feat: 第五轮开发 - 14项未来路线图功能完整实现

W1-W14 全部完成: - W1: 消息搜索 (ILIKE全文检索 + SearchModal) - W2: 对话导出 (JSON/Markdown/TXT三格式) - W3: 记忆时间线 DevTools 可视化 - W4: 通知推送系统 (WebSocket + Browser Notification API) - W5: 定时提醒 (30s轮询 + 重复提醒 + WebSocket推送) - W6: 每日简报 (08:00自动生成: 天气+新闻+提醒+AI摘要) - W7: IoT场景自动化 (规则引擎 10s轮询 + 条件评估 + 场景执行) - W8: 语音输入 (浏览器 Speech Recognition API) - W9: STT服务 (voice-service + whisper.cpp) - W10: TTS服务 (浏览器 Speech Synthesis + edge-tts三档回退) - W11: 文件管理 (上传/下载/缩略图/纯Go bilinear缩放) - W12: 知识库RAG (PostgreSQL tsvector + 文档分块 + 检索) - W13: 多模态 (图片上传+分析: Vision API + 本地Go分析回退) - W14: PWA (Service Worker + 离线页 + install prompt) 总计: 6个Go微服务 + 10+前端组件 + 10+ PostgreSQL表 + 4个后台调度器
2026-05-19 12:01:09 +08:00
parent 78e3f450c2
commit bcf4d4e621
69 changed files with 14599 additions and 150 deletions
@@ -0,0 +1,822 @@
+package store
+
+import (
+	"crypto/rand"
+	"database/sql"
+	"fmt"
+	"log"
+	"strings"
+	"time"
+	"unicode/utf8"
+)
+
+// ========== 模型定义 ==========
+
+// KnowledgeBase 知识库
+type KnowledgeBase struct {
+	ID            string    `json:"id"`
+	UserID        string    `json:"user_id"`
+	Name          string    `json:"name"`
+	Description   string    `json:"description"`
+	DocumentCount int       `json:"document_count"`
+	ChunkCount    int       `json:"chunk_count"`
+	CreatedAt     time.Time `json:"created_at"`
+	UpdatedAt     time.Time `json:"updated_at"`
+}
+
+// KnowledgeDocument 知识库文档
+type KnowledgeDocument struct {
+	ID          string    `json:"id"`
+	KBID        string    `json:"kb_id"`
+	UserID      string    `json:"user_id"`
+	Title       string    `json:"title"`
+	SourceType  string    `json:"source_type"` // "file", "text", "url"
+	SourceRef   string    `json:"source_ref"`  // 文件 ID 或 URL
+	ContentType string    `json:"content_type"` // "text/plain", "text/markdown", "text/html"
+	RawContent  string    `json:"raw_content"`
+	ChunkCount  int       `json:"chunk_count"`
+	CreatedAt   time.Time `json:"created_at"`
+}
+
+// KnowledgeChunk 文档分块
+type KnowledgeChunk struct {
+	ID         string    `json:"id"`
+	DocID      string    `json:"doc_id"`
+	KBID       string    `json:"kb_id"`
+	ChunkIndex int       `json:"chunk_index"`
+	Content    string    `json:"content"`
+	TokenCount int       `json:"token_count"`
+	CreatedAt  time.Time `json:"created_at"`
+}
+
+// SearchChunkResult 搜索结果的块，包含额外上下文信息
+type SearchChunkResult struct {
+	KnowledgeChunk
+	Relevance      float64 `json:"relevance"`
+	DocumentTitle  string  `json:"document_title"`
+	KBName         string  `json:"kb_name"`
+	Headline       string  `json:"headline"`
+}
+
+// ========== KnowledgeStore ==========
+
+// KnowledgeStore 知识库持久化存储
+type KnowledgeStore struct {
+	db *sql.DB
+}
+
+// NewKnowledgeStore 使用已有数据库连接初始化知识库存储并自动建表
+func NewKnowledgeStore(db *sql.DB) (*KnowledgeStore, error) {
+	store := &KnowledgeStore{db: db}
+
+	if err := store.migrate(); err != nil {
+		return nil, fmt.Errorf("知识库表迁移失败: %w", err)
+	}
+
+	log.Println("[KnowledgeStore] 知识库持久化存储已初始化")
+	return store, nil
+}
+
+// migrate 自动创建知识库相关表结构
+func (s *KnowledgeStore) migrate() error {
+	queries := []string{
+		// 知识库表
+		`CREATE TABLE IF NOT EXISTS knowledge_bases (
+			id VARCHAR(64) PRIMARY KEY,
+			user_id VARCHAR(64) NOT NULL,
+			name VARCHAR(255) NOT NULL,
+			description TEXT DEFAULT '',
+			document_count INT DEFAULT 0,
+			chunk_count INT DEFAULT 0,
+			created_at TIMESTAMPTZ DEFAULT NOW(),
+			updated_at TIMESTAMPTZ DEFAULT NOW()
+		)`,
+		`CREATE INDEX IF NOT EXISTS idx_kb_user_id ON knowledge_bases(user_id)`,
+
+		// 文档表
+		`CREATE TABLE IF NOT EXISTS knowledge_documents (
+			id VARCHAR(64) PRIMARY KEY,
+			kb_id VARCHAR(64) NOT NULL REFERENCES knowledge_bases(id) ON DELETE CASCADE,
+			user_id VARCHAR(64) NOT NULL,
+			title VARCHAR(512) NOT NULL,
+			source_type VARCHAR(32) DEFAULT 'text',
+			source_ref VARCHAR(1024) DEFAULT '',
+			content_type VARCHAR(64) DEFAULT 'text/plain',
+			raw_content TEXT DEFAULT '',
+			chunk_count INT DEFAULT 0,
+			created_at TIMESTAMPTZ DEFAULT NOW()
+		)`,
+		`CREATE INDEX IF NOT EXISTS idx_kd_kb_id ON knowledge_documents(kb_id)`,
+		`CREATE INDEX IF NOT EXISTS idx_kd_user_id ON knowledge_documents(user_id)`,
+
+		// 分块表
+		`CREATE TABLE IF NOT EXISTS knowledge_chunks (
+			id VARCHAR(64) PRIMARY KEY,
+			doc_id VARCHAR(64) NOT NULL REFERENCES knowledge_documents(id) ON DELETE CASCADE,
+			kb_id VARCHAR(64) NOT NULL,
+			chunk_index INT NOT NULL,
+			content TEXT NOT NULL,
+			token_count INT DEFAULT 0,
+			tsv TSVECTOR,
+			created_at TIMESTAMPTZ DEFAULT NOW()
+		)`,
+		`CREATE INDEX IF NOT EXISTS idx_kc_doc_id ON knowledge_chunks(doc_id)`,
+		`CREATE INDEX IF NOT EXISTS idx_kc_kb_id ON knowledge_chunks(kb_id)`,
+	}
+
+	for _, q := range queries {
+		if _, err := s.db.Exec(q); err != nil {
+			return fmt.Errorf("迁移SQL执行失败: %w\nSQL: %s", err, q)
+		}
+	}
+
+	// 尝试创建 GIN 索引（可能因权限或扩展问题失败，但不影响功能）
+	_, err := s.db.Exec(`CREATE INDEX IF NOT EXISTS idx_kc_tsv_gin ON knowledge_chunks USING GIN(tsv)`)
+	if err != nil {
+		log.Printf("[KnowledgeStore] ⚠ GIN索引创建失败（将使用ILIKE降级搜索）: %v", err)
+	}
+
+	return nil
+}
+
+// ========== 知识库 CRUD ==========
+
+// CreateKB 创建知识库
+func (s *KnowledgeStore) CreateKB(kb *KnowledgeBase) error {
+	now := time.Now()
+	if kb.CreatedAt.IsZero() {
+		kb.CreatedAt = now
+	}
+	if kb.UpdatedAt.IsZero() {
+		kb.UpdatedAt = now
+	}
+	_, err := s.db.Exec(
+		`INSERT INTO knowledge_bases (id, user_id, name, description, document_count, chunk_count, created_at, updated_at)
+		 VALUES ($1, $2, $3, $4, $5, $6, $7, $8)`,
+		kb.ID, kb.UserID, kb.Name, kb.Description, kb.DocumentCount, kb.ChunkCount, kb.CreatedAt, kb.UpdatedAt,
+	)
+	if err != nil {
+		return fmt.Errorf("创建知识库失败: %w", err)
+	}
+	return nil
+}
+
+// GetKBsByUser 获取用户的所有知识库
+func (s *KnowledgeStore) GetKBsByUser(userID string) ([]KnowledgeBase, error) {
+	rows, err := s.db.Query(
+		`SELECT id, user_id, name, description, document_count, chunk_count, created_at, updated_at
+		 FROM knowledge_bases WHERE user_id = $1
+		 ORDER BY updated_at DESC`,
+		userID,
+	)
+	if err != nil {
+		return nil, fmt.Errorf("查询知识库列表失败: %w", err)
+	}
+	defer rows.Close()
+
+	var kbs []KnowledgeBase
+	for rows.Next() {
+		var kb KnowledgeBase
+		if err := rows.Scan(&kb.ID, &kb.UserID, &kb.Name, &kb.Description,
+			&kb.DocumentCount, &kb.ChunkCount, &kb.CreatedAt, &kb.UpdatedAt); err != nil {
+			return nil, fmt.Errorf("扫描知识库行失败: %w", err)
+		}
+		kbs = append(kbs, kb)
+	}
+	if kbs == nil {
+		kbs = []KnowledgeBase{}
+	}
+	return kbs, rows.Err()
+}
+
+// GetKB 获取单个知识库
+func (s *KnowledgeStore) GetKB(id string) (*KnowledgeBase, error) {
+	var kb KnowledgeBase
+	err := s.db.QueryRow(
+		`SELECT id, user_id, name, description, document_count, chunk_count, created_at, updated_at
+		 FROM knowledge_bases WHERE id = $1`,
+		id,
+	).Scan(&kb.ID, &kb.UserID, &kb.Name, &kb.Description,
+		&kb.DocumentCount, &kb.ChunkCount, &kb.CreatedAt, &kb.UpdatedAt)
+	if err != nil {
+		if err == sql.ErrNoRows {
+			return nil, nil
+		}
+		return nil, fmt.Errorf("查询知识库失败: %w", err)
+	}
+	return &kb, nil
+}
+
+// UpdateKB 更新知识库名称和描述
+func (s *KnowledgeStore) UpdateKB(id string, name, description string) error {
+	_, err := s.db.Exec(
+		`UPDATE knowledge_bases SET name = $1, description = $2, updated_at = NOW() WHERE id = $3`,
+		name, description, id,
+	)
+	if err != nil {
+		return fmt.Errorf("更新知识库失败: %w", err)
+	}
+	return nil
+}
+
+// DeleteKB 删除知识库（级联删除文档和块）
+func (s *KnowledgeStore) DeleteKB(id string) error {
+	_, err := s.db.Exec(`DELETE FROM knowledge_bases WHERE id = $1`, id)
+	if err != nil {
+		return fmt.Errorf("删除知识库失败: %w", err)
+	}
+	return nil
+}
+
+// updateKBStats 更新知识库的统计计数
+func (s *KnowledgeStore) updateKBStats(kbID string) error {
+	_, err := s.db.Exec(
+		`UPDATE knowledge_bases SET
+			document_count = (SELECT COUNT(*) FROM knowledge_documents WHERE kb_id = $1),
+			chunk_count = (SELECT COUNT(*) FROM knowledge_chunks WHERE kb_id = $1),
+			updated_at = NOW()
+		 WHERE id = $1`,
+		kbID,
+	)
+	return err
+}
+
+// ========== 文档 CRUD ==========
+
+// AddDocument 添加文档，返回创建的文档
+func (s *KnowledgeStore) AddDocument(doc *KnowledgeDocument) error {
+	if doc.CreatedAt.IsZero() {
+		doc.CreatedAt = time.Now()
+	}
+	_, err := s.db.Exec(
+		`INSERT INTO knowledge_documents (id, kb_id, user_id, title, source_type, source_ref, content_type, raw_content, chunk_count, created_at)
+		 VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10)`,
+		doc.ID, doc.KBID, doc.UserID, doc.Title, doc.SourceType, doc.SourceRef,
+		doc.ContentType, doc.RawContent, doc.ChunkCount, doc.CreatedAt,
+	)
+	if err != nil {
+		return fmt.Errorf("添加文档失败: %w", err)
+	}
+
+	// 更新知识库统计
+	if err := s.updateKBStats(doc.KBID); err != nil {
+		log.Printf("[KnowledgeStore] 更新知识库统计失败: %v", err)
+	}
+
+	return nil
+}
+
+// GetDocument 获取单个文档
+func (s *KnowledgeStore) GetDocument(id string) (*KnowledgeDocument, error) {
+	var doc KnowledgeDocument
+	err := s.db.QueryRow(
+		`SELECT id, kb_id, user_id, title, source_type, source_ref, content_type, raw_content, chunk_count, created_at
+		 FROM knowledge_documents WHERE id = $1`,
+		id,
+	).Scan(&doc.ID, &doc.KBID, &doc.UserID, &doc.Title, &doc.SourceType, &doc.SourceRef,
+		&doc.ContentType, &doc.RawContent, &doc.ChunkCount, &doc.CreatedAt)
+	if err != nil {
+		if err == sql.ErrNoRows {
+			return nil, nil
+		}
+		return nil, fmt.Errorf("查询文档失败: %w", err)
+	}
+	return &doc, nil
+}
+
+// GetDocumentsByKB 获取知识库中的所有文档
+func (s *KnowledgeStore) GetDocumentsByKB(kbID string) ([]KnowledgeDocument, error) {
+	rows, err := s.db.Query(
+		`SELECT id, kb_id, user_id, title, source_type, source_ref, content_type, raw_content, chunk_count, created_at
+		 FROM knowledge_documents WHERE kb_id = $1
+		 ORDER BY created_at DESC`,
+		kbID,
+	)
+	if err != nil {
+		return nil, fmt.Errorf("查询文档列表失败: %w", err)
+	}
+	defer rows.Close()
+
+	var docs []KnowledgeDocument
+	for rows.Next() {
+		var doc KnowledgeDocument
+		if err := rows.Scan(&doc.ID, &doc.KBID, &doc.UserID, &doc.Title, &doc.SourceType, &doc.SourceRef,
+			&doc.ContentType, &doc.RawContent, &doc.ChunkCount, &doc.CreatedAt); err != nil {
+			return nil, fmt.Errorf("扫描文档行失败: %w", err)
+		}
+		docs = append(docs, doc)
+	}
+	if docs == nil {
+		docs = []KnowledgeDocument{}
+	}
+	return docs, rows.Err()
+}
+
+// UpdateDocumentChunkCount 更新文档的分块计数
+func (s *KnowledgeStore) UpdateDocumentChunkCount(docID string, count int) error {
+	_, err := s.db.Exec(
+		`UPDATE knowledge_documents SET chunk_count = $1 WHERE id = $2`,
+		count, docID,
+	)
+	return err
+}
+
+// DeleteDocument 删除文档（级联删除块）
+func (s *KnowledgeStore) DeleteDocument(id string) error {
+	// 先获取 kb_id 以便后续更新统计
+	var kbID string
+	err := s.db.QueryRow(`SELECT kb_id FROM knowledge_documents WHERE id = $1`, id).Scan(&kbID)
+	if err != nil {
+		if err == sql.ErrNoRows {
+			return nil
+		}
+		return fmt.Errorf("查询文档失败: %w", err)
+	}
+
+	_, err = s.db.Exec(`DELETE FROM knowledge_documents WHERE id = $1`, id)
+	if err != nil {
+		return fmt.Errorf("删除文档失败: %w", err)
+	}
+
+	// 更新知识库统计
+	if err := s.updateKBStats(kbID); err != nil {
+		log.Printf("[KnowledgeStore] 更新知识库统计失败: %v", err)
+	}
+
+	return nil
+}
+
+// ========== 分块操作 ==========
+
+// AddChunk 添加单个分块
+func (s *KnowledgeStore) AddChunk(chunk *KnowledgeChunk) error {
+	if chunk.CreatedAt.IsZero() {
+		chunk.CreatedAt = time.Now()
+	}
+
+	// 尝试使用 to_tsvector('chinese', content) 设置 tsv
+	// 如果中文分词不可用，使用 simple 配置
+	_, err := s.db.Exec(
+		`INSERT INTO knowledge_chunks (id, doc_id, kb_id, chunk_index, content, token_count, tsv, created_at)
+		 VALUES ($1, $2, $3, $4, $5, $6, 
+			CASE WHEN (SELECT count(*) FROM pg_ts_config WHERE cfgname = 'chinese') > 0
+				THEN to_tsvector('chinese', $5)
+				ELSE to_tsvector('simple', $5)
+			END,
+		 $7)`,
+		chunk.ID, chunk.DocID, chunk.KBID, chunk.ChunkIndex, chunk.Content, chunk.TokenCount, chunk.CreatedAt,
+	)
+	if err != nil {
+		// 降级：不使用 tsv
+		_, err = s.db.Exec(
+			`INSERT INTO knowledge_chunks (id, doc_id, kb_id, chunk_index, content, token_count, created_at)
+			 VALUES ($1, $2, $3, $4, $5, $6, $7)`,
+			chunk.ID, chunk.DocID, chunk.KBID, chunk.ChunkIndex, chunk.Content, chunk.TokenCount, chunk.CreatedAt,
+		)
+		if err != nil {
+			return fmt.Errorf("添加分块失败: %w", err)
+		}
+	}
+
+	return nil
+}
+
+// DeleteChunksByDocID 删除文档的所有分块
+func (s *KnowledgeStore) DeleteChunksByDocID(docID string) error {
+	_, err := s.db.Exec(`DELETE FROM knowledge_chunks WHERE doc_id = $1`, docID)
+	return err
+}
+
+// GetChunksByDocID 获取文档的所有分块
+func (s *KnowledgeStore) GetChunksByDocID(docID string) ([]KnowledgeChunk, error) {
+	rows, err := s.db.Query(
+		`SELECT id, doc_id, kb_id, chunk_index, content, token_count, created_at
+		 FROM knowledge_chunks WHERE doc_id = $1
+		 ORDER BY chunk_index ASC`,
+		docID,
+	)
+	if err != nil {
+		return nil, fmt.Errorf("查询分块失败: %w", err)
+	}
+	defer rows.Close()
+
+	var chunks []KnowledgeChunk
+	for rows.Next() {
+		var c KnowledgeChunk
+		if err := rows.Scan(&c.ID, &c.DocID, &c.KBID, &c.ChunkIndex, &c.Content, &c.TokenCount, &c.CreatedAt); err != nil {
+			return nil, fmt.Errorf("扫描分块行失败: %w", err)
+		}
+		chunks = append(chunks, c)
+	}
+	if chunks == nil {
+		chunks = []KnowledgeChunk{}
+	}
+	return chunks, rows.Err()
+}
+
+// ========== 分块逻辑 ==========
+
+// ChunkDocument 将文档分块并存储
+func (s *KnowledgeStore) ChunkDocument(docID string) (int, error) {
+	// 获取文档
+	doc, err := s.GetDocument(docID)
+	if err != nil {
+		return 0, err
+	}
+	if doc == nil {
+		return 0, fmt.Errorf("文档不存在: %s", docID)
+	}
+
+	// 删除旧的分块
+	if err := s.DeleteChunksByDocID(docID); err != nil {
+		return 0, fmt.Errorf("删除旧分块失败: %w", err)
+	}
+
+	// 分块
+	chunks := splitTextIntoChunks(doc.RawContent, 500, 50)
+
+	// 存储分块
+	for i, content := range chunks {
+		chunk := &KnowledgeChunk{
+			ID:         generateUUIDv4(),
+			DocID:      docID,
+			KBID:       doc.KBID,
+			ChunkIndex: i,
+			Content:    content,
+			TokenCount: estimateTokenCount(content),
+		}
+		if err := s.AddChunk(chunk); err != nil {
+			return 0, fmt.Errorf("存储分块 %d 失败: %w", i, err)
+		}
+	}
+
+	// 更新文档的分块计数
+	if err := s.UpdateDocumentChunkCount(docID, len(chunks)); err != nil {
+		log.Printf("[KnowledgeStore] 更新文档分块计数失败: %v", err)
+	}
+
+	// 更新知识库统计
+	if err := s.updateKBStats(doc.KBID); err != nil {
+		log.Printf("[KnowledgeStore] 更新知识库统计失败: %v", err)
+	}
+
+	return len(chunks), nil
+}
+
+// ========== 搜索 ==========
+
+// SearchChunks 在指定知识库中搜索
+func (s *KnowledgeStore) SearchChunks(kbID, query string, limit int) ([]SearchChunkResult, error) {
+	if limit <= 0 {
+		limit = 5
+	}
+
+	// 尝试使用 PostgreSQL 全文搜索
+	results, err := s.searchWithFullText(kbID, query, limit)
+	if err != nil {
+		log.Printf("[KnowledgeStore] 全文搜索失败，降级为ILIKE: %v", err)
+		// 降级为 ILIKE
+		results, err = s.searchWithILike(kbID, query, limit)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	if results == nil {
+		results = []SearchChunkResult{}
+	}
+	return results, nil
+}
+
+// SearchAllKBs 在用户的所有知识库中搜索
+func (s *KnowledgeStore) SearchAllKBs(userID, query string, limit int) ([]SearchChunkResult, error) {
+	if limit <= 0 {
+		limit = 5
+	}
+
+	results, err := s.searchAllWithILike(userID, query, limit)
+	if err != nil {
+		return nil, err
+	}
+
+	if results == nil {
+		results = []SearchChunkResult{}
+	}
+	return results, nil
+}
+
+// searchWithFullText 使用 PostgreSQL ts_rank + plainto_tsquery 搜索
+func (s *KnowledgeStore) searchWithFullText(kbID, query string, limit int) ([]SearchChunkResult, error) {
+	rows, err := s.db.Query(
+		`SELECT kc.id, kc.doc_id, kc.kb_id, kc.chunk_index, kc.content, kc.token_count, kc.created_at,
+				ts_rank(kc.tsv, plainto_tsquery('chinese', $2)) AS relevance,
+				kd.title AS document_title,
+				kb.name AS kb_name
+		 FROM knowledge_chunks kc
+		 JOIN knowledge_documents kd ON kc.doc_id = kd.id
+		 JOIN knowledge_bases kb ON kc.kb_id = kb.id
+		 WHERE kc.kb_id = $1 AND kc.tsv @@ plainto_tsquery('chinese', $2)
+		 ORDER BY relevance DESC
+		 LIMIT $3`,
+		kbID, query, limit,
+	)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	return scanSearchResults(rows)
+}
+
+// searchWithILike 使用 ILIKE 降级搜索
+func (s *KnowledgeStore) searchWithILike(kbID, query string, limit int) ([]SearchChunkResult, error) {
+	// 构建 ILIKE 模式
+	keywords := tokenizeQuery(query)
+	if len(keywords) == 0 {
+		return []SearchChunkResult{}, nil
+	}
+
+	// 对每个关键词构建 ILIKE 条件
+	conditions := make([]string, len(keywords))
+	args := []interface{}{kbID}
+	placeholderIdx := 2
+	for i, kw := range keywords {
+		conditions[i] = fmt.Sprintf("kc.content ILIKE $%d", placeholderIdx)
+		args = append(args, "%"+kw+"%")
+		placeholderIdx++
+	}
+	args = append(args, limit)
+
+	querySQL := fmt.Sprintf(
+		`SELECT kc.id, kc.doc_id, kc.kb_id, kc.chunk_index, kc.content, kc.token_count, kc.created_at,
+				0.0 AS relevance,
+				kd.title AS document_title,
+				kb.name AS kb_name
+		 FROM knowledge_chunks kc
+		 JOIN knowledge_documents kd ON kc.doc_id = kd.id
+		 JOIN knowledge_bases kb ON kc.kb_id = kb.id
+		 WHERE kc.kb_id = $1 AND (%s)
+		 LIMIT $%d`,
+		strings.Join(conditions, " AND "),
+		placeholderIdx,
+	)
+
+	rows, err := s.db.Query(querySQL, args...)
+	if err != nil {
+		return nil, fmt.Errorf("ILIKE搜索失败: %w", err)
+	}
+	defer rows.Close()
+
+	return scanSearchResults(rows)
+}
+
+// searchAllWithILike 跨所有用户知识库使用 ILIKE 搜索
+func (s *KnowledgeStore) searchAllWithILike(userID, query string, limit int) ([]SearchChunkResult, error) {
+	keywords := tokenizeQuery(query)
+	if len(keywords) == 0 {
+		return []SearchChunkResult{}, nil
+	}
+
+	conditions := make([]string, len(keywords))
+	args := []interface{}{userID}
+	placeholderIdx := 2
+	for i, kw := range keywords {
+		conditions[i] = fmt.Sprintf("kc.content ILIKE $%d", placeholderIdx)
+		args = append(args, "%"+kw+"%")
+		placeholderIdx++
+	}
+	args = append(args, limit)
+
+	querySQL := fmt.Sprintf(
+		`SELECT kc.id, kc.doc_id, kc.kb_id, kc.chunk_index, kc.content, kc.token_count, kc.created_at,
+				0.0 AS relevance,
+				kd.title AS document_title,
+				kb.name AS kb_name
+		 FROM knowledge_chunks kc
+		 JOIN knowledge_documents kd ON kc.doc_id = kd.id
+		 JOIN knowledge_bases kb ON kc.kb_id = kb.id
+		 WHERE kb.user_id = $1 AND (%s)
+		 ORDER BY kc.created_at DESC
+		 LIMIT $%d`,
+		strings.Join(conditions, " AND "),
+		placeholderIdx,
+	)
+
+	rows, err := s.db.Query(querySQL, args...)
+	if err != nil {
+		return nil, fmt.Errorf("全知识库ILIKE搜索失败: %w", err)
+	}
+	defer rows.Close()
+
+	return scanSearchResults(rows)
+}
+
+// scanSearchResults 扫描搜索结果
+func scanSearchResults(rows *sql.Rows) ([]SearchChunkResult, error) {
+	var results []SearchChunkResult
+	for rows.Next() {
+		var r SearchChunkResult
+		if err := rows.Scan(&r.ID, &r.DocID, &r.KBID, &r.ChunkIndex, &r.Content,
+			&r.TokenCount, &r.CreatedAt, &r.Relevance, &r.DocumentTitle, &r.KBName); err != nil {
+			return nil, fmt.Errorf("扫描搜索结果行失败: %w", err)
+		}
+		// 生成高亮片段
+		r.Headline = r.Content
+		results = append(results, r)
+	}
+	if results == nil {
+		results = []SearchChunkResult{}
+	}
+	return results, rows.Err()
+}
+
+// ========== 文本分块函数 ==========
+
+// splitTextIntoChunks 将文本按 maxLen 分块，块之间有 overlap 字符重叠
+func splitTextIntoChunks(text string, maxLen int, overlap int) []string {
+	if text == "" {
+		return nil
+	}
+
+	// 按段落分割
+	paragraphs := strings.Split(text, "\n\n")
+	var chunks []string
+	var currentChunk strings.Builder
+
+	for _, para := range paragraphs {
+		para = strings.TrimSpace(para)
+		if para == "" {
+			continue
+		}
+
+		paraLen := utf8.RuneCountInString(para)
+
+		if paraLen <= maxLen {
+			// 如果当前块 + 段落不超过 maxLen，追加到当前块
+			if utf8.RuneCountInString(currentChunk.String()) == 0 {
+				currentChunk.WriteString(para)
+			} else if utf8.RuneCountInString(currentChunk.String())+1+paraLen <= maxLen {
+				currentChunk.WriteString("\n\n")
+				currentChunk.WriteString(para)
+			} else {
+				// 保存当前块，开始新块
+				chunks = append(chunks, currentChunk.String())
+				currentChunk.Reset()
+				currentChunk.WriteString(para)
+			}
+		} else {
+			// 段落超过 maxLen，需要按句子分割
+			// 先保存当前块
+			if currentChunk.Len() > 0 {
+				chunks = append(chunks, currentChunk.String())
+				currentChunk.Reset()
+			}
+
+			// 按句子分割
+			sentences := splitIntoSentences(para)
+			for _, sent := range sentences {
+				sent = strings.TrimSpace(sent)
+				if sent == "" {
+					continue
+				}
+
+				sentLen := utf8.RuneCountInString(sent)
+
+				if sentLen <= maxLen {
+					if utf8.RuneCountInString(currentChunk.String()) == 0 {
+						currentChunk.WriteString(sent)
+					} else if utf8.RuneCountInString(currentChunk.String())+sentLen <= maxLen {
+						currentChunk.WriteString(sent)
+					} else {
+						chunks = append(chunks, currentChunk.String())
+						currentChunk.Reset()
+						currentChunk.WriteString(sent)
+					}
+				} else {
+					// 句子超过 maxLen，按 maxLen 截断
+					if currentChunk.Len() > 0 {
+						chunks = append(chunks, currentChunk.String())
+						currentChunk.Reset()
+					}
+
+					// 按 maxLen 截断，带 overlap
+					runes := []rune(sent)
+					start := 0
+					for start < len(runes) {
+						end := start + maxLen
+						if end > len(runes) {
+							end = len(runes)
+						}
+						chunks = append(chunks, string(runes[start:end]))
+						if end >= len(runes) {
+							break
+						}
+						// 下一块从 end-overlap 开始
+						start = end - overlap
+						if start <= 0 {
+							start = end
+						}
+					}
+				}
+			}
+		}
+	}
+
+	// 保存最后一个块
+	if currentChunk.Len() > 0 {
+		chunks = append(chunks, currentChunk.String())
+	}
+
+	return chunks
+}
+
+// splitIntoSentences 按句子分割文本（中文。！？和英文标点）
+func splitIntoSentences(text string) []string {
+	var sentences []string
+	runes := []rune(text)
+	var current strings.Builder
+
+	for i := 0; i < len(runes); i++ {
+		current.WriteRune(runes[i])
+
+		// 检查句子结束标志
+		if runes[i] == '。' || runes[i] == '！' || runes[i] == '？' ||
+			runes[i] == '!' || runes[i] == '?' ||
+			(runes[i] == '\n' && i+1 < len(runes) && runes[i+1] != '\n') {
+			sentences = append(sentences, current.String())
+			current.Reset()
+		}
+	}
+
+	// 剩余内容
+	if current.Len() > 0 {
+		remaining := strings.TrimSpace(current.String())
+		if remaining != "" {
+			sentences = append(sentences, remaining)
+		}
+	}
+
+	return sentences
+}
+
+// estimateTokenCount 估算 token 数量（中文按每个字符1.2个token，英文按每4个字符1个token）
+func estimateTokenCount(text string) int {
+	runes := []rune(text)
+	total := 0
+	for _, r := range runes {
+		if r >= 0x4e00 && r <= 0x9fff {
+			// 中文字符，约1.2个token
+			total += 1
+		}
+	}
+	// 非中文字符粗略估算：字符数/4
+	nonChinese := len(runes) - total
+	total = int(float64(total)*1.2) + nonChinese/4
+	if total < 1 {
+		total = 1
+	}
+	return total
+}
+
+// tokenizeQuery 将查询字符串分词（简单按空格和标点分割）
+func tokenizeQuery(query string) []string {
+	// 按空格、中文标点、英文标点分割
+	query = strings.TrimSpace(query)
+	if query == "" {
+		return nil
+	}
+
+	// 先用空格分割
+	parts := strings.Fields(query)
+	var tokens []string
+	for _, part := range parts {
+		part = strings.Trim(part, "。！？!?,.;:：；、（）()[]{}《》\"'")
+		if part != "" {
+			tokens = append(tokens, part)
+		}
+	}
+	return tokens
+}
+
+// GenerateUUID 使用 crypto/rand 生成 UUID v4 格式的字符串（导出供其他包使用）
+func GenerateUUID() string {
+	return generateUUIDv4()
+}
+
+// generateUUIDv4 使用 crypto/rand 生成 UUID v4 格式的字符串
+func generateUUIDv4() string {
+	b := make([]byte, 16)
+	if _, err := rand.Read(b); err != nil {
+		// 降级方案：基于时间戳 + 简单随机
+		ts := time.Now().UnixNano()
+		for i := 0; i < 16; i++ {
+			b[i] = byte((ts >> (i * 4)) & 0xFF)
+		}
+	}
+	// 设置 UUID v4 版本位 (version = 4)
+	b[6] = (b[6] & 0x0f) | 0x40
+	// 设置 UUID variant 位 (variant = 10xx)
+	b[8] = (b[8] & 0x3f) | 0x80
+	return fmt.Sprintf("%08x-%04x-%04x-%04x-%012x",
+		b[0:4], b[4:6], b[6:8], b[8:10], b[10:16])
+}