feat: 添加AI问题生成功能，支持配置生成数量并在前端展示生成的问题

config/config.yaml

@@ -518,6 +518,34 @@ conversation:
       }
     ]
 
+  generate_questions_prompt: |
+    你是一个专业的问题生成助手。你的任务是根据给定的【主要内容】生成用户可能会问的相关问题。
+
+    {{.Context}}
+    ## 主要内容（请基于此内容生成问题）
+    文档名称：{{.DocName}}
+    文档内容：
+    {{.Content}}
+
+    ## 核心要求
+    - 生成的问题必须与【主要内容】直接相关
+    - 问题中禁止使用任何代词或指代词（如"它"、"这个"、"该文档"、"本文"、"文中"、"其"等），必须用具体名称替代
+    - 问题必须是完整独立的，脱离上下文也能被理解
+    - 问题应该是用户在实际场景中可能会提出的自然问题
+    - 问题应该多样化，覆盖内容的不同方面
+    - 每个问题应该简洁明了，长度控制在30字以内
+    - 生成的问题数量为 {{.QuestionCount}} 个
+
+    ## 问题类型建议
+    - 定义类：什么是...？...是什么？
+    - 原因类：为什么...？...的原因是什么？
+    - 方法类：如何...？怎样...？
+    - 比较类：...和...有什么区别？
+    - 应用类：...可以用于什么场景？
+
+    ## 输出格式
+    直接输出问题列表，每行一个问题，不要有序号或其他前缀。
+
 # 知识库配置
 knowledge_base:
   chunk_size: 512
@@ -617,4 +645,4 @@ web_search:
 # 租户配置
 tenant:
   # 是否启用跨租户访问功能（内网环境可开启）
-  enable_cross_tenant_access: true
+  enable_cross_tenant_access: false

frontend/src/api/initialization/index.ts

@@ -107,6 +107,10 @@ export interface KBModelConfigRequest {
         nodes: Node[]
         relations: Relation[]
     }
+    questionGeneration?: {
+        enabled: boolean
+        questionCount: number
+    }
 }
 
 export function updateKBConfig(kbId: string, config: KBModelConfigRequest): Promise<any> {

frontend/src/components/doc-content.vue

@@ -185,6 +185,32 @@ const getChunkMeta = (item: any) => {
   return parts.join(' · ');
 };
 
+// 解析生成的问题
+const getGeneratedQuestions = (item: any): string[] => {
+  if (!item || !item.metadata) return [];
+  try {
+    const metadata = typeof item.metadata === 'string' ? JSON.parse(item.metadata) : item.metadata;
+    return metadata.generated_questions || [];
+  } catch {
+    return [];
+  }
+};
+
+// 展开状态管理
+const expandedChunks = ref<Set<number>>(new Set());
+
+const toggleQuestions = (index: number) => {
+  if (expandedChunks.value.has(index)) {
+    expandedChunks.value.delete(index);
+  } else {
+    expandedChunks.value.add(index);
+  }
+  // 触发响应式更新
+  expandedChunks.value = new Set(expandedChunks.value);
+};
+
+const isExpanded = (index: number) => expandedChunks.value.has(index);
+
 const downloadFile = () => {
   downKnowledgeDetails(props.details.id)
     .then((result) => {
@@ -272,17 +298,44 @@ const handleDetailsScroll = () => {
       
       <div v-if="details.md.length == 0" class="no_content">{{ $t('common.noData') }}</div>
       <div v-else class="chunk-list">
-        <div 
-          class="chunk-item" 
+        <div class="chunk-item" 
           v-for="(item, index) in details.md" 
           :key="index"
           :class="getChunkClass(index)"
         >
           <div class="chunk-header">
             <span class="chunk-index">{{ $t('knowledgeBase.segment') || '片段' }} {{ index + 1 }}</span>
-            <span class="chunk-meta">{{ getChunkMeta(item) }}</span>
+            <div class="chunk-header-right">
+              <t-tag 
+                v-if="getGeneratedQuestions(item).length > 0" 
+                size="small" 
+                theme="success" 
+                variant="light"
+              >
+                {{ $t('knowledgeBase.questions') || '问题' }} {{ getGeneratedQuestions(item).length }}
+              </t-tag>
+              <span class="chunk-meta">{{ getChunkMeta(item) }}</span>
+            </div>
           </div>
           <div class="md-content" v-html="processMarkdown(item.content)"></div>
+          
+          <!-- 生成的问题展示 -->
+          <div v-if="getGeneratedQuestions(item).length > 0" class="questions-section">
+            <div class="questions-toggle" @click="toggleQuestions(index)">
+              <t-icon :name="isExpanded(index) ? 'chevron-down' : 'chevron-right'" size="14px" />
+              <span>{{ $t('knowledgeBase.generatedQuestions') || '生成的问题' }} ({{ getGeneratedQuestions(item).length }})</span>
+            </div>
+            <div v-show="isExpanded(index)" class="questions-list">
+              <div 
+                v-for="(question, qIndex) in getGeneratedQuestions(item)" 
+                :key="qIndex" 
+                class="question-item"
+              >
+                <t-icon name="help-circle" size="14px" class="question-icon" />
+                <span>{{ question }}</span>
+              </div>
+            </div>
+          </div>
         </div>
       </div>
       
@@ -493,12 +546,65 @@ const handleDetailsScroll = () => {
     letter-spacing: 0.5px;
   }
   
+  .chunk-header-right {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+  }
+  
   .chunk-meta {
     color: #00000066;
     font-size: 11px;
   }
 }
 
+// 生成的问题样式
+.questions-section {
+  margin-top: 12px;
+  padding-top: 10px;
+  border-top: 1px dashed #e0e0e0;
+}
+
+.questions-toggle {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+  cursor: pointer;
+  color: #059669;
+  font-size: 12px;
+  font-weight: 500;
+  padding: 4px 0;
+  transition: color 0.2s ease;
+  
+  &:hover {
+    color: #07c05f;
+  }
+}
+
+.questions-list {
+  margin-top: 8px;
+  padding-left: 4px;
+}
+
+.question-item {
+  display: flex;
+  align-items: flex-start;
+  gap: 8px;
+  padding: 6px 8px;
+  margin-bottom: 4px;
+  background: #f0fdf4;
+  border-radius: 4px;
+  font-size: 13px;
+  color: #1d2129;
+  line-height: 1.5;
+  
+  .question-icon {
+    color: #059669;
+    flex-shrink: 0;
+    margin-top: 2px;
+  }
+}
+
 .md-content {
   word-break: break-word;
   line-height: 1.6;

frontend/src/i18n/locales/en-US.ts

@@ -92,6 +92,8 @@ export default {
     createTime: 'Create Time',
     characters: 'chars',
     segment: 'Segment',
+    questions: 'Questions',
+    generatedQuestions: 'Generated Questions',
     notInitialized: 'Knowledge base is not initialized. Please configure models in settings before uploading files',
     getInfoFailed: 'Failed to get knowledge base information, file upload is not possible',
     missingId: 'Knowledge base ID is missing',
@@ -1001,7 +1003,13 @@ export default {
     },
     advanced: {
       title: 'Advanced Settings',
-      description: 'Configure multimodal features',
+      description: 'Configure question generation, multimodal features',
+      questionGeneration: {
+        label: 'AI Question Generation',
+        description: 'Generate related questions for each chunk using LLM during document parsing to improve retrieval recall. Enabling this will increase document parsing time.',
+        countLabel: 'Question Count',
+        countDescription: 'Number of questions to generate per document chunk (1-10)',
+      },
       multimodal: {
         label: 'Multimodal Feature',
         description: 'Enable understanding of multimodal content such as images and videos',

frontend/src/i18n/locales/ru-RU.ts

@@ -92,6 +92,8 @@ export default {
     createTime: 'Время создания',
     characters: 'символов',
     segment: 'Фрагмент',
+    questions: 'Вопросы',
+    generatedQuestions: 'Сгенерированные вопросы',
     notInitialized: 'База знаний не инициализирована. Пожалуйста, настройте модели в разделе настроек перед загрузкой файлов',
     getInfoFailed: 'Не удалось получить информацию о базе знаний, загрузка файла невозможна',
     missingId: 'Отсутствует ID базы знаний',
@@ -1095,7 +1097,13 @@ export default {
     },
     advanced: {
       title: 'Расширенные настройки',
-      description: 'Настройте мультимодальные возможности',
+      description: 'Настройте генерацию вопросов и мультимодальные возможности',
+      questionGeneration: {
+        label: 'AI генерация вопросов',
+        description: 'Генерация связанных вопросов для каждого фрагмента с помощью LLM при парсинге документа для улучшения полноты поиска. Включение увеличит время парсинга документа.',
+        countLabel: 'Количество вопросов',
+        countDescription: 'Количество вопросов для генерации на фрагмент документа (1-10)',
+      },
       multimodal: {
         label: 'Мультимодальная функция',
         description: 'Включите понимание мультимедийного контента, такого как изображения и видео',

frontend/src/i18n/locales/zh-CN.ts

@@ -91,6 +91,8 @@ export default {
     createTime: "创建时间",
     characters: "字符",
     segment: "片段",
+    questions: "问题",
+    generatedQuestions: "生成的问题",
     docActionUnsupported: "当前知识库类型不支持该操作",
     notInitialized:
       "该知识库尚未完成初始化配置，请先前往设置页面配置模型信息后再上传文件",
@@ -1334,7 +1336,13 @@ export default {
     },
     advanced: {
       title: "高级设置",
-      description: "配置多模态等高级功能",
+      description: "配置问题生成、多模态等高级功能",
+      questionGeneration: {
+        label: "AI 问题生成",
+        description: "解析文档时调用大模型为每个分块生成相关问题，提高检索召回率。启用后会增加文档解析耗时。",
+        countLabel: "生成问题数量",
+        countDescription: "每个文档分块生成的问题数量（1-10）",
+      },
       multimodal: {
         label: "多模态功能",
         description: "启用图片、视频等多模态内容的理解能力",

frontend/src/views/knowledge/KnowledgeBaseEditorModal.vue

@@ -144,8 +144,10 @@
                     ref="advancedSettingsRef"
                     v-if="formData"
                     :multimodal="formData.multimodalConfig"
+                    :question-generation="formData.questionGenerationConfig"
                     :all-models="allModels"
                     @update:multimodal="handleMultimodalUpdate"
+                    @update:question-generation="handleQuestionGenerationUpdate"
                   />
                 </div>
               </div>
@@ -296,6 +298,10 @@ const initFormData = (type: 'document' | 'faq' = 'document') => {
         type: string
       }>
     },
+    questionGenerationConfig: {
+      enabled: false,
+      questionCount: 3
+    },
   }
 }
 
@@ -377,6 +383,10 @@ const loadKBData = async () => {
         })),
         relations: kb.extract_config?.relations || []
       },
+      questionGenerationConfig: {
+        enabled: kb.question_generation_config?.enabled || false,
+        questionCount: kb.question_generation_config?.question_count || 3
+      },
     }
   } catch (error) {
     console.error('Failed to load knowledge base data:', error)
@@ -406,6 +416,12 @@ const handleMultimodalUpdate = (config: any) => {
   }
 }
 
+const handleQuestionGenerationUpdate = (config: any) => {
+  if (formData.value) {
+    formData.value.questionGenerationConfig = { ...config }
+  }
+}
+
 const handleNodeExtractUpdate = (config: any) => {
   if (formData.value) {
     formData.value.nodeExtractConfig = { ...config }
@@ -514,6 +530,14 @@ const buildSubmitData = () => {
     }
   }
 
+  // 添加问题生成配置
+  if (formData.value.questionGenerationConfig?.enabled) {
+    data.question_generation_config = {
+      enabled: true,
+      question_count: formData.value.questionGenerationConfig.questionCount || 3
+    }
+  }
+
   if (formData.value.type === 'faq') {
     data.faq_config = {
       index_mode: formData.value.faqConfig?.indexMode || 'question_only',
@@ -598,6 +622,10 @@ const handleSubmit = async () => {
           tags: data.extract_config?.tags || [],
           nodes: data.extract_config?.nodes || [],
           relations: data.extract_config?.relations || []
+        },
+        questionGeneration: {
+          enabled: data.question_generation_config?.enabled || false,
+          questionCount: data.question_generation_config?.question_count || 3
         }
       }
 

frontend/src/views/knowledge/settings/KBAdvancedSettings.vue

@@ -6,6 +6,42 @@
     </div>
 
     <div class="settings-group">
+      <!-- Question Generation feature -->
+      <div class="setting-row">
+        <div class="setting-info">
+          <label>{{ $t('knowledgeEditor.advanced.questionGeneration.label') }}</label>
+          <p class="desc">{{ $t('knowledgeEditor.advanced.questionGeneration.description') }}</p>
+        </div>
+        <div class="setting-control">
+          <t-switch
+            v-model="localQuestionGeneration.enabled"
+            @change="handleQuestionGenerationToggle"
+            size="large"
+          />
+        </div>
+      </div>
+
+      <!-- Question Generation configuration -->
+      <div v-if="localQuestionGeneration.enabled" class="subsection">
+        <div class="setting-row">
+          <div class="setting-info">
+            <label>{{ $t('knowledgeEditor.advanced.questionGeneration.countLabel') }}</label>
+            <p class="desc">{{ $t('knowledgeEditor.advanced.questionGeneration.countDescription') }}</p>
+          </div>
+          <div class="setting-control">
+            <t-input-number
+              v-model="localQuestionGeneration.questionCount"
+              :min="1"
+              :max="10"
+              :step="1"
+              theme="normal"
+              @change="handleQuestionGenerationChange"
+              style="width: 120px;"
+            />
+          </div>
+        </div>
+      </div>
+
       <!-- Multimodal feature -->
       <div class="setting-row">
         <div class="setting-info">
@@ -241,8 +277,14 @@ interface MultimodalConfig {
   }
 }
 
+interface QuestionGenerationConfig {
+  enabled: boolean
+  questionCount: number
+}
+
 interface Props {
   multimodal: MultimodalConfig
+  questionGeneration?: QuestionGenerationConfig
   allModels?: any[]
 }
 
@@ -250,9 +292,13 @@ const props = defineProps<Props>()
 
 const emit = defineEmits<{
   'update:multimodal': [value: MultimodalConfig]
+  'update:questionGeneration': [value: QuestionGenerationConfig]
 }>()
 
 const localMultimodal = ref<MultimodalConfig>({ ...props.multimodal })
+const localQuestionGeneration = ref<QuestionGenerationConfig>(
+  props.questionGeneration || { enabled: false, questionCount: 3 }
+)
 
 const vllmSelectorRef = ref()
 const isMinioEnabled = ref(false)
@@ -287,6 +333,25 @@ watch(() => props.multimodal, (newVal) => {
   localMultimodal.value = { ...newVal }
 }, { deep: true })
 
+watch(() => props.questionGeneration, (newVal) => {
+  if (newVal) {
+    localQuestionGeneration.value = { ...newVal }
+  }
+}, { deep: true })
+
+// Handle question generation toggle
+const handleQuestionGenerationToggle = () => {
+  if (!localQuestionGeneration.value.enabled) {
+    localQuestionGeneration.value.questionCount = 3
+  }
+  emit('update:questionGeneration', localQuestionGeneration.value)
+}
+
+// Handle question generation config change
+const handleQuestionGenerationChange = () => {
+  emit('update:questionGeneration', localQuestionGeneration.value)
+}
+
 // Handle multimodal toggle
 const handleMultimodalToggle = () => {
   // Reset related configuration when multimodal is disabled

internal/application/repository/chunk.go

@@ -60,7 +60,6 @@ func (r *chunkRepository) ListChunksByKnowledgeID(
 ) ([]*types.Chunk, error) {
 	var chunks []*types.Chunk
 	if err := r.db.WithContext(ctx).
-		Select("id, content, knowledge_id, knowledge_base_id, start_at, end_at, chunk_index, is_enabled, chunk_type, parent_chunk_id, image_info").
 		Where("tenant_id = ? AND knowledge_id = ? and chunk_type = ?", tenantID, knowledgeID, "text").
 		Order("chunk_index ASC").
 		Find(&chunks).Error; err != nil {
@@ -85,7 +84,7 @@ func (r *chunkRepository) ListPagedChunksByKnowledgeID(
 
 	baseFilter := func(db *gorm.DB) *gorm.DB {
 		db = db.Where("tenant_id = ? AND knowledge_id = ? AND chunk_type IN (?) AND status in (?)",
-			tenantID, knowledgeID, chunkType, []types.ChunkStatus{types.ChunkStatusIndexed, types.ChunkStatusDefault})
+			tenantID, knowledgeID, chunkType, []int{int(types.ChunkStatusIndexed), int(types.ChunkStatusDefault)})
 		if tagID != "" {
 			db = db.Where("tag_id = ?", tagID)
 		}
@@ -106,7 +105,7 @@ func (r *chunkRepository) ListPagedChunksByKnowledgeID(
 	// Then query the paginated data
 	dataQuery := baseFilter(
 		r.db.WithContext(ctx).
-			Select("id, content, knowledge_id, knowledge_base_id, start_at, end_at, chunk_index, is_enabled, chunk_type, parent_chunk_id, image_info, metadata, tag_id"),
+			Select("id, content, knowledge_id, knowledge_base_id, start_at, end_at, chunk_index, is_enabled, chunk_type, parent_chunk_id, image_info, metadata, tag_id, status"),
 	)
 
 	if err := dataQuery.

internal/application/repository/retriever/postgres/repository.go

@@ -291,7 +291,7 @@ func (g *pgRepository) VectorRetrieve(ctx context.Context,
 			ORDER BY embedding::halfvec(%d) <=> $1::halfvec
 			LIMIT $%d
 		) AS candidates
-		WHERE distance < $%d
+		WHERE distance <= $%d
 		ORDER BY distance ASC
 		LIMIT $%d
 	`, dimension, whereClause, dimension, subqueryLimitParam, thresholdParam, finalLimitParam)

internal/application/service/chat_pipline/preprocess.go

@@ -5,34 +5,24 @@ import (
 	"encoding/json"
 	"regexp"
 	"strings"
-	"unicode"
-	"unicode/utf8"
 
 	"github.com/Tencent/WeKnora/internal/config"
 	"github.com/Tencent/WeKnora/internal/models/chat"
 	"github.com/Tencent/WeKnora/internal/types"
 	"github.com/Tencent/WeKnora/internal/types/interfaces"
-	"github.com/yanyiwu/gojieba"
 )
 
 // PluginPreprocess Query preprocessing plugin
 type PluginPreprocess struct {
 	config       *config.Config
-	jieba        *gojieba.Jieba
-	stopwords    map[string]struct{}
 	modelService interfaces.ModelService
 }
 
 // Regular expressions for text cleaning
 var (
-	multiSpaceRegex = regexp.MustCompile(`\s+`)                                 // Multiple spaces
-	urlRegex        = regexp.MustCompile(`https?://\S+`)                        // URLs
-	emailRegex      = regexp.MustCompile(`\b[\w.%+-]+@[\w.-]+\.[a-zA-Z]{2,}\b`) // Email addresses
-	punctRegex      = regexp.MustCompile(`[^\p{L}\p{N}\s]`)                     // Punctuation marks
+	multiSpaceRegex = regexp.MustCompile(`\s+`) // Multiple spaces
 )
 
-const maxProcessedTokens = 12
-
 // NewPluginPreprocess Creates a new query preprocessing plugin
 func NewPluginPreprocess(
 	eventManager *EventManager,
@@ -40,51 +30,15 @@ func NewPluginPreprocess(
 	cleaner interfaces.ResourceCleaner,
 	modelService interfaces.ModelService,
 ) *PluginPreprocess {
-	// Use default dictionary for Jieba tokenizer
-	jieba := gojieba.NewJieba()
-
-	// Load stopwords from built-in stopword library
-	stopwords := loadStopwords()
-
 	res := &PluginPreprocess{
 		config:       config,
-		jieba:        jieba,
-		stopwords:    stopwords,
 		modelService: modelService,
 	}
 
-	// Register resource cleanup function
-	if cleaner != nil {
-		cleaner.RegisterWithName("JiebaPreprocessor", func() error {
-			res.Close()
-			return nil
-		})
-	}
-
 	eventManager.Register(res)
 	return res
 }
 
-// Load stopwords
-func loadStopwords() map[string]struct{} {
-	// Directly use some common stopwords built into Jieba
-	commonStopwords := []string{
-		"的", "了", "和", "是", "在", "我", "你", "他", "她", "它",
-		"这", "那", "什么", "怎么", "如何", "为什么", "哪里", "什么时候",
-		"the", "is", "are", "am", "I", "you", "he", "she", "it", "this",
-		"that", "what", "how", "a", "an", "and", "or", "but", "if", "of",
-		"to", "in", "on", "at", "by", "for", "with", "about", "from",
-		"有", "无", "好", "来", "去", "说", "看", "想", "会", "可以",
-		"吗", "呢", "啊", "吧", "的话", "就是", "只是", "因为", "所以",
-	}
-
-	result := make(map[string]struct{}, len(commonStopwords))
-	for _, word := range commonStopwords {
-		result[word] = struct{}{}
-	}
-	return result
-}
-
 // ActivationEvents Register activation events
 func (p *PluginPreprocess) ActivationEvents() []types.EventType {
 	return []types.EventType{types.PREPROCESS_QUERY}
@@ -107,171 +61,22 @@ func (p *PluginPreprocess) OnEvent(
 		"rewrite_query": rawQuery,
 	})
 
-	normalized := normalizeWhitespace(rawQuery)
-	sanitized := strings.TrimSpace(p.cleanText(normalized))
-	if sanitized == "" {
-		sanitized = normalized
-	}
-
-	var (
-		processed    = sanitized
-		strategy     = "original"
-		tokenPreview string
-		tokenCount   int
-	)
-
-	switch {
-	case containsChineseCharacters(sanitized):
-		segments := p.segmentText(sanitized)
-		tokens := p.selectMeaningfulTokens(segments)
-		tokenCount = len(tokens)
-		if len(tokens) >= 2 {
-			processed = strings.Join(tokens, " ")
-			strategy = "zh_tokens"
-			tokenPreview = strings.Join(tokens, ",")
-		} else {
-			strategy = "fallback_original"
-		}
-	case containsLatinLetters(sanitized):
-		processed = normalizeLatinQuery(sanitized)
-		if processed != sanitized {
-			strategy = "latin_normalize"
-		}
-	default:
-		strategy = "original"
-	}
-
-	if strings.TrimSpace(processed) == "" {
-		processed = rawQuery
-		strategy = "fallback_original"
-	}
+	// Lightweight normalization: just collapse multiple spaces
+	processed := multiSpaceRegex.ReplaceAllString(rawQuery, " ")
+	processed = strings.TrimSpace(processed)
 
 	chatManage.ProcessedQuery = processed
-	chatManage.QueryIntent = p.detectIntentLLM(ctx, chatManage, sanitized)
+	chatManage.QueryIntent = p.detectIntentLLM(ctx, chatManage, processed)
 
 	pipelineInfo(ctx, "Preprocess", "output", map[string]interface{}{
 		"session_id":      chatManage.SessionID,
 		"processed_query": processed,
-		"strategy":        strategy,
-		"token_count":     tokenCount,
-		"token_preview":   tokenPreview,
 		"query_intent":    chatManage.QueryIntent,
 	})
 
 	return next()
 }
 
-// cleanText Basic text cleaning
-func (p *PluginPreprocess) cleanText(text string) string {
-	// Remove URLs
-	text = urlRegex.ReplaceAllString(text, " ")
-
-	// Remove email addresses
-	text = emailRegex.ReplaceAllString(text, " ")
-
-	// Remove excessive spaces
-	text = multiSpaceRegex.ReplaceAllString(text, " ")
-
-	// Remove punctuation marks
-	text = punctRegex.ReplaceAllString(text, " ")
-
-	// Trim leading and trailing spaces
-	text = strings.TrimSpace(text)
-
-	return text
-}
-
-// segmentText Text tokenization
-func (p *PluginPreprocess) segmentText(text string) []string {
-	// Use Jieba tokenizer for tokenization, using search engine mode
-	segments := p.jieba.CutForSearch(text, true)
-	return segments
-}
-
-// filterStopwords Filter stopwords
-func (p *PluginPreprocess) selectMeaningfulTokens(segments []string) []string {
-	var tokens []string
-	seen := make(map[string]struct{})
-
-	for _, word := range segments {
-		word = strings.TrimSpace(word)
-		if word == "" {
-			continue
-		}
-		if _, stop := p.stopwords[word]; stop {
-			continue
-		}
-		if _, exists := seen[word]; exists {
-			continue
-		}
-		if !isInformativeToken(word) {
-			continue
-		}
-
-		seen[word] = struct{}{}
-		tokens = append(tokens, word)
-		if len(tokens) >= maxProcessedTokens {
-			break
-		}
-	}
-
-	return tokens
-}
-
-// isBlank Check if a string is blank
-func isInformativeToken(token string) bool {
-	if token == "" {
-		return false
-	}
-
-	runeCount := utf8.RuneCountInString(token)
-	if runeCount == 1 {
-		r, _ := utf8.DecodeRuneInString(token)
-		if unicode.IsDigit(r) {
-			return true
-		}
-		if r <= unicode.MaxASCII && unicode.IsLetter(r) {
-			return true
-		}
-		return false
-	}
-
-	return true
-}
-
-// containsChineseCharacters checks if a string contains Chinese characters
-func containsChineseCharacters(text string) bool {
-	for _, r := range text {
-		if unicode.Is(unicode.Han, r) {
-			return true
-		}
-	}
-	return false
-}
-
-// containsLatinLetters checks if a string contains Latin letters
-func containsLatinLetters(text string) bool {
-	for _, r := range text {
-		if r <= unicode.MaxASCII && unicode.IsLetter(r) {
-			return true
-		}
-	}
-	return false
-}
-
-// normalizeWhitespace normalizes whitespace in a string
-func normalizeWhitespace(text string) string {
-	text = strings.TrimSpace(text)
-	return multiSpaceRegex.ReplaceAllString(text, " ")
-}
-
-// normalizeLatinQuery normalizes a Latin query
-func normalizeLatinQuery(text string) string {
-	text = strings.ToLower(text)
-	text = multiSpaceRegex.ReplaceAllString(text, " ")
-	return strings.TrimSpace(text)
-}
-
 // intentResp is a response for intent detection
 type intentResp struct {
 	Intent     string  `json:"intent"`
@@ -349,12 +154,8 @@ func extractJSONBody(text string) string {
 	return "{}"
 }
 
-// Ensure resources are properly released
+// Close Releases resources
 func (p *PluginPreprocess) Close() {
-	if p.jieba != nil {
-		p.jieba.Free()
-		p.jieba = nil
-	}
 }
 
 // ShutdownHandler Returns shutdown function

internal/application/service/knowledge.go

@@ -24,7 +24,6 @@ import (
 	"github.com/Tencent/WeKnora/internal/logger"
 	"github.com/Tencent/WeKnora/internal/models/chat"
 	"github.com/Tencent/WeKnora/internal/models/embedding"
-	"github.com/Tencent/WeKnora/internal/models/utils"
 	"github.com/Tencent/WeKnora/internal/tracing"
 	"github.com/Tencent/WeKnora/internal/types"
 	"github.com/Tencent/WeKnora/internal/types/interfaces"
@@ -283,14 +282,26 @@ func (s *knowledgeService) CreateKnowledgeFromFile(ctx context.Context,
 		enableMultimodelValue = kb.VLMConfig.Enabled
 	}
 
+	// Check question generation config
+	enableQuestionGeneration := false
+	questionCount := 3 // default
+	if kb.QuestionGenerationConfig != nil && kb.QuestionGenerationConfig.Enabled {
+		enableQuestionGeneration = true
+		if kb.QuestionGenerationConfig.QuestionCount > 0 {
+			questionCount = kb.QuestionGenerationConfig.QuestionCount
+		}
+	}
+
 	taskPayload := types.DocumentProcessPayload{
-		TenantID:         tenantID,
-		KnowledgeID:      knowledge.ID,
-		KnowledgeBaseID:  kbID,
-		FilePath:         filePath,
-		FileName:         safeFilename,
-		FileType:         getFileType(safeFilename),
-		EnableMultimodel: enableMultimodelValue,
+		TenantID:                 tenantID,
+		KnowledgeID:              knowledge.ID,
+		KnowledgeBaseID:          kbID,
+		FilePath:                 filePath,
+		FileName:                 safeFilename,
+		FileType:                 getFileType(safeFilename),
+		EnableMultimodel:         enableMultimodelValue,
+		EnableQuestionGeneration: enableQuestionGeneration,
+		QuestionCount:            questionCount,
 	}
 
 	payloadBytes, err := json.Marshal(taskPayload)
@@ -405,12 +416,24 @@ func (s *knowledgeService) CreateKnowledgeFromURL(ctx context.Context,
 		enableMultimodelValue = kb.VLMConfig.Enabled
 	}
 
+	// Check question generation config
+	enableQuestionGeneration := false
+	questionCount := 3 // default
+	if kb.QuestionGenerationConfig != nil && kb.QuestionGenerationConfig.Enabled {
+		enableQuestionGeneration = true
+		if kb.QuestionGenerationConfig.QuestionCount > 0 {
+			questionCount = kb.QuestionGenerationConfig.QuestionCount
+		}
+	}
+
 	taskPayload := types.DocumentProcessPayload{
-		TenantID:         tenantID,
-		KnowledgeID:      knowledge.ID,
-		KnowledgeBaseID:  kbID,
-		URL:              url,
-		EnableMultimodel: enableMultimodelValue,
+		TenantID:                 tenantID,
+		KnowledgeID:              knowledge.ID,
+		KnowledgeBaseID:          kbID,
+		URL:                      url,
+		EnableMultimodel:         enableMultimodelValue,
+		EnableQuestionGeneration: enableQuestionGeneration,
+		QuestionCount:            questionCount,
 	}
 
 	payloadBytes, err := json.Marshal(taskPayload)
@@ -595,12 +618,25 @@ func (s *knowledgeService) createKnowledgeFromPassageInternal(ctx context.Contex
 		// Enqueue passage processing task to Asynq
 		logger.Info(ctx, "Enqueuing passage processing task to Asynq")
 		tenantID := ctx.Value(types.TenantIDContextKey).(uint64)
+
+		// Check question generation config
+		enableQuestionGeneration := false
+		questionCount := 3 // default
+		if kb.QuestionGenerationConfig != nil && kb.QuestionGenerationConfig.Enabled {
+			enableQuestionGeneration = true
+			if kb.QuestionGenerationConfig.QuestionCount > 0 {
+				questionCount = kb.QuestionGenerationConfig.QuestionCount
+			}
+		}
+
 		taskPayload := types.DocumentProcessPayload{
-			TenantID:         tenantID,
-			KnowledgeID:      knowledge.ID,
-			KnowledgeBaseID:  kbID,
-			Passages:         safePassages,
-			EnableMultimodel: false, // 文本段落不支持多模态
+			TenantID:                 tenantID,
+			KnowledgeID:              knowledge.ID,
+			KnowledgeBaseID:          kbID,
+			Passages:                 safePassages,
+			EnableMultimodel:         false, // 文本段落不支持多模态
+			EnableQuestionGeneration: enableQuestionGeneration,
+			QuestionCount:            questionCount,
 		}
 
 		payloadBytes, err := json.Marshal(taskPayload)
@@ -917,10 +953,23 @@ func (s *knowledgeService) processDocumentFromPassage(ctx context.Context,
 	s.processChunks(ctx, kb, knowledge, chunks)
 }
 
+// ProcessChunksOptions contains options for processing chunks
+type ProcessChunksOptions struct {
+	EnableQuestionGeneration bool
+	QuestionCount            int
+}
+
 // processChunks processes chunks and creates embeddings for knowledge content
 func (s *knowledgeService) processChunks(ctx context.Context,
 	kb *types.KnowledgeBase, knowledge *types.Knowledge, chunks []*proto.Chunk,
+	opts ...ProcessChunksOptions,
 ) {
+	// Get options
+	var options ProcessChunksOptions
+	if len(opts) > 0 {
+		options = opts[0]
+	}
+
 	ctx, span := tracing.ContextWithSpan(ctx, "knowledgeService.processChunks")
 	defer span.End()
 	span.SetAttributes(
@@ -969,14 +1018,6 @@ func (s *knowledgeService) processChunks(ctx context.Context,
 
 	logger.Infof(ctx, "Cleanup completed, starting to process new chunks")
 
-	// Generate document summary - 只使用文本类型的 Chunk
-	chatModel, err := s.modelService.GetChatModel(ctx, kb.SummaryModelID)
-	if err != nil {
-		logger.GetLogger(ctx).WithField("error", err).Errorf("processChunks get summary model failed")
-		span.RecordError(err)
-		return
-	}
-
 	// Create chunk objects from proto chunks
 	maxSeq := 0
 
@@ -1118,52 +1159,19 @@ func (s *knowledgeService) processChunks(ctx context.Context,
 		}
 	}
 
-	span.AddEvent("extract summary")
-	summary, err := s.getSummary(ctx, chatModel, knowledge, textChunks)
-	if err != nil {
-		logger.GetLogger(ctx).WithField("knowledge_id", knowledge.ID).
-			WithField("error", err).Errorf("processChunks get summary failed, use first chunk as description")
-		if len(textChunks) > 0 {
-			knowledge.Description = textChunks[0].Content
-		}
-	} else {
-		knowledge.Description = summary
-	}
-	span.SetAttributes(attribute.String("summary", knowledge.Description))
-
-	// 批量索引
-	if strings.TrimSpace(knowledge.Description) != "" && len(textChunks) > 0 {
-		sChunk := &types.Chunk{
-			ID:              uuid.New().String(),
-			TenantID:        knowledge.TenantID,
-			KnowledgeID:     knowledge.ID,
-			KnowledgeBaseID: knowledge.KnowledgeBaseID,
-			Content:         fmt.Sprintf("# 文档名称\n%s\n\n# 摘要\n%s", knowledge.FileName, knowledge.Description),
-			ChunkIndex:      maxSeq + 3, // 使用不冲突的索引方式
-			IsEnabled:       true,
-			CreatedAt:       time.Now(),
-			UpdatedAt:       time.Now(),
-			StartAt:         0,
-			EndAt:           0,
-			ChunkType:       types.ChunkTypeSummary,
-			ParentChunkID:   textChunks[0].ID,
-		}
-		logger.GetLogger(ctx).Infof("Created summary chunk for %s with index %d",
-			sChunk.ParentChunkID, sChunk.ChunkIndex)
-		insertChunks = append(insertChunks, sChunk)
-	}
-
-	// Create index information for each chunk
-	indexInfoList := utils.MapSlice(insertChunks, func(chunk *types.Chunk) *types.IndexInfo {
-		return &types.IndexInfo{
+	// Create index information for each chunk (without generated questions for now)
+	indexInfoList := make([]*types.IndexInfo, 0, len(insertChunks))
+	for _, chunk := range insertChunks {
+		// Add original chunk content to index
+		indexInfoList = append(indexInfoList, &types.IndexInfo{
 			Content:         chunk.Content,
 			SourceID:        chunk.ID,
 			SourceType:      types.ChunkSourceType,
 			ChunkID:         chunk.ID,
 			KnowledgeID:     knowledge.ID,
 			KnowledgeBaseID: knowledge.KnowledgeBaseID,
-		}
-	})
+		})
+	}
 
 	// Initialize retrieval engine
 
@@ -1249,6 +1257,23 @@ func (s *knowledgeService) processChunks(ctx context.Context,
 		logger.GetLogger(ctx).WithField("error", err).Errorf("processChunks update knowledge failed")
 	}
 
+	// Enqueue question generation task if enabled (async, non-blocking)
+	if options.EnableQuestionGeneration && len(textChunks) > 0 {
+		questionCount := options.QuestionCount
+		if questionCount <= 0 {
+			questionCount = 3
+		}
+		if questionCount > 10 {
+			questionCount = 10
+		}
+		s.enqueueQuestionGenerationTask(ctx, knowledge.KnowledgeBaseID, knowledge.ID, questionCount)
+	}
+
+	// Enqueue summary generation task (async, non-blocking)
+	if len(textChunks) > 0 {
+		s.enqueueSummaryGenerationTask(ctx, knowledge.KnowledgeBaseID, knowledge.ID)
+	}
+
 	// Update tenant's storage usage
 	tenantInfo.StorageUsed += totalStorageSize
 	if err := s.tenantRepo.AdjustStorageUsed(ctx, tenantInfo.ID, totalStorageSize); err != nil {
@@ -1354,6 +1379,471 @@ func (s *knowledgeService) getSummary(ctx context.Context,
 	return summary.Content, nil
 }
 
+// enqueueQuestionGenerationTask enqueues an async task for question generation
+func (s *knowledgeService) enqueueQuestionGenerationTask(ctx context.Context,
+	kbID, knowledgeID string, questionCount int,
+) {
+	tenantID := ctx.Value(types.TenantIDContextKey).(uint64)
+	payload := types.QuestionGenerationPayload{
+		TenantID:        tenantID,
+		KnowledgeBaseID: kbID,
+		KnowledgeID:     knowledgeID,
+		QuestionCount:   questionCount,
+	}
+
+	payloadBytes, err := json.Marshal(payload)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to marshal question generation payload: %v", err)
+		return
+	}
+
+	task := asynq.NewTask(types.TypeQuestionGeneration, payloadBytes, asynq.Queue("low"), asynq.MaxRetry(3))
+	info, err := s.task.Enqueue(task)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to enqueue question generation task: %v", err)
+		return
+	}
+	logger.Infof(ctx, "Enqueued question generation task: %s for knowledge: %s", info.ID, knowledgeID)
+}
+
+// enqueueSummaryGenerationTask enqueues an async task for summary generation
+func (s *knowledgeService) enqueueSummaryGenerationTask(ctx context.Context,
+	kbID, knowledgeID string,
+) {
+	tenantID := ctx.Value(types.TenantIDContextKey).(uint64)
+	payload := types.SummaryGenerationPayload{
+		TenantID:        tenantID,
+		KnowledgeBaseID: kbID,
+		KnowledgeID:     knowledgeID,
+	}
+
+	payloadBytes, err := json.Marshal(payload)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to marshal summary generation payload: %v", err)
+		return
+	}
+
+	task := asynq.NewTask(types.TypeSummaryGeneration, payloadBytes, asynq.Queue("low"), asynq.MaxRetry(3))
+	info, err := s.task.Enqueue(task)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to enqueue summary generation task: %v", err)
+		return
+	}
+	logger.Infof(ctx, "Enqueued summary generation task: %s for knowledge: %s", info.ID, knowledgeID)
+}
+
+// ProcessSummaryGeneration handles async summary generation task
+func (s *knowledgeService) ProcessSummaryGeneration(ctx context.Context, t *asynq.Task) error {
+	var payload types.SummaryGenerationPayload
+	if err := json.Unmarshal(t.Payload(), &payload); err != nil {
+		logger.Errorf(ctx, "Failed to unmarshal summary generation payload: %v", err)
+		return nil // Don't retry on unmarshal error
+	}
+
+	logger.Infof(ctx, "Processing summary generation for knowledge: %s", payload.KnowledgeID)
+
+	// Set tenant context
+	ctx = context.WithValue(ctx, types.TenantIDContextKey, payload.TenantID)
+
+	// Get knowledge base
+	kb, err := s.kbService.GetKnowledgeBaseByID(ctx, payload.KnowledgeBaseID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get knowledge base: %v", err)
+		return nil
+	}
+
+	// Get knowledge
+	knowledge, err := s.repo.GetKnowledgeByID(ctx, payload.TenantID, payload.KnowledgeID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get knowledge: %v", err)
+		return nil
+	}
+
+	// Get text chunks for this knowledge
+	chunks, err := s.chunkService.ListChunksByKnowledgeID(ctx, payload.KnowledgeID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get chunks: %v", err)
+		return nil
+	}
+
+	// Filter text chunks only
+	textChunks := make([]*types.Chunk, 0)
+	for _, chunk := range chunks {
+		if chunk.ChunkType == types.ChunkTypeText {
+			textChunks = append(textChunks, chunk)
+		}
+	}
+
+	if len(textChunks) == 0 {
+		logger.Infof(ctx, "No text chunks found for knowledge: %s", payload.KnowledgeID)
+		return nil
+	}
+
+	// Sort chunks by ChunkIndex for proper ordering
+	sort.Slice(textChunks, func(i, j int) bool {
+		return textChunks[i].ChunkIndex < textChunks[j].ChunkIndex
+	})
+
+	// Initialize chat model for summary
+	chatModel, err := s.modelService.GetChatModel(ctx, kb.SummaryModelID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get chat model: %v", err)
+		return fmt.Errorf("failed to get chat model: %w", err)
+	}
+
+	// Generate summary
+	summary, err := s.getSummary(ctx, chatModel, knowledge, textChunks)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to generate summary for knowledge %s: %v", payload.KnowledgeID, err)
+		// Use first chunk content as fallback
+		if len(textChunks) > 0 {
+			summary = textChunks[0].Content
+			if len(summary) > 500 {
+				summary = summary[:500]
+			}
+		}
+	}
+
+	// Update knowledge description
+	knowledge.Description = summary
+	knowledge.UpdatedAt = time.Now()
+	if err := s.repo.UpdateKnowledge(ctx, knowledge); err != nil {
+		logger.Errorf(ctx, "Failed to update knowledge description: %v", err)
+		return fmt.Errorf("failed to update knowledge: %w", err)
+	}
+
+	// Create summary chunk and index it
+	if strings.TrimSpace(summary) != "" {
+		// Get max chunk index
+		maxChunkIndex := 0
+		for _, chunk := range chunks {
+			if chunk.ChunkIndex > maxChunkIndex {
+				maxChunkIndex = chunk.ChunkIndex
+			}
+		}
+
+		summaryChunk := &types.Chunk{
+			ID:              uuid.New().String(),
+			TenantID:        knowledge.TenantID,
+			KnowledgeID:     knowledge.ID,
+			KnowledgeBaseID: knowledge.KnowledgeBaseID,
+			Content:         fmt.Sprintf("# 文档名称\n%s\n\n# 摘要\n%s", knowledge.FileName, summary),
+			ChunkIndex:      maxChunkIndex + 1,
+			IsEnabled:       true,
+			CreatedAt:       time.Now(),
+			UpdatedAt:       time.Now(),
+			StartAt:         0,
+			EndAt:           0,
+			ChunkType:       types.ChunkTypeSummary,
+			ParentChunkID:   textChunks[0].ID,
+		}
+
+		// Save summary chunk
+		if err := s.chunkService.CreateChunks(ctx, []*types.Chunk{summaryChunk}); err != nil {
+			logger.Errorf(ctx, "Failed to create summary chunk: %v", err)
+			return fmt.Errorf("failed to create summary chunk: %w", err)
+		}
+
+		// Index summary chunk
+		tenantInfo, err := s.tenantRepo.GetTenantByID(ctx, payload.TenantID)
+		if err != nil {
+			logger.Errorf(ctx, "Failed to get tenant info: %v", err)
+			return fmt.Errorf("failed to get tenant info: %w", err)
+		}
+		ctx = context.WithValue(ctx, types.TenantInfoContextKey, tenantInfo)
+
+		retrieveEngine, err := retriever.NewCompositeRetrieveEngine(s.retrieveEngine, tenantInfo.RetrieverEngines.Engines)
+		if err != nil {
+			logger.Errorf(ctx, "Failed to init retrieve engine: %v", err)
+			return fmt.Errorf("failed to init retrieve engine: %w", err)
+		}
+
+		embeddingModel, err := s.modelService.GetEmbeddingModel(ctx, kb.EmbeddingModelID)
+		if err != nil {
+			logger.Errorf(ctx, "Failed to get embedding model: %v", err)
+			return fmt.Errorf("failed to get embedding model: %w", err)
+		}
+
+		indexInfo := []*types.IndexInfo{{
+			Content:         summaryChunk.Content,
+			SourceID:        summaryChunk.ID,
+			SourceType:      types.ChunkSourceType,
+			ChunkID:         summaryChunk.ID,
+			KnowledgeID:     knowledge.ID,
+			KnowledgeBaseID: knowledge.KnowledgeBaseID,
+		}}
+
+		if err := retrieveEngine.BatchIndex(ctx, embeddingModel, indexInfo); err != nil {
+			logger.Errorf(ctx, "Failed to index summary chunk: %v", err)
+			return fmt.Errorf("failed to index summary chunk: %w", err)
+		}
+
+		logger.Infof(ctx, "Successfully created and indexed summary chunk for knowledge: %s", payload.KnowledgeID)
+	}
+
+	logger.Infof(ctx, "Successfully generated summary for knowledge: %s", payload.KnowledgeID)
+	return nil
+}
+
+// ProcessQuestionGeneration handles async question generation task
+func (s *knowledgeService) ProcessQuestionGeneration(ctx context.Context, t *asynq.Task) error {
+	ctx, span := tracing.ContextWithSpan(ctx, "knowledgeService.ProcessQuestionGeneration")
+	defer span.End()
+
+	var payload types.QuestionGenerationPayload
+	if err := json.Unmarshal(t.Payload(), &payload); err != nil {
+		logger.Errorf(ctx, "Failed to unmarshal question generation payload: %v", err)
+		return nil // Don't retry on unmarshal error
+	}
+
+	logger.Infof(ctx, "Processing question generation for knowledge: %s", payload.KnowledgeID)
+
+	// Set tenant context
+	ctx = context.WithValue(ctx, types.TenantIDContextKey, payload.TenantID)
+
+	// Get knowledge base
+	kb, err := s.kbService.GetKnowledgeBaseByID(ctx, payload.KnowledgeBaseID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get knowledge base: %v", err)
+		return nil
+	}
+
+	// Get knowledge
+	knowledge, err := s.repo.GetKnowledgeByID(ctx, payload.TenantID, payload.KnowledgeID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get knowledge: %v", err)
+		return nil
+	}
+
+	// Get text chunks for this knowledge
+	chunks, err := s.chunkService.ListChunksByKnowledgeID(ctx, payload.KnowledgeID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get chunks: %v", err)
+		return nil
+	}
+
+	// Filter text chunks only
+	textChunks := make([]*types.Chunk, 0)
+	for _, chunk := range chunks {
+		if chunk.ChunkType == types.ChunkTypeText {
+			textChunks = append(textChunks, chunk)
+		}
+	}
+
+	if len(textChunks) == 0 {
+		logger.Infof(ctx, "No text chunks found for knowledge: %s", payload.KnowledgeID)
+		return nil
+	}
+
+	// Sort chunks by StartAt for context building
+	sort.Slice(textChunks, func(i, j int) bool {
+		return textChunks[i].StartAt < textChunks[j].StartAt
+	})
+
+	// Initialize chat model
+	chatModel, err := s.modelService.GetChatModel(ctx, kb.SummaryModelID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get chat model: %v", err)
+		return fmt.Errorf("failed to get chat model: %w", err)
+	}
+
+	// Initialize embedding model and retrieval engine
+	embeddingModel, err := s.modelService.GetEmbeddingModel(ctx, kb.EmbeddingModelID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get embedding model: %v", err)
+		return fmt.Errorf("failed to get embedding model: %w", err)
+	}
+
+	tenantInfo, err := s.tenantRepo.GetTenantByID(ctx, payload.TenantID)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to get tenant info: %v", err)
+		return fmt.Errorf("failed to get tenant info: %w", err)
+	}
+	ctx = context.WithValue(ctx, types.TenantInfoContextKey, tenantInfo)
+
+	retrieveEngine, err := retriever.NewCompositeRetrieveEngine(s.retrieveEngine, tenantInfo.RetrieverEngines.Engines)
+	if err != nil {
+		logger.Errorf(ctx, "Failed to init retrieve engine: %v", err)
+		return fmt.Errorf("failed to init retrieve engine: %w", err)
+	}
+
+	questionCount := payload.QuestionCount
+	if questionCount <= 0 {
+		questionCount = 3
+	}
+	if questionCount > 10 {
+		questionCount = 10
+	}
+
+	// Generate questions for each chunk with context
+	var indexInfoList []*types.IndexInfo
+	for i, chunk := range textChunks {
+		// Build context from adjacent chunks
+		var prevContent, nextContent string
+		if i > 0 {
+			prevContent = textChunks[i-1].Content
+			// Limit context size
+			if len(prevContent) > 500 {
+				prevContent = prevContent[len(prevContent)-500:]
+			}
+		}
+		if i < len(textChunks)-1 {
+			nextContent = textChunks[i+1].Content
+			// Limit context size
+			if len(nextContent) > 500 {
+				nextContent = nextContent[:500]
+			}
+		}
+
+		questions, err := s.generateQuestionsWithContext(ctx, chatModel, chunk.Content, prevContent, nextContent, knowledge.Title, questionCount)
+		if err != nil {
+			logger.Warnf(ctx, "Failed to generate questions for chunk %s: %v", chunk.ID, err)
+			continue
+		}
+
+		if len(questions) == 0 {
+			continue
+		}
+
+		// Update chunk metadata
+		meta := &types.DocumentChunkMetadata{
+			GeneratedQuestions: questions,
+		}
+		if err := chunk.SetDocumentMetadata(meta); err != nil {
+			logger.Warnf(ctx, "Failed to set document metadata for chunk %s: %v", chunk.ID, err)
+			continue
+		}
+
+		// Update chunk in database
+		if err := s.chunkService.UpdateChunk(ctx, chunk); err != nil {
+			logger.Warnf(ctx, "Failed to update chunk %s: %v", chunk.ID, err)
+			continue
+		}
+
+		// Create index entries for generated questions
+		for j, question := range questions {
+			sourceID := fmt.Sprintf("%s-q%d", chunk.ID, j)
+			indexInfoList = append(indexInfoList, &types.IndexInfo{
+				Content:         question,
+				SourceID:        sourceID,
+				SourceType:      types.ChunkSourceType,
+				ChunkID:         chunk.ID,
+				KnowledgeID:     knowledge.ID,
+				KnowledgeBaseID: knowledge.KnowledgeBaseID,
+			})
+		}
+		logger.Debugf(ctx, "Generated %d questions for chunk %s", len(questions), chunk.ID)
+	}
+
+	// Index generated questions
+	if len(indexInfoList) > 0 {
+		if err := retrieveEngine.BatchIndex(ctx, embeddingModel, indexInfoList); err != nil {
+			logger.Errorf(ctx, "Failed to index generated questions: %v", err)
+			return fmt.Errorf("failed to index questions: %w", err)
+		}
+		logger.Infof(ctx, "Successfully indexed %d generated questions for knowledge: %s", len(indexInfoList), payload.KnowledgeID)
+	}
+
+	return nil
+}
+
+// generateQuestionsWithContext generates questions for a chunk with surrounding context
+func (s *knowledgeService) generateQuestionsWithContext(ctx context.Context,
+	chatModel chat.Chat, content, prevContent, nextContent, docName string, questionCount int,
+) ([]string, error) {
+	if content == "" || questionCount <= 0 {
+		return nil, nil
+	}
+
+	// Build prompt with context
+	prompt := s.config.Conversation.GenerateQuestionsPrompt
+	if prompt == "" {
+		prompt = defaultQuestionGenerationPrompt
+	}
+
+	// Build context section
+	var contextSection string
+	if prevContent != "" || nextContent != "" {
+		contextSection = "## 上下文信息（仅供参考，帮助理解主要内容）\n"
+		if prevContent != "" {
+			contextSection += fmt.Sprintf("【前文】%s\n", prevContent)
+		}
+		if nextContent != "" {
+			contextSection += fmt.Sprintf("【后文】%s\n", nextContent)
+		}
+		contextSection += "\n"
+	}
+
+	// Replace placeholders
+	prompt = strings.ReplaceAll(prompt, "{{.QuestionCount}}", fmt.Sprintf("%d", questionCount))
+	prompt = strings.ReplaceAll(prompt, "{{.Content}}", content)
+	prompt = strings.ReplaceAll(prompt, "{{.Context}}", contextSection)
+	prompt = strings.ReplaceAll(prompt, "{{.DocName}}", docName)
+
+	thinking := false
+	response, err := chatModel.Chat(ctx, []chat.Message{
+		{
+			Role:    "user",
+			Content: prompt,
+		},
+	}, &chat.ChatOptions{
+		Temperature: 0.7,
+		MaxTokens:   512,
+		Thinking:    &thinking,
+	})
+	if err != nil {
+		return nil, fmt.Errorf("failed to generate questions: %w", err)
+	}
+
+	// Parse response
+	lines := strings.Split(response.Content, "\n")
+	questions := make([]string, 0, questionCount)
+	for _, line := range lines {
+		line = strings.TrimSpace(line)
+		if line == "" {
+			continue
+		}
+		line = strings.TrimLeft(line, "0123456789.-*) ")
+		line = strings.TrimSpace(line)
+		if line != "" && len(line) > 5 {
+			questions = append(questions, line)
+			if len(questions) >= questionCount {
+				break
+			}
+		}
+	}
+
+	return questions, nil
+}
+
+// Default prompt for question generation with context support
+const defaultQuestionGenerationPrompt = `你是一个专业的问题生成助手。你的任务是根据给定的【主要内容】生成用户可能会问的相关问题。
+
+{{.Context}}
+## 主要内容（请基于此内容生成问题）
+文档名称：{{.DocName}}
+文档内容：
+{{.Content}}
+
+## 核心要求
+- 生成的问题必须与【主要内容】直接相关
+- 问题中禁止使用任何代词或指代词（如"它"、"这个"、"该文档"、"本文"、"文中"、"其"等），必须用具体名称替代
+- 问题必须是完整独立的，脱离上下文也能被理解
+- 问题应该是用户在实际场景中可能会提出的自然问题
+- 问题应该多样化，覆盖内容的不同方面
+- 每个问题应该简洁明了，长度控制在30字以内
+- 生成的问题数量为 {{.QuestionCount}} 个
+
+## 问题类型建议
+- 定义类：什么是...？...是什么？
+- 原因类：为什么...？...的原因是什么？
+- 方法类：如何...？怎样...？
+- 比较类：...和...有什么区别？
+- 应用类：...可以用于什么场景？
+
+## 输出格式
+直接输出问题列表，每行一个问题，不要有序号或其他前缀。`
+
 // GetKnowledgeFile retrieves the physical file associated with a knowledge entry
 func (s *knowledgeService) GetKnowledgeFile(ctx context.Context, id string) (io.ReadCloser, string, error) {
 	// Get knowledge record
@@ -4051,7 +4541,10 @@ func (s *knowledgeService) ProcessDocument(ctx context.Context, t *asynq.Task) e
 	}
 
 	// 处理chunks（这会更新状态为completed）
-	s.processChunks(ctx, kb, knowledge, chunks)
+	s.processChunks(ctx, kb, knowledge, chunks, ProcessChunksOptions{
+		EnableQuestionGeneration: payload.EnableQuestionGeneration,
+		QuestionCount:            payload.QuestionCount,
+	})
 
 	return nil
 }

internal/config/config.go

@@ -57,6 +57,8 @@ type ConversationConfig struct {
 	SimplifyQueryPromptUser    string         `yaml:"simplify_query_prompt_user"    json:"simplify_query_prompt_user"`
 	ExtractEntitiesPrompt      string         `yaml:"extract_entities_prompt"       json:"extract_entities_prompt"`
 	ExtractRelationshipsPrompt string         `yaml:"extract_relationships_prompt"  json:"extract_relationships_prompt"`
+	// GenerateQuestionsPrompt is used to generate questions for document chunks to improve recall
+	GenerateQuestionsPrompt string `yaml:"generate_questions_prompt" json:"generate_questions_prompt"`
 }
 
 // SummaryConfig 摘要配置

internal/handler/initialization.go

@@ -123,6 +123,12 @@ type KBModelConfigRequest struct {
 		Nodes     []types.GraphNode     `json:"nodes"`
 		Relations []types.GraphRelation `json:"relations"`
 	} `json:"nodeExtract"`
+
+	// 问题生成配置
+	QuestionGeneration struct {
+		Enabled       bool `json:"enabled"`
+		QuestionCount int  `json:"questionCount"`
+	} `json:"questionGeneration"`
 }
 
 // InitializationRequest 初始化请求结构
@@ -192,6 +198,11 @@ type InitializationRequest struct {
 			Type  string `json:"type"`
 		} `json:"relations"`
 	} `json:"nodeExtract"`
+
+	QuestionGeneration struct {
+		Enabled       bool `json:"enabled"`
+		QuestionCount int  `json:"questionCount"`
+	} `json:"questionGeneration"`
 }
 
 // UpdateKBConfig 根据知识库ID和模型ID更新配置（简化版）
@@ -333,6 +344,23 @@ func (h *InitializationHandler) UpdateKBConfig(c *gin.Context) {
 		return
 	}
 
+	// 更新问题生成配置
+	if req.QuestionGeneration.Enabled {
+		questionCount := req.QuestionGeneration.QuestionCount
+		if questionCount <= 0 {
+			questionCount = 3
+		}
+		if questionCount > 10 {
+			questionCount = 10
+		}
+		kb.QuestionGenerationConfig = &types.QuestionGenerationConfig{
+			Enabled:       true,
+			QuestionCount: questionCount,
+		}
+	} else {
+		kb.QuestionGenerationConfig = &types.QuestionGenerationConfig{Enabled: false}
+	}
+
 	// 保存更新后的知识库
 	if err := h.kbRepository.UpdateKnowledgeBase(ctx, kb); err != nil {
 		logger.Error(ctx, "Failed to update knowledge base", err)

internal/router/task.go

@@ -67,6 +67,12 @@ func RunAsynqServer(params AsynqTaskParams) *asynq.ServeMux {
 	// Register FAQ import handler
 	mux.HandleFunc(types.TypeFAQImport, params.KnowledgeService.ProcessFAQImport)
 
+	// Register question generation handler
+	mux.HandleFunc(types.TypeQuestionGeneration, params.KnowledgeService.ProcessQuestionGeneration)
+
+	// Register summary generation handler
+	mux.HandleFunc(types.TypeSummaryGeneration, params.KnowledgeService.ProcessSummaryGeneration)
+
 	go func() {
 		// Start the server
 		if err := params.Server.Run(mux); err != nil {

internal/types/extract_graph.go

@@ -1,9 +1,11 @@
 package types
 
 const (
-	TypeChunkExtract    = "chunk:extract"
-	TypeDocumentProcess = "document:process" // 文档处理任务
-	TypeFAQImport       = "faq:import"       // FAQ导入任务
+	TypeChunkExtract        = "chunk:extract"
+	TypeDocumentProcess     = "document:process"     // 文档处理任务
+	TypeFAQImport           = "faq:import"           // FAQ导入任务
+	TypeQuestionGeneration  = "question:generation"  // 问题生成任务
+	TypeSummaryGeneration   = "summary:generation"   // 摘要生成任务
 )
 
 // ExtractChunkPayload represents the extract chunk task payload
@@ -15,16 +17,18 @@ type ExtractChunkPayload struct {
 
 // DocumentProcessPayload represents the document process task payload
 type DocumentProcessPayload struct {
-	RequestId        string   `json:"request_id"`
-	TenantID         uint64   `json:"tenant_id"`
-	KnowledgeID      string   `json:"knowledge_id"`
-	KnowledgeBaseID  string   `json:"knowledge_base_id"`
-	FilePath         string   `json:"file_path,omitempty"` // 文件路径（文件导入时使用）
-	FileName         string   `json:"file_name,omitempty"` // 文件名（文件导入时使用）
-	FileType         string   `json:"file_type,omitempty"` // 文件类型（文件导入时使用）
-	URL              string   `json:"url,omitempty"`       // URL（URL导入时使用）
-	Passages         []string `json:"passages,omitempty"`  // 文本段落（文本导入时使用）
-	EnableMultimodel bool     `json:"enable_multimodel"`
+	RequestId                string   `json:"request_id"`
+	TenantID                 uint64   `json:"tenant_id"`
+	KnowledgeID              string   `json:"knowledge_id"`
+	KnowledgeBaseID          string   `json:"knowledge_base_id"`
+	FilePath                 string   `json:"file_path,omitempty"`  // 文件路径（文件导入时使用）
+	FileName                 string   `json:"file_name,omitempty"`  // 文件名（文件导入时使用）
+	FileType                 string   `json:"file_type,omitempty"`  // 文件类型（文件导入时使用）
+	URL                      string   `json:"url,omitempty"`        // URL（URL导入时使用）
+	Passages                 []string `json:"passages,omitempty"`   // 文本段落（文本导入时使用）
+	EnableMultimodel         bool     `json:"enable_multimodel"`
+	EnableQuestionGeneration bool     `json:"enable_question_generation"` // 是否启用问题生成
+	QuestionCount            int      `json:"question_count,omitempty"`   // 每个chunk生成的问题数量
 }
 
 // FAQImportPayload represents the FAQ import task payload
@@ -37,6 +41,29 @@ type FAQImportPayload struct {
 	Mode        string            `json:"mode"`
 }
 
+// QuestionGenerationPayload represents the question generation task payload
+type QuestionGenerationPayload struct {
+	TenantID        uint64 `json:"tenant_id"`
+	KnowledgeBaseID string `json:"knowledge_base_id"`
+	KnowledgeID     string `json:"knowledge_id"`
+	QuestionCount   int    `json:"question_count"`
+}
+
+// SummaryGenerationPayload represents the summary generation task payload
+type SummaryGenerationPayload struct {
+	TenantID        uint64 `json:"tenant_id"`
+	KnowledgeBaseID string `json:"knowledge_base_id"`
+	KnowledgeID     string `json:"knowledge_id"`
+}
+
+// ChunkContext represents chunk content with surrounding context
+type ChunkContext struct {
+	ChunkID      string `json:"chunk_id"`
+	Content      string `json:"content"`
+	PrevContent  string `json:"prev_content,omitempty"`  // Previous chunk content for context
+	NextContent  string `json:"next_content,omitempty"`  // Next chunk content for context
+}
+
 // PromptTemplateStructured represents the prompt template structured
 type PromptTemplateStructured struct {
 	Description string      `json:"description"`

internal/types/faq.go

@@ -19,6 +19,43 @@ type FAQChunkMetadata struct {
 	Source            string   `json:"source,omitempty"`
 }
 
+// DocumentChunkMetadata 定义文档 Chunk 的元数据结构
+// 用于存储AI生成的问题等增强信息
+type DocumentChunkMetadata struct {
+	// GeneratedQuestions 存储AI为该Chunk生成的相关问题
+	// 这些问题会被独立索引以提高召回率
+	GeneratedQuestions []string `json:"generated_questions,omitempty"`
+}
+
+// DocumentMetadata 解析 Chunk 中的文档元数据
+func (c *Chunk) DocumentMetadata() (*DocumentChunkMetadata, error) {
+	if c == nil || len(c.Metadata) == 0 {
+		return nil, nil
+	}
+	var meta DocumentChunkMetadata
+	if err := json.Unmarshal(c.Metadata, &meta); err != nil {
+		return nil, err
+	}
+	return &meta, nil
+}
+
+// SetDocumentMetadata 设置 Chunk 的文档元数据
+func (c *Chunk) SetDocumentMetadata(meta *DocumentChunkMetadata) error {
+	if c == nil {
+		return nil
+	}
+	if meta == nil {
+		c.Metadata = nil
+		return nil
+	}
+	bytes, err := json.Marshal(meta)
+	if err != nil {
+		return err
+	}
+	c.Metadata = JSON(bytes)
+	return nil
+}
+
 // Normalize 清理空白与重复项
 func (m *FAQChunkMetadata) Normalize() {
 	if m == nil {

internal/types/interfaces/knowledge.go

@@ -99,6 +99,10 @@ type KnowledgeService interface {
 	ProcessDocument(ctx context.Context, t *asynq.Task) error
 	// ProcessFAQImport handles Asynq FAQ import tasks
 	ProcessFAQImport(ctx context.Context, t *asynq.Task) error
+	// ProcessQuestionGeneration handles Asynq question generation tasks
+	ProcessQuestionGeneration(ctx context.Context, t *asynq.Task) error
+	// ProcessSummaryGeneration handles Asynq summary generation tasks
+	ProcessSummaryGeneration(ctx context.Context, t *asynq.Task) error
 }
 
 // KnowledgeRepository defines the interface for knowledge repositories.

internal/types/knowledgebase.go

@@ -65,6 +65,8 @@ type KnowledgeBase struct {
 	ExtractConfig *ExtractConfig `yaml:"extract_config"          json:"extract_config"          gorm:"column:extract_config;type:json"`
 	// FAQConfig stores FAQ specific configuration such as indexing strategy
 	FAQConfig *FAQConfig `yaml:"faq_config"              json:"faq_config"              gorm:"column:faq_config;type:json"`
+	// QuestionGenerationConfig stores question generation configuration for document knowledge bases
+	QuestionGenerationConfig *QuestionGenerationConfig `yaml:"question_generation_config" json:"question_generation_config" gorm:"column:question_generation_config;type:json"`
 	// Creation time of the knowledge base
 	CreatedAt time.Time `yaml:"created_at"              json:"created_at"`
 	// Last updated time of the knowledge base
@@ -180,6 +182,32 @@ type VLMConfig struct {
 	ModelID string `yaml:"model_id" json:"model_id"`
 }
 
+// QuestionGenerationConfig represents the question generation configuration for document knowledge bases
+// When enabled, the system will use LLM to generate questions for each chunk during document parsing
+// These generated questions will be indexed separately to improve recall
+type QuestionGenerationConfig struct {
+	Enabled bool `yaml:"enabled"  json:"enabled"`
+	// Number of questions to generate per chunk (default: 3, max: 10)
+	QuestionCount int `yaml:"question_count" json:"question_count"`
+}
+
+// Value implements the driver.Valuer interface
+func (c QuestionGenerationConfig) Value() (driver.Value, error) {
+	return json.Marshal(c)
+}
+
+// Scan implements the sql.Scanner interface
+func (c *QuestionGenerationConfig) Scan(value interface{}) error {
+	if value == nil {
+		return nil
+	}
+	b, ok := value.([]byte)
+	if !ok {
+		return nil
+	}
+	return json.Unmarshal(b, c)
+}
+
 // Value implements the driver.Valuer interface, used to convert VLMConfig to database value
 func (c VLMConfig) Value() (driver.Value, error) {
 	return json.Marshal(c)

migrations/versioned/000017_add_question_generation_config.down.sql

@@ -0,0 +1,4 @@
+-- Remove question_generation_config column from knowledge_bases table
+
+ALTER TABLE knowledge_bases 
+DROP COLUMN IF EXISTS question_generation_config;

migrations/versioned/000017_add_question_generation_config.up.sql

@@ -0,0 +1,7 @@
+-- Add question_generation_config column to knowledge_bases table
+-- This column stores configuration for AI question generation feature
+-- When enabled, the system generates questions for document chunks to improve recall
+
+ALTER TABLE knowledge_bases 
+ADD COLUMN IF NOT EXISTS question_generation_config JSON NULL 
+COMMENT 'Question generation configuration for document knowledge bases';