fix: mark all unread notifications across all pages

chore: add AGENTS guides for root and submodules

.github/workflows/coffee-bot.yml

@@ -0,0 +1,30 @@
+name: Coffee Bot
+
+on:
+  schedule:
+    - cron: "0 23 * * 0-4"
+  workflow_dispatch:
+
+jobs:
+  run-coffee-bot:
+    environment: Bots
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "20"
+          cache: "npm"
+
+      - name: Install dependencies
+        run: npm install --no-save @openai/agents tsx typescript
+
+      - name: Run coffee bot
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          OPENISLE_TOKEN: ${{ secrets.OPENISLE_TOKEN }}
+          APIFY_API_TOKEN: ${{ secrets.APIFY_API_TOKEN }}
+        run: npx tsx bots/instance/coffee_bot.ts

.github/workflows/news-bot.yml

@@ -0,0 +1,30 @@
+name: Daily News Bot
+
+on:
+  schedule:
+    - cron: "0 22 * * 0-4"
+  workflow_dispatch:
+
+jobs:
+  run-daily-news-bot:
+    environment: Bots
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "20"
+          cache: "npm"
+
+      - name: Install dependencies
+        run: npm install --no-save @openai/agents tsx typescript
+
+      - name: Run daily news bot
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          OPENISLE_TOKEN: ${{ secrets.OPENISLE_TOKEN }}
+          APIFY_API_TOKEN: ${{ secrets.APIFY_API_TOKEN }}
+        run: npx tsx bots/instance/daily_news_bot.ts

.github/workflows/open_source_reply_bot.yml

@@ -0,0 +1,30 @@
+name: Open Source Reply Bot
+
+on:
+  schedule:
+    - cron: "*/30 * * * *"
+  workflow_dispatch:
+
+jobs:
+  run-open-source-reply-bot:
+    environment: Bots
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "20"
+          cache: "npm"
+
+      - name: Install dependencies
+        run: npm install --no-save @openai/agents tsx typescript
+
+      - name: Run open source reply bot
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          OPENISLE_TOKEN: ${{ secrets.OPENISLE_TOKEN_BOT_1 }}
+          APIFY_API_TOKEN: ${{ secrets.APIFY_API_TOKEN }}
+        run: npx tsx bots/instance/open_source_reply_bot.ts

.github/workflows/reply-bots.yml

@@ -0,0 +1,30 @@
+name: Reply Bots
+
+on:
+  schedule:
+    - cron: "*/30 * * * *"
+  workflow_dispatch:
+
+jobs:
+  run-reply-bot:
+    environment: Bots
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "20"
+          cache: "npm"
+
+      - name: Install dependencies
+        run: npm install --no-save @openai/agents tsx typescript
+
+      - name: Run reply bot
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          OPENISLE_TOKEN: ${{ secrets.OPENISLE_TOKEN }}
+          APIFY_API_TOKEN: ${{ secrets.APIFY_API_TOKEN }}
+        run: npx tsx bots/instance/reply_bot.ts

AGENTS.md

@@ -0,0 +1,68 @@
+# OpenIsle 协作指引（根目录）
+
+## 1) 适用范围与优先级
+
+- 本文件作用于整个仓库。
+- 若子目录存在 `AGENTS.md`，以“最近目录”的规则为准（就近覆盖）。
+- 没有子目录规则时，回退到本文件。
+
+## 2) 仓库地图（高频模块）
+
+- `backend/`：Spring Boot 主后端（JPA、Redis、RabbitMQ、OpenSearch、OpenAPI）。
+- `frontend_nuxt/`：Nuxt 3 前端（SSR + WebSocket + OAuth 回调页）。
+- `websocket_service/`：独立 WebSocket/STOMP 服务（消费 RabbitMQ 通知并推送）。
+- `mcp/`：Python MCP Server（封装搜索、发帖、回复、通知等工具）。
+- `docs/`：Fumadocs 文档站（含 OpenAPI 生成）。
+- `deploy/`：生产/预发部署脚本。
+- `bots/`：定时 Bot 脚本（由 GitHub Actions 调度）。
+
+## 3) 全局工作原则
+
+- 小步修改：仅改当前任务相关文件，不做无关重构。
+- 先对齐契约，再改实现：优先保证 API、消息结构、环境变量名称稳定。
+- 变更需可验证：提交前给出最小可执行验证命令与结果。
+- 安全优先：禁止提交密钥、令牌、生产凭证。
+
+## 4) 任务路由（先判定影响面）
+
+- 仅后端业务逻辑：进入 `backend/AGENTS.md`。
+- 仅前端页面交互：进入 `frontend_nuxt/AGENTS.md`。
+- 实时消息/在线通知：同时看 `backend/` + `websocket_service/` + `frontend_nuxt/`。
+- API/DTO/鉴权变更：至少看 `backend/` + `mcp/` + `docs/`。
+- 部署链路变更：进入 `deploy/AGENTS.md`，并附风险与回滚方案。
+
+## 5) 跨服务不变量（必须遵守）
+
+- 环境变量以根目录 `.env.example` 为统一基线；变量改名需同步消费端。
+- 鉴权链路一致：
+  - 后端 JWT：`backend/src/main/java/com/openisle/config/SecurityConfig.java`
+  - WebSocket JWT：`websocket_service/src/main/java/com/openisle/websocket/config/WebSocketAuthInterceptor.java`
+- RabbitMQ 分片一致（16 个十六进制分片 + 兼容遗留队列）：
+  - `backend/src/main/java/com/openisle/config/RabbitMQConfig.java`
+  - `backend/src/main/java/com/openisle/config/ShardingStrategy.java`
+  - `websocket_service/src/main/java/com/openisle/websocket/listener/NotificationListener.java`
+- API 契约变更要同步：
+  - MCP schema/tool：`mcp/src/openisle_mcp/schemas.py`、`mcp/src/openisle_mcp/server.py`
+  - 文档生成：`docs/scripts/generate-docs.ts`、`docs/lib/openapi.ts`
+
+## 6) 最小验证矩阵（按改动类型）
+
+- 后端改动：`cd backend && mvn test`
+- 前端改动：`cd frontend_nuxt && npm run build`
+- WebSocket 服务改动：`cd websocket_service && mvn test`（无测试可退化为 `mvn -DskipTests compile`）
+- MCP 改动：`cd mcp && python -m pip install -e .`
+- Docs/OpenAPI 改动：`cd docs && bun run generate && bun run build`
+- 部署脚本改动：`bash -n deploy/deploy.sh && bash -n deploy/deploy_staging.sh`
+
+## 7) 交付清单（每次任务输出建议）
+
+- 改了什么：按文件列出核心变更点。
+- 为什么改：说明触发原因/缺陷点/一致性要求。
+- 如何验证：给出实际执行命令与结果摘要。
+- 风险与后续：列出剩余风险、可选回归点、回滚建议。
+
+## 8) 禁止事项
+
+- 不提交 `.env`、密钥、生产 token。
+- 不在未明确授权下执行破坏性命令（如大范围删除、强制重置）。
+- 不在无关文件中进行格式化/重排以“顺手优化”。

CONTRIBUTING.md

@@ -57,6 +57,17 @@ cd OpenIsle
      --profile dev up -d --force-recreate
    ```
 
+   仅重启后端容器（不重建镜像、不影响前端）：
+   ```shell
+   docker compose \
+     -f docker/docker-compose.yaml \
+     --env-file .env \
+     --profile dev restart springboot websocket-service 
+   ```
+
+   数据初始化sql会创建几个帐户供大家测试使用
+   > username:admin/user1/user2 password:123456
+
 3. 查看服务状态：
    ```shell
    docker compose -f docker/docker-compose.yaml --env-file .env ps

README.md

@@ -26,7 +26,7 @@ OpenIsle 是一个使用 Spring Boot 和 Vue 3 构建的全栈开源社区平台
 - 集成 OpenAI 提供的 Markdown 格式化功能
 - 通过环境变量可调整密码强度、登录方式、保护码等多种配置
 - 支持图片上传，默认使用腾讯云 COS 扩展
-- 默认头像使用 DiceBear Avatars，可通过 `AVATAR_STYLE` 和 `AVATAR_SIZE` 环境变量自定义主题和大小
+- Bot 集成，可在平台内快速连接自定义机器人，并通过 Telegram 的 BotFather 创建和管理消息机器人，拓展社区互动渠道
 - 浏览器推送通知，离开网站也能及时收到提醒
 
 ## 🌟 项目优势
@@ -41,7 +41,7 @@ OpenIsle 是一个使用 Spring Boot 和 Vue 3 构建的全栈开源社区平台
 
 ## 🏘️ 社区
 
-欢迎彼此交流和使用 OpenIsle，项目以开源方式提供，想了解更多可访问：<https://github.com/nagisa77/OpenIsle>
+- 欢迎彼此交流和使用 OpenIsle，项目以开源方式提供；如果遇到问题请到 GitHub 的 Issues 页面反馈，想发起话题讨论也可以前往源站 <https://www.open-isle.com>，这里提供更完整的社区板块与互动体验。
 
 ## 📋 授权
 

backend/AGENTS.md

@@ -0,0 +1,59 @@
+# Backend 协作指引
+
+## 1) 适用范围
+
+- 作用于 `backend/` 目录及其子目录。
+- 若与根 `AGENTS.md` 冲突，以本文件为准（仅后端范围）。
+
+## 2) 代码结构心智模型
+
+- `controller/`：接口层（入参校验、权限边界、响应格式）。
+- `service/`：业务编排与领域规则（核心逻辑放这里）。
+- `repository/`：JPA 数据访问（基于实体与查询方法）。
+- `model/`：实体模型与枚举。
+- `dto/` + `mapper/`：对外契约和映射转换。
+- `config/`：安全、缓存、MQ、OpenAPI、初始化器等基础设施配置。
+- `search/`：OpenSearch 索引与事件驱动同步。
+
+## 3) 后端修改规则
+
+- 控制器保持“薄”，复杂逻辑下沉到 `service/`。
+- DTO 变更优先考虑兼容性，避免无版本的破坏性字段删除/改名。
+- 新增接口时：
+  - 补齐必要的鉴权规则（`SecurityConfig`）。
+  - 补齐 OpenAPI 注解（`@Operation`、`@ApiResponse` 等）。
+- 涉及缓存时，确认 `CachingConfig` 中缓存名、TTL 与失效策略一致。
+
+## 4) 重点一致性检查
+
+- 鉴权与公开接口：
+  - `src/main/java/com/openisle/config/SecurityConfig.java`
+- 搜索索引同步（实体字段/文案变更时）：
+  - `src/main/java/com/openisle/search/SearchDocumentFactory.java`
+  - `src/main/java/com/openisle/search/SearchIndexEventPublisher.java`
+- 消息通知链路（评论/通知相关）：
+  - `src/main/java/com/openisle/config/RabbitMQConfig.java`
+  - `src/main/java/com/openisle/config/ShardingStrategy.java`
+  - `src/main/java/com/openisle/service/NotificationProducer.java`
+- 环境变量消费面：
+  - `src/main/resources/application.properties`
+  - 根目录 `.env.example`
+
+## 5) 数据与事务注意事项
+
+- 涉及多表写入时，明确事务边界，避免半成功状态。
+- 避免在 Controller 直接操作 Repository。
+- JPA 懒加载对象对外返回前应通过 DTO 映射，避免序列化副作用。
+
+## 6) 测试与验证
+
+- 首选全量：`mvn test`
+- 变更集中时可先跑目标测试（示例）：
+  - `mvn -Dtest=PostControllerTest test`
+  - `mvn -Dtest=SearchServiceTest test`
+- 涉及搜索/MQ 配置时，至少完成一次启动级验证或关键集成测试。
+
+## 7) 输出要求
+
+- 明确列出“接口/字段/权限/事件”是否发生变化。
+- 若影响 `mcp/` 或 `docs/`，在结果中显式提示需同步改动。

backend/src/main/java/com/openisle/controller/AdminUserController.java

@@ -6,10 +6,12 @@ import com.openisle.model.User;
 import com.openisle.repository.NotificationRepository;
 import com.openisle.repository.UserRepository;
 import com.openisle.service.EmailSender;
+import com.openisle.exception.EmailSendException;
 import io.swagger.v3.oas.annotations.Operation;
 import io.swagger.v3.oas.annotations.responses.ApiResponse;
 import io.swagger.v3.oas.annotations.security.SecurityRequirement;
 import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
 import org.springframework.beans.factory.annotation.Value;
 import org.springframework.http.ResponseEntity;
 import org.springframework.web.bind.annotation.*;
@@ -17,6 +19,7 @@ import org.springframework.web.bind.annotation.*;
 @RestController
 @RequestMapping("/api/admin/users")
 @RequiredArgsConstructor
+@Slf4j
 public class AdminUserController {
 
   private final UserRepository userRepository;
@@ -35,11 +38,15 @@ public class AdminUserController {
     user.setApproved(true);
     userRepository.save(user);
     markRegisterRequestNotificationsRead(user);
-    emailSender.sendEmail(
-      user.getEmail(),
-      "您的注册已审核通过",
-      "🎉您的注册已经审核通过, 点击以访问网站: " + websiteUrl
-    );
+    try {
+      emailSender.sendEmail(
+        user.getEmail(),
+        "您的注册已审核通过",
+        "🎉您的注册已经审核通过, 点击以访问网站: " + websiteUrl
+      );
+    } catch (EmailSendException e) {
+      log.warn("Failed to send approve email to {}: {}", user.getEmail(), e.getMessage());
+    }
     return ResponseEntity.ok().build();
   }
 
@@ -52,11 +59,15 @@ public class AdminUserController {
     user.setApproved(false);
     userRepository.save(user);
     markRegisterRequestNotificationsRead(user);
-    emailSender.sendEmail(
-      user.getEmail(),
-      "您的注册已被管理员拒绝",
-      "您的注册被管理员拒绝, 点击链接可以重新填写理由申请: " + websiteUrl
-    );
+    try {
+      emailSender.sendEmail(
+        user.getEmail(),
+        "您的注册已被管理员拒绝",
+        "您的注册被管理员拒绝, 点击链接可以重新填写理由申请: " + websiteUrl
+      );
+    } catch (EmailSendException e) {
+      log.warn("Failed to send reject email to {}: {}", user.getEmail(), e.getMessage());
+    }
     return ResponseEntity.ok().build();
   }
 

backend/src/main/java/com/openisle/controller/AuthController.java

@@ -2,6 +2,7 @@ package com.openisle.controller;
 
 import com.openisle.config.CachingConfig;
 import com.openisle.dto.*;
+import com.openisle.exception.EmailSendException;
 import com.openisle.exception.FieldException;
 import com.openisle.model.RegisterMode;
 import com.openisle.model.User;
@@ -19,6 +20,7 @@ import java.util.concurrent.TimeUnit;
 import lombok.RequiredArgsConstructor;
 import org.springframework.beans.factory.annotation.Value;
 import org.springframework.data.redis.core.RedisTemplate;
+import org.springframework.http.HttpStatus;
 import org.springframework.http.ResponseEntity;
 import org.springframework.web.bind.annotation.*;
 
@@ -83,6 +85,17 @@ public class AuthController {
             "INVITE_APPROVED"
           )
         );
+      } catch (EmailSendException e) {
+        return ResponseEntity
+          .status(HttpStatus.INTERNAL_SERVER_ERROR)
+          .body(
+            Map.of(
+              "error",
+              "邮件发送失败: " + e.getMessage(),
+              "reason_code",
+              "EMAIL_SEND_FAILED"
+            )
+          );
       } catch (FieldException e) {
         return ResponseEntity.badRequest().body(
           Map.of("field", e.getField(), "error", e.getMessage())
@@ -97,7 +110,20 @@ public class AuthController {
       registerModeService.getRegisterMode()
     );
     // 发送确认邮件
-    userService.sendVerifyMail(user, VerifyType.REGISTER);
+    try {
+      userService.sendVerifyMail(user, VerifyType.REGISTER);
+    } catch (EmailSendException e) {
+      return ResponseEntity
+        .status(HttpStatus.INTERNAL_SERVER_ERROR)
+        .body(
+          Map.of(
+            "error",
+            "邮件发送失败: " + e.getMessage(),
+            "reason_code",
+            "EMAIL_SEND_FAILED"
+          )
+        );
+    }
     if (!user.isApproved()) {
       notificationService.createRegisterRequestNotifications(user, user.getRegisterReason());
     }
@@ -169,14 +195,28 @@ public class AuthController {
     }
     User user = userOpt.get();
     if (!user.isVerified()) {
-      user = userService.register(
-        user.getUsername(),
-        user.getEmail(),
-        user.getPassword(),
-        user.getRegisterReason(),
-        registerModeService.getRegisterMode()
-      );
-      userService.sendVerifyMail(user, VerifyType.REGISTER);
+      user =
+        userService.register(
+          user.getUsername(),
+          user.getEmail(),
+          user.getPassword(),
+          user.getRegisterReason(),
+          registerModeService.getRegisterMode()
+        );
+      try {
+        userService.sendVerifyMail(user, VerifyType.REGISTER);
+      } catch (EmailSendException e) {
+        return ResponseEntity
+          .status(HttpStatus.INTERNAL_SERVER_ERROR)
+          .body(
+            Map.of(
+              "error",
+              "Failed to send verification email: " + e.getMessage(),
+              "reason_code",
+              "EMAIL_SEND_FAILED"
+            )
+          );
+      }
       return ResponseEntity.badRequest().body(
         Map.of(
           "error",
@@ -663,7 +703,20 @@ public class AuthController {
     if (userOpt.isEmpty()) {
       return ResponseEntity.badRequest().body(Map.of("error", "User not found"));
     }
-    userService.sendVerifyMail(userOpt.get(), VerifyType.RESET_PASSWORD);
+    try {
+      userService.sendVerifyMail(userOpt.get(), VerifyType.RESET_PASSWORD);
+    } catch (EmailSendException e) {
+      return ResponseEntity
+        .status(HttpStatus.INTERNAL_SERVER_ERROR)
+        .body(
+          Map.of(
+            "error",
+            "邮件发送失败: " + e.getMessage(),
+            "reason_code",
+            "EMAIL_SEND_FAILED"
+          )
+        );
+    }
     return ResponseEntity.ok(Map.of("message", "Verification code sent"));
   }
 

backend/src/main/java/com/openisle/controller/CommentController.java

@@ -112,7 +112,9 @@ public class CommentController {
   )
   public List<TimelineItemDto<?>> listComments(
     @PathVariable Long postId,
-    @RequestParam(value = "sort", required = false, defaultValue = "OLDEST") CommentSort sort
+    @RequestParam(value = "sort", required = false, defaultValue = "OLDEST") CommentSort sort,
+    @RequestParam(value = "page", required = false, defaultValue = "0") int page,
+    @RequestParam(value = "pageSize", required = false, defaultValue = "20") int pageSize
   ) {
     log.debug("listComments called for post {} with sort {}", postId, sort);
     List<CommentDto> commentDtoList = commentService
@@ -183,8 +185,23 @@ public class CommentController {
       createdAtComparator = createdAtComparator.reversed();
     }
     itemDtoList.sort(comparator.thenComparing(createdAtComparator));
-    log.debug("listComments returning {} comments", itemDtoList.size());
-    return itemDtoList;
+
+    int safePage = Math.max(0, page);
+    int safePageSize = Math.max(1, pageSize);
+    int fromIndex = safePage * safePageSize;
+    int toIndex = Math.min(fromIndex + safePageSize, itemDtoList.size());
+    List<TimelineItemDto<?>> pagedItems =
+      fromIndex >= itemDtoList.size() ? List.of() : itemDtoList.subList(fromIndex, toIndex);
+
+    log.debug(
+      "listComments returning {} items for post {} page {} size {} (total {})",
+      pagedItems.size(),
+      postId,
+      safePage,
+      safePageSize,
+      itemDtoList.size()
+    );
+    return pagedItems;
   }
 
   @GetMapping("/comments/{commentId}/context")

backend/src/main/java/com/openisle/controller/PostController.java

@@ -217,11 +217,7 @@ public class PostController {
     //            userVisitService.recordVisit(auth.getName());
     //        }
 
-    return postService
-      .defaultListPosts(ids, tids, page, pageSize)
-      .stream()
-      .map(postMapper::toSummaryDto)
-      .collect(Collectors.toList());
+    return postMapper.toListDtos(postService.defaultListPosts(ids, tids, page, pageSize));
   }
 
   @GetMapping("/recent")
@@ -269,11 +265,7 @@ public class PostController {
     //            userVisitService.recordVisit(auth.getName());
     //        }
 
-    return postService
-      .listPostsByViews(ids, tids, page, pageSize)
-      .stream()
-      .map(postMapper::toSummaryDto)
-      .collect(Collectors.toList());
+    return postMapper.toListDtos(postService.listPostsByViews(ids, tids, page, pageSize));
   }
 
   @GetMapping("/latest-reply")
@@ -305,8 +297,7 @@ public class PostController {
     //            userVisitService.recordVisit(auth.getName());
     //        }
 
-    List<Post> posts = postService.listPostsByLatestReply(ids, tids, page, pageSize);
-    return posts.stream().map(postMapper::toSummaryDto).collect(Collectors.toList());
+    return postMapper.toListDtos(postService.listPostsByLatestReply(ids, tids, page, pageSize));
   }
 
   @GetMapping("/featured")
@@ -333,10 +324,6 @@ public class PostController {
     //        if (auth != null) {
     //            userVisitService.recordVisit(auth.getName());
     //        }
-    return postService
-      .listFeaturedPosts(ids, tids, page, pageSize)
-      .stream()
-      .map(postMapper::toSummaryDto)
-      .collect(Collectors.toList());
+    return postMapper.toListDtos(postService.listFeaturedPosts(ids, tids, page, pageSize));
   }
 }

backend/src/main/java/com/openisle/controller/UserController.java

@@ -34,6 +34,7 @@ public class UserController {
   private final TagService tagService;
   private final SubscriptionService subscriptionService;
   private final LevelService levelService;
+  private final PostReadService postReadService;
   private final JwtService jwtService;
   private final UserMapper userMapper;
   private final TagMapper tagMapper;
@@ -53,6 +54,9 @@ public class UserController {
   @Value("${app.user.tags-limit:50}")
   private int defaultTagsLimit;
 
+  @Value("${app.user.read-posts-limit:50}")
+  private int defaultReadPostsLimit;
+
   @GetMapping("/me")
   @SecurityRequirement(name = "JWT")
   @Operation(summary = "Current user", description = "Get current authenticated user information")
@@ -211,6 +215,33 @@ public class UserController {
       .collect(java.util.stream.Collectors.toList());
   }
 
+  @GetMapping("/{identifier}/read-posts")
+  @SecurityRequirement(name = "JWT")
+  @Operation(summary = "User read posts", description = "Get post read history (self only)")
+  @ApiResponse(
+    responseCode = "200",
+    description = "Post read history",
+    content = @Content(array = @ArraySchema(schema = @Schema(implementation = PostReadDto.class)))
+  )
+  public ResponseEntity<java.util.List<PostReadDto>> userReadPosts(
+    @PathVariable("identifier") String identifier,
+    @RequestParam(value = "limit", required = false) Integer limit,
+    Authentication auth
+  ) {
+    User user = userService.findByIdentifier(identifier).orElseThrow();
+    if (auth == null || !auth.getName().equals(user.getUsername())) {
+      return ResponseEntity.status(403).body(java.util.List.of());
+    }
+    int l = limit != null ? limit : defaultReadPostsLimit;
+    return ResponseEntity.ok(
+      postReadService
+        .getRecentReadsByUser(user.getUsername(), l)
+        .stream()
+        .map(userMapper::toPostReadDto)
+        .collect(java.util.stream.Collectors.toList())
+    );
+  }
+
   @GetMapping("/{identifier}/hot-posts")
   @Operation(summary = "User hot posts", description = "Get most reacted posts by user")
   @ApiResponse(

backend/src/main/java/com/openisle/dto/AuthorDto.java

@@ -13,4 +13,5 @@ public class AuthorDto {
   private String username;
   private String avatar;
   private MedalType displayMedal;
+  private boolean bot;
 }

backend/src/main/java/com/openisle/dto/PostReadDto.java

@@ -0,0 +1,12 @@
+package com.openisle.dto;
+
+import java.time.LocalDateTime;
+import lombok.Data;
+
+/** DTO for a user's post read record. */
+@Data
+public class PostReadDto {
+
+  private PostMetaDto post;
+  private LocalDateTime lastReadAt;
+}

backend/src/main/java/com/openisle/dto/UserDto.java

@@ -28,4 +28,5 @@ public class UserDto {
   private int point;
   private int currentLevel;
   private int nextLevelExp;
+  private boolean bot;
 }

backend/src/main/java/com/openisle/dto/UserSummaryDto.java

@@ -8,4 +8,5 @@ public class UserSummaryDto {
   private Long id;
   private String username;
   private String avatar;
+  private boolean bot;
 }

backend/src/main/java/com/openisle/exception/EmailSendException.java

@@ -0,0 +1,15 @@
+package com.openisle.exception;
+
+/**
+ * Thrown when email sending fails so callers can surface a clear error upstream.
+ */
+public class EmailSendException extends RuntimeException {
+
+  public EmailSendException(String message) {
+    super(message);
+  }
+
+  public EmailSendException(String message, Throwable cause) {
+    super(message, cause);
+  }
+}

backend/src/main/java/com/openisle/mapper/PostMapper.java

@@ -48,6 +48,38 @@ public class PostMapper {
     return dto;
   }
 
+  public List<PostSummaryDto> toListDtos(List<Post> posts) {
+    if (posts == null || posts.isEmpty()) {
+      return List.of();
+    }
+    Map<Long, List<User>> participantsMap = commentService.getParticipantsForPosts(posts, 5);
+    return posts
+      .stream()
+      .map(post -> {
+        PostSummaryDto dto = new PostSummaryDto();
+        applyListFields(post, dto);
+        List<User> participants = participantsMap.get(post.getId());
+        if (participants != null) {
+          dto.setParticipants(
+            participants.stream().map(userMapper::toAuthorDto).collect(Collectors.toList())
+          );
+        } else {
+          dto.setParticipants(List.of());
+        }
+        dto.setReactions(List.of());
+        return dto;
+      })
+      .collect(Collectors.toList());
+  }
+
+  public PostSummaryDto toListDto(Post post) {
+    PostSummaryDto dto = new PostSummaryDto();
+    applyListFields(post, dto);
+    dto.setParticipants(List.of());
+    dto.setReactions(List.of());
+    return dto;
+  }
+
   public PostDetailDto toDetailDto(Post post, String viewer) {
     PostDetailDto dto = new PostDetailDto();
     applyCommon(post, dto);
@@ -61,6 +93,25 @@ public class PostMapper {
     return dto;
   }
 
+  private void applyListFields(Post post, PostSummaryDto dto) {
+    dto.setId(post.getId());
+    dto.setTitle(post.getTitle());
+    dto.setContent(post.getContent());
+    dto.setCreatedAt(post.getCreatedAt());
+    dto.setAuthor(userMapper.toAuthorDto(post.getAuthor()));
+    dto.setCategory(categoryMapper.toDto(post.getCategory()));
+    dto.setTags(post.getTags().stream().map(tagMapper::toDto).collect(Collectors.toList()));
+    dto.setViews(post.getViews());
+    dto.setCommentCount(post.getCommentCount());
+    dto.setStatus(post.getStatus());
+    dto.setPinnedAt(post.getPinnedAt());
+    dto.setLastReplyAt(post.getLastReplyAt());
+    dto.setRssExcluded(post.getRssExcluded() == null || post.getRssExcluded());
+    dto.setClosed(post.isClosed());
+    dto.setVisibleScope(post.getVisibleScope());
+    dto.setType(post.getType());
+  }
+
   private void applyCommon(Post post, PostSummaryDto dto) {
     dto.setId(post.getId());
     dto.setTitle(post.getTitle());

backend/src/main/java/com/openisle/mapper/UserMapper.java

@@ -3,6 +3,7 @@ package com.openisle.mapper;
 import com.openisle.dto.*;
 import com.openisle.model.Comment;
 import com.openisle.model.Post;
+import com.openisle.model.PostRead;
 import com.openisle.model.User;
 import com.openisle.service.*;
 import java.util.stream.Collectors;
@@ -37,6 +38,7 @@ public class UserMapper {
     dto.setUsername(user.getUsername());
     dto.setAvatar(user.getAvatar());
     dto.setDisplayMedal(user.getDisplayMedal());
+    dto.setBot(user.isBot());
     return dto;
   }
 
@@ -63,6 +65,7 @@ public class UserMapper {
     dto.setPoint(user.getPoint());
     dto.setCurrentLevel(levelService.getLevel(user.getExperience()));
     dto.setNextLevelExp(levelService.nextLevelExp(user.getExperience()));
+    dto.setBot(user.isBot());
     if (viewer != null) {
       dto.setSubscribed(subscriptionService.isSubscribed(viewer.getName(), user.getUsername()));
     } else {
@@ -113,4 +116,11 @@ public class UserMapper {
     }
     return dto;
   }
+
+  public PostReadDto toPostReadDto(PostRead read) {
+    PostReadDto dto = new PostReadDto();
+    dto.setPost(toMetaDto(read.getPost()));
+    dto.setLastReadAt(read.getLastReadAt());
+    return dto;
+  }
 }

backend/src/main/java/com/openisle/model/User.java

@@ -62,6 +62,9 @@ public class User {
   @Column(nullable = false)
   private Role role = Role.USER;
 
+  @Column(name = "is_bot", nullable = false)
+  private boolean bot = false;
+
   @Enumerated(EnumType.STRING)
   private MedalType displayMedal;
 

backend/src/main/java/com/openisle/repository/CommentRepository.java

@@ -25,6 +25,13 @@ public interface CommentRepository extends JpaRepository<Comment, Long> {
     @org.springframework.data.repository.query.Param("post") Post post
   );
 
+  @org.springframework.data.jpa.repository.Query(
+    "SELECT DISTINCT c.post.id, c.author FROM Comment c WHERE c.post.id IN :postIds"
+  )
+  java.util.List<Object[]> findDistinctAuthorsByPostIds(
+    @org.springframework.data.repository.query.Param("postIds") java.util.List<Long> postIds
+  );
+
   @org.springframework.data.jpa.repository.Query(
     "SELECT MAX(c.createdAt) FROM Comment c WHERE c.post = :post"
   )

backend/src/main/java/com/openisle/repository/PostReadRepository.java

@@ -3,11 +3,14 @@ package com.openisle.repository;
 import com.openisle.model.Post;
 import com.openisle.model.PostRead;
 import com.openisle.model.User;
+import java.util.List;
 import java.util.Optional;
+import org.springframework.data.domain.Pageable;
 import org.springframework.data.jpa.repository.JpaRepository;
 
 public interface PostReadRepository extends JpaRepository<PostRead, Long> {
   Optional<PostRead> findByUserAndPost(User user, Post post);
+  List<PostRead> findByUserOrderByLastReadAtDesc(User user, Pageable pageable);
   long countByUser(User user);
   void deleteByPost(Post post);
 }

backend/src/main/java/com/openisle/repository/PostRepository.java

@@ -19,6 +19,8 @@ public interface PostRepository extends JpaRepository<Post, Long> {
   List<Post> findByStatusOrderByCreatedAtDesc(PostStatus status, Pageable pageable);
   List<Post> findByStatusOrderByViewsDesc(PostStatus status);
   List<Post> findByStatusOrderByViewsDesc(PostStatus status, Pageable pageable);
+  List<Post> findByStatusOrderByPinnedAtDescViewsDesc(PostStatus status, Pageable pageable);
+  List<Post> findByStatusOrderByPinnedAtDescLastReplyAtDesc(PostStatus status, Pageable pageable);
   List<Post> findByStatusAndCreatedAtGreaterThanEqualOrderByCreatedAtDesc(
     PostStatus status,
     LocalDateTime createdAt
@@ -43,6 +45,16 @@ public interface PostRepository extends JpaRepository<Post, Long> {
     PostStatus status,
     Pageable pageable
   );
+  List<Post> findByCategoryInAndStatusOrderByPinnedAtDescViewsDesc(
+    List<Category> categories,
+    PostStatus status,
+    Pageable pageable
+  );
+  List<Post> findByCategoryInAndStatusOrderByPinnedAtDescLastReplyAtDesc(
+    List<Category> categories,
+    PostStatus status,
+    Pageable pageable
+  );
   List<Post> findDistinctByTagsInAndStatus(List<Tag> tags, PostStatus status);
   List<Post> findDistinctByTagsInAndStatus(List<Tag> tags, PostStatus status, Pageable pageable);
   List<Post> findDistinctByTagsInAndStatusOrderByCreatedAtDesc(List<Tag> tags, PostStatus status);
@@ -132,6 +144,26 @@ public interface PostRepository extends JpaRepository<Post, Long> {
     Pageable pageable
   );
 
+  @Query(
+    "SELECT p FROM Post p JOIN p.tags t WHERE t IN :tags AND p.status = :status GROUP BY p.id HAVING COUNT(DISTINCT t.id) = :tagCount ORDER BY p.pinnedAt DESC, p.views DESC"
+  )
+  List<Post> findByAllTagsOrderByPinnedAtDescViewsDesc(
+    @Param("tags") List<Tag> tags,
+    @Param("status") PostStatus status,
+    @Param("tagCount") long tagCount,
+    Pageable pageable
+  );
+
+  @Query(
+    "SELECT p FROM Post p JOIN p.tags t WHERE t IN :tags AND p.status = :status GROUP BY p.id HAVING COUNT(DISTINCT t.id) = :tagCount ORDER BY p.pinnedAt DESC, p.lastReplyAt DESC"
+  )
+  List<Post> findByAllTagsOrderByPinnedAtDescLastReplyAtDesc(
+    @Param("tags") List<Tag> tags,
+    @Param("status") PostStatus status,
+    @Param("tagCount") long tagCount,
+    Pageable pageable
+  );
+
   @Query(
     "SELECT p FROM Post p JOIN p.tags t WHERE p.category IN :categories AND t IN :tags AND p.status = :status GROUP BY p.id HAVING COUNT(DISTINCT t.id) = :tagCount"
   )
@@ -174,6 +206,28 @@ public interface PostRepository extends JpaRepository<Post, Long> {
     Pageable pageable
   );
 
+  @Query(
+    "SELECT p FROM Post p JOIN p.tags t WHERE p.category IN :categories AND t IN :tags AND p.status = :status GROUP BY p.id HAVING COUNT(DISTINCT t.id) = :tagCount ORDER BY p.pinnedAt DESC, p.views DESC"
+  )
+  List<Post> findByCategoriesAndAllTagsOrderByPinnedAtDescViewsDesc(
+    @Param("categories") List<Category> categories,
+    @Param("tags") List<Tag> tags,
+    @Param("status") PostStatus status,
+    @Param("tagCount") long tagCount,
+    Pageable pageable
+  );
+
+  @Query(
+    "SELECT p FROM Post p JOIN p.tags t WHERE p.category IN :categories AND t IN :tags AND p.status = :status GROUP BY p.id HAVING COUNT(DISTINCT t.id) = :tagCount ORDER BY p.pinnedAt DESC, p.lastReplyAt DESC"
+  )
+  List<Post> findByCategoriesAndAllTagsOrderByPinnedAtDescLastReplyAtDesc(
+    @Param("categories") List<Category> categories,
+    @Param("tags") List<Tag> tags,
+    @Param("status") PostStatus status,
+    @Param("tagCount") long tagCount,
+    Pageable pageable
+  );
+
   @Query(
     "SELECT p FROM Post p JOIN p.tags t WHERE p.category IN :categories AND t IN :tags AND p.status = :status GROUP BY p.id HAVING COUNT(DISTINCT t.id) = :tagCount ORDER BY p.createdAt DESC"
   )

backend/src/main/java/com/openisle/service/ChannelService.java

@@ -105,6 +105,7 @@ public class ChannelService {
     userDto.setId(message.getSender().getId());
     userDto.setUsername(message.getSender().getUsername());
     userDto.setAvatar(message.getSender().getAvatar());
+    userDto.setBot(message.getSender().isBot());
     dto.setSender(userDto);
 
     return dto;

backend/src/main/java/com/openisle/service/CommentService.java

@@ -21,8 +21,12 @@ import com.openisle.service.NotificationService;
 import com.openisle.service.PointService;
 import com.openisle.service.SubscriptionService;
 import java.time.LocalDateTime;
+import java.util.ArrayList;
+import java.util.HashMap;
 import java.util.HashSet;
+import java.util.LinkedHashSet;
 import java.util.List;
+import java.util.Map;
 import java.util.Set;
 import java.util.stream.Collectors;
 import lombok.RequiredArgsConstructor;
@@ -316,6 +320,37 @@ public class CommentService {
     return result;
   }
 
+  public Map<Long, List<User>> getParticipantsForPosts(List<Post> posts, int limit) {
+    if (posts == null || posts.isEmpty()) {
+      return Map.of();
+    }
+    Map<Long, LinkedHashSet<User>> map = new HashMap<>();
+    List<Long> postIds = new ArrayList<>(posts.size());
+    for (Post post : posts) {
+      postIds.add(post.getId());
+      LinkedHashSet<User> set = new LinkedHashSet<>();
+      set.add(post.getAuthor());
+      map.put(post.getId(), set);
+    }
+    for (Object[] row : commentRepository.findDistinctAuthorsByPostIds(postIds)) {
+      Long postId = (Long) row[0];
+      User author = (User) row[1];
+      LinkedHashSet<User> set = map.get(postId);
+      if (set != null) {
+        set.add(author);
+      }
+    }
+    Map<Long, List<User>> result = new HashMap<>(map.size());
+    for (Map.Entry<Long, LinkedHashSet<User>> entry : map.entrySet()) {
+      List<User> list = new ArrayList<>(entry.getValue());
+      if (list.size() > limit) {
+        list = list.subList(0, limit);
+      }
+      result.put(entry.getKey(), list);
+    }
+    return result;
+  }
+
   public java.util.List<Comment> getCommentsByIds(java.util.List<Long> ids) {
     log.debug("getCommentsByIds called for ids {}", ids);
     java.util.List<Comment> comments = commentRepository.findAllById(ids);

backend/src/main/java/com/openisle/service/MessageService.java

@@ -211,6 +211,7 @@ public class MessageService {
     userSummaryDto.setId(message.getSender().getId());
     userSummaryDto.setUsername(message.getSender().getUsername());
     userSummaryDto.setAvatar(message.getSender().getAvatar());
+    userSummaryDto.setBot(message.getSender().isBot());
     dto.setSender(userSummaryDto);
 
     if (message.getReplyTo() != null) {
@@ -222,6 +223,7 @@ public class MessageService {
       replySender.setId(reply.getSender().getId());
       replySender.setUsername(reply.getSender().getUsername());
       replySender.setAvatar(reply.getSender().getAvatar());
+      replySender.setBot(reply.getSender().isBot());
       replyDto.setSender(replySender);
       dto.setReplyTo(replyDto);
     }
@@ -316,6 +318,7 @@ public class MessageService {
           userDto.setId(p.getUser().getId());
           userDto.setUsername(p.getUser().getUsername());
           userDto.setAvatar(p.getUser().getAvatar());
+          userDto.setBot(p.getUser().isBot());
           return userDto;
         })
         .collect(Collectors.toList())
@@ -365,6 +368,7 @@ public class MessageService {
         userDto.setId(p.getUser().getId());
         userDto.setUsername(p.getUser().getUsername());
         userDto.setAvatar(p.getUser().getAvatar());
+        userDto.setBot(p.getUser().isBot());
         return userDto;
       })
       .collect(Collectors.toList());

backend/src/main/java/com/openisle/service/NotificationService.java

@@ -7,6 +7,7 @@ import com.openisle.repository.NotificationRepository;
 import com.openisle.repository.ReactionRepository;
 import com.openisle.repository.UserRepository;
 import com.openisle.service.EmailSender;
+import com.openisle.exception.EmailSendException;
 import java.util.ArrayList;
 import java.util.EnumSet;
 import java.util.HashSet;
@@ -17,6 +18,7 @@ import java.util.concurrent.Executor;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
 import org.springframework.beans.factory.annotation.Value;
 import org.springframework.stereotype.Service;
 import org.springframework.transaction.annotation.Transactional;
@@ -26,6 +28,7 @@ import org.springframework.transaction.support.TransactionSynchronizationManager
 /** Service for creating and retrieving notifications. */
 @Service
 @RequiredArgsConstructor
+@Slf4j
 public class NotificationService {
 
   private final NotificationRepository notificationRepository;
@@ -108,7 +111,11 @@ public class NotificationService {
         post.getId(),
         comment.getId()
       );
-      emailSender.sendEmail(user.getEmail(), "有人回复了你", url);
+      try {
+        emailSender.sendEmail(user.getEmail(), "有人回复了你", url);
+      } catch (EmailSendException e) {
+        log.warn("Failed to send notification email to {}: {}", user.getEmail(), e.getMessage());
+      }
       sendCustomPush(user, "有人回复了你", url);
     } else if (type == NotificationType.REACTION && comment != null) {
       //                long count = reactionRepository.countReceived(comment.getAuthor().getUsername());

backend/src/main/java/com/openisle/service/PostReadService.java

@@ -7,7 +7,10 @@ import com.openisle.repository.PostReadRepository;
 import com.openisle.repository.PostRepository;
 import com.openisle.repository.UserRepository;
 import java.time.LocalDateTime;
+import java.util.List;
 import lombok.RequiredArgsConstructor;
+import org.springframework.data.domain.PageRequest;
+import org.springframework.data.domain.Pageable;
 import org.springframework.stereotype.Service;
 
 @Service
@@ -43,6 +46,14 @@ public class PostReadService {
       );
   }
 
+  public List<PostRead> getRecentReadsByUser(String username, int limit) {
+    User user = userRepository
+      .findByUsername(username)
+      .orElseThrow(() -> new com.openisle.exception.NotFoundException("User not found"));
+    Pageable pageable = PageRequest.of(0, limit);
+    return postReadRepository.findByUserOrderByLastReadAtDesc(user, pageable);
+  }
+
   public long countReads(String username) {
     User user = userRepository
       .findByUsername(username)

backend/src/main/java/com/openisle/service/PostService.java

@@ -19,6 +19,7 @@ import com.openisle.repository.TagRepository;
 import com.openisle.repository.UserRepository;
 import com.openisle.search.SearchIndexEventPublisher;
 import com.openisle.service.EmailSender;
+import com.openisle.exception.EmailSendException;
 import java.time.Duration;
 import java.time.LocalDateTime;
 import java.time.ZoneId;
@@ -338,6 +339,7 @@ public class PostService {
     post.setCategory(category);
     post.setTags(new HashSet<>(tags));
     post.setStatus(publishMode == PublishMode.REVIEW ? PostStatus.PENDING : PostStatus.PUBLISHED);
+    post.setLastReplyAt(LocalDateTime.now());
 
     // 什么都没设置的情况下，默认为ALL
     if (Objects.isNull(postVisibleScopeType)) {
@@ -663,11 +665,15 @@ public class PostService {
             w.getEmail() != null &&
             !w.getDisabledEmailNotificationTypes().contains(NotificationType.LOTTERY_WIN)
           ) {
-            emailSender.sendEmail(
-              w.getEmail(),
-              "你中奖了",
-              "恭喜你在抽奖贴 \"" + lp.getTitle() + "\" 中获奖"
-            );
+            try {
+              emailSender.sendEmail(
+                w.getEmail(),
+                "你中奖了",
+                "恭喜你在抽奖贴 \"" + lp.getTitle() + "\" 中获奖"
+              );
+            } catch (EmailSendException e) {
+              log.warn("Failed to send lottery win email to {}: {}", w.getEmail(), e.getMessage());
+            }
           }
           notificationService.createNotification(
             w,
@@ -693,11 +699,19 @@ public class PostService {
               .getDisabledEmailNotificationTypes()
               .contains(NotificationType.LOTTERY_DRAW)
           ) {
-            emailSender.sendEmail(
-              lp.getAuthor().getEmail(),
-              "抽奖已开奖",
-              "您的抽奖贴 \"" + lp.getTitle() + "\" 已开奖"
-            );
+            try {
+              emailSender.sendEmail(
+                lp.getAuthor().getEmail(),
+                "抽奖已开奖",
+                "您的抽奖贴 \"" + lp.getTitle() + "\" 已开奖"
+              );
+            } catch (EmailSendException e) {
+              log.warn(
+                "Failed to send lottery draw email to {}: {}",
+                lp.getAuthor().getEmail(),
+                e.getMessage()
+              );
+            }
           }
           notificationService.createNotification(
             lp.getAuthor(),
@@ -796,9 +810,10 @@ public class PostService {
     boolean hasTags = tagIds != null && !tagIds.isEmpty();
 
     java.util.List<Post> posts;
+    Pageable pageable = buildPageable(page, pageSize);
 
     if (!hasCategories && !hasTags) {
-      posts = postRepository.findByStatusOrderByViewsDesc(PostStatus.PUBLISHED);
+      posts = postRepository.findByStatusOrderByPinnedAtDescViewsDesc(PostStatus.PUBLISHED, pageable);
     } else if (hasCategories) {
       java.util.List<Category> categories = categoryRepository.findAllById(categoryIds);
       if (categories.isEmpty()) {
@@ -809,16 +824,18 @@ public class PostService {
         if (tags.isEmpty()) {
           return java.util.List.of();
         }
-        posts = postRepository.findByCategoriesAndAllTagsOrderByViewsDesc(
+        posts = postRepository.findByCategoriesAndAllTagsOrderByPinnedAtDescViewsDesc(
           categories,
           tags,
           PostStatus.PUBLISHED,
-          tags.size()
+          tags.size(),
+          pageable
         );
       } else {
-        posts = postRepository.findByCategoryInAndStatusOrderByViewsDesc(
+        posts = postRepository.findByCategoryInAndStatusOrderByPinnedAtDescViewsDesc(
           categories,
-          PostStatus.PUBLISHED
+          PostStatus.PUBLISHED,
+          pageable
         );
       }
     } else {
@@ -826,10 +843,15 @@ public class PostService {
       if (tags.isEmpty()) {
         return java.util.List.of();
       }
-      posts = postRepository.findByAllTagsOrderByViewsDesc(tags, PostStatus.PUBLISHED, tags.size());
+      posts = postRepository.findByAllTagsOrderByPinnedAtDescViewsDesc(
+        tags,
+        PostStatus.PUBLISHED,
+        tags.size(),
+        pageable
+      );
     }
 
-    return paginate(sortByPinnedAndViews(posts), page, pageSize);
+    return posts;
   }
 
   public List<Post> listPostsByLatestReply(Integer page, Integer pageSize) {
@@ -846,9 +868,13 @@ public class PostService {
     boolean hasTags = tagIds != null && !tagIds.isEmpty();
 
     java.util.List<Post> posts;
+    Pageable pageable = buildPageable(page, pageSize);
 
     if (!hasCategories && !hasTags) {
-      posts = postRepository.findByStatusOrderByCreatedAtDesc(PostStatus.PUBLISHED);
+      posts = postRepository.findByStatusOrderByPinnedAtDescLastReplyAtDesc(
+        PostStatus.PUBLISHED,
+        pageable
+      );
     } else if (hasCategories) {
       java.util.List<Category> categories = categoryRepository.findAllById(categoryIds);
       if (categories.isEmpty()) {
@@ -859,16 +885,18 @@ public class PostService {
         if (tags.isEmpty()) {
           return java.util.List.of();
         }
-        posts = postRepository.findByCategoriesAndAllTagsOrderByCreatedAtDesc(
+        posts = postRepository.findByCategoriesAndAllTagsOrderByPinnedAtDescLastReplyAtDesc(
           categories,
           tags,
           PostStatus.PUBLISHED,
-          tags.size()
+          tags.size(),
+          pageable
         );
       } else {
-        posts = postRepository.findByCategoryInAndStatusOrderByCreatedAtDesc(
+        posts = postRepository.findByCategoryInAndStatusOrderByPinnedAtDescLastReplyAtDesc(
           categories,
-          PostStatus.PUBLISHED
+          PostStatus.PUBLISHED,
+          pageable
         );
       }
     } else {
@@ -876,14 +904,15 @@ public class PostService {
       if (tags.isEmpty()) {
         return new ArrayList<>();
       }
-      posts = postRepository.findByAllTagsOrderByCreatedAtDesc(
+      posts = postRepository.findByAllTagsOrderByPinnedAtDescLastReplyAtDesc(
         tags,
         PostStatus.PUBLISHED,
-        tags.size()
+        tags.size(),
+        pageable
       );
     }
 
-    return paginate(sortByPinnedAndLastReply(posts), page, pageSize);
+    return posts;
   }
 
   public List<Post> listPostsByCategories(
@@ -1381,6 +1410,13 @@ public class PostService {
       .toList();
   }
 
+  private Pageable buildPageable(Integer page, Integer pageSize) {
+    if (page == null || pageSize == null) {
+      return Pageable.unpaged();
+    }
+    return PageRequest.of(page, pageSize);
+  }
+
   private List<Post> paginate(List<Post> posts, Integer page, Integer pageSize) {
     if (page == null || pageSize == null) {
       return posts;

backend/src/main/java/com/openisle/service/ResendEmailSender.java

@@ -1,5 +1,6 @@
 package com.openisle.service;
 
+import com.openisle.exception.EmailSendException;
 import java.util.HashMap;
 import java.util.Map;
 import org.springframework.beans.factory.annotation.Value;
@@ -7,8 +8,9 @@ import org.springframework.http.HttpEntity;
 import org.springframework.http.HttpHeaders;
 import org.springframework.http.HttpMethod;
 import org.springframework.http.MediaType;
-import org.springframework.scheduling.annotation.Async;
+import org.springframework.http.ResponseEntity;
 import org.springframework.stereotype.Service;
+import org.springframework.web.client.RestClientException;
 import org.springframework.web.client.RestTemplate;
 
 @Service
@@ -23,7 +25,6 @@ public class ResendEmailSender extends EmailSender {
   private final RestTemplate restTemplate = new RestTemplate();
 
   @Override
-  @Async("notificationExecutor")
   public void sendEmail(String to, String subject, String text) {
     String url = "https://api.resend.com/emails"; // hypothetical endpoint
 
@@ -38,6 +39,20 @@ public class ResendEmailSender extends EmailSender {
     body.put("from", "openisle <" + fromEmail + ">");
 
     HttpEntity<Map<String, String>> entity = new HttpEntity<>(body, headers);
-    restTemplate.exchange(url, HttpMethod.POST, entity, String.class);
+    try {
+      ResponseEntity<String> response = restTemplate.exchange(
+        url,
+        HttpMethod.POST,
+        entity,
+        String.class
+      );
+      if (!response.getStatusCode().is2xxSuccessful()) {
+        throw new EmailSendException(
+          "Email service returned status " + response.getStatusCodeValue()
+        );
+      }
+    } catch (RestClientException e) {
+      throw new EmailSendException("Failed to send email: " + e.getMessage(), e);
+    }
   }
 }

backend/src/main/java/com/openisle/service/UserService.java

@@ -118,7 +118,6 @@ public class UserService {
    * @param user
    */
   public void sendVerifyMail(User user, VerifyType verifyType) {
-    // 缓存验证码
     String code = genCode();
     String key;
     String subject;
@@ -133,8 +132,9 @@ public class UserService {
       subject = "请填写验证码以重置密码(有效期为5分钟)";
     }
 
-    redisTemplate.opsForValue().set(key, code, 5, TimeUnit.MINUTES); // 五分钟后验证码过期
     emailService.sendEmail(user.getEmail(), subject, content);
+    // 邮件发送成功后再缓存验证码，避免发送失败时用户收不到但验证被要求
+    redisTemplate.opsForValue().set(key, code, 5, TimeUnit.MINUTES); // 五分钟后验证码过期
   }
 
   /**

backend/src/main/resources/db/init/01_schema.sql

@@ -20,6 +20,7 @@ CREATE TABLE IF NOT EXISTS `users` (
   `username` varchar(50) NOT NULL,
   `verification_code` varchar(255) DEFAULT NULL,
   `verified` bit(1) DEFAULT NULL,
+  `is_bot` bit(1) NOT NULL DEFAULT b'0',
   PRIMARY KEY (`id`),
   UNIQUE KEY `UK_users_email` (`email`),
   UNIQUE KEY `UK_users_username` (`username`)

backend/src/main/resources/db/init/02_seed_data.sql

@@ -8,10 +8,28 @@ DELETE FROM `users`;
 
 -- 插入用户，两个普通用户，一个管理员
 -- username:admin/user1/user2 password:123456
-INSERT INTO `users` (`id`, `approved`, `avatar`, `created_at`, `display_medal`, `email`, `experience`, `introduction`, `password`, `password_reset_code`, `point`, `register_reason`, `role`, `username`, `verification_code`, `verified`) VALUES
-(1, b'1', 'https://openisle-1307107697.cos.ap-guangzhou.myqcloud.com/assert/image.png', '2025-09-01 16:08:17.426430', 'PIONEER', 'adminmail@openisle.com', 70, NULL, '$2a$10$x7HXjUyJTmrvqjnBlBQZH.vmfsC56NzTSWqQ6WqZqRjUO859EhviS', NULL, 110, '测试测试测试……', 'ADMIN', 'admin', NULL, b'1'),
-(2, b'1', 'https://openisle-1307107697.cos.ap-guangzhou.myqcloud.com/assert/image.png', '2025-09-03 16:08:17.426430', 'PIONEER', 'usermail2@openisle.com', 70, NULL, '$2a$10$x7HXjUyJTmrvqjnBlBQZH.vmfsC56NzTSWqQ6WqZqRjUO859EhviS', NULL, 110, '测试测试测试……', 'USER', 'user1', NULL, b'1'),
-(3, b'1', 'https://openisle-1307107697.cos.ap-guangzhou.myqcloud.com/assert/image.png', '2025-09-02 17:21:21.617666', 'PIONEER', 'usermail1@openisle.com', 40, NULL, '$2a$10$x7HXjUyJTmrvqjnBlBQZH.vmfsC56NzTSWqQ6WqZqRjUO859EhviS', NULL, 40, '测试测试测试……', 'USER', 'user2', NULL, b'1');
+INSERT INTO `users` (
+  `id`,
+  `approved`,
+  `avatar`,
+  `created_at`,
+  `display_medal`,
+  `email`,
+  `experience`,
+  `introduction`,
+  `password`,
+  `password_reset_code`,
+  `point`,
+  `register_reason`,
+  `role`,
+  `username`,
+  `verification_code`,
+  `verified`,
+  `is_bot`
+) VALUES
+(1, b'1', 'https://openisle-1307107697.cos.ap-guangzhou.myqcloud.com/assert/image.png', '2025-09-01 16:08:17.426430', 'PIONEER', 'adminmail@openisle.com', 70, NULL, '$2a$10$x7HXjUyJTmrvqjnBlBQZH.vmfsC56NzTSWqQ6WqZqRjUO859EhviS', NULL, 110, '测试测试测试……', 'ADMIN', 'admin', NULL, b'1', b'0'),
+(2, b'1', 'https://openisle-1307107697.cos.ap-guangzhou.myqcloud.com/assert/image.png', '2025-09-03 16:08:17.426430', 'PIONEER', 'usermail2@openisle.com', 70, NULL, '$2a$10$x7HXjUyJTmrvqjnBlBQZH.vmfsC56NzTSWqQ6WqZqRjUO859EhviS', NULL, 110, '测试测试测试……', 'USER', 'user1', NULL, b'1', b'0'),
+(3, b'1', 'https://openisle-1307107697.cos.ap-guangzhou.myqcloud.com/assert/image.png', '2025-09-02 17:21:21.617666', 'PIONEER', 'usermail1@openisle.com', 40, NULL, '$2a$10$x7HXjUyJTmrvqjnBlBQZH.vmfsC56NzTSWqQ6WqZqRjUO859EhviS', NULL, 40, '测试测试测试……', 'USER', 'user2', NULL, b'1', b'0');
 
 INSERT INTO `categories` (`id`,`description`,`icon`,`name`,`small_icon`) VALUES
 (1,'测试用分类1','star','测试用分类1',NULL),

backend/src/main/resources/db/migration/V10__backfill_post_last_reply_at.sql

@@ -0,0 +1,4 @@
+-- Backfill last_reply_at for posts without comments to preserve latest-reply ordering
+UPDATE posts
+SET last_reply_at = created_at
+WHERE last_reply_at IS NULL;

backend/src/main/resources/db/migration/V8__add_is_bot_to_users.sql

@@ -0,0 +1,2 @@
+ALTER TABLE users
+ADD COLUMN is_bot BIT(1) NOT NULL DEFAULT b'0';

backend/src/main/resources/db/migration/V9__add_post_reads.sql

@@ -0,0 +1,12 @@
+CREATE TABLE IF NOT EXISTS post_reads (
+  id BIGINT NOT NULL AUTO_INCREMENT,
+  user_id BIGINT NOT NULL,
+  post_id BIGINT NOT NULL,
+  last_read_at DATETIME(6) DEFAULT NULL,
+  PRIMARY KEY (id),
+  UNIQUE KEY UK_post_reads_user_post (user_id, post_id),
+  KEY IDX_post_reads_user (user_id),
+  KEY IDX_post_reads_post (post_id),
+  CONSTRAINT FK_post_reads_user FOREIGN KEY (user_id) REFERENCES users (id) ON DELETE CASCADE,
+  CONSTRAINT FK_post_reads_post FOREIGN KEY (post_id) REFERENCES posts (id) ON DELETE CASCADE
+);

bots/AGENTS.md

@@ -0,0 +1,68 @@
+# Bots 协作指引
+
+## 1) 适用范围
+
+- 作用于 `bots/` 目录及其子目录。
+- 本文件用于统一 Bot 脚本开发、调度与发布规范。
+
+## 2) 模块结构与职责
+
+- `bot_father.ts`：Bot 基类，统一 Agent 初始化、MCP 工具接入、CLI 运行入口。
+- `instance/reply_bot.ts`：常规互动回复 Bot（提及/评论自动处理）。
+- `instance/open_source_reply_bot.ts`：开源问答 Bot（偏代码与贡献流程）。
+- `instance/daily_news_bot.ts`：每日新闻帖 Bot。
+- `instance/coffee_bot.ts`：早安抽奖帖 Bot。
+
+## 3) 开发约定（新增/改造 Bot）
+
+- 新 Bot 统一继承 `BotFather`，最少实现：
+  - `getAdditionalInstructions()`
+  - `getCliQuery()`
+- 保持导出约定：`export const runWorkflow = ...`，并保留 `if (require.main === module)` CLI 入口。
+- 不随意改动 `bot_father.ts` 的 MCP 工具白名单；若必须调整，需同步评估 `mcp/` 契约与线上可用性。
+
+## 4) 环境变量与密钥规范
+
+- 必需：`OPENAI_API_KEY`（缺失会直接失败）。
+- 常用：
+  - `OPENISLE_TOKEN`（用于 OpenIsle MCP 鉴权；GitHub Actions 中可映射不同 secret）
+  - `APIFY_API_TOKEN`（天气 MCP 鉴权）
+- News Bot 可选参数：
+  - `DAILY_NEWS_CATEGORY_ID`
+  - `DAILY_NEWS_TAG_IDS`
+- 严禁在代码中硬编码真实 token；仅通过 CI secrets 或本地环境变量注入。
+
+## 5) 工作流同步规则（与 GitHub Actions 对齐）
+
+- 相关工作流：
+  - `.github/workflows/reply-bots.yml`
+  - `.github/workflows/open_source_reply_bot.yml`
+  - `.github/workflows/news-bot.yml`
+  - `.github/workflows/coffee-bot.yml`
+- 若改脚本入口、依赖或 env 键名，必须同步更新对应 workflow。
+- 若改触发节奏（cron）或 Bot 行为边界，需在变更说明中写明影响（频率、成本、风险）。
+
+## 6) 行为约束（防重复/防失控）
+
+- 回复类 Bot 需保持幂等：避免对同一上下文重复回复。
+- 处理未读后应调用 `mark_notifications_read` 清理通知。
+- 批量处理建议保持上限（当前提示词约定为最多 10 条）。
+- 发帖类 Bot（news/coffee）必须控制 `create_post` 调用次数（一次任务最多一次发帖）。
+- Open Source Reply Bot 保持专业技术风格，避免跑题到非开源问答。
+
+## 7) 本地验证建议
+
+- 依赖安装（与 CI 一致）：
+  - `npm install --no-save @openai/agents tsx typescript`
+- 单脚本运行示例：
+  - `npx tsx bots/instance/reply_bot.ts`
+  - `npx tsx bots/instance/open_source_reply_bot.ts`
+  - `npx tsx bots/instance/daily_news_bot.ts`
+  - `npx tsx bots/instance/coffee_bot.ts`
+- 验证时至少确认：可启动、可调用 MCP、异常时退出码非 0。
+
+## 8) 输出要求
+
+- 说明改动影响哪个 Bot、哪个 workflow。
+- 说明是否改变了工具调用边界（MCP tools / 发帖次数 / 回复策略）。
+- 说明是否需要同步更新文档或运维配置。

bots/bot_father.ts

@@ -0,0 +1,186 @@
+import { Agent, Runner, hostedMcpTool, withTrace, webSearchTool } from "@openai/agents";
+
+export type WorkflowInput = { input_as_text: string };
+
+export abstract class BotFather {
+  protected readonly openisleToken = (process.env.OPENISLE_TOKEN ?? "").trim();
+  protected readonly weatherToken = (process.env.APIFY_API_TOKEN ?? "").trim();
+
+  protected readonly openisleMcp = this.createHostedMcpTool();
+  protected readonly weatherMcp = this.createWeatherMcpTool();
+  protected readonly webSearchPreview = this.createWebSearchPreviewTool();
+  protected readonly agent: Agent;
+
+  constructor(protected readonly name: string) {
+    console.log(`✅ ${this.name} starting...`);
+    console.log(
+      this.openisleToken
+        ? "🔑 OPENISLE_TOKEN detected in environment; it will be attached to MCP requests."
+        : "🔓 OPENISLE_TOKEN not set; authenticated MCP tools may be unavailable."
+    );
+
+    console.log(
+      this.weatherToken
+        ? "☁️ APIFY_API_TOKEN detected; weather MCP server will be available."
+        : "🌥️ APIFY_API_TOKEN not set; weather updates will be unavailable."
+    );
+
+    this.agent = new Agent({
+      name: this.name,
+      instructions: this.buildInstructions(),
+      tools: [
+        this.openisleMcp,
+        this.weatherMcp,
+        this.webSearchPreview
+      ],
+      model: this.getModel(),
+      modelSettings: {
+        temperature: 0.7,
+        topP: 1,
+        maxTokens: 2048,
+        toolChoice: "auto",
+        store: true,
+      },
+    });
+  }
+
+  protected buildInstructions(): string {
+    const instructions = [
+      ...this.getBaseInstructions(),
+      ...this.getAdditionalInstructions(),
+    ].filter(Boolean);
+    return instructions.join("\n");
+  }
+
+  protected getBaseInstructions(): string[] {
+    return [
+      "You are a helpful assistant for https://www.open-isle.com.",
+      "Finish tasks end-to-end before replying. If multiple MCP tools are needed, call them sequentially until the task is truly done.",
+      "When presenting the result, reply in Chinese with a concise summary and include any important URLs or IDs.",
+      "After finishing replies, call mark_notifications_read with all processed notification IDs to keep the inbox clean.",
+    ];
+  }
+
+  private createWebSearchPreviewTool() {
+    return webSearchTool({
+      userLocation: {
+        type: "approximate",
+        country: undefined,
+        region: undefined,
+        city: undefined,
+        timezone: undefined
+      },
+      searchContextSize: "medium"
+    })
+  }
+
+  private createHostedMcpTool() {
+    const token = this.openisleToken;
+    const authConfig = token
+      ? {
+          headers: {
+            Authorization: `Bearer ${token}`,
+          },
+        }
+      : {};
+
+    return hostedMcpTool({
+      serverLabel: "openisle_mcp",
+      serverUrl: "https://www.open-isle.com/mcp",
+      allowedTools: [
+        "search", // 用于搜索帖子、内容等
+        "create_post", // 创建新帖子
+        "reply_to_post", // 回复帖子
+        "reply_to_comment", // 回复评论
+        "recent_posts", // 获取最新帖子
+        "get_post", // 获取特定帖子的详细信息
+        "list_unread_messages", // 列出未读消息或通知
+        "mark_notifications_read", // 标记通知为已读
+      ],
+      requireApproval: "never",
+      ...authConfig,
+    });
+  }
+
+  private createWeatherMcpTool(): ReturnType<typeof hostedMcpTool> {
+    return hostedMcpTool({
+      serverLabel: "weather_mcp_server",
+      serverUrl: "https://jiri-spilka--weather-mcp-server.apify.actor/mcp",
+      requireApproval: "never",
+      allowedTools: [
+        "get_current_weather", // 天气 MCP 工具
+      ],
+      headers: {
+        Authorization: `Bearer ${this.weatherToken || ""}`,
+      },
+    });
+  }
+
+  protected getAdditionalInstructions(): string[] {
+    return [];
+  }
+
+  protected getModel(): string {
+    return "gpt-4o-mini";
+  }
+
+  protected createRunner(): Runner {
+    return new Runner({
+      workflowName: this.name,
+      traceMetadata: {
+        __trace_source__: "agent-builder",
+        workflow_id: "wf_69003cbd47e08190928745d3c806c0b50d1a01cfae052be8",
+      },
+    });
+  }
+
+  public async runWorkflow(workflow: WorkflowInput) {
+    if (!process.env.OPENAI_API_KEY) {
+      throw new Error("Missing OPENAI_API_KEY");
+    }
+
+    const runner = this.createRunner();
+
+    return await withTrace(`${this.name} run`, async () => {
+      const preview = workflow.input_as_text.trim();
+      console.log(
+        "📝 Received workflow input (preview):",
+        preview.length > 200 ? `${preview.slice(0, 200)}…` : preview
+      );
+
+      console.log("🚦 Starting agent run with maxTurns=16...");
+      const result = await runner.run(this.agent, workflow.input_as_text, {
+        maxTurns: 16,
+      });
+
+      console.log("📬 Agent run completed. Result keys:", Object.keys(result));
+
+      if (!result.finalOutput) {
+        throw new Error("Agent result is undefined (no final output).");
+      }
+
+      const openisleBotResult = { output_text: String(result.finalOutput) };
+
+      console.log(
+        "🤖 Agent result (length=%d):\n%s",
+        openisleBotResult.output_text.length,
+        openisleBotResult.output_text
+      );
+      return openisleBotResult;
+    });
+  }
+
+  protected abstract getCliQuery(): string;
+
+  public async runCli(): Promise<void> {
+    try {
+      const query = this.getCliQuery();
+      console.log("🔍 Running workflow...");
+      await this.runWorkflow({ input_as_text: query });
+      process.exit(0);
+    } catch (err: any) {
+      console.error("❌ Agent failed:", err?.stack || err);
+      process.exit(1);
+    }
+  }
+}

bots/instance/coffee_bot.ts

@@ -0,0 +1,56 @@
+import { BotFather, WorkflowInput } from "../bot_father";
+
+const WEEKDAY_NAMES = ["日", "一", "二", "三", "四", "五", "六"] as const;
+
+class CoffeeBot extends BotFather {
+  constructor() {
+    super("Coffee Bot");
+  }
+
+  protected override getAdditionalInstructions(): string[] {
+    return [
+      "记住你的系统代号是 system，有需要自称或签名时都要使用这个名字。",
+      "You are responsible for 发布每日抽奖早安贴。",
+      "创建帖子时，确保标题、奖品信息、开奖时间以及领奖方式完全符合 CLI 查询提供的细节。",
+      "正文需亲切友好，简洁明了，鼓励社区成员互动。",
+      "开奖说明需明确告知中奖者需私聊站长 @nagisa 领取奖励。",
+      "确保只发布一个帖子，避免重复调用 create_post。",
+      "使用标签为 weather_mcp_server 的 MCP 工具获取北京、上海、广州、深圳当天的天气信息，并把结果写入早安问候之后。",
+    ];
+  }
+
+  protected override getCliQuery(): string {
+    const now = new Date(Date.now() + 8 * 60 * 60 * 1000);
+    const weekday = WEEKDAY_NAMES[now.getDay()];
+    const drawTime = new Date(now);
+    drawTime.setHours(15, 0, 0, 0);
+
+    return `
+请立即在 https://www.open-isle.com 使用 create_post 发表一篇帖子，遵循以下要求：
+1. 标题固定为「大家星期${weekday}早安--抽一杯咖啡」。
+2. 正文包含：
+   - 亲切的早安问候；
+   - 早安问候后立即列出北京、上海、广州、深圳当天的天气信息，每行格式为“城市：天气描述，最低温~最高温”；天气需调用 weather_mcp_server 获取；
+   - 标注“领奖请私聊站长 @[nagisa]”；
+   - 鼓励大家留言互动。
+3. 奖品信息
+   - 明确奖品写作“Coffee”；
+   - 帖子类型必须为 LOTTERY；
+   - 奖品图片链接：https://openisle-1307107697.cos.accelerate.myqcloud.com/dynamic_assert/0d6a9b33e9ca4fe5a90540187d3f9ecb.png；
+   - 公布开奖时间为 ${drawTime}, 直接传UTC时间给接口，不要考虑时区问题
+   - categoryId 固定为 10，tagIds 设为 [36]。
+4. 帖子语言使用简体中文。
+`.trim();
+  }
+}
+
+const coffeeBot = new CoffeeBot();
+
+export const runWorkflow = async (workflow: WorkflowInput) => {
+  return coffeeBot.runWorkflow(workflow);
+};
+
+if (require.main === module) {
+  coffeeBot.runCli();
+}
+

bots/instance/daily_news_bot.ts

@@ -0,0 +1,69 @@
+import { BotFather, WorkflowInput } from "../bot_father";
+
+const WEEKDAY_NAMES = ["日", "一", "二", "三", "四", "五", "六"] as const;
+
+class DailyNewsBot extends BotFather {
+  constructor() {
+    super("Daily News Bot");
+  }
+
+  protected override getModel(): string {
+    return "gpt-4o";
+  }
+
+  protected override getAdditionalInstructions(): string[] {
+    return [
+      "You are DailyNewsBot，专职在 OpenIsle 发布每日新闻速递。",
+      "始终使用简体中文回复，并以结构化 Markdown 呈现内容。",
+      "发布内容前务必完成资讯核实：分别通过 web_search 调研 CoinDesk 所有要闻、Reuters 重点国际新闻，以及全球 AI 领域的重大进展。",
+      "整合新闻时，将同源资讯合并，突出影响力、涉及主体与潜在影响，保持语句简洁。",
+      "所有新闻要点都要附带来源链接，并在括号中标注来源站点名。",
+      "使用 weather_mcp_server 的 get_current_weather 获取北京、上海、广州、深圳的天气，并在正文中列表展示",
+      "正文结尾补充一个行动建议或提醒，帮助读者快速把握重点。",
+      "严禁发布超过一篇帖子，create_post 只调用一次。",
+    ];
+  }
+
+  protected override getCliQuery(): string {
+    const now = new Date(Date.now() + 8 * 60 * 60 * 1000);
+    const year = now.getFullYear();
+    const month = String(now.getMonth() + 1).padStart(2, "0");
+    const day = String(now.getDate()).padStart(2, "0");
+    const weekday = WEEKDAY_NAMES[now.getDay()];
+    const dateLabel = `${year}年${month}月${day}日 星期${weekday}`;
+    const isoDate = `${year}-${month}-${day}`;
+    const categoryId = Number(process.env.DAILY_NEWS_CATEGORY_ID ?? "6");
+    const tagIdsEnv = process.env.DAILY_NEWS_TAG_IDS ?? "3,33";
+    const tagIds = tagIdsEnv
+      .split(",")
+      .map((id) => Number(id.trim()))
+      .filter((id) => !Number.isNaN(id));
+    const finalTagIds = tagIds.length > 0 ? tagIds : [1];
+    const tagIdsText = `[${finalTagIds.join(", ")}]`;
+
+    return `
+请立即在 https://www.open-isle.com 使用 create_post 发布一篇名为「OpenIsle 每日新闻速递｜${dateLabel}」的帖子，并遵循以下要求：
+1. 发布类型为 NORMAL，categoryId = ${categoryId}，tagIds = ${tagIdsText}。
+2. 正文以简洁问候开头, 不用再重复标题
+3. 使用 web_search 工具按以下顺序收集资讯，并在正文中以 Markdown 小节呈现， 需要调用3次web_search：
+   - 「全球区块链与加密」：汇总 coindesk.com 版面所有重点新闻, 列出至少5条
+   - 「国际新闻速览」：汇总 reuters.com 版面重点头条，关注宏观经济、市场波动或政策变化。列出至少5条
+   - 「AI 行业快讯」：检索今天全球 AI 领域的重要发布或事件（例如 OpenAI、Google、Meta、国内大模型厂商等）。列出至少5条
+4. 每条新闻采用项目符号，先写结论再给出关键数字或细节，末尾添加来源超链接，格式示例：「**结论** —— 关键细节。（来源：[Reuters](URL)）」
+5. 资讯整理完毕后，调用 weather_mcp_server.get_current_weather，列出北京、上海、广州、深圳今日天气，放置在「城市天气」小节下, 本小节可加emoji。
+6. 最后一节为「今日提醒」，给出 2-3 条与新闻或天气相关的行动建议。
+7. 若在资讯搜集过程中发现相互矛盾的信息，须在正文中以「⚠️ 风险提示」说明原因及尚待确认的点。
+9. 发布完成后，不要再次调用 create_post。
+`.trim();
+  }
+}
+
+const dailyNewsBot = new DailyNewsBot();
+
+export const runWorkflow = async (workflow: WorkflowInput) => {
+  return dailyNewsBot.runWorkflow(workflow);
+};
+
+if (require.main === module) {
+  dailyNewsBot.runCli();
+}

bots/instance/open_source_reply_bot.ts

@@ -0,0 +1,65 @@
+import { readFileSync } from "node:fs";
+import path from "node:path";
+import { BotFather, WorkflowInput } from "../bot_father";
+
+class OpenSourceReplyBot extends BotFather {
+  constructor() {
+    super("OpenSource Reply Bot");
+  }
+
+  protected override getAdditionalInstructions(): string[] {
+    const knowledgeBase = this.loadKnowledgeBase();
+
+    return [
+      "You are OpenSourceReplyBot, a professional helper who focuses on answering open-source development and code-related questions for the OpenIsle community.",
+      "Respond in Chinese using well-structured Markdown sections such as 标题、列表、代码块等，让回复清晰易读。",
+      "保持语气专业、耐心、详尽，绝不使用表情符号或颜文字，也不要卖萌。",
+      "优先解答与项目代码、贡献流程、架构设计或排错相关的问题；",
+      "在需要时引用 README.md 与 CONTRIBUTING.md 中的要点，帮助用户快速定位文档位置。",
+      knowledgeBase,
+    ].filter(Boolean);
+  }
+
+  protected override getCliQuery(): string {
+    return `
+【AUTO】每30分钟自动巡检未读提及与评论，严格遵守以下流程：
+1）调用 list_unread_messages 获取待处理的“提及/评论”；
+2）按时间从新到旧逐条处理（最多10条）；如需上下文请调用 get_post；
+3）仅对与开源项目、代码实现或贡献流程直接相关的问题生成详尽的 Markdown 中文回复，
+   若与主题无关则礼貌说明并跳过；
+4）回复时引用 README 或 CONTRIBUTING 中的要点（如适用），并优先给出可执行的排查步骤或代码建议；
+5）回复评论使用 reply_to_comment，回复帖子使用 reply_to_post；
+6）若某通知最后一条已由本 bot 回复，则跳过避免重复；
+7）整理已处理通知 ID 调用 mark_notifications_read；
+8）结束时输出包含处理条目概览（URL或ID）的总结。`.trim();
+  }
+
+  private loadKnowledgeBase(): string {
+    const docs = ["../../README.md", "../../CONTRIBUTING.md"];
+    const sections: string[] = [];
+
+    for (const relativePath of docs) {
+      try {
+        const absolutePath = path.resolve(__dirname, relativePath);
+        const content = readFileSync(absolutePath, "utf-8").trim();
+        if (content) {
+          sections.push(`以下是 ${path.basename(absolutePath)} 的内容：\n${content}`);
+        }
+      } catch (error) {
+        sections.push(`未能加载 ${relativePath}，请检查文件路径或权限。`);
+      }
+    }
+
+    return sections.join("\n\n");
+  }
+}
+
+const openSourceReplyBot = new OpenSourceReplyBot();
+
+export const runWorkflow = async (workflow: WorkflowInput) => {
+  return openSourceReplyBot.runWorkflow(workflow);
+};
+
+if (require.main === module) {
+  openSourceReplyBot.runCli();
+}

bots/instance/reply_bot.ts

@@ -0,0 +1,38 @@
+// reply_bot.ts
+import { BotFather, WorkflowInput } from "../bot_father";
+
+class ReplyBot extends BotFather {
+  constructor() {
+    super("OpenIsle Bot");
+  }
+
+  protected override getAdditionalInstructions(): string[] {
+    return [
+      "记住你的系统代号是 system，任何需要自称、署名或解释身份的时候都使用这个名字。",
+      "以阴阳怪气的方式回复各种互动",
+      "你每天会发布咖啡抽奖贴，跟大家互动",
+    ];
+  }
+
+  protected override getCliQuery(): string {
+    return `
+      【AUTO】无需确认，自动处理所有未读的提及与评论：
+      1）调用 list_unread_messages；
+      2）依次处理每条“提及/评论”：如需上下文则使用 get_post 获取，生成简明中文回复；如有 commentId 则用 reply_to_comment，否则用 reply_to_post；
+      3）跳过关注和系统事件；
+      4）保证幂等性：如该贴最后一条是你自己发的回复，则跳过；
+      5）调用 mark_notifications_read，传入本次已处理的通知 ID 清理已读；
+      6）最多只处理最新10条；结束时仅输出简要摘要（包含URL或ID）。
+    `.trim();
+  }
+}
+
+const replyBot = new ReplyBot();
+
+export const runWorkflow = async (workflow: WorkflowInput) => {
+  return replyBot.runWorkflow(workflow);
+};
+
+if (require.main === module) {
+  replyBot.runCli();
+}

deploy/AGENTS.md

@@ -0,0 +1,40 @@
+# Deploy 协作指引
+
+## 1) 适用范围
+
+- 作用于 `deploy/` 目录及其脚本。
+- 该目录为高风险变更区，默认保守修改。
+
+## 2) 当前部署基线
+
+- 预发：`main` push 触发（见 `.github/workflows/deploy-staging.yml`）。
+- 正式：定时任务触发（见 `.github/workflows/deploy.yml`）。
+- 两者使用同一并发锁 `openisle-server`，避免服务器并发部署冲突。
+
+## 3) 脚本修改原则
+
+- 保留 `set -euo pipefail` 等安全执行特性。
+- 变更服务列表或 `docker compose up` 参数时，必须说明影响范围。
+- 不随意改动 `git fetch/checkout/reset` 逻辑；若改，必须附回滚方案。
+- 任何“可能中断服务”的改动，都要在说明中给出停机/风险评估。
+
+## 4) 环境与参数规则
+
+- 部署依赖根目录 `.env`（由脚本中 `env_file` 与 `ENV_FILE` 传入）。
+- `COMPOSE_PROJECT_NAME`、`NUXT_ENV`、服务名列表需保持可追踪且与 compose 一致。
+- 若新增服务，需同步：
+  - `docker/docker-compose.yaml`
+  - 部署脚本中的 build/up 目标
+  - 必要时更新 workflow 说明
+
+## 5) 验证建议
+
+- 语法检查：
+  - `bash -n deploy/deploy.sh`
+  - `bash -n deploy/deploy_staging.sh`
+- 变更前后做一次 `docker compose config` 思维核对（服务与 profile 是否正确）。
+
+## 6) 输出要求
+
+- 明确：影响环境（预发/正式）、影响服务、是否可能重建容器。
+- 必填：回滚路径（例如切回上一 commit 并重新执行部署脚本）。

docker/docker-compose.yaml

@@ -30,62 +30,62 @@ services:
       - dev_local_backend
       - prod
 
-  # OpenSearch Service
-  opensearch:
-    user: "1000:1000"
-    build:
-      context: .
-      dockerfile: opensearch.Dockerfile
-    container_name: ${COMPOSE_PROJECT_NAME}-opensearch
-    environment:
-      - cluster.name=os-single
-      - node.name=os-node-1
-      - discovery.type=single-node
-      - bootstrap.memory_lock=true
-      - OPENSEARCH_JAVA_OPTS=-Xms1g -Xmx1g
-      - DISABLE_SECURITY_PLUGIN=true
-      - cluster.blocks.create_index=false
-    ulimits:
-      memlock: { soft: -1, hard: -1 }
-      nofile: { soft: 65536, hard: 65536 }
-    volumes:
-      - opensearch-data:/usr/share/opensearch/data
-      - opensearch-snapshots:/snapshots
-    ports:
-      - "${OPENSEARCH_PORT:-9200}:9200"
-      - "${OPENSEARCH_METRICS_PORT:-9600}:9600"
-    restart: unless-stopped
-    healthcheck:
-      test:
-        - CMD-SHELL
-        - curl -fsS http://127.0.0.1:9200/_cluster/health >/dev/null
-      interval: 10s
-      timeout: 5s
-      retries: 30
-      start_period: 60s
-    networks:
-      - openisle-network
-    profiles:
-      - dev
-      - dev_local_backend
+  # # OpenSearch Service
+  # opensearch:
+  #   user: "1000:1000"
+  #   build:
+  #     context: .
+  #     dockerfile: opensearch.Dockerfile
+  #   container_name: ${COMPOSE_PROJECT_NAME}-opensearch
+  #   environment:
+  #     - cluster.name=os-single
+  #     - node.name=os-node-1
+  #     - discovery.type=single-node
+  #     - bootstrap.memory_lock=true
+  #     - OPENSEARCH_JAVA_OPTS=-Xms1g -Xmx1g
+  #     - DISABLE_SECURITY_PLUGIN=true
+  #     - cluster.blocks.create_index=false
+  #   ulimits:
+  #     memlock: { soft: -1, hard: -1 }
+  #     nofile: { soft: 65536, hard: 65536 }
+  #   volumes:
+  #     - opensearch-data:/usr/share/opensearch/data
+  #     - opensearch-snapshots:/snapshots
+  #   ports:
+  #     - "${OPENSEARCH_PORT:-9200}:9200"
+  #     - "${OPENSEARCH_METRICS_PORT:-9600}:9600"
+  #   restart: unless-stopped
+  #   healthcheck:
+  #     test:
+  #       - CMD-SHELL
+  #       - curl -fsS http://127.0.0.1:9200/_cluster/health >/dev/null
+  #     interval: 10s
+  #     timeout: 5s
+  #     retries: 30
+  #     start_period: 60s
+  #   networks:
+  #     - openisle-network
+  #   profiles:
+  #     - dev
+  #     - dev_local_backend
 
-  dashboards:
-    image: opensearchproject/opensearch-dashboards:3.0.0
-    container_name: ${COMPOSE_PROJECT_NAME}-os-dashboards
-    environment:
-      OPENSEARCH_HOSTS: '["http://opensearch:9200"]'
-      DISABLE_SECURITY_DASHBOARDS_PLUGIN: "true"
-    ports:
-      - "${OPENSEARCH_DASHBOARDS_PORT:-5601}:5601"
-    depends_on:
-      - opensearch
-    restart: unless-stopped
-    networks:
-      - openisle-network
-    profiles:
-      - dev
-      - dev_local_backend
-      - prod
+  # dashboards:
+  #   image: opensearchproject/opensearch-dashboards:3.0.0
+  #   container_name: ${COMPOSE_PROJECT_NAME}-os-dashboards
+  #   environment:
+  #     OPENSEARCH_HOSTS: '["http://opensearch:9200"]'
+  #     DISABLE_SECURITY_DASHBOARDS_PLUGIN: "true"
+  #   ports:
+  #     - "${OPENSEARCH_DASHBOARDS_PORT:-5601}:5601"
+  #   depends_on:
+  #     - opensearch
+  #   restart: unless-stopped
+  #   networks:
+  #     - openisle-network
+  #   profiles:
+  #     - dev
+  #     - dev_local_backend
+  #     - prod
 
   rabbitmq:
     image: rabbitmq:3.13-management
@@ -200,7 +200,6 @@ services:
       - openisle-network
     profiles:
       - dev
-      - dev_local_backend
       - prod
 
 

docs/AGENTS.md

@@ -0,0 +1,40 @@
+# Docs 协作指引
+
+## 1) 适用范围
+
+- 作用于 `docs/` 目录及其子目录。
+- 文档需服务“开发者真实使用”，优先准确性与可执行性。
+
+## 2) 文档架构
+
+- 内容目录：`content/docs/`
+- 生成脚本：`scripts/generate-docs.ts`
+- OpenAPI 输入配置：`lib/openapi.ts`
+- 前端框架：Fumadocs + Next.js（Bun 工具链）
+
+## 3) 编辑规则
+
+- 优先修正“与代码不一致”的文档，不复制过时描述。
+- 涉及技术栈说明时，以当前代码为准（例如后端为 JPA/Repository）。
+- OpenAPI 自动生成目录（`content/docs/openapi/(generated)`）不要手工细改，改源头配置与脚本。
+- 结构性改动优先维持导航稳定（`meta.json` 与已有 slug）。
+
+## 4) OpenAPI 同步规则
+
+- 后端 API 变更后，应重新生成文档页面：
+  - `bun run generate`
+- 若接口来源地址或文档聚合策略变化，更新：
+  - `lib/openapi.ts`
+  - `scripts/generate-docs.ts`
+
+## 5) 验证命令
+
+- 安装依赖：`bun install`
+- 生成 API 文档：`bun run generate`
+- 构建校验：`bun run build`
+- 本地预览：`bun dev`
+
+## 6) 输出要求
+
+- 说明更新了哪些文档入口（backend/frontend/openapi）。
+- 说明是否需要后端先部署后再刷新文档产物。

frontend_nuxt/AGENTS.md

@@ -0,0 +1,54 @@
+# Frontend（Nuxt）协作指引
+
+## 1) 适用范围
+
+- 作用于 `frontend_nuxt/` 目录及其子目录。
+- 本文件仅覆盖前端范围；跨服务规则仍遵循根 `AGENTS.md`。
+
+## 2) 代码组织约定
+
+- `pages/`：路由页面与页面级数据获取。
+- `components/`：可复用视图组件。
+- `composables/`：状态与行为复用（如 WebSocket、倒计时）。
+- `plugins/`：运行时插件（鉴权 fetch、主题、第三方库注入）。
+- `utils/`：纯工具函数（时间、鉴权 token、平台适配）。
+- `assets/`、`public/`：静态资源与样式。
+
+## 3) 前端修改规则
+
+- 优先保持现有交互和视觉风格一致，不做无关 UI 重构。
+- 接口字段变更时，先更新调用点，再统一处理回退逻辑与空值分支。
+- SSR 与客户端代码分离：
+  - 涉及 `window`、`localStorage`、WebSocket 的逻辑只在 client 侧运行。
+- 与鉴权相关的 401 处理，保持与 `plugins/auth-fetch.client.ts` 行为一致。
+
+## 4) 环境变量与运行时配置
+
+- 统一通过 `nuxt.config.ts` 的 `runtimeConfig.public` 读取。
+- 关键键位保持一致：
+  - `NUXT_PUBLIC_API_BASE_URL`
+  - `NUXT_PUBLIC_WEBSOCKET_URL`
+  - `NUXT_PUBLIC_WEBSITE_BASE_URL`
+- 变量改动需同步根目录 `.env.example` 与文档说明。
+
+## 5) 实时消息链路注意事项
+
+- WebSocket 入口：
+  - `composables/useWebSocket.js`
+- 若改订阅目标（`/topic/...`、`/user/...`），必须与后端推送目的地保持一致。
+- 重连与重订阅逻辑不可被破坏；避免引入重复订阅和泄漏。
+
+## 6) 构建与验证
+
+- 标准验证：`npm run build`
+- 本地联调：`npm run dev`
+- 涉及 WebSocket/通知的改动，建议至少手工验证：
+  - 登录后连接建立
+  - 收到消息时 UI 状态更新
+  - 断线重连后仍可订阅
+
+## 7) 输出要求
+
+- 标注影响页面/组件路径。
+- 标注是否引入 API 字段兼容处理。
+- 标注是否需要后端或 WebSocket 服务配合发布。

frontend_nuxt/components/CommentItem.vue

@@ -16,6 +16,7 @@
         <div class="info-content-header-left">
           <span class="user-name">{{ comment.userName }}</span>
           <span v-if="isCommentFromPostAuthor" class="op-badge" title="楼主">OP</span>
+          <span v-if="comment.isBot" class="bot-badge" title="Bot">Bot</span>
           <medal-one class="medal-icon" />
           <NuxtLink
             v-if="comment.medal"
@@ -522,6 +523,21 @@ const handleContentClick = (e) => {
   line-height: 1;
 }
 
+.bot-badge {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  margin-left: 6px;
+  padding: 0 6px;
+  height: 18px;
+  border-radius: 9px;
+  background-color: rgba(76, 175, 80, 0.16);
+  color: #2e7d32;
+  font-size: 12px;
+  font-weight: 600;
+  line-height: 1;
+}
+
 .medal-icon {
   font-size: 12px;
   opacity: 0.6;

frontend_nuxt/components/TimelineReadItem.vue

@@ -0,0 +1,149 @@
+<template>
+  <div class="timeline-container">
+    <div class="timeline-header">
+      <div class="timeline-title">浏览了文章</div>
+      <div class="timeline-date">{{ formattedDate }}</div>
+    </div>
+    <div class="article-container">
+      <NuxtLink :to="postLink" class="timeline-article-link">
+        {{ item.post?.title }}
+      </NuxtLink>
+      <div class="timeline-snippet">
+        {{ strippedSnippet }}
+      </div>
+      <div class="article-meta" v-if="hasMeta">
+        <ArticleCategory v-if="item.post?.category" :category="item.post.category" />
+        <ArticleTags :tags="item.post?.tags" />
+        <div class="article-comment-count" v-if="item.post?.commentCount !== undefined">
+          <comment-one class="article-comment-count-icon" />
+          <span>{{ item.post?.commentCount }}</span>
+        </div>
+      </div>
+    </div>
+  </div>
+</template>
+
+<script setup>
+import { computed } from 'vue'
+import { stripMarkdown } from '~/utils/markdown'
+import TimeManager from '~/utils/time'
+
+const props = defineProps({
+  item: { type: Object, required: true },
+})
+
+const postLink = computed(() => {
+  const id = props.item.post?.id
+  return id ? `/posts/${id}` : '#'
+})
+
+const formattedDate = computed(() =>
+  TimeManager.format(props.item.lastReadAt ?? props.item.createdAt),
+)
+const strippedSnippet = computed(() => stripMarkdown(props.item.post?.snippet ?? ''))
+const hasMeta = computed(() => {
+  const tags = props.item.post?.tags ?? []
+  const hasTags = Array.isArray(tags) && tags.length > 0
+  const hasCategory = !!props.item.post?.category
+  const hasCommentCount =
+    props.item.post?.commentCount !== undefined && props.item.post?.commentCount !== null
+  return hasTags || hasCategory || hasCommentCount
+})
+</script>
+
+<style scoped>
+.timeline-container {
+  display: flex;
+  flex-direction: column;
+  padding-top: 5px;
+  gap: 12px;
+  border-radius: 10px;
+  background: var(--timeline-card-background, transparent);
+}
+
+.timeline-header {
+  display: flex;
+  justify-content: space-between;
+  align-items: center;
+}
+
+.timeline-title {
+  font-size: 16px;
+  font-weight: 600;
+}
+
+.timeline-date {
+  font-size: 12px;
+  color: var(--timeline-date-color, #888);
+}
+
+.article-container {
+  display: flex;
+  flex-direction: column;
+  border: 1px solid var(--normal-border-color);
+  border-radius: 10px;
+  padding: 10px;
+  gap: 6px;
+}
+
+.timeline-article-link {
+  font-size: 18px;
+  font-weight: 600;
+  color: var(--link-color);
+  text-decoration: none;
+}
+
+.timeline-article-link:hover {
+  text-decoration: underline;
+}
+
+.timeline-snippet {
+  color: var(--timeline-snippet-color, #666);
+  font-size: 14px;
+  line-height: 1.6;
+}
+
+.article-meta {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 8px;
+  align-items: center;
+}
+
+.article-tags {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 6px;
+}
+
+.article-tag {
+  background-color: var(--article-info-background-color);
+  border-radius: 6px;
+  padding: 2px 6px;
+  font-size: 12px;
+  color: var(--text-color);
+}
+
+.article-comment-count {
+  display: inline-flex;
+  align-items: center;
+  gap: 4px;
+  font-size: 12px;
+  color: var(--text-color);
+}
+
+.article-comment-count-icon {
+  width: 16px;
+  height: 16px;
+}
+
+@media (max-width: 768px) {
+  .timeline-article-link {
+    font-size: 16px;
+  }
+
+  .timeline-snippet {
+    font-size: 13px;
+  }
+}
+</style>

frontend_nuxt/pages/index.vue

@@ -159,6 +159,12 @@ const selectedTags = ref([])
 const route = useRoute()
 const tagOptions = ref([])
 const categoryOptions = ref([])
+const clearFilters = () => {
+  selectedCategory.value = ''
+  selectedTags.value = []
+  selectedCategoryGlobal.value = ''
+  selectedTagsGlobal.value = []
+}
 
 const topics = ref(['最新回复', '最新', '精选', '排行榜' /*, '热门', '类别'*/])
 const selectedTopicCookie = useCookie('homeTab')
@@ -218,8 +224,18 @@ watch(
   (query) => {
     const category = query.category
     const tags = query.tags
-    category && selectedCategorySet(category)
-    tags && selectedTagsSet(tags)
+    if (category) {
+      selectedCategorySet(category)
+    } else {
+      selectedCategory.value = ''
+      selectedCategoryGlobal.value = ''
+    }
+    if (tags) {
+      selectedTagsSet(tags)
+    } else {
+      selectedTags.value = []
+      selectedTagsGlobal.value = []
+    }
   },
 )
 
@@ -367,12 +383,18 @@ watch(selectedTopic, (val) => {
 if (import.meta.server) {
   await loadOptions()
 }
+
+const handleRefreshHome = () => {
+  clearFilters()
+  refreshFirst()
+}
+
 onMounted(() => {
   if (categoryOptions.value.length === 0 && tagOptions.value.length === 0) loadOptions()
-  window.addEventListener('refresh-home', refreshFirst)
+  window.addEventListener('refresh-home', handleRefreshHome)
 })
 onBeforeUnmount(() => {
-  window.removeEventListener('refresh-home', refreshFirst)
+  window.removeEventListener('refresh-home', handleRefreshHome)
 })
 
 /** 供 InfiniteLoadMore 重建用的 key：筛选/Tab 改变即重建内部状态 */

frontend_nuxt/pages/login.vue

@@ -58,12 +58,15 @@ const submitLogin = async () => {
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({ username: username.value, password: password.value }),
     })
-    const data = await res.json()
+    const data = await res.json().catch(() => ({}))
     if (res.ok && data.token) {
       setToken(data.token)
       toast.success('登录成功')
       registerPush()
       await navigateTo('/', { replace: true })
+    } else if (data.reason_code === 'EMAIL_SEND_FAILED') {
+      const msg = data.error || data.message || res.statusText || '登录失败'
+      toast.error(`${res.status} ${msg} (${data.reason_code})`)
     } else if (data.reason_code === 'NOT_VERIFIED') {
       toast.info('当前邮箱未验证，已经为您重新发送验证码')
       await navigateTo(
@@ -76,10 +79,12 @@ const submitLogin = async () => {
     } else if (data.reason_code === 'NOT_APPROVED') {
       await navigateTo({ path: '/signup-reason', query: { token: data.token } }, { replace: true })
     } else {
-      toast.error(data.error || '登录失败')
+      const msg = data.error || data.message || res.statusText || '登录失败'
+      const reason = data.reason_code ? ` (${data.reason_code})` : ''
+      toast.error(`${res.status} ${msg}${reason}`)
     }
   } catch (e) {
-    toast.error('登录失败')
+    toast.error(`登录失败: ${e.message}`)
   } finally {
     isWaitingForLogin.value = false
   }

frontend_nuxt/pages/posts/[id]/index.vue

@@ -181,6 +181,12 @@
             <PostChangeLogItem v-else :log="item" :title="title" />
           </template>
         </BaseTimeline>
+        <InfiniteLoadMore
+          v-if="timelineItems.length > 0"
+          :key="commentSort"
+          :on-load="loadMoreTimeline"
+          :pause="isLoadingMoreComments || isFetchingComments"
+        />
       </div>
     </div>
 
@@ -211,6 +217,7 @@ import CommentEditor from '~/components/CommentEditor.vue'
 import BaseTimeline from '~/components/BaseTimeline.vue'
 import BasePlaceholder from '~/components/BasePlaceholder.vue'
 import PostChangeLogItem from '~/components/PostChangeLogItem.vue'
+import InfiniteLoadMore from '~/components/InfiniteLoadMore.vue'
 import ArticleTags from '~/components/ArticleTags.vue'
 import ArticleCategory from '~/components/ArticleCategory.vue'
 import ReactionsGroup from '~/components/ReactionsGroup.vue'
@@ -276,6 +283,10 @@ const currentIndex = ref(1)
 const subscribed = ref(false)
 const commentSort = ref('NEWEST')
 const isFetchingComments = ref(false)
+const commentPage = ref(0)
+const commentPageSize = 10
+const hasMoreComments = ref(true)
+const isLoadingMoreComments = ref(false)
 const isMobile = useIsMobile()
 const timelineItems = ref([])
 
@@ -377,6 +388,7 @@ const mapComment = (
   text: c.content,
   reactions: c.reactions || [],
   pinned: Boolean(c.pinned ?? c.pinnedAt ?? c.pinned_at),
+  isBot: Boolean(c.author?.bot),
   reply: (c.replies || []).map((r) =>
     mapComment(r, c.author.username, c.author.avatar, c.author.id, level + 1),
   ),
@@ -532,7 +544,7 @@ const {
     } catch (err) {}
   },
   {
-    server: false,
+    server: true,
     lazy: false,
   },
 )
@@ -868,17 +880,33 @@ const fetchCommentSorts = () => {
   ])
 }
 
-const fetchCommentsAndChangeLog = async () => {
-  isFetchingComments.value = true
-  console.info('Fetching comments and chang log', { postId, sort: commentSort.value })
+const fetchCommentsAndChangeLog = async ({ pageNo = 0, append = false } = {}) => {
+  if (isLoadingMoreComments.value) return false
+  if (!append) {
+    hasMoreComments.value = true
+    commentPage.value = 0
+  }
+  if (pageNo === 0) {
+    isFetchingComments.value = true
+  } else {
+    isLoadingMoreComments.value = true
+  }
+  console.info('Fetching comments and chang log', {
+    postId,
+    sort: commentSort.value,
+    page: pageNo,
+    pageSize: commentPageSize,
+  })
+  let done = false
   try {
     const token = getToken()
-    const res = await fetch(
-      `${API_BASE_URL}/api/posts/${postId}/comments?sort=${commentSort.value}`,
-      {
-        headers: { Authorization: token ? `Bearer ${token}` : '' },
-      },
-    )
+    const url = new URL(`${API_BASE_URL}/api/posts/${postId}/comments`)
+    url.searchParams.set('sort', commentSort.value)
+    url.searchParams.set('page', String(pageNo))
+    url.searchParams.set('pageSize', String(commentPageSize))
+    const res = await fetch(url.toString(), {
+      headers: { Authorization: token ? `Bearer ${token}` : '' },
+    })
     console.info('Fetch comments response status', res.status)
     if (res.ok) {
       const data = await res.json()
@@ -901,23 +929,48 @@ const fetchCommentsAndChangeLog = async () => {
         }
       }
 
-      comments.value = commentList
-      changeLogs.value = changeLogList
-      timelineItems.value = newTimelineItemList
+      if (append) {
+        comments.value.push(...commentList)
+        changeLogs.value.push(...changeLogList)
+        timelineItems.value.push(...newTimelineItemList)
+        commentPage.value = pageNo
+      } else {
+        comments.value = commentList
+        changeLogs.value = changeLogList
+        timelineItems.value = newTimelineItemList
+        commentPage.value = 0
+      }
 
-      isFetchingComments.value = false
+      done = data.length < commentPageSize
+      hasMoreComments.value = !done
       await nextTick()
       gatherPostItems()
+      return done
     }
   } catch (e) {
     console.debug('Fetch comments error', e)
+    hasMoreComments.value = false
+    return true
   } finally {
     isFetchingComments.value = false
+    isLoadingMoreComments.value = false
   }
 }
 
 const fetchTimeline = async () => {
-  await fetchCommentsAndChangeLog()
+  hasMoreComments.value = true
+  commentPage.value = 0
+  comments.value = []
+  changeLogs.value = []
+  timelineItems.value = []
+  await fetchCommentsAndChangeLog({ pageNo: 0, append: false })
+}
+
+const loadMoreTimeline = async () => {
+  if (!hasMoreComments.value || isLoadingMoreComments.value) return true
+  const nextPage = commentPage.value + 1
+  const done = await fetchCommentsAndChangeLog({ pageNo: nextPage, append: true })
+  return done || !hasMoreComments.value
 }
 
 watch(commentSort, async () => {
@@ -928,8 +981,17 @@ const jumpToHashComment = async () => {
   const hash = location.hash
   if (hash.startsWith('#comment-')) {
     const id = hash.substring('#comment-'.length)
-    await new Promise((resolve) => setTimeout(resolve, 500))
-    const el = document.getElementById('comment-' + id)
+    await new Promise((resolve) => setTimeout(resolve, 300))
+    let el = document.getElementById('comment-' + id)
+
+    // 若未加载到目标评论，尝试继续分页加载直到找到或无更多
+    while (!el && hasMoreComments.value) {
+      const done = await loadMoreTimeline()
+      await nextTick()
+      el = document.getElementById('comment-' + id)
+      if (done) break
+    }
+
     if (el) {
       const top = el.getBoundingClientRect().top + window.scrollY - headerHeight - 20 // 20 for beauty
       window.scrollTo({ top, behavior: 'smooth' })
@@ -1394,10 +1456,6 @@ onMounted(async () => {
   font-weight: bold;
 }
 
-.reaction-action.copy-link:hover {
-  background-color: #e2e2e2;
-}
-
 .comment-editor-wrapper {
   position: relative;
 }

frontend_nuxt/pages/signup.vue

@@ -139,8 +139,7 @@ const sendVerification = async () => {
         inviteToken: inviteToken.value,
       }),
     })
-    isWaitingForEmailSent.value = false
-    const data = await res.json()
+    const data = await res.json().catch(() => ({}))
     if (res.ok) {
       emailStep.value = 1
       toast.success('验证码已发送，请查看邮箱')
@@ -149,10 +148,14 @@ const sendVerification = async () => {
       if (data.field === 'email') emailError.value = data.error
       if (data.field === 'password') passwordError.value = data.error
     } else {
-      toast.error(data.error || '发送失败')
+      const msg = data.error || data.message || res.statusText || '发送失败'
+      const reason = data.reason_code ? ` (${data.reason_code})` : ''
+      toast.error(`${res.status} ${msg}${reason}`)
     }
   } catch (e) {
-    toast.error('发送失败')
+    toast.error(`发送失败: ${e.message}`)
+  } finally {
+    isWaitingForEmailSent.value = false
   }
 }
 

frontend_nuxt/pages/users/[id].vue

@@ -191,14 +191,25 @@
               >
                 评论和回复
               </div>
+              <div
+                v-if="isMine"
+                :class="['timeline-tab-item', { selected: timelineFilter === 'reads' }]"
+                @click="timelineFilter = 'reads'"
+              >
+                浏览记录
+              </div>
             </div>
             <BasePlaceholder
-              v-if="filteredTimelineItems.length === 0"
-              text="暂无时间线"
+              v-if="
+                timelineFilter === 'reads'
+                  ? readPosts.length === 0
+                  : filteredTimelineItems.length === 0
+              "
+              :text="timelineFilter === 'reads' ? '暂无浏览记录' : '暂无时间线'"
               icon="inbox"
             />
             <div class="timeline-list">
-              <BaseTimeline :items="filteredTimelineItems">
+              <BaseTimeline v-if="timelineFilter !== 'reads'" :items="filteredTimelineItems">
                 <template #item="{ item }">
                   <template v-if="item.type === 'post'">
                     <TimelinePostItem :item="item" />
@@ -214,6 +225,11 @@
                   </template>
                 </template>
               </BaseTimeline>
+              <BaseTimeline v-else :items="readPosts">
+                <template #item="{ item }">
+                  <TimelineReadItem :item="item" />
+                </template>
+              </BaseTimeline>
             </div>
           </div>
 
@@ -276,6 +292,7 @@ import BaseTabs from '~/components/BaseTabs.vue'
 import LevelProgress from '~/components/LevelProgress.vue'
 import TimelineCommentGroup from '~/components/TimelineCommentGroup.vue'
 import TimelinePostItem from '~/components/TimelinePostItem.vue'
+import TimelineReadItem from '~/components/TimelineReadItem.vue'
 import TimelineTagItem from '~/components/TimelineTagItem.vue'
 import BaseUserAvatar from '~/components/BaseUserAvatar.vue'
 import UserList from '~/components/UserList.vue'
@@ -299,12 +316,15 @@ const hotReplies = ref([])
 const hotTags = ref([])
 const favoritePosts = ref([])
 const timelineItems = ref([])
+const readPosts = ref([])
 const timelineFilter = ref('all')
 const filteredTimelineItems = computed(() => {
   if (timelineFilter.value === 'articles') {
     return timelineItems.value.filter((item) => item.type === 'post')
   } else if (timelineFilter.value === 'comments') {
     return timelineItems.value.filter((item) => item.type === 'comment' || item.type === 'reply')
+  } else if (timelineFilter.value === 'reads') {
+    return []
   }
   return timelineItems.value
 })
@@ -477,6 +497,27 @@ const fetchTimeline = async () => {
   timelineItems.value = combineDiscussionItems(mapped)
 }
 
+const fetchReadHistory = async () => {
+  if (!isMine.value) {
+    readPosts.value = []
+    return
+  }
+  const token = getToken()
+  if (!token) {
+    readPosts.value = []
+    return
+  }
+  const res = await fetch(`${API_BASE_URL}/api/users/${username}/read-posts?limit=50`, {
+    headers: { Authorization: `Bearer ${token}` },
+  })
+  if (res.ok) {
+    const data = await res.json()
+    readPosts.value = data.map((r) => ({ ...r, icon: 'file-text' }))
+  } else {
+    readPosts.value = []
+  }
+}
+
 const fetchFollowUsers = async () => {
   const [followerRes, followingRes] = await Promise.all([
     fetch(`${API_BASE_URL}/api/users/${username}/followers`),
@@ -508,6 +549,12 @@ const loadTimeline = async () => {
   tabLoading.value = false
 }
 
+const loadReadHistory = async () => {
+  tabLoading.value = true
+  await fetchReadHistory()
+  tabLoading.value = false
+}
+
 const loadFollow = async () => {
   tabLoading.value = true
   await fetchFollowUsers()
@@ -624,8 +671,14 @@ onMounted(init)
 
 watch(selectedTab, async (val) => {
   // navigateTo({ query: { ...route.query, tab: val } }, { replace: true })
-  if (val === 'timeline' && timelineItems.value.length === 0) {
-    await loadTimeline()
+  if (val === 'timeline') {
+    if (timelineFilter.value === 'reads') {
+      if (readPosts.value.length === 0) {
+        await loadReadHistory()
+      }
+    } else if (timelineItems.value.length === 0) {
+      await loadTimeline()
+    }
   } else if (val === 'following' && followers.value.length === 0 && followings.value.length === 0) {
     await loadFollow()
   } else if (val === 'favorites' && favoritePosts.value.length === 0) {
@@ -634,6 +687,23 @@ watch(selectedTab, async (val) => {
     await loadAchievements()
   }
 })
+
+watch(timelineFilter, async (val) => {
+  if (selectedTab.value !== 'timeline') return
+  if (val === 'reads') {
+    if (readPosts.value.length === 0) {
+      await loadReadHistory()
+    }
+  } else if (timelineItems.value.length === 0) {
+    await loadTimeline()
+  }
+})
+
+watch(isMine, (val) => {
+  if (!val && timelineFilter.value === 'reads') {
+    timelineFilter.value = 'all'
+  }
+})
 </script>
 
 <style scoped>

frontend_nuxt/utils/notification.js

@@ -82,6 +82,56 @@ export async function markNotificationsRead(ids) {
   }
 }
 
+const MARK_ALL_FETCH_SIZE = 100
+const MARK_ALL_CHUNK_SIZE = 200
+const MARK_ALL_MAX_PAGES = 200
+
+async function fetchUnreadNotificationsPage(page, size) {
+  const config = useRuntimeConfig()
+  const API_BASE_URL = config.public.apiBaseUrl
+  const token = getToken()
+  if (!token) throw new Error('NO_TOKEN')
+
+  const res = await fetch(`${API_BASE_URL}/api/notifications/unread?page=${page}&size=${size}`, {
+    headers: {
+      Authorization: `Bearer ${token}`,
+    },
+  })
+
+  if (!res.ok) throw new Error('FETCH_UNREAD_FAILED')
+  const data = await res.json()
+  return Array.isArray(data) ? data : []
+}
+
+async function collectUnreadNotificationIds(excludedTypes = []) {
+  const excludedTypeSet = new Set(excludedTypes)
+  const ids = []
+
+  for (let page = 0; page < MARK_ALL_MAX_PAGES; page++) {
+    const pageData = await fetchUnreadNotificationsPage(page, MARK_ALL_FETCH_SIZE)
+    if (pageData.length === 0) break
+
+    for (const notification of pageData) {
+      if (!notification || excludedTypeSet.has(notification.type)) continue
+      if (typeof notification.id !== 'number') continue
+      ids.push(notification.id)
+    }
+
+    if (pageData.length < MARK_ALL_FETCH_SIZE) break
+  }
+
+  return [...new Set(ids)]
+}
+
+async function markNotificationsReadInChunks(ids) {
+  for (let i = 0; i < ids.length; i += MARK_ALL_CHUNK_SIZE) {
+    const chunk = ids.slice(i, i + MARK_ALL_CHUNK_SIZE)
+    const ok = await markNotificationsRead(chunk)
+    if (!ok) return false
+  }
+  return true
+}
+
 export async function fetchNotificationPreferences() {
   try {
     const config = useRuntimeConfig()
@@ -390,29 +440,37 @@ function createFetchNotifications() {
   }
 
   const markAllRead = async () => {
-    // 除了 REGISTER_REQUEST 类型消息
-    const idsToMark = notifications.value
+    // 为了覆盖分页中的全部未读，先从后端分页拉取全部未读 ID（排除 REGISTER_REQUEST）。
+    const localIdsToMark = notifications.value
       .filter((n) => n.type !== 'REGISTER_REQUEST' && !n.read)
       .map((n) => n.id)
-    if (idsToMark.length === 0) return
+
     notifications.value.forEach((n) => {
       if (n.type !== 'REGISTER_REQUEST') n.read = true
     })
     notificationState.unreadCount = notifications.value.filter((n) => !n.read).length
-    const ok = await markNotificationsRead(idsToMark)
-    if (!ok) {
+
+    try {
+      const idsToMark = await collectUnreadNotificationIds(['REGISTER_REQUEST'])
+      if (idsToMark.length > 0) {
+        const ok = await markNotificationsReadInChunks(idsToMark)
+        if (!ok) throw new Error('MARK_READ_FAILED')
+      }
+
+      await fetchUnreadCount()
+      if (authState.role === 'ADMIN') {
+        toast.success('已读所有消息（注册请求除外）')
+      } else {
+        toast.success('已读所有消息')
+      }
+    } catch (e) {
       notifications.value.forEach((n) => {
-        if (idsToMark.includes(n.id)) n.read = false
+        if (localIdsToMark.includes(n.id)) n.read = false
       })
       await fetchUnreadCount()
+      toast.error('已读操作失败，请稍后重试')
       return
     }
-    fetchUnreadCount()
-    if (authState.role === 'ADMIN') {
-      toast.success('已读所有消息（注册请求除外）')
-    } else {
-      toast.success('已读所有消息')
-    }
   }
   return {
     fetchNotifications,

mcp/AGENTS.md

@@ -0,0 +1,51 @@
+# MCP 服务协作指引
+
+## 1) 适用范围
+
+- 作用于 `mcp/` 目录及其子目录。
+- 本模块对外提供 MCP tools，接口兼容性要求高。
+
+## 2) 模块结构
+
+- `src/openisle_mcp/server.py`：Tool 定义与请求处理入口。
+- `src/openisle_mcp/search_client.py`：调用 OpenIsle 后端 HTTP API。
+- `src/openisle_mcp/schemas.py`：Pydantic 数据契约。
+- `src/openisle_mcp/config.py`：运行配置与环境变量读取。
+
+## 3) 变更原则
+
+- Tool 名称默认视为稳定契约，非必要不重命名。
+- 后端接口字段变化时，优先同步 `schemas.py`，再调整 `server.py` 映射。
+- 对认证接口保持“显式失败”：
+  - 缺 token、401、403 需给出可理解错误信息。
+- 避免吞掉异常上下文，保留足够定位信息（HTTP 状态、接口语义）。
+
+## 4) 与后端契约同步
+
+- 高风险同步点：
+  - `create_post`
+  - `reply_to_post`
+  - `reply_to_comment`
+  - `list_unread_messages`
+  - `mark_notifications_read`
+- 若后端响应结构改动，需同步：
+  - `search_client.py` 的解析逻辑
+  - `schemas.py` 的校验模型
+  - `README.md` 的 tool 说明（如有新增/删减）
+
+## 5) 配置规则
+
+- 环境变量统一使用 `OPENISLE_MCP_*` 前缀。
+- 保持默认值可本地运行（如 `http://localhost:8080` 场景）。
+- 不在代码中硬编码私密 token。
+
+## 6) 验证建议
+
+- 安装校验：`python -m pip install -e .`
+- 启动校验：`openisle-mcp`（或项目内等价启动方式）
+- 如改动 schema/解析逻辑，至少完成一次真实后端联调请求。
+
+## 7) 输出要求
+
+- 说明变更是否影响 tool 输入/输出契约。
+- 说明是否要求调用方更新（参数名、字段、错误语义）。

mcp/README.md

@@ -31,9 +31,12 @@ By default the server listens on port `8085` and serves MCP over Streamable HTTP
 | Tool | Description |
 | --- | --- |
 | `search` | Perform a global search against the OpenIsle backend. |
+| `create_post` | Publish a new post using a JWT token. |
+| `reply_to_post` | Create a new comment on a post using a JWT token. |
 | `reply_to_comment` | Reply to an existing comment using a JWT token. |
 | `recent_posts` | Retrieve posts created within the last *N* minutes. |
 
 The tools return structured data mirroring the backend DTOs, including highlighted snippets for
-search results, the full comment payload for replies, and detailed metadata for recent posts.
+search results, the full comment payload for post replies and comment replies, and detailed
+metadata for recent posts.
 

mcp/src/openisle_mcp/config.py

@@ -5,7 +5,7 @@ from __future__ import annotations
 from functools import lru_cache
 from typing import Literal
 
-from pydantic import Field
+from pydantic import Field, SecretStr
 from pydantic.networks import AnyHttpUrl
 from pydantic_settings import BaseSettings, SettingsConfigDict
 
@@ -36,6 +36,20 @@ class Settings(BaseSettings):
         gt=0,
         description="Timeout (seconds) for backend search requests.",
     )
+    access_token: SecretStr | None = Field(
+        default=None,
+        description=(
+            "Optional JWT bearer token used for authenticated backend calls. "
+            "When set, tools that support authentication will use this token "
+            "automatically unless an explicit token override is provided."
+        ),
+    )
+    log_level: str = Field(
+        "INFO",
+        description=(
+            "Logging level for the MCP server (e.g. DEBUG, INFO, WARNING)."
+        ),
+    )
 
     model_config = SettingsConfigDict(
         env_prefix="OPENISLE_MCP_",

mcp/src/openisle_mcp/schemas.py

@@ -5,7 +5,7 @@ from __future__ import annotations
 from datetime import datetime
 from typing import Any, Optional
 
-from pydantic import BaseModel, Field, ConfigDict
+from pydantic import BaseModel, Field, ConfigDict, field_validator
 
 
 class SearchResultItem(BaseModel):
@@ -170,6 +170,15 @@ class CommentData(BaseModel):
 
     model_config = ConfigDict(populate_by_name=True, extra="allow")
 
+    @field_validator("replies", "reactions", mode="before")
+    @classmethod
+    def _ensure_comment_lists(cls, value: Any) -> list[Any]:
+        """Convert ``None`` payloads to empty lists for comment collections."""
+
+        if value is None:
+            return []
+        return value
+
 
 class CommentReplyResult(BaseModel):
     """Structured response returned when replying to a comment."""
@@ -177,6 +186,18 @@ class CommentReplyResult(BaseModel):
     comment: CommentData = Field(description="Reply comment returned by the backend.")
 
 
+class CommentCreateResult(BaseModel):
+    """Structured response returned when creating a comment on a post."""
+
+    comment: CommentData = Field(description="Comment returned by the backend.")
+
+
+class PostCreateResult(BaseModel):
+    """Structured response returned when creating a new post."""
+
+    post: PostDetail = Field(description="Detailed post payload returned by the backend.")
+
+
 class PostSummary(BaseModel):
     """Summary information for a post."""
 
@@ -247,6 +268,15 @@ class PostSummary(BaseModel):
 
     model_config = ConfigDict(populate_by_name=True, extra="allow")
 
+    @field_validator("tags", "reactions", "participants", mode="before")
+    @classmethod
+    def _ensure_post_lists(cls, value: Any) -> list[Any]:
+        """Normalize ``None`` values returned by the backend to empty lists."""
+
+        if value is None:
+            return []
+        return value
+
 
 class RecentPostsResponse(BaseModel):
     """Structured response for the recent posts tool."""
@@ -260,3 +290,89 @@ class RecentPostsResponse(BaseModel):
 
 
 CommentData.model_rebuild()
+
+
+class PostDetail(PostSummary):
+    """Detailed information for a single post, including comments."""
+
+    comments: list[CommentData] = Field(
+        default_factory=list,
+        description="Comments that belong to the post.",
+    )
+
+    model_config = ConfigDict(populate_by_name=True, extra="allow")
+
+    @field_validator("comments", mode="before")
+    @classmethod
+    def _ensure_comments_list(cls, value: Any) -> list[Any]:
+        """Treat ``None`` comments payloads as empty lists."""
+
+        if value is None:
+            return []
+        return value
+
+
+class NotificationData(BaseModel):
+    """Unread notification payload returned by the backend."""
+
+    id: Optional[int] = Field(default=None, description="Notification identifier.")
+    type: Optional[str] = Field(default=None, description="Type of the notification.")
+    post: Optional[PostSummary] = Field(
+        default=None, description="Post associated with the notification if applicable."
+    )
+    comment: Optional[CommentData] = Field(
+        default=None, description="Comment referenced by the notification when available."
+    )
+    parent_comment: Optional[CommentData] = Field(
+        default=None,
+        alias="parentComment",
+        description="Parent comment for nested replies, when present.",
+    )
+    from_user: Optional[AuthorInfo] = Field(
+        default=None,
+        alias="fromUser",
+        description="User who triggered the notification.",
+    )
+    reaction_type: Optional[str] = Field(
+        default=None,
+        alias="reactionType",
+        description="Reaction type for reaction-based notifications.",
+    )
+    content: Optional[str] = Field(
+        default=None, description="Additional content or message for the notification."
+    )
+    approved: Optional[bool] = Field(
+        default=None, description="Approval status for moderation notifications."
+    )
+    read: Optional[bool] = Field(default=None, description="Whether the notification is read.")
+    created_at: Optional[datetime] = Field(
+        default=None,
+        alias="createdAt",
+        description="Timestamp when the notification was created.",
+    )
+
+    model_config = ConfigDict(populate_by_name=True, extra="allow")
+
+
+class UnreadNotificationsResponse(BaseModel):
+    """Structured response for unread notification queries."""
+
+    page: int = Field(description="Requested page index for the unread notifications.")
+    size: int = Field(description="Requested page size for the unread notifications.")
+    total: int = Field(description="Number of unread notifications returned in this page.")
+    notifications: list[NotificationData] = Field(
+        default_factory=list,
+        description="Unread notifications returned by the backend.",
+    )
+
+
+class NotificationCleanupResult(BaseModel):
+    """Structured response returned after marking notifications as read."""
+
+    processed_ids: list[int] = Field(
+        default_factory=list,
+        description="Identifiers that were marked as read in the backend.",
+    )
+    total_marked: int = Field(
+        description="Total number of notifications successfully marked as read.",
+    )

mcp/src/openisle_mcp/search_client.py

@@ -3,38 +3,109 @@
 from __future__ import annotations
 
 import json
+import logging
 from typing import Any
 
 import httpx
 
 
+logger = logging.getLogger(__name__)
+
+
 class SearchClient:
     """Client for calling the OpenIsle HTTP APIs used by the MCP server."""
 
-    def __init__(self, base_url: str, *, timeout: float = 10.0) -> None:
+    def __init__(
+        self,
+        base_url: str,
+        *,
+        timeout: float = 10.0,
+        access_token: str | None = None,
+    ) -> None:
         self._base_url = base_url.rstrip("/")
         self._timeout = timeout
         self._client: httpx.AsyncClient | None = None
+        self._access_token = self._sanitize_token(access_token)
 
     def _get_client(self) -> httpx.AsyncClient:
         if self._client is None:
-            self._client = httpx.AsyncClient(base_url=self._base_url, timeout=self._timeout)
+            logger.debug(
+                "Creating httpx.AsyncClient for base URL %s with timeout %.2fs",
+                self._base_url,
+                self._timeout,
+            )
+            self._client = httpx.AsyncClient(
+                base_url=self._base_url,
+                timeout=self._timeout,
+            )
         return self._client
 
+    @staticmethod
+    def _sanitize_token(token: str | None) -> str | None:
+        if token is None:
+            return None
+        stripped = token.strip()
+        return stripped or None
+
+    def update_access_token(self, token: str | None) -> None:
+        """Update the default access token used for authenticated requests."""
+
+        self._access_token = self._sanitize_token(token)
+        if self._access_token:
+            logger.debug("Configured default access token for SearchClient requests.")
+        else:
+            logger.debug("Cleared default access token for SearchClient requests.")
+
+    def _resolve_token(self, token: str | None) -> str | None:
+        candidate = self._sanitize_token(token)
+        if candidate is not None:
+            return candidate
+        return self._access_token
+
+    def _require_token(self, token: str | None) -> str:
+        resolved = self._resolve_token(token)
+        if resolved is None:
+            raise ValueError(
+                "Authenticated request requires an access token. Provide a Bearer token "
+                "via the MCP Authorization header or configure a default token for the server."
+            )
+        return resolved
+
+    def _build_headers(
+        self,
+        *,
+        token: str,
+        accept: str = "application/json",
+        include_json: bool = False,
+    ) -> dict[str, str]:
+        headers: dict[str, str] = {"Accept": accept}
+        resolved = self._resolve_token(token)
+        if resolved:
+            headers["Authorization"] = f"Bearer {resolved}"
+        if include_json:
+            headers["Content-Type"] = "application/json"
+        return headers
+
     async def global_search(self, keyword: str) -> list[dict[str, Any]]:
         """Call the global search endpoint and return the parsed JSON payload."""
 
         client = self._get_client()
+        logger.debug("Calling global search with keyword=%s", keyword)
         response = await client.get(
             "/api/search/global",
             params={"keyword": keyword},
-            headers={"Accept": "application/json"},
+            headers=self._build_headers(),
         )
         response.raise_for_status()
         payload = response.json()
         if not isinstance(payload, list):
             formatted = json.dumps(payload, ensure_ascii=False)[:200]
             raise ValueError(f"Unexpected response format from search endpoint: {formatted}")
+        logger.info(
+            "Global search returned %d results for keyword '%s'",
+            len(payload),
+            keyword,
+        )
         return [self._ensure_dict(entry) for entry in payload]
 
     async def reply_to_comment(
@@ -47,33 +118,101 @@ class SearchClient:
         """Reply to an existing comment and return the created reply."""
 
         client = self._get_client()
-        headers = {
-            "Accept": "application/json",
-            "Content-Type": "application/json",
-            "Authorization": f"Bearer {token}",
-        }
+        resolved_token = self._require_token(token)
+        headers = self._build_headers(token=resolved_token, include_json=True)
         payload: dict[str, Any] = {"content": content}
         if captcha is not None:
             stripped_captcha = captcha.strip()
             if stripped_captcha:
                 payload["captcha"] = stripped_captcha
 
+        logger.debug(
+            "Posting reply to comment_id=%s (captcha=%s)",
+            comment_id,
+            bool(captcha),
+        )
         response = await client.post(
             f"/api/comments/{comment_id}/replies",
             json=payload,
             headers=headers,
         )
         response.raise_for_status()
-        return self._ensure_dict(response.json())
+        body = self._ensure_dict(response.json())
+        logger.info("Reply to comment_id=%s succeeded with id=%s", comment_id, body.get("id"))
+        return body
+
+    async def reply_to_post(
+        self,
+        post_id: int,
+        token: str,
+        content: str,
+        captcha: str | None = None,
+    ) -> dict[str, Any]:
+        """Create a comment on a post and return the backend payload."""
+
+        client = self._get_client()
+        resolved_token = self._require_token(token)
+        headers = self._build_headers(token=resolved_token, include_json=True)
+        payload: dict[str, Any] = {"content": content}
+        if captcha is not None:
+            stripped_captcha = captcha.strip()
+            if stripped_captcha:
+                payload["captcha"] = stripped_captcha
+
+        logger.debug(
+            "Posting comment to post_id=%s (captcha=%s)",
+            post_id,
+            bool(captcha),
+        )
+        response = await client.post(
+            f"/api/posts/{post_id}/comments",
+            json=payload,
+            headers=headers,
+        )
+        response.raise_for_status()
+        body = self._ensure_dict(response.json())
+        logger.info("Reply to post_id=%s succeeded with id=%s", post_id, body.get("id"))
+        return body
+
+    async def create_post(
+        self,
+        payload: dict[str, Any],
+        *,
+        token: str,
+    ) -> dict[str, Any]:
+        """Create a new post and return the detailed backend payload."""
+
+        client = self._get_client()
+        resolved_token = self._require_token(token)
+        headers = self._build_headers(token=resolved_token, include_json=True)
+
+        logger.debug(
+            "Creating post with category_id=%s and %d tag(s)",
+            payload.get("categoryId"),
+            len(payload.get("tagIds", []) if isinstance(payload.get("tagIds"), list) else []),
+        )
+        response = await client.post(
+            "/api/posts",
+            json=payload,
+            headers=headers,
+        )
+        response.raise_for_status()
+        body = self._ensure_dict(response.json())
+        logger.info("Post creation succeeded with id=%s, token=%s", body.get("id"), token)
+        return body
 
     async def recent_posts(self, minutes: int) -> list[dict[str, Any]]:
         """Return posts created within the given timeframe."""
 
         client = self._get_client()
+        logger.debug(
+            "Fetching recent posts within last %s minutes",
+            minutes,
+        )
         response = await client.get(
             "/api/posts/recent",
             params={"minutes": minutes},
-            headers={"Accept": "application/json"},
+            headers=self._build_headers(),
         )
         response.raise_for_status()
         payload = response.json()
@@ -82,14 +221,120 @@ class SearchClient:
             raise ValueError(
                 f"Unexpected response format from recent posts endpoint: {formatted}"
             )
+        logger.info(
+            "Fetched %d recent posts for window=%s minutes",
+            len(payload),
+            minutes,
+        )
         return [self._ensure_dict(entry) for entry in payload]
 
+    async def get_post(self, post_id: int, token: str | None = None) -> dict[str, Any]:
+        """Retrieve the detailed payload for a single post."""
+
+        client = self._get_client()
+        headers = self._build_headers(token=token)
+        logger.debug("Fetching post details for post_id=%s", post_id)
+        response = await client.get(f"/api/posts/{post_id}", headers=headers)
+        response.raise_for_status()
+        body = self._ensure_dict(response.json())
+        logger.info(
+            "Retrieved post_id=%s successfully with %d top-level comments",
+            post_id,
+            len(body.get("comments", []) if isinstance(body.get("comments"), list) else []),
+        )
+        return body
+
+    async def list_unread_notifications(
+        self,
+        *,
+        page: int = 0,
+        size: int = 30,
+        token: str,
+    ) -> list[dict[str, Any]]:
+        """Return unread notifications for the authenticated user."""
+
+        client = self._get_client()
+        resolved_token = self._require_token(token)
+        logger.debug(
+            "Fetching unread notifications with page=%s, size=%s",
+            page,
+            size,
+        )
+        response = await client.get(
+            "/api/notifications/unread",
+            params={"page": page, "size": size},
+            headers=self._build_headers(token=resolved_token),
+        )
+        response.raise_for_status()
+        payload = response.json()
+        if not isinstance(payload, list):
+            formatted = json.dumps(payload, ensure_ascii=False)[:200]
+            raise ValueError(
+                "Unexpected response format from unread notifications endpoint: "
+                f"{formatted}"
+            )
+        logger.info(
+            "Fetched %d unread notifications (page=%s, size=%s)",
+            len(payload),
+            page,
+            size,
+        )
+        return [self._ensure_dict(entry) for entry in payload]
+
+    async def mark_notifications_read(
+        self,
+        ids: list[int],
+        *,
+        token: str
+    ) -> None:
+        """Mark the provided notifications as read for the authenticated user."""
+
+        if not ids:
+            raise ValueError(
+                "At least one notification identifier must be provided to mark as read."
+            )
+
+        sanitized_ids: list[int] = []
+        for value in ids:
+            if isinstance(value, bool):
+                raise ValueError("Notification identifiers must be integers, not booleans.")
+            try:
+                converted = int(value)
+            except (TypeError, ValueError) as exc:  # pragma: no cover - defensive
+                raise ValueError(
+                    "Notification identifiers must be integers."
+                ) from exc
+            if converted <= 0:
+                raise ValueError(
+                    "Notification identifiers must be positive integers."
+                )
+            sanitized_ids.append(converted)
+
+        client = self._get_client()
+        resolved_token = self._require_token(token)
+        logger.debug(
+            "Marking %d notifications as read: ids=%s",
+            len(sanitized_ids),
+            sanitized_ids,
+        )
+        response = await client.post(
+            "/api/notifications/read",
+            json={"ids": sanitized_ids},
+            headers=self._build_headers(token=resolved_token, include_json=True),
+        )
+        response.raise_for_status()
+        logger.info(
+            "Successfully marked %d notifications as read.",
+            len(sanitized_ids),
+        )
+
     async def aclose(self) -> None:
         """Dispose of the underlying HTTP client."""
 
         if self._client is not None:
             await self._client.aclose()
             self._client = None
+            logger.debug("Closed httpx.AsyncClient for SearchClient.")
 
     @staticmethod
     def _ensure_dict(entry: Any) -> dict[str, Any]:

mcp/src/openisle_mcp/server.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import logging
 from contextlib import asynccontextmanager
 from typing import Annotated
 
@@ -12,8 +13,14 @@ from pydantic import Field as PydanticField
 
 from .config import get_settings
 from .schemas import (
+    CommentCreateResult,
     CommentData,
     CommentReplyResult,
+    NotificationData,
+    NotificationCleanupResult,
+    UnreadNotificationsResponse,
+    PostDetail,
+    PostCreateResult,
     PostSummary,
     RecentPostsResponse,
     SearchResponse,
@@ -22,26 +29,79 @@ from .schemas import (
 from .search_client import SearchClient
 
 settings = get_settings()
+if not logging.getLogger().handlers:
+    logging.basicConfig(
+        level=getattr(logging, settings.log_level.upper(), logging.INFO),
+        format="%(asctime)s | %(levelname)s | %(name)s | %(message)s",
+    )
+else:
+    logging.getLogger().setLevel(
+        getattr(logging, settings.log_level.upper(), logging.INFO)
+    )
+
+logger = logging.getLogger(__name__)
+
 search_client = SearchClient(
-    str(settings.backend_base_url), timeout=settings.request_timeout
+    str(settings.backend_base_url),
+    timeout=settings.request_timeout,
+    access_token=(
+        settings.access_token.get_secret_value()
+        if settings.access_token is not None
+        else None
+    ),
 )
 
 
+def _extract_authorization_token(ctx: Context | None) -> str | None:
+    """Return the Bearer token from the incoming MCP request headers."""
+
+    if ctx is None:
+        return None
+
+    try:
+        request_context = ctx.request_context
+    except ValueError:
+        return None
+
+    request = getattr(request_context, "request", None)
+    if request is None:
+        return None
+
+    headers = getattr(request, "headers", None)
+    if headers is None:
+        return None
+
+    authorization = headers.get("authorization")
+    if not authorization:
+        return None
+
+    scheme, _, token = authorization.partition(" ")
+    if scheme.lower() != "bearer":
+        return None
+
+    stripped = token.strip()
+    return stripped or None
+
+
 @asynccontextmanager
 async def lifespan(_: FastMCP):
     """Lifecycle hook that disposes shared resources when the server stops."""
 
     try:
+        logger.debug("OpenIsle MCP server lifespan started.")
         yield
     finally:
+        logger.debug("Disposing shared SearchClient instance.")
         await search_client.aclose()
 
 
 app = FastMCP(
     name="openisle-mcp",
     instructions=(
-        "Use this server to search OpenIsle content, reply to comments with an authentication "
-        "token, and list posts created within a recent time window."
+        "Use this server to search OpenIsle content, create new posts, reply to posts and "
+        "comments using the Authorization header or configured access token, retrieve details "
+        "for a specific post, list posts created within a recent time window, and review "
+        "unread notification messages."
     ),
     host=settings.host,
     port=settings.port,
@@ -65,6 +125,7 @@ async def search(
         raise ValueError("Keyword must not be empty.")
 
     try:
+        logger.info("Received search request for keyword='%s'", sanitized)
         raw_results = await search_client.global_search(sanitized)
     except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
         message = (
@@ -90,13 +151,122 @@ async def search(
 
     if ctx is not None:
         await ctx.info(f"Search keyword '{sanitized}' returned {len(results)} results.")
+    logger.debug(
+        "Validated %d search results for keyword='%s'",
+        len(results),
+        sanitized,
+    )
 
     return SearchResponse(keyword=sanitized, total=len(results), results=results)
 
 
+@app.tool(
+    name="reply_to_post",
+    description=(
+        "Create a comment on a post using the request Authorization header or the configured "
+        "access token."
+    ),
+    structured_output=True,
+)
+async def reply_to_post(
+    post_id: Annotated[
+        int,
+        PydanticField(ge=1, description="Identifier of the post being replied to."),
+    ],
+    content: Annotated[
+        str,
+        PydanticField(description="Markdown content of the reply."),
+    ],
+    captcha: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Optional captcha solution if the backend requires it.",
+        ),
+    ] = None,
+    ctx: Context | None = None,
+) -> CommentCreateResult:
+    """Create a comment on a post and return the backend payload."""
+
+    sanitized_content = content.strip()
+    if not sanitized_content:
+        raise ValueError("Reply content must not be empty.")
+
+    sanitized_captcha = captcha.strip() if isinstance(captcha, str) else None
+
+    request_token = _extract_authorization_token(ctx)
+
+    try:
+        logger.info(
+            "Creating reply for post_id=%s (captcha=%s)",
+            post_id,
+            bool(sanitized_captcha),
+        )
+        raw_comment = await search_client.reply_to_post(
+            post_id,
+            token=request_token,
+            content=sanitized_content,
+            captcha=sanitized_captcha,
+        )
+    except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
+        status_code = exc.response.status_code
+        if status_code == 401:
+            message = (
+                "Authentication failed while replying to post "
+                f"{post_id}. Please verify the Authorization header or configured token."
+            )
+        elif status_code == 403:
+            message = (
+                "The provided Authorization token is not authorized to reply to post "
+                f"{post_id}."
+            )
+        elif status_code == 404:
+            message = f"Post {post_id} was not found."
+        else:
+            message = (
+                "OpenIsle backend returned HTTP "
+                f"{status_code} while replying to post {post_id}."
+            )
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+    except httpx.RequestError as exc:  # pragma: no cover - network errors
+        message = (
+            "Unable to reach OpenIsle backend comment service: "
+            f"{exc}."
+        )
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    try:
+        comment = CommentData.model_validate(raw_comment)
+    except ValidationError as exc:
+        message = "Received malformed data from the post comment endpoint."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    if ctx is not None:
+        await ctx.info(
+            "Reply created successfully for post "
+            f"{post_id}."
+        )
+    logger.debug(
+        "Validated reply comment payload for post_id=%s (comment_id=%s)",
+        post_id,
+        comment.id,
+    )
+
+    return CommentCreateResult(comment=comment)
+
+
 @app.tool(
     name="reply_to_comment",
-    description="Reply to an existing comment using an authentication token.",
+    description=(
+        "Reply to an existing comment using the request Authorization header or the configured "
+        "access token."
+    ),
     structured_output=True,
 )
 async def reply_to_comment(
@@ -104,7 +274,6 @@ async def reply_to_comment(
         int,
         PydanticField(ge=1, description="Identifier of the comment being replied to."),
     ],
-    token: Annotated[str, PydanticField(description="JWT bearer token for the user performing the reply.")],
     content: Annotated[
         str,
         PydanticField(description="Markdown content of the reply."),
@@ -124,29 +293,32 @@ async def reply_to_comment(
     if not sanitized_content:
         raise ValueError("Reply content must not be empty.")
 
-    sanitized_token = token.strip()
-    if not sanitized_token:
-        raise ValueError("Authentication token must not be empty.")
-
     sanitized_captcha = captcha.strip() if isinstance(captcha, str) else None
 
+    request_token = _extract_authorization_token(ctx)
+
     try:
+        logger.info(
+            "Creating reply for comment_id=%s (captcha=%s)",
+            comment_id,
+            bool(sanitized_captcha),
+        )
         raw_comment = await search_client.reply_to_comment(
             comment_id,
-            sanitized_token,
-            sanitized_content,
-            sanitized_captcha,
+            token=request_token,
+            content=sanitized_content,
+            captcha=sanitized_captcha,
         )
     except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
         status_code = exc.response.status_code
         if status_code == 401:
             message = (
                 "Authentication failed while replying to comment "
-                f"{comment_id}. Please verify the token."
+                f"{comment_id}. Please verify the Authorization header or configured token."
             )
         elif status_code == 403:
             message = (
-                "The provided token is not authorized to reply to comment "
+                "The provided Authorization token is not authorized to reply to comment "
                 f"{comment_id}."
             )
         else:
@@ -179,10 +351,351 @@ async def reply_to_comment(
             "Reply created successfully for comment "
             f"{comment_id}."
         )
+    logger.debug(
+        "Validated reply payload for comment_id=%s (reply_id=%s)",
+        comment_id,
+        comment.id,
+    )
 
     return CommentReplyResult(comment=comment)
 
 
+@app.tool(
+    name="create_post",
+    description=(
+        "Publish a new post using the request Authorization header or the configured access "
+        "token."
+    ),
+    structured_output=True,
+)
+async def create_post(
+    title: Annotated[
+        str,
+        PydanticField(description="Title of the post to be created."),
+    ],
+    content: Annotated[
+        str,
+        PydanticField(description="Markdown content of the post."),
+    ],
+    category_id: Annotated[
+        int | None,
+        PydanticField(
+            default=None,
+            ge=1,
+            description="Optional category identifier for the post.",
+        ),
+    ] = None,
+    tag_ids: Annotated[
+        list[int] | None,
+        PydanticField(
+            default=None,
+            min_length=1,
+            description="Optional list of tag identifiers to assign to the post.",
+        ),
+    ] = None,
+    post_type: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Optional post type value (e.g. LOTTERY, POLL).",
+        ),
+    ] = None,
+    visible_scope: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Optional visibility scope for the post.",
+        ),
+    ] = None,
+    prize_description: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Description of the prize for lottery posts.",
+        ),
+    ] = None,
+    prize_icon: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Icon URL for the lottery prize.",
+        ),
+    ] = None,
+    prize_count: Annotated[
+        int | None,
+        PydanticField(
+            default=None,
+            ge=1,
+            description="Total number of prizes available for lottery posts.",
+        ),
+    ] = None,
+    point_cost: Annotated[
+        int | None,
+        PydanticField(
+            default=None,
+            ge=0,
+            description="Point cost required to participate in the post, when applicable.",
+        ),
+    ] = None,
+    start_time: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="ISO 8601 start time for lottery or poll posts.",
+        ),
+    ] = None,
+    end_time: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="ISO 8601 end time for lottery or poll posts.",
+        ),
+    ] = None,
+    options: Annotated[
+        list[str] | None,
+        PydanticField(
+            default=None,
+            min_length=1,
+            description="Poll options when creating a poll post.",
+        ),
+    ] = None,
+    multiple: Annotated[
+        bool | None,
+        PydanticField(
+            default=None,
+            description="Whether the poll allows selecting multiple options.",
+        ),
+    ] = None,
+    proposed_name: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Proposed category name for suggestion posts.",
+        ),
+    ] = None,
+    proposal_description: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Supporting description for the proposed category.",
+        ),
+    ] = None,
+    captcha: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Captcha solution if the backend requires one to create posts.",
+        ),
+    ] = None,
+    ctx: Context | None = None,
+) -> PostCreateResult:
+    """Create a new post in OpenIsle and return the detailed backend payload."""
+
+    sanitized_title = title.strip()
+    if not sanitized_title:
+        raise ValueError("Post title must not be empty.")
+
+    sanitized_content = content.strip()
+    if not sanitized_content:
+        raise ValueError("Post content must not be empty.")
+
+    sanitized_category_id: int | None = None
+    if category_id is not None:
+        if isinstance(category_id, bool):
+            raise ValueError("Category identifier must be an integer, not a boolean.")
+        try:
+            sanitized_category_id = int(category_id)
+        except (TypeError, ValueError) as exc:
+            raise ValueError("Category identifier must be an integer.") from exc
+        if sanitized_category_id <= 0:
+            raise ValueError("Category identifier must be a positive integer.")
+    if sanitized_category_id is None:
+        raise ValueError("A category identifier is required to create a post.")
+
+    sanitized_tag_ids: list[int] | None = None
+    if tag_ids is not None:
+        sanitized_tag_ids = []
+        for value in tag_ids:
+            if isinstance(value, bool):
+                raise ValueError("Tag identifiers must be integers, not booleans.")
+            try:
+                converted = int(value)
+            except (TypeError, ValueError) as exc:
+                raise ValueError("Tag identifiers must be integers.") from exc
+            if converted <= 0:
+                raise ValueError("Tag identifiers must be positive integers.")
+            sanitized_tag_ids.append(converted)
+        if not sanitized_tag_ids:
+            sanitized_tag_ids = None
+    if not sanitized_tag_ids:
+        raise ValueError("At least one tag identifier is required to create a post.")
+    if len(sanitized_tag_ids) > 2:
+        raise ValueError("At most two tag identifiers can be provided for a post.")
+
+    sanitized_post_type = post_type.strip() if isinstance(post_type, str) else None
+    if sanitized_post_type == "":
+        sanitized_post_type = None
+
+    sanitized_visible_scope = (
+        visible_scope.strip() if isinstance(visible_scope, str) else None
+    )
+    if sanitized_visible_scope == "":
+        sanitized_visible_scope = None
+
+    sanitized_prize_description = (
+        prize_description.strip() if isinstance(prize_description, str) else None
+    )
+    if sanitized_prize_description == "":
+        sanitized_prize_description = None
+
+    sanitized_prize_icon = prize_icon.strip() if isinstance(prize_icon, str) else None
+    if sanitized_prize_icon == "":
+        sanitized_prize_icon = None
+
+    sanitized_prize_count: int | None = None
+    if prize_count is not None:
+        if isinstance(prize_count, bool):
+            raise ValueError("Prize count must be an integer, not a boolean.")
+        try:
+            sanitized_prize_count = int(prize_count)
+        except (TypeError, ValueError) as exc:
+            raise ValueError("Prize count must be an integer.") from exc
+        if sanitized_prize_count <= 0:
+            raise ValueError("Prize count must be a positive integer.")
+
+    sanitized_point_cost: int | None = None
+    if point_cost is not None:
+        if isinstance(point_cost, bool):
+            raise ValueError("Point cost must be an integer, not a boolean.")
+        try:
+            sanitized_point_cost = int(point_cost)
+        except (TypeError, ValueError) as exc:
+            raise ValueError("Point cost must be an integer.") from exc
+        if sanitized_point_cost < 0:
+            raise ValueError("Point cost cannot be negative.")
+
+    sanitized_start_time = start_time.strip() if isinstance(start_time, str) else None
+    if sanitized_start_time == "":
+        sanitized_start_time = None
+
+    sanitized_end_time = end_time.strip() if isinstance(end_time, str) else None
+    if sanitized_end_time == "":
+        sanitized_end_time = None
+
+    sanitized_options: list[str] | None = None
+    if options is not None:
+        sanitized_options = []
+        for option in options:
+            if option is None:
+                continue
+            stripped_option = option.strip()
+            if stripped_option:
+                sanitized_options.append(stripped_option)
+        if not sanitized_options:
+            sanitized_options = None
+
+    sanitized_multiple = bool(multiple) if isinstance(multiple, bool) else None
+
+    sanitized_proposed_name = (
+        proposed_name.strip() if isinstance(proposed_name, str) else None
+    )
+    if sanitized_proposed_name == "":
+        sanitized_proposed_name = None
+
+    sanitized_proposal_description = (
+        proposal_description.strip() if isinstance(proposal_description, str) else None
+    )
+    if sanitized_proposal_description == "":
+        sanitized_proposal_description = None
+
+    sanitized_captcha = captcha.strip() if isinstance(captcha, str) else None
+    if sanitized_captcha == "":
+        sanitized_captcha = None
+
+    payload: dict[str, object] = {
+        "title": sanitized_title,
+        "content": sanitized_content,
+    }
+    if sanitized_category_id is not None:
+        payload["categoryId"] = sanitized_category_id
+    if sanitized_tag_ids is not None:
+        payload["tagIds"] = sanitized_tag_ids
+    if sanitized_post_type is not None:
+        payload["type"] = sanitized_post_type
+    if sanitized_visible_scope is not None:
+        payload["postVisibleScopeType"] = sanitized_visible_scope
+    if sanitized_prize_description is not None:
+        payload["prizeDescription"] = sanitized_prize_description
+    if sanitized_prize_icon is not None:
+        payload["prizeIcon"] = sanitized_prize_icon
+    if sanitized_prize_count is not None:
+        payload["prizeCount"] = sanitized_prize_count
+    if sanitized_point_cost is not None:
+        payload["pointCost"] = sanitized_point_cost
+    if sanitized_start_time is not None:
+        payload["startTime"] = sanitized_start_time
+    if sanitized_end_time is not None:
+        payload["endTime"] = sanitized_end_time
+    if sanitized_options is not None:
+        payload["options"] = sanitized_options
+    if sanitized_multiple is not None:
+        payload["multiple"] = sanitized_multiple
+    if sanitized_proposed_name is not None:
+        payload["proposedName"] = sanitized_proposed_name
+    if sanitized_proposal_description is not None:
+        payload["proposalDescription"] = sanitized_proposal_description
+    if sanitized_captcha is not None:
+        payload["captcha"] = sanitized_captcha
+
+    try:
+        logger.info("Creating post with title='%s'", sanitized_title)
+        raw_post = await search_client.create_post(payload, token=_extract_authorization_token(ctx))
+    except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
+        status_code = exc.response.status_code
+        if status_code == 400:
+            message = (
+                "Post creation failed due to invalid input or captcha verification errors."
+            )
+        elif status_code == 401:
+            message = (
+                "Authentication failed while creating the post. Please verify the "
+                "Authorization header or configured token."
+            )
+        elif status_code == 403:
+            message = "The provided Authorization token is not authorized to create posts."
+        else:
+            message = (
+                "OpenIsle backend returned HTTP "
+                f"{status_code} while creating the post."
+            )
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+    except httpx.RequestError as exc:  # pragma: no cover - network errors
+        message = f"Unable to reach OpenIsle backend post service: {exc}."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    try:
+        post = PostDetail.model_validate(raw_post)
+    except ValidationError as exc:
+        message = "Received malformed data from the post creation endpoint."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    if ctx is not None:
+        await ctx.info(f"Post '{post.title}' created successfully.")
+    logger.debug(
+        "Validated created post payload with id=%s and title='%s'",
+        post.id,
+        post.title,
+    )
+
+    return PostCreateResult(post=post)
+
+
 @app.tool(
     name="recent_posts",
     description="Retrieve posts created in the last N minutes.",
@@ -198,6 +711,7 @@ async def recent_posts(
     """Fetch recent posts from the backend and return structured data."""
 
     try:
+        logger.info("Fetching recent posts for last %s minutes", minutes)
         raw_posts = await search_client.recent_posts(minutes)
     except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
         message = (
@@ -225,10 +739,219 @@ async def recent_posts(
         await ctx.info(
             f"Found {len(posts)} posts created within the last {minutes} minutes."
         )
+    logger.debug(
+        "Validated %d recent posts for window=%s minutes",
+        len(posts),
+        minutes,
+    )
 
     return RecentPostsResponse(minutes=minutes, total=len(posts), posts=posts)
 
 
+@app.tool(
+    name="get_post",
+    description="Retrieve detailed information for a single post.",
+    structured_output=True,
+)
+async def get_post(
+    post_id: Annotated[
+        int,
+        PydanticField(ge=1, description="Identifier of the post to retrieve."),
+    ],
+    ctx: Context | None = None,
+) -> PostDetail:
+    """Fetch post details from the backend and validate the response."""
+
+    try:
+        logger.info("Fetching post details for post_id=%s", post_id)
+        raw_post = await search_client.get_post(
+            post_id, _extract_authorization_token(ctx)
+        )
+    except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
+        status_code = exc.response.status_code
+        if status_code == 404:
+            message = f"Post {post_id} was not found."
+        elif status_code == 401:
+            message = "Authentication failed while retrieving the post."
+        elif status_code == 403:
+            message = "The provided Authorization token is not authorized to view this post."
+        else:
+            message = (
+                "OpenIsle backend returned HTTP "
+                f"{status_code} while retrieving post {post_id}."
+            )
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+    except httpx.RequestError as exc:  # pragma: no cover - network errors
+        message = f"Unable to reach OpenIsle backend post service: {exc}."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    try:
+        post = PostDetail.model_validate(raw_post)
+    except ValidationError as exc:
+        message = "Received malformed data from the post detail endpoint."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    if ctx is not None:
+        await ctx.info(f"Retrieved post {post_id} successfully.")
+    logger.debug(
+        "Validated post payload for post_id=%s with %d comments",
+        post_id,
+        len(post.comments),
+    )
+
+    return post
+
+
+@app.tool(
+    name="list_unread_messages",
+    description="List unread notification messages for the authenticated user.",
+    structured_output=True,
+)
+async def list_unread_messages(
+    page: Annotated[
+        int,
+        PydanticField(
+            default=0,
+            ge=0,
+            description="Page number of unread notifications to retrieve.",
+        ),
+    ] = 0,
+    size: Annotated[
+        int,
+        PydanticField(
+            default=30,
+            ge=1,
+            le=100,
+            description="Number of unread notifications to include per page.",
+        ),
+    ] = 30,
+    ctx: Context | None = None,
+) -> UnreadNotificationsResponse:
+    """Retrieve unread notifications and return structured data."""
+
+    try:
+        logger.info(
+            "Fetching unread notifications (page=%s, size=%s)",
+            page,
+            size,
+        )
+        raw_notifications = await search_client.list_unread_notifications(
+            page=page,
+            size=size,
+            token=_extract_authorization_token(ctx),
+        )
+    except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
+        message = (
+            "OpenIsle backend returned HTTP "
+            f"{exc.response.status_code} while fetching unread notifications."
+        )
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+    except httpx.RequestError as exc:  # pragma: no cover - network errors
+        message = f"Unable to reach OpenIsle backend notification service: {exc}."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    try:
+        notifications = [
+            NotificationData.model_validate(entry) for entry in raw_notifications
+        ]
+    except ValidationError as exc:
+        message = "Received malformed data from the unread notifications endpoint."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    total = len(notifications)
+    if ctx is not None:
+        await ctx.info(
+            f"Retrieved {total} unread notifications (page {page}, size {size})."
+        )
+    logger.debug(
+        "Validated %d unread notifications for page=%s size=%s",
+        total,
+        page,
+        size,
+    )
+
+    return UnreadNotificationsResponse(
+        page=page,
+        size=size,
+        total=total,
+        notifications=notifications,
+    )
+
+
+@app.tool(
+    name="mark_notifications_read",
+    description="Mark specific notification messages as read to remove them from the unread list.",
+    structured_output=True,
+)
+async def mark_notifications_read(
+    ids: Annotated[
+        list[int],
+        PydanticField(
+            min_length=1,
+            description="Notification identifiers that should be marked as read.",
+        ),
+    ],
+    ctx: Context | None = None,
+) -> NotificationCleanupResult:
+    """Mark the supplied notifications as read and report the processed identifiers."""
+
+    try:
+        logger.info(
+            "Marking %d notifications as read",  # pragma: no branch - logging
+            len(ids),
+        )
+        await search_client.mark_notifications_read(
+            ids, token=_extract_authorization_token(ctx)
+        )
+    except httpx.HTTPStatusError as exc:  # pragma: no cover - network errors
+        message = (
+            "OpenIsle backend returned HTTP "
+            f"{exc.response.status_code} while marking notifications as read."
+        )
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+    except httpx.RequestError as exc:  # pragma: no cover - network errors
+        message = f"Unable to reach OpenIsle backend notification service: {exc}."
+        if ctx is not None:
+            await ctx.error(message)
+        raise ValueError(message) from exc
+
+    processed_ids: list[int] = []
+    for value in ids:
+        if isinstance(value, bool):
+            raise ValueError("Notification identifiers must be integers, not booleans.")
+        converted = int(value)
+        if converted <= 0:
+            raise ValueError("Notification identifiers must be positive integers.")
+        processed_ids.append(converted)
+    if ctx is not None:
+        await ctx.info(
+            f"Marked {len(processed_ids)} notifications as read.",
+        )
+    logger.debug(
+        "Successfully marked notifications as read: ids=%s",
+        processed_ids,
+    )
+
+    return NotificationCleanupResult(
+        processed_ids=processed_ids,
+        total_marked=len(processed_ids),
+    )
+
+
 def main() -> None:
     """Run the MCP server using the configured transport."""
 

websocket_service/AGENTS.md

@@ -0,0 +1,48 @@
+# WebSocket Service 协作指引
+
+## 1) 适用范围
+
+- 作用于 `websocket_service/` 目录及其子目录。
+- 本服务是实时通知链路关键节点，改动需谨慎。
+
+## 2) 服务职责
+
+- 通过 STOMP 维护客户端实时连接（`/api/ws`、`/api/sockjs`）。
+- 从 RabbitMQ 队列消费通知并转发至用户/会话目的地。
+- 在连接阶段执行 JWT 鉴权。
+
+## 3) 关键一致性规则
+
+- JWT 密钥与后端保持一致（同一 `JWT_SECRET` 语义）。
+- 队列配置与后端分片策略同步：
+  - 后端声明：`backend/.../RabbitMQConfig.java`
+  - 后端分片：`backend/.../ShardingStrategy.java`
+  - 本服务监听：`src/main/java/com/openisle/websocket/listener/NotificationListener.java`
+- 监听队列当前约定：16 个十六进制分片队列 + 遗留 `notifications-queue`。
+
+## 4) 修改规则
+
+- 不随意变更 STOMP 目的地命名（`/topic/...`、`/user/...`）。
+- 若必须调整目的地，需同步前端 `frontend_nuxt/composables/useWebSocket.js` 与相关消费代码。
+- `WebSocketAuthInterceptor` 中 CONNECT 鉴权失败策略（拒绝连接）应保持清晰一致。
+- Allowed origins 改动需考虑本地、预发、正式环境域名。
+
+## 5) 配置与可观测性
+
+- 配置入口：`src/main/resources/application.properties`
+- 健康检查：`/actuator/health`（部署与 compose 依赖该路径）
+- 日志级别改动需避免在生产产生高噪声。
+
+## 6) 验证建议
+
+- 首选：`mvn test`
+- 若暂无测试覆盖：`mvn -DskipTests compile`
+- 变更消息推送逻辑时，至少完成一次端到端验证：
+  - 生产者发送消息
+  - RabbitMQ 消费成功
+  - 客户端收到对应目的地消息
+
+## 7) 输出要求
+
+- 说明是否影响队列名、路由键、目的地或鉴权逻辑。
+- 说明是否需要前端/后端同步改动。