diff --git a/electron/main/ai/agent.ts b/electron/main/ai/agent.ts
index 33d94d2..2d51041 100644
--- a/electron/main/ai/agent.ts
+++ b/electron/main/ai/agent.ts
@@ -9,6 +9,7 @@ import { getAllToolDefinitions, executeToolCalls } from './tools'
 import type { ToolContext, OwnerInfo } from './tools/types'
 import { aiLogger } from './logger'
 import { randomUUID } from 'crypto'
+import { t as i18nT } from '../i18n'
 
 // 思考类标签列表（可按需扩展）
 const THINK_TAGS = ['think', 'analysis', 'reasoning', 'reflection', 'thought', 'thinking']
@@ -311,78 +312,11 @@ export interface PromptConfig {
   responseRules: string
 }
 
-// 国际化内容
-const i18nContent = {
-  'zh-CN': {
-    currentDateIs: '当前日期是',
-    chatTypeDesc: { private: '私聊记录', group: '群聊记录' },
-    chatContext: { private: '对话', group: '群聊' },
-    ownerNote: (displayName: string, platformId: string, chatContext: string) => `当前用户身份：
-- 用户在${chatContext}中的身份是「${displayName}」（platformId: ${platformId}）
-- 当用户提到"我"、"我的"时，指的就是「${displayName}」
-- 查询"我"的发言时，使用 sender_id 参数筛选该成员
-`,
-    memberNotePrivate: `成员查询策略：
-- 私聊只有两个人，可以直接获取成员列表
-- 当用户提到"对方"、"他/她"时，通过 get_group_members 获取另一方信息
-`,
-    memberNoteGroup: `成员查询策略：
-- 当用户提到特定群成员（如"张三说过什么"、"小明的发言"等）时，应先调用 get_group_members 获取成员列表
-- 群成员有三种名称：accountName（原始昵称）、groupNickname（群昵称）、aliases（用户自定义别名）
-- 通过 get_group_members 的 search 参数可以模糊搜索这三种名称
-- 找到成员后，使用其 id 字段作为 search_messages 的 sender_id 参数来获取该成员的发言
-`,
-    timeParamsIntro: '时间参数：按用户提到的精度组合 year/month/day/hour',
-    timeParamExample1: (year: number) => `"10月" → year: ${year}, month: 10`,
-    timeParamExample2: (year: number) => `"10月1号" → year: ${year}, month: 10, day: 1`,
-    timeParamExample3: (year: number) => `"10月1号下午3点" → year: ${year}, month: 10, day: 1, hour: 15`,
-    defaultYearNote: (year: number, prevYear: number) => `未指定年份默认${year}年，若该月份未到则用${prevYear}年`,
-    responseInstruction: '根据用户的问题，选择合适的工具获取数据，然后基于数据给出回答。',
-    responseRulesTitle: '回答要求：',
-    // Fallback 角色定义：适中幽默，允许 B 站/网络热梗与表情
-    fallbackRoleDefinition: (chatType: string) => `你是一个专业但风格轻松的${chatType}记录分析助手。
-你的任务是帮助用户理解和分析他们的${chatType}记录数据，同时可以适度使用 B 站/网络热梗和表情/颜文字活跃气氛，但不影响结论的准确性。`,
-    // Fallback 回答要求：强调严谨优先，适度玩梗
-    fallbackResponseRules: `1. 基于工具返回的数据回答，不要编造信息
-2. 如果数据不足以回答问题，请说明
-3. 回答要简洁明了，使用 Markdown 格式
-4. 可以适度加入 B 站/网络热梗、表情/颜文字（强度适中）
-5. 玩梗不得影响事实准确与结论清晰，避免低俗或冒犯性表达`,
-  },
-  'en-US': {
-    currentDateIs: 'Current date is',
-    chatTypeDesc: { private: 'private chat records', group: 'group chat records' },
-    chatContext: { private: 'conversation', group: 'group chat' },
-    ownerNote: (displayName: string, platformId: string, chatContext: string) => `Current user identity:
-- The user's identity in this ${chatContext} is "${displayName}" (platformId: ${platformId})
-- When the user refers to "I" or "my", it refers to "${displayName}"
-- When querying "my" messages, use the sender_id parameter to filter for this member
-`,
-    memberNotePrivate: `Member query strategy:
-- Private chats only have two participants, so the member list can be directly obtained
-- When the user refers to "the other party" or "he/she", get the other participant's information via get_group_members
-`,
-    memberNoteGroup: `Member query strategy:
-- When the user refers to specific group members (e.g., "what did John say", "Mary's messages"), first call get_group_members to get the member list
-- Group members have three names: accountName (original nickname), groupNickname (group nickname), aliases (user-defined aliases)
-- The search parameter of get_group_members can be used for fuzzy searching these three names
-- Once a member is found, use their id field as the sender_id parameter for search_messages to retrieve their messages
-`,
-    timeParamsIntro: 'Time parameters: combine year/month/day/hour based on user mention',
-    timeParamExample1: (year: number) => `"October" → year: ${year}, month: 10`,
-    timeParamExample2: (year: number) => `"October 1st" → year: ${year}, month: 10, day: 1`,
-    timeParamExample3: (year: number) => `"October 1st 3 PM" → year: ${year}, month: 10, day: 1, hour: 15`,
-    defaultYearNote: (year: number, prevYear: number) =>
-      `If year is not specified, defaults to ${year}. If the month has not yet occurred, ${prevYear} is used.`,
-    responseInstruction:
-      "Based on the user's question, select appropriate tools to retrieve data, then provide an answer based on the data.",
-    responseRulesTitle: 'Response requirements:',
-    fallbackRoleDefinition: (chatType: string) => `You are a professional ${chatType} analysis assistant.
-Your task is to help users understand and analyze their ${chatType} data.`,
-    fallbackResponseRules: `1. Answer based on data returned by tools, do not fabricate information
-2. If data is insufficient to answer, please state so
-3. Keep answers concise and clear, use Markdown format`,
-  },
+// ==================== 国际化辅助（使用 i18next） ====================
+
+/** 获取 Agent 翻译，根据传入的 locale 参数 */
+function agentT(key: string, locale: string, options?: Record<string, unknown>): string {
+  return i18nT(key, { lng: locale, ...options })
 }
 
 /**
@@ -400,9 +334,8 @@ function getLockedPromptSection(
   ownerInfo?: OwnerInfo,
   locale: string = 'zh-CN'
 ): string {
-  const content = i18nContent[locale as keyof typeof i18nContent] || i18nContent['zh-CN']
   const now = new Date()
-  const dateLocale = locale === 'zh-CN' ? 'zh-CN' : 'en-US'
+  const dateLocale = locale.startsWith('zh') ? 'zh-CN' : 'en-US'
   const currentDate = now.toLocaleDateString(dateLocale, {
     year: 'numeric',
     month: 'long',
@@ -411,27 +344,35 @@ function getLockedPromptSection(
   })
 
   const isPrivate = chatType === 'private'
-  const chatContext = content.chatContext[chatType]
+  const chatContext = agentT(`ai.agent.chatContext.${chatType}`, locale)
 
   // Owner 说明（当用户设置了"我是谁"时）
-  const ownerNote = ownerInfo ? content.ownerNote(ownerInfo.displayName, ownerInfo.platformId, chatContext) : ''
+  const ownerNote = ownerInfo
+    ? agentT('ai.agent.ownerNote', locale, {
+        displayName: ownerInfo.displayName,
+        platformId: ownerInfo.platformId,
+        chatContext,
+      })
+    : ''
 
   // 成员说明（私聊只有2人）
-  const memberNote = isPrivate ? content.memberNotePrivate : content.memberNoteGroup
+  const memberNote = isPrivate
+    ? agentT('ai.agent.memberNotePrivate', locale)
+    : agentT('ai.agent.memberNoteGroup', locale)
 
   const year = now.getFullYear()
   const prevYear = year - 1
 
-  return `${content.currentDateIs} ${currentDate}。
+  return `${agentT('ai.agent.currentDateIs', locale)} ${currentDate}。
 ${ownerNote}
 ${memberNote}
-${content.timeParamsIntro}
-- ${content.timeParamExample1(year)}
-- ${content.timeParamExample2(year)}
-- ${content.timeParamExample3(year)}
-${content.defaultYearNote(year, prevYear)}
+${agentT('ai.agent.timeParamsIntro', locale)}
+- ${agentT('ai.agent.timeParamExample1', locale, { year })}
+- ${agentT('ai.agent.timeParamExample2', locale, { year })}
+- ${agentT('ai.agent.timeParamExample3', locale, { year })}
+${agentT('ai.agent.defaultYearNote', locale, { year, prevYear })}
 
-${content.responseInstruction}`
+${agentT('ai.agent.responseInstruction', locale)}`
 }
 
 /**
@@ -439,10 +380,7 @@ ${content.responseInstruction}`
  * 仅在前端未传递 promptConfig 时使用
  */
 function getFallbackRoleDefinition(chatType: 'group' | 'private', locale: string = 'zh-CN'): string {
-  const content = i18nContent[locale as keyof typeof i18nContent] || i18nContent['zh-CN']
-  const chatTypeDesc =
-    chatType === 'private' ? (locale === 'zh-CN' ? '私聊' : 'private chat') : locale === 'zh-CN' ? '群聊' : 'group chat'
-  return content.fallbackRoleDefinition(chatTypeDesc)
+  return agentT(`ai.agent.fallbackRoleDefinition.${chatType}`, locale)
 }
 
 /**
@@ -450,8 +388,7 @@ function getFallbackRoleDefinition(chatType: 'group' | 'private', locale: string
  * 仅在前端未传递 promptConfig 时使用
  */
 function getFallbackResponseRules(locale: string = 'zh-CN'): string {
-  const content = i18nContent[locale as keyof typeof i18nContent] || i18nContent['zh-CN']
-  return content.fallbackResponseRules
+  return agentT('ai.agent.fallbackResponseRules', locale)
 }
 
 /**
@@ -471,8 +408,6 @@ function buildSystemPrompt(
   ownerInfo?: OwnerInfo,
   locale: string = 'zh-CN'
 ): string {
-  const content = i18nContent[locale as keyof typeof i18nContent] || i18nContent['zh-CN']
-
   // 使用用户配置或 fallback
   const roleDefinition = promptConfig?.roleDefinition || getFallbackRoleDefinition(chatType, locale)
   const responseRules = promptConfig?.responseRules || getFallbackResponseRules(locale)
@@ -485,7 +420,7 @@ function buildSystemPrompt(
 
 ${lockedSection}
 
-${content.responseRulesTitle}
+${agentT('ai.agent.responseRulesTitle', locale)}
 ${responseRules}`
 }
 
@@ -634,7 +569,7 @@ export class Agent {
     aiLogger.warn('Agent', '达到最大工具调用轮数', { maxRounds: this.config.maxToolRounds })
     this.messages.push({
       role: 'user',
-      content: '请根据已获取的信息给出回答，不要再调用工具。',
+      content: agentT('ai.agent.answerWithoutTools', this.locale),
     })
 
     const finalResponse = await chat(this.messages, this.config.llmOptions)
@@ -877,7 +812,7 @@ export class Agent {
 
     this.messages.push({
       role: 'user',
-      content: '请根据已获取的信息给出回答，不要再调用工具。',
+      content: agentT('ai.agent.answerWithoutTools', this.locale),
     })
 
     // 最后一轮不带 tools（传入 abortSignal）
@@ -993,7 +928,9 @@ export class Agent {
       // 添加工具结果消息
       this.messages.push({
         role: 'tool',
-        content: result.success ? JSON.stringify(result.result) : `错误: ${result.error}`,
+        content: result.success
+          ? JSON.stringify(result.result)
+          : agentT('ai.agent.toolError', this.locale, { error: result.error }),
         tool_call_id: tc.id,
       })
     }
diff --git a/electron/main/ai/llm/index.ts b/electron/main/ai/llm/index.ts
index d56cc56..4bd1164 100644
--- a/electron/main/ai/llm/index.ts
+++ b/electron/main/ai/llm/index.ts
@@ -23,6 +23,7 @@ import { GeminiService, GEMINI_INFO } from './gemini'
 import { OpenAICompatibleService, OPENAI_COMPATIBLE_INFO } from './openai-compatible'
 import { aiLogger, extractErrorInfo, extractErrorStack } from '../logger'
 import { encryptApiKey, decryptApiKey, isEncrypted } from './crypto'
+import { t } from '../../i18n'
 
 // 导出类型
 export * from './types'
@@ -299,7 +300,7 @@ export function addConfig(config: Omit<AIServiceConfig, 'id' | 'createdAt' | 'up
   const store = loadConfigStore()
 
   if (store.configs.length >= MAX_CONFIG_COUNT) {
-    return { success: false, error: `最多只能添加 ${MAX_CONFIG_COUNT} 个配置` }
+    return { success: false, error: t('llm.maxConfigs', { count: MAX_CONFIG_COUNT }) }
   }
 
   const now = Date.now()
@@ -332,7 +333,7 @@ export function updateConfig(
   const index = store.configs.findIndex((c) => c.id === id)
 
   if (index === -1) {
-    return { success: false, error: '配置不存在' }
+    return { success: false, error: t('llm.configNotFound') }
   }
 
   store.configs[index] = {
@@ -353,7 +354,7 @@ export function deleteConfig(id: string): { success: boolean; error?: string } {
   const index = store.configs.findIndex((c) => c.id === id)
 
   if (index === -1) {
-    return { success: false, error: '配置不存在' }
+    return { success: false, error: t('llm.configNotFound') }
   }
 
   store.configs.splice(index, 1)
@@ -375,7 +376,7 @@ export function setActiveConfig(id: string): { success: boolean; error?: string
   const config = store.configs.find((c) => c.id === id)
 
   if (!config) {
-    return { success: false, error: '配置不存在' }
+    return { success: false, error: t('llm.configNotFound') }
   }
 
   store.activeConfigId = id
@@ -542,7 +543,7 @@ export async function chat(
   const service = getCurrentLLMService()
   if (!service) {
     aiLogger.error('LLM', '服务未配置')
-    throw new Error('LLM 服务未配置，请先在设置中配置 API Key')
+    throw new Error(t('llm.notConfigured'))
   }
 
   try {
@@ -597,7 +598,7 @@ export async function* chatStream(messages: ChatMessage[], options?: ChatOptions
   const service = getCurrentLLMService()
   if (!service) {
     aiLogger.error('LLM', '服务未配置（流式）')
-    throw new Error('LLM 服务未配置，请先在设置中配置 API Key')
+    throw new Error(t('llm.notConfigured'))
   }
 
   let chunkCount = 0
diff --git a/electron/main/ai/summary/index.ts b/electron/main/ai/summary/index.ts
index c9b2f03..b8ff5d9 100644
--- a/electron/main/ai/summary/index.ts
+++ b/electron/main/ai/summary/index.ts
@@ -11,6 +11,7 @@ import Database from 'better-sqlite3'
 import { chat } from '../llm'
 import { getDbPath, openDatabase } from '../../database/core'
 import { aiLogger } from '../logger'
+import { t } from '../../i18n'
 
 /** 最小消息数阈值（少于此数量不生成摘要） */
 const MIN_MESSAGE_COUNT = 3
@@ -122,6 +123,7 @@ function getSummaryLengthLimit(messageCount: number): number {
 
 /**
  * 判断消息是否有意义（用于过滤）
+ * 支持中英文内容过滤
  */
 function isValidMessage(content: string): boolean {
   const trimmed = content.trim()
@@ -129,19 +131,45 @@ function isValidMessage(content: string): boolean {
   // 过滤空内容
   if (!trimmed) return false
 
-  // 过滤单字/双字无意义回复
+  // 过滤单字/双字无意义回复（中文）
   if (trimmed.length <= 2) {
-    // 允许一些有意义的短词
-    const meaningfulShort = ['好的', '不是', '是的', '可以', '不行', '好吧', '明白', '知道', '同意']
-    if (!meaningfulShort.includes(trimmed)) return false
+    const meaningfulShortZh = ['好的', '不是', '是的', '可以', '不行', '好吧', '明白', '知道', '同意']
+    if (!meaningfulShortZh.includes(trimmed)) return false
   }
 
+  // 过滤短无意义回复（英文，不区分大小写）
+  const lowerTrimmed = trimmed.toLowerCase()
+  const meaninglessShortEn = [
+    'ok',
+    'k',
+    'yes',
+    'no',
+    'ya',
+    'yep',
+    'nope',
+    'lol',
+    'haha',
+    'hehe',
+    'hmm',
+    'ah',
+    'oh',
+    'wow',
+    'thx',
+    'ty',
+    'np',
+    'gg',
+    'brb',
+    'idk',
+  ]
+  if (meaninglessShortEn.includes(lowerTrimmed)) return false
+
   // 过滤纯表情消息
   const emojiOnlyPattern = /^[\p{Emoji}\s[\]（）()]+$/u
   if (emojiOnlyPattern.test(trimmed)) return false
 
-  // 过滤占位符文本
+  // 过滤占位符文本（中文 + 英文）
   const placeholders = [
+    // 中文占位符（QQ/微信导出格式）
     '[图片]',
     '[语音]',
     '[视频]',
@@ -153,12 +181,38 @@ function isValidMessage(content: string): boolean {
     '[红包]',
     '[转账]',
     '[撤回消息]',
+    // 英文占位符
+    '[image]',
+    '[voice]',
+    '[video]',
+    '[file]',
+    '[sticker]',
+    '[animated sticker]',
+    '[location]',
+    '[contact]',
+    '[red packet]',
+    '[transfer]',
+    '[recalled message]',
+    '[photo]',
+    '[audio]',
+    '[gif]',
   ]
-  if (placeholders.some((p) => trimmed === p)) return false
+  if (placeholders.some((p) => lowerTrimmed === p.toLowerCase())) return false
 
-  // 过滤系统消息（入群、退群等）
-  const systemPatterns = [/^.*邀请.*加入了群聊$/, /^.*退出了群聊$/, /^.*撤回了一条消息$/, /^你撤回了一条消息$/]
-  if (systemPatterns.some((p) => p.test(trimmed))) return false
+  // 过滤系统消息（中文：入群、退群等）
+  const systemPatternsZh = [/^.*邀请.*加入了群聊$/, /^.*退出了群聊$/, /^.*撤回了一条消息$/, /^你撤回了一条消息$/]
+  if (systemPatternsZh.some((p) => p.test(trimmed))) return false
+
+  // 过滤系统消息（英文）
+  const systemPatternsEn = [
+    /^.*invited.*to the group$/i,
+    /^.*left the group$/i,
+    /^.*recalled a message$/i,
+    /^you recalled a message$/i,
+    /^.*joined the group$/i,
+    /^.*has been removed$/i,
+  ]
+  if (systemPatternsEn.some((p) => p.test(trimmed))) return false
 
   return true
 }
@@ -282,17 +336,14 @@ export async function generateSessionSummary(
     // 2. 获取会话消息
     const sessionData = getSessionMessagesForSummary(dbSessionId, chatSessionId)
     if (!sessionData) {
-      return { success: false, error: '会话不存在或数据库打开失败' }
+      return { success: false, error: t('summary.sessionNotFound') }
     }
 
     // 3. 检查消息数量
     if (sessionData.messageCount < MIN_MESSAGE_COUNT) {
       return {
         success: false,
-        error:
-          locale === 'zh-CN'
-            ? `消息数量少于${MIN_MESSAGE_COUNT}条，无需生成摘要`
-            : `Message count less than ${MIN_MESSAGE_COUNT}, no need to generate summary`,
+        error: t('summary.tooFewMessages', { count: MIN_MESSAGE_COUNT }),
       }
     }
 
@@ -301,10 +352,7 @@ export async function generateSessionSummary(
     if (validMessages.length < MIN_MESSAGE_COUNT) {
       return {
         success: false,
-        error:
-          locale === 'zh-CN'
-            ? `有效消息数量少于${MIN_MESSAGE_COUNT}条，无需生成摘要`
-            : `Valid message count less than ${MIN_MESSAGE_COUNT}, no need to generate summary`,
+        error: t('summary.tooFewValidMessages', { count: MIN_MESSAGE_COUNT }),
       }
     }
 
@@ -362,7 +410,7 @@ export async function generateSessionSummary(
 async function generateDirectSummary(content: string, lengthLimit: number, locale: string): Promise<string> {
   const response = await chat(
     [
-      { role: 'system', content: '你是一个对话摘要专家，擅长用简洁的语言总结对话内容。' },
+      { role: 'system', content: t('summary.systemPromptDirect') },
       { role: 'user', content: buildSummaryPrompt(content, lengthLimit, locale) },
     ],
     {
@@ -391,7 +439,7 @@ async function generateMapReduceSummary(
     const segmentContent = formatMessages(segments[i])
     const response = await chat(
       [
-        { role: 'system', content: '你是一个对话摘要专家，擅长用简洁的语言总结对话内容。' },
+        { role: 'system', content: t('summary.systemPromptDirect') },
         { role: 'user', content: buildSubSummaryPrompt(segmentContent, locale) },
       ],
       {
@@ -409,7 +457,7 @@ async function generateMapReduceSummary(
 
   const mergeResponse = await chat(
     [
-      { role: 'system', content: '你是一个对话摘要专家，擅长将多个摘要合并成一个连贯的总结。' },
+      { role: 'system', content: t('summary.systemPromptMerge') },
       { role: 'user', content: buildMergePrompt(subSummaries, lengthLimit, locale) },
     ],
     {
@@ -447,7 +495,7 @@ export async function generateSessionSummaries(
 
     if (result.success) {
       success++
-    } else if (result.error?.includes('少于') || result.error?.includes('less than')) {
+    } else if (result.error?.includes('少于') || result.error?.includes('less than') || result.error?.includes('few')) {
       skipped++
     } else {
       failed++
@@ -479,20 +527,20 @@ export function checkSessionsCanGenerateSummary(
     const sessionData = getSessionMessagesForSummary(dbSessionId, chatSessionId)
 
     if (!sessionData) {
-      results.set(chatSessionId, { canGenerate: false, reason: '会话不存在' })
+      results.set(chatSessionId, { canGenerate: false, reason: t('summary.sessionNotExist') })
       continue
     }
 
     // 检查原始消息数量
     if (sessionData.messageCount < MIN_MESSAGE_COUNT) {
-      results.set(chatSessionId, { canGenerate: false, reason: '消息太少' })
+      results.set(chatSessionId, { canGenerate: false, reason: t('summary.messagesTooFew') })
       continue
     }
 
     // 预处理：过滤无意义消息
     const validMessages = preprocessMessages(sessionData.messages)
     if (validMessages.length < MIN_MESSAGE_COUNT) {
-      results.set(chatSessionId, { canGenerate: false, reason: '有效消息太少' })
+      results.set(chatSessionId, { canGenerate: false, reason: t('summary.validMessagesTooFew') })
       continue
     }
 
diff --git a/electron/main/ai/tools/index.ts b/electron/main/ai/tools/index.ts
index 722f257..3b6c31b 100644
--- a/electron/main/ai/tools/index.ts
+++ b/electron/main/ai/tools/index.ts
@@ -6,6 +6,7 @@
 import type { ToolDefinition, ToolCall } from '../llm/types'
 import type { ToolRegistry, RegisteredTool, ToolContext, ToolExecutionResult, ToolExecutor } from './types'
 import { isEmbeddingEnabled } from '../rag'
+import { t as i18nT } from '../../i18n'
 
 // 导出类型
 export * from './types'
@@ -44,23 +45,65 @@ export async function ensureToolsInitialized(): Promise<void> {
   return initPromise
 }
 
+/**
+ * 翻译工具定义的 description 和参数 description
+ * 使用 i18next 查找翻译，如果未找到则保留原始文本（中文）
+ *
+ * i18n 键命名规则：
+ * - 工具描述：ai.tools.{toolName}.desc
+ * - 参数描述：ai.tools.{toolName}.params.{paramName}
+ */
+function translateToolDefinition(tool: ToolDefinition): ToolDefinition {
+  const name = tool.function.name
+  const descKey = `ai.tools.${name}.desc`
+  const translatedDesc = i18nT(descKey)
+
+  // 深拷贝并翻译参数描述
+  const translatedProperties: typeof tool.function.parameters.properties = {}
+  for (const [paramName, param] of Object.entries(tool.function.parameters.properties)) {
+    const paramKey = `ai.tools.${name}.params.${paramName}`
+    const translatedParamDesc = i18nT(paramKey)
+    translatedProperties[paramName] = {
+      ...param,
+      // 如果 i18next 返回的是 key 本身，说明没有找到翻译，保留原始文本
+      description: translatedParamDesc !== paramKey ? translatedParamDesc : param.description,
+    }
+  }
+
+  return {
+    type: tool.type,
+    function: {
+      name: tool.function.name,
+      // 如果 i18next 返回的是 key 本身，说明没有找到翻译，保留原始文本
+      description: translatedDesc !== descKey ? translatedDesc : tool.function.description,
+      parameters: {
+        type: tool.function.parameters.type,
+        properties: translatedProperties,
+        required: tool.function.parameters.required,
+      },
+    },
+  }
+}
+
 /**
  * 获取所有已注册的工具定义
  * 根据配置动态过滤工具（如：语义搜索工具仅在启用 Embedding 时可用）
+ * 根据当前 locale 动态翻译工具描述（解决"响应式"陷阱：每次调用时实时翻译）
  * @returns 工具定义数组（用于传给 LLM）
  */
 export async function getAllToolDefinitions(): Promise<ToolDefinition[]> {
   await ensureToolsInitialized()
 
-  const allTools = Array.from(toolRegistry.values()).map((t) => t.definition)
+  const allTools = Array.from(toolRegistry.values()).map((reg) => reg.definition)
 
   // 根据 Embedding 配置决定是否包含语义搜索工具
   const embeddingEnabled = isEmbeddingEnabled()
-  if (!embeddingEnabled) {
-    return allTools.filter((t) => t.function.name !== 'semantic_search_messages')
-  }
+  const filteredTools = embeddingEnabled
+    ? allTools
+    : allTools.filter((tool) => tool.function.name !== 'semantic_search_messages')
 
-  return allTools
+  // 所有 locale 统一走翻译层，确保 locale 文件同构
+  return filteredTools.map(translateToolDefinition)
 }
 
 /**
@@ -87,7 +130,7 @@ export async function executeToolCall(toolCall: ToolCall, context: ToolContext):
     return {
       toolName,
       success: false,
-      error: `工具 "${toolName}" 未注册`,
+      error: i18nT('tools.notRegistered', { toolName }),
     }
   }
 
diff --git a/electron/main/database/migrations.ts b/electron/main/database/migrations.ts
index 7f741a4..8653a12 100644
--- a/electron/main/database/migrations.ts
+++ b/electron/main/database/migrations.ts
@@ -11,15 +11,16 @@
  */
 
 import type Database from 'better-sqlite3'
+import { t } from '../i18n'
 
 /** 迁移脚本接口 */
 interface Migration {
   /** 版本号（必须递增） */
   version: number
-  /** 迁移描述（技术说明） */
-  description: string
-  /** 用户可读的升级原因（显示在弹窗中） */
-  userMessage: string
+  /** 迁移描述 i18n key（技术说明） */
+  descriptionKey: string
+  /** 用户可读的升级原因 i18n key（显示在弹窗中） */
+  userMessageKey: string
   /** 迁移执行函数 */
   up: (db: Database.Database) => void
 }
@@ -43,8 +44,8 @@ export const CURRENT_SCHEMA_VERSION = 3
 const migrations: Migration[] = [
   {
     version: 1,
-    description: '添加 owner_id 字段到 meta 表',
-    userMessage: '支持「Owner」功能，可在成员列表中设置自己的身份',
+    descriptionKey: 'database.migrationV1Desc',
+    userMessageKey: 'database.migrationV1Message',
     up: (db) => {
       // 检查 owner_id 列是否已存在（防止重复执行）
       const tableInfo = db.prepare('PRAGMA table_info(meta)').all() as Array<{ name: string }>
@@ -56,8 +57,8 @@ const migrations: Migration[] = [
   },
   {
     version: 2,
-    description: '添加 roles、reply_to_message_id、platform_message_id 字段',
-    userMessage: '支持成员角色、消息回复关系和回复内容预览',
+    descriptionKey: 'database.migrationV2Desc',
+    userMessageKey: 'database.migrationV2Message',
     up: (db) => {
       // 检查 roles 列是否已存在（防止重复执行）
       const memberTableInfo = db.prepare('PRAGMA table_info(member)').all() as Array<{ name: string }>
@@ -91,8 +92,8 @@ const migrations: Migration[] = [
   },
   {
     version: 3,
-    description: '添加会话索引相关表（chat_session、message_context）和 session_gap_threshold 字段',
-    userMessage: '支持会话时间轴浏览和 AI 增强分析功能',
+    descriptionKey: 'database.migrationV3Desc',
+    userMessageKey: 'database.migrationV3Message',
     up: (db) => {
       // 创建 chat_session 会话表
       db.exec(`
@@ -192,14 +193,14 @@ function checkDatabaseIntegrity(db: Database.Database): { valid: boolean; error?
     if (tables.length === 0) {
       return {
         valid: false,
-        error: '数据库结构不完整：缺少 meta 表。建议删除此数据库文件后重新导入。',
+        error: t('database.integrityError'),
       }
     }
     return { valid: true }
   } catch (error) {
     return {
       valid: false,
-      error: `数据库检查失败: ${error instanceof Error ? error.message : String(error)}`,
+      error: t('database.checkFailed', { error: error instanceof Error ? error.message : String(error) }),
     }
   }
 }
@@ -264,7 +265,7 @@ export function getPendingMigrationInfos(fromVersion = 0): MigrationInfo[] {
     .filter((m) => m.version > fromVersion)
     .map((m) => ({
       version: m.version,
-      description: m.description,
-      userMessage: m.userMessage,
+      description: t(m.descriptionKey),
+      userMessage: t(m.userMessageKey),
     }))
 }
diff --git a/electron/main/i18n/index.ts b/electron/main/i18n/index.ts
new file mode 100644
index 0000000..2d701a7
--- /dev/null
+++ b/electron/main/i18n/index.ts
@@ -0,0 +1,99 @@
+/**
+ * 主进程国际化模块
+ *
+ * 基于 i18next，提供主进程的多语言支持。
+ * 语言设置持久化在 settings/locale.json 中，
+ * 并通过 IPC 'locale:change' 与渲染进程同步。
+ */
+
+import i18next from 'i18next'
+import { app, ipcMain } from 'electron'
+import * as fs from 'fs'
+import * as path from 'path'
+import { getSettingsDir, ensureDir } from '../paths'
+import zhCN from './locales/zh-CN'
+import enUS from './locales/en-US'
+
+const LOCALE_FILE = 'locale.json'
+
+/**
+ * 获取 locale 配置文件路径
+ */
+function getLocaleFilePath(): string {
+  return path.join(getSettingsDir(), LOCALE_FILE)
+}
+
+/**
+ * 保存语言设置到文件
+ */
+function saveLocale(lng: string): void {
+  try {
+    ensureDir(getSettingsDir())
+    fs.writeFileSync(getLocaleFilePath(), JSON.stringify({ locale: lng }, null, 2), 'utf-8')
+  } catch (err) {
+    console.error('[i18n] Failed to save locale:', err)
+  }
+}
+
+/**
+ * 初始化主进程国际化
+ *
+ * 优先级：settings/locale.json > app.getLocale() 系统检测 > en-US 默认
+ * 同时注册 IPC 监听器接收渲染进程的语言切换请求
+ */
+export async function initLocale(): Promise<void> {
+  let lng = 'en-US' // 默认回退
+
+  try {
+    const filePath = getLocaleFilePath()
+    if (fs.existsSync(filePath)) {
+      // 读取用户保存的语言偏好
+      const data = JSON.parse(fs.readFileSync(filePath, 'utf-8'))
+      if (data.locale) lng = data.locale
+    } else {
+      // 无配置文件，探测系统语言
+      const sysLocale = app.getLocale()
+      lng = sysLocale.startsWith('zh') ? 'zh-CN' : 'en-US'
+    }
+  } catch (e) {
+    console.error('[i18n] Error loading locale config:', e)
+  }
+
+  await i18next.init({
+    lng,
+    fallbackLng: 'en-US',
+    resources: {
+      'zh-CN': { translation: zhCN },
+      'en-US': { translation: enUS },
+    },
+    interpolation: { escapeValue: false }, // Node 环境不需要防 XSS
+  })
+
+  console.log(`[i18n] Initialized with locale: ${lng}`)
+
+  // 监听渲染进程的语言切换请求（补全半完成的 IPC 机制）
+  ipcMain.on('locale:change', async (_event, newLocale: string) => {
+    if (newLocale !== i18next.language) {
+      await i18next.changeLanguage(newLocale)
+      saveLocale(newLocale)
+      console.log(`[i18n] Locale changed to: ${newLocale}`)
+    }
+  })
+}
+
+/**
+ * 翻译函数
+ * @param key 翻译 key，如 'update.newVersionTitle'
+ * @param options 插值参数，如 { version: '1.0.0' }
+ */
+export const t = (key: string, options?: Record<string, unknown>): string => i18next.t(key, options)
+
+/**
+ * 获取当前 locale
+ */
+export const getLocale = (): string => i18next.language
+
+/**
+ * 判断当前是否为中文环境（兼容现有 isChineseLocale 模式）
+ */
+export const isChineseLocale = (): boolean => i18next.language.startsWith('zh')
diff --git a/electron/main/i18n/locales/en-US.ts b/electron/main/i18n/locales/en-US.ts
new file mode 100644
index 0000000..3e75b64
--- /dev/null
+++ b/electron/main/i18n/locales/en-US.ts
@@ -0,0 +1,266 @@
+/**
+ * 主进程英文翻译
+ */
+export default {
+  // ===== Common =====
+  common: {
+    error: 'Error',
+  },
+
+  // ===== P0: Update dialogs =====
+  update: {
+    newVersionTitle: 'New version v{{version}} available',
+    newVersionMessage: 'New version v{{version}} available',
+    newVersionDetail: 'Would you like to download and install the new version?',
+    downloadNow: 'Download Now',
+    cancel: 'Cancel',
+    downloadComplete: 'Download Complete',
+    readyToInstall: 'The new version is ready. Install now?',
+    install: 'Install',
+    remindLater: 'Remind Later',
+    installOnQuit: 'Later (auto-install on quit)',
+    upToDate: 'You are up to date',
+  },
+
+  // ===== P0: File/directory dialogs =====
+  dialog: {
+    selectChatFile: 'Select Chat Record File',
+    chatRecords: 'Chat Records',
+    allFiles: 'All Files',
+    import: 'Import',
+    selectDirectory: 'Select Directory',
+    selectFolder: 'Select Folder',
+    selectFolderError: 'Error selecting folder: ',
+  },
+
+  // ===== P1: Database migrations =====
+  database: {
+    migrationV1Desc: 'Add owner_id field to meta table',
+    migrationV1Message: 'Support "Owner" feature to set your identity in the member list',
+    migrationV2Desc: 'Add roles, reply_to_message_id, platform_message_id fields',
+    migrationV2Message: 'Support member roles, message reply relationships and reply preview',
+    migrationV3Desc: 'Add session index tables (chat_session, message_context) and session_gap_threshold field',
+    migrationV3Message: 'Support session timeline browsing and AI-enhanced analysis',
+    integrityError:
+      'Database structure is incomplete: missing meta table. Please delete this database file and re-import.',
+    checkFailed: 'Database check failed: {{error}}',
+  },
+
+  // ===== Tool system =====
+  tools: {
+    notRegistered: 'Tool "{{toolName}}" is not registered',
+  },
+
+  // ===== P2: AI Tool definitions (Function Calling) =====
+  ai: {
+    tools: {
+      search_messages: {
+        desc: 'Search group chat records by keywords. Suitable for finding specific topics or keyword-related chat content. Can specify time range and sender to filter messages. Supports minute-level time queries.',
+        params: {
+          keywords:
+            'List of search keywords, using OR logic to match messages containing any keyword. Pass an empty array [] to filter by sender only',
+          sender_id:
+            'Sender member ID, used to filter messages from a specific member. Can be obtained via the get_group_members tool',
+          limit: 'Message count limit, default 1000, max 50000',
+          year: 'Filter messages by year, e.g. 2024',
+          month: 'Filter messages by month (1-12), use with year',
+          day: 'Filter messages by day (1-31), use with year and month',
+          hour: 'Filter messages by hour (0-23), use with year, month, and day',
+          start_time:
+            'Start time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 14:00". Overrides year/month/day/hour when specified',
+          end_time:
+            'End time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 18:30". Overrides year/month/day/hour when specified',
+        },
+      },
+      get_recent_messages: {
+        desc: 'Get chat messages within a specified time period. Suitable for overview questions like "what has everyone been chatting about recently" or "what was discussed in month X". Supports minute-level time queries.',
+        params: {
+          limit: 'Message count limit, default 100 (saves tokens, can be increased as needed)',
+          year: 'Filter messages by year, e.g. 2024',
+          month: 'Filter messages by month (1-12), use with year',
+          day: 'Filter messages by day (1-31), use with year and month',
+          hour: 'Filter messages by hour (0-23), use with year, month, and day',
+          start_time:
+            'Start time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 14:00". Overrides year/month/day/hour when specified',
+          end_time:
+            'End time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 18:30". Overrides year/month/day/hour when specified',
+        },
+      },
+      get_member_stats: {
+        desc: 'Get member activity statistics. Suitable for questions like "who is the most active" or "who sends the most messages".',
+        params: {
+          top_n: 'Return top N members, default 10',
+        },
+      },
+      get_time_stats: {
+        desc: 'Get time distribution statistics of chat activity. Suitable for questions like "when is the group most active" or "what time do people usually chat".',
+        params: {
+          type: 'Statistics type: hourly (by hour), weekday (by day of week), daily (by date)',
+        },
+      },
+      get_group_members: {
+        desc: 'Get group member list, including basic info, aliases, and message statistics. Suitable for queries like "who is in the group", "what is someone\'s alias", or "whose ID is xxx".',
+        params: {
+          search: 'Optional search keyword to filter by member nickname, alias, or platform ID',
+          limit: 'Member count limit, returns all by default',
+        },
+      },
+      get_member_name_history: {
+        desc: 'Get member name change history. Suitable for questions like "what was someone\'s previous name", "name changes", or "former names". Requires member ID from get_group_members tool first.',
+        params: {
+          member_id: 'Member database ID, can be obtained via get_group_members tool',
+        },
+      },
+      get_conversation_between: {
+        desc: 'Get conversation records between two group members. Suitable for questions like "what did A and B talk about" or "view the conversation between two people". Requires member IDs from get_group_members first. Supports minute-level time queries.',
+        params: {
+          member_id_1: 'Database ID of the first member',
+          member_id_2: 'Database ID of the second member',
+          limit: 'Message count limit, default 100',
+          year: 'Filter messages by year',
+          month: 'Filter messages by month (1-12), use with year',
+          day: 'Filter messages by day (1-31), use with year and month',
+          hour: 'Filter messages by hour (0-23), use with year, month, and day',
+          start_time:
+            'Start time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 14:00". Overrides year/month/day/hour when specified',
+          end_time:
+            'End time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 18:30". Overrides year/month/day/hour when specified',
+        },
+      },
+      get_message_context: {
+        desc: 'Get surrounding context messages for a given message ID. Suitable for viewing what was discussed before and after a specific message. Supports single or batch message IDs.',
+        params: {
+          message_ids:
+            'List of message IDs to query context for. Can be single or multiple IDs. Message IDs can be obtained from search_messages and other tool results',
+          context_size: 'Context size, i.e. how many messages before and after to retrieve, default 20',
+        },
+      },
+      search_sessions: {
+        desc: 'Search chat sessions (conversation segments). Sessions are conversation units automatically split by message time intervals. Suitable for finding discussions on specific topics or understanding how many conversations occurred in a time period. Returns matching sessions with a 5-message preview each.',
+        params: {
+          keywords: 'Optional keyword list, only returns sessions containing these keywords (OR logic)',
+          limit: 'Session count limit, default 20',
+          year: 'Filter sessions by year, e.g. 2024',
+          month: 'Filter sessions by month (1-12), use with year',
+          day: 'Filter sessions by day (1-31), use with year and month',
+          start_time: 'Start time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 14:00"',
+          end_time: 'End time, format "YYYY-MM-DD HH:mm", e.g. "2024-03-15 18:30"',
+        },
+      },
+      get_session_messages: {
+        desc: 'Get the complete message list for a specific session. Used to get full context after finding a relevant session via search_sessions. Returns all messages and participant information.',
+        params: {
+          session_id: 'Session ID, can be obtained from search_sessions results',
+          limit: 'Message count limit, default 1000. Can be limited for very long sessions to save tokens',
+        },
+      },
+      get_session_summaries: {
+        desc: `Get session summary list to quickly understand discussion topics in chat history.
+
+Use cases:
+1. Understand what topics have been discussed recently
+2. Search for discussed topics by keyword
+3. Overview questions like "has the group discussed travel"
+
+Returned summaries are brief descriptions of each session, helping quickly locate sessions of interest. Use get_session_messages for details.`,
+        params: {
+          keywords: 'Keyword list to search within summaries (OR logic)',
+          limit: 'Session count limit, default 20',
+          year: 'Filter sessions by year',
+          month: 'Filter sessions by month (1-12)',
+          day: 'Filter sessions by day (1-31)',
+          start_time: 'Start time, format "YYYY-MM-DD HH:mm"',
+          end_time: 'End time, format "YYYY-MM-DD HH:mm"',
+        },
+      },
+      semantic_search_messages: {
+        desc: `Search historical conversations using Embedding vector similarity, understanding semantics rather than keyword matching.
+
+⚠️ Use cases (prefer search_messages for keyword search, consider this tool for the following):
+1. Finding "similar expressions": e.g. "has anyone said something like 'I miss you'"
+2. Insufficient keyword results: when search_messages returns too few or irrelevant results
+3. Vague sentiment/relationship analysis: e.g. "how does the other person feel about me"
+
+❌ Not suitable for (use search_messages):
+- Searches with specific keywords (e.g. "travel", "birthday")
+- Finding specific person's messages
+- Finding messages in specific time periods`,
+        params: {
+          query: 'Semantic search query, describe in natural language what type of content you are looking for',
+          top_k: 'Number of results to return, default 10 (recommended 5-20)',
+          candidate_limit: 'Number of candidate sessions, default 50 (larger is slower but potentially more accurate)',
+          year: 'Filter sessions by year',
+          month: 'Filter sessions by month (1-12)',
+          day: 'Filter sessions by day (1-31)',
+          start_time: 'Start time, format "YYYY-MM-DD HH:mm"',
+          end_time: 'End time, format "YYYY-MM-DD HH:mm"',
+        },
+      },
+    },
+
+    // ===== AI Agent system prompts =====
+    agent: {
+      answerWithoutTools: 'Please answer based on the information already retrieved, do not call any more tools.',
+      toolError: 'Error: {{error}}',
+      currentDateIs: 'Current date is',
+      chatContext: {
+        private: 'conversation',
+        group: 'group chat',
+      },
+      ownerNote: `Current user identity:
+- The user's identity in this {{chatContext}} is "{{displayName}}" (platformId: {{platformId}})
+- When the user refers to "I" or "my", it refers to "{{displayName}}"
+- When querying "my" messages, use the sender_id parameter to filter for this member
+`,
+      memberNotePrivate: `Member query strategy:
+- Private chats only have two participants, so the member list can be directly obtained
+- When the user refers to "the other party" or "he/she", get the other participant's information via get_group_members
+`,
+      memberNoteGroup: `Member query strategy:
+- When the user refers to specific group members (e.g., "what did John say", "Mary's messages"), first call get_group_members to get the member list
+- Group members have three names: accountName (original nickname), groupNickname (group nickname), aliases (user-defined aliases)
+- The search parameter of get_group_members can be used for fuzzy searching these three names
+- Once a member is found, use their id field as the sender_id parameter for search_messages to retrieve their messages
+`,
+      timeParamsIntro: 'Time parameters: combine year/month/day/hour based on user mention',
+      timeParamExample1: '"October" → year: {{year}}, month: 10',
+      timeParamExample2: '"October 1st" → year: {{year}}, month: 10, day: 1',
+      timeParamExample3: '"October 1st 3 PM" → year: {{year}}, month: 10, day: 1, hour: 15',
+      defaultYearNote:
+        'If year is not specified, defaults to {{year}}. If the month has not yet occurred, {{prevYear}} is used.',
+      responseInstruction:
+        "Based on the user's question, select appropriate tools to retrieve data, then provide an answer based on the data.",
+      responseRulesTitle: 'Response requirements:',
+      fallbackRoleDefinition: {
+        group: `You are a professional group chat analysis assistant.
+Your task is to help users understand and analyze their group chat data.`,
+        private: `You are a professional private chat analysis assistant.
+Your task is to help users understand and analyze their private chat data.`,
+      },
+      fallbackResponseRules: `1. Answer based on data returned by tools, do not fabricate information
+2. If data is insufficient to answer, please state so
+3. Keep answers concise and clear, use Markdown format`,
+    },
+  },
+
+  // ===== P3: LLM config =====
+  llm: {
+    notConfigured: 'LLM service not configured. Please set up an API Key in settings first.',
+    maxConfigs: 'Maximum of {{count}} configurations allowed',
+    configNotFound: 'Configuration not found',
+    noActiveConfig: 'No active configuration',
+  },
+
+  // ===== P4: Summary generation =====
+  summary: {
+    sessionNotFound: 'Session not found or database could not be opened',
+    tooFewMessages: 'Message count less than {{count}}, no need to generate summary',
+    tooFewValidMessages: 'Valid message count less than {{count}}, no need to generate summary',
+    sessionNotExist: 'Session not found',
+    messagesTooFew: 'Too few messages',
+    validMessagesTooFew: 'Too few valid messages',
+    systemPromptDirect: 'You are a conversation summarization expert. Summarize conversations concisely.',
+    systemPromptMerge:
+      'You are a conversation summarization expert skilled at merging multiple summaries into a coherent overview.',
+  },
+}
diff --git a/electron/main/i18n/locales/zh-CN.ts b/electron/main/i18n/locales/zh-CN.ts
new file mode 100644
index 0000000..5420823
--- /dev/null
+++ b/electron/main/i18n/locales/zh-CN.ts
@@ -0,0 +1,262 @@
+/**
+ * 主进程中文翻译
+ */
+export default {
+  // ===== 通用 =====
+  common: {
+    error: '错误',
+  },
+
+  // ===== P0: 更新弹窗 =====
+  update: {
+    newVersionTitle: '发现新版本 v{{version}}',
+    newVersionMessage: '发现新版本 v{{version}}',
+    newVersionDetail: '是否立即下载并安装新版本？',
+    downloadNow: '立即下载',
+    cancel: '取消',
+    downloadComplete: '下载完成',
+    readyToInstall: '新版本已准备就绪，是否现在安装？',
+    install: '安装',
+    remindLater: '之后提醒',
+    installOnQuit: '稍后（应用退出后自动安装）',
+    upToDate: '已是最新版本',
+  },
+
+  // ===== P0: 文件/目录对话框 =====
+  dialog: {
+    selectChatFile: '选择聊天记录文件',
+    chatRecords: '聊天记录',
+    allFiles: '所有文件',
+    import: '导入',
+    selectDirectory: '选择目录',
+    selectFolder: '选择文件夹',
+    selectFolderError: '选择文件夹时发生错误：',
+  },
+
+  // ===== P1: 数据库迁移 =====
+  database: {
+    migrationV1Desc: '添加 owner_id 字段到 meta 表',
+    migrationV1Message: '支持「Owner」功能，可在成员列表中设置自己的身份',
+    migrationV2Desc: '添加 roles、reply_to_message_id、platform_message_id 字段',
+    migrationV2Message: '支持成员角色、消息回复关系和回复内容预览',
+    migrationV3Desc: '添加会话索引相关表（chat_session、message_context）和 session_gap_threshold 字段',
+    migrationV3Message: '支持会话时间轴浏览和 AI 增强分析功能',
+    integrityError: '数据库结构不完整：缺少 meta 表。建议删除此数据库文件后重新导入。',
+    checkFailed: '数据库检查失败: {{error}}',
+  },
+
+  // ===== 工具系统 =====
+  tools: {
+    notRegistered: '工具 "{{toolName}}" 未注册',
+  },
+
+  // ===== P2: AI 工具描述（Function Calling） =====
+  ai: {
+    tools: {
+      search_messages: {
+        desc: '根据关键词搜索群聊记录。适用于用户想要查找特定话题、关键词相关的聊天内容。可以指定时间范围和发送者来筛选消息。支持精确到分钟级别的时间查询。',
+        params: {
+          keywords: '搜索关键词列表，会用 OR 逻辑匹配包含任一关键词的消息。如果只需要按发送者筛选，可以传空数组 []',
+          sender_id: '发送者的成员 ID，用于筛选特定成员发送的消息。可以通过 get_group_members 工具获取成员 ID',
+          limit: '返回消息数量限制，默认 1000，最大 50000',
+          year: '筛选指定年份的消息，如 2024',
+          month: '筛选指定月份的消息（1-12），需要配合 year 使用',
+          day: '筛选指定日期的消息（1-31），需要配合 year 和 month 使用',
+          hour: '筛选指定小时的消息（0-23），需要配合 year、month 和 day 使用',
+          start_time:
+            '开始时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 14:00"。指定后会覆盖 year/month/day/hour 参数',
+          end_time:
+            '结束时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 18:30"。指定后会覆盖 year/month/day/hour 参数',
+        },
+      },
+      get_recent_messages: {
+        desc: '获取指定时间段内的群聊消息。适用于回答"最近大家聊了什么"、"X月群里聊了什么"等概览性问题。支持精确到分钟级别的时间查询。',
+        params: {
+          limit: '返回消息数量限制，默认 100（节省 token，可根据需要增加）',
+          year: '筛选指定年份的消息，如 2024',
+          month: '筛选指定月份的消息（1-12），需要配合 year 使用',
+          day: '筛选指定日期的消息（1-31），需要配合 year 和 month 使用',
+          hour: '筛选指定小时的消息（0-23），需要配合 year、month 和 day 使用',
+          start_time:
+            '开始时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 14:00"。指定后会覆盖 year/month/day/hour 参数',
+          end_time:
+            '结束时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 18:30"。指定后会覆盖 year/month/day/hour 参数',
+        },
+      },
+      get_member_stats: {
+        desc: '获取群成员的活跃度统计数据。适用于回答"谁最活跃"、"发言最多的是谁"等问题。',
+        params: {
+          top_n: '返回前 N 名成员，默认 10',
+        },
+      },
+      get_time_stats: {
+        desc: '获取群聊的时间分布统计。适用于回答"什么时候最活跃"、"大家一般几点聊天"等问题。',
+        params: {
+          type: '统计类型：hourly（按小时）、weekday（按星期）、daily（按日期）',
+        },
+      },
+      get_group_members: {
+        desc: '获取群成员列表，包括成员的基本信息、别名和消息统计。适用于查询"群里有哪些人"、"某人的别名是什么"、"谁的QQ号是xxx"等问题。',
+        params: {
+          search: '可选的搜索关键词，用于筛选成员昵称、别名或QQ号',
+          limit: '返回成员数量限制，默认返回全部',
+        },
+      },
+      get_member_name_history: {
+        desc: '获取成员的昵称变更历史记录。适用于回答"某人以前叫什么名字"、"某人的昵称变化"、"某人曾用名"等问题。需要先通过 get_group_members 工具获取成员 ID。',
+        params: {
+          member_id: '成员的数据库 ID，可以通过 get_group_members 工具获取',
+        },
+      },
+      get_conversation_between: {
+        desc: '获取两个群成员之间的对话记录。适用于回答"A和B之间聊了什么"、"查看两人的对话"等问题。需要先通过 get_group_members 获取成员 ID。支持精确到分钟级别的时间查询。',
+        params: {
+          member_id_1: '第一个成员的数据库 ID',
+          member_id_2: '第二个成员的数据库 ID',
+          limit: '返回消息数量限制，默认 100',
+          year: '筛选指定年份的消息',
+          month: '筛选指定月份的消息（1-12），需要配合 year 使用',
+          day: '筛选指定日期的消息（1-31），需要配合 year 和 month 使用',
+          hour: '筛选指定小时的消息（0-23），需要配合 year、month 和 day 使用',
+          start_time:
+            '开始时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 14:00"。指定后会覆盖 year/month/day/hour 参数',
+          end_time:
+            '结束时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 18:30"。指定后会覆盖 year/month/day/hour 参数',
+        },
+      },
+      get_message_context: {
+        desc: '根据消息 ID 获取前后的上下文消息。适用于需要查看某条消息前后聊天内容的场景，比如"这条消息的前后在聊什么"、"查看某条消息的上下文"等。支持单个或批量消息 ID。',
+        params: {
+          message_ids:
+            '要查询上下文的消息 ID 列表，可以是单个 ID 或多个 ID。消息 ID 可以从 search_messages 等工具的返回结果中获取',
+          context_size: '上下文大小，即获取前后各多少条消息，默认 20',
+        },
+      },
+      search_sessions: {
+        desc: '搜索聊天会话（对话段落）。会话是根据消息时间间隔自动切分的对话单元。适用于查找特定话题的讨论、了解某个时间段内发生了几次对话等场景。返回匹配的会话列表及每个会话的前5条消息预览。',
+        params: {
+          keywords: '可选的搜索关键词列表，只返回包含这些关键词的会话（OR 逻辑匹配）',
+          limit: '返回会话数量限制，默认 20',
+          year: '筛选指定年份的会话，如 2024',
+          month: '筛选指定月份的会话（1-12），需要配合 year 使用',
+          day: '筛选指定日期的会话（1-31），需要配合 year 和 month 使用',
+          start_time: '开始时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 14:00"',
+          end_time: '结束时间，格式 "YYYY-MM-DD HH:mm"，如 "2024-03-15 18:30"',
+        },
+      },
+      get_session_messages: {
+        desc: '获取指定会话的完整消息列表。用于在 search_sessions 找到相关会话后，获取该会话的完整上下文。返回会话的所有消息及参与者信息。',
+        params: {
+          session_id: '会话 ID，可以从 search_sessions 的返回结果中获取',
+          limit: '返回消息数量限制，默认 1000。对于超长会话可以限制返回数量以节省 token',
+        },
+      },
+      get_session_summaries: {
+        desc: `获取会话摘要列表，快速了解群聊历史讨论的主题。
+
+适用场景：
+1. 了解群里最近在聊什么话题
+2. 按关键词搜索讨论过的话题
+3. 概览性问题如"群里有没有讨论过旅游"
+
+返回的摘要是对每个会话的简短总结，可以帮助快速定位感兴趣的会话，然后用 get_session_messages 获取详情。`,
+        params: {
+          keywords: '在摘要中搜索的关键词列表（OR 逻辑匹配）',
+          limit: '返回会话数量限制，默认 20',
+          year: '筛选指定年份的会话',
+          month: '筛选指定月份的会话（1-12）',
+          day: '筛选指定日期的会话（1-31）',
+          start_time: '开始时间，格式 "YYYY-MM-DD HH:mm"',
+          end_time: '结束时间，格式 "YYYY-MM-DD HH:mm"',
+        },
+      },
+      semantic_search_messages: {
+        desc: `使用 Embedding 向量相似度搜索历史对话，理解语义而非关键词匹配。
+
+⚠️ 使用场景（优先使用 search_messages 关键词搜索，以下场景再考虑本工具）：
+1. 找"类似的话"或"类似的表达"：如"有没有说过类似'我想你了'这样的话"
+2. 关键词搜索结果不足：当 search_messages 返回结果太少或不相关时，可用本工具补充
+3. 模糊的情感/关系分析：如"对方对我的态度是怎样的"、"我们之间的氛围"
+
+❌ 不适合的场景（请用 search_messages）：
+- 有明确关键词的搜索（如"旅游"、"生日"、"加班"）
+- 查找特定人物的发言
+- 查找特定时间段的消息`,
+        params: {
+          query: '语义检索查询，用自然语言描述你想要找的内容类型',
+          top_k: '返回结果数量，默认 10（建议 5-20）',
+          candidate_limit: '候选会话数量，默认 50（越大越慢但可能更准确）',
+          year: '筛选指定年份的会话',
+          month: '筛选指定月份的会话（1-12）',
+          day: '筛选指定日期的会话（1-31）',
+          start_time: '开始时间，格式 "YYYY-MM-DD HH:mm"',
+          end_time: '结束时间，格式 "YYYY-MM-DD HH:mm"',
+        },
+      },
+    },
+
+    // ===== AI Agent 系统提示词 =====
+    agent: {
+      answerWithoutTools: '请根据已获取的信息给出回答，不要再调用工具。',
+      toolError: '错误: {{error}}',
+      currentDateIs: '当前日期是',
+      chatContext: {
+        private: '对话',
+        group: '群聊',
+      },
+      ownerNote: `当前用户身份：
+- 用户在{{chatContext}}中的身份是「{{displayName}}」（platformId: {{platformId}}）
+- 当用户提到"我"、"我的"时，指的就是「{{displayName}}」
+- 查询"我"的发言时，使用 sender_id 参数筛选该成员
+`,
+      memberNotePrivate: `成员查询策略：
+- 私聊只有两个人，可以直接获取成员列表
+- 当用户提到"对方"、"他/她"时，通过 get_group_members 获取另一方信息
+`,
+      memberNoteGroup: `成员查询策略：
+- 当用户提到特定群成员（如"张三说过什么"、"小明的发言"等）时，应先调用 get_group_members 获取成员列表
+- 群成员有三种名称：accountName（原始昵称）、groupNickname（群昵称）、aliases（用户自定义别名）
+- 通过 get_group_members 的 search 参数可以模糊搜索这三种名称
+- 找到成员后，使用其 id 字段作为 search_messages 的 sender_id 参数来获取该成员的发言
+`,
+      timeParamsIntro: '时间参数：按用户提到的精度组合 year/month/day/hour',
+      timeParamExample1: '"10月" → year: {{year}}, month: 10',
+      timeParamExample2: '"10月1号" → year: {{year}}, month: 10, day: 1',
+      timeParamExample3: '"10月1号下午3点" → year: {{year}}, month: 10, day: 1, hour: 15',
+      defaultYearNote: '未指定年份默认{{year}}年，若该月份未到则用{{prevYear}}年',
+      responseInstruction: '根据用户的问题，选择合适的工具获取数据，然后基于数据给出回答。',
+      responseRulesTitle: '回答要求：',
+      fallbackRoleDefinition: {
+        group: `你是一个专业但风格轻松的群聊记录分析助手。
+你的任务是帮助用户理解和分析他们的群聊记录数据，同时可以适度使用 B 站/网络热梗和表情/颜文字活跃气氛，但不影响结论的准确性。`,
+        private: `你是一个专业但风格轻松的私聊记录分析助手。
+你的任务是帮助用户理解和分析他们的私聊记录数据，同时可以适度使用 B 站/网络热梗和表情/颜文字活跃气氛，但不影响结论的准确性。`,
+      },
+      fallbackResponseRules: `1. 基于工具返回的数据回答，不要编造信息
+2. 如果数据不足以回答问题，请说明
+3. 回答要简洁明了，使用 Markdown 格式
+4. 可以适度加入 B 站/网络热梗、表情/颜文字（强度适中）
+5. 玩梗不得影响事实准确与结论清晰，避免低俗或冒犯性表达`,
+    },
+  },
+
+  // ===== P3: LLM 配置 =====
+  llm: {
+    notConfigured: 'LLM 服务未配置，请先在设置中配置 API Key',
+    maxConfigs: '最多只能添加 {{count}} 个配置',
+    configNotFound: '配置不存在',
+    noActiveConfig: '没有激活的配置',
+  },
+
+  // ===== P4: 摘要生成 =====
+  summary: {
+    sessionNotFound: '会话不存在或数据库打开失败',
+    tooFewMessages: '消息数量少于{{count}}条，无需生成摘要',
+    tooFewValidMessages: '有效消息数量少于{{count}}条，无需生成摘要',
+    sessionNotExist: '会话不存在',
+    messagesTooFew: '消息太少',
+    validMessagesTooFew: '有效消息太少',
+    systemPromptDirect: '你是一个对话摘要专家，擅长用简洁的语言总结对话内容。',
+    systemPromptMerge: '你是一个对话摘要专家，擅长将多个摘要合并成一个连贯的总结。',
+  },
+}
diff --git a/electron/main/index.ts b/electron/main/index.ts
index a05a04a..f9dafeb 100644
--- a/electron/main/index.ts
+++ b/electron/main/index.ts
@@ -7,6 +7,7 @@ import { initAnalytics, trackDailyActive } from './analytics'
 import { initProxy } from './network/proxy'
 import { needsLegacyMigration, migrateFromLegacyDir, ensureAppDirs, cleanupPendingDeleteDir } from './paths'
 import { migrateAllDatabases, checkMigrationNeeded } from './database/core'
+import { initLocale } from './i18n'
 
 class MainProcess {
   mainWindow: BrowserWindow | null
@@ -58,6 +59,9 @@ class MainProcess {
     // 确保应用目录存在
     ensureAppDirs()
 
+    // 初始化主进程国际化（在 ensureAppDirs 之后，确保 settings 目录存在）
+    await initLocale()
+
     // 执行数据库 schema 迁移（确保所有数据库在 Worker 查询前已是最新 schema）
     this.migrateDatabasesIfNeeded()
 
diff --git a/electron/main/ipc/ai.ts b/electron/main/ipc/ai.ts
index c6703f7..0513ab8 100644
--- a/electron/main/ipc/ai.ts
+++ b/electron/main/ipc/ai.ts
@@ -8,6 +8,7 @@ import * as rag from '../ai/rag'
 import { aiLogger } from '../ai/logger'
 import { getLogsDir } from '../paths'
 import { Agent, type AgentStreamChunk, type PromptConfig } from '../ai/agent'
+import { t } from '../i18n'
 import type { ToolContext } from '../ai/tools/types'
 import type { IpcContext } from './types'
 
@@ -374,7 +375,7 @@ export function registerAIHandlers({ win }: IpcContext): void {
         if (activeConfig) {
           return llm.deleteConfig(activeConfig.id)
         }
-        return { success: false, error: '没有激活的配置' }
+        return { success: false, error: t('llm.noActiveConfig') }
       }
       return llm.deleteConfig(id)
     } catch (error) {
diff --git a/electron/main/ipc/chat.ts b/electron/main/ipc/chat.ts
index 114ed18..68a3af1 100644
--- a/electron/main/ipc/chat.ts
+++ b/electron/main/ipc/chat.ts
@@ -10,6 +10,7 @@ import { detectFormat, diagnoseFormat, scanMultiChatFile, type ParseProgress } f
 import type { IpcContext } from './types'
 import { CURRENT_SCHEMA_VERSION, getPendingMigrationInfos, type MigrationInfo } from '../database/migrations'
 import { exportSessionToTempFile, cleanupTempExportFiles } from '../merger'
+import { t } from '../i18n'
 
 /**
  * 注册聊天记录相关 IPC 处理器
@@ -60,14 +61,14 @@ export function registerChatHandlers(ctx: IpcContext): void {
   ipcMain.handle('chat:selectFile', async () => {
     try {
       const { canceled, filePaths } = await dialog.showOpenDialog({
-        title: '选择聊天记录文件',
+        title: t('dialog.selectChatFile'),
         defaultPath: app.getPath('documents'),
         properties: ['openFile'],
         filters: [
-          { name: '聊天记录', extensions: ['json', 'jsonl', 'txt'] },
-          { name: '所有文件', extensions: ['*'] },
+          { name: t('dialog.chatRecords'), extensions: ['json', 'jsonl', 'txt'] },
+          { name: t('dialog.allFiles'), extensions: ['*'] },
         ],
-        buttonLabel: '导入',
+        buttonLabel: t('dialog.import'),
       })
 
       if (canceled || filePaths.length === 0) {
diff --git a/electron/main/ipc/window.ts b/electron/main/ipc/window.ts
index d3e4052..f63ced5 100644
--- a/electron/main/ipc/window.ts
+++ b/electron/main/ipc/window.ts
@@ -6,6 +6,7 @@ import { ipcMain, app, dialog, clipboard, shell } from 'electron'
 import * as fs from 'fs/promises'
 import type { IpcContext } from './types'
 import { simulateUpdateDialog, manualCheckForUpdates } from '../update'
+import { t } from '../i18n'
 
 /**
  * 注册窗口和文件系统操作 IPC 处理器
@@ -153,17 +154,17 @@ export function registerWindowHandlers(ctx: IpcContext): void {
   ipcMain.handle('selectDir', async (_, defaultPath = '') => {
     try {
       const { canceled, filePaths } = await dialog.showOpenDialog({
-        title: '选择目录',
+        title: t('dialog.selectDirectory'),
         defaultPath: defaultPath || app.getPath('documents'),
         properties: ['openDirectory', 'createDirectory'],
-        buttonLabel: '选择文件夹',
+        buttonLabel: t('dialog.selectFolder'),
       })
       if (!canceled) {
         return filePaths[0]
       }
       return null
     } catch (err) {
-      console.error('选择文件夹时发生错误：', err)
+      console.error(t('dialog.selectFolderError'), err)
       throw err
     }
   })
diff --git a/electron/main/update.ts b/electron/main/update.ts
index 1d8c9f4..3bf23a3 100644
--- a/electron/main/update.ts
+++ b/electron/main/update.ts
@@ -4,6 +4,7 @@ import { platform } from '@electron-toolkit/utils'
 import { logger } from './logger'
 import { getActiveProxyUrl } from './network/proxy'
 import { closeWorkerAsync } from './worker/workerManager'
+import { t } from './i18n'
 
 // R2 镜像源 URL（速度更快，作为主要更新源）
 const R2_MIRROR_URL = 'https://chatlab.1app.top/releases/download'
@@ -143,10 +144,10 @@ const checkUpdate = (win) => {
 
     dialog
       .showMessageBox({
-        title: '发现新版本 v' + info.version,
-        message: '发现新版本 v' + info.version,
-        detail: '是否立即下载并安装新版本？',
-        buttons: ['立即下载', '取消'],
+        title: t('update.newVersionTitle', { version: info.version }),
+        message: t('update.newVersionMessage', { version: info.version }),
+        detail: t('update.newVersionDetail'),
+        buttons: [t('update.downloadNow'), t('update.cancel')],
         defaultId: 0,
         cancelId: 1,
         type: 'question',
@@ -178,9 +179,9 @@ const checkUpdate = (win) => {
   autoUpdater.on('update-downloaded', () => {
     dialog
       .showMessageBox({
-        title: '下载完成',
-        message: '新版本已准备就绪，是否现在安装？',
-        buttons: ['安装', platform.isMacOS ? '之后提醒' : '稍后（应用退出后自动安装）'],
+        title: t('update.downloadComplete'),
+        message: t('update.readyToInstall'),
+        buttons: [t('update.install'), platform.isMacOS ? t('update.remindLater') : t('update.installOnQuit')],
         defaultId: 1,
         cancelId: 2,
         type: 'question',
@@ -219,7 +220,7 @@ const checkUpdate = (win) => {
     } else {
       win.webContents.send('show-message', {
         type: 'success',
-        message: '已是最新版本',
+        message: t('update.upToDate'),
       })
     }
   })
@@ -279,10 +280,10 @@ const manualCheckForUpdates = () => {
  */
 const simulateUpdateDialog = (win) => {
   dialog.showMessageBox({
-    title: '发现新版本 v9.9.9',
-    message: '发现新版本 v9.9.9',
-    detail: '是否立即下载并安装新版本？',
-    buttons: ['立即下载', '取消'],
+    title: t('update.newVersionTitle', { version: '9.9.9' }),
+    message: t('update.newVersionMessage', { version: '9.9.9' }),
+    detail: t('update.newVersionDetail'),
+    buttons: [t('update.downloadNow'), t('update.cancel')],
     defaultId: 0,
     cancelId: 1,
     type: 'question',
diff --git a/package.json b/package.json
index a9f73e8..715488c 100644
--- a/package.json
+++ b/package.json
@@ -43,6 +43,7 @@
     "echarts": "^6.0.0",
     "echarts-wordcloud": "^2.1.0",
     "electron-updater": "^6.6.2",
+    "i18next": "^25.8.5",
     "markdown-it": "^14.1.0",
     "node-machine-id": "^1.1.12",
     "stream-json": "^1.9.1",
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml
index a90b7b1..9b5af01 100644
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -50,6 +50,9 @@ importers:
       electron-updater:
         specifier: ^6.6.2
         version: 6.7.3
+      i18next:
+        specifier: ^25.8.5
+        version: 25.8.5(typescript@5.9.3)
       markdown-it:
         specifier: ^14.1.0
         version: 14.1.0
@@ -258,6 +261,10 @@ packages:
     peerDependencies:
       '@babel/core': ^7.0.0-0
 
+  '@babel/runtime@7.28.6':
+    resolution: {integrity: sha512-05WQkdpL9COIMz4LjTxGpPNCdlpyimKppYNoJ5Di5EUObifl8t4tuLuUBBZEpoLYOmfvIWrsp9fCl0HoPRVTdA==}
+    engines: {node: '>=6.9.0'}
+
   '@babel/template@7.28.6':
     resolution: {integrity: sha512-YA6Ma2KsCdGb+WC6UpBVFJGXL58MDA6oyONbjyF/+5sBgxY/dwkhLogbMT2GXXyU84/IhRw/2D1Os1B/giz+BQ==}
     engines: {node: '>=6.9.0'}
@@ -2578,6 +2585,14 @@ packages:
     resolution: {integrity: sha512-AXcZb6vzzrFAUE61HnN4mpLqd/cSIwNQjtNWR0euPm6y0iqx3G4gOXaIDdtdDwZmhwe82LA6+zinmW4UBWVePQ==}
     engines: {node: '>=16.17.0'}
 
+  i18next@25.8.5:
+    resolution: {integrity: sha512-TApjhgqQU6P7BQlpCTv6zQuXrYAP9rjYWgx2Nm8dsq+Zg9yJlAz+iR16/w7uVtTlSoULbqPTfqYjMK/DAQI+Ng==}
+    peerDependencies:
+      typescript: ^5
+    peerDependenciesMeta:
+      typescript:
+        optional: true
+
   iconv-corefoundation@1.1.7:
     resolution: {integrity: sha512-T10qvkw0zz4wnm560lOEg0PovVqUXuOFhhHAkixw8/sycy7TJt7v/RrkEKEQnAw2viPSJu6iAkErxnzR0g8PpQ==}
     engines: {node: ^8.11.2 || >=10}
@@ -4304,6 +4319,8 @@ snapshots:
       '@babel/core': 7.29.0
       '@babel/helper-plugin-utils': 7.28.6
 
+  '@babel/runtime@7.28.6': {}
+
   '@babel/template@7.28.6':
     dependencies:
       '@babel/code-frame': 7.29.0
@@ -7018,6 +7035,12 @@ snapshots:
 
   human-signals@5.0.0: {}
 
+  i18next@25.8.5(typescript@5.9.3):
+    dependencies:
+      '@babel/runtime': 7.28.6
+    optionalDependencies:
+      typescript: 5.9.3
+
   iconv-corefoundation@1.1.7:
     dependencies:
       cli-truncate: 2.1.0