feat: v1.0.0 - 唤醒词、性别配置、身份强化、移动端适配、Bug修复

.github/workflows/cd.yaml

@@ -6,7 +6,24 @@ on:
       - published
 
 jobs:
+  # Pre-cleanup: delete all existing assets from the release to avoid gh release upload failure
+  pre-release-cleanup:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Delete Existing Release Assets
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          TAG_NAME="${{ github.event.release.tag_name }}"
+          # List all existing assets and delete them
+          gh release view "$TAG_NAME" --json assets --jq '.assets[].name' 2>/dev/null | while read -r filename; do
+            echo "Deleting existing asset: $filename"
+            gh release delete-asset "$TAG_NAME" "$filename" --yes 2>/dev/null || true
+          done
+        shell: bash
+
   cd:
+    needs: pre-release-cleanup
     uses: halo-sigs/reusable-workflows/.github/workflows/plugin-cd.yaml@v4
     permissions:
       contents: write

README.md

@@ -4,10 +4,11 @@
 
 ## 功能特性
 
-- **多 AI 角色** — 支持创建多个 AI 角色，每个角色有独立的昵称、人格提示词和 Gravatar 头像，可为不同文章指定不同角色
+- **多 AI 角色** — 支持创建多个 AI 角色，每个角色有独立的昵称、人格提示词、性别、语气风格和 Gravatar 头像，可为不同文章指定不同角色
+- **唤醒词** — 评论以唤醒词开头可唤醒指定角色回复，支持自定义唤醒词，可在未启用AI回评的页面使用唤醒词召唤AI
 - **自动回复** — 监听新评论，自动调用 AI 生成回复，支持多轮对话上下文
 - **多语言适配** — 根据评论语言自动用对应语言回复
-- **情感分析** — 分析评论情感倾向（正面/中性/负面），根据情感调整回复语气
+- **情感分析** — 分析评论情感倾向（非常正面/正面/中性/负面/非常负面），根据情感调整回复语气
 - **草稿模式** — AI 回复先存为草稿，管理员审核后再发布，支持批量操作
 - **失败重试** — AI 生成失败时自动重试，指数退避策略
 - **对话轮次限制** — 同一评论线程中限制 AI 最多回复轮次，防止无限对话
@@ -30,16 +31,18 @@
 
 ## 安装
 
+### 应用商店安装
+
+进入 **插件** → **安装** → 应用市场搜索 **AI回评** → 安装，或前往 [Halo 应用商店](https://www.halo.run/store/apps/app-mo5tivjt) 一键安装。
+
+### 手动安装
+
 1. 前往 [Releases](https://github.com/sunny-335/plugin-comment-ai-autopilot/releases) 下载最新的 `.jar` 文件
 2. 登录 Halo 管理后台
-3. 进入 **插件** → **已安装** → 点击右上角 **安装** 按钮
+3. 进入 **插件** → **安装** → **本地上传**
 4. 选择下载的 `.jar` 文件上传
 5. 安装完成后启用插件
 
-::: tip
-安装本插件后，应用市场会推荐安装 AI Foundation 插件（本插件的必要依赖）。
-:::
-
 ## 从源码构建
 
 ```bash
@@ -67,65 +70,7 @@ pnpm dev
 
 ## 文档
 
-完整文档请访问 [AI回评文档站](https://nxxy335.top/comment-ai-autopilot)
+完整文档及更新日志请访问 [AI回评文档站](https://nxxy335.top/comment-ai-autopilot)
-
-## 更新日志
-
-### v1.0.0-beta.2
-
-**改进**
-
-- 通过 Halo 官方推荐的 `ExtensionGetter` 获取 AI Foundation 的 `AiModelService`，替换原先的跨 ClassLoader 反射调用方式
-- 在 `plugin.yaml` 中声明可选插件依赖 `ai-foundation?: "*"`，建立正确的插件依赖关系
-- 新增 `store.halo.run/recommended-apps` 注解，安装后可在应用市场推荐安装 AI Foundation 插件
-- 情感分析和内容审核改用 AI Foundation 结构化输出（`OutputSpec.choice`），分类更可靠
-- AI 调用改用 `GenerateTextRequest` 并设置 `maxRetries=2`，由 SDK 自动重试瞬时错误
-- AI 对话续接时自动获取之前的回复历史并注入到 Prompt 中，AI 能更好地理解对话上下文
-- 优化 AI 自审核评分机制：改为两阶段评估（安全检查 + 质量评分 1-5 分），评分映射到 0-100 分，替代原先的二值评分
-- 精简仪表盘：移除情感分布、近7日回复趋势、平均审核评分三个卡片，快捷操作精简为回复日志、插件设置、刷新数据
-- 重做设置页面：基本设置、AI角色设置、模型设置、Prompt设置、数据清理各为独立页面，通过标签栏切换
-- 优化设置页面布局：按钮统一排版并添加图标，侧边栏新增"未保存"状态指示器
-- 优化日志页面：评分增加等级标签（优秀/良好/一般/较差），筛选下拉框修复文本与箭头重叠
-- 优化 AI Foundation 状态提示宽度，与内容区宽度一致
-- 优化插件文档：修复版本要求、变量名、链接等错误
-
-**Bug 修复**
-
-- 修复 RateLimitService 清理线程在插件停止时未关闭导致线程泄漏
-- 修复内容审核提示词要求"重新生成"但代码未使用重新生成内容的问题
-- 修复设置页面按钮图标和文字未在同一行显示的问题
-- 修复设置页面标签栏无法点击切换的问题
-- 修复设置页面配置区域布局错误的问题
-
-### v1.0.0-beta.1
-
-**新功能**
-
-- 草稿模式：关闭"自动发布"后，AI 回复将保存为草稿，需站长审核后才发布
-- 多 AI 角色支持：支持配置多个 AI 虚拟角色，每个角色有独立的提示词、头像和模型
-- 提示词预设：内置多种回复风格预设（专业型、幽默型、简洁型等），可自由组合
-- 对话历史查看：支持查看 AI 回复的完整对话上下文
-
-**Bug 修复**
-
-- 修复草稿模式下审批失败（"AI回复已存在，无法重复发布"）的问题
-- 修复批量审批时同样的去重检查冲突问题
-- 修复 AI Foundation 不可用的问题（`PluginManager` 无法通过 Spring 依赖注入获取）
-- 修复 `DefaultSpringPlugin` 包级私有类反射访问权限问题
-- 修复 CI 构建失败（`gradlew` 缺少执行权限）
-
-**改进**
-
-- 审批逻辑优化：先查找已有 Reply 扩展再决定创建或更新
-- 移除不必要的 `AiFoundationConfiguration` 配置类
-- 前端 UI 优化：移除编辑功能、简化角色排序逻辑、清理无用代码
-
-### v0.0.3
-
-- 增强插件可靠性与可用性
-- 多 AI 角色支持
-- 草稿模式初步实现
-- 文档全面更新
 
 ## 许可证
 

build.gradle

@@ -5,7 +5,7 @@ plugins {
 }
 
 group 'top.nxxy335.commentaiautopilot'
-version '1.0.0-beta.2'
+version '1.0.0'
 
 repositories {
     mavenCentral()

docs/.vitepress/config.ts

@@ -41,13 +41,14 @@ export default defineConfig({
         text: "其他",
         items: [
           { text: "常见问题", link: "/guide/faq" },
+          { text: "更新日志", link: "/CHANGELOG" },
         ],
       },
     ],
     socialLinks: [
       {
         icon: "github",
-        link: "https://github.com/nxxy335/plugin-comment-ai-autopilot",
+        link: "https://github.com/sunny-335/plugin-comment-ai-autopilot",
       },
     ],
     search: {

docs/CHANGELOG.md

@@ -1,5 +1,42 @@
 # 更新日志
 
+## v1.0.0
+
+> 2026-06-18
+
+### 新功能
+
+- **唤醒词**：评论以唤醒词开头可唤醒指定角色回复，支持自定义唤醒词，可在未启用AI回评的页面使用唤醒词召唤AI，二级评论同样支持
+- **性别配置**：AI角色支持性别设置（男/女），AI回复时会保持对应性别身份
+- **语气风格**：支持中性语气复选框，勾选后使用中性语气，取消勾选则跟随性别语气（女性温柔细腻/男性沉稳理性）
+- **身份提示词强化**：角色身份信息前置到Prompt最开头（【核心身份】），安全规范中增加身份约束，确保AI始终保持角色身份
+
+### 改进
+
+- **优化情感分析系统**：从 3 级分类（正面/中性/负面）升级为 5 级分类（非常正面/正面/中性/负面/非常负面），情感判断更精细
+- **优化日志页面 UI**：批量操作按钮重写样式，确保底色和白色文字清晰可见；搜索框添加搜索图标；重置按钮添加图标和底色
+- **优化评分显示**：评分数字与等级标签之间添加间距，等级标签增加底色背景（优秀/良好/一般/较差）
+- **优化状态标签**：通过状态、发布状态、情感标签统一使用带底色的标签样式
+- **支持页面链接显示**：日志中新增独立页面（SinglePage）链接显示，之前仅支持文章链接
+- **移动端适配**：仪表盘、配置、日志页面全面适配移动端
+- **ObjectMapper 统一注入**：FilterService 和 PromptBuilder 中的 `new ObjectMapper()` 改为 Spring 构造函数注入
+- **服务端过滤优化**：日志列表查询改用 `Queries.equal()` 服务端过滤 status/sentiment，减少内存过滤开销
+- **新增索引**：为 AiCommentReply 扩展添加 `spec.sentiment`、`spec.published`、`spec.postKind` 索引
+- **新增 postKind 字段**：区分关联内容类型（Post/SinglePage），支持页面评论的链接生成
+- **PromptBuilder 情感提示**：适配 5 级情感分类，新增 VERY_POSITIVE 和 VERY_NEGATIVE 的语气提示
+
+### Bug 修复
+
+- **修复 ObjectMapper Bean 不存在**：Halo 插件上下文中没有自动注册 ObjectMapper Bean，创建 ObjectMapperConfiguration 手动注册
+- **修复 AI 回复仍说没有性别**：将身份信息前置到 Prompt 最开头，安全规范中删除"作为AI助手"措辞，新增身份约束
+- **修复唤醒词无法唤醒**：评论内容提取时对 raw 也做 HTML strip（Jsoup.clean），所有内容做 trim()，wakeWord 也做 trim()
+- **修复二级评论唤醒词检查位置错误**：唤醒词检查提前到 isReplyToAi 判断之前
+- **修复 SinglePage 内容获取 404**：PostContentService 不能用于 SinglePage，改用 SinglePage.getStatus().getExcerpt()
+- **修复 Post/SinglePage 404 容错**：fetch 添加 onErrorResume 降级为空上下文继续处理
+- **修复 Sort 参数 null 警告**：listAll 调用改为 Sort.unsorted()
+
+---
+
 ## v1.0.0-beta.2
 
 > 2026-06-17

docs/guide/auto-reply.md

@@ -21,6 +21,32 @@
 
 插件启动时间之前的评论不会触发自动回复，避免安装插件后对大量历史评论批量回复。
 
+## 唤醒词机制
+
+唤醒词允许用户在评论中通过特定词语唤醒指定AI角色回复，即使该页面未启用AI回评。
+
+### 工作方式
+
+1. 用户发表以唤醒词开头的评论（如"小回小回你好"）
+2. 插件检测到唤醒词匹配，自动唤醒对应角色
+3. 唤醒词后的内容作为实际评论内容传递给AI
+4. AI生成回复时自动获取上下文（文章内容、对话历史等）
+
+### 唤醒词特点
+
+- **跨页面唤醒**：在未启用AI回评的页面也能使用唤醒词召唤AI
+- **二级评论支持**：回复中同样可以使用唤醒词
+- **独立唤醒**：每个角色有独立的唤醒词，可以唤醒不同角色
+- **绕过限制**：唤醒词触发时绕过页面级启用检查和"必须是回复给AI"的检查，但仍检查黑名单
+
+### 配置唤醒词
+
+在 **AI回评** → **插件设置** → **AI角色设置** 中，为每个角色配置唤醒词。唤醒词留空则不启用该角色的唤醒功能。
+
+::: tip
+唤醒词建议设置为容易记忆且不易与正常评论混淆的词语。
+:::
+
 ## 对话式回复
 
 当评论者回复AI的评论时，插件会自动提取对话上下文（最近5条回复），让AI的回复更连贯自然。

docs/guide/installation.md

@@ -7,19 +7,19 @@
 
 ## 安装
 
-### 方式一：从 Release 下载
+### 方式一：应用商店安装
+
+进入 **插件** → **安装** → 应用市场搜索 **AI回评** → 安装，或前往 [Halo 应用商店](https://www.halo.run/store/apps/app-mo5tivjt) 一键安装。
+
+### 方式二：从 Release 下载
 
 1. 前往 [GitHub Releases](https://github.com/sunny-335/plugin-comment-ai-autopilot/releases) 下载最新的 `.jar` 文件
 2. 登录 Halo 管理后台
-3. 进入 **插件** → **已安装** → 点击右上角 **安装** 按钮
+3. 进入 **插件** → **安装** → **本地上传**
 4. 选择下载的 `.jar` 文件上传
 5. 安装完成后启用插件
 
-::: tip 推荐安装
+### 方式三：从源码构建
-安装本插件后，应用市场会推荐安装 AI Foundation 插件（本插件的必要依赖）。
-:::
-
-### 方式二：从源码构建
 
 ```bash
 # 克隆仓库

docs/guide/introduction.md

@@ -5,9 +5,11 @@ AI回评（Comment AI Autopilot）是一个 Halo 博客系统的插件，能够
 ## 核心功能
 
 - **自动回复** — 监听新评论，自动调用AI生成回复，支持多轮对话上下文
-- **多 AI 角色** — 支持创建多个 AI 角色，每个角色有独立的昵称、人格提示词和 Gravatar 头像，可为不同文章指定不同角色
+- **多 AI 角色** — 支持创建多个 AI 角色，每个角色有独立的昵称、人格提示词、性别、语气风格和 Gravatar 头像，可为不同文章指定不同角色
+- **唤醒词** — 评论以唤醒词开头可唤醒指定角色回复，支持自定义唤醒词，可在未启用AI回评的页面使用唤醒词召唤AI
+- **性别与语气** — AI角色支持性别配置（男/女）和中性语气风格，AI回复时保持对应性别身份
 - **多语言适配** — 根据评论语言自动用对应语言回复
-- **情感分析** — 分析评论情感倾向（正面/中性/负面），根据情感调整回复语气
+- **情感分析** — 分析评论情感倾向（非常正面/正面/中性/负面/非常负面），根据情感调整回复语气
 - **草稿模式** — AI回复先存为草稿，管理员审核后再发布
 - **失败重试** — AI生成失败时自动重试，指数退避策略
 - **批量操作** — 草稿模式下支持批量通过/拒绝/删除
@@ -25,19 +27,20 @@ AI回评（Comment AI Autopilot）是一个 Halo 博客系统的插件，能够
 ## 工作流程
 
 ```
-新评论 → 过滤检查 → 情感分析 → 构建Prompt → AI生成 → 安全审核 → 发布/草稿
+新评论 → 唤醒词检查 → 过滤检查 → 情感分析 → 构建Prompt → AI生成 → 安全审核 → 发布/草稿
                                                           ↓ (失败)
                                                         重试 → ... → 最终失败
 ```
 
 1. **新评论到达** — Reconciler 监听到新评论创建事件
-2. **过滤检查** — 检查文章/页面是否启用AI回复、评论者是否在黑名单中
+2. **唤醒词检查** — 检查评论是否以某个角色的唤醒词开头，匹配则唤醒对应角色
-3. **情感分析** — 调用AI分析评论情感倾向
+3. **过滤检查** — 检查文章/页面是否启用AI回复、评论者是否在黑名单中（唤醒词触发时绕过页面级启用检查）
-4. **构建Prompt** — 结合AI角色人格、情感提示、文章内容、评论上下文构建Prompt
+4. **情感分析** — 调用AI分析评论情感倾向
-5. **AI生成** — 调用AI模型生成回复内容
+5. **构建Prompt** — 结合AI角色人格、情感提示、文章内容、评论上下文构建Prompt
-6. **安全审核** — 对生成内容进行两阶段审核（安全检查 + 质量评分 1-5 分映射到 0-100）
+6. **AI生成** — 调用AI模型生成回复内容
-7. **发布/草稿** — 根据设置自动发布或存为草稿等待审核
+7. **安全审核** — 对生成内容进行两阶段审核（安全检查 + 质量评分 1-5 分映射到 0-100）
-8. **重试** — 如果AI生成失败，系统会自动重试（最多 maxRetryCount 次），每次重试间隔递增
+8. **发布/草稿** — 根据设置自动发布或存为草稿等待审核
+9. **重试** — 如果AI生成失败，系统会自动重试（最多 maxRetryCount 次），每次重试间隔递增
 
 ## 前置要求
 

docs/guide/persona.md

@@ -1,6 +1,6 @@
 # AI角色
 
-AI角色定义了回复评论的虚拟身份，包括昵称、人格和头像。
+AI角色定义了回复评论的虚拟身份，包括昵称、人格、性别、语气风格和头像。
 
 ## 角色配置
 
@@ -8,6 +8,28 @@ AI角色定义了回复评论的虚拟身份，包括昵称、人格和头像。
 
 AI回复者的显示名称，默认为「小回」。修改后新回复将使用新昵称，已有回复不受影响。
 
+### 性别与语气
+
+每个角色可以设置性别（男/女），AI回复时会保持对应性别身份。语气风格通过"中性语气"复选框控制：
+
+- **勾选中性语气**：AI使用中性语气回复
+- **取消勾选**：AI根据性别使用对应语气风格（女性→温柔细腻，男性→沉稳理性）
+
+默认角色「小回」的性别为女，勾选中性语气。
+
+### 唤醒词
+
+评论以唤醒词开头可唤醒该角色回复。唤醒词功能的特点：
+
+- **跨页面唤醒**：在未启用AI回评的页面，使用唤醒词也能召唤AI回复
+- **二级评论支持**：回复中同样可以使用唤醒词唤醒指定角色
+- **独立唤醒**：每个角色有独立的唤醒词，可以唤醒不同角色
+- **留空不启用**：唤醒词留空则不启用该角色的唤醒功能
+
+::: tip
+唤醒词匹配时，插件会自动去除评论中的HTML标签并去除首尾空格，确保匹配准确。
+:::
+
 ### 人格提示词
 
 人格提示词定义了AI角色的性格和回复风格，是影响回复质量的关键配置。

docs/guide/sentiment.md

@@ -6,14 +6,16 @@
 
 | 分类 | 说明 | AI回复语气 |
 |------|------|-----------|
-| 正面 | 评论情绪积极、友好、感谢 | 热情友好，表达感谢和共鸣 |
+| 非常正面 | 强烈的感谢、赞美、认同（如"太棒了"、"非常感谢"） | 热情洋溢，表达真诚的感谢和共鸣 |
-| 中性 | 评论情绪平淡、普通提问 | 正常语气回复，不加额外提示 |
+| 正面 | 友好、肯定、支持（如"不错"、"学习了"） | 热情友好，表达感谢和共鸣 |
-| 负面 | 评论情绪偏消极、不满、批评 | 理性温和，避免激化矛盾 |
+| 中性 | 提问、讨论、陈述事实（如"请问..."、"这个怎么用"） | 正常语气回复，不加额外提示 |
+| 负面 | 不满、质疑、批评（如"不好用"、"有问题"） | 理性温和，避免激化矛盾 |
+| 非常负面 | 攻击、辱骂、极端情绪（如"垃圾"、"骗子"） | 非常温和理性，避免激化矛盾，展现理解和耐心 |
 
 ## 工作原理
 
 1. 评论通过过滤检查后，调用AI对评论内容进行情感分析
-2. AI 使用结构化输出（`OutputSpec.choice`）返回情感分类结果（POSITIVE / NEUTRAL / NEGATIVE）
+2. AI 使用结构化输出（`OutputSpec.choice`）返回情感分类结果（VERY_POSITIVE / POSITIVE / NEUTRAL / NEGATIVE / VERY_NEGATIVE）
 3. 如果情感分析失败（如AI不可用），默认降级为 NEUTRAL
 4. 情感结果传入 PromptBuilder，在生成Prompt时追加对应的语气提示
 5. 情感结果同时记录在 `AiCommentReply` 的 `sentiment` 字段中
@@ -22,9 +24,11 @@
 
 在AI回复日志页面，每条记录会显示情感标签：
 
-- 🟢 **正面** — 绿色标签
+- 🟢 **非常正面** — 深绿色标签
+- 🟩 **正面** — 浅绿色标签
 - ⚪ **中性** — 灰色标签
-- 🔴 **负面** — 红色标签
+- 🟥 **负面** — 浅红色标签
+- 🔴 **非常负面** — 深红色标签
 
 ## 性能影响
 

docs/guide/settings.md

@@ -30,11 +30,13 @@
 
 ## AI角色设置
 
-AI角色定义了回复评论的虚拟身份。支持创建多个角色，每个角色有独立的昵称、人格提示词和 Gravatar 头像，可指定一个为默认角色。
+AI角色定义了回复评论的虚拟身份。支持创建多个角色，每个角色有独立的昵称、人格提示词、性别、语气风格和 Gravatar 头像，可指定一个为默认角色。
 
 | 配置项 | 说明 | 默认值 |
 |--------|------|--------|
 | 角色昵称 | AI回复者的显示名称 | 小回 |
+| 性别与语气 | 角色性别（男/女）+ 中性语气复选框（勾选=中性语气，取消勾选=跟随性别语气） | 女 + 中性语气 |
+| 唤醒词 | 评论以此词开头则唤醒该角色回复，留空不启用 | 空 |
 | 人格提示词 | 定义AI角色的人格和回复风格 | 见下方 |
 | 邮箱 | 用于 Gravatar 头像服务展示头像 | 空 |
 | 设为默认 | 将该角色设为默认角色 | 第一个角色默认 |
@@ -90,8 +92,10 @@ AI角色定义了回复评论的虚拟身份。支持创建多个角色，每个
 
 ::: tip 情感提示
 情感提示由插件根据情感分析结果自动追加到 Prompt 末尾，不需要在模板中手动添加：
+- **非常正面** → 追加热情洋溢的语气提示
 - **正面** → 追加热情友好的语气提示
 - **负面** → 追加理性温和的语气提示
+- **非常负面** → 追加冷静关怀的语气提示
 - **中性** → 不追加额外提示
 :::
 

src/main/java/top/nxxy335/commentaiautopilot/CommentAiAutopilotPlugin.java

@@ -43,6 +43,12 @@ public class CommentAiAutopilotPlugin extends BasePlugin {
                 .indexFunc(ext -> ext.getSpec().getPostId()));
             indexSpecs.add(IndexSpecs.<AiCommentReply, String>single("spec.status", String.class)
                 .indexFunc(ext -> ext.getSpec().getStatus()));
+            indexSpecs.add(IndexSpecs.<AiCommentReply, String>single("spec.sentiment", String.class)
+                .indexFunc(ext -> ext.getSpec().getSentiment()));
+            indexSpecs.add(IndexSpecs.<AiCommentReply, String>single("spec.published", String.class)
+                .indexFunc(ext -> String.valueOf(ext.getSpec().getPublished())));
+            indexSpecs.add(IndexSpecs.<AiCommentReply, String>single("spec.postKind", String.class)
+                .indexFunc(ext -> ext.getSpec().getPostKind()));
         });
         schemeManager.register(AiPersona.class);
 
@@ -61,6 +67,9 @@ public class CommentAiAutopilotPlugin extends BasePlugin {
                 spec.setDisplayName("小回");
                 spec.setPrompt("你是一个友善的评论者，回复简洁自然，像朋友聊天一样。");
                 spec.setEmail("");
+                spec.setGender("female");
+                spec.setNeutralVoice(true);
+                spec.setWakeWord("小回小回");
                 spec.setIsDefault(true);
                 persona.setSpec(spec);
                 return client.create(persona);

src/main/java/top/nxxy335/commentaiautopilot/ObjectMapperConfiguration.java

@@ -0,0 +1,14 @@
+package top.nxxy335.commentaiautopilot;
+
+import com.fasterxml.jackson.databind.ObjectMapper;
+import org.springframework.context.annotation.Bean;
+import org.springframework.context.annotation.Configuration;
+
+@Configuration
+public class ObjectMapperConfiguration {
+
+    @Bean
+    ObjectMapper objectMapper() {
+        return new ObjectMapper();
+    }
+}

src/main/java/top/nxxy335/commentaiautopilot/endpoint/CommentAiAutopilotEndpoint.java

@@ -8,7 +8,6 @@ import org.springframework.web.reactive.function.server.ServerResponse;
 import reactor.core.publisher.Flux;
 import reactor.core.publisher.Mono;
 import run.halo.app.core.extension.content.Comment;
-import run.halo.app.core.extension.content.Post;
 import run.halo.app.core.extension.content.Reply;
 import run.halo.app.core.extension.endpoint.CustomEndpoint;
 import run.halo.app.extension.ConfigMap;
@@ -16,6 +15,7 @@ import run.halo.app.extension.Metadata;
 import run.halo.app.extension.GroupVersion;
 import run.halo.app.extension.ListOptions;
 import run.halo.app.extension.ListResult;
+import run.halo.app.extension.index.query.Queries;
 import run.halo.app.extension.ReactiveExtensionClient;
 import run.halo.app.extension.PageRequestImpl;
 import top.nxxy335.commentaiautopilot.extension.AiCommentReply;
@@ -24,6 +24,8 @@ import top.nxxy335.commentaiautopilot.service.AiFoundationClient;
 import top.nxxy335.commentaiautopilot.service.AiReplyCleanupService;
 import top.nxxy335.commentaiautopilot.service.AiReplyOrchestrator;
 import top.nxxy335.commentaiautopilot.service.CommentReplyPublisher;
+import top.nxxy335.commentaiautopilot.service.PersonaResolver;
+import top.nxxy335.commentaiautopilot.util.GravatarUtil;
 
 import com.fasterxml.jackson.databind.JsonNode;
 import com.fasterxml.jackson.databind.ObjectMapper;
@@ -56,16 +58,18 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
     private final AiFoundationClient aiFoundationClient;
     private final CommentReplyPublisher commentReplyPublisher;
     private final ObjectMapper objectMapper;
+    private final PersonaResolver personaResolver;
 
     private static final String CONFIG_MAP_NAME = "comment-ai-autopilot-configmap";
 
-    public CommentAiAutopilotEndpoint(ReactiveExtensionClient client, AiReplyOrchestrator orchestrator, AiReplyCleanupService cleanupService, AiFoundationClient aiFoundationClient, CommentReplyPublisher commentReplyPublisher) {
+    public CommentAiAutopilotEndpoint(ReactiveExtensionClient client, AiReplyOrchestrator orchestrator, AiReplyCleanupService cleanupService, AiFoundationClient aiFoundationClient, CommentReplyPublisher commentReplyPublisher, ObjectMapper objectMapper, PersonaResolver personaResolver) {
         this.client = client;
         this.orchestrator = orchestrator;
         this.cleanupService = cleanupService;
         this.aiFoundationClient = aiFoundationClient;
         this.commentReplyPublisher = commentReplyPublisher;
-        this.objectMapper = new ObjectMapper();
+        this.objectMapper = objectMapper;
+        this.personaResolver = personaResolver;
     }
 
     @Override
@@ -133,19 +137,26 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
         final Instant finalStartInstant = startInstant;
         final Instant finalEndInstant = endInstant;
 
-        // Check if we need in-memory filtering (keyword, date range, status, or sentiment)
+        // Check if we need in-memory filtering (keyword or date range)
-        boolean needsMemoryFilter = !keywordFilter.isBlank() || finalStartInstant != null || finalEndInstant != null
+        boolean needsMemoryFilter = !keywordFilter.isBlank() || finalStartInstant != null || finalEndInstant != null;
-            || !statusFilter.isBlank() || !sentimentFilter.isBlank();
+
+        // Build server-side query for status and sentiment (indexed fields)
+        var listOptionsBuilder = ListOptions.builder();
+        if (!statusFilter.isBlank()) {
+            listOptionsBuilder.andQuery(Queries.equal("spec.status", statusFilter));
+        }
+        if (!sentimentFilter.isBlank()) {
+            listOptionsBuilder.andQuery(Queries.equal("spec.sentiment", sentimentFilter));
+        }
+        var listOptions = listOptionsBuilder.build();
 
         if (needsMemoryFilter) {
-            // Fall back to listAll + in-memory filter for complex queries
+            // Fall back to listAll + in-memory filter for keyword/date queries
-            return client.listAll(AiCommentReply.class, ListOptions.builder().build(), Sort.unsorted())
+            return client.listAll(AiCommentReply.class, listOptions, Sort.unsorted())
                 .collectList()
                 .map(replies -> {
                     var filtered = replies.stream()
                         .filter(r -> {
-                            if (!statusFilter.isBlank() && !statusFilter.equals(r.getSpec().getStatus())) return false;
-                            if (!sentimentFilter.isBlank() && !sentimentFilter.equals(r.getSpec().getSentiment())) return false;
                             if (!keywordFilter.isBlank()) {
                                 String reply = r.getSpec().getReply();
                                 if (reply == null || !reply.contains(keywordFilter)) return false;
@@ -184,13 +195,11 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
                 .flatMap(result -> ServerResponse.ok().bodyValue(result));
         }
 
-        // No filters - use server-side pagination directly
+        // No memory filters needed - use server-side pagination directly
         Sort sort = "asc".equalsIgnoreCase(sortOrder)
             ? Sort.by(Sort.Order.asc("metadata.creationTimestamp"))
             : Sort.by(Sort.Order.desc("metadata.creationTimestamp"));
 
-        var listOptions = ListOptions.builder().build();
-
         return client.listBy(AiCommentReply.class, listOptions,
                 PageRequestImpl.of(page - 1, size, sort))
             .map(listResult -> {
@@ -246,18 +255,7 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
             .next()
             .flatMap(persona -> {
                 String email = persona.getSpec().getEmail();
-                String avatarUrl = "";
+                String avatarUrl = GravatarUtil.generateUrl(email);
-                if (email != null && !email.isBlank()) {
-                    try {
-                        var digest = java.security.MessageDigest.getInstance("SHA-256");
-                        var hashBytes = digest.digest(email.trim().toLowerCase().getBytes(java.nio.charset.StandardCharsets.UTF_8));
-                        var hexString = new StringBuilder();
-                        for (byte b : hashBytes) {
-                            hexString.append(String.format("%02x", b));
-                        }
-                        avatarUrl = "https://cn.cravatar.com/avatar/" + hexString;
-                    } catch (Exception ignored) {}
-                }
                 return ServerResponse.ok().bodyValue(Map.of(
                     "name", persona.getSpec().getDisplayName(),
                     "prompt", persona.getSpec().getPrompt() != null ? persona.getSpec().getPrompt() : "",
@@ -297,8 +295,9 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
                     "comment", commentOwner, commentContent, commentTime, isCommentAi
                 );
 
-                return client.listAll(Reply.class, ListOptions.builder().build(), Sort.unsorted())
+                return client.list(Reply.class,
-                    .filter(reply -> commentName.equals(reply.getSpec().getCommentName()))
+                        reply -> commentName.equals(reply.getSpec().getCommentName()),
+                        null)
                     .sort(Comparator.comparing(r -> r.getMetadata().getCreationTimestamp()))
                     .map(reply -> {
                         var replyOwner = extractOwnerName(reply.getSpec().getOwner());
@@ -642,9 +641,9 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
                         .bodyValue(Map.of("message", "该评论已有AI回复记录"));
                 }
                 // Read persona name from post annotations
-                return getPersonaNameFromComment(commentName)
+                return personaResolver.getPersonaNameFromComment(commentName)
                     .flatMap(personaName ->
-                        orchestrator.processComment(commentName, null, false, personaName)
+                        orchestrator.processComment(commentName, null, false, personaName, false)
                             .then(ServerResponse.ok().bodyValue(Map.of("message", "已触发AI回复")))
                     );
             });
@@ -669,9 +668,9 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
                             return ServerResponse.badRequest()
                                 .bodyValue(Map.of("message", "该回复已有AI对话记录"));
                         }
-                        return getPersonaNameFromComment(commentName)
+                        return personaResolver.getPersonaNameFromComment(commentName)
                             .flatMap(personaName ->
-                                orchestrator.processComment(commentName, replyName, true, personaName)
+                                orchestrator.processComment(commentName, replyName, true, personaName, false)
                                     .then(ServerResponse.ok().bodyValue(Map.of("message", "已触发AI对话回复")))
                             );
                     });
@@ -679,31 +678,6 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
             .switchIfEmpty(ServerResponse.notFound().build());
     }
 
-    private static final String AI_PERSONA_ANNOTATION = "comment-ai-autopilot.nxxy335.top/ai-persona";
-
-    private Mono<String> getPersonaNameFromComment(String commentName) {
-        return client.fetch(Comment.class, commentName)
-            .flatMap(comment -> {
-                var subjectRef = comment.getSpec().getSubjectRef();
-                if (subjectRef == null || !"Post".equals(subjectRef.getKind())) {
-                    return Mono.justOrEmpty(null);
-                }
-                String postName = subjectRef.getName();
-                return client.fetch(Post.class, postName)
-                    .mapNotNull(post -> {
-                        var annotations = post.getMetadata().getAnnotations();
-                        if (annotations != null) {
-                            String persona = annotations.get(AI_PERSONA_ANNOTATION);
-                            if (persona != null && !persona.isBlank()) {
-                                return persona;
-                            }
-                        }
-                        return null;
-                    });
-            })
-            .defaultIfEmpty("");
-    }
-
     private Mono<Reply> findReplyForRecord(AiCommentReply record) {
         // First try using replyName if available
         String replyName = record.getSpec().getReplyName();
@@ -748,7 +722,8 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
     ) {}
 
     private Mono<ServerResponse> listCommenters(ServerRequest request) {
-        return client.listAll(Comment.class, ListOptions.builder().build(), Sort.unsorted())
+        return client.list(Comment.class, null, null)
+            .take(1000)
             .collectList()
             .map(comments -> {
                 Set<String> seen = new HashSet<>();
@@ -766,7 +741,7 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
                         if (owner.getAnnotations() != null && owner.getAnnotations().get(Comment.CommentOwner.AVATAR_ANNO) != null) {
                             avatarUrl = owner.getAnnotations().get(Comment.CommentOwner.AVATAR_ANNO);
                         } else if (!email.isBlank()) {
-                            avatarUrl = generateGravatarUrl(email);
+                            avatarUrl = GravatarUtil.generateUrl(email);
                         }
                         result.add(new CommenterInfo(displayName, email, avatarUrl));
                     }
@@ -776,26 +751,11 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
             .flatMap(commenters -> ServerResponse.ok().bodyValue(commenters));
     }
 
-    private String generateGravatarUrl(String email) {
-        try {
-            var digest = java.security.MessageDigest.getInstance("SHA-256");
-            var hashBytes = digest.digest(email.trim().toLowerCase().getBytes(java.nio.charset.StandardCharsets.UTF_8));
-            var hexString = new StringBuilder();
-            for (byte b : hashBytes) {
-                hexString.append(String.format("%02x", b));
-            }
-            return "https://cn.cravatar.com/avatar/" + hexString;
-        } catch (Exception e) {
-            return "";
-        }
-    }
-
     private Mono<ServerResponse> triggerCleanup(ServerRequest request) {
-        return Mono.fromCallable(() -> {
+        return cleanupService.getRetentionDays()
-                int retentionDays = cleanupService.getRetentionDays();
+            .flatMap(retentionDays -> cleanupService.executeCleanup(retentionDays)
-                long deleted = cleanupService.executeCleanup(retentionDays);
+                .map(deleted -> Map.of("deletedCount", deleted, "retentionDays", retentionDays))
-                return Map.of("deletedCount", deleted, "retentionDays", retentionDays);
+            )
-            })
             .flatMap(result -> ServerResponse.ok().bodyValue(result))
             .onErrorResume(e -> {
                 log.warn("Failed to trigger cleanup: {}", e.getMessage());
@@ -961,7 +921,6 @@ public class CommentAiAutopilotEndpoint implements CustomEndpoint {
                     for (var personaData : personasList) {
                         importMono = importMono.then(Mono.defer(() -> {
                             try {
-                                var objectMapper = new com.fasterxml.jackson.databind.ObjectMapper();
                                 var personaJson = objectMapper.writeValueAsString(personaData);
                                 var persona = objectMapper.readValue(personaJson, AiPersona.class);
                                 var personaName = persona.getMetadata().getName();

src/main/java/top/nxxy335/commentaiautopilot/extension/AiCommentReply.java

@@ -33,6 +33,9 @@ public class AiCommentReply extends AbstractExtension {
         @Schema(description = "关联文章Slug，用于生成文章链接")
         private String postSlug;
 
+        @Schema(description = "关联内容类型: Post/SinglePage")
+        private String postKind;
+
         @Schema(description = "AI回复内容")
         private String reply;
 

src/main/java/top/nxxy335/commentaiautopilot/extension/AiPersona.java

@@ -34,6 +34,16 @@ public class AiPersona extends AbstractExtension {
         @Schema(description = "邮箱（用于Gravatar头像）")
         private String email;
 
+        @Schema(description = "角色性别（male/female）")
+        private String gender;
+
+        @Schema(description = "是否使用中性语气，默认false即跟随性别语气")
+        @JsonProperty("neutralVoice")
+        private Boolean neutralVoice;
+
+        @Schema(description = "唤醒词，评论以此开头则唤醒该角色回复，留空则不启用唤醒")
+        private String wakeWord;
+
         @Schema(description = "是否为默认角色")
         @JsonProperty("isDefault")
         private Boolean isDefault;

src/main/java/top/nxxy335/commentaiautopilot/listener/AiPersonaReconciler.java

@@ -1,31 +0,0 @@
-package top.nxxy335.commentaiautopilot.listener;
-
-import lombok.RequiredArgsConstructor;
-import lombok.extern.slf4j.Slf4j;
-import org.springframework.stereotype.Component;
-import run.halo.app.extension.ReactiveExtensionClient;
-import run.halo.app.extension.controller.Controller;
-import run.halo.app.extension.controller.ControllerBuilder;
-import run.halo.app.extension.controller.Reconciler;
-import top.nxxy335.commentaiautopilot.extension.AiPersona;
-
-@Component
-@Slf4j
-@RequiredArgsConstructor
-public class AiPersonaReconciler implements Reconciler<Reconciler.Request> {
-
-    private final ReactiveExtensionClient client;
-
-    @Override
-    public Result reconcile(Request request) {
-        return new Result(false, null);
-    }
-
-    @Override
-    public Controller setupWith(ControllerBuilder builder) {
-        return builder
-            .extension(new AiPersona())
-            .syncAllOnStart(false)
-            .build();
-    }
-}

src/main/java/top/nxxy335/commentaiautopilot/listener/CommentReconciler.java

@@ -4,22 +4,19 @@ import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.stereotype.Component;
 import reactor.core.scheduler.Schedulers;
-import run.halo.app.core.extension.content.Category;
 import run.halo.app.core.extension.content.Comment;
-import run.halo.app.core.extension.content.Post;
-import run.halo.app.core.extension.content.Tag;
 import run.halo.app.extension.ExtensionClient;
 import run.halo.app.extension.controller.Controller;
 import run.halo.app.extension.controller.ControllerBuilder;
 import run.halo.app.extension.controller.Reconciler;
 import top.nxxy335.commentaiautopilot.extension.AiCommentReply;
 import top.nxxy335.commentaiautopilot.service.AiReplyOrchestrator;
+import top.nxxy335.commentaiautopilot.service.PersonaResolver;
+import top.nxxy335.commentaiautopilot.service.WakeWordService;
 
 import java.time.Instant;
 import java.util.HashMap;
 import java.util.Map;
-import java.util.concurrent.ConcurrentHashMap;
-import java.util.concurrent.atomic.AtomicBoolean;
 
 @Component
 @Slf4j
@@ -28,97 +25,92 @@ public class CommentReconciler implements Reconciler<Reconciler.Request> {
 
     private final ExtensionClient client;
     private final AiReplyOrchestrator orchestrator;
+    private final PersonaResolver personaResolver;
+    private final WakeWordService wakeWordService;
 
     private static final String PROCESSED_ANNOTATION = "comment-ai-autopilot.nxxy335.top/processed";
-    private static final String AI_MARKER_ANNOTATION = "comment-ai-autopilot.nxxy335.top/is-ai";
     private static final String AI_PERSONA_OWNER_PREFIX = "ai-persona-";
-    private static final String AI_PERSONA_ANNOTATION = "comment-ai-autopilot.nxxy335.top/ai-persona";
+    private static final String AI_MARKER_ANNOTATION = "comment-ai-autopilot.nxxy335.top/is-ai";
 
     // Record the time when this bean was created (plugin startup time)
     private final Instant pluginStartTime = Instant.now();
 
-    // In-memory dedup lock: prevents the same comment from being processed multiple times
-    // even if reconcile is triggered concurrently
-    private final ConcurrentHashMap<String, Boolean> processingLocks = new ConcurrentHashMap<>();
-
     @Override
     public Result reconcile(Request request) {
         var name = request.name();
 
-        // Acquire lock at the very beginning to prevent any concurrent processing
+        client.fetch(Comment.class, name).ifPresent(comment -> {
-        if (processingLocks.putIfAbsent(name, Boolean.TRUE) != null) {
+            if (isProcessed(comment.getMetadata().getAnnotations())) {
-            log.debug("[CommentReconciler] Already processing comment: {}, skipping", name);
+                return;
-            return Result.doNotRetry();
+            }
-        }
 
-        AtomicBoolean asyncStarted = new AtomicBoolean(false);
+            // Skip comments created before plugin startup (historical comments)
-        try {
+            var creationTime = comment.getMetadata().getCreationTimestamp();
-            client.fetch(Comment.class, name).ifPresent(comment -> {
+            if (creationTime != null && creationTime.isBefore(pluginStartTime)) {
-                if (isProcessed(comment.getMetadata().getAnnotations())) {
+                log.debug("[CommentReconciler] Skipping historical comment: {} (created before plugin startup)", name);
-                    return;
-                }
-
-                // Skip comments created before plugin startup (historical comments)
-                var creationTime = comment.getMetadata().getCreationTimestamp();
-                if (creationTime != null && creationTime.isBefore(pluginStartTime)) {
-                    log.debug("[CommentReconciler] Skipping historical comment: {} (created before plugin startup)", name);
-                    markProcessed(comment);
-                    client.update(comment);
-                    return;
-                }
-
-                // Skip comments from AI persona itself
-                if (isAiComment(comment)) {
-                    markProcessed(comment);
-                    client.update(comment);
-                    return;
-                }
-
-                // Dedup: check if we already have an AiCommentReply record for this comment
-                boolean alreadyHasRecord = !client.list(AiCommentReply.class,
-                        record -> name.equals(record.getSpec().getCommentId())
-                            && !Boolean.TRUE.equals(record.getSpec().getIsAiConversation()),
-                        null)
-                    .isEmpty();
-
-                if (alreadyHasRecord) {
-                    log.debug("[CommentReconciler] Already have AiCommentReply record for: {}, skipping", name);
-                    markProcessed(comment);
-                    client.update(comment);
-                    return;
-                }
-
-                // Mark as processed first to avoid re-processing
                 markProcessed(comment);
                 client.update(comment);
+                return;
+            }
 
-                // Read persona name from the post's annotations
+            // Skip comments from AI persona itself
-                String personaName = getPersonaNameFromComment(comment);
+            if (isAiComment(comment)) {
+                markProcessed(comment);
+                client.update(comment);
+                return;
+            }
+
+            // Dedup: check if we already have an AiCommentReply record for this comment
+            boolean alreadyHasRecord = !client.list(AiCommentReply.class,
+                    record -> name.equals(record.getSpec().getCommentId())
+                        && !Boolean.TRUE.equals(record.getSpec().getIsAiConversation()),
+                    null)
+                .isEmpty();
+
+            if (alreadyHasRecord) {
+                log.debug("[CommentReconciler] Already have AiCommentReply record for: {}, skipping", name);
+                markProcessed(comment);
+                client.update(comment);
+                return;
+            }
+
+            // Mark as processed first to avoid re-processing
+            markProcessed(comment);
+            client.update(comment);
+
+            // Check for wake word in comment content
+            String commentContent = getCommentContent(comment);
+            log.info("[CommentReconciler] Wake word check for comment {}: content='{}'",
+                name, commentContent.length() > 80 ? commentContent.substring(0, 80) + "..." : commentContent);
+            var wakeMatch = wakeWordService.checkWakeWordBlocking(client, commentContent);
+
+            if (wakeMatch != null) {
+                // Wake word matched: trigger AI reply with the matched persona,
+                // bypassing normal page-level enable check
+                log.info("[CommentReconciler] Wake word '{}' matched for persona '{}', triggering reply for: {}",
+                    wakeMatch.wakeWord(), wakeMatch.personaName(), name);
+                orchestrator.processComment(name, null, false, wakeMatch.personaName(), true)
+                    .subscribeOn(Schedulers.boundedElastic())
+                    .subscribe(
+                        null,
+                        e -> log.error("[CommentReconciler] Error processing wake word comment {}: {}", name, e.getMessage(), e),
+                        () -> log.info("[CommentReconciler] Wake word processing completed for comment: {}", name)
+                    );
+            } else {
+                // Normal flow: read persona name from the post's annotations
+                String personaName = personaResolver.getPersonaNameFromCommentBlocking(client, comment);
 
                 // Top-level comment → always trigger AI reply
                 log.info("[CommentReconciler] New top-level comment detected: {}, personaName: {}", name, personaName);
-                asyncStarted.set(true);
+                orchestrator.processComment(name, null, false, personaName, false)
-                orchestrator.processComment(name, null, false, personaName)
                     .subscribeOn(Schedulers.boundedElastic())
-                    .doFinally(signal -> {
-                        processingLocks.remove(name);
-                        log.debug("[CommentReconciler] Released processing lock for: {}", name);
-                    })
                     .subscribe(
                         null,
                         e -> log.error("[CommentReconciler] Error processing comment {}: {}", name, e.getMessage(), e),
                         () -> log.info("[CommentReconciler] Processing completed for comment: {}", name)
                     );
-            });
-        } catch (Exception e) {
-            log.error("[CommentReconciler] Error in reconcile for {}: {}", name, e.getMessage(), e);
-        } finally {
-            // Only release lock here if async processing was NOT started
-            // (async path releases lock in doFinally)
-            if (!asyncStarted.get()) {
-                processingLocks.remove(name);
             }
-        }
+        });
 
         return Result.doNotRetry();
     }
@@ -139,61 +131,6 @@ public class CommentReconciler implements Reconciler<Reconciler.Request> {
         return false;
     }
 
-    /**
-     * Read persona name from the post's annotations associated with this comment.
-     */
-    private String getPersonaNameFromComment(Comment comment) {
-        var subjectRef = comment.getSpec().getSubjectRef();
-        if (subjectRef == null || !"Post".equals(subjectRef.getKind())) {
-            return null;
-        }
-        String postName = subjectRef.getName();
-        return client.fetch(Post.class, postName)
-            .map(post -> {
-                // 1. 文章注解优先
-                var annotations = post.getMetadata().getAnnotations();
-                if (annotations != null) {
-                    String persona = annotations.get(AI_PERSONA_ANNOTATION);
-                    if (persona != null && !persona.isBlank()) {
-                        return persona;
-                    }
-                }
-                // 2. 分类注解
-                var spec = post.getSpec();
-                if (spec != null && spec.getCategories() != null) {
-                    for (String categoryName : spec.getCategories()) {
-                        var cat = client.fetch(Category.class, categoryName).orElse(null);
-                        if (cat != null) {
-                            var catAnnotations = cat.getMetadata().getAnnotations();
-                            if (catAnnotations != null) {
-                                String catPersona = catAnnotations.get(AI_PERSONA_ANNOTATION);
-                                if (catPersona != null && !catPersona.isBlank()) {
-                                    return catPersona;
-                                }
-                            }
-                        }
-                    }
-                }
-                // 3. 标签注解
-                if (spec != null && spec.getTags() != null) {
-                    for (String tagName : spec.getTags()) {
-                        var tag = client.fetch(Tag.class, tagName).orElse(null);
-                        if (tag != null) {
-                            var tagAnnotations = tag.getMetadata().getAnnotations();
-                            if (tagAnnotations != null) {
-                                String tagPersona = tagAnnotations.get(AI_PERSONA_ANNOTATION);
-                                if (tagPersona != null && !tagPersona.isBlank()) {
-                                    return tagPersona;
-                                }
-                            }
-                        }
-                    }
-                }
-                return null;
-            })
-            .orElse(null);
-    }
-
     private boolean isProcessed(Map<String, String> annotations) {
         return annotations != null && "true".equals(annotations.get(PROCESSED_ANNOTATION));
     }
@@ -207,6 +144,22 @@ public class CommentReconciler implements Reconciler<Reconciler.Request> {
         annotations.put(PROCESSED_ANNOTATION, "true");
     }
 
+    private String getCommentContent(Comment comment) {
+        if (comment.getSpec() == null) return "";
+        // Always strip HTML to get plain text for wake word matching
+        String raw = comment.getSpec().getRaw();
+        if (raw != null && !raw.isBlank()) {
+            // raw might still contain HTML in some cases, always strip
+            String plain = org.jsoup.Jsoup.clean(raw, org.jsoup.safety.Safelist.none()).trim();
+            if (!plain.isBlank()) return plain;
+        }
+        String content = comment.getSpec().getContent();
+        if (content != null && !content.isBlank()) {
+            return org.jsoup.Jsoup.clean(content, org.jsoup.safety.Safelist.none()).trim();
+        }
+        return "";
+    }
+
     @Override
     public Controller setupWith(ControllerBuilder builder) {
         return builder

src/main/java/top/nxxy335/commentaiautopilot/listener/ReplyReconciler.java

@@ -4,17 +4,16 @@ import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.stereotype.Component;
 import reactor.core.scheduler.Schedulers;
-import run.halo.app.core.extension.content.Category;
 import run.halo.app.core.extension.content.Comment;
-import run.halo.app.core.extension.content.Post;
 import run.halo.app.core.extension.content.Reply;
-import run.halo.app.core.extension.content.Tag;
 import run.halo.app.extension.ExtensionClient;
 import run.halo.app.extension.controller.Controller;
 import run.halo.app.extension.controller.ControllerBuilder;
 import run.halo.app.extension.controller.Reconciler;
 import top.nxxy335.commentaiautopilot.extension.AiCommentReply;
 import top.nxxy335.commentaiautopilot.service.AiReplyOrchestrator;
+import top.nxxy335.commentaiautopilot.service.PersonaResolver;
+import top.nxxy335.commentaiautopilot.service.WakeWordService;
 
 import java.time.Instant;
 import java.util.HashMap;
@@ -27,11 +26,12 @@ public class ReplyReconciler implements Reconciler<Reconciler.Request> {
 
     private final ExtensionClient client;
     private final AiReplyOrchestrator orchestrator;
+    private final PersonaResolver personaResolver;
+    private final WakeWordService wakeWordService;
 
     private static final String PROCESSED_ANNOTATION = "comment-ai-autopilot.nxxy335.top/processed";
     private static final String AI_PERSONA_OWNER_PREFIX = "ai-persona-";
     private static final String AI_MARKER_ANNOTATION = "comment-ai-autopilot.nxxy335.top/is-ai";
-    private static final String AI_PERSONA_ANNOTATION = "comment-ai-autopilot.nxxy335.top/ai-persona";
 
     // Record the time when this bean was created (plugin startup time)
     private final Instant pluginStartTime = Instant.now();
@@ -77,25 +77,21 @@ public class ReplyReconciler implements Reconciler<Reconciler.Request> {
                 return;
             }
 
-            // Check if this reply is specifically replying to an AI reply
+            // Check for wake word FIRST - wake word can bypass "must be reply to AI" check
-            // by checking the quoteReply field
+            String replyContent = getReplyContent(reply);
-            String quoteReply = reply.getSpec().getQuoteReply();
+            log.info("[ReplyReconciler] Wake word check for reply {}: content='{}'",
-
+                name, replyContent.length() > 80 ? replyContent.substring(0, 80) + "..." : replyContent);
-            if (quoteReply == null || quoteReply.isBlank()) {
+            var wakeMatch = wakeWordService.checkWakeWordBlocking(client, replyContent);
-                // No quoteReply - this is a direct reply to the top-level comment,
-                // NOT a reply to AI. Skip it (CommentReconciler handles top-level comments).
-                log.debug("[ReplyReconciler] Reply {} has no quoteReply, skipping (not a reply to AI)", name);
-                markProcessed(reply);
-                client.update(reply);
-                return;
-            }
 
             // This reply quotes another reply - check if the quoted reply is from AI
-            boolean isReplyToAi = isAiReply(quoteReply);
+            String quoteReply = reply.getSpec().getQuoteReply();
+            boolean isReplyToAi = quoteReply != null && !quoteReply.isBlank() && isAiReply(quoteReply);
             log.debug("[ReplyReconciler] Reply {} quotes {}, isAiReply={}", name, quoteReply, isReplyToAi);
 
-            if (!isReplyToAi) {
+            // Skip if not a reply to AI AND no wake word matched
-                log.debug("[ReplyReconciler] Not a reply to AI, skipping: {}", name);
+            if (!isReplyToAi && wakeMatch == null) {
+                // No quoteReply or not replying to AI, and no wake word - skip
+                log.debug("[ReplyReconciler] Not a reply to AI and no wake word, skipping: {}", name);
                 markProcessed(reply);
                 client.update(reply);
                 return;
@@ -119,16 +115,32 @@ public class ReplyReconciler implements Reconciler<Reconciler.Request> {
             markProcessed(reply);
             client.update(reply);
 
-            // Reply to AI → trigger AI reply (conversation continuation)
+            if (wakeMatch != null) {
-            String personaName = getPersonaNameFromComment(parentCommentName);
+                // Wake word matched: trigger AI reply with the matched persona,
-            log.info("[ReplyReconciler] Reply to AI detected: {}, triggering conversation, personaName: {}", name, personaName);
+                // bypassing the "must be reply to AI" check and page-level enable check
-            orchestrator.processComment(parentCommentName, name, true, personaName)
+                log.info("[ReplyReconciler] Wake word '{}' matched for persona '{}', triggering reply for: {}",
-                .subscribeOn(Schedulers.boundedElastic())
+                    wakeMatch.wakeWord(), wakeMatch.personaName(), name);
-                .subscribe(
+                orchestrator.processComment(parentCommentName, name, true, wakeMatch.personaName(), true)
-                    null,
+                    .subscribeOn(Schedulers.boundedElastic())
-                    e -> log.error("[ReplyReconciler] Error processing reply {}: {}", name, e.getMessage(), e),
+                    .subscribe(
-                    () -> log.info("[ReplyReconciler] Processing completed for reply: {}", name)
+                        null,
-                );
+                        e -> log.error("[ReplyReconciler] Error processing wake word reply {}: {}", name, e.getMessage(), e),
+                        () -> log.info("[ReplyReconciler] Wake word processing completed for reply: {}", name)
+                    );
+            } else if (isReplyToAi) {
+                // Normal flow: reply to AI → trigger AI reply (conversation continuation)
+                String personaName = client.fetch(Comment.class, parentCommentName)
+                    .map(comment -> personaResolver.getPersonaNameFromCommentBlocking(client, comment))
+                    .orElse(null);
+                log.info("[ReplyReconciler] Reply to AI detected: {}, triggering conversation, personaName: {}", name, personaName);
+                orchestrator.processComment(parentCommentName, name, true, personaName, false)
+                    .subscribeOn(Schedulers.boundedElastic())
+                    .subscribe(
+                        null,
+                        e -> log.error("[ReplyReconciler] Error processing reply {}: {}", name, e.getMessage(), e),
+                        () -> log.info("[ReplyReconciler] Processing completed for reply: {}", name)
+                    );
+            }
         });
 
         return Result.doNotRetry();
@@ -154,65 +166,6 @@ public class ReplyReconciler implements Reconciler<Reconciler.Request> {
             .orElse(false);
     }
 
-    /**
-     * Read persona name from the post's annotations associated with the parent comment.
-     */
-    private String getPersonaNameFromComment(String commentName) {
-        return client.fetch(Comment.class, commentName)
-            .map(comment -> {
-                var subjectRef = comment.getSpec().getSubjectRef();
-                if (subjectRef == null || !"Post".equals(subjectRef.getKind())) {
-                    return null;
-                }
-                String postName = subjectRef.getName();
-                return client.fetch(Post.class, postName)
-                    .map(post -> {
-                        // 1. 文章注解优先
-                        var annotations = post.getMetadata().getAnnotations();
-                        if (annotations != null) {
-                            String persona = annotations.get(AI_PERSONA_ANNOTATION);
-                            if (persona != null && !persona.isBlank()) {
-                                return persona;
-                            }
-                        }
-                        // 2. 分类注解
-                        var spec = post.getSpec();
-                        if (spec != null && spec.getCategories() != null) {
-                            for (String categoryName : spec.getCategories()) {
-                                var cat = client.fetch(Category.class, categoryName).orElse(null);
-                                if (cat != null) {
-                                    var catAnnotations = cat.getMetadata().getAnnotations();
-                                    if (catAnnotations != null) {
-                                        String catPersona = catAnnotations.get(AI_PERSONA_ANNOTATION);
-                                        if (catPersona != null && !catPersona.isBlank()) {
-                                            return catPersona;
-                                        }
-                                    }
-                                }
-                            }
-                        }
-                        // 3. 标签注解
-                        if (spec != null && spec.getTags() != null) {
-                            for (String tagName : spec.getTags()) {
-                                var tag = client.fetch(Tag.class, tagName).orElse(null);
-                                if (tag != null) {
-                                    var tagAnnotations = tag.getMetadata().getAnnotations();
-                                    if (tagAnnotations != null) {
-                                        String tagPersona = tagAnnotations.get(AI_PERSONA_ANNOTATION);
-                                        if (tagPersona != null && !tagPersona.isBlank()) {
-                                            return tagPersona;
-                                        }
-                                    }
-                                }
-                            }
-                        }
-                        return null;
-                    })
-                    .orElse(null);
-            })
-            .orElse(null);
-    }
-
     private boolean isProcessed(Map<String, String> annotations) {
         return annotations != null && "true".equals(annotations.get(PROCESSED_ANNOTATION));
     }
@@ -226,6 +179,22 @@ public class ReplyReconciler implements Reconciler<Reconciler.Request> {
         annotations.put(PROCESSED_ANNOTATION, "true");
     }
 
+    private String getReplyContent(Reply reply) {
+        if (reply.getSpec() == null) return "";
+        // Always strip HTML to get plain text for wake word matching
+        String raw = reply.getSpec().getRaw();
+        if (raw != null && !raw.isBlank()) {
+            // raw might still contain HTML in some cases, always strip
+            String plain = org.jsoup.Jsoup.clean(raw, org.jsoup.safety.Safelist.none()).trim();
+            if (!plain.isBlank()) return plain;
+        }
+        String content = reply.getSpec().getContent();
+        if (content != null && !content.isBlank()) {
+            return org.jsoup.Jsoup.clean(content, org.jsoup.safety.Safelist.none()).trim();
+        }
+        return "";
+    }
+
     @Override
     public Controller setupWith(ControllerBuilder builder) {
         return builder

src/main/java/top/nxxy335/commentaiautopilot/service/AiReplyCleanupService.java

@@ -17,6 +17,9 @@ import java.util.concurrent.Executors;
 import java.util.concurrent.ScheduledExecutorService;
 import java.util.concurrent.TimeUnit;
 
+import reactor.core.publisher.Mono;
+import reactor.core.publisher.Flux;
+
 @Component
 @Slf4j
 public class AiReplyCleanupService implements DisposableBean {
@@ -27,9 +30,9 @@ public class AiReplyCleanupService implements DisposableBean {
 
     private static final String CONFIG_MAP_NAME = "comment-ai-autopilot-configmap";
 
-    public AiReplyCleanupService(ReactiveExtensionClient client) {
+    public AiReplyCleanupService(ReactiveExtensionClient client, ObjectMapper objectMapper) {
         this.client = client;
-        this.objectMapper = new ObjectMapper();
+        this.objectMapper = objectMapper;
         this.scheduler = Executors.newSingleThreadScheduledExecutor(r -> {
             Thread t = new Thread(r, "ai-reply-cleanup");
             t.setDaemon(true);
@@ -40,85 +43,85 @@ public class AiReplyCleanupService implements DisposableBean {
     }
 
     public void dailyCleanup() {
-        try {
+        isCleanupEnabled()
-            Boolean enabled = client.fetch(ConfigMap.class, CONFIG_MAP_NAME)
+            .flatMap(enabled -> {
-                .mapNotNull(cm -> {
+                if (!Boolean.TRUE.equals(enabled)) {
-                    var data = cm.getData();
+                    log.debug("[Cleanup] Auto cleanup is disabled, skipping");
-                    if (data == null) return false;
+                    return Mono.empty();
-                    String cleanupJson = data.get("cleanup");
+                }
-                    if (cleanupJson == null || cleanupJson.isBlank()) return true;
+                return getRetentionDays()
-                    try {
+                    .flatMap(retentionDays -> executeCleanup(retentionDays)
-                        JsonNode node = objectMapper.readTree(cleanupJson);
+                        .doOnNext(deleted -> log.info("[Cleanup] Auto cleanup completed, deleted {} records older than {} days", deleted, retentionDays))
-                        return !node.has("cleanupEnabled") || node.get("cleanupEnabled").asBoolean(true);
+                    );
-                    } catch (Exception e) {
+            })
-                        log.warn("[Cleanup] Failed to parse cleanup config: {}", e.getMessage());
+            .subscribe(
-                        return true;
+                null,
-                    }
+                e -> log.error("[Cleanup] Error during daily cleanup: {}", e.getMessage(), e)
-                })
+            );
-                .defaultIfEmpty(true)
-                .block();
-
-            if (!Boolean.TRUE.equals(enabled)) {
-                log.debug("[Cleanup] Auto cleanup is disabled, skipping");
-                return;
-            }
-
-            int retentionDays = getRetentionDays();
-            long deleted = executeCleanup(retentionDays);
-            log.info("[Cleanup] Auto cleanup completed, deleted {} records older than {} days", deleted, retentionDays);
-        } catch (Exception e) {
-            log.error("[Cleanup] Error during daily cleanup: {}", e.getMessage(), e);
-        }
     }
 
-    public long executeCleanup(int retentionDays) {
+    private Mono<Boolean> isCleanupEnabled() {
+        return client.fetch(ConfigMap.class, CONFIG_MAP_NAME)
+            .mapNotNull(cm -> {
+                var data = cm.getData();
+                if (data == null) return false;
+                String cleanupJson = data.get("cleanup");
+                if (cleanupJson == null || cleanupJson.isBlank()) return true;
+                try {
+                    JsonNode node = objectMapper.readTree(cleanupJson);
+                    return !node.has("cleanupEnabled") || node.get("cleanupEnabled").asBoolean(true);
+                } catch (Exception e) {
+                    log.warn("[Cleanup] Failed to parse cleanup config: {}", e.getMessage());
+                    return true;
+                }
+            })
+            .defaultIfEmpty(true);
+    }
+
+    public Mono<Long> executeCleanup(int retentionDays) {
         Instant cutoff = Instant.now().minus(retentionDays, ChronoUnit.DAYS);
 
-        var oldRecords = client.listAll(AiCommentReply.class, ListOptions.builder().build(), Sort.unsorted())
+        return client.listAll(AiCommentReply.class, ListOptions.builder().build(), Sort.unsorted())
             .filter(r -> {
                 Instant created = r.getMetadata().getCreationTimestamp();
                 return created != null && created.isBefore(cutoff);
             })
             .collectList()
-            .block();
+            .flatMap(oldRecords -> {
-
+                if (oldRecords.isEmpty()) {
-        if (oldRecords == null || oldRecords.isEmpty()) {
+                    return Mono.just(0L);
-            return 0;
+                }
-        }
+                return Flux.fromIterable(oldRecords)
-
+                    .flatMap(record -> client.delete(record)
-        long deleted = 0;
+                        .thenReturn(1L)
-        for (var record : oldRecords) {
+                        .onErrorResume(e -> {
-            try {
+                            log.warn("[Cleanup] Failed to delete record {}: {}", record.getMetadata().getName(), e.getMessage());
-                client.delete(record).block();
+                            return Mono.just(0L);
-                deleted++;
+                        })
-            } catch (Exception e) {
+                    )
-                log.warn("[Cleanup] Failed to delete record {}: {}", record.getMetadata().getName(), e.getMessage());
+                    .reduce(0L, Long::sum);
-            }
+            });
-        }
-        return deleted;
     }
 
-    public int getRetentionDays() {
+    public Mono<Integer> getRetentionDays() {
-        try {
+        return client.fetch(ConfigMap.class, CONFIG_MAP_NAME)
-            return client.fetch(ConfigMap.class, CONFIG_MAP_NAME)
+            .mapNotNull(cm -> {
-                .mapNotNull(cm -> {
+                var data = cm.getData();
-                    var data = cm.getData();
+                if (data == null) return 30;
-                    if (data == null) return 30;
+                String cleanupJson = data.get("cleanup");
-                    String cleanupJson = data.get("cleanup");
+                if (cleanupJson == null || cleanupJson.isBlank()) return 30;
-                    if (cleanupJson == null || cleanupJson.isBlank()) return 30;
+                try {
-                    try {
+                    JsonNode node = objectMapper.readTree(cleanupJson);
-                        JsonNode node = objectMapper.readTree(cleanupJson);
+                    return node.has("retentionDays") ? node.get("retentionDays").asInt(30) : 30;
-                        return node.has("retentionDays") ? node.get("retentionDays").asInt(30) : 30;
+                } catch (Exception e) {
-                    } catch (Exception e) {
+                    return 30;
-                        return 30;
+                }
-                    }
+            })
-                })
+            .defaultIfEmpty(30)
-                .defaultIfEmpty(30)
+            .onErrorResume(e -> {
-                .block();
+                log.warn("[Cleanup] Failed to read retentionDays config: {}", e.getMessage());
-        } catch (Exception e) {
+                return Mono.just(30);
-            log.warn("[Cleanup] Failed to read retentionDays config: {}", e.getMessage());
+            });
-            return 30;
-        }
     }
 
     @Override

src/main/java/top/nxxy335/commentaiautopilot/service/AiReplyOrchestrator.java

@@ -49,7 +49,8 @@ public class AiReplyOrchestrator {
                                CommentReplyPublisher commentReplyPublisher,
                                FilterService filterService,
                                RateLimitService rateLimitService,
-                               ReactiveExtensionClient client) {
+                               ReactiveExtensionClient client,
+                               ObjectMapper objectMapper) {
         this.contextExtractor = contextExtractor;
         this.promptBuilder = promptBuilder;
         this.aiReplyService = aiReplyService;
@@ -59,7 +60,7 @@ public class AiReplyOrchestrator {
         this.filterService = filterService;
         this.rateLimitService = rateLimitService;
         this.client = client;
-        this.objectMapper = new ObjectMapper();
+        this.objectMapper = objectMapper;
     }
 
     /**
@@ -69,9 +70,10 @@ public class AiReplyOrchestrator {
      * @param replyName       the Reply name that triggered this (null for top-level comments)
      * @param isAiConversation true when someone replied to AI's reply (conversation continuation)
      * @param personaName     the persona name to use (null for default persona)
+     * @param wakeWordTriggered true when triggered by a wake word (bypasses page-level enable check)
      */
     public Mono<Void> processComment(String commentName, String replyName, boolean isAiConversation,
-                                      String personaName) {
+                                      String personaName, boolean wakeWordTriggered) {
         String lockKey = isAiConversation ? commentName + ":conv:" + replyName : commentName + ":top";
 
         // Clean up stale locks before acquiring new one
@@ -83,12 +85,12 @@ public class AiReplyOrchestrator {
             return Mono.empty();
         }
 
-        log.info("[Orchestrator] Start processing: comment={}, replyName={}, isAiConversation={}, personaName={}",
+        log.info("[Orchestrator] Start processing: comment={}, replyName={}, isAiConversation={}, personaName={}, wakeWordTriggered={}",
-            commentName, replyName, isAiConversation, personaName);
+            commentName, replyName, isAiConversation, personaName, wakeWordTriggered);
 
         return isAutoReplyEnabled()
             .flatMap(enabled -> {
-                if (!enabled) {
+                if (!enabled && !wakeWordTriggered) {
                     log.info("[Orchestrator] Auto reply disabled, skipping: {}", commentName);
                     return Mono.empty();
                 }
@@ -98,42 +100,24 @@ public class AiReplyOrchestrator {
                             log.info("[Orchestrator] 速率限制，跳过: {}", commentName);
                             return Mono.empty();
                         }
+                        // Wake word triggered: skip page-level annotation check
+                        if (wakeWordTriggered) {
+                            return checkBlockedCommenters(commentName)
+                                .flatMap(blocked -> {
+                                    if (blocked) {
+                                        log.info("[Orchestrator] Commenter blocked, skipping wake word: {}", commentName);
+                                        return Mono.empty();
+                                    }
+                                    return proceedWithProcess(commentName, replyName, isAiConversation, personaName);
+                                });
+                        }
                         return filterService.shouldProcess(commentName)
                             .flatMap(shouldProcess -> {
                                 if (!shouldProcess) {
                                     log.info("[Orchestrator] Filtered out by rules: {}", commentName);
                                     return Mono.empty();
                                 }
-                        // For top-level comments: skip if we already have ANY reply record
+                                return proceedWithProcess(commentName, replyName, isAiConversation, personaName);
-                        // For AI conversation: skip if we already replied to THIS specific reply
-                        if (!isAiConversation) {
-                            return hasExistingReply(commentName)
-                                .flatMap(hasReply -> {
-                                    if (hasReply) {
-                                        log.info("[Orchestrator] Already have reply record for: {}, skipping", commentName);
-                                        return Mono.empty();
-                                    }
-                                    return doProcess(commentName, replyName, isAiConversation, personaName);
-                                });
-                        }
-                        // Check conversation rounds limit
-                        return getMaxConversationRounds()
-                            .flatMap(maxRounds -> getConversationRounds(commentName)
-                                .flatMap(rounds -> {
-                                    if (rounds >= maxRounds) {
-                                        log.info("[Orchestrator] 对话轮次已达上限({}/{}), 跳过: {}", rounds, maxRounds, commentName);
-                                        return Mono.empty();
-                                    }
-                                    return hasExistingConversationReply(replyName)
-                                        .flatMap(hasReply -> {
-                                            if (hasReply) {
-                                                log.info("[Orchestrator] Already replied to reply: {}, skipping", replyName);
-                                                return Mono.empty();
-                                            }
-                                            return doProcess(commentName, replyName, isAiConversation, personaName);
-                                        });
-                                })
-                            );
                             });
                     });
             })
@@ -146,6 +130,81 @@ public class AiReplyOrchestrator {
             .then();
     }
 
+    /**
+     * Proceed with processing after all checks have passed.
+     * Handles dedup checks and conversation round limits.
+     */
+    private Mono<Void> proceedWithProcess(String commentName, String replyName,
+                                           boolean isAiConversation, String personaName) {
+        // For top-level comments: skip if we already have ANY reply record
+        // For AI conversation: skip if we already replied to THIS specific reply
+        if (!isAiConversation) {
+            return hasExistingReply(commentName)
+                .flatMap(hasReply -> {
+                    if (hasReply) {
+                        log.info("[Orchestrator] Already have reply record for: {}, skipping", commentName);
+                        return Mono.empty();
+                    }
+                    return doProcess(commentName, replyName, isAiConversation, personaName);
+                });
+        }
+        // Check conversation rounds limit
+        return getMaxConversationRounds()
+            .flatMap(maxRounds -> getConversationRounds(commentName)
+                .flatMap(rounds -> {
+                    if (rounds >= maxRounds) {
+                        log.info("[Orchestrator] 对话轮次已达上限({}/{}), 跳过: {}", rounds, maxRounds, commentName);
+                        return Mono.empty();
+                    }
+                    return hasExistingConversationReply(replyName)
+                        .flatMap(hasReply -> {
+                            if (hasReply) {
+                                log.info("[Orchestrator] Already replied to reply: {}, skipping", replyName);
+                                return Mono.empty();
+                            }
+                            return doProcess(commentName, replyName, isAiConversation, personaName);
+                        });
+                })
+            );
+    }
+
+    /**
+     * Check if the commenter is in the blocked list.
+     */
+    private Mono<Boolean> checkBlockedCommenters(String commentName) {
+        return client.fetch(run.halo.app.core.extension.content.Comment.class, commentName)
+            .flatMap(comment -> {
+                var owner = comment.getSpec().getOwner();
+                if (owner == null) return Mono.just(false);
+                String displayName = owner.getDisplayName();
+                String email = run.halo.app.core.extension.content.Comment.CommentOwner.KIND_EMAIL.equals(owner.getKind())
+                    ? owner.getName() : "";
+                return client.fetch(ConfigMap.class, CONFIG_MAP_NAME)
+                    .mapNotNull(cm -> {
+                        var data = cm.getData();
+                        if (data == null) return false;
+                        String basicJson = data.get("basic");
+                        if (basicJson == null || basicJson.isBlank()) return false;
+                        try {
+                            JsonNode node = objectMapper.readTree(basicJson);
+                            String blockedStr = node.has("blockedCommenters") ? node.get("blockedCommenters").asText("") : "";
+                            if (blockedStr.isBlank()) return false;
+                            for (String item : blockedStr.split(",")) {
+                                String trimmed = item.trim();
+                                if (!trimmed.isEmpty() && (trimmed.equalsIgnoreCase(displayName) || trimmed.equalsIgnoreCase(email))) {
+                                    return true;
+                                }
+                            }
+                            return false;
+                        } catch (Exception e) {
+                            return false;
+                        }
+                    })
+                    .defaultIfEmpty(false);
+            })
+            .defaultIfEmpty(false);
+    }
+
     private Mono<Void> doProcess(String commentName, String replyName, boolean isAiConversation,
                                   String personaName) {
         return getModelName().flatMap(modelName ->
@@ -524,6 +583,7 @@ public class AiReplyOrchestrator {
         record.getSpec().setCommentId(context.commentId());
         record.getSpec().setPostId(context.postId());
         record.getSpec().setPostSlug(context.postSlug());
+        record.getSpec().setPostKind(context.postKind());
         record.getSpec().setReply("");
         record.getSpec().setScore(0);
         record.getSpec().setStatus("PENDING");

src/main/java/top/nxxy335/commentaiautopilot/service/CommentReplyPublisher.java

@@ -8,9 +8,8 @@ import run.halo.app.core.extension.content.Reply;
 import run.halo.app.extension.Metadata;
 import run.halo.app.extension.ReactiveExtensionClient;
 import top.nxxy335.commentaiautopilot.extension.AiPersona;
+import top.nxxy335.commentaiautopilot.util.GravatarUtil;
 
-import java.nio.charset.StandardCharsets;
-import java.security.MessageDigest;
 import java.time.Instant;
 import java.util.HashMap;
 import java.util.Map;
@@ -120,7 +119,7 @@ public class CommentReplyPublisher {
             ownerAnnotations.put("comment-ai-autopilot.nxxy335.top/is-ai", "true");
             // 使用Gravatar邮箱头像
             if (email != null && !email.isBlank()) {
-                String gravatarUrl = generateGravatarUrl(email);
+                String gravatarUrl = GravatarUtil.generateUrl(email);
                 ownerAnnotations.put(Comment.CommentOwner.AVATAR_ANNO, gravatarUrl);
             }
             owner.setAnnotations(ownerAnnotations);
@@ -181,22 +180,4 @@ public class CommentReplyPublisher {
     private String generateReplyName() {
         return "ai-comment-reply-" + UUID.randomUUID().toString().substring(0, 8);
     }
-
-    /**
-     * Generate Gravatar URL from email address using SHA-256 hash.
-     */
-    private String generateGravatarUrl(String email) {
-        try {
-            var digest = MessageDigest.getInstance("SHA-256");
-            var hashBytes = digest.digest(email.trim().toLowerCase().getBytes(StandardCharsets.UTF_8));
-            var hexString = new StringBuilder();
-            for (byte b : hashBytes) {
-                hexString.append(String.format("%02x", b));
-            }
-            return "https://cn.cravatar.com/avatar/" + hexString;
-        } catch (Exception e) {
-            log.error("[Publisher] Failed to generate Gravatar URL: {}", e.getMessage());
-            return "";
-        }
-    }
 }

src/main/java/top/nxxy335/commentaiautopilot/service/ContextExtractor.java

@@ -10,6 +10,7 @@ import run.halo.app.content.ContentWrapper;
 import run.halo.app.content.PostContentService;
 import run.halo.app.core.extension.content.Comment;
 import run.halo.app.core.extension.content.Post;
+import run.halo.app.core.extension.content.SinglePage;
 import run.halo.app.core.extension.content.Reply;
 import run.halo.app.extension.ReactiveExtensionClient;
 
@@ -113,10 +114,15 @@ public class ContextExtractor {
                             isAiConversation,
                             formatPostDate(post),
                             commentCount,
-                            ""
+                            "",
+                            "Post"
                         ))
                     )
                 )
+                .onErrorResume(e -> {
+                    log.warn("[ContextExtractor] Failed to fetch Post {}: {}", postName, e.getMessage());
+                    return Mono.empty();
+                })
                 .defaultIfEmpty(new CommentContext(
                     comment.getMetadata().getName(),
                     postName,
@@ -129,7 +135,51 @@ public class ContextExtractor {
                     isAiConversation,
                     "",
                     0,
-                    ""
+                    "",
+                    "Post"
+                ));
+        }
+
+        if (subjectRef != null && "SinglePage".equals(subjectRef.getKind())) {
+            String postName = subjectRef.getName();
+            return client.fetch(SinglePage.class, postName)
+                .flatMap(singlePage -> getSinglePageContent(postName)
+                    .flatMap(content -> getCommentCount(comment.getMetadata().getName())
+                        .map(commentCount -> new CommentContext(
+                            comment.getMetadata().getName(),
+                            postName,
+                            singlePage.getSpec().getSlug(),
+                            commentContent,
+                            commentOwner,
+                            singlePage.getSpec().getTitle(),
+                            content,
+                            null,
+                            isAiConversation,
+                            formatSinglePageDate(singlePage),
+                            commentCount,
+                            "",
+                            "SinglePage"
+                        ))
+                    )
+                )
+                .onErrorResume(e -> {
+                    log.warn("[ContextExtractor] Failed to fetch SinglePage {}: {}", postName, e.getMessage());
+                    return Mono.empty();
+                })
+                .defaultIfEmpty(new CommentContext(
+                    comment.getMetadata().getName(),
+                    postName,
+                    "",
+                    commentContent,
+                    commentOwner,
+                    "",
+                    "",
+                    null,
+                    isAiConversation,
+                    "",
+                    0,
+                    "",
+                    "SinglePage"
                 ));
         }
 
@@ -145,6 +195,7 @@ public class ContextExtractor {
             isAiConversation,
             "",
             0,
+            "",
             ""
         ));
     }
@@ -179,11 +230,16 @@ public class ContextExtractor {
                                 isAiConversation,
                                 formatPostDate(post),
                                 commentCount,
-                                history
+                                history,
+                                "Post"
                             ))
                         )
                     )
                 )
+                .onErrorResume(e -> {
+                    log.warn("[ContextExtractor] Failed to fetch Post {} for reply: {}", postName, e.getMessage());
+                    return Mono.empty();
+                })
                 .defaultIfEmpty(new CommentContext(
                     commentName,
                     postName,
@@ -196,7 +252,53 @@ public class ContextExtractor {
                     isAiConversation,
                     "",
                     0,
-                    ""
+                    "",
+                    "Post"
+                ));
+        }
+
+        if (subjectRef != null && "SinglePage".equals(subjectRef.getKind())) {
+            String postName = subjectRef.getName();
+            return client.fetch(SinglePage.class, postName)
+                .flatMap(singlePage -> getSinglePageContent(postName)
+                    .flatMap(content -> getCommentCount(commentName)
+                        .flatMap(commentCount -> historyMono
+                            .map(history -> new CommentContext(
+                                commentName,
+                                postName,
+                                singlePage.getSpec().getSlug(),
+                                replyContent,
+                                replyOwner,
+                                singlePage.getSpec().getTitle(),
+                                content,
+                                replyName,
+                                isAiConversation,
+                                formatSinglePageDate(singlePage),
+                                commentCount,
+                                history,
+                                "SinglePage"
+                            ))
+                        )
+                    )
+                )
+                .onErrorResume(e -> {
+                    log.warn("[ContextExtractor] Failed to fetch SinglePage {} for reply: {}", postName, e.getMessage());
+                    return Mono.empty();
+                })
+                .defaultIfEmpty(new CommentContext(
+                    commentName,
+                    postName,
+                    "",
+                    replyContent,
+                    replyOwner,
+                    "",
+                    "",
+                    replyName,
+                    isAiConversation,
+                    "",
+                    0,
+                    "",
+                    "SinglePage"
                 ));
         }
 
@@ -213,7 +315,8 @@ public class ContextExtractor {
                 isAiConversation,
                 "",
                 0,
-                history
+                history,
+                ""
             ));
     }
 
@@ -278,6 +381,23 @@ public class ContextExtractor {
             .defaultIfEmpty("");
     }
 
+    private Mono<String> getSinglePageContent(String pageName) {
+        // SinglePage doesn't have a dedicated ContentService in Halo API,
+        // and Snapshot content requires patch merging which is too complex.
+        // Use the excerpt from status as a fallback for context.
+        return client.fetch(SinglePage.class, pageName)
+            .mapNotNull(page -> {
+                if (page.getStatus() != null && page.getStatus().getExcerpt() != null) {
+                    String excerpt = page.getStatus().getExcerpt();
+                    if (excerpt != null && !excerpt.isBlank()) {
+                        return excerpt;
+                    }
+                }
+                return "";
+            })
+            .defaultIfEmpty("");
+    }
+
     private String formatPostDate(Post post) {
         var publishTime = post.getSpec().getPublishTime();
         if (publishTime != null) {
@@ -290,6 +410,18 @@ public class ContextExtractor {
         return "";
     }
 
+    private String formatSinglePageDate(SinglePage singlePage) {
+        var publishTime = singlePage.getSpec().getPublishTime();
+        if (publishTime != null) {
+            return publishTime.toString().substring(0, 10);
+        }
+        var creationTimestamp = singlePage.getMetadata().getCreationTimestamp();
+        if (creationTimestamp != null) {
+            return creationTimestamp.toString().substring(0, 10);
+        }
+        return "";
+    }
+
     private Mono<Integer> getCommentCount(String commentName) {
         return client.list(Reply.class,
                 reply -> commentName.equals(reply.getSpec().getCommentName()),
@@ -311,6 +443,7 @@ public class ContextExtractor {
         boolean isAiConversation,
         String postDate,
         int commentCount,
-        String conversationHistory
+        String conversationHistory,
+        String postKind
     ) {}
 }

src/main/java/top/nxxy335/commentaiautopilot/service/FilterService.java

@@ -28,9 +28,9 @@ public class FilterService {
     private static final String ANNOTATION_KEY = "comment-ai-autopilot.nxxy335.top/ai-reply-enabled";
     private static final String GROUP_CONTENT = "content.halo.run";
 
-    public FilterService(ReactiveExtensionClient client) {
+    public FilterService(ReactiveExtensionClient client, ObjectMapper objectMapper) {
         this.client = client;
-        this.objectMapper = new ObjectMapper();
+        this.objectMapper = objectMapper;
     }
 
     public Mono<Boolean> shouldProcess(Comment comment) {

src/main/java/top/nxxy335/commentaiautopilot/service/PersonaResolver.java

@@ -0,0 +1,175 @@
+package top.nxxy335.commentaiautopilot.service;
+
+import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.stereotype.Component;
+import run.halo.app.core.extension.content.Category;
+import run.halo.app.core.extension.content.Comment;
+import run.halo.app.core.extension.content.Post;
+import run.halo.app.core.extension.content.Tag;
+import run.halo.app.extension.ExtensionClient;
+import run.halo.app.extension.ReactiveExtensionClient;
+import reactor.core.publisher.Mono;
+
+/**
+ * Shared service for resolving AI persona name from a comment's associated
+ * post/category/tag annotations.
+ *
+ * <p>Priority: Post annotation &gt; Category annotation &gt; Tag annotation
+ */
+@Component
+@Slf4j
+@RequiredArgsConstructor
+public class PersonaResolver {
+
+    private static final String AI_PERSONA_ANNOTATION = "comment-ai-autopilot.nxxy335.top/ai-persona";
+
+    private final ReactiveExtensionClient reactiveClient;
+
+    /**
+     * Resolve persona name from a comment (reactive version).
+     * Reads the post's annotations, then falls back to category and tag annotations.
+     *
+     * @param commentName the Comment metadata.name
+     * @return the persona name, or empty string if none found
+     */
+    public Mono<String> getPersonaNameFromComment(String commentName) {
+        return reactiveClient.fetch(Comment.class, commentName)
+            .flatMap(comment -> {
+                var subjectRef = comment.getSpec().getSubjectRef();
+                if (subjectRef == null || !"Post".equals(subjectRef.getKind())) {
+                    return Mono.just("");
+                }
+                String postName = subjectRef.getName();
+                return resolveFromPost(postName);
+            })
+            .defaultIfEmpty("");
+    }
+
+    private Mono<String> resolveFromPost(String postName) {
+        return reactiveClient.fetch(Post.class, postName)
+            .flatMap(post -> {
+                // 1. Post annotation takes priority
+                var annotations = post.getMetadata().getAnnotations();
+                if (annotations != null) {
+                    String persona = annotations.get(AI_PERSONA_ANNOTATION);
+                    if (persona != null && !persona.isBlank()) {
+                        return Mono.just(persona);
+                    }
+                }
+                // 2. Category annotations
+                var spec = post.getSpec();
+                if (spec != null && spec.getCategories() != null) {
+                    for (String categoryName : spec.getCategories()) {
+                        var persona = resolveFromCategory(categoryName);
+                        if (persona != null) return Mono.just(persona);
+                    }
+                }
+                // 3. Tag annotations
+                if (spec != null && spec.getTags() != null) {
+                    for (String tagName : spec.getTags()) {
+                        var persona = resolveFromTag(tagName);
+                        if (persona != null) return Mono.just(persona);
+                    }
+                }
+                return Mono.just("");
+            })
+            .defaultIfEmpty("");
+    }
+
+    private String resolveFromCategory(String categoryName) {
+        // Use block() here because this is called from a Reconciler (sync context)
+        // For reactive context, the caller should use the reactive version
+        try {
+            return reactiveClient.fetch(Category.class, categoryName)
+                .mapNotNull(cat -> {
+                    var catAnnotations = cat.getMetadata().getAnnotations();
+                    if (catAnnotations != null) {
+                        String catPersona = catAnnotations.get(AI_PERSONA_ANNOTATION);
+                        if (catPersona != null && !catPersona.isBlank()) {
+                            return catPersona;
+                        }
+                    }
+                    return null;
+                })
+                .block();
+        } catch (Exception e) {
+            return null;
+        }
+    }
+
+    private String resolveFromTag(String tagName) {
+        try {
+            return reactiveClient.fetch(Tag.class, tagName)
+                .mapNotNull(tag -> {
+                    var tagAnnotations = tag.getMetadata().getAnnotations();
+                    if (tagAnnotations != null) {
+                        String tagPersona = tagAnnotations.get(AI_PERSONA_ANNOTATION);
+                        if (tagPersona != null && !tagPersona.isBlank()) {
+                            return tagPersona;
+                        }
+                    }
+                    return null;
+                })
+                .block();
+        } catch (Exception e) {
+            return null;
+        }
+    }
+
+    /**
+     * Resolve persona name from a comment using blocking ExtensionClient
+     * (for use in Reconciler sync context).
+     */
+    public String getPersonaNameFromCommentBlocking(ExtensionClient client, Comment comment) {
+        var subjectRef = comment.getSpec().getSubjectRef();
+        if (subjectRef == null || !"Post".equals(subjectRef.getKind())) {
+            return null;
+        }
+        String postName = subjectRef.getName();
+        return client.fetch(Post.class, postName)
+            .map(post -> {
+                // 1. Post annotation
+                var annotations = post.getMetadata().getAnnotations();
+                if (annotations != null) {
+                    String persona = annotations.get(AI_PERSONA_ANNOTATION);
+                    if (persona != null && !persona.isBlank()) {
+                        return persona;
+                    }
+                }
+                // 2. Category annotations
+                var spec = post.getSpec();
+                if (spec != null && spec.getCategories() != null) {
+                    for (String categoryName : spec.getCategories()) {
+                        var cat = client.fetch(Category.class, categoryName).orElse(null);
+                        if (cat != null) {
+                            var catAnnotations = cat.getMetadata().getAnnotations();
+                            if (catAnnotations != null) {
+                                String catPersona = catAnnotations.get(AI_PERSONA_ANNOTATION);
+                                if (catPersona != null && !catPersona.isBlank()) {
+                                    return catPersona;
+                                }
+                            }
+                        }
+                    }
+                }
+                // 3. Tag annotations
+                if (spec != null && spec.getTags() != null) {
+                    for (String tagName : spec.getTags()) {
+                        var tag = client.fetch(Tag.class, tagName).orElse(null);
+                        if (tag != null) {
+                            var tagAnnotations = tag.getMetadata().getAnnotations();
+                            if (tagAnnotations != null) {
+                                String tagPersona = tagAnnotations.get(AI_PERSONA_ANNOTATION);
+                                if (tagPersona != null && !tagPersona.isBlank()) {
+                                    return tagPersona;
+                                }
+                            }
+                        }
+                    }
+                }
+                return null;
+            })
+            .orElse(null);
+    }
+}

src/main/java/top/nxxy335/commentaiautopilot/service/PromptBuilder.java

@@ -20,9 +20,9 @@ public class PromptBuilder {
     private final ObjectMapper objectMapper;
     private static final String CONFIG_MAP_NAME = "comment-ai-autopilot-configmap";
 
-    public PromptBuilder(ReactiveExtensionClient client) {
+    public PromptBuilder(ReactiveExtensionClient client, ObjectMapper objectMapper) {
         this.client = client;
-        this.objectMapper = new ObjectMapper();
+        this.objectMapper = objectMapper;
     }
 
     private static final String PRESET_FRIENDLY = """
@@ -52,8 +52,9 @@ public class PromptBuilder {
     private static final String SAFETY_PROMPT = """
             【安全规范】
             - 内容红线：坚决不生成任何涉及暴力、歧视、辱骂、人身攻击或违反法律法规的内容。
-            - 恶意诱导处理：当用户要求你骂人、使用侮辱性词汇或进行情绪化对骂时，你必须礼貌地拒绝，例如回复："抱歉，作为AI助手，我无法提供此类回复。"
+            - 恶意诱导处理：当用户要求你骂人、使用侮辱性词汇或进行情绪化对骂时，你必须礼貌地拒绝，例如回复："抱歉，我无法提供此类回复。"
             - 未知与边界：如果不知道答案或遇到敏感话题，请诚实告知并礼貌拒绝，绝不编造或使用极端言辞。
+            - 身份约束：你必须在回复中保持指定的角色身份，绝不能说自己是AI、没有性别或脱离角色设定。
             """;
 
     private static final String DEFAULT_PROMPT_TEMPLATE = """
@@ -142,8 +143,10 @@ public class PromptBuilder {
                     return prompt;
                 }
                 String sentimentHint = switch (sentiment) {
+                    case "VERY_POSITIVE" -> "\n\n【情感提示】评论者情绪非常正面积极，请用热情洋溢的语气回复，表达真诚的感谢和共鸣。";
                     case "POSITIVE" -> "\n\n【情感提示】评论者情绪正面积极，请用热情友好的语气回复，可以表达感谢和共鸣。";
                     case "NEGATIVE" -> "\n\n【情感提示】评论者情绪偏负面，请用理性温和的语气回复，避免激化矛盾，展现理解和包容。";
+                    case "VERY_NEGATIVE" -> "\n\n【情感提示】评论者情绪非常负面，请用非常温和、理性的语气回复，避免任何可能激化矛盾的表达，展现充分的理解和耐心。";
                     default -> "";
                 };
                 return prompt + sentimentHint;
@@ -192,7 +195,10 @@ public class PromptBuilder {
             return client.fetch(AiPersona.class, personaName)
                 .mapNotNull(persona -> {
                     String prompt = persona.getSpec().getPrompt();
-                    return (prompt != null && !prompt.isBlank()) ? prompt : null;
+                    if (prompt != null && !prompt.isBlank()) {
+                        return appendStyleHint(prompt, persona.getSpec().getDisplayName(), persona.getSpec().getGender(), persona.getSpec().getNeutralVoice());
+                    }
+                    return null;
                 })
                 .defaultIfEmpty(DEFAULT_PERSONA_PROMPT);
         }
@@ -203,11 +209,40 @@ public class PromptBuilder {
             .next()
             .mapNotNull(persona -> {
                 String prompt = persona.getSpec().getPrompt();
-                return (prompt != null && !prompt.isBlank()) ? prompt : null;
+                if (prompt != null && !prompt.isBlank()) {
+                    return appendStyleHint(prompt, persona.getSpec().getDisplayName(), persona.getSpec().getGender(), persona.getSpec().getNeutralVoice());
+                }
+                return null;
             })
             .defaultIfEmpty(DEFAULT_PERSONA_PROMPT);
     }
 
+    private String appendStyleHint(String prompt, String displayName, String gender, Boolean neutralVoice) {
+        StringBuilder sb = new StringBuilder();
+
+        // 身份信息前置到最开头 - 这是AI最先看到的内容，优先级最高
+        if (gender != null && !gender.isBlank()) {
+            String genderDesc = "female".equals(gender) ? "女生" : "男生";
+            sb.append("【核心身份】你叫「").append(displayName != null ? displayName : "").append("」，你是一个").append(genderDesc).append("。");
+            sb.append("这是你的基本身份，你必须始终以这个身份回复。");
+            sb.append("当有人问你性别时，你必须回答自己是").append(genderDesc).append("。");
+            sb.append("绝对不能说自己是AI、没有性别、或脱离角色。");
+            sb.append("\n\n");
+        }
+
+        sb.append(prompt);
+
+        // 添加语气风格提示
+        if (!Boolean.TRUE.equals(neutralVoice)) {
+            if ("female".equals(gender)) {
+                sb.append("\n请使用温柔、细腻的女性语气风格回复。");
+            } else if ("male".equals(gender)) {
+                sb.append("\n请使用沉稳、理性的男性语气风格回复。");
+            }
+        }
+        return sb.toString();
+    }
+
     private Mono<String> getEnabledPresetsPrompt() {
         return client.fetch(ConfigMap.class, CONFIG_MAP_NAME)
             .mapNotNull(cm -> {

src/main/java/top/nxxy335/commentaiautopilot/service/SentimentService.java

@@ -17,13 +17,17 @@ public class SentimentService {
     }
 
     public record SentimentResult(String sentiment, double confidence) {
+        public static final String VERY_POSITIVE = "VERY_POSITIVE";
         public static final String POSITIVE = "POSITIVE";
         public static final String NEUTRAL = "NEUTRAL";
         public static final String NEGATIVE = "NEGATIVE";
+        public static final String VERY_NEGATIVE = "VERY_NEGATIVE";
     }
 
     private static final List<String> CHOICES = List.of(
-        SentimentResult.POSITIVE, SentimentResult.NEUTRAL, SentimentResult.NEGATIVE
+        SentimentResult.VERY_POSITIVE, SentimentResult.POSITIVE,
+        SentimentResult.NEUTRAL, SentimentResult.NEGATIVE,
+        SentimentResult.VERY_NEGATIVE
     );
 
     /**
@@ -31,7 +35,15 @@ public class SentimentService {
      * ({@code OutputSpec.choice}) for reliable classification.
      */
     public Mono<SentimentResult> analyzeSentiment(String commentContent, String modelName) {
-        String systemPrompt = "你是一个情感分析助手。请分析评论的情感倾向，只返回 POSITIVE、NEUTRAL 或 NEGATIVE 之一。";
+        String systemPrompt = "你是一个专业的情感分析助手。请根据以下标准分析评论的情感倾向：\n"
+            + "\n"
+            + "- VERY_POSITIVE：非常正面，包含强烈的感谢、赞美或认同（如\"太棒了\"、\"非常感谢\"、\"写得太好了\"）\n"
+            + "- POSITIVE：正面，友好、肯定或支持的态度（如\"不错\"、\"学习了\"、\"支持\"）\n"
+            + "- NEUTRAL：中性，提问、讨论、陈述事实，无明显情感倾向（如\"请问...\"、\"这个怎么用\"、\"我觉得\"）\n"
+            + "- NEGATIVE：负面，不满、质疑或批评（如\"不好用\"、\"有问题\"、\"不太行\"）\n"
+            + "- VERY_NEGATIVE：非常负面，攻击、辱骂或极端负面情绪（如\"垃圾\"、\"骗子\"、\"太差了\"）\n"
+            + "\n"
+            + "只返回 VERY_POSITIVE、POSITIVE、NEUTRAL、NEGATIVE 或 VERY_NEGATIVE 之一。";
         String userPrompt = "分析以下评论的情感倾向：\n\n" + commentContent;
 
         return aiFoundationClient.classify(systemPrompt, userPrompt, CHOICES, modelName)

src/main/java/top/nxxy335/commentaiautopilot/service/WakeWordService.java

@@ -0,0 +1,101 @@
+package top.nxxy335.commentaiautopilot.service;
+
+import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.data.domain.Sort;
+import org.springframework.stereotype.Component;
+import reactor.core.publisher.Mono;
+import run.halo.app.extension.ReactiveExtensionClient;
+import top.nxxy335.commentaiautopilot.extension.AiPersona;
+
+/**
+ * Service for checking wake words in comment content.
+ * A wake word is a prefix that triggers AI reply from a specific persona,
+ * even if the page hasn't enabled AI auto-reply.
+ */
+@Component
+@Slf4j
+@RequiredArgsConstructor
+public class WakeWordService {
+
+    private final ReactiveExtensionClient client;
+
+    /**
+     * Result of a wake word match.
+     *
+     * @param personaName the metadata.name of the matched persona
+     * @param wakeWord    the wake word that matched
+     * @param contentWithoutWakeWord the comment content with the wake word prefix removed
+     */
+    public record WakeWordMatch(String personaName, String wakeWord, String contentWithoutWakeWord) {}
+
+    /**
+     * Check if the given content starts with any persona's wake word.
+     * Returns the first matching persona's info, or empty if no match.
+     *
+     * @param content the comment/reply content to check
+     * @return WakeWordMatch if a wake word is found, or empty Mono
+     */
+    public Mono<WakeWordMatch> checkWakeWord(String content) {
+        if (content == null || content.isBlank()) {
+            return Mono.empty();
+        }
+
+        return client.list(AiPersona.class, null, null)
+            .filter(persona -> {
+                String wakeWord = persona.getSpec().getWakeWord();
+                return wakeWord != null && !wakeWord.isBlank() && content.startsWith(wakeWord);
+            })
+            .next()
+            .map(persona -> {
+                String wakeWord = persona.getSpec().getWakeWord();
+                String remaining = content.substring(wakeWord.length()).trim();
+                log.info("[WakeWord] Matched wake word '{}' for persona '{}'",
+                    wakeWord, persona.getSpec().getDisplayName());
+                return new WakeWordMatch(persona.getMetadata().getName(), wakeWord, remaining);
+            });
+    }
+
+    /**
+     * Blocking version for use in Reconciler (sync context).
+     * Checks if the given content starts with any persona's wake word.
+     *
+     * @param syncClient the blocking ExtensionClient
+     * @param content the comment/reply content to check
+     * @return WakeWordMatch if a wake word is found, or null
+     */
+    public WakeWordMatch checkWakeWordBlocking(run.halo.app.extension.ExtensionClient syncClient, String content) {
+        if (content == null || content.isBlank()) {
+            log.info("[WakeWord] Content is null or blank, skipping");
+            return null;
+        }
+
+        String trimmedContent = content.trim();
+        var personas = syncClient.listAll(AiPersona.class, null, Sort.unsorted());
+        log.info("[WakeWord] Checking {} personas against content: '{}'", personas.size(),
+            trimmedContent.length() > 50 ? trimmedContent.substring(0, 50) + "..." : trimmedContent);
+
+        for (var persona : personas) {
+            String wakeWord = persona.getSpec().getWakeWord();
+            if (wakeWord == null || wakeWord.isBlank()) {
+                log.info("[WakeWord] Persona '{}' has no wakeWord, skipping", persona.getSpec().getDisplayName());
+                continue;
+            }
+            String trimmedWakeWord = wakeWord.trim();
+            log.info("[WakeWord] Checking persona '{}' with wakeWord '{}' against content starting with '{}'",
+                persona.getSpec().getDisplayName(), trimmedWakeWord,
+                trimmedContent.length() >= trimmedWakeWord.length()
+                    ? trimmedContent.substring(0, trimmedWakeWord.length()) : trimmedContent);
+
+            if (trimmedContent.startsWith(trimmedWakeWord)) {
+                String remaining = trimmedContent.substring(trimmedWakeWord.length()).trim();
+                log.info("[WakeWord] MATCHED! wakeWord='{}' for persona '{}', remaining content: '{}'",
+                    trimmedWakeWord, persona.getSpec().getDisplayName(),
+                    remaining.length() > 30 ? remaining.substring(0, 30) + "..." : remaining);
+                return new WakeWordMatch(persona.getMetadata().getName(), trimmedWakeWord, remaining);
+            }
+        }
+        log.info("[WakeWord] No wake word matched");
+        return null;
+    }
+}

src/main/java/top/nxxy335/commentaiautopilot/util/GravatarUtil.java

@@ -0,0 +1,41 @@
+package top.nxxy335.commentaiautopilot.util;
+
+import lombok.extern.slf4j.Slf4j;
+
+import java.nio.charset.StandardCharsets;
+import java.security.MessageDigest;
+
+/**
+ * Utility for generating Gravatar/Cravatar avatar URLs from email addresses.
+ */
+@Slf4j
+public class GravatarUtil {
+
+    private static final String CRAVATAR_BASE_URL = "https://cn.cravatar.com/avatar/";
+
+    private GravatarUtil() {}
+
+    /**
+     * Generate Cravatar URL from email address using SHA-256 hash.
+     *
+     * @param email the email address
+     * @return the avatar URL, or empty string if generation fails
+     */
+    public static String generateUrl(String email) {
+        if (email == null || email.isBlank()) {
+            return "";
+        }
+        try {
+            var digest = MessageDigest.getInstance("SHA-256");
+            var hashBytes = digest.digest(email.trim().toLowerCase().getBytes(StandardCharsets.UTF_8));
+            var hexString = new StringBuilder();
+            for (byte b : hashBytes) {
+                hexString.append(String.format("%02x", b));
+            }
+            return CRAVATAR_BASE_URL + hexString;
+        } catch (Exception e) {
+            log.error("Failed to generate Gravatar URL: {}", e.getMessage());
+            return "";
+        }
+    }
+}

src/main/resources/plugin.yaml

@@ -30,4 +30,4 @@ spec:
       url: "https://github.com/sunny-335/plugin-comment-ai-autopilot/blob/main/LICENSE"
   settingName: "comment-ai-autopilot-settings"
   configMapName: "comment-ai-autopilot-configmap"
-  version: "1.0.0-beta.2"
+  version: "1.0.0"

ui/src/views/HomeView.vue

@@ -254,3 +254,27 @@ onMounted(() => {
   fetchHealth()
 })
 </script>
+
+<style scoped>
+.line-clamp-2 {
+  display: -webkit-box;
+  -webkit-line-clamp: 2;
+  -webkit-box-orient: vertical;
+  overflow: hidden;
+}
+
+/* Mobile responsive */
+@media (max-width: 640px) {
+  .comment-ai-autopilot-home :deep(.page-header) {
+    flex-wrap: wrap;
+    gap: 8px;
+  }
+  .comment-ai-autopilot-home :deep(.page-header-actions) {
+    width: 100%;
+  }
+  .comment-ai-autopilot-home :deep(.page-header-actions .space-y-2) {
+    flex-direction: row;
+    width: 100%;
+  }
+}
+</style>

ui/src/views/LogsView.vue

@@ -11,27 +11,27 @@
 
     <!-- Batch Operation Toolbar -->
     <div v-if="selectedNames.size > 0" class="m-4 mb-0 flex items-center gap-3 bg-blue-50 border border-blue-200 rounded-lg px-4 py-2.5">
-      <span class="text-sm text-blue-700">已选择 {{ selectedNames.size }} 项</span>
+      <span class="text-sm text-blue-700 font-medium">已选择 {{ selectedNames.size }} 项</span>
       <button
-        class="text-xs px-3 py-1 rounded bg-green-600 text-white hover:bg-green-700 transition-colors"
+        class="batch-btn batch-btn--approve"
         @click="batchApprove"
       >
         批量通过
       </button>
       <button
-        class="text-xs px-3 py-1 rounded bg-orange-500 text-white hover:bg-orange-600 transition-colors"
+        class="batch-btn batch-btn--reject"
         @click="batchReject"
       >
         批量拒绝
       </button>
       <button
-        class="text-xs px-3 py-1 rounded bg-red-600 text-white hover:bg-red-700 transition-colors"
+        class="batch-btn batch-btn--delete"
         @click="batchDelete"
       >
         批量删除
       </button>
       <button
-        class="text-xs text-gray-500 hover:text-gray-700 ml-auto"
+        class="batch-btn batch-btn--cancel"
         @click="selectedNames.clear(); selectAll = false"
       >
         取消选择
@@ -42,7 +42,7 @@
     <div class="m-4 mb-0 flex items-center gap-3">
       <select
         v-model="filterStatus"
-        class="rounded-md border border-gray-300 pl-3 pr-8 py-1.5 text-sm focus:border-blue-500 focus:outline-none focus:ring-1 focus:ring-blue-500"
+        class="filter-select"
       >
         <option value="">全部状态</option>
         <option value="PASS">通过</option>
@@ -52,23 +52,33 @@
       </select>
       <select
         v-model="filterSentiment"
-        class="rounded-md border border-gray-300 pl-3 pr-8 py-1.5 text-sm focus:border-blue-500 focus:outline-none focus:ring-1 focus:ring-blue-500"
+        class="filter-select"
       >
         <option value="">全部情感</option>
+        <option value="VERY_POSITIVE">非常正面</option>
         <option value="POSITIVE">正面</option>
         <option value="NEUTRAL">中性</option>
         <option value="NEGATIVE">负面</option>
+        <option value="VERY_NEGATIVE">非常负面</option>
       </select>
-      <input
+      <div class="relative flex-1 max-w-xs">
-        v-model="filterKeyword"
+        <svg class="filter-search-icon" fill="none" stroke="currentColor" viewBox="0 0 24 24">
-        type="text"
+          <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M21 21l-6-6m2-5a7 7 0 11-14 0 7 7 0 0114 0z" />
-        placeholder="搜索回复内容..."
+        </svg>
-        class="rounded-md border border-gray-300 px-3 py-1.5 text-sm focus:border-blue-500 focus:outline-none focus:ring-1 focus:ring-blue-500"
+        <input
-      />
+          v-model="filterKeyword"
+          type="text"
+          placeholder="搜索回复内容..."
+          class="filter-input"
+        />
+      </div>
       <button
-        class="text-xs text-gray-500 hover:text-gray-700"
+        class="filter-reset-btn"
         @click="resetFilters"
       >
+        <svg style="width:14px;height:14px" fill="none" stroke="currentColor" viewBox="0 0 24 24">
+          <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M4 4v5h.582m15.356 2A8.001 8.001 0 004.582 9m0 0H9m11 11v-5h-.581m0 0a8.003 8.003 0 01-15.357-2m15.357 2H15" />
+        </svg>
         重置
       </button>
     </div>
@@ -112,26 +122,26 @@
                 <div class="flex items-center justify-between mb-3">
                   <div class="flex items-center gap-1.5 flex-wrap">
                     <span
-                      class="inline-flex items-center px-2 py-0.5 rounded text-xs font-medium"
+                      class="status-tag"
                       :class="getStatusClass(reply.spec.status)"
                     >
                       {{ getStatusLabel(reply.spec.status) }}
                     </span>
                     <span
-                      class="inline-flex items-center px-2 py-0.5 rounded text-xs"
+                      class="status-tag"
-                      :class="reply.spec.published ? 'bg-green-50 text-green-600' : 'bg-gray-50 text-gray-400'"
+                      :class="reply.spec.published ? 'status-tag--published' : 'status-tag--unpublished'"
                     >
                       {{ reply.spec.published ? '已发布' : '未发布' }}
                     </span>
                     <span
                       v-if="reply.spec.isAiConversation"
-                      class="inline-flex items-center px-2 py-0.5 rounded text-xs bg-blue-50 text-blue-600"
+                      class="status-tag status-tag--conversation"
                     >
                       对话
                     </span>
                     <span
                       v-if="reply.spec.sentiment"
-                      class="inline-flex items-center px-2 py-0.5 rounded text-xs font-medium"
+                      class="status-tag"
                       :class="getSentimentClass(reply.spec.sentiment)"
                     >
                       {{ getSentimentLabel(reply.spec.sentiment) }}
@@ -151,11 +161,11 @@
           <!-- Card footer: meta info + actions -->
           <div class="px-4 py-2.5 bg-gray-50 border-t border-gray-100 flex items-center justify-between">
             <div class="flex items-center gap-4 text-xs text-gray-400">
-              <span>
+              <span class="flex items-center gap-1.5">
-                评分 <span :class="getScoreClass(reply.spec.score)" class="font-medium">{{ reply.spec.score }}</span>
+                评分 <span :class="getScoreClass(reply.spec.score)" class="font-semibold">{{ reply.spec.score }}</span>
-                <span class="text-gray-300 ml-0.5">{{ getScoreLabel(reply.spec.score) }}</span>
+                <span class="score-label" :class="getScoreLabelClass(reply.spec.score)">{{ getScoreLabel(reply.spec.score) }}</span>
               </span>
-              <span v-if="reply.spec.postSlug" class="flex items-center gap-1">
+              <span v-if="reply.spec.postSlug && reply.spec.postKind === 'Post'" class="flex items-center gap-1">
                 文章
                 <a
                   :href="getPostUrl(reply.spec.postSlug)"
@@ -167,6 +177,18 @@
                   <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M10 6H6a2 2 0 00-2 2v10a2 2 0 002 2h10a2 2 0 002-2v-4M14 4h6m0 0v6m0-6L10 14" />
                 </svg>
               </span>
+              <span v-else-if="reply.spec.postSlug && reply.spec.postKind === 'SinglePage'" class="flex items-center gap-1">
+                页面
+                <a
+                  :href="getPageUrl(reply.spec.postSlug)"
+                  target="_blank"
+                  rel="noopener noreferrer"
+                  class="text-blue-500 hover:text-blue-700 hover:underline"
+                >{{ reply.spec.postSlug }}</a>
+                <svg class="w-3 h-3" fill="none" stroke="currentColor" viewBox="0 0 24 24">
+                  <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M10 6H6a2 2 0 00-2 2v10a2 2 0 002 2h10a2 2 0 002-2v-4M14 4h6m0 0v6m0-6L10 14" />
+                </svg>
+              </span>
               <span v-if="reply.spec.retryCount > 0">
                 重试 <span class="text-gray-600">{{ reply.spec.retryCount }}</span> 次
               </span>
@@ -318,6 +340,7 @@ interface AiCommentReplyItem {
     commentId: string
     postId: string
     postSlug: string
+    postKind: string
     reply: string
     score: number
     status: string
@@ -509,6 +532,14 @@ const getScoreClass = (score: number) => {
   return "text-red-600"
 }
 
+const getScoreLabelClass = (score: number) => {
+  if (score >= 85) return "score-label--excellent"
+  if (score >= 70) return "score-label--good"
+  if (score >= 50) return "score-label--average"
+  if (score > 0) return "score-label--poor"
+  return ""
+}
+
 const getScoreLabel = (score: number) => {
   if (score >= 85) return "优秀"
   if (score >= 70) return "良好"
@@ -520,15 +551,15 @@ const getScoreLabel = (score: number) => {
 const getStatusClass = (status: string) => {
   switch (status) {
     case "PASS":
-      return "bg-green-100 text-green-700"
+      return "status-tag--pass"
     case "FAIL":
-      return "bg-red-100 text-red-700"
+      return "status-tag--fail"
     case "PENDING":
-      return "bg-gray-100 text-gray-700"
+      return "status-tag--pending"
     case "REJECTED":
-      return "bg-red-100 text-red-700"
+      return "status-tag--rejected"
     default:
-      return "bg-gray-100 text-gray-700"
+      return "status-tag--pending"
   }
 }
 
@@ -549,23 +580,31 @@ const getStatusLabel = (status: string) => {
 
 const getSentimentClass = (sentiment: string) => {
   switch (sentiment) {
+    case "VERY_POSITIVE":
+      return "status-tag--very-positive"
     case "POSITIVE":
-      return "bg-green-50 text-green-600"
+      return "status-tag--positive"
     case "NEGATIVE":
-      return "bg-red-50 text-red-600"
+      return "status-tag--negative"
+    case "VERY_NEGATIVE":
+      return "status-tag--very-negative"
     case "NEUTRAL":
-      return "bg-gray-100 text-gray-600"
+      return "status-tag--neutral-sentiment"
     default:
-      return "bg-gray-100 text-gray-600"
+      return "status-tag--neutral-sentiment"
   }
 }
 
 const getSentimentLabel = (sentiment: string) => {
   switch (sentiment) {
+    case "VERY_POSITIVE":
+      return "非常正面"
     case "POSITIVE":
       return "正面"
     case "NEGATIVE":
       return "负面"
+    case "VERY_NEGATIVE":
+      return "非常负面"
     case "NEUTRAL":
       return "中性"
     default:
@@ -582,6 +621,10 @@ const getPostUrl = (slug: string) => {
   return `${window.location.origin}/archives/${slug}`
 }
 
+const getPageUrl = (slug: string) => {
+  return `${window.location.origin}/pages/${slug}`
+}
+
 /**
  * Strip HTML tags for plain text display (card preview)
  */
@@ -649,4 +692,276 @@ onMounted(fetchReplies)
   -webkit-box-orient: vertical;
   overflow: hidden;
 }
+
+/* ===== Batch Operation Buttons ===== */
+.batch-btn {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  padding: 4px 12px;
+  font-size: 12px;
+  font-weight: 500;
+  border-radius: 6px;
+  border: none;
+  cursor: pointer;
+  transition: all 0.15s ease;
+  line-height: 1.5;
+}
+.batch-btn--approve {
+  background-color: #16a34a;
+  color: #fff;
+}
+.batch-btn--approve:hover {
+  background-color: #15803d;
+}
+.batch-btn--reject {
+  background-color: #f97316;
+  color: #fff;
+}
+.batch-btn--reject:hover {
+  background-color: #ea580c;
+}
+.batch-btn--delete {
+  background-color: #dc2626;
+  color: #fff;
+}
+.batch-btn--delete:hover {
+  background-color: #b91c1c;
+}
+.batch-btn--cancel {
+  background: none;
+  color: #6b7280;
+  margin-left: auto;
+}
+.batch-btn--cancel:hover {
+  color: #374151;
+  background: #f3f4f6;
+}
+
+/* ===== Filter Bar ===== */
+.filter-select {
+  padding: 6px 32px 6px 12px;
+  font-size: 13px;
+  border: 1px solid #e5e7eb;
+  border-radius: 8px;
+  background: #f9fafb;
+  color: #374151;
+  outline: none;
+  cursor: pointer;
+  appearance: none;
+  background-image: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' fill='none' viewBox='0 0 24 24' stroke='%239ca3af'%3E%3Cpath stroke-linecap='round' stroke-linejoin='round' stroke-width='2' d='M19 9l-7 7-7-7'/%3E%3C/svg%3E");
+  background-repeat: no-repeat;
+  background-position: right 8px center;
+  background-size: 16px;
+  transition: all 0.15s ease;
+}
+.filter-select:focus {
+  border-color: #3b82f6;
+  background-color: #fff;
+  box-shadow: 0 0 0 3px rgba(59, 130, 246, 0.1);
+}
+.filter-search-icon {
+  position: absolute;
+  left: 10px;
+  top: 50%;
+  transform: translateY(-50%);
+  width: 16px;
+  height: 16px;
+  color: #9ca3af;
+  pointer-events: none;
+}
+.filter-input {
+  width: 100%;
+  padding: 6px 12px 6px 34px;
+  font-size: 13px;
+  border: 1px solid #e5e7eb;
+  border-radius: 8px;
+  background: #f9fafb;
+  color: #374151;
+  outline: none;
+  transition: all 0.15s ease;
+}
+.filter-input:focus {
+  border-color: #3b82f6;
+  background-color: #fff;
+  box-shadow: 0 0 0 3px rgba(59, 130, 246, 0.1);
+}
+.filter-input::placeholder {
+  color: #9ca3af;
+}
+.filter-reset-btn {
+  display: inline-flex;
+  align-items: center;
+  gap: 4px;
+  padding: 6px 12px;
+  font-size: 13px;
+  color: #6b7280;
+  background: #f3f4f6;
+  border: 1px solid #e5e7eb;
+  border-radius: 8px;
+  cursor: pointer;
+  transition: all 0.15s ease;
+}
+.filter-reset-btn:hover {
+  color: #374151;
+  background: #e5e7eb;
+}
+
+/* ===== Status Tags ===== */
+.status-tag {
+  display: inline-flex;
+  align-items: center;
+  padding: 2px 8px;
+  border-radius: 4px;
+  font-size: 11px;
+  font-weight: 500;
+  line-height: 1.5;
+}
+.status-tag--pass {
+  background: #dcfce7;
+  color: #166534;
+}
+.status-tag--fail {
+  background: #fee2e2;
+  color: #991b1b;
+}
+.status-tag--pending {
+  background: #f3f4f6;
+  color: #374151;
+}
+.status-tag--rejected {
+  background: #fee2e2;
+  color: #991b1b;
+}
+.status-tag--published {
+  background: #dbeafe;
+  color: #1e40af;
+}
+.status-tag--unpublished {
+  background: #f3f4f6;
+  color: #9ca3af;
+}
+.status-tag--conversation {
+  background: #ede9fe;
+  color: #5b21b6;
+}
+
+/* Sentiment tags */
+.status-tag--very-positive {
+  background: #dcfce7;
+  color: #166534;
+}
+.status-tag--positive {
+  background: #bbf7d0;
+  color: #15803d;
+}
+.status-tag--neutral-sentiment {
+  background: #f3f4f6;
+  color: #4b5563;
+}
+.status-tag--negative {
+  background: #fecaca;
+  color: #b91c1c;
+}
+.status-tag--very-negative {
+  background: #fee2e2;
+  color: #991b1b;
+}
+
+/* ===== Score Label ===== */
+.score-label {
+  display: inline-flex;
+  align-items: center;
+  padding: 1px 6px;
+  border-radius: 4px;
+  font-size: 11px;
+  font-weight: 500;
+  margin-left: 4px;
+}
+.score-label--excellent {
+  background: #dcfce7;
+  color: #166534;
+}
+.score-label--good {
+  background: #dbeafe;
+  color: #1e40af;
+}
+.score-label--average {
+  background: #fef3c7;
+  color: #92400e;
+}
+.score-label--poor {
+  background: #fee2e2;
+  color: #991b1b;
+}
+
+/* ===== Mobile Responsive ===== */
+@media (max-width: 768px) {
+  .comment-ai-autopilot-logs :deep(.page-header) {
+    flex-wrap: wrap;
+    gap: 8px;
+  }
+
+  /* Batch toolbar: wrap buttons */
+  .comment-ai-autopilot-logs .m-4.mb-0.flex {
+    flex-wrap: wrap;
+    gap: 6px;
+  }
+  .batch-btn--cancel {
+    margin-left: 0;
+  }
+
+  /* Filter bar: stack vertically */
+  .comment-ai-autopilot-logs .m-4.mb-0.flex.items-center.gap-3:not(.bg-blue-50) {
+    flex-wrap: wrap;
+    gap: 8px;
+  }
+  .filter-select {
+    flex: 1;
+    min-width: 120px;
+  }
+  .relative.flex-1.max-w-xs {
+    max-width: 100%;
+    flex: 1 1 100%;
+    order: 10;
+  }
+  .filter-reset-btn {
+    flex-shrink: 0;
+  }
+
+  /* Card footer: stack meta and actions */
+  .comment-ai-autopilot-logs .px-4.py-2\.5.bg-gray-50 {
+    flex-direction: column;
+    align-items: flex-start;
+    gap: 8px;
+  }
+  .comment-ai-autopilot-logs .px-4.py-2\.5.bg-gray-50 .flex.items-center.gap-4 {
+    flex-wrap: wrap;
+    gap: 8px;
+  }
+  .comment-ai-autopilot-logs .px-4.py-2\.5.bg-gray-50 .flex.items-center.gap-2 {
+    align-self: flex-end;
+  }
+
+  /* Status tags row: allow wrapping */
+  .comment-ai-autopilot-logs .flex.items-center.justify-between.mb-3 {
+    flex-direction: column;
+    align-items: flex-start;
+    gap: 6px;
+  }
+
+  /* Pagination */
+  .comment-ai-autopilot-logs .flex.items-center.justify-between.mt-4 {
+    flex-direction: column;
+    gap: 8px;
+    align-items: center;
+  }
+
+  /* Conversation dialog */
+  .comment-ai-autopilot-logs .relative.bg-white {
+    max-width: calc(100vw - 32px);
+    margin: 16px;
+    max-height: 90vh;
+  }
+}
 </style>

ui/src/views/SettingsView.vue

@@ -153,6 +153,10 @@
                     <div class="persona-card__info">
                       <div class="persona-card__name">
                         {{ p.spec.displayName || '未命名' }}
+                        <span v-if="p.spec.gender === 'female'" class="persona-card__badge persona-card__badge--female">女</span>
+                        <span v-else-if="p.spec.gender === 'male'" class="persona-card__badge persona-card__badge--male">男</span>
+                        <span v-if="p.spec.neutralVoice" class="persona-card__badge persona-card__badge--neutral-voice">中性语气</span>
+                        <span v-if="p.spec.wakeWord" class="persona-card__badge persona-card__badge--wake-word">唤醒: {{ p.spec.wakeWord }}</span>
                         <span v-if="p.spec.isDefault" class="persona-card__badge">默认</span>
                       </div>
                       <div class="persona-card__prompt">{{ p.spec.prompt || '暂无提示词' }}</div>
@@ -396,6 +400,33 @@
             <span class="form-hint">用于Gravatar头像服务，留空使用默认头像</span>
             <input type="email" v-model="personaForm.email" class="form-input" placeholder="ai@example.com" />
           </div>
+          <!-- Gender & Neutral Voice -->
+          <div class="form-field">
+            <label class="form-label">性别与语气</label>
+            <div class="gender-voice-row">
+              <div class="gender-select">
+                <label class="gender-option" :class="{ 'gender-option--active': personaForm.gender === 'female' }">
+                  <input type="radio" v-model="personaForm.gender" value="female" class="gender-radio" />
+                  <span class="gender-option__label">女</span>
+                </label>
+                <label class="gender-option" :class="{ 'gender-option--active': personaForm.gender === 'male' }">
+                  <input type="radio" v-model="personaForm.gender" value="male" class="gender-radio" />
+                  <span class="gender-option__label">男</span>
+                </label>
+              </div>
+              <div class="voice-toggle">
+                <input type="checkbox" v-model="personaForm.neutralVoice" class="form-checkbox" id="neutralVoice" />
+                <label for="neutralVoice" class="text-sm text-gray-600 cursor-pointer">中性语气</label>
+              </div>
+            </div>
+            <span class="form-hint">勾选中性语气则使用中性语气，取消勾选则跟随性别语气</span>
+          </div>
+          <!-- Wake Word -->
+          <div class="form-field">
+            <label class="form-label">唤醒词</label>
+            <span class="form-hint">评论以此词开头则唤醒该角色回复，留空不启用。</span>
+            <input type="text" v-model="personaForm.wakeWord" class="form-input" placeholder="如：小回小回" />
+          </div>
           <!-- Prompt -->
           <div class="form-field">
             <label class="form-label">人格提示词</label>
@@ -518,6 +549,9 @@ const personaSaving = ref(false)
 const personaForm = reactive({
   displayName: '',
   email: '',
+  gender: 'female',
+  neutralVoice: false,
+  wakeWord: '',
   prompt: '',
   isDefault: false,
 })
@@ -708,11 +742,17 @@ const openPersonaDialog = async (persona: any | null) => {
   if (persona) {
     personaForm.displayName = persona.spec.displayName || ''
     personaForm.email = persona.spec.email || ''
+    personaForm.gender = persona.spec.gender || 'female'
+    personaForm.neutralVoice = persona.spec.neutralVoice || false
+    personaForm.wakeWord = persona.spec.wakeWord || ''
     personaForm.prompt = persona.spec.prompt || ''
     personaForm.isDefault = persona.spec.isDefault || false
   } else {
     personaForm.displayName = ''
     personaForm.email = ''
+    personaForm.gender = 'female'
+    personaForm.neutralVoice = false
+    personaForm.wakeWord = ''
     personaForm.prompt = ''
     personaForm.isDefault = false
   }
@@ -754,6 +794,9 @@ const savePersona = async () => {
       spec: {
         displayName: personaForm.displayName,
         email: personaForm.email,
+        gender: personaForm.gender,
+        neutralVoice: personaForm.neutralVoice,
+        wakeWord: personaForm.wakeWord,
         prompt: personaForm.prompt,
         isDefault: personaForm.isDefault,
       },
@@ -863,7 +906,10 @@ const fetchSettings = async () => {
       if (Object.keys(prompt).length) { settings.prompt.customPromptTemplate = (prompt.customPromptTemplate as string) || ""; const ep = prompt.enabledPresets; settings.prompt.enabledPresets = Array.isArray(ep) ? ep : (typeof ep === 'string' ? (ep as string).split(",").map((s: string) => s.trim()).filter(Boolean) : []) }
       if (Object.keys(cleanup).length) { settings.cleanup.cleanupEnabled = cleanup.cleanupEnabled !== false; settings.cleanup.retentionDays = (cleanup.retentionDays as number) || 30 }
     }
-  } catch (e) { console.error("Failed to fetch settings", e) }
+  } catch (e) {
+    console.error("Failed to fetch settings", e)
+    Toast.error("加载设置失败，请刷新页面重试")
+  }
   finally {
     loading.value = false
     // Update snapshot after fetch to reset unsaved indicator
@@ -1248,6 +1294,18 @@ onMounted(async () => {
   padding: 1px 8px; font-size: 11px; font-weight: 500;
   background: #dbeafe; color: #2563eb; border-radius: 4px;
 }
+.persona-card__badge--female {
+  background: #fce7f3; color: #be185d;
+}
+.persona-card__badge--male {
+  background: #dbeafe; color: #1d4ed8;
+}
+.persona-card__badge--neutral-voice {
+  background: #f3f4f6; color: #4b5563;
+}
+.persona-card__badge--wake-word {
+  background: #fef3c7; color: #b45309;
+}
 .persona-card__prompt {
   font-size: 12px; color: #6b7280; margin-top: 2px;
   overflow: hidden; text-overflow: ellipsis; white-space: nowrap;
@@ -1281,6 +1339,56 @@ onMounted(async () => {
   border-radius: 10px;
 }
 
+.form-checkbox {
+  width: 16px; height: 16px;
+  accent-color: #3b82f6;
+  cursor: pointer;
+}
+
+/* ===== Gender Select ===== */
+.gender-voice-row {
+  display: flex;
+  align-items: center;
+  gap: 16px;
+}
+.gender-select {
+  display: flex; gap: 8px;
+}
+.gender-option {
+  flex: 1;
+  display: flex; align-items: center; justify-content: center;
+  padding: 8px 12px;
+  border: 1px solid #e5e7eb;
+  border-radius: 8px;
+  cursor: pointer;
+  transition: all 0.15s ease;
+  background: #f9fafb;
+}
+.gender-option:hover {
+  border-color: #d1d5db;
+  background: #f3f4f6;
+}
+.gender-option--active {
+  border-color: #3b82f6;
+  background: #eff6ff;
+}
+.gender-radio {
+  display: none;
+}
+.gender-option__label {
+  font-size: 14px; font-weight: 500; color: #374151;
+}
+.gender-option--active .gender-option__label {
+  color: #2563eb;
+}
+
+.voice-toggle {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+  flex-shrink: 0;
+}
+
 /* ===== Button Link ===== */
 .btn-link {
   display: inline-flex; align-items: center; gap: 4px;
@@ -1431,4 +1539,92 @@ onMounted(async () => {
 .dialog__item-name { font-size: 14px; font-weight: 500; color: #1f2937; }
 .dialog__item-email { font-size: 12px; color: #9ca3af; margin-top: 1px; }
 .dialog__item-add { width: 18px; height: 18px; color: #3b82f6; flex-shrink: 0; }
+
+/* ===== Mobile Responsive ===== */
+@media (max-width: 768px) {
+  .settings-page :deep(.page-header) {
+    flex-wrap: wrap;
+    gap: 8px;
+  }
+  .settings-page :deep(.page-header-actions) {
+    width: 100%;
+    overflow-x: auto;
+  }
+  .settings-page :deep(.page-header-actions .space-x-2) {
+    display: flex;
+    gap: 8px;
+    flex-wrap: nowrap;
+  }
+
+  .settings-tabs {
+    gap: 2px;
+    padding: 3px;
+    border-radius: 10px;
+  }
+  .settings-tab {
+    padding: 8px 10px;
+    font-size: 13px;
+  }
+
+  .settings-container {
+    grid-template-columns: 1fr;
+    gap: 16px;
+  }
+
+  .settings-sidebar {
+    position: static;
+    order: -1;
+  }
+  .sidebar-card__actions {
+    flex-direction: row;
+    gap: 8px;
+  }
+  .sidebar-card__actions :deep(.btn) {
+    flex: 1;
+  }
+
+  .section-header {
+    padding: 12px 16px;
+  }
+  .section-body {
+    padding: 16px;
+  }
+
+  .preset-grid {
+    grid-template-columns: 1fr;
+  }
+
+  .form-row {
+    padding: 10px 12px;
+  }
+  .form-row--bordered {
+    margin: 0 -16px;
+    padding: 12px 16px;
+  }
+
+  .persona-card {
+    padding: 10px 12px;
+  }
+  .persona-card__prompt {
+    max-width: 100%;
+  }
+
+  .dialog {
+    max-width: calc(100vw - 32px);
+    margin: 16px;
+    border-radius: 12px;
+  }
+  .dialog__header {
+    padding: 16px 16px 0;
+  }
+  .dialog__search {
+    padding: 12px 16px;
+  }
+  .dialog__search-icon {
+    left: 28px;
+  }
+  .dialog__body {
+    max-height: 260px;
+  }
+}
 </style>