    <!-- 更新文档状态 -->
    <update id="updateStatus">
-        UPDATE ai_document
+        UPDATE vector_store
        SET status = #{status},
            update_time = NOW()
-        WHERE doc_id = #{docId}
+        WHERE id = #{docId}
    </update>

    <!-- 更新文档错误信息 -->
    <update id="updateErrorMessage">
-        UPDATE ai_document
+        UPDATE vector_store
        SET error_message = #{errorMessage},
            status = 'error',
            update_time = NOW()
-        WHERE doc_id = #{docId}
+        WHERE id = #{docId}
    </update>

    <!-- 增加文档查看次数 -->
    <update id="incrementViewCount">
-        UPDATE ai_document
+        UPDATE vector_store
        SET view_count = view_count + 1
        WHERE id = #{id}
    </update>

    <!-- 增加文档查询次数 -->
    <update id="incrementQueryCount">
-        UPDATE ai_document
+        UPDATE vector_store
        SET query_count = query_count + 1,
            last_queried_at = NOW()
        WHERE id = #{id}