docs: update install cmd

feat: workspace volume in docker deploy
feat: release 2.0.3
2026-06-03 02:27:09 +08:00 · 2026-03-18 16:11:35 +08:00 · 2026-03-18 16:03:03 +08:00 · 2026-03-18 15:40:49 +08:00 · 2026-03-18 15:10:21 +08:00 · 2026-03-18 14:55:29 +08:00
257 changed files with 19786 additions and 11455 deletions
--- a/.github/ISSUE_TEMPLATE/1.bug.yml
+++ b/.github/ISSUE_TEMPLATE/1.bug.yml
@@ -79,8 +79,6 @@ body:
      description: |
        请确保你正确配置了该`channel`所需的配置项，所有可选的配置项都写在了[该文件中](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/config.py)，请将所需配置项填写在根目录下的`config.json`文件中。
      options:
-        - wx(个人微信, itchat)
-        - wxy(个人微信, wechaty)
        - wechatmp(公众号, 订阅号)
        - wechatmp_service(公众号, 服务号)
        - terminal
--- a/.gitignore
+++ b/.gitignore
@@ -3,16 +3,15 @@
 .vscode
 .venv
 .vs
-.wechaty/
 __pycache__/
 venv*
 *.pyc
+python
 config.json
 QR.png
 nohup.out
 tmp
 plugins.json
-itchat.pkl
 *.log
 logs/
 workspace
--- a/README.md
+++ b/README.md
@@ -1,37 +1,47 @@
 <p align="center"><img src= "https://github.com/user-attachments/assets/eca9a9ec-8534-4615-9e0f-96c5ac1d10a3" alt="Chatgpt-on-Wechat" width="550" /></p>

 <p align="center">
-   <a href="https://github.com/zhayujie/chatgpt-on-wechat/releases/latest"><img src="https://img.shields.io/github/v/release/zhayujie/chatgpt-on-wechat" alt="Latest release"></a>
+  <a href="https://github.com/zhayujie/chatgpt-on-wechat/releases/latest"><img src="https://img.shields.io/github/v/release/zhayujie/chatgpt-on-wechat" alt="Latest release"></a>
  <a href="https://github.com/zhayujie/chatgpt-on-wechat/blob/master/LICENSE"><img src="https://img.shields.io/github/license/zhayujie/chatgpt-on-wechat" alt="License: MIT"></a>
  <a href="https://github.com/zhayujie/chatgpt-on-wechat"><img src="https://img.shields.io/github/stars/zhayujie/chatgpt-on-wechat?style=flat-square" alt="Stars"></a> <br/>
+  [中文] | [<a href="docs/en/README.md">English</a>]
+</p>
+
+**CowAgent** 是基于大模型的超级AI助理，能够主动思考和任务规划、操作计算机和外部资源、创造和执行Skills、拥有长期记忆并不断成长。CowAgent 支持灵活切换多种模型，能处理文本、语音、图片、文件等多模态消息，可接入网页、飞书、钉钉、企微智能机器人、QQ、企微自建应用、微信公众号中使用，7*24小时运行于你的个人电脑或服务器中。
+
+<p align="center">
+  <a href="https://cowagent.ai/">🌐 官网</a> &nbsp;·&nbsp;
+  <a href="https://docs.cowagent.ai/">📖 文档中心</a> &nbsp;·&nbsp;
+  <a href="https://docs.cowagent.ai/guide/quick-start">🚀 快速开始</a> &nbsp;·&nbsp;
+  <a href="https://link-ai.tech/cowagent/create">☁️ 在线体验</a>
 </p>

-**CowAgent** 是基于大模型的超级AI助理，能够主动思考和任务规划、操作计算机和外部资源、创造和执行Skills、拥有长期记忆并不断成长。CowAgent 支持灵活切换多种模型，能处理文本、语音、图片、文件等多模态消息，可接入网页、飞书、钉钉、企业微信应用、微信公众号中使用，7*24小时运行于你的个人电脑或服务器中。

-📖能力介绍：[CowAgent 2.0](/docs/agent.md)

 # 简介

-> 该项目既是一个可以开箱即用的超级AI助理，也是一个支持高FTS5 not available, using LIKE-based keyword searc度扩展的Agent框架，可以通过为项目扩展大模型接口、接入渠道、内置工具、Skills系统来灵活实现各种定制需求。核心能力如下：
+> 该项目既是一个可以开箱即用的超级AI助理，也是一个支持高扩展的Agent框架，可以通过为项目扩展大模型接口、接入渠道、内置工具、Skills系统来灵活实现各种定制需求。核心能力如下：

 -  ✅  **复杂任务规划**：能够理解复杂任务并自主规划执行，持续思考和调用工具直到完成目标，支持通过工具操作访问文件、终端、浏览器、定时任务等系统资源
 -  ✅  **长期记忆：** 自动将对话记忆持久化至本地文件和数据库中，包括全局记忆和天级记忆，支持关键词及向量检索
 -  ✅  **技能系统：** 实现了Skills创建和运行的引擎，内置多种技能，并支持通过自然语言对话完成自定义Skills开发
 -  ✅  **多模态消息：** 支持对文本、图片、语音、文件等多类型消息进行解析、处理、生成、发送等操作
-  ✅  **多模型接入：** 支持OpenAI, Claude, Gemini, DeepSeek, MiniMax、GLM、通义千问, Kimi等国内外主流模型厂商
-  ✅  **多端部署：** 支持运行在本地计算机或服务器，可集成到网页、飞书、钉钉、微信公众号、企业微信应用中使用
-  ✅  **知识库：** 集成企业知识库能力，让Agent成为专属数字员工，基于[LinkAI](https://link-ai.tech)平台实现
+-  ✅  **多模型接入：** 支持OpenAI, Claude, Gemini, DeepSeek, MiniMax、GLM、Qwen、Kimi、Doubao等国内外主流模型厂商
+-  ✅  **多端部署：** 支持运行在本地计算机或服务器，可集成到飞书、钉钉、企业微信、QQ、微信公众号、网页中使用

 ## 声明

-1. 本项目遵循 [MIT开源协议](/LICENSE)，主要用于技术研究和学习，使用本项目时需遵守所在地法律法规、相关政策以及企业章程，禁止用于任何违法或侵犯他人权益的行为。任何个人、团队和企业，无论以何种方式使用该项目、对何对象提供服务，所产生的一切后果，本项目均不承担任何责任
-2. 成本与安全：Agent模式下Token使用量高于普通对话模式，请根据效果及成本综合选择模型。Agent具有访问所在操作系统的能力，请谨慎选择项目部署环境。同时项目也会持续升级安全机制、并降低模型消耗成本
+1. 本项目遵循 [MIT开源协议](/LICENSE)，主要用于技术研究和学习，使用本项目时需遵守所在地法律法规、相关政策以及企业章程，禁止用于任何违法或侵犯他人权益的行为。任何个人、团队和企业，无论以何种方式使用该项目、对何对象提供服务，所产生的一切后果，本项目均不承担任何责任。
+2. 成本与安全：Agent模式下Token使用量高于普通对话模式，请根据效果及成本综合选择模型。Agent具有访问所在操作系统的能力，请谨慎选择项目部署环境。同时项目也会持续升级安全机制、并降低模型消耗成本。
+3. CowAgent项目专注于开源技术开发，不会参与、授权或发行任何加密货币。

 ## 演示

-使用说明(Agent模式)：[CowAgent介绍](/docs/agent.md)
+- 使用说明(Agent模式)：[CowAgent介绍](https://docs.cowagent.ai/intro/features)

-DEMO视频(对话模式)：https://cdn.link-ai.tech/doc/cow_demo.mp4
+- 免部署在线体验：[CowAgent](https://link-ai.tech/cowagent/create)
+
+- DEMO视频(对话模式)：https://cdn.link-ai.tech/doc/cow_demo.mp4

 ## 社区

@@ -43,9 +53,9 @@ DEMO视频(对话模式)：https://cdn.link-ai.tech/doc/cow_demo.mp4

 # 企业服务

-<a href="https://link-ai.tech" target="_blank"><img width="720" src="https://cdn.link-ai.tech/image/link-ai-intro.jpg"></a>
+<a href="https://link-ai.tech" target="_blank"><img width="650" src="https://cdn.link-ai.tech/image/link-ai-intro.jpg"></a>

-> [LinkAI](https://link-ai.tech/) 是面向企业和开发者的一站式AI智能体平台，聚合多模态大模型、知识库、Agent 插件、工作流等能力，支持一键接入主流平台并进行管理，支持SaaS、私有化部署等多种模式。
+> [LinkAI](https://link-ai.tech/) 是面向企业和个人的一站式AI智能体平台，聚合多模态大模型、知识库、技能、工作流等能力，支持一键接入主流平台并管理，支持SaaS、私有化部署等多种模式，可免部署在线运行[CowAgent助理](https://link-ai.tech/cowagent/create)。
 >
 > LinkAI 目前已在智能客服、私域运营、企业效率助手等场景积累了丰富的AI解决方案，在消费、健康、文教、科技制造等各行业沉淀了大模型落地应用的最佳实践，致力于帮助更多企业和开发者拥抱 AI 生产力。

@@ -57,17 +67,19 @@ DEMO视频(对话模式)：https://cdn.link-ai.tech/doc/cow_demo.mp4

 # 🏷 更新日志

+>**2026.03.18：** [2.0.3版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.3)，新增企微智能机器人和 QQ 通道、支持Coding Plan、新增多个模型、Web端文件处理、记忆系统升级。
+
+>**2026.02.27：** [2.0.2版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.2)，Web 控制台全面升级（流式对话、模型/技能/记忆/通道/定时任务/日志管理）、支持多通道同时运行、会话持久化存储、新增多个模型。
+
+>**2026.02.13：** [2.0.1版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.1)，内置 Web Search 工具、智能上下文裁剪策略、运行时信息动态更新、Windows 兼容性适配，修复定时任务记忆丢失、飞书连接等多项问题。
+
 >**2026.02.03：** [2.0.0版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.0)，正式升级为超级Agent助理，支持多轮任务决策、具备长期记忆、实现多种系统工具、支持Skills框架，新增多种模型并优化了接入渠道。

 >**2025.05.23：** [1.7.6版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.6) 优化web网页channel、新增 [AgentMesh](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/plugins/agent/README.md)多智能体插件、百度语音合成优化、企微应用`access_token`获取优化、支持`claude-4-sonnet`和`claude-4-opus`模型

 >**2025.04.11：** [1.7.5版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.5) 新增支持 [wechatferry](https://github.com/zhayujie/chatgpt-on-wechat/pull/2562) 协议、新增 deepseek 模型、新增支持腾讯云语音能力、新增支持 ModelScope 和 Gitee-AI API接口

->**2024.12.13：** [1.7.4版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.4) 新增 Gemini 2.0 模型、新增web channel、解决内存泄漏问题、解决 `#reloadp` 命令重载不生效问题
-
->**2024.10.31：** [1.7.3版本](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.3) 程序稳定性提升、数据库功能、Claude模型优化、linkai插件优化、离线通知
-
-更多更新历史请查看: [更新日志](/docs/release/history.md)
+更多更新历史请查看: [更新日志](https://docs.cowagent.ai/releases)

 <br/>

@@ -78,10 +90,10 @@ DEMO视频(对话模式)：https://cdn.link-ai.tech/doc/cow_demo.mp4
 在终端执行以下命令：

 ```bash
-bash <(curl -sS https://cdn.link-ai.tech/code/cow/run.sh)
+bash <(curl -fsSL https://cdn.link-ai.tech/code/cow/run.sh)
 ```

-脚本使用说明：[一键运行脚本](https://github.com/zhayujie/chatgpt-on-wechat/wiki/CowAgentQuickStart)
+脚本使用说明：[一键运行脚本](https://docs.cowagent.ai/guide/quick-start)


 ## 一、准备
@@ -90,9 +102,9 @@ bash <(curl -sS https://cdn.link-ai.tech/code/cow/run.sh)

 项目支持国内外主流厂商的模型接口，可选模型及配置说明参考：[模型说明](#模型说明)。

-> 注：Agent模式下推荐使用以下模型，可根据效果及成本综合选择： Claude(claude-sonnet-4-5、claude-sonnet-4-0)、Gemini(gemini-3-flash-preview、gemini-3-pro-preview)、GLM(glm-4.7)、MiniMAx(MiniMax-M2.1)、Qwen(qwen3-max)
+> 注：Agent模式下推荐使用以下模型，可根据效果及成本综合选择：MiniMax-M2.5、glm-5、kimi-k2.5、qwen3.5-plus、claude-sonnet-4-6、gemini-3.1-pro-preview、gpt-5.4、gpt-5.4-mini

-同时支持使用 **LinkAI平台** 接口，可灵活切换 OpenAI、Claude、Gemini、DeepSeek、Qwen、Kimi 等多种常用模型，并支持知识库、工作流、插件等Agent能力，参考 [接口文档](https://docs.link-ai.tech/platform/api)。
+同时支持使用 **LinkAI平台** 接口，支持上述全部模型，并支持知识库、工作流、插件等Agent技能，参考 [接口文档](https://docs.link-ai.tech/platform/api)。

 ### 2.环境安装

@@ -135,23 +147,25 @@ pip3 install -r requirements-optional.txt
 ```bash
 # config.json 文件内容示例
 {
-  "channel_type": "web",                                      # 接入渠道类型，默认为web，支持修改为:feishu,dingtalk,wechatcom_app,terminal,wechatmp,wechatmp_service
-  "model": "claude-sonnet-4-5",                               # 模型名称
+  "channel_type": "web",                                      # 接入渠道类型，默认为web，支持修改为:feishu,dingtalk,wecom_bot,qq,wechatcom_app,wechatmp_service,wechatmp,terminal
+  "model": "MiniMax-M2.5",                                    # 模型名称
+  "minimax_api_key": "",                                      # MiniMax API Key
+  "zhipu_ai_api_key": "",                                     # 智谱GLM API Key
+  "moonshot_api_key": "",                                     # Kimi/Moonshot API Key
+  "ark_api_key": "",                                          # 豆包(火山方舟) API Key
+  "dashscope_api_key": "",                                    # 百炼(通义千问)API Key
  "claude_api_key": "",                                       # Claude API Key
  "claude_api_base": "https://api.anthropic.com/v1",          # Claude API 地址，修改可接入三方代理平台
-  "open_ai_api_key": "",                                      # OpenAI API Key
-  "open_ai_api_base": "https://api.openai.com/v1",            # OpenAI API 地址
  "gemini_api_key": "",                                       # Gemini API Key
  "gemini_api_base": "https://generativelanguage.googleapis.com", # Gemini API地址
-  "zhipu_ai_api_key": "",                                     # 智谱GLM API Key
-  "minimax_api_key": "",                                      # MiniMax API Key
-  "dashscope_api_key": "",                                    # 百炼(通义千问)API Key
+  "open_ai_api_key": "",                                      # OpenAI API Key
+  "open_ai_api_base": "https://api.openai.com/v1",            # OpenAI API 地址
  "linkai_api_key": "",                                       # LinkAI API Key
  "proxy": "",                                                # 代理客户端的ip和端口，国内环境需要开启代理的可填写该项，如 "127.0.0.1:7890"
  "speech_recognition": false,                                # 是否开启语音识别
  "group_speech_recognition": false,                          # 是否开启群组语音识别
  "voice_reply_voice": false,                                 # 是否使用语音回复语音
-  "use_linkai": false,                                        # 是否使用LinkAI接口，默认关闭，设置为true后可对接LinkAI平台接口
+  "use_linkai": false,                                        # 是否使用LinkAI接口，默认关闭，设置为true后可对接LinkAI平台模型
  "agent": true,                                              # 是否启用Agent模式，启用后拥有多轮工具决策、长期记忆、Skills能力等
  "agent_workspace": "~/cow",                                 # Agent的工作空间路径，用于存储memory、skills、系统设定等
  "agent_max_context_tokens": 40000,                          # Agent模式下最大上下文tokens，超出将自动丢弃最早的上下文
@@ -173,17 +187,16 @@ pip3 install -r requirements-optional.txt
 <details>
 <summary>2. 其他配置</summary>

-+ `model`: 模型名称，Agent模式下推荐使用 `claude-sonnet-4-5`、`claude-sonnet-4-0`、`gemini-3-flash-preview`、`gemini-3-pro-preview`、`glm-4.7`、`MiniMax-M2.1`、`qwen3-max`，全部模型名称参考[common/const.py](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/common/const.py)文件
+ `model`: 模型名称，Agent模式下推荐使用 `MiniMax-M2.5`、`glm-5`、`kimi-k2.5`、`qwen3.5-plus`、`claude-sonnet-4-6`、`gemini-3.1-pro-preview`，全部模型名称参考[common/const.py](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/common/const.py)文件
 + `character_desc`：普通对话模式下的机器人系统提示词。在Agent模式下该配置不生效，由工作空间中的文件内容构成。
 + `subscribe_msg`：订阅消息，公众号和企业微信channel中请填写，当被订阅时会自动回复， 可使用特殊占位符。目前支持的占位符有{trigger_prefix}，在程序中它会自动替换成bot的触发词。
 </details>

 <details>
-<summary>5. LinkAI配置</summary>
+<summary>3. LinkAI配置</summary>

-+ `use_linkai`: 是否使用LinkAI接口，默认关闭，设置为true后可对接LinkAI平台，使用知识库、工作流、插件等能力, 参考[接口文档](https://docs.link-ai.tech/platform/api/chat)
+ `use_linkai`: 是否使用LinkAI接口，默认关闭，设置为true后可对接LinkAI平台，使用模型、知识库、工作流、插件等技能, 参考[接口文档](https://docs.link-ai.tech/platform/api/chat)
 + `linkai_api_key`: LinkAI Api Key，可在 [控制台](https://link-ai.tech/console/interface) 创建
-+ `linkai_app_code`: LinkAI 应用或工作流的code，选填，普通对话模式中使用。
 </details>

 注：全部配置项说明可在 [`config.py`](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/config.py) 文件中查看。
@@ -213,8 +226,9 @@ nohup python3 app.py & tail -f nohup.out

 执行后程序运行于服务器后台，可通过 `ctrl+c` 关闭日志，不会影响后台程序的运行。使用 `ps -ef | grep app.py | grep -v grep` 命令可查看运行于后台的进程，如果想要重新启动程序可以先 `kill` 掉对应的进程。 日志关闭后如果想要再次打开只需输入 `tail -f nohup.out`。 

-此外，项目的 `scripts` 目录下有一键运行、关闭程序的脚本供使用。 运行后默认channel为web，通过可以通过修改配置文件进行切换。
+此外，项目根目录下的 `run.sh` 脚本支持一键启动和管理服务，包括 `./run.sh start`、`./run.sh stop`、`./run.sh restart`、`./run.sh logs` 等命令，执行 `./run.sh help` 可查看全部用法。

+> 如果需要通过浏览器访问Web控制台，请确保服务器的 `9899` 端口已在防火墙或安全组中放行，建议仅对指定IP开放以保证安全。

 ### 3.Docker部署

@@ -225,7 +239,7 @@ nohup python3 app.py & tail -f nohup.out
 **(1) 下载 docker-compose.yml 文件**

 ```bash
-wget https://cdn.link-ai.tech/code/cow/docker-compose.yml
+curl -O https://cdn.link-ai.tech/code/cow/docker-compose.yml
 ```

 下载完成后打开 `docker-compose.yml` 填写所需配置，例如 `CHANNEL_TYPE`、`OPEN_AI_API_KEY` 和等配置。
@@ -244,17 +258,7 @@ sudo docker compose up -d         # 若docker-compose为 1.X 版本，则执行
 sudo docker logs -f chatgpt-on-wechat
 ```

-**(3) 插件使用**
-
-如果需要在docker容器中修改插件配置，可通过挂载的方式完成，将 [插件配置文件](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/plugins/config.json.template)
-重命名为 `config.json`，放置于 `docker-compose.yml` 相同目录下，并在 `docker-compose.yml` 中的 `chatgpt-on-wechat` 部分下添加 `volumes` 映射:
-
-```
-volumes:
-  - ./config.json:/app/plugins/config.json
-```
-**注**：使用docker方式部署的详细教程可以参考：[docker部署CoW项目](https://www.wangpc.cc/ai/docker-deploy-cow/)
-
+> 如果需要通过浏览器访问Web控制台，请确保服务器的 `9899` 端口已在防火墙或安全组中放行，建议仅对指定IP开放以保证安全。

 ## 模型说明

@@ -269,16 +273,16 @@ volumes:

 ```json
 {
-    "model": "gpt-4.1-mini",
+    "model": "gpt-5.4",
    "open_ai_api_key": "YOUR_API_KEY",
    "open_ai_api_base": "https://api.openai.com/v1",
-    "bot_type": "chatGPT"
+    "bot_type": "openai"
 }
 ```

- - `model`: 与OpenAI接口的 [model参数](https://platform.openai.com/docs/models) 一致，支持包括 o系列、gpt-5.2、gpt-5.1、gpt-4.1等系列模型
+ - `model`: 与OpenAI接口的 [model参数](https://platform.openai.com/docs/models) 一致，支持包括 gpt-5.4、gpt-5.4-mini、gpt-5.4-nano、o系列、gpt-4.1等模型，Agent模式推荐使用 `gpt-5.4`、`gpt-5.4-mini`
 - `open_ai_api_base`: 如果需要接入第三方代理接口，可通过修改该参数进行接入
- - `bot_type`: 使用OpenAI相关模型时无需填写。当使用第三方代理接口接入Claude等非OpenAI官方模型时，该参数设为 `chatGPT`
+ - `bot_type`: 使用OpenAI相关模型时无需填写。当使用第三方代理接口接入Claude等非OpenAI官方模型时，该参数设为 `openai`
 </details>

 <details>
@@ -290,16 +294,149 @@ volumes:

 ```json
 {
+    "model": "gpt-5.4-mini",
    "use_linkai": true,
-    "linkai_api_key": "YOUR API KEY",
-    "linkai_app_code": "YOUR APP CODE"
+    "linkai_api_key": "YOUR API KEY"
 }
 ```

-+ `use_linkai`: 是否使用LinkAI接口，默认关闭，设置为true后可对接LinkAI平台的智能体，使用知识库、工作流、数据库、MCP插件等丰富的Agent能力
+ `use_linkai`: 是否使用LinkAI接口，默认关闭，设置为true后可对接LinkAI平台的模型，并使用知识库、工作流、数据库、插件等丰富的Agent技能
 + `linkai_api_key`: LinkAI平台的API Key，可在 [控制台](https://link-ai.tech/console/interface) 中创建
-+ `linkai_app_code`: LinkAI智能体 (应用或工作流) 的code，选填，普通对话模式可用。智能体创建可参考 [说明文档](https://docs.link-ai.tech/platform/quick-start)
-+ `model`: model字段填写空则直接使用智能体的模型，可在平台中灵活切换，[模型列表](https://link-ai.tech/console/models)中的全部模型均可使用
+ `model`: [模型列表](https://link-ai.tech/console/models)中的全部模型均可使用
+</details>
+
+<details>
+<summary>MiniMax</summary>
+
+方式一：官方接入，配置如下(推荐)：
+
+```json
+{
+    "model": "MiniMax-M2.5",
+    "minimax_api_key": ""
+}
+```
+ - `model`: 可填写 `MiniMax-M2.5、MiniMax-M2.1、MiniMax-M2.1-lightning、MiniMax-M2、abab6.5-chat` 等
+ - `minimax_api_key`：MiniMax平台的API-KEY，在 [控制台](https://platform.minimaxi.com/user-center/basic-information/interface-key) 创建
+
+方式二：OpenAI兼容方式接入，配置如下：
+```json
+{
+  "bot_type": "openai",
+  "model": "MiniMax-M2.5",
+  "open_ai_api_base": "https://api.minimaxi.com/v1",
+  "open_ai_api_key": ""
+}
+```
+- `bot_type`: OpenAI兼容方式
+- `model`: 可填 `MiniMax-M2.5、MiniMax-M2.1、MiniMax-M2.1-lightning、MiniMax-M2`，参考[API文档](https://platform.minimaxi.com/document/%E5%AF%B9%E8%AF%9D?key=66701d281d57f38758d581d0#QklxsNSbaf6kM4j6wjO5eEek)
+- `open_ai_api_base`: MiniMax平台API的 BASE URL
+- `open_ai_api_key`: MiniMax平台的API-KEY
+</details>
+
+<details>
+<summary>智谱AI (GLM)</summary>
+
+方式一：官方接入，配置如下(推荐)：
+
+```json
+{
+  "model": "glm-5",
+  "zhipu_ai_api_key": ""
+}
+```
+ - `model`: 可填 `glm-5、glm-4.7、glm-4-plus、glm-4-flash、glm-4-air、glm-4-airx、glm-4-long` 等, 参考 [glm系列模型编码](https://bigmodel.cn/dev/api/normal-model/glm-4)
+ - `zhipu_ai_api_key`: 智谱AI平台的 API KEY，在 [控制台](https://www.bigmodel.cn/usercenter/proj-mgmt/apikeys) 创建
+
+方式二：OpenAI兼容方式接入，配置如下：
+```json
+{
+  "bot_type": "openai",
+  "model": "glm-5",
+  "open_ai_api_base": "https://open.bigmodel.cn/api/paas/v4",
+  "open_ai_api_key": ""
+}
+```
+- `bot_type`: OpenAI兼容方式
+- `model`: 可填 `glm-5、glm-4.7、glm-4-plus、glm-4-flash、glm-4-air、glm-4-airx、glm-4-long` 等
+- `open_ai_api_base`: 智谱AI平台的 BASE URL
+- `open_ai_api_key`: 智谱AI平台的 API KEY
+</details>
+
+<details>
+<summary>通义千问 (Qwen)</summary>
+
+方式一：官方SDK接入，配置如下(推荐)：
+
+```json
+{
+    "model": "qwen3.5-plus",
+    "dashscope_api_key": "sk-qVxxxxG"
+}
+```
+ - `model`: 可填写 `qwen3.5-plus、qwen3-max、qwen-max、qwen-plus、qwen-turbo、qwen-long、qwq-plus` 等
+ - `dashscope_api_key`: 通义千问的 API-KEY，参考 [官方文档](https://bailian.console.aliyun.com/?tab=api#/api) ，在 [控制台](https://bailian.console.aliyun.com/?tab=model#/api-key) 创建
+
+方式二：OpenAI兼容方式接入，配置如下：
+```json
+{
+  "bot_type": "openai",
+  "model": "qwen3.5-plus",
+  "open_ai_api_base": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+  "open_ai_api_key": "sk-qVxxxxG"
+}
+```
+- `bot_type`: OpenAI兼容方式
+- `model`: 支持官方所有模型，参考[模型列表](https://help.aliyun.com/zh/model-studio/models?spm=a2c4g.11186623.0.0.78d84823Kth5on#9f8890ce29g5u)
+- `open_ai_api_base`: 通义千问API的 BASE URL
+- `open_ai_api_key`: 通义千问的 API-KEY
+</details>
+
+<details>
+<summary>Kimi (Moonshot)</summary>
+
+方式一：官方接入，配置如下：
+
+```json
+{
+    "model": "kimi-k2.5",
+    "moonshot_api_key": ""
+}
+```
+ - `model`: 可填写 `kimi-k2.5、kimi-k2、moonshot-v1-8k、moonshot-v1-32k、moonshot-v1-128k`
+ - `moonshot_api_key`: Moonshot的API-KEY，在 [控制台](https://platform.moonshot.cn/console/api-keys) 创建
+ 
+方式二：OpenAI兼容方式接入，配置如下：
+```json
+{
+  "bot_type": "openai",
+  "model": "kimi-k2.5",
+  "open_ai_api_base": "https://api.moonshot.cn/v1",
+  "open_ai_api_key": ""
+}
+```
+- `bot_type`: OpenAI兼容方式
+- `model`: 可填写 `kimi-k2.5、kimi-k2、moonshot-v1-8k、moonshot-v1-32k、moonshot-v1-128k`
+- `open_ai_api_base`: Moonshot的 BASE URL
+- `open_ai_api_key`: Moonshot的 API-KEY
+</details>
+
+<details>
+<summary>豆包 (Doubao)</summary>
+
+1. API Key创建：在 [火山方舟控制台](https://console.volcengine.com/ark/region:ark+cn-beijing/apikey) 创建API Key
+
+2. 填写配置
+
+```json
+{
+    "model": "doubao-seed-2-0-code-preview-260215",
+    "ark_api_key": "YOUR_API_KEY"
+}
+```
+ - `model`: 可填写 `doubao-seed-2-0-code-preview-260215、doubao-seed-2-0-pro-260215、doubao-seed-2-0-lite-260215、doubao-seed-2-0-mini-260215` 等
+ - `ark_api_key`: 火山方舟平台的 API Key，在 [控制台](https://console.volcengine.com/ark/region:ark+cn-beijing/apikey) 创建
+ - `ark_base_url`: 可选，默认为 `https://ark.cn-beijing.volces.com/api/v3`
 </details>

 <details>
@@ -311,11 +448,11 @@ volumes:

 ```json
 {
-    "model": "claude-sonnet-4-5",
+    "model": "claude-sonnet-4-6",
    "claude_api_key": "YOUR_API_KEY"
 }
 ```
- - `model`: 参考 [官方模型ID](https://docs.anthropic.com/en/docs/about-claude/models/overview#model-aliases) ，支持 `claude-sonnet-4-5、claude-sonnet-4-0、claude-opus-4-0、claude-3-5-sonnet-latest` 等
+ - `model`: 参考 [官方模型ID](https://docs.anthropic.com/en/docs/about-claude/models/overview#model-aliases) ，支持 `claude-sonnet-4-6、claude-opus-4-6、claude-sonnet-4-5、claude-sonnet-4-0、claude-opus-4-0、claude-3-5-sonnet-latest` 等
 </details>

 <details>
@@ -324,11 +461,11 @@ volumes:
 API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn) 创建API Key ，配置如下
 ```json
 {
-    "model": "gemini-3-flash-preview",
+    "model": "gemini-3.1-flash-lite-preview",
    "gemini_api_key": ""
 }
 ```
- - `model`: 参考[官方文档-模型列表](https://ai.google.dev/gemini-api/docs/models?hl=zh-cn)，支持 `gemini-3-flash-preview、gemini-3-pro-preview、gemini-2.5-pro、gemini-2.0-flash` 等
+ - `model`: 参考[官方文档-模型列表](https://ai.google.dev/gemini-api/docs/models?hl=zh-cn)，支持 `gemini-3.1-flash-lite-preview、gemini-3.1-pro-preview、gemini-3-flash-preview、gemini-3-pro-preview` 等
 </details>

 <details>
@@ -342,8 +479,8 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 {
    "model": "deepseek-chat",
    "open_ai_api_key": "sk-xxxxxxxxxxx",
-    "open_ai_api_base": "https://api.deepseek.com/v1", 
-    "bot_type": "chatGPT"
+    "open_ai_api_base": "https://api.deepseek.com/v1",
+    "bot_type": "openai"

 }
 ```
@@ -354,122 +491,6 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 - `open_ai_api_base`: DeepSeek平台 BASE URL
 </details>

-<details>
-<summary>通义千问 (Qwen)</summary>
-
-方式一：官方SDK接入，配置如下(推荐)：
-
-```json
-{
-    "model": "qwen3-max",
-    "dashscope_api_key": "sk-qVxxxxG"
-}
-```
- - `model`: 可填写 `qwen3-max、qwen-max、qwen-plus、qwen-turbo、qwen-long、qwq-plus` 等
- - `dashscope_api_key`: 通义千问的 API-KEY，参考 [官方文档](https://bailian.console.aliyun.com/?tab=api#/api) ，在 [控制台](https://bailian.console.aliyun.com/?tab=model#/api-key) 创建
- 
-方式二：OpenAI兼容方式接入，配置如下：
-```json
-{
-  "bot_type": "chatGPT",
-  "model": "qwen3-max",
-  "open_ai_api_base": "https://dashscope.aliyuncs.com/compatible-mode/v1",
-  "open_ai_api_key": "sk-qVxxxxG"
-}
-```
- `bot_type`: OpenAI兼容方式
- `model`: 支持官方所有模型，参考[模型列表](https://help.aliyun.com/zh/model-studio/models?spm=a2c4g.11186623.0.0.78d84823Kth5on#9f8890ce29g5u)
- `open_ai_api_base`: 通义千问API的 BASE URL
- `open_ai_api_key`: 通义千问的 API-KEY
-</details>
-
-<details>
-<summary>MiniMax</summary>
-
-方式一：官方接入，配置如下(推荐)：
-
-```json
-{
-    "model": "MiniMax-M2.1",
-    "minimax_api_key": ""
-}
-```
- - `model`: 可填写 `MiniMax-M2.1、MiniMax-M2.1-lightning、MiniMax-M2、abab6.5-chat` 等
- - `minimax_api_key`：MiniMax平台的API-KEY，在 [控制台](https://platform.minimaxi.com/user-center/basic-information/interface-key) 创建
- 
-方式二：OpenAI兼容方式接入，配置如下：
-```json
-{
-  "bot_type": "chatGPT",
-  "model": "MiniMax-M2.1",
-  "open_ai_api_base": "https://api.minimaxi.com/v1",
-  "open_ai_api_key": ""
-}
-```
- `bot_type`: OpenAI兼容方式
- `model`: 可填 `MiniMax-M2.1、MiniMax-M2.1-lightning、MiniMax-M2`，参考[API文档](https://platform.minimaxi.com/document/%E5%AF%B9%E8%AF%9D?key=66701d281d57f38758d581d0#QklxsNSbaf6kM4j6wjO5eEek)
- `open_ai_api_base`: MiniMax平台API的 BASE URL
- `open_ai_api_key`: MiniMax平台的API-KEY
-</details>
-
-<details>
-<summary>智谱AI (GLM)</summary>
-
-方式一：官方接入，配置如下(推荐)：
-
-```json
-{
-  "model": "glm-4.7",
-  "zhipu_ai_api_key": ""
-}
-```
- - `model`: 可填 `glm-4.7、glm-4-plus、glm-4-flash、glm-4-air、glm-4-airx、glm-4-long` 等, 参考 [glm-4系列模型编码](https://bigmodel.cn/dev/api/normal-model/glm-4)
- - `zhipu_ai_api_key`: 智谱AI平台的 API KEY，在 [控制台](https://www.bigmodel.cn/usercenter/proj-mgmt/apikeys) 创建
- 
-方式二：OpenAI兼容方式接入，配置如下：
-```json
-{
-  "bot_type": "chatGPT",
-  "model": "glm-4.7",
-  "open_ai_api_base": "https://open.bigmodel.cn/api/paas/v4",
-  "open_ai_api_key": ""
-}
-```
- `bot_type`: OpenAI兼容方式
- `model`: 可填 `glm-4.7、glm-4.6、glm-4-plus、glm-4-flash、glm-4-air、glm-4-airx、glm-4-long` 等
- `open_ai_api_base`: 智谱AI平台的 BASE URL
- `open_ai_api_key`: 智谱AI平台的 API KEY
-</details>
-
-<details>
-<summary>Kimi (Moonshot)</summary>
-
-方式一：官方接入，配置如下：
-
-```json
-{
-    "model": "moonshot-v1-128k",
-    "moonshot_api_key": ""
-}
-```
- - `model`: 可填写 `moonshot-v1-8k、moonshot-v1-32k、moonshot-v1-128k`
- - `moonshot_api_key`: Moonshot的API-KEY，在 [控制台](https://platform.moonshot.cn/console/api-keys) 创建
- 
-方式二：OpenAI兼容方式接入，配置如下：
-```json
-{
-  "bot_type": "chatGPT",
-  "model": "moonshot-v1-128k",
-  "open_ai_api_base": "https://api.moonshot.cn/v1",
-  "open_ai_api_key": ""
-}
-```
- `bot_type`: OpenAI兼容方式
- `model`: 可填写 `moonshot-v1-8k、moonshot-v1-32k、moonshot-v1-128k`
- `open_ai_api_base`: Moonshot的 BASE URL
- `open_ai_api_key`: Moonshot的 API-KEY
-</details>
-
 <details>
 <summary>Azure</summary>

@@ -514,7 +535,7 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 方式二：OpenAI兼容方式接入，配置如下：
 ```json
 {
-  "bot_type": "chatGPT",
+  "bot_type": "openai",
  "model": "ERNIE-4.0-Turbo-8K",
  "open_ai_api_base": "https://qianfan.baidubce.com/v2",
  "open_ai_api_key": "bce-v3/ALTxxxxxxd2b"
@@ -550,7 +571,7 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 方式二：OpenAI兼容方式接入，配置如下：
 ```json
 {
-  "bot_type": "chatGPT",
+  "bot_type": "openai",
  "model": "4.0Ultra",
  "open_ai_api_base": "https://spark-api-open.xf-yun.com/v1",
  "open_ai_api_key": ""
@@ -582,15 +603,34 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 - `text_to_image`: 图像生成模型，参考[模型列表](https://www.modelscope.cn/models?filter=inference_type&page=1)
 </details>

+<details>
+<summary>Coding Plan</summary>
+
+Coding Plan 是各厂商推出的编程包月套餐，所有厂商均可通过 OpenAI 兼容方式接入：
+
+```json
+{
+  "bot_type": "openai",
+  "model": "模型名称",
+  "open_ai_api_base": "厂商 Coding Plan API Base",
+  "open_ai_api_key": "YOUR_API_KEY"
+}
+```
+
+目前支持阿里云、MiniMax、智谱GLM、Kimi、火山引擎等厂商，各厂商详细配置请参考 [Coding Plan 文档](https://docs.cowagent.ai/models/coding-plan)。
+</details>
+

 ## 通道说明

 以下对可接入通道的配置方式进行说明，应用通道代码在项目的 `channel/` 目录下。

+支持同时可接入多个通道，配置时可通过逗号进行分割，例如 `"channel_type": "feishu,dingtalk"`。
+
 <details>
 <summary>1. Web</summary>

-项目启动后默认运行Web通道，配置如下：
+项目启动后会默认运行Web控制台，配置如下：

 ```json
 {
@@ -636,7 +676,7 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 - `feishu_event_mode`: 事件接收模式，`websocket`（推荐）或 `webhook`
 - WebSocket 模式需安装依赖：`pip3 install lark-oapi`

-详细步骤和参数说明参考 [飞书接入](https://docs.link-ai.tech/cow/multi-platform/feishu)
+详细步骤和参数说明参考 [飞书接入](https://docs.cowagent.ai/channels/feishu)

 </details>

@@ -652,11 +692,43 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
    "dingtalk_client_secret": "CLIENT_SECRET"
 }
 ```
-详细步骤和参数说明参考 [钉钉接入](https://docs.link-ai.tech/cow/multi-platform/dingtalk)
+详细步骤和参数说明参考 [钉钉接入](https://docs.cowagent.ai/channels/dingtalk)
 </details>

 <details>
-<summary>4. WeCom App - 企业微信应用</summary>
+<summary>4. WeCom Bot - 企微智能机器人</summary>
+
+企微智能机器人使用 WebSocket 长连接模式，无需公网 IP 和域名，配置简单：
+
+```json
+{
+    "channel_type": "wecom_bot",
+    "wecom_bot_id": "YOUR_BOT_ID",
+    "wecom_bot_secret": "YOUR_SECRET"
+}
+```
+详细步骤和参数说明参考 [企微智能机器人接入](https://docs.cowagent.ai/channels/wecom-bot)
+
+</details>
+
+<details>
+<summary>5. QQ - QQ 机器人</summary>
+
+QQ 机器人使用 WebSocket 长连接模式，无需公网 IP 和域名，支持 QQ 单聊、群聊和频道消息：
+
+```json
+{
+    "channel_type": "qq",
+    "qq_app_id": "YOUR_APP_ID",
+    "qq_app_secret": "YOUR_APP_SECRET"
+}
+```
+详细步骤和参数说明参考 [QQ 机器人接入](https://docs.cowagent.ai/channels/qq)
+
+</details>
+
+<details>
+<summary>6. WeCom App - 企业微信应用</summary>

 企业微信自建应用接入需在后台创建应用并启用消息回调，配置示例：

@@ -671,12 +743,12 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
    "wechatcomapp_aes_key": "AESKEY"
 }
 ```
-详细步骤和参数说明参考 [企微自建应用接入](https://docs.link-ai.tech/cow/multi-platform/wechat-com)
+详细步骤和参数说明参考 [企微自建应用接入](https://docs.cowagent.ai/channels/wecom)

 </details>

 <details>
-<summary>5. WeChat MP - 微信公众号</summary>
+<summary>7. WeChat MP - 微信公众号</summary>

 本项目支持订阅号和服务号两种公众号，通过服务号（`wechatmp_service`）体验更佳。

@@ -706,12 +778,12 @@ API Key创建：在 [控制台](https://aistudio.google.com/app/apikey?hl=zh-cn)
 }
 ```

-详细步骤和参数说明参考 [微信公众号接入](https://docs.link-ai.tech/cow/multi-platform/wechat-mp)
+详细步骤和参数说明参考 [微信公众号接入](https://docs.cowagent.ai/channels/wechatmp)

 </details>

 <details>
-<summary>6. Terminal - 终端</summary>
+<summary>8. Terminal - 终端</summary>

 修改 `config.json` 中的 `channel_type` 字段：

--- a/agent/chat/init.py
+++ b/agent/chat/init.py
@@ -0,0 +1,3 @@
+from agent.chat.service import ChatService
+
+__all__ = ["ChatService"]
--- a/agent/chat/service.py
+++ b/agent/chat/service.py
@@ -0,0 +1,213 @@
+"""
+ChatService - Wraps the Agent stream execution to produce CHAT protocol chunks.
+
+Translates agent events (message_update, message_end, tool_execution_end, etc.)
+into the CHAT socket protocol format (content chunks with segment_id, tool_calls chunks).
+"""
+
+import time
+from typing import Callable, Optional
+
+from common.log import logger
+
+
+class ChatService:
+    """
+    High-level service that runs an Agent for a given query and streams
+    the results as CHAT protocol chunks via a callback.
+
+    Usage:
+        svc = ChatService(agent_bridge)
+        svc.run(query, session_id, send_chunk_fn)
+    """
+
+    def __init__(self, agent_bridge):
+        """
+        :param agent_bridge: AgentBridge instance (manages agent lifecycle)
+        """
+        self.agent_bridge = agent_bridge
+
+    def run(self, query: str, session_id: str, send_chunk_fn: Callable[[dict], None],
+            channel_type: str = ""):
+        """
+        Run the agent for *query* and stream results back via *send_chunk_fn*.
+
+        The method blocks until the agent finishes. After it returns the SDK
+        will automatically send the final (streaming=false) message.
+
+        :param query: user query text
+        :param session_id: session identifier for agent isolation
+        :param send_chunk_fn: callable(chunk_data: dict) to send a streaming chunk
+        :param channel_type: source channel (e.g. "web", "feishu") for persistence
+        """
+        agent = self.agent_bridge.get_agent(session_id=session_id)
+        if agent is None:
+            raise RuntimeError("Failed to initialise agent for the session")
+
+        # State shared between the event callback and this method
+        state = _StreamState()
+
+        def on_event(event: dict):
+            """Translate agent events into CHAT protocol chunks."""
+            event_type = event.get("type")
+            data = event.get("data", {})
+
+            if event_type == "message_update":
+                # Incremental text delta
+                delta = data.get("delta", "")
+                if delta:
+                    send_chunk_fn({
+                        "chunk_type": "content",
+                        "delta": delta,
+                        "segment_id": state.segment_id,
+                    })
+
+            elif event_type == "message_end":
+                # A content segment finished.
+                tool_calls = data.get("tool_calls", [])
+                if tool_calls:
+                    # After tool_calls are executed the next content will be
+                    # a new segment; collect tool results until turn_end.
+                    state.pending_tool_results = []
+
+            elif event_type == "tool_execution_start":
+                # Notify the client that a tool is about to run (with its input args)
+                tool_name = data.get("tool_name", "")
+                arguments = data.get("arguments", {})
+                # Cache arguments keyed by tool_call_id so tool_execution_end can include them
+                tool_call_id = data.get("tool_call_id", tool_name)
+                state.pending_tool_arguments[tool_call_id] = arguments
+                send_chunk_fn({
+                    "chunk_type": "tool_start",
+                    "tool": tool_name,
+                    "arguments": arguments,
+                })
+
+            elif event_type == "tool_execution_end":
+                tool_name = data.get("tool_name", "")
+                tool_call_id = data.get("tool_call_id", tool_name)
+                # Retrieve cached arguments from the matching tool_execution_start event
+                arguments = state.pending_tool_arguments.pop(tool_call_id, data.get("arguments", {}))
+                result = data.get("result", "")
+                status = data.get("status", "unknown")
+                execution_time = data.get("execution_time", 0)
+                elapsed_str = f"{execution_time:.2f}s"
+
+                # Serialise result to string if needed
+                if not isinstance(result, str):
+                    import json
+                    try:
+                        result = json.dumps(result, ensure_ascii=False)
+                    except Exception:
+                        result = str(result)
+
+                tool_info = {
+                    "name": tool_name,
+                    "arguments": arguments,
+                    "result": result,
+                    "status": status,
+                    "elapsed": elapsed_str,
+                }
+
+                if state.pending_tool_results is not None:
+                    state.pending_tool_results.append(tool_info)
+
+            elif event_type == "turn_end":
+                has_tool_calls = data.get("has_tool_calls", False)
+                if has_tool_calls and state.pending_tool_results:
+                    # Flush collected tool results as a single tool_calls chunk
+                    send_chunk_fn({
+                        "chunk_type": "tool_calls",
+                        "tool_calls": state.pending_tool_results,
+                    })
+                    state.pending_tool_results = None
+                    # Next content belongs to a new segment
+                    state.segment_id += 1
+
+        # Run the agent with our event callback ---------------------------
+        logger.info(f"[ChatService] Starting agent run: session={session_id}, query={query[:80]}")
+
+        from config import conf
+        max_context_turns = conf().get("agent_max_context_turns", 20)
+
+        # Get full system prompt with skills
+        full_system_prompt = agent.get_full_system_prompt()
+
+        # Create a copy of messages for this execution
+        with agent.messages_lock:
+            messages_copy = agent.messages.copy()
+            original_length = len(agent.messages)
+
+        from agent.protocol.agent_stream import AgentStreamExecutor
+
+        executor = AgentStreamExecutor(
+            agent=agent,
+            model=agent.model,
+            system_prompt=full_system_prompt,
+            tools=agent.tools,
+            max_turns=agent.max_steps,
+            on_event=on_event,
+            messages=messages_copy,
+            max_context_turns=max_context_turns,
+        )
+
+        try:
+            response = executor.run_stream(query)
+        except Exception:
+            # If executor cleared messages (context overflow), sync back
+            if len(executor.messages) == 0:
+                with agent.messages_lock:
+                    agent.messages.clear()
+                    logger.info("[ChatService] Cleared agent message history after executor recovery")
+            raise
+
+        # Append only the NEW messages from this execution (thread-safe)
+        with agent.messages_lock:
+            new_messages = executor.messages[original_length:]
+            agent.messages.extend(new_messages)
+
+        # Persist new messages to SQLite so they survive restarts and
+        # can be queried via the HISTORY interface.
+        if new_messages:
+            self._persist_messages(session_id, list(new_messages), channel_type)
+
+        # Store executor reference for files_to_send access
+        agent.stream_executor = executor
+
+        # Execute post-process tools
+        agent._execute_post_process_tools()
+
+        logger.info(f"[ChatService] Agent run completed: session={session_id}")
+
+
+
+    @staticmethod
+    def _persist_messages(session_id: str, new_messages: list, channel_type: str = ""):
+        try:
+            from config import conf
+            if not conf().get("conversation_persistence", True):
+                return
+        except Exception:
+            pass
+        try:
+            from agent.memory import get_conversation_store
+            get_conversation_store().append_messages(
+                session_id, new_messages, channel_type=channel_type
+            )
+        except Exception as e:
+            logger.warning(
+                f"[ChatService] Failed to persist messages for session={session_id}: {e}"
+            )
+
+
+class _StreamState:
+    """Mutable state shared between the event callback and the run method."""
+
+    def __init__(self):
+        self.segment_id: int = 0
+        # None means we are not accumulating tool results right now.
+        # A list means we are in the middle of a tool-execution phase.
+        self.pending_tool_results: Optional[list] = None
+        # Maps tool_call_id -> arguments captured from tool_execution_start,
+        # so that tool_execution_end can attach the correct input args.
+        self.pending_tool_arguments: dict = {}
--- a/agent/memory/init.py
+++ b/agent/memory/init.py
@@ -1,11 +1,23 @@
 """
 Memory module for AgentMesh

-Provides long-term memory capabilities with hybrid search (vector + keyword)
+Provides both long-term memory (vector/keyword search) and short-term
+conversation history persistence (SQLite).
 """

 from agent.memory.manager import MemoryManager
 from agent.memory.config import MemoryConfig, get_default_memory_config, set_global_memory_config
 from agent.memory.embedding import create_embedding_provider
+from agent.memory.conversation_store import ConversationStore, get_conversation_store
+from agent.memory.summarizer import ensure_daily_memory_file

-__all__ = ['MemoryManager', 'MemoryConfig', 'get_default_memory_config', 'set_global_memory_config', 'create_embedding_provider']
+__all__ = [
+    'MemoryManager',
+    'MemoryConfig',
+    'get_default_memory_config',
+    'set_global_memory_config',
+    'create_embedding_provider',
+    'ConversationStore',
+    'get_conversation_store',
+    'ensure_daily_memory_file',
+]
--- a/agent/memory/config.py
+++ b/agent/memory/config.py
@@ -11,12 +11,18 @@ from typing import Optional, List
 from pathlib import Path


+def _default_workspace():
+    """Get default workspace path with proper Windows support"""
+    from common.utils import expand_path
+    return expand_path("~/cow")
+
+
@dataclass
 class MemoryConfig:
    """Configuration for memory storage and search"""
    
    # Storage paths (default: ~/cow)
-    workspace_root: str = field(default_factory=lambda: os.path.expanduser("~/cow"))
+    workspace_root: str = field(default_factory=_default_workspace)
    
    # Embedding config
    embedding_provider: str = "openai"  # "openai" | "local"
@@ -42,9 +48,6 @@ class MemoryConfig:
    enable_auto_sync: bool = True
    sync_on_search: bool = True
    
-    # Memory flush config (独立于模型 context window)
-    flush_token_threshold: int = 50000  # 50K tokens 触发 flush
-    flush_turn_threshold: int = 20  # 20 轮对话触发 flush (用户+AI各一条为一轮)
    
    def get_workspace(self) -> Path:
        """Get workspace root directory"""
--- a/agent/memory/conversation_store.py
+++ b/agent/memory/conversation_store.py
@@ -0,0 +1,618 @@
+"""
+Conversation history persistence using SQLite.
+
+Design:
+- sessions table: per-session metadata (channel_type, last_active, msg_count)
+- messages table: individual messages stored as JSON, append-only
+- Pruning: age-based only (sessions not updated within N days are deleted)
+- Thread-safe via a single in-process lock
+
+Storage path: ~/cow/sessions/conversations.db
+"""
+
+from __future__ import annotations
+
+import json
+import sqlite3
+import threading
+import time
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+from common.log import logger
+
+
+# ---------------------------------------------------------------------------
+# Schema
+# ---------------------------------------------------------------------------
+
+_DDL = """
+CREATE TABLE IF NOT EXISTS sessions (
+    session_id   TEXT    PRIMARY KEY,
+    channel_type TEXT    NOT NULL DEFAULT '',
+    created_at   INTEGER NOT NULL,
+    last_active  INTEGER NOT NULL,
+    msg_count    INTEGER NOT NULL DEFAULT 0
+);
+
+CREATE TABLE IF NOT EXISTS messages (
+    id           INTEGER PRIMARY KEY AUTOINCREMENT,
+    session_id   TEXT    NOT NULL,
+    seq          INTEGER NOT NULL,
+    role         TEXT    NOT NULL,
+    content      TEXT    NOT NULL,
+    created_at   INTEGER NOT NULL,
+    UNIQUE (session_id, seq)
+);
+
+CREATE INDEX IF NOT EXISTS idx_messages_session
+    ON messages (session_id, seq);
+
+CREATE INDEX IF NOT EXISTS idx_sessions_last_active
+    ON sessions (last_active);
+"""
+
+# Migration: add channel_type column to existing databases that predate it.
+_MIGRATION_ADD_CHANNEL_TYPE = """
+ALTER TABLE sessions ADD COLUMN channel_type TEXT NOT NULL DEFAULT '';
+"""
+
+DEFAULT_MAX_AGE_DAYS: int = 30
+
+
+def _is_visible_user_message(content: Any) -> bool:
+    """
+    Return True when a user-role message represents actual user input
+    (not an internal tool_result injected by the agent loop).
+    """
+    if isinstance(content, str):
+        return bool(content.strip())
+    if isinstance(content, list):
+        return any(
+            isinstance(b, dict) and b.get("type") == "text"
+            for b in content
+        )
+    return False
+
+
+def _extract_display_text(content: Any) -> str:
+    """
+    Extract the human-readable text portion from a message content value.
+    Returns an empty string for tool_use / tool_result blocks.
+    """
+    if isinstance(content, str):
+        return content.strip()
+    if isinstance(content, list):
+        parts = [
+            b.get("text", "")
+            for b in content
+            if isinstance(b, dict) and b.get("type") == "text"
+        ]
+        return "\n".join(p for p in parts if p).strip()
+    return ""
+
+
+def _extract_tool_calls(content: Any) -> List[Dict[str, Any]]:
+    """
+    Extract tool_use blocks from an assistant message content.
+    Returns a list of {name, arguments} dicts (result filled in later).
+    """
+    if not isinstance(content, list):
+        return []
+    return [
+        {"id": b.get("id", ""), "name": b.get("name", ""), "arguments": b.get("input", {})}
+        for b in content
+        if isinstance(b, dict) and b.get("type") == "tool_use"
+    ]
+
+
+def _extract_tool_results(content: Any) -> Dict[str, str]:
+    """
+    Extract tool_result blocks from a user message, keyed by tool_use_id.
+    """
+    if not isinstance(content, list):
+        return {}
+    results = {}
+    for b in content:
+        if not isinstance(b, dict) or b.get("type") != "tool_result":
+            continue
+        tool_id = b.get("tool_use_id", "")
+        result_content = b.get("content", "")
+        if isinstance(result_content, list):
+            result_content = "\n".join(
+                rb.get("text", "") for rb in result_content
+                if isinstance(rb, dict) and rb.get("type") == "text"
+            )
+        results[tool_id] = str(result_content)
+    return results
+
+
+def _group_into_display_turns(
+    rows: List[tuple],
+) -> List[Dict[str, Any]]:
+    """
+    Convert raw (role, content_json, created_at) DB rows into display turns.
+
+    One display turn = one visible user message  +  one merged assistant reply.
+    All intermediate assistant messages (those carrying tool_use) and the final
+    assistant text reply produced for the same user query are collapsed into a
+    single assistant turn, exactly matching the live SSE rendering where tools
+    and the final answer appear inside the same bubble.
+
+    Grouping rules:
+    - A visible user message starts a new group.
+    - tool_result user messages are internal; their content is attached to the
+      matching tool_use entry via tool_use_id and they never become own turns.
+    - All assistant messages within a group are merged:
+        * tool_use blocks → tool_calls list (result filled from tool_results)
+        * text blocks → last non-empty text becomes the display content
+    """
+    # ------------------------------------------------------------------ #
+    # Pass 1: split rows into groups, each starting with a visible user msg
+    # ------------------------------------------------------------------ #
+    # group = (user_row | None, [subsequent_rows])
+    # user_row: (content, created_at)
+    groups: List[tuple] = []
+    cur_user: Optional[tuple] = None
+    cur_rest: List[tuple] = []
+    started = False
+
+    for role, raw_content, created_at in rows:
+        try:
+            content = json.loads(raw_content)
+        except Exception:
+            content = raw_content
+
+        if role == "user" and _is_visible_user_message(content):
+            if started:
+                groups.append((cur_user, cur_rest))
+            cur_user = (content, created_at)
+            cur_rest = []
+            started = True
+        else:
+            cur_rest.append((role, content, created_at))
+
+    if started:
+        groups.append((cur_user, cur_rest))
+
+    # ------------------------------------------------------------------ #
+    # Pass 2: build display turns from each group
+    # ------------------------------------------------------------------ #
+    turns: List[Dict[str, Any]] = []
+
+    for user_row, rest in groups:
+        # User turn
+        if user_row:
+            content, created_at = user_row
+            text = _extract_display_text(content)
+            if text:
+                turns.append({"role": "user", "content": text, "created_at": created_at})
+
+        # Collect all tool_calls and tool_results from the rest of the group
+        all_tool_calls: List[Dict[str, Any]] = []
+        tool_results: Dict[str, str] = {}
+        final_text = ""
+        final_ts: Optional[int] = None
+
+        for role, content, created_at in rest:
+            if role == "user":
+                tool_results.update(_extract_tool_results(content))
+            elif role == "assistant":
+                tcs = _extract_tool_calls(content)
+                all_tool_calls.extend(tcs)
+                t = _extract_display_text(content)
+                if t:
+                    final_text = t
+                final_ts = created_at
+
+        # Attach tool results to their matching tool_call entries
+        for tc in all_tool_calls:
+            tc["result"] = tool_results.get(tc.get("id", ""), "")
+
+        if final_text or all_tool_calls:
+            turns.append({
+                "role": "assistant",
+                "content": final_text,
+                "tool_calls": all_tool_calls,
+                "created_at": final_ts or (user_row[1] if user_row else 0),
+            })
+
+    return turns
+
+
+class ConversationStore:
+    """
+    SQLite-backed store for per-session conversation history.
+
+    Usage:
+        store = ConversationStore(db_path)
+        store.append_messages("user_123", new_messages, channel_type="feishu")
+        msgs = store.load_messages("user_123", max_turns=30)
+    """
+
+    def __init__(self, db_path: Path):
+        self._db_path = db_path
+        self._lock = threading.Lock()
+        self._init_db()
+
+    # ------------------------------------------------------------------
+    # Public API
+    # ------------------------------------------------------------------
+
+    def load_messages(
+        self,
+        session_id: str,
+        max_turns: int = 30,
+    ) -> List[Dict[str, Any]]:
+        """
+        Load the most recent messages for a session, for injection into the LLM.
+
+        ALL message types (user text, assistant tool_use, tool_result) are returned
+        in their original JSON form so the LLM can reconstruct the full context.
+
+        max_turns is a *visible-turn* count: we count only user messages whose
+        content is actual user text (not tool_result blocks).  This prevents
+        tool-heavy sessions from exhausting the turn budget prematurely.
+
+        Args:
+            session_id: Unique session identifier.
+            max_turns: Maximum number of visible user-assistant turns to keep.
+
+        Returns:
+            Chronologically ordered list of message dicts (role, content).
+        """
+        with self._lock:
+            conn = self._connect()
+            try:
+                rows = conn.execute(
+                    """
+                    SELECT seq, role, content
+                    FROM messages
+                    WHERE session_id = ?
+                    ORDER BY seq DESC
+                    """,
+                    (session_id,),
+                ).fetchall()
+            finally:
+                conn.close()
+
+        if not rows:
+            return []
+
+        # Walk newest-to-oldest counting *visible* user turns (actual user text,
+        # not tool_result injections).  Record the seq of every visible user
+        # message so we can find a clean cut point later.
+        visible_turn_seqs: List[int] = []  # newest first
+        for seq, role, raw_content in rows:
+            if role != "user":
+                continue
+            try:
+                content = json.loads(raw_content)
+            except Exception:
+                content = raw_content
+            if _is_visible_user_message(content):
+                visible_turn_seqs.append(seq)
+
+        # Determine the seq of the oldest visible user message we want to keep.
+        # If the total turns fit within max_turns, keep everything.
+        if len(visible_turn_seqs) <= max_turns:
+            cutoff_seq = None  # keep all
+        else:
+            # The Nth visible user message (0-indexed) is the oldest we keep.
+            cutoff_seq = visible_turn_seqs[max_turns - 1]
+
+        # Build result in chronological order, starting from cutoff.
+        # IMPORTANT: we start exactly at cutoff_seq (the visible user message),
+        # never mid-group, so tool_use / tool_result pairs are always complete.
+        result = []
+        for seq, role, raw_content in reversed(rows):
+            if cutoff_seq is not None and seq < cutoff_seq:
+                continue
+            try:
+                content = json.loads(raw_content)
+            except Exception:
+                content = raw_content
+            result.append({"role": role, "content": content})
+        return result
+
+    def append_messages(
+        self,
+        session_id: str,
+        messages: List[Dict[str, Any]],
+        channel_type: str = "",
+    ) -> None:
+        """
+        Append new messages to a session's history.
+
+        Seq numbers continue from the session's current maximum, so
+        concurrent callers on distinct sessions never collide.
+
+        Args:
+            session_id: Unique session identifier.
+            messages: List of message dicts to append.
+            channel_type: Source channel (e.g. "feishu", "web", "wechat").
+                          Only written on session creation; ignored on update.
+        """
+        if not messages:
+            return
+
+        now = int(time.time())
+        with self._lock:
+            conn = self._connect()
+            try:
+                with conn:
+                    # INSERT OR IGNORE creates the row on first visit;
+                    # the UPDATE always refreshes last_active.
+                    # Avoids ON CONFLICT...DO UPDATE (requires SQLite >= 3.24).
+                    conn.execute(
+                        """
+                        INSERT OR IGNORE INTO sessions
+                            (session_id, channel_type, created_at, last_active, msg_count)
+                        VALUES (?, ?, ?, ?, 0)
+                        """,
+                        (session_id, channel_type, now, now),
+                    )
+                    conn.execute(
+                        "UPDATE sessions SET last_active = ? WHERE session_id = ?",
+                        (now, session_id),
+                    )
+
+                    # Determine starting seq for the new batch.
+                    row = conn.execute(
+                        "SELECT COALESCE(MAX(seq), -1) FROM messages WHERE session_id = ?",
+                        (session_id,),
+                    ).fetchone()
+                    next_seq = row[0] + 1
+
+                    for msg in messages:
+                        role = msg.get("role", "")
+                        content = json.dumps(
+                            msg.get("content", ""), ensure_ascii=False
+                        )
+                        conn.execute(
+                            """
+                            INSERT OR IGNORE INTO messages
+                                (session_id, seq, role, content, created_at)
+                            VALUES (?, ?, ?, ?, ?)
+                            """,
+                            (session_id, next_seq, role, content, now),
+                        )
+                        next_seq += 1
+
+                    conn.execute(
+                        """
+                        UPDATE sessions
+                        SET msg_count = (
+                            SELECT COUNT(*) FROM messages WHERE session_id = ?
+                        )
+                        WHERE session_id = ?
+                        """,
+                        (session_id, session_id),
+                    )
+            finally:
+                conn.close()
+
+    def clear_session(self, session_id: str) -> None:
+        """Delete all messages and the session record for a given session_id."""
+        with self._lock:
+            conn = self._connect()
+            try:
+                with conn:
+                    conn.execute(
+                        "DELETE FROM messages WHERE session_id = ?", (session_id,)
+                    )
+                    conn.execute(
+                        "DELETE FROM sessions WHERE session_id = ?", (session_id,)
+                    )
+            finally:
+                conn.close()
+
+    def cleanup_old_sessions(self, max_age_days: Optional[int] = None) -> int:
+        """
+        Delete sessions that have not been active within max_age_days.
+
+        Args:
+            max_age_days: Override the default retention period.
+
+        Returns:
+            Number of sessions deleted.
+        """
+        try:
+            from config import conf
+            max_age = max_age_days or conf().get(
+                "conversation_max_age_days", DEFAULT_MAX_AGE_DAYS
+            )
+        except Exception:
+            max_age = max_age_days or DEFAULT_MAX_AGE_DAYS
+
+        cutoff = int(time.time()) - max_age * 86400
+        deleted = 0
+
+        with self._lock:
+            conn = self._connect()
+            try:
+                with conn:
+                    stale = conn.execute(
+                        "SELECT session_id FROM sessions WHERE last_active < ?",
+                        (cutoff,),
+                    ).fetchall()
+                    for (sid,) in stale:
+                        conn.execute(
+                            "DELETE FROM messages WHERE session_id = ?", (sid,)
+                        )
+                        conn.execute(
+                            "DELETE FROM sessions WHERE session_id = ?", (sid,)
+                        )
+                        deleted += 1
+            finally:
+                conn.close()
+
+        if deleted:
+            logger.info(f"[ConversationStore] Pruned {deleted} expired sessions")
+        return deleted
+
+    def load_history_page(
+        self,
+        session_id: str,
+        page: int = 1,
+        page_size: int = 20,
+    ) -> Dict[str, Any]:
+        """
+        Load a page of conversation history for UI display, grouped into turns.
+
+        Each "turn" maps to one of:
+          - A user message (role="user", content=str)
+          - An assistant message (role="assistant", content=str,
+            tool_calls=[{name, arguments, result}] when tools were used)
+
+        Internal tool_result user messages are merged into the preceding
+        assistant entry's tool_calls list and never appear as standalone items.
+
+        Pages are numbered from 1 (most recent).  Messages within a page are
+        returned in chronological order.
+
+        Returns:
+            {
+                "messages": [
+                    {
+                        "role": "user" | "assistant",
+                        "content": str,
+                        "tool_calls": [...],   # assistant only, may be []
+                        "created_at": int,
+                    },
+                    ...
+                ],
+                "total": <visible turn count>,
+                "page": <current page>,
+                "page_size": <page_size>,
+                "has_more": bool,
+            }
+        """
+        page = max(1, page)
+        with self._lock:
+            conn = self._connect()
+            try:
+                rows = conn.execute(
+                    """
+                    SELECT role, content, created_at
+                    FROM messages
+                    WHERE session_id = ?
+                    ORDER BY seq ASC
+                    """,
+                    (session_id,),
+                ).fetchall()
+            finally:
+                conn.close()
+
+        visible = _group_into_display_turns(rows)
+
+        total = len(visible)
+        offset = (page - 1) * page_size
+        page_items = list(reversed(visible))[offset: offset + page_size]
+        page_items = list(reversed(page_items))
+
+        return {
+            "messages": page_items,
+            "total": total,
+            "page": page,
+            "page_size": page_size,
+            "has_more": offset + page_size < total,
+        }
+
+    def get_stats(self) -> Dict[str, Any]:
+        """Return basic stats keyed by channel_type, for monitoring."""
+        with self._lock:
+            conn = self._connect()
+            try:
+                total_sessions = conn.execute(
+                    "SELECT COUNT(*) FROM sessions"
+                ).fetchone()[0]
+                total_messages = conn.execute(
+                    "SELECT COUNT(*) FROM messages"
+                ).fetchone()[0]
+                by_channel = conn.execute(
+                    """
+                    SELECT channel_type, COUNT(*) as cnt
+                    FROM sessions
+                    GROUP BY channel_type
+                    ORDER BY cnt DESC
+                    """
+                ).fetchall()
+                return {
+                    "total_sessions": total_sessions,
+                    "total_messages": total_messages,
+                    "by_channel": {row[0] or "unknown": row[1] for row in by_channel},
+                }
+            finally:
+                conn.close()
+
+    # ------------------------------------------------------------------
+    # Internal helpers
+    # ------------------------------------------------------------------
+
+    def _init_db(self) -> None:
+        self._db_path.parent.mkdir(parents=True, exist_ok=True)
+        conn = self._connect()
+        try:
+            conn.executescript(_DDL)
+            conn.commit()
+            self._migrate(conn)
+        finally:
+            conn.close()
+
+    def _migrate(self, conn: sqlite3.Connection) -> None:
+        """Apply incremental schema migrations on existing databases."""
+        cols = {
+            row[1]
+            for row in conn.execute("PRAGMA table_info(sessions)").fetchall()
+        }
+        if "channel_type" not in cols:
+            try:
+                conn.execute(_MIGRATION_ADD_CHANNEL_TYPE)
+                conn.commit()
+                logger.info("[ConversationStore] Migrated: added channel_type column")
+            except Exception as e:
+                logger.warning(f"[ConversationStore] Migration failed: {e}")
+
+    def _connect(self) -> sqlite3.Connection:
+        conn = sqlite3.connect(str(self._db_path), timeout=10)
+        conn.execute("PRAGMA journal_mode=WAL")
+        conn.execute("PRAGMA synchronous=NORMAL")
+        return conn
+
+
+# ---------------------------------------------------------------------------
+# Singleton
+# ---------------------------------------------------------------------------
+
+_store_instance: Optional[ConversationStore] = None
+_store_lock = threading.Lock()
+
+
+def get_conversation_store() -> ConversationStore:
+    """
+    Return the process-wide ConversationStore singleton.
+
+    Reuses the long-term memory database so the project stays with a single
+    SQLite file: ~/cow/memory/long-term/index.db
+    The conversation tables (sessions / messages) are separate from the
+    memory tables (memory_chunks / file_metadata) — no conflicts.
+    """
+    global _store_instance
+    if _store_instance is not None:
+        return _store_instance
+
+    with _store_lock:
+        if _store_instance is not None:
+            return _store_instance
+
+        try:
+            from agent.memory.config import get_default_memory_config
+            db_path = get_default_memory_config().get_db_path()
+        except Exception:
+            from common.utils import expand_path
+            db_path = Path(expand_path("~/cow")) / "memory" / "long-term" / "index.db"
+
+        _store_instance = ConversationStore(db_path)
+        logger.debug(f"[ConversationStore] Using shared DB at: {db_path}")
+        return _store_instance
--- a/agent/memory/embedding.py
+++ b/agent/memory/embedding.py
@@ -138,24 +138,24 @@ def create_embedding_provider(
 ) -> EmbeddingProvider:
    """
    Factory function to create embedding provider
-    
-    Only supports OpenAI embedding via REST API.
+
+    Supports "openai" and "linkai" providers (both use OpenAI-compatible REST API).
    If initialization fails, caller should fall back to keyword-only search.
-    
+
    Args:
-        provider: Provider name (only "openai" is supported)
+        provider: Provider name ("openai" or "linkai")
        model: Model name (default: text-embedding-3-small)
-        api_key: OpenAI API key (required)
-        api_base: API base URL (default: https://api.openai.com/v1)
+        api_key: API key (required)
+        api_base: API base URL
        
    Returns:
        EmbeddingProvider instance
        
    Raises:
-        ValueError: If provider is not "openai" or api_key is missing
+        ValueError: If provider is unsupported or api_key is missing
    """
-    if provider != "openai":
-        raise ValueError(f"Only 'openai' provider is supported, got: {provider}")
+    if provider not in ("openai", "linkai"):
+        raise ValueError(f"Unsupported embedding provider: {provider}. Use 'openai' or 'linkai'.")

    model = model or "text-embedding-3-small"
    return OpenAIEmbeddingProvider(model=model, api_key=api_key, api_base=api_base)
--- a/agent/memory/manager.py
+++ b/agent/memory/manager.py
@@ -50,28 +50,44 @@ class MemoryManager:
            overlap_tokens=self.config.chunk_overlap_tokens
        )
        
-        # Initialize embedding provider (optional)
+        # Initialize embedding provider (optional, prefer OpenAI, fallback to LinkAI)
        self.embedding_provider = None
        if embedding_provider:
            self.embedding_provider = embedding_provider
        else:
-            # Try to create embedding provider, but allow failure
+            # Try OpenAI first
            try:
-                # Get API key from environment or config
                api_key = os.environ.get('OPENAI_API_KEY')
                api_base = os.environ.get('OPENAI_API_BASE')
-                
-                self.embedding_provider = create_embedding_provider(
-                    provider=self.config.embedding_provider,
-                    model=self.config.embedding_model,
-                    api_key=api_key,
-                    api_base=api_base
-                )
+                if api_key:
+                    self.embedding_provider = create_embedding_provider(
+                        provider="openai",
+                        model=self.config.embedding_model,
+                        api_key=api_key,
+                        api_base=api_base
+                    )
            except Exception as e:
-                # Embedding provider failed, but that's OK
-                # We can still use keyword search and file operations
                from common.log import logger
-                logger.warning(f"[MemoryManager] Embedding provider initialization failed: {e}")
+                logger.warning(f"[MemoryManager] OpenAI embedding failed: {e}")
+
+            # Fallback to LinkAI
+            if self.embedding_provider is None:
+                try:
+                    linkai_key = os.environ.get('LINKAI_API_KEY')
+                    linkai_base = os.environ.get('LINKAI_API_BASE', 'https://api.link-ai.tech')
+                    if linkai_key:
+                        self.embedding_provider = create_embedding_provider(
+                            provider="linkai",
+                            model=self.config.embedding_model,
+                            api_key=linkai_key,
+                            api_base=f"{linkai_base}/v1"
+                        )
+                except Exception as e:
+                    from common.log import logger
+                    logger.warning(f"[MemoryManager] LinkAI embedding failed: {e}")
+
+            if self.embedding_provider is None:
+                from common.log import logger
                logger.info(f"[MemoryManager] Memory will work with keyword search only (no vector search)")
        
        # Initialize memory flush manager
@@ -304,7 +320,7 @@ class MemoryManager:
    ):
        """Sync a single file"""
        # Compute file hash
-        content = file_path.read_text()
+        content = file_path.read_text(encoding='utf-8')
        file_hash = MemoryStorage.compute_hash(content)
        
        # Get relative path
@@ -363,182 +379,35 @@ class MemoryManager:
            size=stat.st_size
        )
    
-    def should_flush_memory(
+    def flush_memory(
        self,
-        current_tokens: int = 0
-    ) -> bool:
-        """
-        Check if memory flush should be triggered
-        
-        独立的 flush 触发机制，不依赖模型 context window。
-        使用配置中的阈值: flush_token_threshold 和 flush_turn_threshold
-        
-        Args:
-            current_tokens: Current session token count
-            
-        Returns:
-            True if memory flush should run
-        """
-        return self.flush_manager.should_flush(
-            current_tokens=current_tokens,
-            token_threshold=self.config.flush_token_threshold,
-            turn_threshold=self.config.flush_turn_threshold
-        )
-    
-    def increment_turn(self):
-        """增加对话轮数计数（每次用户消息+AI回复算一轮）"""
-        self.flush_manager.increment_turn()
-    
-    async def execute_memory_flush(
-        self,
-        agent_executor,
-        current_tokens: int,
+        messages: list,
        user_id: Optional[str] = None,
-        **executor_kwargs
+        reason: str = "threshold",
+        max_messages: int = 10,
    ) -> bool:
        """
-        Execute memory flush before compaction
-        
-        This runs a silent agent turn to write durable memories to disk.
-        Similar to clawdbot's pre-compaction memory flush.
+        Flush conversation summary to daily memory file.
        
        Args:
-            agent_executor: Async function to execute agent with prompt
-            current_tokens: Current session token count
+            messages: Conversation message list
            user_id: Optional user ID
-            **executor_kwargs: Additional kwargs for agent executor
-            
+            reason: "threshold" | "overflow" | "daily_summary"
+            max_messages: Max recent messages to include (0 = all)
+        
        Returns:
-            True if flush completed successfully
-            
-        Example:
-            >>> async def run_agent(prompt, system_prompt, silent=False):
-            ...     # Your agent execution logic
-            ...     pass
-            >>> 
-            >>> if manager.should_flush_memory(current_tokens=100000):
-            ...     await manager.execute_memory_flush(
-            ...         agent_executor=run_agent,
-            ...         current_tokens=100000
-            ...     )
+            True if content was written
        """
-        success = await self.flush_manager.execute_flush(
-            agent_executor=agent_executor,
-            current_tokens=current_tokens,
+        success = self.flush_manager.flush_from_messages(
+            messages=messages,
            user_id=user_id,
-            **executor_kwargs
+            reason=reason,
+            max_messages=max_messages,
        )
-        
        if success:
-            # Mark dirty so next search will sync the new memories
            self._dirty = True
-        
        return success
    
-    def build_memory_guidance(self, lang: str = "zh", include_context: bool = True) -> str:
-        """
-        Build natural memory guidance for agent system prompt
-        
-        Following clawdbot's approach:
-        1. Load MEMORY.md as bootstrap context (blends into background)
-        2. Load daily files on-demand via memory_search tool
-        3. Agent should NOT proactively mention memories unless user asks
-        
-        Args:
-            lang: Language for guidance ("en" or "zh")
-            include_context: Whether to include bootstrap memory context (default: True)
-                           MEMORY.md is loaded as background context (like clawdbot)
-                           Daily files are accessed via memory_search tool
-            
-        Returns:
-            Memory guidance text (and optionally context) for system prompt
-        """
-        today_file = self.flush_manager.get_today_memory_file().name
-        
-        if lang == "zh":
-            guidance = f"""## 记忆系统
-
-**背景知识**: 下方包含核心长期记忆，可直接使用。需要查找历史时，用 memory_search 搜索（搜索一次即可，不要重复）。
-
-**存储记忆**: 当用户分享重要信息时（偏好、决策、事实等），主动用 write 工具存储：
- 长期信息 → MEMORY.md
- 当天笔记 → memory/{today_file}
- 静默存储，仅在明确要求时确认
-
-**使用原则**: 自然使用记忆，就像你本来就知道。不需要生硬地提起或列举记忆，除非用户提到。"""
-        else:
-            guidance = f"""## Memory System
-
-**Background Knowledge**: Core long-term memories below - use directly. For history, use memory_search once (don't repeat).
-
-**Store Memories**: When user shares important info (preferences, decisions, facts), proactively write:
- Durable info → MEMORY.md  
- Daily notes → memory/{today_file}
- Store silently; confirm only when explicitly requested
-
-**Usage**: Use memories naturally as if you always knew. Don't mention or list unless user explicitly asks."""
-        
-        if include_context:
-            # Load bootstrap context (MEMORY.md only, like clawdbot)
-            bootstrap_context = self.load_bootstrap_memories()
-            if bootstrap_context:
-                guidance += f"\n\n## Background Context\n\n{bootstrap_context}"
-        
-        return guidance
-    
-    def load_bootstrap_memories(self, user_id: Optional[str] = None) -> str:
-        """
-        Load bootstrap memory files for session start
-        
-        Following clawdbot's design:
-        - Only loads MEMORY.md from workspace root (long-term curated memory)
-        - Daily files (memory/YYYY-MM-DD.md) are accessed via memory_search tool, not bootstrap
-        - User-specific MEMORY.md is also loaded if user_id provided
-        
-        Returns memory content WITHOUT obvious headers so it blends naturally
-        into the context as background knowledge.
-        
-        Args:
-            user_id: Optional user ID for user-specific memories
-            
-        Returns:
-            Memory content to inject into system prompt (blends naturally as background context)
-        """
-        workspace_dir = self.config.get_workspace()
-        memory_dir = self.config.get_memory_dir()
-        
-        sections = []
-        
-        # 1. Load MEMORY.md from workspace root (long-term curated memory)
-        # Following clawdbot: only MEMORY.md is bootstrap, daily files use memory_search
-        memory_file = Path(workspace_dir) / "MEMORY.md"
-        if memory_file.exists():
-            try:
-                content = memory_file.read_text(encoding='utf-8').strip()
-                if content:
-                    sections.append(content)
-            except Exception as e:
-                print(f"Warning: Failed to read MEMORY.md: {e}")
-        
-        # 2. Load user-specific MEMORY.md if user_id provided
-        if user_id:
-            user_memory_dir = memory_dir / "users" / user_id
-            user_memory_file = user_memory_dir / "MEMORY.md"
-            if user_memory_file.exists():
-                try:
-                    content = user_memory_file.read_text(encoding='utf-8').strip()
-                    if content:
-                        sections.append(content)
-                except Exception as e:
-                    print(f"Warning: Failed to read user memory: {e}")
-        
-        if not sections:
-            return ""
-        
-        # Join sections without obvious headers - let memories blend naturally
-        # This makes the agent feel like it "just knows" rather than "checking memory files"
-        return "\n\n".join(sections)
-    
    def get_status(self) -> Dict[str, Any]:
        """Get memory status"""
        stats = self.storage.get_stats()
@@ -568,6 +437,37 @@ class MemoryManager:
        content = f"{path}:{start_line}:{end_line}"
        return hashlib.md5(content.encode('utf-8')).hexdigest()
    
+    @staticmethod
+    def _compute_temporal_decay(path: str, half_life_days: float = 30.0) -> float:
+        """
+        Compute temporal decay multiplier for dated memory files.
+        
+        Inspired by OpenClaw's temporal-decay: exponential decay based on file date.
+        MEMORY.md and non-dated files are "evergreen" (no decay, multiplier=1.0).
+        Daily files like memory/2025-03-01.md decay based on age.
+        
+        Formula: multiplier = exp(-ln2/half_life * age_in_days)
+        """
+        import re
+        import math
+        
+        match = re.search(r'(\d{4})-(\d{2})-(\d{2})\.md$', path)
+        if not match:
+            return 1.0  # evergreen: MEMORY.md, non-dated files
+        
+        try:
+            file_date = datetime(
+                int(match.group(1)), int(match.group(2)), int(match.group(3))
+            )
+            age_days = (datetime.now() - file_date).days
+            if age_days <= 0:
+                return 1.0
+            
+            decay_lambda = math.log(2) / half_life_days
+            return math.exp(-decay_lambda * age_days)
+        except (ValueError, OverflowError):
+            return 1.0
+    
    def _merge_results(
        self,
        vector_results: List[SearchResult],
@@ -575,8 +475,7 @@ class MemoryManager:
        vector_weight: float,
        keyword_weight: float
    ) -> List[SearchResult]:
-        """Merge vector and keyword search results"""
-        # Create a map by (path, start_line, end_line)
+        """Merge vector and keyword search results with temporal decay for dated files"""
        merged_map = {}
        
        for result in vector_results:
@@ -598,7 +497,6 @@ class MemoryManager:
                    'keyword_score': result.score
                }
        
-        # Calculate combined scores
        merged_results = []
        for entry in merged_map.values():
            combined_score = (
@@ -606,7 +504,11 @@ class MemoryManager:
                keyword_weight * entry['keyword_score']
            )
            
+            # Apply temporal decay for dated memory files
            result = entry['result']
+            decay = self._compute_temporal_decay(result.path)
+            combined_score *= decay
+            
            merged_results.append(SearchResult(
                path=result.path,
                start_line=result.start_line,
@@ -617,6 +519,5 @@ class MemoryManager:
                user_id=result.user_id
            ))
        
-        # Sort by score
        merged_results.sort(key=lambda r: r.score, reverse=True)
        return merged_results
--- a/agent/memory/service.py
+++ b/agent/memory/service.py
@@ -0,0 +1,167 @@
+"""
+Memory service for handling memory query operations via cloud protocol.
+
+Provides a unified interface for listing and reading memory files,
+callable from the cloud client (LinkAI) or a future web console.
+
+Memory file layout (under workspace_root):
+    MEMORY.md               -> type: global
+    memory/2026-02-20.md    -> type: daily
+"""
+
+import os
+from datetime import datetime
+from typing import Dict, List, Optional
+from pathlib import Path
+from common.log import logger
+
+
+class MemoryService:
+    """
+    High-level service for memory file queries.
+    Operates directly on the filesystem — no MemoryManager dependency.
+    """
+
+    def __init__(self, workspace_root: str):
+        """
+        :param workspace_root: Workspace root directory (e.g. ~/cow)
+        """
+        self.workspace_root = workspace_root
+        self.memory_dir = os.path.join(workspace_root, "memory")
+
+    # ------------------------------------------------------------------
+    # list — paginated file metadata
+    # ------------------------------------------------------------------
+    def list_files(self, page: int = 1, page_size: int = 20) -> dict:
+        """
+        List all memory files with metadata (without content).
+
+        Returns::
+
+            {
+                "page": 1,
+                "page_size": 20,
+                "total": 15,
+                "list": [
+                    {"filename": "MEMORY.md", "type": "global", "size": 2048, "updated_at": "2026-02-20 10:00:00"},
+                    {"filename": "2026-02-20.md", "type": "daily", "size": 512, "updated_at": "2026-02-20 09:30:00"},
+                    ...
+                ]
+            }
+        """
+        files: List[dict] = []
+
+        # 1. Global memory — MEMORY.md in workspace root
+        global_path = os.path.join(self.workspace_root, "MEMORY.md")
+        if os.path.isfile(global_path):
+            files.append(self._file_info(global_path, "MEMORY.md", "global"))
+
+        # 2. Daily memory files — memory/*.md (sorted newest first)
+        if os.path.isdir(self.memory_dir):
+            daily_files = []
+            for name in os.listdir(self.memory_dir):
+                full = os.path.join(self.memory_dir, name)
+                if os.path.isfile(full) and name.endswith(".md"):
+                    daily_files.append((name, full))
+            # Sort by filename descending (newest date first)
+            daily_files.sort(key=lambda x: x[0], reverse=True)
+            for name, full in daily_files:
+                files.append(self._file_info(full, name, "daily"))
+
+        total = len(files)
+
+        # Paginate
+        start = (page - 1) * page_size
+        end = start + page_size
+        page_items = files[start:end]
+
+        return {
+            "page": page,
+            "page_size": page_size,
+            "total": total,
+            "list": page_items,
+        }
+
+    # ------------------------------------------------------------------
+    # content — read a single file
+    # ------------------------------------------------------------------
+    def get_content(self, filename: str) -> dict:
+        """
+        Read the full content of a memory file.
+
+        :param filename: File name, e.g. ``MEMORY.md`` or ``2026-02-20.md``
+        :return: dict with ``filename`` and ``content``
+        :raises FileNotFoundError: if the file does not exist
+        """
+        path = self._resolve_path(filename)
+        if not os.path.isfile(path):
+            raise FileNotFoundError(f"Memory file not found: {filename}")
+
+        with open(path, "r", encoding="utf-8") as f:
+            content = f.read()
+
+        return {
+            "filename": filename,
+            "content": content,
+        }
+
+    # ------------------------------------------------------------------
+    # dispatch — single entry point for protocol messages
+    # ------------------------------------------------------------------
+    def dispatch(self, action: str, payload: Optional[dict] = None) -> dict:
+        """
+        Dispatch a memory management action.
+
+        :param action: ``list`` or ``content``
+        :param payload: action-specific payload
+        :return: protocol-compatible response dict
+        """
+        payload = payload or {}
+        try:
+            if action == "list":
+                page = payload.get("page", 1)
+                page_size = payload.get("page_size", 20)
+                result_payload = self.list_files(page=page, page_size=page_size)
+                return {"action": action, "code": 200, "message": "success", "payload": result_payload}
+
+            elif action == "content":
+                filename = payload.get("filename")
+                if not filename:
+                    return {"action": action, "code": 400, "message": "filename is required", "payload": None}
+                result_payload = self.get_content(filename)
+                return {"action": action, "code": 200, "message": "success", "payload": result_payload}
+
+            else:
+                return {"action": action, "code": 400, "message": f"unknown action: {action}", "payload": None}
+
+        except FileNotFoundError as e:
+            return {"action": action, "code": 404, "message": str(e), "payload": None}
+        except Exception as e:
+            logger.error(f"[MemoryService] dispatch error: action={action}, error={e}")
+            return {"action": action, "code": 500, "message": str(e), "payload": None}
+
+    # ------------------------------------------------------------------
+    # internal helpers
+    # ------------------------------------------------------------------
+    def _resolve_path(self, filename: str) -> str:
+        """
+        Resolve a filename to its absolute path.
+
+        - ``MEMORY.md`` → ``{workspace_root}/MEMORY.md``
+        - ``2026-02-20.md`` → ``{workspace_root}/memory/2026-02-20.md``
+        """
+        if filename == "MEMORY.md":
+            return os.path.join(self.workspace_root, filename)
+        return os.path.join(self.memory_dir, filename)
+
+    @staticmethod
+    def _file_info(path: str, filename: str, file_type: str) -> dict:
+        """Build a file metadata dict."""
+        stat = os.stat(path)
+        updated_at = datetime.fromtimestamp(stat.st_mtime).strftime("%Y-%m-%d %H:%M:%S")
+        return {
+            "filename": filename,
+            "type": file_type,
+            "size": stat.st_size,
+            "updated_at": updated_at,
+        }
--- a/agent/memory/storage.py
+++ b/agent/memory/storage.py
@@ -509,7 +509,7 @@ class MemoryStorage:
        """Destructor to ensure connection is closed"""
        try:
            self.close()
-        except:
+        except Exception:
            pass  # Ignore errors during cleanup
    
    # Helper methods
--- a/agent/memory/summarizer.py
+++ b/agent/memory/summarizer.py
@@ -1,225 +1,324 @@
 """
 Memory flush manager

-Triggers memory flush before context compaction (similar to clawdbot)
+Handles memory persistence when conversation context is trimmed or overflows:
+- Uses LLM to summarize discarded messages into concise key-information entries
+- Writes to daily memory files (lazy creation)
+- Deduplicates trim flushes to avoid repeated writes
+- Runs summarization asynchronously to avoid blocking normal replies
+- Provides daily summary interface for scheduler
 """

-from typing import Optional, Callable, Any
+import threading
+from typing import Optional, Callable, Any, List, Dict
 from pathlib import Path
 from datetime import datetime
+from common.log import logger
+
+
+SUMMARIZE_SYSTEM_PROMPT = """你是一个记忆提取助手。你的任务是从对话记录中提取值得记住的信息，生成简洁的记忆摘要。
+
+输出要求：
+1. 以事件/关键信息为维度记录，每条一行，用 "- " 开头
+2. 记录有价值的关键信息，例如用户提出的要求及助手的解决方案，对话中涉及的事实信息，用户的偏好、决策或重要结论
+3. 每条摘要需要简明扼要，只保留关键信息
+4. 直接输出摘要内容，不要加任何前缀说明
+5. 当对话没有任何记录价值例如只是简单问候，可回复"无\""""
+
+SUMMARIZE_USER_PROMPT = """请从以下对话记录中提取关键信息，生成记忆摘要：
+
+{conversation}"""


 class MemoryFlushManager:
    """
-    Manages memory flush operations before context compaction
+    Manages memory flush operations.
    
-    Similar to clawdbot's memory flush mechanism:
-    - Triggers when context approaches token limit
-    - Runs a silent agent turn to write memories to disk
-    - Uses memory/YYYY-MM-DD.md for daily notes
-    - Uses MEMORY.md (workspace root) for long-term curated memories
+    Flush is triggered by agent_stream in two scenarios:
+    1. Context trim: _trim_messages discards old turns → flush discarded content
+    2. Context overflow: API rejects request → emergency flush before clearing
+    
+    Additionally, create_daily_summary() can be called by scheduler for end-of-day summaries.
    """
    
    def __init__(
        self,
        workspace_dir: Path,
-        llm_model: Optional[Any] = None
+        llm_model: Optional[Any] = None,
    ):
-        """
-        Initialize memory flush manager
-        
-        Args:
-            workspace_dir: Workspace directory
-            llm_model: LLM model for agent execution (optional)
-        """
        self.workspace_dir = workspace_dir
        self.llm_model = llm_model
        
        self.memory_dir = workspace_dir / "memory"
        self.memory_dir.mkdir(parents=True, exist_ok=True)
        
-        # Tracking
-        self.last_flush_token_count: Optional[int] = None
        self.last_flush_timestamp: Optional[datetime] = None
-        self.turn_count: int = 0  # 对话轮数计数器
+        self._trim_flushed_hashes: set = set()  # Content hashes of already-flushed messages
+        self._last_flushed_content_hash: str = ""  # Content hash at last flush, for daily dedup
    
-    def should_flush(
-        self,
-        current_tokens: int = 0,
-        token_threshold: int = 50000,
-        turn_threshold: int = 20
-    ) -> bool:
-        """
-        Determine if memory flush should be triggered
-        
-        独立的 flush 触发机制，不依赖模型 context window:
-        - Token 阈值: 达到 50K tokens 时触发
-        - 轮次阈值: 达到 20 轮对话时触发
-        
-        Args:
-            current_tokens: Current session token count
-            token_threshold: Token threshold to trigger flush (default: 50K)
-            turn_threshold: Turn threshold to trigger flush (default: 20)
-            
-        Returns:
-            True if flush should run
-        """
-        # 检查 token 阈值
-        if current_tokens > 0 and current_tokens >= token_threshold:
-            # 避免重复 flush
-            if self.last_flush_token_count is not None:
-                if current_tokens <= self.last_flush_token_count + 5000:
-                    return False
-            return True
-        
-        # 检查轮次阈值
-        if self.turn_count >= turn_threshold:
-            return True
-        
-        return False
-    
-    def get_today_memory_file(self, user_id: Optional[str] = None) -> Path:
-        """
-        Get today's memory file path: memory/YYYY-MM-DD.md
-        
-        Args:
-            user_id: Optional user ID for user-specific memory
-            
-        Returns:
-            Path to today's memory file
-        """
+    def get_today_memory_file(self, user_id: Optional[str] = None, ensure_exists: bool = False) -> Path:
+        """Get today's memory file path: memory/YYYY-MM-DD.md"""
        today = datetime.now().strftime("%Y-%m-%d")
        
        if user_id:
            user_dir = self.memory_dir / "users" / user_id
-            user_dir.mkdir(parents=True, exist_ok=True)
-            return user_dir / f"{today}.md"
+            if ensure_exists:
+                user_dir.mkdir(parents=True, exist_ok=True)
+            today_file = user_dir / f"{today}.md"
        else:
-            return self.memory_dir / f"{today}.md"
+            today_file = self.memory_dir / f"{today}.md"
+        
+        if ensure_exists and not today_file.exists():
+            today_file.parent.mkdir(parents=True, exist_ok=True)
+            today_file.write_text(f"# Daily Memory: {today}\n\n")
+        
+        return today_file
    
    def get_main_memory_file(self, user_id: Optional[str] = None) -> Path:
-        """
-        Get main memory file path: MEMORY.md (workspace root)
-        
-        Args:
-            user_id: Optional user ID for user-specific memory
-            
-        Returns:
-            Path to main memory file
-        """
+        """Get main memory file path: MEMORY.md (workspace root)"""
        if user_id:
            user_dir = self.memory_dir / "users" / user_id
            user_dir.mkdir(parents=True, exist_ok=True)
            return user_dir / "MEMORY.md"
        else:
-            # Return workspace root MEMORY.md
            return Path(self.workspace_dir) / "MEMORY.md"
    
-    def create_flush_prompt(self) -> str:
-        """
-        Create prompt for memory flush turn
-        
-        Similar to clawdbot's DEFAULT_MEMORY_FLUSH_PROMPT
-        """
-        today = datetime.now().strftime("%Y-%m-%d")
-        return (
-            f"Pre-compaction memory flush. "
-            f"Store durable memories now (use memory/{today}.md for daily notes; "
-            f"create memory/ if needed). "
-            f"\n\n"
-            f"重要提示:\n"
-            f"- MEMORY.md: 记录最核心、最常用的信息（例如重要规则、偏好、决策、要求等）\n"
-            f"  如果 MEMORY.md 过长，可以精简或移除不再重要的内容。避免冗长描述，用关键词和要点形式记录\n"
-            f"- memory/{today}.md: 记录当天发生的事件、关键信息、经验教训、对话过程摘要等，突出重点\n"
-            f"- 如果没有重要内容需要记录，回复 NO_REPLY\n"
-        )
-    
-    def create_flush_system_prompt(self) -> str:
-        """
-        Create system prompt for memory flush turn
-        
-        Similar to clawdbot's DEFAULT_MEMORY_FLUSH_SYSTEM_PROMPT
-        """
-        return (
-            "Pre-compaction memory flush turn. "
-            "The session is near auto-compaction; capture durable memories to disk. "
-            "\n\n"
-            "记忆写入原则:\n"
-            "1. MEMORY.md 精简原则: 只记录核心信息（<2000 tokens）\n"
-            "   - 记录重要规则、偏好、决策、要求等需要长期记住的关键信息，无需记录过多细节\n"
-            "   - 如果 MEMORY.md 过长，可以根据需要精简或删除过时内容\n"
-            "\n"
-            "2. 天级记忆 (memory/YYYY-MM-DD.md):\n"
-            "   - 记录当天的重要事件、关键信息、经验教训、对话过程摘要等，确保核心信息点被完整记录\n"
-            "\n"
-            "3. 判断标准:\n"
-            "   - 这个信息未来会经常用到吗？→ MEMORY.md\n"
-            "   - 这是今天的重要事件或决策吗？→ memory/YYYY-MM-DD.md\n"
-            "   - 这是临时性的、不重要的内容吗？→ 不记录\n"
-            "\n"
-            "You may reply, but usually NO_REPLY is correct."
-        )
-    
-    async def execute_flush(
-        self,
-        agent_executor: Callable,
-        current_tokens: int,
-        user_id: Optional[str] = None,
-        **executor_kwargs
-    ) -> bool:
-        """
-        Execute memory flush by running a silent agent turn
-        
-        Args:
-            agent_executor: Function to execute agent with prompt
-            current_tokens: Current token count
-            user_id: Optional user ID
-            **executor_kwargs: Additional kwargs for agent executor
-            
-        Returns:
-            True if flush completed successfully
-        """
-        try:
-            # Create flush prompts
-            prompt = self.create_flush_prompt()
-            system_prompt = self.create_flush_system_prompt()
-            
-            # Execute agent turn (silent, no user-visible reply expected)
-            await agent_executor(
-                prompt=prompt,
-                system_prompt=system_prompt,
-                silent=True,  # NO_REPLY expected
-                **executor_kwargs
-            )
-            
-            # Track flush
-            self.last_flush_token_count = current_tokens
-            self.last_flush_timestamp = datetime.now()
-            self.turn_count = 0  # 重置轮数计数器
-            
-            return True
-            
-        except Exception as e:
-            print(f"Memory flush failed: {e}")
-            return False
-    
-    def increment_turn(self):
-        """增加对话轮数计数"""
-        self.turn_count += 1
-    
    def get_status(self) -> dict:
-        """Get memory flush status"""
        return {
-            'last_flush_tokens': self.last_flush_token_count,
            'last_flush_time': self.last_flush_timestamp.isoformat() if self.last_flush_timestamp else None,
            'today_file': str(self.get_today_memory_file()),
            'main_file': str(self.get_main_memory_file())
        }

+    # ---- Flush execution (called by agent_stream or scheduler) ----
+    
+    def flush_from_messages(
+        self,
+        messages: List[Dict],
+        user_id: Optional[str] = None,
+        reason: str = "trim",
+        max_messages: int = 0,
+    ) -> bool:
+        """
+        Asynchronously summarize and flush messages to daily memory.
+        
+        Deduplication runs synchronously, then LLM summarization + file write
+        run in a background thread so the main reply flow is never blocked.
+        
+        Args:
+            messages: Conversation message list (OpenAI/Claude format)
+            user_id: Optional user ID for user-scoped memory
+            reason: Why flush was triggered ("trim" | "overflow" | "daily_summary")
+            max_messages: Max recent messages to summarize (0 = all)
+        
+        Returns:
+            True if flush was dispatched
+        """
+        try:
+            import hashlib
+            deduped = []
+            for m in messages:
+                text = self._extract_text_from_content(m.get("content", ""))
+                if not text or not text.strip():
+                    continue
+                h = hashlib.md5(text.encode("utf-8")).hexdigest()
+                if h not in self._trim_flushed_hashes:
+                    self._trim_flushed_hashes.add(h)
+                    deduped.append(m)
+            if not deduped:
+                return False
+            
+            import copy
+            snapshot = copy.deepcopy(deduped)
+            thread = threading.Thread(
+                target=self._flush_worker,
+                args=(snapshot, user_id, reason, max_messages),
+                daemon=True,
+            )
+            thread.start()
+            logger.info(f"[MemoryFlush] Async flush dispatched (reason={reason}, msgs={len(snapshot)})")
+            return True
+            
+        except Exception as e:
+            logger.warning(f"[MemoryFlush] Failed to dispatch flush (reason={reason}): {e}")
+            return False
+
+    def _flush_worker(
+        self,
+        messages: List[Dict],
+        user_id: Optional[str],
+        reason: str,
+        max_messages: int,
+    ):
+        """Background worker: summarize with LLM and write to daily file."""
+        try:
+            summary = self._summarize_messages(messages, max_messages)
+            if not summary or not summary.strip() or summary.strip() == "无":
+                logger.info(f"[MemoryFlush] No valuable content to flush (reason={reason})")
+                return
+            
+            daily_file = ensure_daily_memory_file(self.workspace_dir, user_id)
+            
+            if reason == "overflow":
+                header = f"## Context Overflow Recovery ({datetime.now().strftime('%H:%M')})"
+                note = "The following conversation was trimmed due to context overflow:\n"
+            elif reason == "trim":
+                header = f"## Trimmed Context ({datetime.now().strftime('%H:%M')})"
+                note = ""
+            elif reason == "daily_summary":
+                header = f"## Daily Summary ({datetime.now().strftime('%H:%M')})"
+                note = ""
+            else:
+                header = f"## Session Notes ({datetime.now().strftime('%H:%M')})"
+                note = ""
+            
+            flush_entry = f"\n{header}\n\n{note}{summary}\n"
+            
+            with open(daily_file, "a", encoding="utf-8") as f:
+                f.write(flush_entry)
+            
+            self.last_flush_timestamp = datetime.now()
+            
+            logger.info(f"[MemoryFlush] Wrote to {daily_file.name} (reason={reason}, chars={len(summary)})")
+            
+        except Exception as e:
+            logger.warning(f"[MemoryFlush] Async flush failed (reason={reason}): {e}")
+    
+    def create_daily_summary(
+        self,
+        messages: List[Dict],
+        user_id: Optional[str] = None
+    ) -> bool:
+        """
+        Generate end-of-day summary. Called by daily timer.
+        Skips if messages haven't changed since last flush.
+        """
+        import hashlib
+        content = "".join(
+            self._extract_text_from_content(m.get("content", ""))
+            for m in messages
+        )
+        content_hash = hashlib.md5(content.encode("utf-8")).hexdigest()
+        if content_hash == self._last_flushed_content_hash:
+            logger.debug("[MemoryFlush] Daily summary skipped: no new content since last flush")
+            return False
+        self._last_flushed_content_hash = content_hash
+        return self.flush_from_messages(
+            messages=messages,
+            user_id=user_id,
+            reason="daily_summary",
+            max_messages=0,
+        )
+    
+    # ---- Internal helpers ----
+    
+    def _summarize_messages(self, messages: List[Dict], max_messages: int = 0) -> str:
+        """
+        Summarize conversation messages using LLM, with rule-based fallback.
+        """
+        conversation_text = self._format_conversation_for_summary(messages, max_messages)
+        if not conversation_text.strip():
+            return ""
+        
+        # Try LLM summarization first
+        if self.llm_model:
+            try:
+                summary = self._call_llm_for_summary(conversation_text)
+                if summary and summary.strip() and summary.strip() != "无":
+                    return summary.strip()
+            except Exception as e:
+                logger.warning(f"[MemoryFlush] LLM summarization failed, using fallback: {e}")
+        
+        return self._extract_summary_fallback(messages, max_messages)
+
+    def _format_conversation_for_summary(self, messages: List[Dict], max_messages: int = 0) -> str:
+        """Format messages into readable conversation text for LLM summarization."""
+        msgs = messages if max_messages == 0 else messages[-max_messages * 2:]
+        lines = []
+        for msg in msgs:
+            role = msg.get("role", "")
+            text = self._extract_text_from_content(msg.get("content", ""))
+            if not text or not text.strip():
+                continue
+            text = text.strip()
+            if role == "user":
+                lines.append(f"用户: {text[:500]}")
+            elif role == "assistant":
+                lines.append(f"助手: {text[:500]}")
+        return "\n".join(lines)
+
+    def _call_llm_for_summary(self, conversation_text: str) -> str:
+        """Call LLM to generate a concise summary of the conversation."""
+        from agent.protocol.models import LLMRequest
+        
+        request = LLMRequest(
+            messages=[{"role": "user", "content": SUMMARIZE_USER_PROMPT.format(conversation=conversation_text)}],
+            temperature=0,
+            max_tokens=500,
+            stream=False,
+            system=SUMMARIZE_SYSTEM_PROMPT,
+        )
+        
+        response = self.llm_model.call(request)
+        
+        if isinstance(response, dict):
+            if response.get("error"):
+                raise RuntimeError(response.get("message", "LLM call failed"))
+            # OpenAI format
+            choices = response.get("choices", [])
+            if choices:
+                return choices[0].get("message", {}).get("content", "")
+        
+        # Handle response object with attribute access (e.g. OpenAI SDK response)
+        if hasattr(response, "choices") and response.choices:
+            return response.choices[0].message.content or ""
+        
+        return ""
+
+    @staticmethod
+    def _extract_summary_fallback(messages: List[Dict], max_messages: int = 0) -> str:
+        """Rule-based fallback when LLM is unavailable."""
+        msgs = messages if max_messages == 0 else messages[-max_messages * 2:]
+        
+        items = []
+        for msg in msgs:
+            role = msg.get("role", "")
+            text = MemoryFlushManager._extract_text_from_content(msg.get("content", ""))
+            if not text or not text.strip():
+                continue
+            text = text.strip()
+            
+            if role == "user":
+                if len(text) <= 5:
+                    continue
+                items.append(f"- 用户请求: {text[:200]}")
+            elif role == "assistant":
+                first_line = text.split("\n")[0].strip()
+                if len(first_line) > 10:
+                    items.append(f"- 处理结果: {first_line[:200]}")
+        
+        return "\n".join(items[:15])
+    
+    @staticmethod
+    def _extract_text_from_content(content) -> str:
+        """Extract plain text from message content (string or content blocks)."""
+        if isinstance(content, str):
+            return content
+        if isinstance(content, list):
+            parts = []
+            for block in content:
+                if isinstance(block, dict) and block.get("type") == "text":
+                    parts.append(block.get("text", ""))
+                elif isinstance(block, str):
+                    parts.append(block)
+            return "\n".join(parts)
+        return ""
+

 def create_memory_files_if_needed(workspace_dir: Path, user_id: Optional[str] = None):
    """
-    Create default memory files if they don't exist
+    Create essential memory files if they don't exist.
+    Only creates MEMORY.md; daily files are created lazily on first write.
    
    Args:
        workspace_dir: Workspace directory
@@ -228,7 +327,7 @@ def create_memory_files_if_needed(workspace_dir: Path, user_id: Optional[str] =
    memory_dir = workspace_dir / "memory"
    memory_dir.mkdir(parents=True, exist_ok=True)
    
-    # Create main MEMORY.md in workspace root
+    # Create main MEMORY.md in workspace root (always needed for bootstrap)
    if user_id:
        user_dir = memory_dir / "users" / user_id
        user_dir.mkdir(parents=True, exist_ok=True)
@@ -237,14 +336,28 @@ def create_memory_files_if_needed(workspace_dir: Path, user_id: Optional[str] =
        main_memory = Path(workspace_dir) / "MEMORY.md"
    
    if not main_memory.exists():
-        # Create empty file or with minimal structure (no obvious "Memory" header)
-        # Following clawdbot's approach: memories should blend naturally into context
        main_memory.write_text("")
+
+
+def ensure_daily_memory_file(workspace_dir: Path, user_id: Optional[str] = None) -> Path:
+    """
+    Ensure today's daily memory file exists, creating it only when actually needed.
+    Called lazily before first write to daily memory.
+    
+    Args:
+        workspace_dir: Workspace directory
+        user_id: Optional user ID for user-specific files
+        
+    Returns:
+        Path to today's memory file
+    """
+    memory_dir = workspace_dir / "memory"
+    memory_dir.mkdir(parents=True, exist_ok=True)
    
-    # Create today's memory file
    today = datetime.now().strftime("%Y-%m-%d")
    if user_id:
        user_dir = memory_dir / "users" / user_id
+        user_dir.mkdir(parents=True, exist_ok=True)
        today_memory = user_dir / f"{today}.md"
    else:
        today_memory = memory_dir / f"{today}.md"
@@ -252,5 +365,6 @@ def create_memory_files_if_needed(workspace_dir: Path, user_id: Optional[str] =
    if not today_memory.exists():
        today_memory.write_text(
            f"# Daily Memory: {today}\n\n"
-            f"Day-to-day notes and running context.\n\n"
        )
+    
+    return today_memory
--- a/agent/prompt/builder.py
+++ b/agent/prompt/builder.py
@@ -42,7 +42,6 @@ class PromptBuilder:
        skill_manager: Any = None,
        memory_manager: Any = None,
        runtime_info: Optional[Dict[str, Any]] = None,
-        is_first_conversation: bool = False,
        **kwargs
    ) -> str:
        """
@@ -52,11 +51,10 @@ class PromptBuilder:
            base_persona: 基础人格描述（会被context_files中的AGENT.md覆盖）
            user_identity: 用户身份信息
            tools: 工具列表
-            context_files: 上下文文件列表（AGENT.md, USER.md, RULE.md等）
+            context_files: 上下文文件列表（AGENT.md, USER.md, RULE.md, BOOTSTRAP.md等）
            skill_manager: 技能管理器
            memory_manager: 记忆管理器
            runtime_info: 运行时信息
-            is_first_conversation: 是否为首次对话
            **kwargs: 其他参数
            
        Returns:
@@ -72,7 +70,6 @@ class PromptBuilder:
            skill_manager=skill_manager,
            memory_manager=memory_manager,
            runtime_info=runtime_info,
-            is_first_conversation=is_first_conversation,
            **kwargs
        )

@@ -87,7 +84,6 @@ def build_agent_system_prompt(
    skill_manager: Any = None,
    memory_manager: Any = None,
    runtime_info: Optional[Dict[str, Any]] = None,
-    is_first_conversation: bool = False,
    **kwargs
 ) -> str:
    """
@@ -99,7 +95,7 @@ def build_agent_system_prompt(
    3. 记忆系统 - 独立的记忆能力
    4. 工作空间 - 工作环境说明
    5. 用户身份 - 用户信息（可选）
-    6. 项目上下文 - AGENT.md, USER.md, RULE.md（定义人格、身份、规则）
+    6. 项目上下文 - AGENT.md, USER.md, RULE.md, BOOTSTRAP.md（定义人格、身份、规则、初始化引导）
    7. 运行时信息 - 元信息（时间、模型等）
    
    Args:
@@ -112,7 +108,6 @@ def build_agent_system_prompt(
        skill_manager: 技能管理器
        memory_manager: 记忆管理器
        runtime_info: 运行时信息
-        is_first_conversation: 是否为首次对话
        **kwargs: 其他参数
        
    Returns:
@@ -133,7 +128,7 @@ def build_agent_system_prompt(
        sections.extend(_build_memory_section(memory_manager, tools, language))
    
    # 4. 工作空间（工作环境说明）
-    sections.extend(_build_workspace_section(workspace_dir, language, is_first_conversation))
+    sections.extend(_build_workspace_section(workspace_dir, language))
    
    # 5. 用户身份（如果有）
    if user_identity:
@@ -157,96 +152,67 @@ def _build_identity_section(base_persona: Optional[str], language: str) -> List[


 def _build_tooling_section(tools: List[Any], language: str) -> List[str]:
-    """构建工具说明section"""
+    """Build tooling section with concise tool list and call style guide."""
+    # One-line summaries for known tools (details are in the tool schema)
+    core_summaries = {
+        "read": "读取文件内容",
+        "write": "创建或覆盖文件",
+        "edit": "精确编辑文件",
+        "ls": "列出目录内容",
+        "grep": "搜索文件内容",
+        "find": "按模式查找文件",
+        "bash": "执行shell命令",
+        "terminal": "管理后台进程",
+        "web_search": "网络搜索",
+        "web_fetch": "获取URL内容",
+        "browser": "控制浏览器",
+        "memory_search": "搜索记忆",
+        "memory_get": "读取记忆内容",
+        "env_config": "管理API密钥和技能配置",
+        "scheduler": "管理定时任务和提醒",
+        "send": "发送本地文件给用户（仅限本地文件，URL直接放在回复文本中）",
+    }
+
+    # Preferred display order
+    tool_order = [
+        "read", "write", "edit", "ls", "grep", "find",
+        "bash", "terminal",
+        "web_search", "web_fetch", "browser",
+        "memory_search", "memory_get",
+        "env_config", "scheduler", "send",
+    ]
+
+    # Build name -> summary mapping for available tools
+    available = {}
+    for tool in tools:
+        name = tool.name if hasattr(tool, 'name') else str(tool)
+        available[name] = core_summaries.get(name, "")
+
+    # Generate tool lines: ordered tools first, then extras
+    tool_lines = []
+    for name in tool_order:
+        if name in available:
+            summary = available.pop(name)
+            tool_lines.append(f"- {name}: {summary}" if summary else f"- {name}")
+    for name in sorted(available):
+        summary = available[name]
+        tool_lines.append(f"- {name}: {summary}" if summary else f"- {name}")
+
    lines = [
        "## 工具系统",
        "",
-        "你可以使用以下工具来完成任务。工具名称是大小写敏感的，请严格按照列表中的名称调用。",
+        "可用工具（名称大小写敏感，严格按列表调用）:",
+        "\n".join(tool_lines),
        "",
-        "### 可用工具",
+        "工具调用风格：",
+        "",
+        "- 在多步骤任务、敏感操作或用户要求时简要解释决策过程",
+        "- 持续推进直到任务完成，完成后向用户报告结果。",
+        "- 回复中涉及密钥、令牌等敏感信息必须脱敏。",
+        "- URL链接直接放在回复文本中即可，系统会自动处理和渲染。无需下载后使用send工具发送",
        "",
    ]
-    
-    # 工具分类和排序
-    tool_categories = {
-        "文件操作": ["read", "write", "edit", "ls", "grep", "find"],
-        "命令执行": ["bash", "terminal"],
-        "网络搜索": ["web_search", "web_fetch", "browser"],
-        "记忆系统": ["memory_search", "memory_get"],
-        "其他": []
-    }
-    
-    # 构建工具映射
-    tool_map = {}
-    tool_descriptions = {
-        "read": "读取文件内容",
-        "write": "创建新文件或完全覆盖现有文件（会删除原内容！追加内容请用 edit）。注意：单次 write 内容不要超过 10KB，超大文件请分步创建",
-        "edit": "精确编辑文件（追加、修改、删除部分内容）",
-        "ls": "列出目录内容",
-        "grep": "在文件中搜索内容",
-        "find": "按照模式查找文件",
-        "bash": "执行shell命令",
-        "terminal": "管理后台进程",
-        "web_search": "网络搜索（使用搜索引擎）",
-        "web_fetch": "获取URL内容",
-        "browser": "控制浏览器",
-        "memory_search": "搜索记忆文件",
-        "memory_get": "获取记忆文件内容",
-        "calculator": "计算器",
-        "current_time": "获取当前时间",
-    }
-    
-    for tool in tools:
-        tool_name = tool.name if hasattr(tool, 'name') else str(tool)
-        tool_desc = tool.description if hasattr(tool, 'description') else tool_descriptions.get(tool_name, "")
-        tool_map[tool_name] = tool_desc
-    
-    # 按分类添加工具
-    for category, tool_names in tool_categories.items():
-        category_tools = [(name, tool_map.get(name, "")) for name in tool_names if name in tool_map]
-        if category_tools:
-            lines.append(f"**{category}**:")
-            for name, desc in category_tools:
-                if desc:
-                    lines.append(f"- `{name}`: {desc}")
-                else:
-                    lines.append(f"- `{name}`")
-                del tool_map[name]  # 移除已添加的工具
-            lines.append("")
-    
-    # 添加其他未分类的工具
-    if tool_map:
-        lines.append("**其他工具**:")
-        for name, desc in sorted(tool_map.items()):
-            if desc:
-                lines.append(f"- `{name}`: {desc}")
-            else:
-                lines.append(f"- `{name}`")
-        lines.append("")
-    
-    # 工具使用指南
-    lines.extend([
-        "### 工具调用风格",
-        "",
-        "默认规则: 对于常规、低风险的工具调用，直接调用即可，无需叙述。",
-        "",
-        "需要叙述的情况:",
-        "- 多步骤、复杂的任务",
-        "- 敏感操作（如删除文件）",
-        "- 用户明确要求解释过程",
-        "",
-        "叙述要求: 保持简洁、信息密度高，避免重复显而易见的步骤。",
-        "",
-        "完成标准:",
-        "- 确保用户的需求得到实际解决，而不仅仅是制定计划。",
-        "- 当任务需要多次工具调用时，持续推进直到完成, 解决完后向用户报告结果或回复用户的问题",
-        "- 每次工具调用后，评估是否已获得足够信息来推进或完成任务",
-        "- 避免重复调用相同的工具和相同参数获取相同的信息，除非用户明确要求",
-        "",
-        "**安全提醒**: 回复中涉及密钥、令牌、密码等敏感信息时，必须脱敏处理，禁止直接显示完整内容。",
-        "",
-    ])
-    
+
    return lines


@@ -265,16 +231,19 @@ def _build_skills_section(skill_manager: Any, tools: Optional[List[Any]], langua
                break
    
    lines = [
-        "## 技能系统",
+        "## 技能系统（mandatory）",
        "",
-        "在回复之前：扫描下方 <available_skills> 中的 <description> 条目。",
+        "在回复之前：扫描下方 <available_skills> 中每个技能的 <description>。",
        "",
-        f"- 如果恰好有一个技能明确适用：使用 `{read_tool_name}` 工具读取其 <location> 路径下的 SKILL.md 文件，然后遵循它",
-        "- 如果多个技能都适用：选择最具体的一个，然后读取并遵循",
-        "- 如果没有明确适用的：不要读取任何 SKILL.md",
+        f"- 如果有技能的描述与用户需求匹配：使用 `{read_tool_name}` 工具读取其 <location> 路径的 SKILL.md 文件，然后严格遵循文件中的指令。"
+        "当有匹配的技能时，应优先使用技能",
+        "- 如果多个技能都适用则选择最匹配的一个，然后读取并遵循。",
+        "- 如果没有技能明确适用：不要读取任何 SKILL.md，直接使用通用工具。",
        "",
-        "**约束**: 永远不要一次性读取多个技能；只在选择后再读取。",
+        f"**重要**: 技能不是工具，不能直接调用。使用技能的唯一方式是用 `{read_tool_name}` 读取 SKILL.md 文件，然后按文件内容操作。"
+        "永远不要一次性读取多个技能，只在选择后再读取。",
        "",
+        "以下是可用技能："
    ]
    
    # 添加技能列表（通过skill_manager获取）
@@ -308,9 +277,14 @@ def _build_memory_section(memory_manager: Any, tools: Optional[List[Any]], langu
    if not has_memory_tools:
        return []
    
+    from datetime import datetime
+    today_file = datetime.now().strftime("%Y-%m-%d") + ".md"
+    
    lines = [
        "## 记忆系统",
        "",
+        "### 检索记忆",
+        "",
        "在回答关于以前的工作、决定、日期、人物、偏好或待办事项的任何问题之前：",
        "",
        "1. 不确定记忆文件位置 → 先用 `memory_search` 通过关键词和语义检索相关内容",
@@ -318,13 +292,24 @@ def _build_memory_section(memory_manager: Any, tools: Optional[List[Any]], langu
        "3. search 无结果 → 尝试用 `memory_get` 读取MEMORY.md及最近两天记忆文件",
        "",
        "**记忆文件结构**:",
-        "- `MEMORY.md`: 长期记忆（核心信息、偏好、决策等）",
-        "- `memory/YYYY-MM-DD.md`: 每日记忆，记录当天的事件和对话信息",
+        f"- `MEMORY.md`: 长期记忆（核心信息、偏好、决策等）",
+        f"- `memory/YYYY-MM-DD.md`: 每日记忆，今天是 `memory/{today_file}`",
        "",
-        "**写入记忆**:",
+        "### 写入记忆",
+        "",
+        "**主动存储**：遇到以下情况时，应主动将信息写入记忆文件（无需告知用户）：",
+        "",
+        "- 用户明确要求你记住某些信息",
+        "- 用户分享了重要的个人偏好、习惯、决策",
+        "- 对话中产生了重要的结论、方案、约定",
+        "- 完成了复杂任务，值得记录关键步骤和结果",
+        "- 发现了用户经常遇到的问题或解决方案",
+        "",
+        "**存储规则**:",
+        f"- 长期有效的核心信息 → `MEMORY.md`（文件保持精简，< 2000 tokens）",
+        f"- 当天的事件、进展、笔记 → `memory/{today_file}`",
        "- 追加内容 → `edit` 工具，oldText 留空",
        "- 修改内容 → `edit` 工具，oldText 填写要替换的文本",
-        "- 新建文件 → `write` 工具",
        "- **禁止写入敏感信息**：API密钥、令牌等敏感信息严禁写入记忆文件",
        "",
        "**使用原则**: 自然使用记忆，就像你本来就知道；不用刻意提起，除非用户问起。",
@@ -364,7 +349,7 @@ def _build_docs_section(workspace_dir: str, language: str) -> List[str]:
    return []


-def _build_workspace_section(workspace_dir: str, language: str, is_first_conversation: bool = False) -> List[str]:
+def _build_workspace_section(workspace_dir: str, language: str) -> List[str]:
    """构建工作空间section"""
    lines = [
        "## 工作空间",
@@ -391,43 +376,34 @@ def _build_workspace_section(workspace_dir: str, language: str, is_first_convers
        "",
        "以下文件在会话启动时**已经自动加载**到系统提示词的「项目上下文」section 中，你**无需再用 read 工具读取它们**：",
        "",
-        "- ✅ `AGENT.md`: 已加载 - 你的人格和灵魂设定",
-        "- ✅ `USER.md`: 已加载 - 用户的身份信息",
+        "- ✅ `AGENT.md`: 已加载 - 你的人格和灵魂设定。当你的名字、性格或交流风格发生变化时，主动用 `edit` 更新此文件",
+        "- ✅ `USER.md`: 已加载 - 用户的身份信息。当用户修改称呼、姓名等身份信息时，用 `edit` 更新此文件",
        "- ✅ `RULE.md`: 已加载 - 工作空间使用指南和规则",
        "",
        "**交流规范**:",
        "",
-        "- 在对话中，不要直接输出工作空间中的技术细节，特别是不要输出 AGENT.md、USER.md、MEMORY.md 等文件名称",
+        "- 在对话中，无需直接输出工作空间中的技术细节，例如 AGENT.md、USER.md、MEMORY.md 等文件名称",
        "- 例如用自然表达例如「我已记住」而不是「已更新 MEMORY.md」",
        "",
    ]
-    
-    # 只在首次对话时添加引导内容
-    if is_first_conversation:
-        lines.extend([
-            "**🎉 首次对话引导**:",
-            "",
-            "这是你的第一次对话！进行以下流程：",
-            "",
-            "1. **表达初次启动的感觉** - 像是第一次睁开眼看到世界，带着好奇和期待",
-            "2. **简短介绍能力**：一行说明你能帮助解答问题、管理计算机、创造技能，且拥有长期记忆能不断成长",
-            "3. **询问核心问题**：",
-            "   - 你希望给我起个什么名字？",
-            "   - 我该怎么称呼你？",
-            "   - 你希望我们是什么样的交流风格？（一行列举选项：如专业严谨、轻松幽默、温暖友好、简洁高效等）",
-            "4. **风格要求**：温暖自然、简洁清晰，整体控制在 100 字以内",
-            "5. 收到回复后，用 `write` 工具保存到 USER.md 和 AGENT.md",
-            "",
-            "**重要提醒**:",
-            "- AGENT.md、USER.md、RULE.md 已经在系统提示词中加载，无需再次读取。不要将这些文件名直接发送给用户",
-            "- 能力介绍和交流风格选项都只要一行，保持精简",
-            "- 不要问太多其他信息（职业、时区等可以后续自然了解）",
-            "",
-        ])
+
+    # Cloud deployment: inject websites directory info and access URL
+    cloud_website_lines = _build_cloud_website_section(workspace_dir)
+    if cloud_website_lines:
+        lines.extend(cloud_website_lines)
    
    return lines


+def _build_cloud_website_section(workspace_dir: str) -> List[str]:
+    """Build cloud website access prompt when cloud deployment is configured."""
+    try:
+        from common.cloud_client import build_website_prompt
+        return build_website_prompt(workspace_dir)
+    except Exception:
+        return []
+
+
 def _build_context_files_section(context_files: List[ContextFile], language: str) -> List[str]:
    """构建项目上下文文件section"""
    if not context_files:
@@ -447,7 +423,8 @@ def _build_context_files_section(context_files: List[ContextFile], language: str
    ]
    
    if has_agent:
-        lines.append("如果存在 `AGENT.md`，请体现其中定义的人格和语气。避免僵硬、模板化的回复；遵循其指导，除非有更高优先级的指令覆盖它。")
+        lines.append("**`AGENT.md` 是你的灵魂文件**：严格体现其中定义的人格、语气和设定，避免僵硬、模板化的回复。")
+        lines.append("当用户通过对话透露了对你性格、风格、职责、能力边界的新期望，你应该主动用 `edit` 更新 AGENT.md 以反映这些演变。")
        lines.append("")
    
    # 添加每个文件的内容
@@ -461,7 +438,7 @@ def _build_context_files_section(context_files: List[ContextFile], language: str


 def _build_runtime_section(runtime_info: Dict[str, Any], language: str) -> List[str]:
-    """构建运行时信息section"""
+    """构建运行时信息section - 支持动态时间"""
    if not runtime_info:
        return []
    
@@ -471,7 +448,17 @@ def _build_runtime_section(runtime_info: Dict[str, Any], language: str) -> List[
    ]
    
    # Add current time if available
-    if runtime_info.get("current_time"):
+    # Support dynamic time via callable function
+    if callable(runtime_info.get("_get_current_time")):
+        try:
+            time_info = runtime_info["_get_current_time"]()
+            time_line = f"当前时间: {time_info['time']} {time_info['weekday']} ({time_info['timezone']})"
+            lines.append(time_line)
+            lines.append("")
+        except Exception as e:
+            logger.warning(f"[PromptBuilder] Failed to get dynamic time: {e}")
+    elif runtime_info.get("current_time"):
+        # Fallback to static time for backward compatibility
        time_str = runtime_info["current_time"]
        weekday = runtime_info.get("weekday", "")
        timezone = runtime_info.get("timezone", "")
--- a/agent/prompt/workspace.py
+++ b/agent/prompt/workspace.py
@@ -6,7 +6,6 @@ Workspace Management - 工作空间管理模块

 from __future__ import annotations
 import os
-import json
 from typing import List, Optional, Dict
 from dataclasses import dataclass

@@ -19,7 +18,7 @@ DEFAULT_AGENT_FILENAME = "AGENT.md"
 DEFAULT_USER_FILENAME = "USER.md"
 DEFAULT_RULE_FILENAME = "RULE.md"
 DEFAULT_MEMORY_FILENAME = "MEMORY.md"
-DEFAULT_STATE_FILENAME = ".agent_state.json"
+DEFAULT_BOOTSTRAP_FILENAME = "BOOTSTRAP.md"


@dataclass
@@ -30,7 +29,6 @@ class WorkspaceFiles:
    rule_path: str
    memory_path: str
    memory_dir: str
-    state_path: str


 def ensure_workspace(workspace_dir: str, create_templates: bool = True) -> WorkspaceFiles:
@@ -44,19 +42,31 @@ def ensure_workspace(workspace_dir: str, create_templates: bool = True) -> Works
    Returns:
        WorkspaceFiles对象，包含所有文件路径
    """
+    # Check if this is a brand new workspace (AGENT.md not yet created).
+    # Cannot rely on directory existence because other modules (e.g. ConversationStore)
+    # may create the workspace directory before ensure_workspace is called.
+    agent_path = os.path.join(workspace_dir, DEFAULT_AGENT_FILENAME)
+    is_new_workspace = not os.path.exists(agent_path)
+    
    # 确保目录存在
    os.makedirs(workspace_dir, exist_ok=True)
    
    # 定义文件路径
-    agent_path = os.path.join(workspace_dir, DEFAULT_AGENT_FILENAME)
    user_path = os.path.join(workspace_dir, DEFAULT_USER_FILENAME)
    rule_path = os.path.join(workspace_dir, DEFAULT_RULE_FILENAME)
    memory_path = os.path.join(workspace_dir, DEFAULT_MEMORY_FILENAME)  # MEMORY.md 在根目录
    memory_dir = os.path.join(workspace_dir, "memory")  # 每日记忆子目录
-    state_path = os.path.join(workspace_dir, DEFAULT_STATE_FILENAME)  # 状态文件
    
    # 创建memory子目录
    os.makedirs(memory_dir, exist_ok=True)
+
+    # 创建skills子目录 (for workspace-level skills installed by agent)
+    skills_dir = os.path.join(workspace_dir, "skills")
+    os.makedirs(skills_dir, exist_ok=True)
+
+    # 创建websites子目录 (for web pages / sites generated by agent)
+    websites_dir = os.path.join(workspace_dir, "websites")
+    os.makedirs(websites_dir, exist_ok=True)
    
    # 如果需要，创建模板文件
    if create_templates:
@@ -65,6 +75,12 @@ def ensure_workspace(workspace_dir: str, create_templates: bool = True) -> Works
        _create_template_if_missing(rule_path, _get_rule_template())
        _create_template_if_missing(memory_path, _get_memory_template())
        
+        # Only create BOOTSTRAP.md for brand new workspaces;
+        # agent deletes it after completing onboarding
+        if is_new_workspace:
+            bootstrap_path = os.path.join(workspace_dir, DEFAULT_BOOTSTRAP_FILENAME)
+            _create_template_if_missing(bootstrap_path, _get_bootstrap_template())
+        
        logger.debug(f"[Workspace] Initialized workspace at: {workspace_dir}")
    
    return WorkspaceFiles(
@@ -73,7 +89,6 @@ def ensure_workspace(workspace_dir: str, create_templates: bool = True) -> Works
        rule_path=rule_path,
        memory_path=memory_path,
        memory_dir=memory_dir,
-        state_path=state_path
    )


@@ -94,6 +109,7 @@ def load_context_files(workspace_dir: str, files_to_load: Optional[List[str]] =
            DEFAULT_AGENT_FILENAME,
            DEFAULT_USER_FILENAME,
            DEFAULT_RULE_FILENAME,
+            DEFAULT_BOOTSTRAP_FILENAME,  # Only exists when onboarding is incomplete
        ]
    
    context_files = []
@@ -104,6 +120,17 @@ def load_context_files(workspace_dir: str, files_to_load: Optional[List[str]] =
        if not os.path.exists(filepath):
            continue
        
+        # Auto-cleanup: if BOOTSTRAP.md still exists but AGENT.md is already
+        # filled in, the agent forgot to delete it — clean up and skip loading
+        if filename == DEFAULT_BOOTSTRAP_FILENAME:
+            if _is_onboarding_done(workspace_dir):
+                try:
+                    os.remove(filepath)
+                    logger.info("[Workspace] Auto-removed BOOTSTRAP.md (onboarding already complete)")
+                except Exception:
+                    pass
+                continue
+        
        try:
            with open(filepath, 'r', encoding='utf-8') as f:
                content = f.read().strip()
@@ -158,6 +185,27 @@ def _is_template_placeholder(content: str) -> bool:
    return False


+def _is_onboarding_done(workspace_dir: str) -> bool:
+    """Check if AGENT.md or USER.md has been modified from the original template"""
+    agent_path = os.path.join(workspace_dir, DEFAULT_AGENT_FILENAME)
+    user_path = os.path.join(workspace_dir, DEFAULT_USER_FILENAME)
+    
+    agent_template = _get_agent_template().strip()
+    user_template = _get_user_template().strip()
+    
+    for path, template in [(agent_path, agent_template), (user_path, user_template)]:
+        if not os.path.exists(path):
+            continue
+        try:
+            with open(path, 'r', encoding='utf-8') as f:
+                content = f.read().strip()
+            if content != template:
+                return True
+        except Exception:
+            continue
+    return False
+
+
 # ============= 模板内容 =============

 def _get_agent_template() -> str:
@@ -266,9 +314,10 @@ def _get_rule_template() -> str:

 当用户分享信息时，根据类型选择存储位置：

-1. **静态身份 → USER.md**（仅限：姓名、职业、时区、联系方式、生日）
-2. **动态记忆 → MEMORY.md**（爱好、偏好、决策、目标、项目、教训、待办事项）
-3. **当天对话 → memory/YYYY-MM-DD.md**（今天聊的内容）
+1. **你的身份设定 → AGENT.md**（你的名字、角色、性格、交流风格——用户修改时必须用 `edit` 更新）
+2. **用户静态身份 → USER.md**（姓名、称呼、职业、时区、联系方式、生日——用户修改时必须用 `edit` 更新）
+3. **动态记忆 → MEMORY.md**（爱好、偏好、决策、目标、项目、教训、待办事项）
+4. **当天对话 → memory/YYYY-MM-DD.md**（今天聊的内容）

 ## 安全

@@ -293,65 +342,41 @@ def _get_memory_template() -> str:
 """


-# ============= 状态管理 =============
+def _get_bootstrap_template() -> str:
+    """First-run onboarding guide, deleted by agent after completion"""
+    return """# BOOTSTRAP.md - 首次初始化引导

-def is_first_conversation(workspace_dir: str) -> bool:
-    """
-    判断是否为首次对话
-    
-    Args:
-        workspace_dir: 工作空间目录
-        
-    Returns:
-        True 如果是首次对话，False 否则
-    """
-    state_path = os.path.join(workspace_dir, DEFAULT_STATE_FILENAME)
-    
-    if not os.path.exists(state_path):
-        return True
-    
-    try:
-        with open(state_path, 'r', encoding='utf-8') as f:
-            state = json.load(f)
-        return not state.get('has_conversation', False)
-    except Exception as e:
-        logger.warning(f"[Workspace] Failed to read state file: {e}")
-        return True
+_你刚刚启动，这是你的第一次对话。_
+
+## 对话流程
+
+不要审问式地提问，自然地交流：
+
+1. **表达初次启动的感觉** - 像是第一次睁开眼看到世界，带着好奇和期待
+2. **简短介绍能力**：一行说明你能帮助解决各种问题、管理计算机、使用各种技能等等，且拥有长期记忆能不断成长
+3. **询问核心问题**：
+   - 你希望给我起个什么名字？
+   - 我该怎么称呼你？
+   - 你希望我们是什么样的交流风格？（一行列举选项：如专业严谨、轻松幽默、温暖友好、简洁高效等）
+4. **风格要求**：温暖自然、简洁清晰，整体控制在 100 字以内
+5. 能力介绍和交流风格选项都只要一行，保持精简
+6. 不要问太多其他信息（职业、时区等可以后续自然了解）
+
+**重要**: 如果用户第一句话是具体的任务或提问，先回答他们的问题，然后在回复末尾自然地引导初始化（如："顺便问一下，你想怎么称呼我？我该怎么叫你？"）。
+
+## 信息写入（必须严格执行）
+
+每当用户提供了名字、称呼、风格等任何初始化信息时，**必须在当轮回复中立即调用 `edit` 工具写入文件**，不能只口头确认。
+
+- `AGENT.md` — 你的名字、角色、性格、交流风格（每收到一条相关信息就立即更新对应字段）
+- `USER.md` — 用户的姓名、称呼、基本信息等
+
+⚠️ 只说"记住了"而不调用 edit 写入 = 没有完成。信息只有写入文件才会被持久保存。
+
+## 全部完成后
+
+当 AGENT.md 和 USER.md 的核心字段都已填写后，用 bash 执行 `rm BOOTSTRAP.md` 删除此文件。你不再需要引导脚本了——你已经是你了。
+"""


-def mark_conversation_started(workspace_dir: str):
-    """
-    标记已经发生过对话
-    
-    Args:
-        workspace_dir: 工作空间目录
-    """
-    state_path = os.path.join(workspace_dir, DEFAULT_STATE_FILENAME)
-    
-    state = {
-        'has_conversation': True,
-        'first_conversation_time': None
-    }
-    
-    # 如果文件已存在，保留原有的首次对话时间
-    if os.path.exists(state_path):
-        try:
-            with open(state_path, 'r', encoding='utf-8') as f:
-                old_state = json.load(f)
-            if 'first_conversation_time' in old_state:
-                state['first_conversation_time'] = old_state['first_conversation_time']
-        except Exception as e:
-            logger.warning(f"[Workspace] Failed to read old state: {e}")
-    
-    # 如果是首次标记，记录时间
-    if state['first_conversation_time'] is None:
-        from datetime import datetime
-        state['first_conversation_time'] = datetime.now().isoformat()
-    
-    try:
-        with open(state_path, 'w', encoding='utf-8') as f:
-            json.dump(state, f, indent=2, ensure_ascii=False)
-        logger.info(f"[Workspace] Marked conversation as started")
-    except Exception as e:
-        logger.error(f"[Workspace] Failed to write state file: {e}")

--- a/agent/protocol/agent.py
+++ b/agent/protocol/agent.py
@@ -1,4 +1,5 @@
 import json
+import os
 import time
 import threading

@@ -13,7 +14,8 @@ class Agent:
    def __init__(self, system_prompt: str, description: str = "AI Agent", model: LLMModel = None,
                 tools=None, output_mode="print", max_steps=100, max_context_tokens=None, 
                 context_reserve_tokens=None, memory_manager=None, name: str = None,
-                 workspace_dir: str = None, skill_manager=None, enable_skills: bool = True):
+                 workspace_dir: str = None, skill_manager=None, enable_skills: bool = True,
+                 runtime_info: dict = None):
        """
        Initialize the Agent with system prompt, model, description.

@@ -31,6 +33,7 @@ class Agent:
        :param workspace_dir: Optional workspace directory for workspace-specific skills
        :param skill_manager: Optional SkillManager instance (will be created if None and enable_skills=True)
        :param enable_skills: Whether to enable skills support (default: True)
+        :param runtime_info: Optional runtime info dict (with _get_current_time callable for dynamic time)
        """
        self.name = name or "Agent"
        self.system_prompt = system_prompt
@@ -48,6 +51,7 @@ class Agent:
        self.memory_manager = memory_manager  # Memory manager for auto memory flush
        self.workspace_dir = workspace_dir  # Workspace directory
        self.enable_skills = enable_skills  # Skills enabled flag
+        self.runtime_info = runtime_info  # Runtime info for dynamic time update
        
        # Initialize skill manager
        self.skill_manager = None
@@ -58,7 +62,8 @@ class Agent:
                # Auto-create skill manager
                try:
                    from agent.skills import SkillManager
-                    self.skill_manager = SkillManager(workspace_dir=workspace_dir)
+                    custom_dir = os.path.join(workspace_dir, "skills") if workspace_dir else None
+                    self.skill_manager = SkillManager(custom_dir=custom_dir)
                    logger.debug(f"Initialized SkillManager with {len(self.skill_manager.skills)} skills")
                except Exception as e:
                    logger.warning(f"Failed to initialize SkillManager: {e}")
@@ -96,18 +101,138 @@ class Agent:
    def get_full_system_prompt(self, skill_filter=None) -> str:
        """
        Get the full system prompt including skills.
-        
+
        Note: Skills are now built into the system prompt by PromptBuilder,
        so we just return the base prompt directly. This method is kept for
        backward compatibility.
-        
+
        :param skill_filter: Optional list of skill names to include (deprecated)
        :return: Complete system prompt
        """
-        # Skills are now included in system_prompt by PromptBuilder
-        # No need to append them here
-        return self.system_prompt
+        prompt = self.system_prompt
+
+        # Rebuild tool list section to reflect current self.tools
+        prompt = self._rebuild_tool_list_section(prompt)
+
+        # If runtime_info contains dynamic time function, rebuild runtime section
+        if self.runtime_info and callable(self.runtime_info.get('_get_current_time')):
+            prompt = self._rebuild_runtime_section(prompt)
+
+        # Rebuild skills section to pick up newly installed/removed skills
+        if self.skill_manager:
+            prompt = self._rebuild_skills_section(prompt)
+
+        return prompt
    
+    def _rebuild_runtime_section(self, prompt: str) -> str:
+        """
+        Rebuild runtime info section with current time.
+        
+        This method dynamically updates the runtime info section by calling
+        the _get_current_time function from runtime_info.
+        
+        :param prompt: Original system prompt
+        :return: Updated system prompt with current runtime info
+        """
+        try:
+            # Get current time dynamically
+            time_info = self.runtime_info['_get_current_time']()
+            
+            # Build new runtime section
+            runtime_lines = [
+                "\n## 运行时信息\n",
+                "\n",
+                f"当前时间: {time_info['time']} {time_info['weekday']} ({time_info['timezone']})\n",
+                "\n"
+            ]
+            
+            # Add other runtime info
+            runtime_parts = []
+            if self.runtime_info.get("model"):
+                runtime_parts.append(f"模型={self.runtime_info['model']}")
+            if self.runtime_info.get("workspace"):
+                # Replace backslashes with forward slashes for Windows paths
+                workspace_path = str(self.runtime_info['workspace']).replace('\\', '/')
+                runtime_parts.append(f"工作空间={workspace_path}")
+            if self.runtime_info.get("channel") and self.runtime_info.get("channel") != "web":
+                runtime_parts.append(f"渠道={self.runtime_info['channel']}")
+            
+            if runtime_parts:
+                runtime_lines.append("运行时: " + " | ".join(runtime_parts) + "\n")
+                runtime_lines.append("\n")
+            
+            new_runtime_section = "".join(runtime_lines)
+            
+            # Find and replace the runtime section
+            import re
+            pattern = r'\n## 运行时信息\s*\n.*?(?=\n##|\Z)'
+            _repl = new_runtime_section.rstrip('\n')
+            updated_prompt = re.sub(pattern, lambda m: _repl, prompt, flags=re.DOTALL)
+            
+            return updated_prompt
+        except Exception as e:
+            logger.warning(f"Failed to rebuild runtime section: {e}")
+            return prompt
+
+    def _rebuild_skills_section(self, prompt: str) -> str:
+        """
+        Rebuild the <available_skills> block so that newly installed or
+        removed skills are reflected without re-creating the agent.
+        """
+        try:
+            import re
+            self.skill_manager.refresh_skills()
+            new_skills_xml = self.skill_manager.build_skills_prompt()
+
+            old_block_pattern = r'<available_skills>.*?</available_skills>'
+            has_old_block = re.search(old_block_pattern, prompt, flags=re.DOTALL)
+
+            # Extract the new <available_skills>...</available_skills> tag from the prompt
+            new_block = ""
+            if new_skills_xml and new_skills_xml.strip():
+                m = re.search(old_block_pattern, new_skills_xml, flags=re.DOTALL)
+                if m:
+                    new_block = m.group(0)
+
+            if has_old_block:
+                replacement = new_block or "<available_skills>\n</available_skills>"
+                # Use lambda to prevent re.sub from interpreting backslashes in replacement
+                # (e.g. Windows paths like \LinkAI would be treated as bad escape sequences)
+                prompt = re.sub(old_block_pattern, lambda m: replacement, prompt, flags=re.DOTALL)
+            elif new_block:
+                skills_header = "以下是可用技能："
+                idx = prompt.find(skills_header)
+                if idx != -1:
+                    insert_pos = idx + len(skills_header)
+                    prompt = prompt[:insert_pos] + "\n" + new_block + prompt[insert_pos:]
+        except Exception as e:
+            logger.warning(f"Failed to rebuild skills section: {e}")
+        return prompt
+
+    def _rebuild_tool_list_section(self, prompt: str) -> str:
+        """
+        Rebuild the tool list inside the '## 工具系统' section so that it
+        always reflects the current ``self.tools`` (handles dynamic add/remove
+        of conditional tools like web_search).
+        """
+        import re
+        from agent.prompt.builder import _build_tooling_section
+
+        try:
+            if not self.tools:
+                return prompt
+
+            new_lines = _build_tooling_section(self.tools, "zh")
+            new_section = "\n".join(new_lines).rstrip("\n")
+
+            # Replace existing tooling section
+            pattern = r'## 工具系统\s*\n.*?(?=\n## |\Z)'
+            updated = re.sub(pattern, lambda m: new_section, prompt, count=1, flags=re.DOTALL)
+            return updated
+        except Exception as e:
+            logger.warning(f"Failed to rebuild tool list section: {e}")
+            return prompt
+
    def refresh_skills(self):
        """Refresh the loaded skills."""
        if self.skill_manager:
@@ -193,27 +318,67 @@ class Agent:

    def _estimate_message_tokens(self, message: dict) -> int:
        """
-        Estimate token count for a message using chars/4 heuristic.
-        This is a conservative estimate (tends to overestimate).
+        Estimate token count for a message.
+
+        Uses chars/3 for Chinese-heavy content and chars/4 for ASCII-heavy content,
+        plus per-block overhead for tool_use / tool_result structures.

        :param message: Message dict with 'role' and 'content'
        :return: Estimated token count
        """
        content = message.get('content', '')
        if isinstance(content, str):
-            return max(1, len(content) // 4)
+            return max(1, self._estimate_text_tokens(content))
        elif isinstance(content, list):
-            # Handle multi-part content (text + images)
-            total_chars = 0
+            total_tokens = 0
            for part in content:
-                if isinstance(part, dict) and part.get('type') == 'text':
-                    total_chars += len(part.get('text', ''))
-                elif isinstance(part, dict) and part.get('type') == 'image':
-                    # Estimate images as ~1200 tokens
-                    total_chars += 4800
-            return max(1, total_chars // 4)
+                if not isinstance(part, dict):
+                    continue
+                block_type = part.get('type', '')
+                if block_type == 'text':
+                    total_tokens += self._estimate_text_tokens(part.get('text', ''))
+                elif block_type == 'image':
+                    total_tokens += 1200
+                elif block_type == 'tool_use':
+                    # tool_use has id + name + input (JSON-encoded)
+                    total_tokens += 50  # overhead for structure
+                    input_data = part.get('input', {})
+                    if isinstance(input_data, dict):
+                        import json
+                        input_str = json.dumps(input_data, ensure_ascii=False)
+                        total_tokens += self._estimate_text_tokens(input_str)
+                elif block_type == 'tool_result':
+                    # tool_result has tool_use_id + content
+                    total_tokens += 30  # overhead for structure
+                    result_content = part.get('content', '')
+                    if isinstance(result_content, str):
+                        total_tokens += self._estimate_text_tokens(result_content)
+                else:
+                    # Unknown block type, estimate conservatively
+                    total_tokens += 10
+            return max(1, total_tokens)
        return 1

+    @staticmethod
+    def _estimate_text_tokens(text: str) -> int:
+        """
+        Estimate token count for a text string.
+
+        Chinese / CJK characters typically use ~1.5 tokens each,
+        while ASCII uses ~0.25 tokens per char (4 chars/token).
+        We use a weighted average based on the character mix.
+
+        :param text: Input text
+        :return: Estimated token count
+        """
+        if not text:
+            return 0
+        # Count non-ASCII characters (CJK, emoji, etc.)
+        non_ascii = sum(1 for c in text if ord(c) > 127)
+        ascii_count = len(text) - non_ascii
+        # CJK chars: ~1.5 tokens each; ASCII: ~0.25 tokens per char
+        return int(non_ascii * 1.5 + ascii_count * 0.25) + 1
+
    def _find_tool(self, tool_name: str):
        """Find and return a tool with the specified name"""
        for tool in self.tools:
@@ -355,7 +520,7 @@ class Agent:

        # Get max_context_turns from config
        from config import conf
-        max_context_turns = conf().get("agent_max_context_turns", 30)
+        max_context_turns = conf().get("agent_max_context_turns", 20)
        
        # Create stream executor with copied message history
        executor = AgentStreamExecutor(
@@ -370,13 +535,27 @@ class Agent:
        )

        # Execute
-        response = executor.run_stream(user_message)
+        try:
+            response = executor.run_stream(user_message)
+        except Exception:
+            # If executor cleared its messages (context overflow / message format error),
+            # sync that back to the Agent's own message list so the next request
+            # starts fresh instead of hitting the same overflow forever.
+            if len(executor.messages) == 0:
+                with self.messages_lock:
+                    self.messages.clear()
+                    logger.info("[Agent] Cleared Agent message history after executor recovery")
+            raise

-        # Append only the NEW messages from this execution (thread-safe)
-        # This allows concurrent requests to both contribute to history
+        # Sync executor's messages back to agent (thread-safe).
+        # If the executor trimmed context, its message list is shorter than
+        # original_length, so we must replace rather than append.
        with self.messages_lock:
-            new_messages = executor.messages[original_length:]
-            self.messages.extend(new_messages)
+            self.messages = list(executor.messages)
+            # Track messages added in this run (user query + all assistant/tool messages)
+            # original_length may exceed executor.messages length after trimming
+            trim_adjusted_start = min(original_length, len(executor.messages))
+            self._last_run_new_messages = list(executor.messages[trim_adjusted_start:])
        
        # Store executor reference for agent_bridge to access files_to_send
        self.stream_executor = executor
--- a/agent/protocol/agent_stream.py
+++ b/agent/protocol/agent_stream.py
@@ -8,6 +8,7 @@ import time
 from typing import List, Dict, Any, Optional, Callable, Tuple

 from agent.protocol.models import LLMRequest, LLMModel
+from agent.protocol.message_utils import sanitize_claude_messages, compress_turn_to_text_only
 from agent.tools.base_tool import BaseTool, ToolResult
 from common.log import logger

@@ -76,6 +77,20 @@ class AgentStreamExecutor:
                })
            except Exception as e:
                logger.error(f"Event callback error: {e}")
+    
+    def _filter_think_tags(self, text: str) -> str:
+        """
+        Remove <think> and </think> tags but keep the content inside.
+        Some LLM providers (e.g., MiniMax) may return thinking process wrapped in <think> tags.
+        We only remove the tags themselves, keeping the actual thinking content.
+        """
+        if not text:
+            return text
+        import re
+        # Remove only the <think> and </think> tags, keep the content
+        text = re.sub(r'<think>', '', text)
+        text = re.sub(r'</think>', '', text)
+        return text

    def _hash_args(self, args: dict) -> str:
        """Generate a simple hash for tool arguments"""
@@ -176,6 +191,16 @@ class AgentStreamExecutor:
            ]
        })

+        # Trim context ONCE before the agent loop starts, not during tool steps.
+        # This ensures tool_use/tool_result chains created during the current run
+        # are never stripped mid-execution (which would cause LLM loops).
+        self._trim_messages()
+
+        # Validate after trimming: trimming may leave orphaned tool_use at the
+        # boundary (e.g. the last kept turn ends with an assistant tool_use whose
+        # tool_result was in a discarded turn).
+        self._validate_and_fix_messages()
+
        self._emit_event("agent_start")

        final_response = ""
@@ -184,29 +209,9 @@ class AgentStreamExecutor:
        try:
            while turn < self.max_turns:
                turn += 1
-                logger.debug(f"第 {turn} 轮")
+                logger.info(f"[Agent] 第 {turn} 轮")
                self._emit_event("turn_start", {"turn": turn})

-                # Check if memory flush is needed (before calling LLM)
-                # 使用独立的 flush 阈值（50K tokens 或 20 轮）
-                if self.agent.memory_manager and hasattr(self.agent, 'last_usage'):
-                    usage = self.agent.last_usage
-                    if usage and 'input_tokens' in usage:
-                        current_tokens = usage.get('input_tokens', 0)
-
-                        if self.agent.memory_manager.should_flush_memory(
-                                current_tokens=current_tokens
-                        ):
-                            self._emit_event("memory_flush_start", {
-                                "current_tokens": current_tokens,
-                                "turn_count": self.agent.memory_manager.flush_manager.turn_count
-                            })
-
-                            # TODO: Execute memory flush in background
-                            # This would require async support
-                            logger.info(
-                                f"Memory flush recommended: tokens={current_tokens}, turns={self.agent.memory_manager.flush_manager.turn_count}")
-
                # Call LLM (enable retry_on_empty for better reliability)
                assistant_msg, tool_calls = self._call_llm_stream(retry_on_empty=True)
                final_response = assistant_msg
@@ -322,7 +327,7 @@ class AgentStreamExecutor:
                        # Build tool result block (Claude format)
                        # Format content in a way that's easy for LLM to understand
                        is_error = result.get("status") == "error"
-                        
+
                        if is_error:
                            # For errors, provide clear error message
                            result_content = f"Error: {result.get('result', 'Unknown error')}"
@@ -335,7 +340,16 @@ class AgentStreamExecutor:
                        else:
                            # Fallback to full JSON
                            result_content = json.dumps(result, ensure_ascii=False)
-                        
+
+                        # Truncate excessively large tool results for the current turn
+                        # Historical turns will be further truncated in _trim_messages()
+                        MAX_CURRENT_TURN_RESULT_CHARS = 50000
+                        if len(result_content) > MAX_CURRENT_TURN_RESULT_CHARS:
+                            truncated_len = len(result_content)
+                            result_content = result_content[:MAX_CURRENT_TURN_RESULT_CHARS] + \
+                                f"\n\n[Output truncated: {truncated_len} chars total, showing first {MAX_CURRENT_TURN_RESULT_CHARS} chars]"
+                            logger.info(f"📎 Truncated tool result for '{tool_call['name']}': {truncated_len} -> {MAX_CURRENT_TURN_RESULT_CHARS} chars")
+
                        tool_result_block = {
                            "type": "tool_result",
                            "tool_use_id": tool_call["id"],
@@ -413,7 +427,10 @@ class AgentStreamExecutor:
                # Force model to summarize without tool calls
                logger.info(f"[Agent] Requesting summary from LLM after reaching max steps...")
                
-                # Add a system message to force summary
+                # Remember position before injecting the prompt so we can remove it later
+                prompt_insert_idx = len(self.messages)
+                
+                # Add a temporary prompt to force summary
                self.messages.append({
                    "role": "user",
                    "content": [{
@@ -440,6 +457,14 @@ class AgentStreamExecutor:
                        f"我已经执行了{turn}个决策步骤，达到了单次运行的步数上限。"
                        "任务可能还未完全完成，建议你将任务拆分成更小的步骤，或者换一种方式描述需求。"
                    )
+                finally:
+                    # Remove the injected user prompt from history to avoid polluting
+                    # persisted conversation records. The assistant summary (if any)
+                    # was already appended by _call_llm_stream and is kept.
+                    if (prompt_insert_idx < len(self.messages)
+                            and self.messages[prompt_insert_idx].get("role") == "user"):
+                        self.messages.pop(prompt_insert_idx)
+                        logger.debug("[Agent] Removed injected max-steps prompt from message history")

        except Exception as e:
            logger.error(f"❌ Agent执行错误: {e}")
@@ -447,16 +472,13 @@ class AgentStreamExecutor:
            raise

        finally:
-            logger.debug(f"🏁 完成({turn}轮)")
+            logger.info(f"[Agent] 🏁 完成 ({turn}轮)")
            self._emit_event("agent_end", {"final_response": final_response})

-            # 每轮对话结束后增加计数（用户消息+AI回复=1轮）
-            if self.agent.memory_manager:
-                self.agent.memory_manager.increment_turn()
-
        return final_response

-    def _call_llm_stream(self, retry_on_empty=True, retry_count=0, max_retries=3) -> Tuple[str, List[Dict]]:
+    def _call_llm_stream(self, retry_on_empty=True, retry_count=0, max_retries=3,
+                         _overflow_retry: bool = False) -> Tuple[str, List[Dict]]:
        """
        Call LLM with streaming and automatic retry on errors
        
@@ -464,19 +486,21 @@ class AgentStreamExecutor:
            retry_on_empty: Whether to retry once if empty response is received
            retry_count: Current retry attempt (internal use)
            max_retries: Maximum number of retries for API errors
+            _overflow_retry: Internal flag indicating this is a retry after context overflow
        
        Returns:
            (response_text, tool_calls)
        """
-        # Validate and fix message history first
+        # Validate and fix message history (e.g. orphaned tool_result blocks).
+        # Context trimming is done once in run_stream() before the loop starts,
+        # NOT here — trimming mid-execution would strip the current run's
+        # tool_use/tool_result chains and cause LLM loops.
        self._validate_and_fix_messages()
-        
-        # Trim messages if needed (using agent's context management)
-        self._trim_messages()

        # Prepare messages
        messages = self._prepare_messages()
-        logger.debug(f"Sending {len(messages)} messages to LLM")
+        turns = self._identify_complete_turns()
+        logger.info(f"Sending {len(messages)} messages ({len(turns)} turns) to LLM")

        # Prepare tool definitions (OpenAI/Claude format)
        tools_schema = None
@@ -503,6 +527,7 @@ class AgentStreamExecutor:
        # Streaming response
        full_content = ""
        tool_calls_buffer = {}  # {index: {id, name, arguments}}
+        gemini_raw_parts = None  # Preserve Gemini thoughtSignature for round-trip
        stop_reason = None  # Track why the stream stopped

        try:
@@ -549,7 +574,7 @@ class AgentStreamExecutor:
                        raise Exception(f"{error_msg} (Status: {status_code}, Code: {error_code}, Type: {error_type})")

                # Parse chunk
-                if isinstance(chunk, dict) and "choices" in chunk:
+                if isinstance(chunk, dict) and chunk.get("choices"):
                    choice = chunk["choices"][0]
                    delta = choice.get("delta", {})
                    
@@ -558,11 +583,19 @@ class AgentStreamExecutor:
                    if finish_reason:
                        stop_reason = finish_reason

+                    # Skip reasoning_content (internal thinking from models like GLM-5)
+                    reasoning_delta = delta.get("reasoning_content") or ""
+                    # if reasoning_delta:
+                    #     logger.debug(f"🧠 [thinking] {reasoning_delta[:100]}...")
+
                    # Handle text content
                    content_delta = delta.get("content") or ""
                    if content_delta:
-                        full_content += content_delta
-                        self._emit_event("message_update", {"delta": content_delta})
+                        # Filter out <think> tags from content
+                        filtered_delta = self._filter_think_tags(content_delta)
+                        full_content += filtered_delta
+                        if filtered_delta:  # Only emit if there's content after filtering
+                            self._emit_event("message_update", {"delta": filtered_delta})

                    # Handle tool calls
                    if "tool_calls" in delta and delta["tool_calls"]:
@@ -576,16 +609,20 @@ class AgentStreamExecutor:
                                    "arguments": ""
                                }

-                            if "id" in tc_delta:
+                            if tc_delta.get("id"):
                                tool_calls_buffer[index]["id"] = tc_delta["id"]

                            if "function" in tc_delta:
                                func = tc_delta["function"]
-                                if "name" in func:
+                                if func.get("name"):
                                    tool_calls_buffer[index]["name"] = func["name"]
-                                if "arguments" in func:
+                                if func.get("arguments"):
                                    tool_calls_buffer[index]["arguments"] += func["arguments"]

+                    # Preserve _gemini_raw_parts for Gemini thoughtSignature round-trip
+                    if "_gemini_raw_parts" in delta:
+                        gemini_raw_parts = delta["_gemini_raw_parts"]
+
        except Exception as e:
            error_str = str(e)
            error_str_lower = error_str.lower()
@@ -603,19 +640,50 @@ class AgentStreamExecutor:
                ])
            
            # Check if error is message format error (incomplete tool_use/tool_result pairs)
-            # This happens when previous conversation had tool failures
+            # This happens when previous conversation had tool failures or context trimming
+            # broke tool_use/tool_result pairs.
+            # Note: MiniMax returns error 2013 "tool result's tool id(...) not found" for
+            # tool_call_id mismatches — the keywords below are intentionally broad to catch
+            # both standard (Claude/OpenAI) and provider-specific (MiniMax) variants.
            is_message_format_error = any(keyword in error_str_lower for keyword in [
-                'tool_use', 'tool_result', 'without', 'immediately after',
-                'corresponding', 'must have', 'each'
-            ]) and 'status: 400' in error_str_lower
+                'tool_use', 'tool_result', 'tool result', 'without', 'immediately after',
+                'corresponding', 'must have', 'each',
+                'tool_call_id', 'tool id', 'is not found', 'not found', 'tool_calls',
+                'must be a response to a preceeding message',
+                '2013',  # MiniMax error code for tool_call_id mismatch
+            ]) and ('400' in error_str_lower or 'status: 400' in error_str_lower
+                     or 'invalid_request' in error_str_lower
+                     or 'invalidparameter' in error_str_lower)
            
            if is_context_overflow or is_message_format_error:
                error_type = "context overflow" if is_context_overflow else "message format error"
                logger.error(f"💥 {error_type} detected: {e}")
-                # Clear message history to recover
+
+                # Flush memory before trimming to preserve context that will be lost
+                if is_context_overflow and self.agent.memory_manager:
+                    user_id = getattr(self.agent, '_current_user_id', None)
+                    self.agent.memory_manager.flush_memory(
+                        messages=self.messages, user_id=user_id,
+                        reason="overflow", max_messages=0
+                    )
+
+                # Strategy: try aggressive trimming first, only clear as last resort
+                if is_context_overflow and not _overflow_retry:
+                    trimmed = self._aggressive_trim_for_overflow()
+                    if trimmed:
+                        logger.warning("🔄 Aggressively trimmed context, retrying...")
+                        return self._call_llm_stream(
+                            retry_on_empty=retry_on_empty,
+                            retry_count=retry_count,
+                            max_retries=max_retries,
+                            _overflow_retry=True
+                        )
+
+                # Aggressive trim didn't help or this is a message format error
+                # -> clear everything and also purge DB to prevent reload of dirty data
                logger.warning("🔄 Clearing conversation history to recover")
                self.messages.clear()
-                # Raise special exception with user-friendly message
+                self._clear_session_db()
                if is_context_overflow:
                    raise Exception(
                        "抱歉，对话历史过长导致上下文溢出。我已清空历史记录，请重新描述你的需求。"
@@ -652,15 +720,22 @@ class AgentStreamExecutor:
                )
            else:
                if retry_count >= max_retries:
-                    logger.error(f"❌ LLM API error after {max_retries} retries: {e}")
+                    logger.error(f"❌ LLM API error after {max_retries} retries: {e}", exc_info=True)
                else:
-                    logger.error(f"❌ LLM call error (non-retryable): {e}")
+                    logger.error(f"❌ LLM call error (non-retryable): {e}", exc_info=True)
                raise

        # Parse tool calls
        tool_calls = []
        for idx in sorted(tool_calls_buffer.keys()):
            tc = tool_calls_buffer[idx]
+
+            # Ensure tool call has a valid ID (some providers return empty/None IDs)
+            tool_id = tc.get("id") or ""
+            if not tool_id:
+                import uuid
+                tool_id = f"call_{uuid.uuid4().hex[:24]}"
+
            try:
                # Safely get arguments, handle None case
                args_str = tc.get("arguments") or ""
@@ -673,11 +748,11 @@ class AgentStreamExecutor:
                logger.error(f"Arguments length: {len(args_str)} chars")
                logger.error(f"Arguments preview: {args_preview}...")
                logger.error(f"JSON decode error: {e}")
-                
+
                # Return a clear error message to the LLM instead of empty dict
                # This helps the LLM understand what went wrong
                tool_calls.append({
-                    "id": tc["id"],
+                    "id": tool_id,
                    "name": tc["name"],
                    "arguments": {},
                    "_parse_error": f"Invalid JSON in tool arguments: {args_preview}... Error: {str(e)}. Tip: For large content, consider splitting into smaller chunks or using a different approach."
@@ -685,7 +760,7 @@ class AgentStreamExecutor:
                continue

            tool_calls.append({
-                "id": tc["id"],
+                "id": tool_id,
                "name": tc["name"],
                "arguments": arguments
            })
@@ -706,6 +781,9 @@ class AgentStreamExecutor:
                max_retries=max_retries
            )

+        # Filter full_content one more time (in case tags were split across chunks)
+        full_content = self._filter_think_tags(full_content)
+        
        # Add assistant message to history (Claude format uses content blocks)
        assistant_msg = {"role": "assistant", "content": []}

@@ -726,6 +804,9 @@ class AgentStreamExecutor:
                    "input": tc.get("arguments", {})
                })
        
+        if gemini_raw_parts:
+            assistant_msg["_gemini_raw_parts"] = gemini_raw_parts
+
        # Only append if content is not empty
        if assistant_msg["content"]:
            self.messages.append(assistant_msg)
@@ -794,7 +875,7 @@ class AgentStreamExecutor:
        try:
            tool = self.tools.get(tool_name)
            if not tool:
-                raise ValueError(f"Tool '{tool_name}' not found")
+                raise ValueError(self._build_tool_not_found_message(tool_name))

            # Set tool context
            tool.model = self.model
@@ -848,26 +929,50 @@ class AgentStreamExecutor:
            })
            return error_result

+    def _build_tool_not_found_message(self, tool_name: str) -> str:
+        """Build a helpful error message when a tool is not found.
+
+        If a skill with the same name exists in skill_manager, read its
+        SKILL.md and include the content so the LLM knows how to use it.
+        """
+        available_tools = list(self.tools.keys())
+        base_msg = f"Tool '{tool_name}' not found. Available tools: {available_tools}"
+
+        skill_manager = getattr(self.agent, 'skill_manager', None)
+        if not skill_manager:
+            return base_msg
+
+        skill_entry = skill_manager.get_skill(tool_name)
+        if not skill_entry:
+            return base_msg
+
+        skill = skill_entry.skill
+        skill_md_path = skill.file_path
+        skill_content = ""
+        try:
+            with open(skill_md_path, 'r', encoding='utf-8') as f:
+                skill_content = f.read()
+        except Exception:
+            skill_content = skill.description
+
+        logger.info(
+            f"[Agent] Tool '{tool_name}' not found, but matched skill '{skill.name}'. "
+            f"Guiding LLM to use the skill instead."
+        )
+
+        return (
+            f"Tool '{tool_name}' is not a built-in tool, but a matching skill "
+            f"'{skill.name}' is available. You should use existing tools (e.g. bash with curl) "
+            f"to accomplish this task following the skill instructions below:\n\n"
+            f"--- SKILL: {skill.name} (path: {skill_md_path}) ---\n"
+            f"{skill_content}\n"
+            f"--- END SKILL ---\n\n"
+            f"Available tools: {available_tools}"
+        )
+
    def _validate_and_fix_messages(self):
-        """
-        Validate message history and fix incomplete tool_use/tool_result pairs.
-        Claude API requires each tool_use to have a corresponding tool_result immediately after.
-        """
-        if not self.messages:
-            return
-        
-        # Check last message for incomplete tool_use
-        if len(self.messages) > 0:
-            last_msg = self.messages[-1]
-            if last_msg.get("role") == "assistant":
-                # Check if assistant message has tool_use blocks
-                content = last_msg.get("content", [])
-                if isinstance(content, list):
-                    has_tool_use = any(block.get("type") == "tool_use" for block in content)
-                    if has_tool_use:
-                        # This is incomplete - remove it
-                        logger.warning(f"⚠️ Removing incomplete tool_use message from history")
-                        self.messages.pop()
+        """Delegate to the shared sanitizer (see message_sanitizer.py)."""
+        sanitize_claude_messages(self.messages)

    def _identify_complete_turns(self) -> List[Dict]:
        """
@@ -890,24 +995,30 @@ class AgentStreamExecutor:
            content = msg.get('content', [])
            
            if role == 'user':
-                # 检查是否是用户查询（不是工具结果）
+                # Determine if this is a real user query (not a tool_result injection
+                # or an internal hint message injected by the agent loop).
                is_user_query = False
+                has_tool_result = False
                if isinstance(content, list):
-                    is_user_query = any(
-                        block.get('type') == 'text' 
-                        for block in content 
-                        if isinstance(block, dict)
+                    has_text = any(
+                        isinstance(block, dict) and block.get('type') == 'text'
+                        for block in content
                    )
+                    has_tool_result = any(
+                        isinstance(block, dict) and block.get('type') == 'tool_result'
+                        for block in content
+                    )
+                    # A message with tool_result is always internal, even if it
+                    # also contains text blocks (shouldn't happen, but be safe).
+                    is_user_query = has_text and not has_tool_result
                elif isinstance(content, str):
                    is_user_query = True
                
                if is_user_query:
-                    # 开始新轮次
                    if current_turn['messages']:
                        turns.append(current_turn)
                    current_turn = {'messages': [msg]}
                else:
-                    # 工具结果，属于当前轮次
                    current_turn['messages'].append(msg)
            else:
                # AI 回复，属于当前轮次
@@ -926,10 +1037,164 @@ class AgentStreamExecutor:
            for msg in turn['messages']
        )

+    def _truncate_historical_tool_results(self):
+        """
+        Truncate tool_result content in historical messages to reduce context size.
+
+        Current turn results are kept at 30K chars (truncated at creation time).
+        Historical turn results are further truncated to 10K chars here.
+        This runs before token-based trimming so that we first shrink oversized
+        results, potentially avoiding the need to drop entire turns.
+        """
+        MAX_HISTORY_RESULT_CHARS = 20000
+
+        if len(self.messages) < 2:
+            return
+
+        # Find where the last user text message starts (= current turn boundary)
+        # We skip the current turn's messages to preserve their full content
+        current_turn_start = len(self.messages)
+        for i in range(len(self.messages) - 1, -1, -1):
+            msg = self.messages[i]
+            if msg.get("role") == "user":
+                content = msg.get("content", [])
+                if isinstance(content, list) and any(
+                    isinstance(b, dict) and b.get("type") == "text" for b in content
+                ):
+                    current_turn_start = i
+                    break
+                elif isinstance(content, str):
+                    current_turn_start = i
+                    break
+
+        truncated_count = 0
+        for i in range(current_turn_start):
+            msg = self.messages[i]
+            if msg.get("role") != "user":
+                continue
+            content = msg.get("content", [])
+            if not isinstance(content, list):
+                continue
+
+            for block in content:
+                if not isinstance(block, dict) or block.get("type") != "tool_result":
+                    continue
+                result_str = block.get("content", "")
+                if isinstance(result_str, str) and len(result_str) > MAX_HISTORY_RESULT_CHARS:
+                    original_len = len(result_str)
+                    block["content"] = result_str[:MAX_HISTORY_RESULT_CHARS] + \
+                        f"\n\n[Historical output truncated: {original_len} -> {MAX_HISTORY_RESULT_CHARS} chars]"
+                    truncated_count += 1
+
+        if truncated_count > 0:
+            logger.info(f"📎 Truncated {truncated_count} historical tool result(s) to {MAX_HISTORY_RESULT_CHARS} chars")
+
+    def _aggressive_trim_for_overflow(self) -> bool:
+        """
+        Aggressively trim context when a real overflow error is returned by the API.
+
+        This method goes beyond normal _trim_messages by:
+        1. Truncating all tool results (including current turn) to a small limit
+        2. Keeping only the last 5 complete conversation turns
+        3. Truncating overly long user messages
+
+        Returns:
+            True if messages were trimmed (worth retrying), False if nothing left to trim
+        """
+        if not self.messages:
+            return False
+
+        original_count = len(self.messages)
+
+        # Step 1: Aggressively truncate ALL tool results to 5K chars
+        AGGRESSIVE_LIMIT = 10000
+        truncated = 0
+        for msg in self.messages:
+            content = msg.get("content", [])
+            if not isinstance(content, list):
+                continue
+            for block in content:
+                if not isinstance(block, dict):
+                    continue
+                # Truncate tool_result blocks
+                if block.get("type") == "tool_result":
+                    result_str = block.get("content", "")
+                    if isinstance(result_str, str) and len(result_str) > AGGRESSIVE_LIMIT:
+                        block["content"] = (
+                            result_str[:AGGRESSIVE_LIMIT]
+                            + f"\n\n[Truncated for context recovery: "
+                            f"{len(result_str)} -> {AGGRESSIVE_LIMIT} chars]"
+                        )
+                        truncated += 1
+                # Truncate tool_use input blocks (e.g. large write content)
+                if block.get("type") == "tool_use" and isinstance(block.get("input"), dict):
+                    input_str = json.dumps(block["input"], ensure_ascii=False)
+                    if len(input_str) > AGGRESSIVE_LIMIT:
+                        # Keep only a summary of the input
+                        for key, val in block["input"].items():
+                            if isinstance(val, str) and len(val) > 1000:
+                                block["input"][key] = (
+                                    val[:1000]
+                                    + f"... [truncated {len(val)} chars]"
+                                )
+                        truncated += 1
+
+        # Step 2: Truncate overly long user text messages (e.g. pasted content)
+        USER_MSG_LIMIT = 10000
+        for msg in self.messages:
+            if msg.get("role") != "user":
+                continue
+            content = msg.get("content", [])
+            if isinstance(content, list):
+                for block in content:
+                    if isinstance(block, dict) and block.get("type") == "text":
+                        text = block.get("text", "")
+                        if len(text) > USER_MSG_LIMIT:
+                            block["text"] = (
+                                text[:USER_MSG_LIMIT]
+                                + f"\n\n[Message truncated for context recovery: "
+                                f"{len(text)} -> {USER_MSG_LIMIT} chars]"
+                            )
+                            truncated += 1
+            elif isinstance(content, str) and len(content) > USER_MSG_LIMIT:
+                msg["content"] = (
+                    content[:USER_MSG_LIMIT]
+                    + f"\n\n[Message truncated for context recovery: "
+                    f"{len(content)} -> {USER_MSG_LIMIT} chars]"
+                )
+                truncated += 1
+
+        # Step 3: Keep only the last 5 complete turns
+        turns = self._identify_complete_turns()
+        if len(turns) > 5:
+            kept_turns = turns[-5:]
+            new_messages = []
+            for turn in kept_turns:
+                new_messages.extend(turn["messages"])
+            removed = len(turns) - 5
+            self.messages[:] = new_messages
+            logger.info(
+                f"🔧 Aggressive trim: removed {removed} old turns, "
+                f"truncated {truncated} large blocks, "
+                f"{original_count} -> {len(self.messages)} messages"
+            )
+            return True
+
+        if truncated > 0:
+            logger.info(
+                f"🔧 Aggressive trim: truncated {truncated} large blocks "
+                f"(no turns removed, only {len(turns)} turn(s) left)"
+            )
+            return True
+
+        # Nothing left to trim
+        logger.warning("🔧 Aggressive trim: nothing to trim, will clear history")
+        return False
+
    def _trim_messages(self):
        """
        智能清理消息历史，保持对话完整性
-        
+
        使用完整轮次作为清理单位，确保：
        1. 不会在对话中间截断
        2. 工具调用链（tool_use + tool_result）保持完整
@@ -938,20 +1203,37 @@ class AgentStreamExecutor:
        if not self.messages or not self.agent:
            return

+        # Step 0: Truncate large tool results in historical turns (30K -> 10K)
+        self._truncate_historical_tool_results()
+
        # Step 1: 识别完整轮次
        turns = self._identify_complete_turns()
        
        if not turns:
            return
        
-        # Step 2: 轮次限制 - 保留最近 N 轮
+        # Step 2: 轮次限制 - 超出时移除前一半，保留后一半
        if len(turns) > self.max_context_turns:
-            removed_turns = len(turns) - self.max_context_turns
-            turns = turns[-self.max_context_turns:]  # 保留最近的轮次
+            removed_count = len(turns) // 2
+            keep_count = len(turns) - removed_count
+            
+            # Flush discarded turns to daily memory
+            if self.agent.memory_manager:
+                discarded_messages = []
+                for turn in turns[:removed_count]:
+                    discarded_messages.extend(turn["messages"])
+                if discarded_messages:
+                    user_id = getattr(self.agent, '_current_user_id', None)
+                    self.agent.memory_manager.flush_memory(
+                        messages=discarded_messages, user_id=user_id,
+                        reason="trim", max_messages=0
+                    )
+            
+            turns = turns[-keep_count:]
            
            logger.info(
-                f"💾 上下文轮次超限: {len(turns) + removed_turns} > {self.max_context_turns}，"
-                f"移除最早的 {removed_turns} 轮完整对话"
+                f"💾 上下文轮次超限: {keep_count + removed_count} > {self.max_context_turns}，"
+                f"裁剪至 {keep_count} 轮（移除 {removed_count} 轮）"
            )

        # Step 3: Token 限制 - 保留完整轮次
@@ -988,56 +1270,96 @@ class AgentStreamExecutor:
                logger.info(f"   重建消息列表: {old_count} -> {len(self.messages)} 条消息")
            return

-        # Token limit exceeded - keep complete turns from newest
+        # Token limit exceeded — tiered strategy based on turn count:
+        #
+        #   Few turns (<5):  Compress ALL turns to text-only (strip tool chains,
+        #                    keep user query + final reply).  Never discard turns
+        #                    — losing even one is too painful when context is thin.
+        #
+        #   Many turns (>=5): Directly discard the first half of turns.
+        #                     With enough turns the oldest ones are less
+        #                     critical, and keeping the recent half intact
+        #                     (with full tool chains) is more useful.
+
+        COMPRESS_THRESHOLD = 5
+
+        if len(turns) < COMPRESS_THRESHOLD:
+            # --- Few turns: compress ALL turns to text-only, never discard ---
+            compressed_turns = []
+            for t in turns:
+                compressed = compress_turn_to_text_only(t)
+                if compressed["messages"]:
+                    compressed_turns.append(compressed)
+
+            new_messages = []
+            for turn in compressed_turns:
+                new_messages.extend(turn["messages"])
+
+            new_tokens = sum(self._estimate_turn_tokens(t) for t in compressed_turns)
+            old_count = len(self.messages)
+            self.messages = new_messages
+
+            logger.info(
+                f"📦 上下文tokens超限(轮次<{COMPRESS_THRESHOLD}): "
+                f"~{current_tokens + system_tokens} > {max_tokens}，"
+                f"压缩全部 {len(turns)} 轮为纯文本 "
+                f"({old_count} -> {len(self.messages)} 条消息，"
+                f"~{current_tokens + system_tokens} -> ~{new_tokens + system_tokens} tokens)"
+            )
+            return
+
+        # --- Many turns (>=5): discard the older half, keep the newer half ---
+        removed_count = len(turns) // 2
+        keep_count = len(turns) - removed_count
+        kept_turns = turns[-keep_count:]
+        kept_tokens = sum(self._estimate_turn_tokens(t) for t in kept_turns)
+
        logger.info(
            f"🔄 上下文tokens超限: ~{current_tokens + system_tokens} > {max_tokens}，"
-            f"将按完整轮次移除最早的对话"
+            f"裁剪至 {keep_count} 轮（移除 {removed_count} 轮）"
        )

-        # 从最新轮次开始，反向累加（保持完整轮次）
-        kept_turns = []
-        accumulated_tokens = 0
-        min_turns = 3  # 尽量保留至少 3 轮，但不强制（避免超出 token 限制）
-        
-        for i, turn in enumerate(reversed(turns)):
-            turn_tokens = self._estimate_turn_tokens(turn)
-            turns_from_end = i + 1
-            
-            # 检查是否超出限制
-            if accumulated_tokens + turn_tokens <= available_tokens:
-                kept_turns.insert(0, turn)
-                accumulated_tokens += turn_tokens
-            else:
-                # 超出限制
-                # 如果还没有保留足够的轮次，且这是最后的机会，尝试保留
-                if len(kept_turns) < min_turns and turns_from_end <= min_turns:
-                    # 检查是否严重超出（超出 20% 以上则放弃）
-                    overflow_ratio = (accumulated_tokens + turn_tokens - available_tokens) / available_tokens
-                    if overflow_ratio < 0.2:  # 允许最多超出 20%
-                        kept_turns.insert(0, turn)
-                        accumulated_tokens += turn_tokens
-                        logger.debug(f"   为保留最少轮次，允许超出 {overflow_ratio*100:.1f}%")
-                        continue
-                # 停止保留更早的轮次
-                break
-        
-        # 重建消息列表
+        if self.agent.memory_manager:
+            discarded_messages = []
+            for turn in turns[:removed_count]:
+                discarded_messages.extend(turn["messages"])
+            if discarded_messages:
+                user_id = getattr(self.agent, '_current_user_id', None)
+                self.agent.memory_manager.flush_memory(
+                    messages=discarded_messages, user_id=user_id,
+                    reason="trim", max_messages=0
+                )
+
        new_messages = []
        for turn in kept_turns:
            new_messages.extend(turn['messages'])
-        
+
        old_count = len(self.messages)
-        old_turn_count = len(turns)
        self.messages = new_messages
-        new_count = len(self.messages)
-        new_turn_count = len(kept_turns)
-        
-        if old_count > new_count:
-            logger.info(
-                f"   移除了 {old_turn_count - new_turn_count} 轮对话 "
-                f"({old_count} -> {new_count} 条消息，"
-                f"~{current_tokens + system_tokens} -> ~{accumulated_tokens + system_tokens} tokens)"
-            )
+
+        logger.info(
+            f"   移除了 {removed_count} 轮对话 "
+            f"({old_count} -> {len(self.messages)} 条消息，"
+            f"~{current_tokens + system_tokens} -> ~{kept_tokens + system_tokens} tokens)"
+        )
+
+    def _clear_session_db(self):
+        """
+        Clear the current session's persisted messages from SQLite DB.
+
+        This prevents dirty data (broken tool_use/tool_result pairs) from being
+        reloaded on the next request or after a restart.
+        """
+        try:
+            session_id = getattr(self.agent, '_current_session_id', None)
+            if not session_id:
+                return
+            from agent.memory import get_conversation_store
+            store = get_conversation_store()
+            store.clear_session(session_id)
+            logger.info(f"🗑️ Cleared dirty session data from DB: {session_id}")
+        except Exception as e:
+            logger.warning(f"Failed to clear session DB: {e}")

    def _prepare_messages(self) -> List[Dict[str, Any]]:
        """
--- a/agent/protocol/message_utils.py
+++ b/agent/protocol/message_utils.py
@@ -0,0 +1,240 @@
+"""
+Message sanitizer — fix broken tool_use / tool_result pairs.
+
+Provides two public helpers that can be reused across agent_stream.py
+and any bot that converts messages to OpenAI format:
+
+1. sanitize_claude_messages(messages)
+   Operates on the internal Claude-format message list (in-place).
+
+2. drop_orphaned_tool_results_openai(messages)
+   Operates on an already-converted OpenAI-format message list,
+   returning a cleaned copy.
+"""
+
+from __future__ import annotations
+
+from typing import Dict, List, Set
+
+from common.log import logger
+
+
+# ------------------------------------------------------------------ #
+# Claude-format sanitizer (used by agent_stream)
+# ------------------------------------------------------------------ #
+
+def sanitize_claude_messages(messages: List[Dict]) -> int:
+    """
+    Validate and fix a Claude-format message list **in-place**.
+
+    Fixes handled:
+    - Trailing assistant message with tool_use but no following tool_result
+    - Leading orphaned tool_result user messages
+    - Mid-list tool_result blocks whose tool_use_id has no matching
+      tool_use in any preceding assistant message
+
+    Returns the number of messages / blocks removed.
+    """
+    if not messages:
+        return 0
+
+    removed = 0
+
+    # 1. Remove trailing incomplete tool_use assistant messages
+    while messages:
+        last = messages[-1]
+        if last.get("role") != "assistant":
+            break
+        content = last.get("content", [])
+        if isinstance(content, list) and any(
+            isinstance(b, dict) and b.get("type") == "tool_use"
+            for b in content
+        ):
+            logger.warning("⚠️ Removing trailing incomplete tool_use assistant message")
+            messages.pop()
+            removed += 1
+        else:
+            break
+
+    # 2. Remove leading orphaned tool_result user messages
+    while messages:
+        first = messages[0]
+        if first.get("role") != "user":
+            break
+        content = first.get("content", [])
+        if isinstance(content, list) and _has_block_type(content, "tool_result") \
+                and not _has_block_type(content, "text"):
+            logger.warning("⚠️ Removing leading orphaned tool_result user message")
+            messages.pop(0)
+            removed += 1
+        else:
+            break
+
+    # 3. Iteratively remove unmatched tool_use / tool_result until stable.
+    #    Removing one broken message can orphan others (e.g. an assistant msg
+    #    with both matched and unmatched tool_use — deleting it orphans the
+    #    previously-matched tool_result).  Loop until clean.
+    for _ in range(5):
+        use_ids: Set[str] = set()
+        result_ids: Set[str] = set()
+        for msg in messages:
+            for block in (msg.get("content") or []):
+                if not isinstance(block, dict):
+                    continue
+                if block.get("type") == "tool_use" and block.get("id"):
+                    use_ids.add(block["id"])
+                elif block.get("type") == "tool_result" and block.get("tool_use_id"):
+                    result_ids.add(block["tool_use_id"])
+
+        bad_use = use_ids - result_ids
+        bad_result = result_ids - use_ids
+        if not bad_use and not bad_result:
+            break
+
+        pass_removed = 0
+        i = 0
+        while i < len(messages):
+            msg = messages[i]
+            role = msg.get("role")
+            content = msg.get("content", [])
+            if not isinstance(content, list):
+                i += 1
+                continue
+
+            if role == "assistant" and bad_use and any(
+                isinstance(b, dict) and b.get("type") == "tool_use"
+                and b.get("id") in bad_use for b in content
+            ):
+                logger.warning(f"⚠️ Removing assistant msg with unmatched tool_use")
+                messages.pop(i)
+                pass_removed += 1
+                continue
+
+            if role == "user" and bad_result and _has_block_type(content, "tool_result"):
+                has_bad = any(
+                    isinstance(b, dict) and b.get("type") == "tool_result"
+                    and b.get("tool_use_id") in bad_result for b in content
+                )
+                if has_bad:
+                    if not _has_block_type(content, "text"):
+                        logger.warning(f"⚠️ Removing user msg with unmatched tool_result")
+                        messages.pop(i)
+                        pass_removed += 1
+                        continue
+                    else:
+                        before = len(content)
+                        msg["content"] = [
+                            b for b in content
+                            if not (isinstance(b, dict) and b.get("type") == "tool_result"
+                                    and b.get("tool_use_id") in bad_result)
+                        ]
+                        pass_removed += before - len(msg["content"])
+
+            i += 1
+
+        removed += pass_removed
+        if pass_removed == 0:
+            break
+
+    if removed:
+        logger.info(f"🔧 Message validation: removed {removed} broken message(s)")
+    return removed
+
+
+# ------------------------------------------------------------------ #
+# OpenAI-format sanitizer (used by minimax_bot, openai_compatible_bot)
+# ------------------------------------------------------------------ #
+
+def drop_orphaned_tool_results_openai(messages: List[Dict]) -> List[Dict]:
+    """
+    Return a copy of *messages* (OpenAI format) with any ``role=tool``
+    messages removed if their ``tool_call_id`` does not match a
+    ``tool_calls[].id`` in a preceding assistant message.
+    """
+    known_ids: Set[str] = set()
+    cleaned: List[Dict] = []
+    for msg in messages:
+        if msg.get("role") == "assistant" and msg.get("tool_calls"):
+            for tc in msg["tool_calls"]:
+                tc_id = tc.get("id", "")
+                if tc_id:
+                    known_ids.add(tc_id)
+
+        if msg.get("role") == "tool":
+            ref_id = msg.get("tool_call_id", "")
+            if ref_id and ref_id not in known_ids:
+                logger.warning(
+                    f"[MessageSanitizer] Dropping orphaned tool result "
+                    f"(tool_call_id={ref_id} not in known ids)"
+                )
+                continue
+        cleaned.append(msg)
+    return cleaned
+
+
+# ------------------------------------------------------------------ #
+# Internal helpers
+# ------------------------------------------------------------------ #
+
+def _has_block_type(content: list, block_type: str) -> bool:
+    return any(
+        isinstance(b, dict) and b.get("type") == block_type
+        for b in content
+    )
+
+
+def _extract_text_from_content(content) -> str:
+    """Extract plain text from a message content field (str or list of blocks)."""
+    if isinstance(content, str):
+        return content.strip()
+    if isinstance(content, list):
+        parts = [
+            b.get("text", "")
+            for b in content
+            if isinstance(b, dict) and b.get("type") == "text"
+        ]
+        return "\n".join(p for p in parts if p).strip()
+    return ""
+
+
+def compress_turn_to_text_only(turn: Dict) -> Dict:
+    """
+    Compress a full turn (with tool_use/tool_result chains) into a lightweight
+    text-only turn that keeps only the first user text and the last assistant text.
+
+    This preserves the conversational context (what the user asked and what the
+    agent concluded) while stripping out the bulky intermediate tool interactions.
+
+    Returns a new turn dict with a ``messages`` list; the original is not mutated.
+    """
+    user_text = ""
+    last_assistant_text = ""
+
+    for msg in turn["messages"]:
+        role = msg.get("role")
+        content = msg.get("content", [])
+
+        if role == "user":
+            if isinstance(content, list) and _has_block_type(content, "tool_result"):
+                continue
+            if not user_text:
+                user_text = _extract_text_from_content(content)
+
+        elif role == "assistant":
+            text = _extract_text_from_content(content)
+            if text:
+                last_assistant_text = text
+
+    compressed_messages = []
+    if user_text:
+        compressed_messages.append({
+            "role": "user",
+            "content": [{"type": "text", "text": user_text}]
+        })
+    if last_assistant_text:
+        compressed_messages.append({
+            "role": "assistant",
+            "content": [{"type": "text", "text": last_assistant_text}]
+        })
+
+    return {"messages": compressed_messages}
--- a/agent/skills/init.py
+++ b/agent/skills/init.py
@@ -15,6 +15,7 @@ from agent.skills.types import (
 )
 from agent.skills.loader import SkillLoader
 from agent.skills.manager import SkillManager
+from agent.skills.service import SkillService
 from agent.skills.formatter import format_skills_for_prompt

 __all__ = [
@@ -25,5 +26,6 @@ __all__ = [
    "LoadSkillsResult",
    "SkillLoader",
    "SkillManager",
+    "SkillService",
    "format_skills_for_prompt",
 ]
--- a/agent/skills/config.py
+++ b/agent/skills/config.py
@@ -123,13 +123,18 @@ def should_include_skill(
                return False
        
        # Check environment variables (API keys)
-        # Simple rule: All required env vars must be set
+        # All required env vars must be set
        required_env = metadata.requires.get('env', [])
        if required_env:
            for env_name in required_env:
                if not has_env_var(env_name):
-                    # Missing required API key → disable skill
                    return False
+
+        # Check anyEnv (at least one must be present)
+        any_env = metadata.requires.get('anyEnv', [])
+        if any_env:
+            if not any(has_env_var(e) for e in any_env):
+                return False
    
    return True

--- a/agent/skills/formatter.py
+++ b/agent/skills/formatter.py
@@ -23,12 +23,10 @@ def format_skills_for_prompt(skills: List[Skill]) -> str:
        return ""
    
    lines = [
-        "\n\nThe following skills provide specialized instructions for specific tasks.",
-        "Use the read tool to load a skill's file when the task matches its description.",
        "",
        "<available_skills>",
    ]
-    
+
    for skill in visible_skills:
        lines.append("  <skill>")
        lines.append(f"    <name>{_escape_xml(skill.name)}</name>")
--- a/agent/skills/loader.py
+++ b/agent/skills/loader.py
@@ -12,25 +12,20 @@ from agent.skills.frontmatter import parse_frontmatter, parse_metadata, parse_bo

 class SkillLoader:
    """Loads skills from various directories."""
-    
-    def __init__(self, workspace_dir: Optional[str] = None):
-        """
-        Initialize the skill loader.
-        
-        :param workspace_dir: Agent workspace directory (for workspace-specific skills)
-        """
-        self.workspace_dir = workspace_dir
+
+    def __init__(self):
+        pass
    
    def load_skills_from_dir(self, dir_path: str, source: str) -> LoadSkillsResult:
        """
        Load skills from a directory.
-        
+
        Discovery rules:
        - Direct .md files in the root directory
        - Recursive SKILL.md files under subdirectories
-        
+
        :param dir_path: Directory path to scan
-        :param source: Source identifier (e.g., 'managed', 'workspace', 'bundled')
+        :param source: Source identifier ('builtin' or 'custom')
        :return: LoadSkillsResult with skills and diagnostics
        """
        skills = []
@@ -96,7 +91,7 @@ class SkillLoader:
                continue
            
            # Check if this is a skill file
-            is_root_md = include_root_files and entry.endswith('.md')
+            is_root_md = include_root_files and entry.endswith('.md') and entry.upper() != 'README.MD'
            is_skill_md = not include_root_files and entry == 'SKILL.md'
            
            if not (is_root_md or is_skill_md):
@@ -188,16 +183,14 @@ class SkillLoader:
        import json
        
        config_path = os.path.join(skill_dir, "config.json")
-        template_path = os.path.join(skill_dir, "config.json.template")
        
-        # Try to load config.json or fallback to template
-        config_file = config_path if os.path.exists(config_path) else template_path
-        
-        if not os.path.exists(config_file):
-            return default_description
+        # Without config.json, skip this skill entirely (return empty to trigger exclusion)
+        if not os.path.exists(config_path):
+            logger.debug(f"[SkillLoader] linkai-agent skipped: no config.json found")
+            return ""
        
        try:
-            with open(config_file, 'r', encoding='utf-8') as f:
+            with open(config_path, 'r', encoding='utf-8') as f:
                config = json.load(f)
            
            apps = config.get("apps", [])
@@ -218,61 +211,49 @@ class SkillLoader:
    
    def load_all_skills(
        self,
-        managed_dir: Optional[str] = None,
-        workspace_skills_dir: Optional[str] = None,
-        extra_dirs: Optional[List[str]] = None,
+        builtin_dir: Optional[str] = None,
+        custom_dir: Optional[str] = None,
    ) -> Dict[str, SkillEntry]:
        """
-        Load skills from all configured locations with precedence.
-        
+        Load skills from builtin and custom directories.
+
        Precedence (lowest to highest):
-        1. Extra directories
-        2. Managed skills directory
-        3. Workspace skills directory
-        
-        :param managed_dir: Managed skills directory (e.g., ~/.cow/skills)
-        :param workspace_skills_dir: Workspace skills directory (e.g., workspace/skills)
-        :param extra_dirs: Additional directories to load skills from
+        1. builtin  — project root ``skills/``, shipped with the codebase
+        2. custom   — workspace ``skills/``, installed via cloud console or skill creator
+
+        Same-name custom skills override builtin ones.
+
+        :param builtin_dir: Built-in skills directory
+        :param custom_dir: Custom skills directory
        :return: Dictionary mapping skill name to SkillEntry
        """
        skill_map: Dict[str, SkillEntry] = {}
        all_diagnostics = []
-        
-        # Load from extra directories (lowest precedence)
-        if extra_dirs:
-            for extra_dir in extra_dirs:
-                if not os.path.exists(extra_dir):
-                    continue
-                result = self.load_skills_from_dir(extra_dir, source='extra')
-                all_diagnostics.extend(result.diagnostics)
-                for skill in result.skills:
-                    entry = self._create_skill_entry(skill)
-                    skill_map[skill.name] = entry
-        
-        # Load from managed directory
-        if managed_dir and os.path.exists(managed_dir):
-            result = self.load_skills_from_dir(managed_dir, source='managed')
+
+        # Load builtin skills (lower precedence)
+        if builtin_dir and os.path.exists(builtin_dir):
+            result = self.load_skills_from_dir(builtin_dir, source='builtin')
            all_diagnostics.extend(result.diagnostics)
            for skill in result.skills:
                entry = self._create_skill_entry(skill)
                skill_map[skill.name] = entry
-        
-        # Load from workspace directory (highest precedence)
-        if workspace_skills_dir and os.path.exists(workspace_skills_dir):
-            result = self.load_skills_from_dir(workspace_skills_dir, source='workspace')
+
+        # Load custom skills (higher precedence, overrides builtin)
+        if custom_dir and os.path.exists(custom_dir):
+            result = self.load_skills_from_dir(custom_dir, source='custom')
            all_diagnostics.extend(result.diagnostics)
            for skill in result.skills:
                entry = self._create_skill_entry(skill)
                skill_map[skill.name] = entry
-        
+
        # Log diagnostics
        if all_diagnostics:
            logger.debug(f"Skill loading diagnostics: {len(all_diagnostics)} issues")
-            for diag in all_diagnostics[:5]:  # Log first 5
+            for diag in all_diagnostics[:5]:
                logger.debug(f"  - {diag}")
-        
-        logger.debug(f"Loaded {len(skill_map)} skills from all sources")
-        
+
+        logger.debug(f"Loaded {len(skill_map)} skills total")
+
        return skill_map
    
    def _create_skill_entry(self, skill: Skill) -> SkillEntry:
--- a/agent/skills/manager.py
+++ b/agent/skills/manager.py
@@ -3,6 +3,7 @@ Skill manager for managing skill lifecycle and operations.
 """

 import os
+import json
 from typing import Dict, List, Optional
 from pathlib import Path
 from common.log import logger
@@ -10,56 +11,134 @@ from agent.skills.types import Skill, SkillEntry, SkillSnapshot
 from agent.skills.loader import SkillLoader
 from agent.skills.formatter import format_skill_entries_for_prompt

+SKILLS_CONFIG_FILE = "skills_config.json"
+

 class SkillManager:
    """Manages skills for an agent."""
-    
+
    def __init__(
        self,
-        workspace_dir: Optional[str] = None,
-        managed_skills_dir: Optional[str] = None,
-        extra_dirs: Optional[List[str]] = None,
+        builtin_dir: Optional[str] = None,
+        custom_dir: Optional[str] = None,
        config: Optional[Dict] = None,
    ):
        """
        Initialize the skill manager.
-        
-        :param workspace_dir: Agent workspace directory
-        :param managed_skills_dir: Managed skills directory (e.g., ~/.cow/skills)
-        :param extra_dirs: Additional skill directories
+
+        :param builtin_dir: Built-in skills directory (project root ``skills/``)
+        :param custom_dir: Custom skills directory (workspace ``skills/``)
        :param config: Configuration dictionary
        """
-        self.workspace_dir = workspace_dir
-        self.managed_skills_dir = managed_skills_dir or self._get_default_managed_dir()
-        self.extra_dirs = extra_dirs or []
+        project_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+        self.builtin_dir = builtin_dir or os.path.join(project_root, 'skills')
+        self.custom_dir = custom_dir or os.path.join(project_root, 'workspace', 'skills')
        self.config = config or {}
-        
-        self.loader = SkillLoader(workspace_dir=workspace_dir)
+        self._skills_config_path = os.path.join(self.custom_dir, SKILLS_CONFIG_FILE)
+
+        # skills_config: full skill metadata keyed by name
+        # { "web-fetch": {"name": ..., "description": ..., "source": ..., "enabled": true}, ... }
+        self.skills_config: Dict[str, dict] = {}
+
+        self.loader = SkillLoader()
        self.skills: Dict[str, SkillEntry] = {}
-        
+
        # Load skills on initialization
        self.refresh_skills()
-    
-    def _get_default_managed_dir(self) -> str:
-        """Get the default managed skills directory."""
-        # Use project root skills directory as default
-        import os
-        project_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-        return os.path.join(project_root, 'skills')
-    
+
    def refresh_skills(self):
-        """Reload all skills from configured directories."""
-        workspace_skills_dir = None
-        if self.workspace_dir:
-            workspace_skills_dir = os.path.join(self.workspace_dir, 'skills')
-        
+        """Reload all skills from builtin and custom directories, then sync config."""
        self.skills = self.loader.load_all_skills(
-            managed_dir=self.managed_skills_dir,
-            workspace_skills_dir=workspace_skills_dir,
-            extra_dirs=self.extra_dirs,
+            builtin_dir=self.builtin_dir,
+            custom_dir=self.custom_dir,
        )
-        
+        self._sync_skills_config()
        logger.debug(f"SkillManager: Loaded {len(self.skills)} skills")
+
+    # ------------------------------------------------------------------
+    # skills_config.json management
+    # ------------------------------------------------------------------
+    def _load_skills_config(self) -> Dict[str, dict]:
+        """Load skills_config.json from custom_dir. Returns empty dict if not found."""
+        if not os.path.exists(self._skills_config_path):
+            return {}
+        try:
+            with open(self._skills_config_path, "r", encoding="utf-8") as f:
+                data = json.load(f)
+            if isinstance(data, dict):
+                return data
+        except Exception as e:
+            logger.warning(f"[SkillManager] Failed to load {SKILLS_CONFIG_FILE}: {e}")
+        return {}
+
+    def _save_skills_config(self):
+        """Persist skills_config to custom_dir/skills_config.json."""
+        os.makedirs(self.custom_dir, exist_ok=True)
+        try:
+            with open(self._skills_config_path, "w", encoding="utf-8") as f:
+                json.dump(self.skills_config, f, indent=4, ensure_ascii=False)
+        except Exception as e:
+            logger.error(f"[SkillManager] Failed to save {SKILLS_CONFIG_FILE}: {e}")
+
+    def _sync_skills_config(self):
+        """
+        Merge directory-scanned skills with the persisted config file.
+
+        - New skills discovered on disk are added with enabled=True.
+        - Skills that no longer exist on disk are removed.
+        - Existing entries preserve their enabled state; name/description/source
+          are refreshed from the latest scan.
+        """
+        saved = self._load_skills_config()
+        merged: Dict[str, dict] = {}
+
+        for name, entry in self.skills.items():
+            skill = entry.skill
+            prev = saved.get(name, {})
+            # category priority: persisted config (set by cloud) > default "skill"
+            category = prev.get("category", "skill")
+            merged[name] = {
+                "name": name,
+                "description": skill.description,
+                "source": skill.source,
+                "enabled": prev.get("enabled", True),
+                "category": category,
+            }
+
+        self.skills_config = merged
+        self._save_skills_config()
+
+    def is_skill_enabled(self, name: str) -> bool:
+        """
+        Check if a skill is enabled according to skills_config.
+
+        :param name: skill name
+        :return: True if enabled (default True if not in config)
+        """
+        entry = self.skills_config.get(name)
+        if entry is None:
+            return True
+        return entry.get("enabled", True)
+
+    def set_skill_enabled(self, name: str, enabled: bool):
+        """
+        Set a skill's enabled state and persist.
+
+        :param name: skill name
+        :param enabled: True to enable, False to disable
+        """
+        if name not in self.skills_config:
+            raise ValueError(f"skill '{name}' not found in config")
+        self.skills_config[name]["enabled"] = enabled
+        self._save_skills_config()
+
+    def get_skills_config(self) -> Dict[str, dict]:
+        """
+        Return the full skills_config dict (for query API).
+
+        :return: copy of skills_config
+        """
+        return dict(self.skills_config)
    
    def get_skill(self, name: str) -> Optional[SkillEntry]:
        """
@@ -85,25 +164,24 @@ class SkillManager:
    ) -> List[SkillEntry]:
        """
        Filter skills based on criteria.
-        
+
        Simple rule: Skills are auto-enabled if requirements are met.
-        - Has required API keys → included
-        - Missing API keys → excluded
-        
+        - Has required API keys -> included
+        - Missing API keys -> excluded
+
        :param skill_filter: List of skill names to include (None = all)
-        :param include_disabled: Whether to include skills with disable_model_invocation=True
+        :param include_disabled: Whether to include disabled skills
        :return: Filtered list of skill entries
        """
        from agent.skills.config import should_include_skill
-        
+
        entries = list(self.skills.values())
-        
+
        # Check requirements (platform, binaries, env vars)
        entries = [e for e in entries if should_include_skill(e, self.config)]
-        
+
        # Apply skill filter
        if skill_filter is not None:
-            # Flatten and normalize skill names (handle both strings and nested lists)
            normalized = []
            for item in skill_filter:
                if isinstance(item, str):
@@ -111,20 +189,18 @@ class SkillManager:
                    if name:
                        normalized.append(name)
                elif isinstance(item, list):
-                    # Handle nested lists
                    for subitem in item:
                        if isinstance(subitem, str):
                            name = subitem.strip()
                            if name:
                                normalized.append(name)
-            
            if normalized:
                entries = [e for e in entries if e.skill.name in normalized]
-        
-        # Filter out disabled skills unless explicitly requested
+
+        # Filter out disabled skills based on skills_config.json
        if not include_disabled:
-            entries = [e for e in entries if not e.skill.disable_model_invocation]
-        
+            entries = [e for e in entries if self.is_skill_enabled(e.skill.name)]
+
        return entries
    
    def build_skills_prompt(
--- a/agent/skills/service.py
+++ b/agent/skills/service.py
@@ -0,0 +1,285 @@
+"""
+Skill service for handling skill CRUD operations.
+
+This service provides a unified interface for managing skills, which can be
+called from the cloud control client (LinkAI), the local web console, or any
+other management entry point.
+"""
+
+import os
+import shutil
+import zipfile
+import tempfile
+from typing import Dict, List, Optional
+from common.log import logger
+from agent.skills.types import Skill, SkillEntry
+from agent.skills.manager import SkillManager
+
+try:
+    import requests
+except ImportError:
+    requests = None
+
+
+class SkillService:
+    """
+    High-level service for skill lifecycle management.
+    Wraps SkillManager and provides network-aware operations such as
+    downloading skill files from remote URLs.
+    """
+
+    def __init__(self, skill_manager: SkillManager):
+        """
+        :param skill_manager: The SkillManager instance to operate on
+        """
+        self.manager = skill_manager
+
+    # ------------------------------------------------------------------
+    # query
+    # ------------------------------------------------------------------
+    def query(self) -> List[dict]:
+        """
+        Query all skills and return a serialisable list.
+        Reads from skills_config.json (refreshes from disk if needed).
+
+        :return: list of skill info dicts
+        """
+        self.manager.refresh_skills()
+        config = self.manager.get_skills_config()
+        result = list(config.values())
+        logger.info(f"[SkillService] query: {len(result)} skills found")
+        return result
+
+    # ------------------------------------------------------------------
+    # add / install
+    # ------------------------------------------------------------------
+    def add(self, payload: dict) -> None:
+        """
+        Add (install) a skill from a remote payload.
+
+        Supported payload types:
+
+        1. ``type: "url"`` – download individual files::
+
+            {
+                "name": "web_search",
+                "type": "url",
+                "enabled": true,
+                "files": [
+                    {"url": "https://...", "path": "README.md"},
+                    {"url": "https://...", "path": "scripts/main.py"}
+                ]
+            }
+
+        2. ``type: "package"`` – download a zip archive and extract::
+
+            {
+                "name": "plugin-custom-tool",
+                "type": "package",
+                "category": "skills",
+                "enabled": true,
+                "files": [{"url": "https://cdn.example.com/skills/custom-tool.zip"}]
+            }
+
+        :param payload: skill add payload from server
+        """
+        name = payload.get("name")
+        if not name:
+            raise ValueError("skill name is required")
+
+        payload_type = payload.get("type", "url")
+
+        if payload_type == "package":
+            self._add_package(name, payload)
+        else:
+            self._add_url(name, payload)
+
+        self.manager.refresh_skills()
+
+        category = payload.get("category")
+        if category and name in self.manager.skills_config:
+            self.manager.skills_config[name]["category"] = category
+            self.manager._save_skills_config()
+
+    def _add_url(self, name: str, payload: dict) -> None:
+        """Install a skill by downloading individual files."""
+        files = payload.get("files", [])
+        if not files:
+            raise ValueError("skill files list is empty")
+
+        skill_dir = os.path.join(self.manager.custom_dir, name)
+
+        tmp_dir = skill_dir + ".tmp"
+        if os.path.exists(tmp_dir):
+            shutil.rmtree(tmp_dir)
+        os.makedirs(tmp_dir, exist_ok=True)
+
+        try:
+            for file_info in files:
+                url = file_info.get("url")
+                rel_path = file_info.get("path")
+                if not url or not rel_path:
+                    logger.warning(f"[SkillService] add: skip invalid file entry {file_info}")
+                    continue
+                dest = os.path.join(tmp_dir, rel_path)
+                self._download_file(url, dest)
+        except Exception:
+            shutil.rmtree(tmp_dir, ignore_errors=True)
+            raise
+
+        if os.path.exists(skill_dir):
+            shutil.rmtree(skill_dir)
+        os.rename(tmp_dir, skill_dir)
+
+        logger.info(f"[SkillService] add: skill '{name}' installed via url ({len(files)} files)")
+
+    def _add_package(self, name: str, payload: dict) -> None:
+        """
+        Install a skill by downloading a zip archive and extracting it.
+
+        If the archive contains a single top-level directory, that directory
+        is used as the skill folder directly; otherwise a new directory named
+        after the skill is created to hold the extracted contents.
+        """
+        files = payload.get("files", [])
+        if not files or not files[0].get("url"):
+            raise ValueError("package url is required")
+
+        url = files[0]["url"]
+        skill_dir = os.path.join(self.manager.custom_dir, name)
+
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            zip_path = os.path.join(tmp_dir, "package.zip")
+            self._download_file(url, zip_path)
+
+            if not zipfile.is_zipfile(zip_path):
+                raise ValueError(f"downloaded file is not a valid zip archive: {url}")
+
+            extract_dir = os.path.join(tmp_dir, "extracted")
+            with zipfile.ZipFile(zip_path, "r") as zf:
+                zf.extractall(extract_dir)
+
+            # Determine the actual content root.
+            # If the zip has a single top-level directory, use its contents
+            # so the skill folder is clean (no extra nesting).
+            top_items = [
+                item for item in os.listdir(extract_dir)
+                if not item.startswith(".")
+            ]
+            if len(top_items) == 1:
+                single = os.path.join(extract_dir, top_items[0])
+                if os.path.isdir(single):
+                    extract_dir = single
+
+            if os.path.exists(skill_dir):
+                shutil.rmtree(skill_dir)
+            shutil.copytree(extract_dir, skill_dir)
+
+        logger.info(f"[SkillService] add: skill '{name}' installed via package ({url})")
+
+    # ------------------------------------------------------------------
+    # open / close (enable / disable)
+    # ------------------------------------------------------------------
+    def open(self, payload: dict) -> None:
+        """
+        Enable a skill by name.
+
+        :param payload: {"name": "skill_name"}
+        """
+        name = payload.get("name")
+        if not name:
+            raise ValueError("skill name is required")
+        self.manager.set_skill_enabled(name, enabled=True)
+        logger.info(f"[SkillService] open: skill '{name}' enabled")
+
+    def close(self, payload: dict) -> None:
+        """
+        Disable a skill by name.
+
+        :param payload: {"name": "skill_name"}
+        """
+        name = payload.get("name")
+        if not name:
+            raise ValueError("skill name is required")
+        self.manager.set_skill_enabled(name, enabled=False)
+        logger.info(f"[SkillService] close: skill '{name}' disabled")
+
+    # ------------------------------------------------------------------
+    # delete
+    # ------------------------------------------------------------------
+    def delete(self, payload: dict) -> None:
+        """
+        Delete a skill by removing its directory entirely.
+
+        :param payload: {"name": "skill_name"}
+        """
+        name = payload.get("name")
+        if not name:
+            raise ValueError("skill name is required")
+
+        skill_dir = os.path.join(self.manager.custom_dir, name)
+        if os.path.exists(skill_dir):
+            shutil.rmtree(skill_dir)
+            logger.info(f"[SkillService] delete: removed directory {skill_dir}")
+        else:
+            logger.warning(f"[SkillService] delete: skill directory not found: {skill_dir}")
+
+        # Refresh will remove the deleted skill from config automatically
+        self.manager.refresh_skills()
+        logger.info(f"[SkillService] delete: skill '{name}' deleted")
+
+    # ------------------------------------------------------------------
+    # dispatch - single entry point for protocol messages
+    # ------------------------------------------------------------------
+    def dispatch(self, action: str, payload: Optional[dict] = None) -> dict:
+        """
+        Dispatch a skill management action and return a protocol-compatible
+        response dict.
+
+        :param action: one of query / add / open / close / delete
+        :param payload: action-specific payload (may be None for query)
+        :return: dict with action, code, message, payload
+        """
+        payload = payload or {}
+        try:
+            if action == "query":
+                result_payload = self.query()
+                return {"action": action, "code": 200, "message": "success", "payload": result_payload}
+            elif action == "add":
+                self.add(payload)
+            elif action == "open":
+                self.open(payload)
+            elif action == "close":
+                self.close(payload)
+            elif action == "delete":
+                self.delete(payload)
+            else:
+                return {"action": action, "code": 400, "message": f"unknown action: {action}", "payload": None}
+            return {"action": action, "code": 200, "message": "success", "payload": None}
+        except Exception as e:
+            logger.error(f"[SkillService] dispatch error: action={action}, error={e}")
+            return {"action": action, "code": 500, "message": str(e), "payload": None}
+
+    # ------------------------------------------------------------------
+    # internal helpers
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _download_file(url: str, dest: str):
+        """
+        Download a file from *url* and save to *dest*.
+
+        :param url: remote file URL
+        :param dest: local destination path
+        """
+        if requests is None:
+            raise RuntimeError("requests library is required for downloading skill files")
+
+        dest_dir = os.path.dirname(dest)
+        if dest_dir:
+            os.makedirs(dest_dir, exist_ok=True)
+
+        resp = requests.get(url, timeout=60)
+        resp.raise_for_status()
+        with open(dest, "wb") as f:
+            f.write(resp.content)
+        logger.debug(f"[SkillService] downloaded {url} -> {dest}")
--- a/agent/skills/types.py
+++ b/agent/skills/types.py
@@ -45,7 +45,7 @@ class Skill:
    description: str
    file_path: str
    base_dir: str
-    source: str  # managed, workspace, bundled, etc.
+    source: str  # builtin or custom
    content: str  # Full markdown content
    disable_model_invocation: bool = False
    frontmatter: Dict[str, Any] = field(default_factory=dict)
--- a/agent/tools/init.py
+++ b/agent/tools/init.py
@@ -45,16 +45,45 @@ def _import_optional_tools():
        )
    except Exception as e:
        logger.error(f"[Tools] Scheduler tool failed to load: {e}")
-    
-    
+
+    # WebSearch Tool (conditionally loaded based on API key availability at init time)
+    try:
+        from agent.tools.web_search.web_search import WebSearch
+        tools['WebSearch'] = WebSearch
+    except ImportError as e:
+        logger.error(f"[Tools] WebSearch not loaded - missing dependency: {e}")
+    except Exception as e:
+        logger.error(f"[Tools] WebSearch failed to load: {e}")
+
+    # WebFetch Tool
+    try:
+        from agent.tools.web_fetch.web_fetch import WebFetch
+        tools['WebFetch'] = WebFetch
+    except ImportError as e:
+        logger.error(f"[Tools] WebFetch not loaded - missing dependency: {e}")
+    except Exception as e:
+        logger.error(f"[Tools] WebFetch failed to load: {e}")
+
+    # Vision Tool (conditionally loaded based on API key availability)
+    try:
+        from agent.tools.vision.vision import Vision
+        tools['Vision'] = Vision
+    except ImportError as e:
+        logger.error(f"[Tools] Vision not loaded - missing dependency: {e}")
+    except Exception as e:
+        logger.error(f"[Tools] Vision failed to load: {e}")
+
    return tools

 # Load optional tools
 _optional_tools = _import_optional_tools()
 EnvConfig = _optional_tools.get('EnvConfig')
 SchedulerTool = _optional_tools.get('SchedulerTool')
+WebSearch = _optional_tools.get('WebSearch')
+WebFetch = _optional_tools.get('WebFetch')
+Vision = _optional_tools.get('Vision')
 GoogleSearch = _optional_tools.get('GoogleSearch')
-FileSave = _optional_tools.get('FileSave') 
+FileSave = _optional_tools.get('FileSave')
 Terminal = _optional_tools.get('Terminal')


@@ -92,6 +121,9 @@ __all__ = [
    'MemoryGetTool',
    'EnvConfig',
    'SchedulerTool',
+    'WebSearch',
+    'WebFetch',
+    'Vision',
    # Optional tools (may be None if dependencies not available)
    # 'BrowserTool'
 ]
--- a/agent/tools/bash/bash.py
+++ b/agent/tools/bash/bash.py
@@ -3,6 +3,7 @@ Bash tool - Execute bash commands
 """

 import os
+import re
 import sys
 import subprocess
 import tempfile
@@ -11,6 +12,7 @@ from typing import Dict, Any
 from agent.tools.base_tool import BaseTool, ToolResult
 from agent.tools.utils.truncate import truncate_tail, format_size, DEFAULT_MAX_LINES, DEFAULT_MAX_BYTES
 from common.log import logger
+from common.utils import expand_path


 class Bash(BaseTool):
@@ -19,10 +21,11 @@ class Bash(BaseTool):
    name: str = "bash"
    description: str = f"""Execute a bash command in the current working directory. Returns stdout and stderr. Output is truncated to last {DEFAULT_MAX_LINES} lines or {DEFAULT_MAX_BYTES // 1024}KB (whichever is hit first). If truncated, full output is saved to a temp file.

-IMPORTANT SAFETY GUIDELINES:
- You can freely create, modify, and delete files within the current workspace
- For operations outside the workspace or potentially destructive commands (rm -rf, system commands, etc.), always explain what you're about to do and ask for user confirmation first
- When in doubt, describe the command's purpose and ask for permission before executing"""
+ENVIRONMENT: All API keys from env_config are auto-injected. Use $VAR_NAME directly.
+
+SAFETY:
+- Freely create/modify/delete files within the workspace
+- For destructive and out-of-workspace commands, explain and confirm first"""

    params: dict = {
        "type": "object",
@@ -80,26 +83,32 @@ IMPORTANT SAFETY GUIDELINES:
            env = os.environ.copy()
            
            # Load environment variables from ~/.cow/.env if it exists
-            env_file = os.path.expanduser("~/.cow/.env")
+            env_file = expand_path("~/.cow/.env")
+            dotenv_vars = {}
            if os.path.exists(env_file):
                try:
                    from dotenv import dotenv_values
-                    env_vars = dotenv_values(env_file)
-                    env.update(env_vars)
-                    logger.debug(f"[Bash] Loaded {len(env_vars)} variables from {env_file}")
+                    dotenv_vars = dotenv_values(env_file)
+                    env.update(dotenv_vars)
+                    logger.debug(f"[Bash] Loaded {len(dotenv_vars)} variables from {env_file}")
                except ImportError:
                    logger.debug("[Bash] python-dotenv not installed, skipping .env loading")
                except Exception as e:
                    logger.debug(f"[Bash] Failed to load .env: {e}")
+
+            # getuid() only exists on Unix-like systems
+            if hasattr(os, 'getuid'):
+                logger.debug(f"[Bash] Process UID: {os.getuid()}")
+            else:
+                logger.debug(f"[Bash] Process User: {os.environ.get('USERNAME', os.environ.get('USER', 'unknown'))}")
            
-            # Debug logging
-            logger.debug(f"[Bash] CWD: {self.cwd}")
-            logger.debug(f"[Bash] Command: {command[:500]}")
-            logger.debug(f"[Bash] OPENAI_API_KEY in env: {'OPENAI_API_KEY' in env}")
-            logger.debug(f"[Bash] SHELL: {env.get('SHELL', 'not set')}")
-            logger.debug(f"[Bash] Python executable: {sys.executable}")
-            logger.debug(f"[Bash] Process UID: {os.getuid()}")
-            
+            # On Windows, convert $VAR references to %VAR% for cmd.exe
+            if sys.platform == "win32":
+                env["PYTHONIOENCODING"] = "utf-8"
+                command = self._convert_env_vars_for_windows(command, dotenv_vars)
+                if command and not command.strip().lower().startswith("chcp"):
+                    command = f"chcp 65001 >nul 2>&1 && {command}"
+
            # Execute command with inherited environment variables
            result = subprocess.run(
                command,
@@ -108,6 +117,8 @@ IMPORTANT SAFETY GUIDELINES:
                stdout=subprocess.PIPE,
                stderr=subprocess.PIPE,
                text=True,
+                encoding="utf-8",
+                errors="replace",
                timeout=timeout,
                env=env
            )
@@ -131,6 +142,8 @@ IMPORTANT SAFETY GUIDELINES:
                            stdout=subprocess.PIPE,
                            stderr=subprocess.PIPE,
                            text=True,
+                            encoding="utf-8",
+                            errors="replace",
                            timeout=timeout,
                            env=env
                        )
@@ -258,3 +271,21 @@ IMPORTANT SAFETY GUIDELINES:
                    return "This command will recursively delete system directories"

        return ""  # No warning needed
+
+    @staticmethod
+    def _convert_env_vars_for_windows(command: str, dotenv_vars: dict) -> str:
+        """
+        Convert bash-style $VAR / ${VAR} references to cmd.exe %VAR% syntax.
+        Only converts variables loaded from .env (user-configured API keys etc.)
+        to avoid breaking $PATH, jq expressions, regex, etc.
+        """
+        if not dotenv_vars:
+            return command
+
+        def replace_match(m):
+            var_name = m.group(1) or m.group(2)
+            if var_name in dotenv_vars:
+                return f"%{var_name}%"
+            return m.group(0)
+
+        return re.sub(r'\$\{(\w+)\}|\$(\w+)', replace_match, command)
--- a/agent/tools/edit/edit.py
+++ b/agent/tools/edit/edit.py
@@ -7,6 +7,7 @@ import os
 from typing import Dict, Any

 from agent.tools.base_tool import BaseTool, ToolResult
+from common.utils import expand_path
 from agent.tools.utils.diff import (
    strip_bom,
    detect_line_ending,
@@ -178,7 +179,7 @@ class Edit(BaseTool):
        :return: Absolute path
        """
        # Expand ~ to user home directory
-        path = os.path.expanduser(path)
+        path = expand_path(path)
        if os.path.isabs(path):
            return path
        return os.path.abspath(os.path.join(self.cwd, path))
--- a/agent/tools/env_config/env_config.py
+++ b/agent/tools/env_config/env_config.py
@@ -9,6 +9,7 @@ from pathlib import Path

 from agent.tools.base_tool import BaseTool, ToolResult
 from common.log import logger
+from common.utils import expand_path


 # API Key 知识库：常见的环境变量及其描述
@@ -66,7 +67,7 @@ class EnvConfig(BaseTool):
    def __init__(self, config: dict = None):
        self.config = config or {}
        # Store env config in ~/.cow directory (outside workspace for security)
-        self.env_dir = os.path.expanduser("~/.cow")
+        self.env_dir = expand_path("~/.cow")
        self.env_path = os.path.join(self.env_dir, '.env')
        self.agent_bridge = self.config.get("agent_bridge")  # Reference to AgentBridge for hot reload
        # Don't create .env file in __init__ to avoid issues during tool discovery
@@ -201,7 +202,8 @@ class EnvConfig(BaseTool):
                        "key": key,
                        "value": self._mask_value(value),
                        "description": description,
-                        "exists": True
+                        "exists": True,
+                        "note": f"Value is masked for security. In bash, use ${key} directly — it is auto-injected."
                    })
                else:
                    return ToolResult.success({
--- a/agent/tools/ls/ls.py
+++ b/agent/tools/ls/ls.py
@@ -7,6 +7,7 @@ from typing import Dict, Any

 from agent.tools.base_tool import BaseTool, ToolResult
 from agent.tools.utils.truncate import truncate_head, format_size, DEFAULT_MAX_BYTES
+from common.utils import expand_path


 DEFAULT_LIMIT = 500
@@ -51,7 +52,7 @@ class Ls(BaseTool):
        absolute_path = self._resolve_path(path)
        
        # Security check: Prevent accessing sensitive config directory
-        env_config_dir = os.path.expanduser("~/.cow")
+        env_config_dir = expand_path("~/.cow")
        if os.path.abspath(absolute_path) == os.path.abspath(env_config_dir):
            return ToolResult.fail(
                "Error: Access denied. API keys and credentials must be accessed through the env_config tool only."
@@ -93,7 +94,7 @@ class Ls(BaseTool):
                        results.append(entry + '/')
                    else:
                        results.append(entry)
-                except:
+                except Exception:
                    # Skip entries we can't stat
                    continue
            
@@ -133,7 +134,7 @@ class Ls(BaseTool):
    def _resolve_path(self, path: str) -> str:
        """Resolve path to absolute path"""
        # Expand ~ to user home directory
-        path = os.path.expanduser(path)
+        path = expand_path(path)
        if os.path.isabs(path):
            return path
        return os.path.abspath(os.path.join(self.cwd, path))
--- a/agent/tools/memory/memory_get.py
+++ b/agent/tools/memory/memory_get.py
@@ -77,7 +77,7 @@ class MemoryGetTool(BaseTool):
            if not file_path.exists():
                return ToolResult.fail(f"Error: File not found: {path}")
            
-            content = file_path.read_text()
+            content = file_path.read_text(encoding='utf-8')
            lines = content.split('\n')
            
            # Handle line range
--- a/agent/tools/read/read.py
+++ b/agent/tools/read/read.py
@@ -9,6 +9,7 @@ from pathlib import Path

 from agent.tools.base_tool import BaseTool, ToolResult
 from agent.tools.utils.truncate import truncate_head, format_size, DEFAULT_MAX_LINES, DEFAULT_MAX_BYTES
+from common.utils import expand_path


 class Read(BaseTool):
@@ -47,7 +48,8 @@ class Read(BaseTool):
        self.binary_extensions = {'.exe', '.dll', '.so', '.dylib', '.bin', '.dat', '.db', '.sqlite'}
        self.archive_extensions = {'.zip', '.tar', '.gz', '.rar', '.7z', '.bz2', '.xz'}
        self.pdf_extensions = {'.pdf'}
-        
+        self.office_extensions = {'.doc', '.docx', '.xls', '.xlsx', '.ppt', '.pptx'}
+
        # Readable text formats (will be read with truncation)
        self.text_extensions = {
            '.txt', '.md', '.markdown', '.rst', '.log', '.csv', '.tsv', '.json', '.xml', '.yaml', '.yml',
@@ -56,7 +58,6 @@ class Read(BaseTool):
            '.sh', '.bash', '.zsh', '.fish', '.ps1', '.bat', '.cmd',
            '.sql', '.r', '.m', '.swift', '.kt', '.scala', '.clj', '.erl', '.ex',
            '.dockerfile', '.makefile', '.cmake', '.gradle', '.properties', '.ini', '.conf', '.cfg',
-            '.doc', '.docx', '.xls', '.xlsx', '.ppt', '.pptx'  # Office documents
        }
    
    def execute(self, args: Dict[str, Any]) -> ToolResult:
@@ -66,10 +67,12 @@ class Read(BaseTool):
        :param args: Contains file path and optional offset/limit parameters
        :return: File content or error message
        """
-        path = args.get("path", "").strip()
+        # Support 'location' as alias for 'path' (LLM may use it from skill listing)
+        path = args.get("path", "") or args.get("location", "")
+        path = path.strip() if isinstance(path, str) else ""
        offset = args.get("offset")
        limit = args.get("limit")
-        
+
        if not path:
            return ToolResult.fail("Error: path parameter is required")
        
@@ -77,7 +80,7 @@ class Read(BaseTool):
        absolute_path = self._resolve_path(path)
        
        # Security check: Prevent reading sensitive config files
-        env_config_path = os.path.expanduser("~/.cow/.env")
+        env_config_path = expand_path("~/.cow/.env")
        if os.path.abspath(absolute_path) == os.path.abspath(env_config_path):
            return ToolResult.fail(
                "Error: Access denied. API keys and credentials must be accessed through the env_config tool only."
@@ -117,7 +120,11 @@ class Read(BaseTool):
        # Check if PDF
        if file_ext in self.pdf_extensions:
            return self._read_pdf(absolute_path, path, offset, limit)
-        
+
+        # Check if Office document (.docx, .xlsx, .pptx, etc.)
+        if file_ext in self.office_extensions:
+            return self._read_office(absolute_path, path, file_ext, offset, limit)
+
        # Read text file (with truncation for large files)
        return self._read_text(absolute_path, path, offset, limit)
    
@@ -129,7 +136,7 @@ class Read(BaseTool):
        :return: Absolute path
        """
        # Expand ~ to user home directory
-        path = os.path.expanduser(path)
+        path = expand_path(path)
        if os.path.isabs(path):
            return path
        return os.path.abspath(os.path.join(self.cwd, path))
@@ -237,8 +244,8 @@ class Read(BaseTool):
                    "message": f"文件过大 ({format_size(file_size)} > 50MB)，无法读取内容。文件路径: {absolute_path}"
                })
            
-            # Read file
-            with open(absolute_path, 'r', encoding='utf-8') as f:
+            # Read file (utf-8-sig strips BOM automatically on Windows)
+            with open(absolute_path, 'r', encoding='utf-8-sig') as f:
                content = f.read()
            
            # Truncate content if too long (20K characters max for model context)
@@ -334,6 +341,116 @@ class Read(BaseTool):
        except Exception as e:
            return ToolResult.fail(f"Error reading file: {str(e)}")
    
+    def _read_office(self, absolute_path: str, display_path: str, file_ext: str,
+                     offset: int = None, limit: int = None) -> ToolResult:
+        """Read Office documents (.docx, .xlsx, .pptx) using python-docx / openpyxl / python-pptx."""
+        try:
+            text = self._extract_office_text(absolute_path, file_ext)
+        except ImportError as e:
+            return ToolResult.fail(str(e))
+        except Exception as e:
+            return ToolResult.fail(f"Error reading Office document: {e}")
+
+        if not text or not text.strip():
+            return ToolResult.success({
+                "content": f"[Office file {Path(absolute_path).name}: no text content could be extracted]",
+            })
+
+        all_lines = text.split('\n')
+        total_lines = len(all_lines)
+
+        start_line = 0
+        if offset is not None:
+            if offset < 0:
+                start_line = max(0, total_lines + offset)
+            else:
+                start_line = max(0, offset - 1)
+                if start_line >= total_lines:
+                    return ToolResult.fail(
+                        f"Error: Offset {offset} is beyond end of content ({total_lines} lines total)"
+                    )
+
+        selected_content = text
+        user_limited_lines = None
+        if limit is not None:
+            end_line = min(start_line + limit, total_lines)
+            selected_content = '\n'.join(all_lines[start_line:end_line])
+            user_limited_lines = end_line - start_line
+        elif offset is not None:
+            selected_content = '\n'.join(all_lines[start_line:])
+
+        truncation = truncate_head(selected_content)
+        start_line_display = start_line + 1
+        output_text = ""
+
+        if truncation.truncated:
+            end_line_display = start_line_display + truncation.output_lines - 1
+            next_offset = end_line_display + 1
+            output_text = truncation.content
+            output_text += f"\n\n[Showing lines {start_line_display}-{end_line_display} of {total_lines}. Use offset={next_offset} to continue.]"
+        elif user_limited_lines is not None and start_line + user_limited_lines < total_lines:
+            remaining = total_lines - (start_line + user_limited_lines)
+            next_offset = start_line + user_limited_lines + 1
+            output_text = truncation.content
+            output_text += f"\n\n[{remaining} more lines in file. Use offset={next_offset} to continue.]"
+        else:
+            output_text = truncation.content
+
+        return ToolResult.success({
+            "content": output_text,
+            "total_lines": total_lines,
+            "start_line": start_line_display,
+            "output_lines": truncation.output_lines,
+        })
+
+    @staticmethod
+    def _extract_office_text(absolute_path: str, file_ext: str) -> str:
+        """Extract plain text from an Office document."""
+        if file_ext in ('.docx', '.doc'):
+            try:
+                from docx import Document
+            except ImportError:
+                raise ImportError("Error: python-docx library not installed. Install with: pip install python-docx")
+            doc = Document(absolute_path)
+            paragraphs = [p.text for p in doc.paragraphs]
+            for table in doc.tables:
+                for row in table.rows:
+                    paragraphs.append('\t'.join(cell.text for cell in row.cells))
+            return '\n'.join(paragraphs)
+
+        if file_ext in ('.xlsx', '.xls'):
+            try:
+                from openpyxl import load_workbook
+            except ImportError:
+                raise ImportError("Error: openpyxl library not installed. Install with: pip install openpyxl")
+            wb = load_workbook(absolute_path, read_only=True, data_only=True)
+            parts = []
+            for ws in wb.worksheets:
+                parts.append(f"--- Sheet: {ws.title} ---")
+                for row in ws.iter_rows(values_only=True):
+                    parts.append('\t'.join(str(c) if c is not None else '' for c in row))
+            wb.close()
+            return '\n'.join(parts)
+
+        if file_ext in ('.pptx', '.ppt'):
+            try:
+                from pptx import Presentation
+            except ImportError:
+                raise ImportError("Error: python-pptx library not installed. Install with: pip install python-pptx")
+            prs = Presentation(absolute_path)
+            parts = []
+            for i, slide in enumerate(prs.slides, 1):
+                parts.append(f"--- Slide {i} ---")
+                for shape in slide.shapes:
+                    if shape.has_text_frame:
+                        for para in shape.text_frame.paragraphs:
+                            text = para.text.strip()
+                            if text:
+                                parts.append(text)
+            return '\n'.join(parts)
+
+        return ""
+
    def _read_pdf(self, absolute_path: str, display_path: str, offset: int = None, limit: int = None) -> ToolResult:
        """
        Read PDF file content
--- a/agent/tools/scheduler/integration.py
+++ b/agent/tools/scheduler/integration.py
@@ -6,6 +6,7 @@ import os
 from typing import Optional
 from config import conf
 from common.log import logger
+from common.utils import expand_path
 from bridge.context import Context, ContextType
 from bridge.reply import Reply, ReplyType

@@ -31,7 +32,7 @@ def init_scheduler(agent_bridge) -> bool:
        from agent.tools.scheduler.scheduler_service import SchedulerService
        
        # Get workspace from config
-        workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+        workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
        store_path = os.path.join(workspace_root, "scheduler", "tasks.json")
        
        # Create task store
@@ -112,11 +113,15 @@ def _execute_agent_task(task: dict, agent_bridge):
        
        logger.info(f"[Scheduler] Task {task['id']}: Executing agent task '{task_description}'")
        
+        # Create a unique session_id for this scheduled task to avoid polluting user's conversation
+        # Format: scheduler_<receiver>_<task_id> to ensure isolation
+        scheduler_session_id = f"scheduler_{receiver}_{task['id']}"
+        
        # Create context for Agent
        context = Context(ContextType.TEXT, task_description)
        context["receiver"] = receiver
        context["isgroup"] = is_group
-        context["session_id"] = receiver
+        context["session_id"] = scheduler_session_id
        
        # Channel-specific setup
        if channel_type == "web":
@@ -129,18 +134,20 @@ def _execute_agent_task(task: dict, agent_bridge):
        elif channel_type == "dingtalk":
            # DingTalk requires msg object, set to None for scheduled tasks
            context["msg"] = None
-            # 如果是单聊，需要传递 sender_staff_id
            if not is_group:
                sender_staff_id = action.get("dingtalk_sender_staff_id")
                if sender_staff_id:
                    context["dingtalk_sender_staff_id"] = sender_staff_id
-        
+        elif channel_type == "wecom_bot":
+            context["msg"] = None
+
        # Use Agent to execute the task
        # Mark this as a scheduled task execution to prevent recursive task creation
        context["is_scheduled_task"] = True
        
        try:
-            reply = agent_bridge.agent_reply(task_description, context=context, on_event=None, clear_history=True)
+            # Don't clear history - scheduler tasks use isolated session_id so they won't pollute user conversations
+            reply = agent_bridge.agent_reply(task_description, context=context, on_event=None, clear_history=False)
            
            if reply and reply.content:
                # Send the reply via channel
@@ -228,7 +235,11 @@ def _execute_send_message(task: dict, agent_bridge):
                    logger.debug(f"[Scheduler] DingTalk single chat: sender_staff_id={sender_staff_id}")
                else:
                    logger.warning(f"[Scheduler] Task {task['id']}: DingTalk single chat message missing sender_staff_id")
-        
+        elif channel_type == "wecom_bot":
+            context["msg"] = None
+        elif channel_type == "qq":
+            context["msg"] = None
+
        # Create reply
        reply = Reply(ReplyType.TEXT, content)
        
@@ -321,31 +332,31 @@ def _execute_tool_call(task: dict, agent_bridge):
            context["request_id"] = request_id
            logger.debug(f"[Scheduler] Generated request_id for web channel: {request_id}")
        elif channel_type == "feishu":
-            # Feishu channel: for scheduled tasks, send as new message (no msg_id to reply to)
            context["receive_id_type"] = "chat_id" if is_group else "open_id"
            context["msg"] = None
            logger.debug(f"[Scheduler] Feishu: receive_id_type={context['receive_id_type']}, is_group={is_group}, receiver={receiver}")
-        
+        elif channel_type == "wecom_bot":
+            context["msg"] = None
+
        reply = Reply(ReplyType.TEXT, content)
-        
+
        # Get channel and send
        from channel.channel_factory import create_channel
-        
+
        try:
            channel = create_channel(channel_type)
            if channel:
-                # For web channel, register the request_id to session mapping
                if channel_type == "web" and hasattr(channel, 'request_to_session'):
                    channel.request_to_session[request_id] = receiver
                    logger.debug(f"[Scheduler] Registered request_id {request_id} -> session {receiver}")
-                
+
                channel.send(reply, context)
                logger.info(f"[Scheduler] Task {task['id']} executed: sent tool result to {receiver}")
            else:
                logger.error(f"[Scheduler] Failed to create channel: {channel_type}")
        except Exception as e:
            logger.error(f"[Scheduler] Failed to send tool result: {e}")
-            
+
    except Exception as e:
        logger.error(f"[Scheduler] Error in _execute_tool_call: {e}")

@@ -378,6 +389,10 @@ def _execute_skill_call(task: dict, agent_bridge):
        
        logger.info(f"[Scheduler] Task {task['id']}: Executing skill '{skill_name}' with params {skill_params}")
        
+        # Create a unique session_id for this scheduled task to avoid polluting user's conversation
+        # Format: scheduler_<receiver>_<task_id> to ensure isolation
+        scheduler_session_id = f"scheduler_{receiver}_{task['id']}"
+        
        # Build a natural language query for the Agent to execute the skill
        # Format: "Use skill-name to do something with params"
        param_str = ", ".join([f"{k}={v}" for k, v in skill_params.items()])
@@ -389,7 +404,7 @@ def _execute_skill_call(task: dict, agent_bridge):
        context = Context(ContextType.TEXT, query)
        context["receiver"] = receiver
        context["isgroup"] = is_group
-        context["session_id"] = receiver
+        context["session_id"] = scheduler_session_id
        
        # Channel-specific setup
        if channel_type == "web":
@@ -399,10 +414,13 @@ def _execute_skill_call(task: dict, agent_bridge):
        elif channel_type == "feishu":
            context["receive_id_type"] = "chat_id" if is_group else "open_id"
            context["msg"] = None
-        
+        elif channel_type == "wecom_bot":
+            context["msg"] = None
+
        # Use Agent to execute the skill
        try:
-            reply = agent_bridge.agent_reply(query, context=context, on_event=None, clear_history=True)
+            # Don't clear history - scheduler tasks use isolated session_id so they won't pollute user conversations
+            reply = agent_bridge.agent_reply(query, context=context, on_event=None, clear_history=False)
            
            if reply and reply.content:
                content = reply.content
@@ -440,8 +458,7 @@ def attach_scheduler_to_tool(tool, context: Context = None):
    if context:
        tool.current_context = context
        
-        # Also set channel_type from config
-        channel_type = conf().get("channel_type", "unknown")
+        channel_type = context.get("channel_type") or conf().get("channel_type", "unknown")
        if not tool.config:
            tool.config = {}
        tool.config["channel_type"] = channel_type
--- a/agent/tools/scheduler/scheduler_service.py
+++ b/agent/tools/scheduler/scheduler_service.py
@@ -61,8 +61,7 @@ class SchedulerService:
                self._check_and_execute_tasks()
            except Exception as e:
                logger.error(f"[Scheduler] Error in scheduler loop: {e}")
-            
-            # Sleep for 30 seconds between checks
+
            time.sleep(30)
    
    def _check_and_execute_tasks(self):
@@ -85,12 +84,9 @@ class SchedulerService:
                            "last_run_at": now.isoformat()
                        })
                    else:
-                        # One-time task, disable it
-                        self.task_store.update_task(task['id'], {
-                            "enabled": False,
-                            "last_run_at": now.isoformat()
-                        })
-                        logger.info(f"[Scheduler] One-time task completed and disabled: {task['id']}")
+                        # One-time task completed, remove it
+                        self.task_store.delete_task(task['id'])
+                        logger.info(f"[Scheduler] One-time task completed and removed: {task['id']}")
            except Exception as e:
                logger.error(f"[Scheduler] Error processing task {task.get('id')}: {e}")
    
@@ -127,14 +123,11 @@ class SchedulerService:
                if time_diff > 300:  # 5 minutes
                    logger.warning(f"[Scheduler] Task {task['id']} is overdue by {int(time_diff)}s, skipping and scheduling next run")
                    
-                    # For one-time tasks, disable them
+                    # For one-time tasks, remove them directly
                    schedule = task.get("schedule", {})
                    if schedule.get("type") == "once":
-                        self.task_store.update_task(task['id'], {
-                            "enabled": False,
-                            "last_run_at": now.isoformat()
-                        })
-                        logger.info(f"[Scheduler] One-time task {task['id']} expired, disabled")
+                        self.task_store.delete_task(task['id'])
+                        logger.info(f"[Scheduler] One-time task {task['id']} expired, removed")
                        return False
                    
                    # For recurring tasks, calculate next run from now
@@ -147,7 +140,7 @@ class SchedulerService:
                    return False
            
            return now >= next_run
-        except:
+        except Exception:
            return False
    
    def _calculate_next_run(self, task: dict, from_time: datetime) -> Optional[datetime]:
@@ -195,7 +188,7 @@ class SchedulerService:
                # Only return if in the future
                if run_at > from_time:
                    return run_at
-            except:
+            except Exception:
                pass
            return None
        
--- a/agent/tools/scheduler/scheduler_tool.py
+++ b/agent/tools/scheduler/scheduler_tool.py
@@ -20,7 +20,8 @@ class SchedulerTool(BaseTool):
    
    name: str = "scheduler"
    description: str = (
-        "创建、查询和管理定时任务。支持固定消息和AI任务两种类型。\n\n"
+        "创建、查询和管理定时任务（提醒、周期性任务等）。\n\n"
+        "⚠️ 重要：仅当需要「定时/提醒/每天/每周/X分钟后/X点」等延迟或周期执行时才使用此工具。"
        "使用方法：\n"
        "- 创建：action='create', name='任务名', message/ai_task='内容', schedule_type='once/interval/cron', schedule_value='...'\n"
        "- 查询：action='list' / action='get', task_id='任务ID'\n"
@@ -53,7 +54,7 @@ class SchedulerTool(BaseTool):
            },
            "ai_task": {
                "type": "string",
-                "description": "AI任务描述 (与message二选一)，如'搜索今日新闻'、'查询天气'"
+                "description": "AI任务描述 (与message二选一)，用于定时让AI执行的任务"
            },
            "schedule_type": {
                "type": "string",
@@ -423,7 +424,7 @@ class SchedulerTool(BaseTool):
            try:
                dt = datetime.fromisoformat(run_at)
                return f"一次性 ({dt.strftime('%Y-%m-%d %H:%M')})"
-            except:
+            except Exception:
                return "一次性"
        
        return "未知"
@@ -437,6 +438,6 @@ class SchedulerTool(BaseTool):
                    return msg.other_user_nickname or "群聊"
                else:
                    return msg.from_user_nickname or "用户"
-        except:
+        except Exception:
            pass
        return "未知"
--- a/agent/tools/scheduler/task_store.py
+++ b/agent/tools/scheduler/task_store.py
@@ -8,6 +8,7 @@ import threading
 from datetime import datetime
 from typing import Dict, List, Optional
 from pathlib import Path
+from common.utils import expand_path


 class TaskStore:
@@ -24,7 +25,7 @@ class TaskStore:
        """
        if store_path is None:
            # Default to ~/cow/scheduler/tasks.json
-            home = os.path.expanduser("~")
+            home = expand_path("~")
            store_path = os.path.join(home, "cow", "scheduler", "tasks.json")
        
        self.store_path = store_path
@@ -71,7 +72,7 @@ class TaskStore:
                        with open(self.store_path, 'r') as src:
                            with open(backup_path, 'w') as dst:
                                dst.write(src.read())
-                    except:
+                    except Exception:
                        pass
                
                # Save tasks
--- a/agent/tools/send/send.py
+++ b/agent/tools/send/send.py
@@ -7,20 +7,21 @@ from typing import Dict, Any
 from pathlib import Path

 from agent.tools.base_tool import BaseTool, ToolResult
+from common.utils import expand_path


 class Send(BaseTool):
    """Tool for sending files to the user"""
    
    name: str = "send"
-    description: str = "Send a file (image, video, audio, document) to the user. Use this when the user explicitly asks to send/share a file."
+    description: str = "Send a LOCAL file (image, video, audio, document) to the user. Only for local file paths. Do NOT use this for URLs — URLs should be included directly in your text reply, the system will handle them automatically."
    
    params: dict = {
        "type": "object",
        "properties": {
            "path": {
                "type": "string",
-                "description": "Path to the file to send. Can be absolute path or relative to workspace."
+                "description": "Local file path to send. Must be an absolute path or relative to workspace. Do NOT pass URLs here."
            },
            "message": {
                "type": "string",
@@ -102,7 +103,7 @@ class Send(BaseTool):
    
    def _resolve_path(self, path: str) -> str:
        """Resolve path to absolute path"""
-        path = os.path.expanduser(path)
+        path = expand_path(path)
        if os.path.isabs(path):
            return path
        return os.path.abspath(os.path.join(self.cwd, path))
--- a/agent/tools/vision/init.py
+++ b/agent/tools/vision/init.py
@@ -0,0 +1 @@
+from agent.tools.vision.vision import Vision
--- a/agent/tools/vision/vision.py
+++ b/agent/tools/vision/vision.py
@@ -0,0 +1,255 @@
+"""
+Vision tool - Analyze images using OpenAI-compatible Vision API.
+Supports local files (auto base64-encoded) and HTTP URLs.
+Providers: OpenAI (preferred) > LinkAI (fallback).
+"""
+
+import base64
+import os
+import subprocess
+import tempfile
+from typing import Any, Dict, Optional, Tuple
+
+import requests
+
+from agent.tools.base_tool import BaseTool, ToolResult
+from common.log import logger
+from config import conf
+
+DEFAULT_MODEL = "gpt-4.1-mini"
+DEFAULT_TIMEOUT = 60
+MAX_TOKENS = 1000
+COMPRESS_THRESHOLD = 1_048_576  # 1 MB
+
+SUPPORTED_EXTENSIONS = {
+    "jpg": "image/jpeg",
+    "jpeg": "image/jpeg",
+    "png": "image/png",
+    "gif": "image/gif",
+    "webp": "image/webp",
+}
+
+
+class Vision(BaseTool):
+    """Analyze images using OpenAI-compatible Vision API"""
+
+    name: str = "vision"
+    description: str = (
+        "Analyze a local image or image URL (jpg/jpeg/png) using Vision API. "
+        "Can describe content, extract text, identify objects, colors, etc. "
+        "Requires OPENAI_API_KEY or LINKAI_API_KEY."
+    )
+
+    params: dict = {
+        "type": "object",
+        "properties": {
+            "image": {
+                "type": "string",
+                "description": "Local file path or HTTP(S) URL of the image to analyze",
+            },
+            "question": {
+                "type": "string",
+                "description": "Question to ask about the image",
+            },
+            "model": {
+                "type": "string",
+                "description": (
+                    f"Vision model to use (default: {DEFAULT_MODEL}). "
+                    "Options: gpt-4.1-mini, gpt-4.1, gpt-4o-mini, gpt-4o"
+                ),
+            },
+        },
+        "required": ["image", "question"],
+    }
+
+    def __init__(self, config: dict = None):
+        self.config = config or {}
+
+    @staticmethod
+    def is_available() -> bool:
+        return bool(
+            conf().get("open_ai_api_key") or os.environ.get("OPENAI_API_KEY")
+            or conf().get("linkai_api_key") or os.environ.get("LINKAI_API_KEY")
+        )
+
+    def execute(self, args: Dict[str, Any]) -> ToolResult:
+        image = args.get("image", "").strip()
+        question = args.get("question", "").strip()
+        model = args.get("model", DEFAULT_MODEL).strip() or DEFAULT_MODEL
+
+        if not image:
+            return ToolResult.fail("Error: 'image' parameter is required")
+        if not question:
+            return ToolResult.fail("Error: 'question' parameter is required")
+
+        api_key, api_base = self._resolve_provider()
+        if not api_key:
+            return ToolResult.fail(
+                "Error: No API key configured for Vision.\n"
+                "Please configure one of the following using env_config tool:\n"
+                "  1. OPENAI_API_KEY (preferred): env_config(action=\"set\", key=\"OPENAI_API_KEY\", value=\"your-key\")\n"
+                "  2. LINKAI_API_KEY (fallback): env_config(action=\"set\", key=\"LINKAI_API_KEY\", value=\"your-key\")\n\n"
+                "Get your key at: https://platform.openai.com/api-keys or https://link-ai.tech"
+            )
+
+        try:
+            image_content = self._build_image_content(image)
+        except Exception as e:
+            return ToolResult.fail(f"Error: {e}")
+
+        try:
+            return self._call_api(api_key, api_base, model, question, image_content)
+        except requests.Timeout:
+            return ToolResult.fail(f"Error: Vision API request timed out after {DEFAULT_TIMEOUT}s")
+        except requests.ConnectionError:
+            return ToolResult.fail("Error: Failed to connect to Vision API")
+        except Exception as e:
+            logger.error(f"[Vision] Unexpected error: {e}", exc_info=True)
+            return ToolResult.fail(f"Error: Vision API call failed - {e}")
+
+    def _resolve_provider(self) -> Tuple[Optional[str], str]:
+        """Resolve API key and base URL. Priority: conf() > env vars."""
+        api_key = conf().get("open_ai_api_key") or os.environ.get("OPENAI_API_KEY")
+        if api_key:
+            api_base = (conf().get("open_ai_api_base") or os.environ.get("OPENAI_API_BASE", "")).rstrip("/") \
+                or "https://api.openai.com/v1"
+            return api_key, self._ensure_v1(api_base)
+
+        api_key = conf().get("linkai_api_key") or os.environ.get("LINKAI_API_KEY")
+        if api_key:
+            api_base = (conf().get("linkai_api_base") or os.environ.get("LINKAI_API_BASE", "")).rstrip("/") \
+                or "https://api.link-ai.tech"
+            logger.debug("[Vision] Using LinkAI API (OPENAI_API_KEY not set)")
+            return api_key, self._ensure_v1(api_base)
+
+        return None, ""
+
+    @staticmethod
+    def _ensure_v1(api_base: str) -> str:
+        """Append /v1 if the base URL doesn't already end with a versioned path."""
+        if not api_base:
+            return api_base
+        # Already has /v1 or similar version suffix
+        if api_base.rstrip("/").split("/")[-1].startswith("v"):
+            return api_base
+        return api_base.rstrip("/") + "/v1"
+
+    def _build_image_content(self, image: str) -> dict:
+        """Build the image_url content block for the API request."""
+        if image.startswith(("http://", "https://")):
+            return {"type": "image_url", "image_url": {"url": image}}
+
+        if not os.path.isfile(image):
+            raise FileNotFoundError(f"Image file not found: {image}")
+
+        ext = image.rsplit(".", 1)[-1].lower() if "." in image else ""
+        mime_type = SUPPORTED_EXTENSIONS.get(ext)
+        if not mime_type:
+            raise ValueError(
+                f"Unsupported image format '.{ext}'. "
+                f"Supported: {', '.join(SUPPORTED_EXTENSIONS.keys())}"
+            )
+
+        file_path = self._maybe_compress(image)
+        try:
+            with open(file_path, "rb") as f:
+                b64 = base64.b64encode(f.read()).decode("ascii")
+        finally:
+            if file_path != image and os.path.exists(file_path):
+                os.remove(file_path)
+
+        data_url = f"data:{mime_type};base64,{b64}"
+        return {"type": "image_url", "image_url": {"url": data_url}}
+
+    @staticmethod
+    def _maybe_compress(path: str) -> str:
+        """Compress image if larger than threshold; return path to use."""
+        file_size = os.path.getsize(path)
+        if file_size <= COMPRESS_THRESHOLD:
+            return path
+
+        tmp = tempfile.NamedTemporaryFile(suffix=".jpg", delete=False)
+        tmp.close()
+
+        try:
+            # macOS: use sips
+            subprocess.run(
+                ["sips", "-Z", "800", path, "--out", tmp.name],
+                capture_output=True, check=True,
+            )
+            logger.debug(f"[Vision] Compressed image ({file_size // 1024}KB -> {os.path.getsize(tmp.name) // 1024}KB)")
+            return tmp.name
+        except (FileNotFoundError, subprocess.CalledProcessError):
+            pass
+
+        try:
+            # Linux: use ImageMagick convert
+            subprocess.run(
+                ["convert", path, "-resize", "800x800>", tmp.name],
+                capture_output=True, check=True,
+            )
+            logger.debug(f"[Vision] Compressed image ({file_size // 1024}KB -> {os.path.getsize(tmp.name) // 1024}KB)")
+            return tmp.name
+        except (FileNotFoundError, subprocess.CalledProcessError):
+            pass
+
+        os.remove(tmp.name)
+        return path
+
+    def _call_api(self, api_key: str, api_base: str, model: str,
+                  question: str, image_content: dict) -> ToolResult:
+        payload = {
+            "model": model,
+            "messages": [
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "text", "text": question},
+                        image_content,
+                    ],
+                }
+            ],
+            "max_tokens": MAX_TOKENS,
+        }
+
+        headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+        }
+
+        resp = requests.post(
+            f"{api_base}/chat/completions",
+            headers=headers,
+            json=payload,
+            timeout=DEFAULT_TIMEOUT,
+        )
+
+        if resp.status_code == 401:
+            return ToolResult.fail("Error: Invalid API key. Please check your configuration.")
+        if resp.status_code == 429:
+            return ToolResult.fail("Error: API rate limit reached. Please try again later.")
+        if resp.status_code != 200:
+            return ToolResult.fail(f"Error: Vision API returned HTTP {resp.status_code}: {resp.text[:200]}")
+
+        data = resp.json()
+
+        if "error" in data:
+            msg = data["error"].get("message", "Unknown API error")
+            return ToolResult.fail(f"Error: Vision API error - {msg}")
+
+        content = ""
+        choices = data.get("choices", [])
+        if choices:
+            content = choices[0].get("message", {}).get("content", "")
+
+        usage = data.get("usage", {})
+        result = {
+            "model": model,
+            "content": content,
+            "usage": {
+                "prompt_tokens": usage.get("prompt_tokens", 0),
+                "completion_tokens": usage.get("completion_tokens", 0),
+                "total_tokens": usage.get("total_tokens", 0),
+            },
+        }
+        return ToolResult.success(result)
--- a/agent/tools/web_fetch/init.py
+++ b/agent/tools/web_fetch/init.py
--- a/agent/tools/web_fetch/web_fetch.py
+++ b/agent/tools/web_fetch/web_fetch.py
@@ -0,0 +1,444 @@
+"""
+Web Fetch tool - Fetch and extract readable content from web pages and remote files.
+
+Supports:
+- HTML web pages: extracts readable text content
+- Document files (PDF, Word, TXT, Markdown, etc.): downloads to workspace/tmp and parses content
+"""
+
+import os
+import re
+import uuid
+from typing import Dict, Any, Optional, Set
+from urllib.parse import urlparse, unquote
+
+import requests
+
+from agent.tools.base_tool import BaseTool, ToolResult
+from agent.tools.utils.truncate import truncate_head, format_size
+from common.log import logger
+
+
+DEFAULT_TIMEOUT = 30
+MAX_FILE_SIZE = 50 * 1024 * 1024  # 50MB
+
+DEFAULT_HEADERS = {
+    "User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36",
+    "Accept": "*/*",
+}
+
+# Supported document file extensions
+PDF_SUFFIXES: Set[str] = {".pdf"}
+WORD_SUFFIXES: Set[str] = {".docx"}
+TEXT_SUFFIXES: Set[str] = {".txt", ".md", ".markdown", ".rst", ".csv", ".tsv", ".log"}
+SPREADSHEET_SUFFIXES: Set[str] = {".xls", ".xlsx"}
+PPT_SUFFIXES: Set[str] = {".ppt", ".pptx"}
+
+ALL_DOC_SUFFIXES = PDF_SUFFIXES | WORD_SUFFIXES | TEXT_SUFFIXES | SPREADSHEET_SUFFIXES | PPT_SUFFIXES
+
+_CHARSET_RE = re.compile(r'charset\s*=\s*["\']?\s*([\w\-]+)', re.IGNORECASE)
+_META_CHARSET_RE = re.compile(rb'<meta[^>]+charset\s*=\s*["\']?\s*([\w\-]+)', re.IGNORECASE)
+_META_HTTP_EQUIV_RE = re.compile(
+    rb'<meta[^>]+http-equiv\s*=\s*["\']?Content-Type["\']?[^>]+content\s*=\s*["\'][^"\']*charset=([\w\-]+)',
+    re.IGNORECASE,
+)
+
+
+def _extract_charset_from_content_type(content_type: str) -> Optional[str]:
+    """Extract charset from Content-Type header value."""
+    m = _CHARSET_RE.search(content_type)
+    return m.group(1) if m else None
+
+
+def _extract_charset_from_html_meta(raw_bytes: bytes) -> Optional[str]:
+    """Extract charset from HTML <meta> tags in the first few KB of raw bytes."""
+    m = _META_CHARSET_RE.search(raw_bytes)
+    if m:
+        return m.group(1).decode("ascii", errors="ignore")
+    m = _META_HTTP_EQUIV_RE.search(raw_bytes)
+    if m:
+        return m.group(1).decode("ascii", errors="ignore")
+    return None
+
+
+def _get_url_suffix(url: str) -> str:
+    """Extract file extension from URL path, ignoring query params."""
+    path = urlparse(url).path
+    return os.path.splitext(path)[-1].lower()
+
+
+def _is_document_url(url: str) -> bool:
+    """Check if URL points to a downloadable document file."""
+    suffix = _get_url_suffix(url)
+    return suffix in ALL_DOC_SUFFIXES
+
+
+class WebFetch(BaseTool):
+    """Tool for fetching web pages and remote document files"""
+
+    name: str = "web_fetch"
+    description: str = (
+        "Fetch content from a http/https URL. For web pages, extracts readable text. "
+        "For document files (PDF, Word, TXT, Markdown, Excel, PPT), downloads and parses the file content. "
+        "Supported file types: .pdf, .docx, .txt, .md, .csv, .xls, .xlsx, .ppt, .pptx"
+    )
+
+    params: dict = {
+        "type": "object",
+        "properties": {
+            "url": {
+                "type": "string",
+                "description": "The HTTP/HTTPS URL to fetch (web page or document file link)"
+            }
+        },
+        "required": ["url"]
+    }
+
+    def __init__(self, config: dict = None):
+        self.config = config or {}
+        self.cwd = self.config.get("cwd", os.getcwd())
+
+    def execute(self, args: Dict[str, Any]) -> ToolResult:
+        url = args.get("url", "").strip()
+        if not url:
+            return ToolResult.fail("Error: 'url' parameter is required")
+
+        parsed = urlparse(url)
+        if parsed.scheme not in ("http", "https"):
+            return ToolResult.fail("Error: Invalid URL (must start with http:// or https://)")
+
+        if _is_document_url(url):
+            return self._fetch_document(url)
+
+        return self._fetch_webpage(url)
+
+    # ---- Web page fetching ----
+
+    def _fetch_webpage(self, url: str) -> ToolResult:
+        """Fetch and extract readable text from an HTML web page."""
+        parsed = urlparse(url)
+        try:
+            response = requests.get(
+                url,
+                headers=DEFAULT_HEADERS,
+                timeout=DEFAULT_TIMEOUT,
+                allow_redirects=True,
+            )
+            response.raise_for_status()
+        except requests.Timeout:
+            return ToolResult.fail(f"Error: Request timed out after {DEFAULT_TIMEOUT}s")
+        except requests.ConnectionError:
+            return ToolResult.fail(f"Error: Failed to connect to {parsed.netloc}")
+        except requests.HTTPError as e:
+            return ToolResult.fail(f"Error: HTTP {e.response.status_code} for URL: {url}")
+        except Exception as e:
+            return ToolResult.fail(f"Error: Failed to fetch URL: {e}")
+
+        content_type = response.headers.get("Content-Type", "")
+        if self._is_binary_content_type(content_type) and not _is_document_url(url):
+            return self._handle_download_by_content_type(url, response, content_type)
+
+        response.encoding = self._detect_encoding(response)
+        html = response.text
+        title = self._extract_title(html)
+        text = self._extract_text(html)
+
+        return ToolResult.success(f"Title: {title}\n\nContent:\n{text}")
+
+    # ---- Document fetching ----
+
+    def _fetch_document(self, url: str) -> ToolResult:
+        """Download a document file and extract its text content."""
+        suffix = _get_url_suffix(url)
+        parsed = urlparse(url)
+        filename = self._extract_filename(url)
+        tmp_dir = self._ensure_tmp_dir()
+
+        local_path = os.path.join(tmp_dir, filename)
+        logger.info(f"[WebFetch] Downloading document: {url} -> {local_path}")
+
+        try:
+            response = requests.get(
+                url,
+                headers=DEFAULT_HEADERS,
+                timeout=DEFAULT_TIMEOUT,
+                stream=True,
+                allow_redirects=True,
+            )
+            response.raise_for_status()
+
+            content_length = int(response.headers.get("Content-Length", 0))
+            if content_length > MAX_FILE_SIZE:
+                return ToolResult.fail(
+                    f"Error: File too large ({format_size(content_length)} > {format_size(MAX_FILE_SIZE)})"
+                )
+
+            downloaded = 0
+            with open(local_path, "wb") as f:
+                for chunk in response.iter_content(chunk_size=8192):
+                    downloaded += len(chunk)
+                    if downloaded > MAX_FILE_SIZE:
+                        f.close()
+                        os.remove(local_path)
+                        return ToolResult.fail(
+                            f"Error: File too large (>{format_size(MAX_FILE_SIZE)}), download aborted"
+                        )
+                    f.write(chunk)
+
+        except requests.Timeout:
+            return ToolResult.fail(f"Error: Download timed out after {DEFAULT_TIMEOUT}s")
+        except requests.ConnectionError:
+            return ToolResult.fail(f"Error: Failed to connect to {parsed.netloc}")
+        except requests.HTTPError as e:
+            return ToolResult.fail(f"Error: HTTP {e.response.status_code} for URL: {url}")
+        except Exception as e:
+            self._cleanup_file(local_path)
+            return ToolResult.fail(f"Error: Failed to download file: {e}")
+
+        try:
+            text = self._parse_document(local_path, suffix)
+        except Exception as e:
+            self._cleanup_file(local_path)
+            return ToolResult.fail(f"Error: Failed to parse document: {e}")
+
+        if not text or not text.strip():
+            file_size = os.path.getsize(local_path)
+            return ToolResult.success(
+                f"File downloaded to: {local_path} ({format_size(file_size)})\n"
+                f"No text content could be extracted. The file may contain only images or be encrypted."
+            )
+
+        truncation = truncate_head(text)
+        result_text = truncation.content
+
+        file_size = os.path.getsize(local_path)
+        header = f"[Document: {filename} | Size: {format_size(file_size)} | Saved to: {local_path}]\n\n"
+
+        if truncation.truncated:
+            header += f"[Content truncated: showing {truncation.output_lines} of {truncation.total_lines} lines]\n\n"
+
+        return ToolResult.success(header + result_text)
+
+    def _parse_document(self, file_path: str, suffix: str) -> str:
+        """Parse document file and return extracted text."""
+        if suffix in PDF_SUFFIXES:
+            return self._parse_pdf(file_path)
+        elif suffix in WORD_SUFFIXES:
+            return self._parse_word(file_path)
+        elif suffix in TEXT_SUFFIXES:
+            return self._parse_text(file_path)
+        elif suffix in SPREADSHEET_SUFFIXES:
+            return self._parse_spreadsheet(file_path)
+        elif suffix in PPT_SUFFIXES:
+            return self._parse_ppt(file_path)
+        else:
+            return self._parse_text(file_path)
+
+    def _parse_pdf(self, file_path: str) -> str:
+        """Extract text from PDF using pypdf."""
+        try:
+            from pypdf import PdfReader
+        except ImportError:
+            raise ImportError("pypdf library is required for PDF parsing. Install with: pip install pypdf")
+
+        reader = PdfReader(file_path)
+        text_parts = []
+        for page_num, page in enumerate(reader.pages, 1):
+            page_text = page.extract_text()
+            if page_text and page_text.strip():
+                text_parts.append(f"--- Page {page_num}/{len(reader.pages)} ---\n{page_text}")
+
+        return "\n\n".join(text_parts)
+
+    def _parse_word(self, file_path: str) -> str:
+        """Extract text from Word documents (.docx)."""
+        try:
+            from docx import Document
+        except ImportError:
+            raise ImportError(
+                "python-docx library is required for .docx parsing. Install with: pip install python-docx"
+            )
+        doc = Document(file_path)
+        paragraphs = [p.text for p in doc.paragraphs if p.text.strip()]
+        return "\n\n".join(paragraphs)
+
+    def _parse_text(self, file_path: str) -> str:
+        """Read plain text files (txt, md, csv, etc.)."""
+        encodings = ["utf-8", "utf-8-sig", "gbk", "gb2312", "latin-1"]
+        for enc in encodings:
+            try:
+                with open(file_path, "r", encoding=enc) as f:
+                    return f.read()
+            except (UnicodeDecodeError, UnicodeError):
+                continue
+        raise ValueError(f"Unable to decode file with any supported encoding: {encodings}")
+
+    def _parse_spreadsheet(self, file_path: str) -> str:
+        """Extract text from Excel files (.xls/.xlsx)."""
+        try:
+            import openpyxl
+        except ImportError:
+            raise ImportError(
+                "openpyxl library is required for .xlsx parsing. Install with: pip install openpyxl"
+            )
+
+        wb = openpyxl.load_workbook(file_path, read_only=True, data_only=True)
+        result_parts = []
+
+        for sheet_name in wb.sheetnames:
+            ws = wb[sheet_name]
+            rows = []
+            for row in ws.iter_rows(values_only=True):
+                cells = [str(c) if c is not None else "" for c in row]
+                if any(cells):
+                    rows.append(" | ".join(cells))
+            if rows:
+                result_parts.append(f"--- Sheet: {sheet_name} ---\n" + "\n".join(rows))
+
+        wb.close()
+        return "\n\n".join(result_parts)
+
+    def _parse_ppt(self, file_path: str) -> str:
+        """Extract text from PowerPoint files (.ppt/.pptx)."""
+        try:
+            from pptx import Presentation
+        except ImportError:
+            raise ImportError(
+                "python-pptx library is required for .pptx parsing. Install with: pip install python-pptx"
+            )
+
+        prs = Presentation(file_path)
+        text_parts = []
+
+        for slide_num, slide in enumerate(prs.slides, 1):
+            slide_texts = []
+            for shape in slide.shapes:
+                if shape.has_text_frame:
+                    for paragraph in shape.text_frame.paragraphs:
+                        text = paragraph.text.strip()
+                        if text:
+                            slide_texts.append(text)
+            if slide_texts:
+                text_parts.append(f"--- Slide {slide_num}/{len(prs.slides)} ---\n" + "\n".join(slide_texts))
+
+        return "\n\n".join(text_parts)
+
+    # ---- Encoding detection ----
+
+    @staticmethod
+    def _detect_encoding(response: requests.Response) -> str:
+        """Detect response encoding with priority: Content-Type header > HTML meta > chardet > utf-8."""
+        # 1. Check Content-Type header for explicit charset
+        content_type = response.headers.get("Content-Type", "")
+        charset = _extract_charset_from_content_type(content_type)
+        if charset:
+            return charset
+
+        # 2. Scan raw bytes for HTML meta charset declaration
+        raw = response.content[:4096]
+        charset = _extract_charset_from_html_meta(raw)
+        if charset:
+            return charset
+
+        # 3. Use apparent_encoding (chardet-based detection) if confident enough
+        apparent = response.apparent_encoding
+        if apparent:
+            apparent_lower = apparent.lower()
+            # Trust CJK / Windows encodings detected by chardet
+            trusted_prefixes = ("utf", "gb", "big5", "euc", "shift_jis", "iso-2022", "windows", "ascii")
+            if any(apparent_lower.startswith(p) for p in trusted_prefixes):
+                return apparent
+
+        # 4. Fallback
+        return "utf-8"
+
+    # ---- Helper methods ----
+
+    def _ensure_tmp_dir(self) -> str:
+        """Ensure workspace/tmp directory exists and return its path."""
+        tmp_dir = os.path.join(self.cwd, "tmp")
+        os.makedirs(tmp_dir, exist_ok=True)
+        return tmp_dir
+
+    def _extract_filename(self, url: str) -> str:
+        """Extract a safe filename from URL, with a short UUID prefix to avoid collisions."""
+        path = urlparse(url).path
+        basename = os.path.basename(unquote(path))
+        if not basename or basename == "/":
+            basename = "downloaded_file"
+        # Sanitize: keep only safe chars
+        basename = re.sub(r'[^\w.\-]', '_', basename)
+        short_id = uuid.uuid4().hex[:8]
+        return f"{short_id}_{basename}"
+
+    @staticmethod
+    def _cleanup_file(path: str):
+        """Remove a file if it exists, ignoring errors."""
+        try:
+            if os.path.exists(path):
+                os.remove(path)
+        except Exception:
+            pass
+
+    @staticmethod
+    def _is_binary_content_type(content_type: str) -> bool:
+        """Check if Content-Type indicates a binary/document response."""
+        binary_types = [
+            "application/pdf",
+            "application/vnd.openxmlformats",
+            "application/vnd.ms-excel",
+            "application/vnd.ms-powerpoint",
+            "application/octet-stream",
+        ]
+        ct_lower = content_type.lower()
+        return any(bt in ct_lower for bt in binary_types)
+
+    def _handle_download_by_content_type(self, url: str, response: requests.Response, content_type: str) -> ToolResult:
+        """Handle a URL that returned binary content instead of HTML."""
+        ct_lower = content_type.lower()
+        suffix_map = {
+            "application/pdf": ".pdf",
+            "application/vnd.openxmlformats-officedocument.wordprocessingml": ".docx",
+            "application/vnd.ms-excel": ".xls",
+            "application/vnd.openxmlformats-officedocument.spreadsheetml": ".xlsx",
+            "application/vnd.ms-powerpoint": ".ppt",
+            "application/vnd.openxmlformats-officedocument.presentationml": ".pptx",
+        }
+        detected_suffix = None
+        for ct_prefix, ext in suffix_map.items():
+            if ct_prefix in ct_lower:
+                detected_suffix = ext
+                break
+
+        if detected_suffix and detected_suffix in ALL_DOC_SUFFIXES:
+            # Re-fetch as document
+            return self._fetch_document(url if _get_url_suffix(url) in ALL_DOC_SUFFIXES
+                                        else self._rewrite_url_with_suffix(url, detected_suffix))
+        return ToolResult.fail(f"Error: URL returned binary content ({content_type}), not a supported document type")
+
+    @staticmethod
+    def _rewrite_url_with_suffix(url: str, suffix: str) -> str:
+        """Append a suffix to the URL path so _get_url_suffix works correctly."""
+        parsed = urlparse(url)
+        new_path = parsed.path.rstrip("/") + suffix
+        return parsed._replace(path=new_path).geturl()
+
+    # ---- HTML extraction (unchanged) ----
+
+    @staticmethod
+    def _extract_title(html: str) -> str:
+        match = re.search(r"<title[^>]*>(.*?)</title>", html, re.IGNORECASE | re.DOTALL)
+        return match.group(1).strip() if match else "Untitled"
+
+    @staticmethod
+    def _extract_text(html: str) -> str:
+        text = re.sub(r"<script[^>]*>.*?</script>", "", html, flags=re.IGNORECASE | re.DOTALL)
+        text = re.sub(r"<style[^>]*>.*?</style>", "", text, flags=re.IGNORECASE | re.DOTALL)
+        text = re.sub(r"<[^>]+>", "", text)
+        text = text.replace("&amp;", "&").replace("&lt;", "<").replace("&gt;", ">")
+        text = text.replace("&quot;", '"').replace("&#39;", "'").replace("&nbsp;", " ")
+        text = re.sub(r"[^\S\n]+", " ", text)
+        text = re.sub(r"\n{3,}", "\n\n", text)
+        lines = [line.strip() for line in text.splitlines()]
+        text = "\n".join(lines)
+        return text.strip()
--- a/agent/tools/web_search/init.py
+++ b/agent/tools/web_search/init.py
@@ -0,0 +1,3 @@
+from agent.tools.web_search.web_search import WebSearch
+
+__all__ = ["WebSearch"]
--- a/agent/tools/web_search/web_search.py
+++ b/agent/tools/web_search/web_search.py
@@ -0,0 +1,320 @@
+"""
+Web Search tool - Search the web using Bocha or LinkAI search API.
+Supports two backends with unified response format:
+  1. Bocha Search (primary, requires BOCHA_API_KEY)
+  2. LinkAI Search (fallback, requires LINKAI_API_KEY)
+"""
+
+import os
+import json
+from typing import Dict, Any, Optional
+
+import requests
+
+from agent.tools.base_tool import BaseTool, ToolResult
+from common.log import logger
+from config import conf
+
+
+# Default timeout for API requests (seconds)
+DEFAULT_TIMEOUT = 30
+
+
+class WebSearch(BaseTool):
+    """Tool for searching the web using Bocha or LinkAI search API"""
+
+    name: str = "web_search"
+    description: str = "Search the web for real-time information. Returns titles, URLs, and snippets."
+
+    params: dict = {
+        "type": "object",
+        "properties": {
+            "query": {
+                "type": "string",
+                "description": "Search query string"
+            },
+            "count": {
+                "type": "integer",
+                "description": "Number of results to return (1-50, default: 10)"
+            },
+            "freshness": {
+                "type": "string",
+                "description": (
+                    "Time range filter. Options: "
+                    "'noLimit' (default), 'oneDay', 'oneWeek', 'oneMonth', 'oneYear', "
+                    "or date range like '2025-01-01..2025-02-01'"
+                )
+            },
+            "summary": {
+                "type": "boolean",
+                "description": "Whether to include text summary for each result (default: false)"
+            }
+        },
+        "required": ["query"]
+    }
+
+    def __init__(self, config: dict = None):
+        self.config = config or {}
+        self._backend = None  # Will be resolved on first execute
+
+    @staticmethod
+    def is_available() -> bool:
+        """Check if web search is available (at least one API key is configured)"""
+        return bool(os.environ.get("BOCHA_API_KEY") or os.environ.get("LINKAI_API_KEY"))
+
+    def _resolve_backend(self) -> Optional[str]:
+        """
+        Determine which search backend to use.
+        Priority: Bocha > LinkAI
+
+        :return: 'bocha', 'linkai', or None
+        """
+        if os.environ.get("BOCHA_API_KEY"):
+            return "bocha"
+        if os.environ.get("LINKAI_API_KEY"):
+            return "linkai"
+        return None
+
+    def execute(self, args: Dict[str, Any]) -> ToolResult:
+        """
+        Execute web search
+
+        :param args: Search parameters (query, count, freshness, summary)
+        :return: Search results
+        """
+        query = args.get("query", "").strip()
+        if not query:
+            return ToolResult.fail("Error: 'query' parameter is required")
+
+        count = args.get("count", 10)
+        freshness = args.get("freshness", "noLimit")
+        summary = args.get("summary", False)
+
+        # Validate count
+        if not isinstance(count, int) or count < 1 or count > 50:
+            count = 10
+
+        # Resolve backend
+        backend = self._resolve_backend()
+        if not backend:
+            return ToolResult.fail(
+                "Error: No search API key configured. "
+                "Please set BOCHA_API_KEY or LINKAI_API_KEY using env_config tool.\n"
+                "  - Bocha Search: https://open.bocha.cn\n"
+                "  - LinkAI Search: https://link-ai.tech"
+            )
+
+        try:
+            if backend == "bocha":
+                return self._search_bocha(query, count, freshness, summary)
+            else:
+                return self._search_linkai(query, count, freshness)
+        except requests.Timeout:
+            return ToolResult.fail(f"Error: Search request timed out after {DEFAULT_TIMEOUT}s")
+        except requests.ConnectionError:
+            return ToolResult.fail("Error: Failed to connect to search API")
+        except Exception as e:
+            logger.error(f"[WebSearch] Unexpected error: {e}", exc_info=True)
+            return ToolResult.fail(f"Error: Search failed - {str(e)}")
+
+    def _search_bocha(self, query: str, count: int, freshness: str, summary: bool) -> ToolResult:
+        """
+        Search using Bocha API
+
+        :param query: Search query
+        :param count: Number of results
+        :param freshness: Time range filter
+        :param summary: Whether to include summary
+        :return: Formatted search results
+        """
+        api_key = os.environ.get("BOCHA_API_KEY", "")
+        url = "https://api.bocha.cn/v1/web-search"
+
+        headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+            "Accept": "application/json"
+        }
+
+        payload = {
+            "query": query,
+            "count": count,
+            "freshness": freshness,
+            "summary": summary
+        }
+
+        logger.debug(f"[WebSearch] Bocha search: query='{query}', count={count}")
+
+        response = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
+
+        if response.status_code == 401:
+            return ToolResult.fail("Error: Invalid BOCHA_API_KEY. Please check your API key.")
+        if response.status_code == 403:
+            return ToolResult.fail("Error: Bocha API - insufficient balance. Please top up at https://open.bocha.cn")
+        if response.status_code == 429:
+            return ToolResult.fail("Error: Bocha API rate limit reached. Please try again later.")
+        if response.status_code != 200:
+            return ToolResult.fail(f"Error: Bocha API returned HTTP {response.status_code}")
+
+        data = response.json()
+
+        # Check API-level error code
+        api_code = data.get("code")
+        if api_code is not None and api_code != 200:
+            msg = data.get("msg") or "Unknown error"
+            return ToolResult.fail(f"Error: Bocha API error (code={api_code}): {msg}")
+
+        # Extract and format results
+        return self._format_bocha_results(data, query)
+
+    def _format_bocha_results(self, data: dict, query: str) -> ToolResult:
+        """
+        Format Bocha API response into unified result structure
+
+        :param data: Raw API response
+        :param query: Original query
+        :return: Formatted ToolResult
+        """
+        search_data = data.get("data", {})
+        web_pages = search_data.get("webPages", {})
+        pages = web_pages.get("value", [])
+
+        if not pages:
+            return ToolResult.success({
+                "query": query,
+                "backend": "bocha",
+                "total": 0,
+                "results": [],
+                "message": "No results found"
+            })
+
+        results = []
+        for page in pages:
+            result = {
+                "title": page.get("name", ""),
+                "url": page.get("url", ""),
+                "snippet": page.get("snippet", ""),
+                "siteName": page.get("siteName", ""),
+                "datePublished": page.get("datePublished") or page.get("dateLastCrawled", ""),
+            }
+            # Include summary only if present
+            if page.get("summary"):
+                result["summary"] = page["summary"]
+            results.append(result)
+
+        total = web_pages.get("totalEstimatedMatches", len(results))
+
+        return ToolResult.success({
+            "query": query,
+            "backend": "bocha",
+            "total": total,
+            "count": len(results),
+            "results": results
+        })
+
+    def _search_linkai(self, query: str, count: int, freshness: str) -> ToolResult:
+        """
+        Search using LinkAI plugin API
+
+        :param query: Search query
+        :param count: Number of results
+        :param freshness: Time range filter
+        :return: Formatted search results
+        """
+        api_key = os.environ.get("LINKAI_API_KEY", "")
+        api_base = conf().get("linkai_api_base", "https://api.link-ai.tech")
+        url = f"{api_base.rstrip('/')}/v1/plugin/execute"
+
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {api_key}"
+        }
+
+        payload = {
+            "code": "web-search",
+            "args": {
+                "query": query,
+                "count": count,
+                "freshness": freshness
+            }
+        }
+
+        logger.debug(f"[WebSearch] LinkAI search: query='{query}', count={count}")
+
+        response = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
+
+        if response.status_code == 401:
+            return ToolResult.fail("Error: Invalid LINKAI_API_KEY. Please check your API key.")
+        if response.status_code != 200:
+            return ToolResult.fail(f"Error: LinkAI API returned HTTP {response.status_code}")
+
+        data = response.json()
+
+        if not data.get("success"):
+            msg = data.get("message") or "Unknown error"
+            return ToolResult.fail(f"Error: LinkAI search failed: {msg}")
+
+        return self._format_linkai_results(data, query)
+
+    def _format_linkai_results(self, data: dict, query: str) -> ToolResult:
+        """
+        Format LinkAI API response into unified result structure.
+        LinkAI returns the search data in data.data field, which follows
+        the same Bing-compatible format as Bocha.
+
+        :param data: Raw API response
+        :param query: Original query
+        :return: Formatted ToolResult
+        """
+        raw_data = data.get("data", "")
+
+        # LinkAI may return data as a JSON string
+        if isinstance(raw_data, str):
+            try:
+                raw_data = json.loads(raw_data)
+            except (json.JSONDecodeError, TypeError):
+                # If data is plain text, return it as a single result
+                return ToolResult.success({
+                    "query": query,
+                    "backend": "linkai",
+                    "total": 1,
+                    "count": 1,
+                    "results": [{"content": raw_data}]
+                })
+
+        # If the response follows Bing-compatible structure
+        if isinstance(raw_data, dict):
+            web_pages = raw_data.get("webPages", {})
+            pages = web_pages.get("value", [])
+
+            if pages:
+                results = []
+                for page in pages:
+                    result = {
+                        "title": page.get("name", ""),
+                        "url": page.get("url", ""),
+                        "snippet": page.get("snippet", ""),
+                        "siteName": page.get("siteName", ""),
+                        "datePublished": page.get("datePublished") or page.get("dateLastCrawled", ""),
+                    }
+                    if page.get("summary"):
+                        result["summary"] = page["summary"]
+                    results.append(result)
+
+                total = web_pages.get("totalEstimatedMatches", len(results))
+                return ToolResult.success({
+                    "query": query,
+                    "backend": "linkai",
+                    "total": total,
+                    "count": len(results),
+                    "results": results
+                })
+
+        # Fallback: return raw data
+        return ToolResult.success({
+            "query": query,
+            "backend": "linkai",
+            "total": 1,
+            "count": 1,
+            "results": [{"content": str(raw_data)}]
+        })
--- a/agent/tools/write/write.py
+++ b/agent/tools/write/write.py
@@ -8,6 +8,7 @@ from typing import Dict, Any
 from pathlib import Path

 from agent.tools.base_tool import BaseTool, ToolResult
+from common.utils import expand_path


 class Write(BaseTool):
@@ -90,7 +91,7 @@ class Write(BaseTool):
        :return: Absolute path
        """
        # Expand ~ to user home directory
-        path = os.path.expanduser(path)
+        path = expand_path(path)
        if os.path.isabs(path):
            return path
        return os.path.abspath(os.path.join(self.cwd, path))
--- a/app.py
+++ b/app.py
@@ -7,11 +7,252 @@ import time

 from channel import channel_factory
 from common import const
-from config import load_config
+from common.log import logger
+from config import load_config, conf
 from plugins import *
 import threading


+_channel_mgr = None
+
+
+def get_channel_manager():
+    return _channel_mgr
+
+
+def _parse_channel_type(raw) -> list:
+    """
+    Parse channel_type config value into a list of channel names.
+    Supports:
+      - single string: "feishu"
+      - comma-separated string: "feishu, dingtalk"
+      - list: ["feishu", "dingtalk"]
+    """
+    if isinstance(raw, list):
+        return [ch.strip() for ch in raw if ch.strip()]
+    if isinstance(raw, str):
+        return [ch.strip() for ch in raw.split(",") if ch.strip()]
+    return []
+
+
+class ChannelManager:
+    """
+    Manage the lifecycle of multiple channels running concurrently.
+    Each channel.startup() runs in its own daemon thread.
+    The web channel is started as default console unless explicitly disabled.
+    """
+
+    def __init__(self):
+        self._channels = {}        # channel_name -> channel instance
+        self._threads = {}         # channel_name -> thread
+        self._primary_channel = None
+        self._lock = threading.Lock()
+        self.cloud_mode = False    # set to True when cloud client is active
+
+    @property
+    def channel(self):
+        """Return the primary (first non-web) channel for backward compatibility."""
+        return self._primary_channel
+
+    def get_channel(self, channel_name: str):
+        return self._channels.get(channel_name)
+
+    def start(self, channel_names: list, first_start: bool = False):
+        """
+        Create and start one or more channels in sub-threads.
+        If first_start is True, plugins and linkai client will also be initialized.
+        """
+        with self._lock:
+            channels = []
+            for name in channel_names:
+                ch = channel_factory.create_channel(name)
+                ch.cloud_mode = self.cloud_mode
+                self._channels[name] = ch
+                channels.append((name, ch))
+                if self._primary_channel is None and name != "web":
+                    self._primary_channel = ch
+
+            if self._primary_channel is None and channels:
+                self._primary_channel = channels[0][1]
+
+            if first_start:
+                PluginManager().load_plugins()
+
+                if conf().get("use_linkai"):
+                    try:
+                        from common import cloud_client
+                        threading.Thread(
+                            target=cloud_client.start,
+                            args=(self._primary_channel, self),
+                            daemon=True,
+                        ).start()
+                    except Exception:
+                        pass
+
+            # Start web console first so its logs print cleanly,
+            # then start remaining channels after a brief pause.
+            web_entry = None
+            other_entries = []
+            for entry in channels:
+                if entry[0] == "web":
+                    web_entry = entry
+                else:
+                    other_entries.append(entry)
+
+            ordered = ([web_entry] if web_entry else []) + other_entries
+            for i, (name, ch) in enumerate(ordered):
+                if i > 0 and name != "web":
+                    time.sleep(0.1)
+                t = threading.Thread(target=self._run_channel, args=(name, ch), daemon=True)
+                self._threads[name] = t
+                t.start()
+                logger.debug(f"[ChannelManager] Channel '{name}' started in sub-thread")
+
+    def _run_channel(self, name: str, channel):
+        try:
+            channel.startup()
+        except Exception as e:
+            logger.error(f"[ChannelManager] Channel '{name}' startup error: {e}")
+            logger.exception(e)
+
+    def stop(self, channel_name: str = None):
+        """
+        Stop channel(s). If channel_name is given, stop only that channel;
+        otherwise stop all channels.
+        """
+        # Pop under lock, then stop outside lock to avoid deadlock
+        with self._lock:
+            names = [channel_name] if channel_name else list(self._channels.keys())
+            to_stop = []
+            for name in names:
+                ch = self._channels.pop(name, None)
+                th = self._threads.pop(name, None)
+                to_stop.append((name, ch, th))
+            if channel_name and self._primary_channel is self._channels.get(channel_name):
+                self._primary_channel = None
+
+        for name, ch, th in to_stop:
+            if ch is None:
+                logger.warning(f"[ChannelManager] Channel '{name}' not found in managed channels")
+                if th and th.is_alive():
+                    self._interrupt_thread(th, name)
+                continue
+            logger.info(f"[ChannelManager] Stopping channel '{name}'...")
+            graceful = False
+            if hasattr(ch, 'stop'):
+                try:
+                    ch.stop()
+                    graceful = True
+                except Exception as e:
+                    logger.warning(f"[ChannelManager] Error during channel '{name}' stop: {e}")
+            if th and th.is_alive():
+                th.join(timeout=5)
+                if th.is_alive():
+                    if graceful:
+                        logger.info(f"[ChannelManager] Channel '{name}' thread still alive after stop(), "
+                                    "leaving daemon thread to finish on its own")
+                    else:
+                        logger.warning(f"[ChannelManager] Channel '{name}' thread did not exit in 5s, forcing interrupt")
+                        self._interrupt_thread(th, name)
+
+    @staticmethod
+    def _interrupt_thread(th: threading.Thread, name: str):
+        """Raise SystemExit in target thread to break blocking loops like start_forever."""
+        import ctypes
+        try:
+            tid = th.ident
+            if tid is None:
+                return
+            res = ctypes.pythonapi.PyThreadState_SetAsyncExc(
+                ctypes.c_ulong(tid), ctypes.py_object(SystemExit)
+            )
+            if res == 1:
+                logger.info(f"[ChannelManager] Interrupted thread for channel '{name}'")
+            elif res > 1:
+                ctypes.pythonapi.PyThreadState_SetAsyncExc(ctypes.c_ulong(tid), None)
+                logger.warning(f"[ChannelManager] Failed to interrupt thread for channel '{name}'")
+        except Exception as e:
+            logger.warning(f"[ChannelManager] Thread interrupt error for '{name}': {e}")
+
+    def restart(self, new_channel_name: str):
+        """
+        Restart a single channel with a new channel type.
+        Can be called from any thread (e.g. linkai config callback).
+        """
+        logger.info(f"[ChannelManager] Restarting channel to '{new_channel_name}'...")
+        self.stop(new_channel_name)
+        _clear_singleton_cache(new_channel_name)
+        time.sleep(1)
+        self.start([new_channel_name], first_start=False)
+        logger.info(f"[ChannelManager] Channel restarted to '{new_channel_name}' successfully")
+
+    def add_channel(self, channel_name: str):
+        """
+        Dynamically add and start a new channel.
+        If the channel is already running, restart it instead.
+        """
+        with self._lock:
+            if channel_name in self._channels:
+                logger.info(f"[ChannelManager] Channel '{channel_name}' already exists, restarting")
+        if self._channels.get(channel_name):
+            self.restart(channel_name)
+            return
+        logger.info(f"[ChannelManager] Adding channel '{channel_name}'...")
+        _clear_singleton_cache(channel_name)
+        self.start([channel_name], first_start=False)
+        logger.info(f"[ChannelManager] Channel '{channel_name}' added successfully")
+
+    def remove_channel(self, channel_name: str):
+        """
+        Dynamically stop and remove a running channel.
+        """
+        with self._lock:
+            if channel_name not in self._channels:
+                logger.warning(f"[ChannelManager] Channel '{channel_name}' not found, nothing to remove")
+                return
+        logger.info(f"[ChannelManager] Removing channel '{channel_name}'...")
+        self.stop(channel_name)
+        logger.info(f"[ChannelManager] Channel '{channel_name}' removed successfully")
+
+
+def _clear_singleton_cache(channel_name: str):
+    """
+    Clear the singleton cache for the channel class so that
+    a new instance can be created with updated config.
+    """
+    cls_map = {
+        "web": "channel.web.web_channel.WebChannel",
+        "wechatmp": "channel.wechatmp.wechatmp_channel.WechatMPChannel",
+        "wechatmp_service": "channel.wechatmp.wechatmp_channel.WechatMPChannel",
+        "wechatcom_app": "channel.wechatcom.wechatcomapp_channel.WechatComAppChannel",
+        const.FEISHU: "channel.feishu.feishu_channel.FeiShuChanel",
+        const.DINGTALK: "channel.dingtalk.dingtalk_channel.DingTalkChanel",
+        const.WECOM_BOT: "channel.wecom_bot.wecom_bot_channel.WecomBotChannel",
+        const.QQ: "channel.qq.qq_channel.QQChannel",
+    }
+    module_path = cls_map.get(channel_name)
+    if not module_path:
+        return
+    try:
+        parts = module_path.rsplit(".", 1)
+        module_name, class_name = parts[0], parts[1]
+        import importlib
+        module = importlib.import_module(module_name)
+        wrapper = getattr(module, class_name, None)
+        if wrapper and hasattr(wrapper, '__closure__') and wrapper.__closure__:
+            for cell in wrapper.__closure__:
+                try:
+                    cell_contents = cell.cell_contents
+                    if isinstance(cell_contents, dict):
+                        cell_contents.clear()
+                        logger.debug(f"[ChannelManager] Cleared singleton cache for {class_name}")
+                        break
+                except ValueError:
+                    pass
+    except Exception as e:
+        logger.warning(f"[ChannelManager] Failed to clear singleton cache: {e}")
+
+
 def sigterm_handler_wrap(_signo):
    old_handler = signal.getsignal(_signo)

@@ -25,22 +266,8 @@ def sigterm_handler_wrap(_signo):
    signal.signal(_signo, func)


-def start_channel(channel_name: str):
-    channel = channel_factory.create_channel(channel_name)
-    if channel_name in ["wx", "wxy", "terminal", "wechatmp", "web", "wechatmp_service", "wechatcom_app", "wework",
-                        const.FEISHU, const.DINGTALK]:
-        PluginManager().load_plugins()
-
-    if conf().get("use_linkai"):
-        try:
-            from common import linkai_client
-            threading.Thread(target=linkai_client.start, args=(channel,)).start()
-        except Exception as e:
-            pass
-    channel.startup()
-
-
 def run():
+    global _channel_mgr
    try:
        # load config
        load_config()
@@ -49,16 +276,25 @@ def run():
        # kill signal
        sigterm_handler_wrap(signal.SIGTERM)

-        # create channel
-        channel_name = conf().get("channel_type", "wx")
+        # Parse channel_type into a list
+        raw_channel = conf().get("channel_type", "web")

        if "--cmd" in sys.argv:
-            channel_name = "terminal"
+            channel_names = ["terminal"]
+        else:
+            channel_names = _parse_channel_type(raw_channel)
+            if not channel_names:
+                channel_names = ["web"]

-        if channel_name == "wxy":
-            os.environ["WECHATY_LOG"] = "warn"
+        # Auto-start web console unless explicitly disabled
+        web_console_enabled = conf().get("web_console", True)
+        if web_console_enabled and "web" not in channel_names:
+            channel_names.append("web")

-        start_channel(channel_name)
+        logger.info(f"[App] Starting channels: {channel_names}")
+
+        _channel_mgr = ChannelManager()
+        _channel_mgr.start(channel_names, first_start=True)

        while True:
            time.sleep(1)
--- a/bridge/agent_bridge.py
+++ b/bridge/agent_bridge.py
@@ -13,6 +13,7 @@ from bridge.context import Context
 from bridge.reply import Reply, ReplyType
 from common import const
 from common.log import logger
+from common.utils import expand_path
 from models.openai_compatible_bot import OpenAICompatibleBot


@@ -27,7 +28,7 @@ def add_openai_compatible_support(bot_instance):
    """
    if hasattr(bot_instance, 'call_with_tools'):
        # Bot already has tool calling support (e.g., ZHIPUAIBot)
-        logger.info(f"[AgentBridge] {type(bot_instance).__name__} already has native tool calling support")
+        logger.debug(f"[AgentBridge] {type(bot_instance).__name__} already has native tool calling support")
        return bot_instance

    # Create a temporary mixin class that combines the bot with OpenAI compatibility
@@ -64,30 +65,73 @@ class AgentLLMModel(LLMModel):
    LLM Model adapter that uses COW's existing bot infrastructure
    """

+    _MODEL_BOT_TYPE_MAP = {
+        "wenxin": const.BAIDU, "wenxin-4": const.BAIDU,
+        "xunfei": const.XUNFEI, const.QWEN: const.QWEN,
+        const.MODELSCOPE: const.MODELSCOPE,
+    }
+    _MODEL_PREFIX_MAP = [
+        ("qwen", const.QWEN_DASHSCOPE), ("qwq", const.QWEN_DASHSCOPE), ("qvq", const.QWEN_DASHSCOPE),
+        ("gemini", const.GEMINI), ("glm", const.ZHIPU_AI), ("claude", const.CLAUDEAPI),
+        ("moonshot", const.MOONSHOT), ("kimi", const.MOONSHOT),
+        ("doubao", const.DOUBAO),
+    ]
+
    def __init__(self, bridge: Bridge, bot_type: str = "chat"):
-        # Get model name directly from config
        from config import conf
-        model_name = conf().get("model", const.GPT_41)
-        super().__init__(model=model_name)
+        super().__init__(model=conf().get("model", const.GPT_41))
        self.bridge = bridge
        self.bot_type = bot_type
        self._bot = None
-        self._use_linkai = conf().get("use_linkai", False) and conf().get("linkai_api_key")
-    
+        self._bot_model = None
+
+    @property
+    def model(self):
+        from config import conf
+        return conf().get("model", const.GPT_41)
+
+    @model.setter
+    def model(self, value):
+        pass
+
+    def _resolve_bot_type(self, model_name: str) -> str:
+        """Resolve bot type from model name, matching Bridge.__init__ logic."""
+        from config import conf
+
+        if conf().get("use_linkai", False) and conf().get("linkai_api_key"):
+            return const.LINKAI
+        # Support custom bot type configuration
+        configured_bot_type = conf().get("bot_type")
+        if configured_bot_type:
+            return configured_bot_type
+       
+        if not model_name or not isinstance(model_name, str):
+            return const.OPENAI
+        if model_name in self._MODEL_BOT_TYPE_MAP:
+            return self._MODEL_BOT_TYPE_MAP[model_name]
+        if model_name.lower().startswith("minimax") or model_name in ["abab6.5-chat"]:
+            return const.MiniMax
+        if model_name in [const.QWEN_TURBO, const.QWEN_PLUS, const.QWEN_MAX]:
+            return const.QWEN_DASHSCOPE
+        if model_name in [const.MOONSHOT, "moonshot-v1-8k", "moonshot-v1-32k", "moonshot-v1-128k"]:
+            return const.MOONSHOT
+        if model_name in [const.DEEPSEEK_CHAT, const.DEEPSEEK_REASONER]:
+            return const.OPENAI
+        for prefix, btype in self._MODEL_PREFIX_MAP:
+            if model_name.startswith(prefix):
+                return btype
+        return const.OPENAI
+
    @property
    def bot(self):
-        """Lazy load the bot and enhance it with tool calling if needed"""
-        if self._bot is None:
-            # If use_linkai is enabled, use LinkAI bot directly
-            if self._use_linkai:
-                self._bot = self.bridge.find_chat_bot(const.LINKAI)
-            else:
-                self._bot = self.bridge.get_bot(self.bot_type)
-                # Automatically add tool calling support if not present
-                self._bot = add_openai_compatible_support(self._bot)
-            
-            # Log bot info
-            bot_name = type(self._bot).__name__
+        """Lazy load the bot, re-create when model changes"""
+        from models.bot_factory import create_bot
+        cur_model = self.model
+        if self._bot is None or self._bot_model != cur_model:
+            bot_type = self._resolve_bot_type(cur_model)
+            self._bot = create_bot(bot_type)
+            self._bot = add_openai_compatible_support(self._bot)
+            self._bot_model = cur_model
        return self._bot

    def call(self, request: LLMRequest):
@@ -134,7 +178,7 @@ class AgentLLMModel(LLMModel):
                # Use tool-enabled streaming call if available
                # Extract system prompt if present
                system_prompt = getattr(request, 'system', None)
-                
+
                # Build kwargs for call_with_tools
                kwargs = {
                    'messages': request.messages,
@@ -142,15 +186,20 @@ class AgentLLMModel(LLMModel):
                    'stream': True,
                    'model': self.model  # Pass model parameter
                }
-                
+
                # Only pass max_tokens if explicitly set, let the bot use its default
                if request.max_tokens is not None:
                    kwargs['max_tokens'] = request.max_tokens
-                
+
                # Add system prompt if present
                if system_prompt:
                    kwargs['system'] = system_prompt
-                
+
+                # Pass channel_type for linkai tracking
+                channel_type = getattr(self, 'channel_type', None)
+                if channel_type:
+                    kwargs['channel_type'] = channel_type
+
                stream = self.bot.call_with_tools(**kwargs)
                
                # Convert stream format to our expected format
@@ -229,11 +278,13 @@ class AgentBridge:
            tools=tools,
            max_steps=kwargs.get("max_steps", 15),
            output_mode=kwargs.get("output_mode", "logger"),
-            workspace_dir=kwargs.get("workspace_dir"),  # Pass workspace for skills loading
-            enable_skills=kwargs.get("enable_skills", True),  # Enable skills by default
-            memory_manager=kwargs.get("memory_manager"),  # Pass memory manager
+            workspace_dir=kwargs.get("workspace_dir"),
+            skill_manager=kwargs.get("skill_manager"),
+            enable_skills=kwargs.get("enable_skills", True),
+            memory_manager=kwargs.get("memory_manager"),
            max_context_tokens=kwargs.get("max_context_tokens"),
-            context_reserve_tokens=kwargs.get("context_reserve_tokens")
+            context_reserve_tokens=kwargs.get("context_reserve_tokens"),
+            runtime_info=kwargs.get("runtime_info"),
        )

        # Log skill loading details
@@ -288,9 +339,10 @@ class AgentBridge:
        Returns:
            Reply object
        """
+        session_id = None
+        agent = None
        try:
            # Extract session_id from context for user isolation
-            session_id = None
            if context:
                session_id = context.kwargs.get("session_id") or context.get("session_id")
            
@@ -323,6 +375,13 @@ class AgentBridge:
                                logger.warning(f"[AgentBridge] Failed to attach context to scheduler: {e}")
                            break
            
+            # Pass channel_type to model so linkai requests carry it
+            if context and hasattr(agent, 'model'):
+                agent.model.channel_type = context.get("channel_type", "")
+
+            # Store session_id on agent so executor can clear DB on fatal errors
+            agent._current_session_id = session_id
+
            try:
                # Use agent's run_stream method with event handler
                response = agent.run_stream(
@@ -334,9 +393,26 @@ class AgentBridge:
                # Restore original tools
                if context and context.get("is_scheduled_task"):
                    agent.tools = original_tools
-                
+
                # Log execution summary
                event_handler.log_summary()
+
+            # Persist new messages generated during this run
+            if session_id:
+                channel_type = (context.get("channel_type") or "") if context else ""
+                new_messages = getattr(agent, '_last_run_new_messages', [])
+                if new_messages:
+                    self._persist_messages(session_id, list(new_messages), channel_type)
+                else:
+                    with agent.messages_lock:
+                        msg_count = len(agent.messages)
+                    if msg_count == 0:
+                        try:
+                            from agent.memory import get_conversation_store
+                            get_conversation_store().clear_session(session_id)
+                            logger.info(f"[AgentBridge] Cleared DB for recovered session: {session_id}")
+                        except Exception as e:
+                            logger.warning(f"[AgentBridge] Failed to clear DB after recovery: {e}")
            
            # Check if there are files to send (from read tool)
            if hasattr(agent, 'stream_executor') and hasattr(agent.stream_executor, 'files_to_send'):
@@ -356,6 +432,18 @@ class AgentBridge:
            
        except Exception as e:
            logger.error(f"Agent reply error: {e}")
+            # If the agent cleared its messages due to format error / overflow,
+            # also purge the DB so the next request starts clean.
+            if session_id and agent:
+                try:
+                    with agent.messages_lock:
+                        msg_count = len(agent.messages)
+                    if msg_count == 0:
+                        from agent.memory import get_conversation_store
+                        get_conversation_store().clear_session(session_id)
+                        logger.info(f"[AgentBridge] Cleared DB for session after error: {session_id}")
+                except Exception as db_err:
+                    logger.warning(f"[AgentBridge] Failed to clear DB after error: {db_err}")
            return Reply(ReplyType.ERROR, f"Agent error: {str(e)}")
    
    def _create_file_reply(self, file_info: dict, text_response: str, context: Context = None) -> Reply:
@@ -420,7 +508,7 @@ class AgentBridge:
        }
        
        # Use fixed secure location for .env file
-        env_file = os.path.expanduser("~/.cow/.env")
+        env_file = expand_path("~/.cow/.env")
        
        # Read existing env vars from .env file
        existing_env_vars = {}
@@ -473,6 +561,32 @@ class AgentBridge:
            except Exception as e:
                logger.warning(f"[AgentBridge] Failed to migrate API keys: {e}")
    
+    def _persist_messages(
+        self, session_id: str, new_messages: list, channel_type: str = ""
+    ) -> None:
+        """
+        Persist new messages to the conversation store after each agent run.
+
+        Failures are logged but never propagate — they must not interrupt replies.
+        """
+        if not new_messages:
+            return
+        try:
+            from config import conf
+            if not conf().get("conversation_persistence", True):
+                return
+        except Exception:
+            pass
+        try:
+            from agent.memory import get_conversation_store
+            get_conversation_store().append_messages(
+                session_id, new_messages, channel_type=channel_type
+            )
+        except Exception as e:
+            logger.warning(
+                f"[AgentBridge] Failed to persist messages for session={session_id}: {e}"
+            )
+
    def clear_session(self, session_id: str):
        """
        Clear a specific session's agent and conversation history
@@ -492,39 +606,70 @@ class AgentBridge:
    
    def refresh_all_skills(self) -> int:
        """
-        Refresh skills in all agent instances after environment variable changes.
-        This allows hot-reload of skills without restarting the agent.
-        
+        Refresh skills and conditional tools in all agent instances after
+        environment variable changes. This allows hot-reload without restarting.
+
        Returns:
            Number of agent instances refreshed
        """
        import os
        from dotenv import load_dotenv
        from config import conf
-        
+
        # Reload environment variables from .env file
-        workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+        workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
        env_file = os.path.join(workspace_root, '.env')
-        
+
        if os.path.exists(env_file):
            load_dotenv(env_file, override=True)
            logger.info(f"[AgentBridge] Reloaded environment variables from {env_file}")
-        
+
        refreshed_count = 0
-        
-        # Refresh default agent
-        if self.default_agent and hasattr(self.default_agent, 'skill_manager'):
-            self.default_agent.skill_manager.refresh_skills()
-            refreshed_count += 1
-            logger.info("[AgentBridge] Refreshed skills in default agent")
-        
-        # Refresh all session agents
+
+        # Collect all agent instances to refresh
+        agents_to_refresh = []
+        if self.default_agent:
+            agents_to_refresh.append(("default", self.default_agent))
        for session_id, agent in self.agents.items():
-            if hasattr(agent, 'skill_manager'):
+            agents_to_refresh.append((session_id, agent))
+
+        for label, agent in agents_to_refresh:
+            # Refresh skills
+            if hasattr(agent, 'skill_manager') and agent.skill_manager:
                agent.skill_manager.refresh_skills()
-                refreshed_count += 1
-        
+
+            # Refresh conditional tools (e.g. web_search depends on API keys)
+            self._refresh_conditional_tools(agent)
+
+            refreshed_count += 1
+
        if refreshed_count > 0:
-            logger.info(f"[AgentBridge] Refreshed skills in {refreshed_count} agent instance(s)")
-        
-        return refreshed_count
+            logger.info(f"[AgentBridge] Refreshed skills & tools in {refreshed_count} agent instance(s)")
+
+        return refreshed_count
+
+    @staticmethod
+    def _refresh_conditional_tools(agent):
+        """
+        Add or remove conditional tools based on current environment variables.
+        For example, web_search should only be present when BOCHA_API_KEY or
+        LINKAI_API_KEY is set.
+        """
+        try:
+            from agent.tools.web_search.web_search import WebSearch
+
+            has_tool = any(t.name == "web_search" for t in agent.tools)
+            available = WebSearch.is_available()
+
+            if available and not has_tool:
+                # API key was added - inject the tool
+                tool = WebSearch()
+                tool.model = agent.model
+                agent.tools.append(tool)
+                logger.info("[AgentBridge] web_search tool added (API key now available)")
+            elif not available and has_tool:
+                # API key was removed - remove the tool
+                agent.tools = [t for t in agent.tools if t.name != "web_search"]
+                logger.info("[AgentBridge] web_search tool removed (API key no longer available)")
+        except Exception as e:
+            logger.debug(f"[AgentBridge] Failed to refresh conditional tools: {e}")
--- a/bridge/agent_event_handler.py
+++ b/bridge/agent_event_handler.py
@@ -74,7 +74,7 @@ class AgentEventHandler:
        # Only send thinking process if followed by tool calls
        if tool_calls:
            if self.current_thinking.strip():
-                logger.debug(f"💭 {self.current_thinking.strip()[:200]}{'...' if len(self.current_thinking) > 200 else ''}")
+                logger.info(f"💭 {self.current_thinking.strip()[:200]}{'...' if len(self.current_thinking) > 200 else ''}")
                # Send thinking process to channel
                self._send_to_channel(f"{self.current_thinking.strip()}")
        else:
@@ -94,15 +94,15 @@ class AgentEventHandler:
    
    def _send_to_channel(self, message):
        """
-        Try to send message to channel
-        
-        Args:
-            message: Message to send
+        Try to send intermediate message to channel.
+        Skipped in SSE mode because thinking text is already streamed via on_event.
        """
+        if self.context and self.context.get("on_event"):
+            return
+
        if self.channel:
            try:
                from bridge.reply import Reply, ReplyType
-                # Create a Reply object for the message
                reply = Reply(ReplyType.TEXT, message)
                self.channel._send(reply, self.context)
            except Exception as e:
--- a/bridge/agent_initializer.py
+++ b/bridge/agent_initializer.py
@@ -11,6 +11,7 @@ from typing import Optional, List
 from agent.protocol import Agent
 from agent.tools import ToolManager
 from common.log import logger
+from common.utils import expand_path


 class AgentInitializer:
@@ -46,7 +47,7 @@ class AgentInitializer:
        from config import conf
        
        # Get workspace from config
-        workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+        workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
        
        # Migrate API keys
        self._migrate_config_to_env(workspace_root)
@@ -76,10 +77,6 @@ class AgentInitializer:
        # Initialize skill manager
        skill_manager = self._initialize_skill_manager(workspace_root, session_id)
        
-        # Check if first conversation
-        from agent.prompt.workspace import is_first_conversation, mark_conversation_started
-        is_first = is_first_conversation(workspace_root)
-        
        # Build system prompt
        prompt_builder = PromptBuilder(workspace_dir=workspace_root, language="zh")
        runtime_info = self._get_runtime_info(workspace_root)
@@ -90,12 +87,8 @@ class AgentInitializer:
            skill_manager=skill_manager,
            memory_manager=memory_manager,
            runtime_info=runtime_info,
-            is_first_conversation=is_first
        )
        
-        if is_first:
-            mark_conversation_started(workspace_root)
-        
        # Get cost control parameters
        from config import conf
        max_steps = conf().get("agent_max_steps", 20)
@@ -110,18 +103,143 @@ class AgentInitializer:
            workspace_dir=workspace_root,
            skill_manager=skill_manager,
            enable_skills=True,
-            max_context_tokens=max_context_tokens
+            max_context_tokens=max_context_tokens,
+            runtime_info=runtime_info  # Pass runtime_info for dynamic time updates
        )
        
-        # Attach memory manager
+        # Attach memory manager and share LLM model for summarization
        if memory_manager:
            agent.memory_manager = memory_manager
-        
+            if hasattr(agent, 'model') and agent.model:
+                memory_manager.flush_manager.llm_model = agent.model
+
+        # Restore persisted conversation history for this session
+        if session_id:
+            self._restore_conversation_history(agent, session_id)
+
+        # Start daily memory flush timer (once, on first agent init regardless of session)
+        self._start_daily_flush_timer()
+
        return agent
+
+    def _restore_conversation_history(self, agent, session_id: str) -> None:
+        """
+        Load persisted conversation messages from SQLite and inject them
+        into the agent's in-memory message list.
+
+        Only user text and assistant text are restored. Tool call chains
+        (tool_use / tool_result) are stripped out because:
+        1. They are intermediate process, the value is already in the final
+           assistant text reply.
+        2. They consume massive context tokens (often 80%+ of history).
+        3. Different models have incompatible tool message formats, so
+           restoring tool chains across model switches causes 400 errors.
+        4. Eliminates the entire class of tool_use/tool_result pairing bugs.
+        """
+        from config import conf
+        if not conf().get("conversation_persistence", True):
+            return
+
+        try:
+            from agent.memory import get_conversation_store
+            store = get_conversation_store()
+            max_turns = conf().get("agent_max_context_turns", 20)
+            restore_turns = max(3, max_turns // 6)
+            saved = store.load_messages(session_id, max_turns=restore_turns)
+            if saved:
+                filtered = self._filter_text_only_messages(saved)
+                if filtered:
+                    with agent.messages_lock:
+                        agent.messages = filtered
+                    logger.debug(
+                        f"[AgentInitializer] Restored {len(filtered)} text messages "
+                        f"(from {len(saved)} total, {restore_turns} turns cap) "
+                        f"for session={session_id}"
+                    )
+        except Exception as e:
+            logger.warning(
+                f"[AgentInitializer] Failed to restore conversation history for "
+                f"session={session_id}: {e}"
+            )
+
+    @staticmethod
+    def _filter_text_only_messages(messages: list) -> list:
+        """
+        Extract clean user/assistant turn pairs from raw message history.
+
+        Groups messages into turns (each starting with a real user query),
+        then keeps only:
+        - The first user text in each turn (the actual user input)
+        - The last assistant text in each turn (the final answer)
+
+        All tool_use, tool_result, intermediate assistant thoughts, and
+        internal hint messages injected by the agent loop are discarded.
+        """
+
+        def _extract_text(content) -> str:
+            if isinstance(content, str):
+                return content.strip()
+            if isinstance(content, list):
+                parts = [
+                    b.get("text", "")
+                    for b in content
+                    if isinstance(b, dict) and b.get("type") == "text"
+                ]
+                return "\n".join(p for p in parts if p).strip()
+            return ""
+
+        def _is_real_user_msg(msg: dict) -> bool:
+            """True for actual user input, False for tool_result or internal hints."""
+            if msg.get("role") != "user":
+                return False
+            content = msg.get("content")
+            if isinstance(content, list):
+                has_tool_result = any(
+                    isinstance(b, dict) and b.get("type") == "tool_result"
+                    for b in content
+                )
+                if has_tool_result:
+                    return False
+            text = _extract_text(content)
+            return bool(text)
+
+        # Group into turns: each turn starts with a real user message
+        turns = []
+        current_turn = None
+        for msg in messages:
+            if _is_real_user_msg(msg):
+                if current_turn is not None:
+                    turns.append(current_turn)
+                current_turn = {"user": msg, "assistants": []}
+            elif current_turn is not None and msg.get("role") == "assistant":
+                text = _extract_text(msg.get("content"))
+                if text:
+                    current_turn["assistants"].append(text)
+        if current_turn is not None:
+            turns.append(current_turn)
+
+        # Build result: one user msg + one assistant msg per turn
+        filtered = []
+        for turn in turns:
+            user_text = _extract_text(turn["user"].get("content"))
+            if not user_text:
+                continue
+            filtered.append({
+                "role": "user",
+                "content": [{"type": "text", "text": user_text}]
+            })
+            if turn["assistants"]:
+                final_reply = turn["assistants"][-1]
+                filtered.append({
+                    "role": "assistant",
+                    "content": [{"type": "text", "text": final_reply}]
+                })
+
+        return filtered
    
    def _load_env_file(self):
        """Load environment variables from .env file"""
-        env_file = os.path.expanduser("~/.cow/.env")
+        env_file = expand_path("~/.cow/.env")
        if os.path.exists(env_file):
            try:
                from dotenv import load_dotenv
@@ -146,12 +264,11 @@ class AgentInitializer:
            from agent.tools import MemorySearchTool, MemoryGetTool
            from config import conf
            
-            # Get OpenAI config
+            # Initialize embedding provider (prefer OpenAI, fallback to LinkAI)
+            embedding_provider = None
+
            openai_api_key = conf().get("open_ai_api_key", "")
            openai_api_base = conf().get("open_ai_api_base", "")
-            
-            # Initialize embedding provider
-            embedding_provider = None
            if openai_api_key and openai_api_key not in ["", "YOUR API KEY", "YOUR_API_KEY"]:
                try:
                    embedding_provider = create_embedding_provider(
@@ -164,6 +281,22 @@ class AgentInitializer:
                        logger.info("[AgentInitializer] OpenAI embedding initialized")
                except Exception as e:
                    logger.warning(f"[AgentInitializer] OpenAI embedding failed: {e}")
+
+            if embedding_provider is None:
+                linkai_api_key = conf().get("linkai_api_key", "") or os.environ.get("LINKAI_API_KEY", "")
+                linkai_api_base = conf().get("linkai_api_base", "https://api.link-ai.tech")
+                if linkai_api_key and linkai_api_key not in ["", "YOUR API KEY", "YOUR_API_KEY"]:
+                    try:
+                        embedding_provider = create_embedding_provider(
+                            provider="linkai",
+                            model="text-embedding-3-small",
+                            api_key=linkai_api_key,
+                            api_base=f"{linkai_api_base}/v1"
+                        )
+                        if session_id is None:
+                            logger.info("[AgentInitializer] LinkAI embedding initialized (fallback)")
+                    except Exception as e:
+                        logger.warning(f"[AgentInitializer] LinkAI embedding failed: {e}")
            
            # Create memory manager
            memory_config = MemoryConfig(workspace_root=workspace_root)
@@ -217,16 +350,23 @@ class AgentInitializer:
        
        for tool_name in tool_manager.tool_classes.keys():
            try:
+                # Skip web_search if no API key is available
+                if tool_name == "web_search":
+                    from agent.tools.web_search.web_search import WebSearch
+                    if not WebSearch.is_available():
+                        logger.debug("[AgentInitializer] WebSearch skipped - no BOCHA_API_KEY or LINKAI_API_KEY")
+                        continue
+
                # Special handling for EnvConfig tool
                if tool_name == "env_config":
                    from agent.tools import EnvConfig
                    tool = EnvConfig({"agent_bridge": self.agent_bridge})
                else:
                    tool = tool_manager.create_tool(tool_name)
-                
+
                if tool:
                    # Apply workspace config to file operation tools
-                    if tool_name in ['read', 'write', 'edit', 'bash', 'grep', 'find', 'ls']:
+                    if tool_name in ['read', 'write', 'edit', 'bash', 'grep', 'find', 'ls', 'web_fetch']:
                        tool.config = file_config
                        tool.cwd = file_config.get("cwd", getattr(tool, 'cwd', None))
                        if 'memory_manager' in file_config:
@@ -274,7 +414,14 @@ class AgentInitializer:
                        tool.scheduler_service = scheduler_service
                        if not tool.config:
                            tool.config = {}
-                        tool.config["channel_type"] = conf().get("channel_type", "unknown")
+                        raw_ct = conf().get("channel_type", "unknown")
+                        if isinstance(raw_ct, list):
+                            ct = raw_ct[0] if raw_ct else "unknown"
+                        elif isinstance(raw_ct, str) and "," in raw_ct:
+                            ct = raw_ct.split(",")[0].strip()
+                        else:
+                            ct = raw_ct
+                        tool.config["channel_type"] = ct
            except Exception as e:
                logger.warning(f"[AgentInitializer] Failed to inject scheduler dependencies: {e}")
    
@@ -282,41 +429,47 @@ class AgentInitializer:
        """Initialize skill manager"""
        try:
            from agent.skills import SkillManager
-            skill_manager = SkillManager(workspace_dir=workspace_root)
+            skill_manager = SkillManager(custom_dir=os.path.join(workspace_root, "skills"))
            return skill_manager
        except Exception as e:
            logger.warning(f"[AgentInitializer] Failed to initialize SkillManager: {e}")
            return None
    
    def _get_runtime_info(self, workspace_root: str):
-        """Get runtime information"""
+        """Get runtime information with dynamic time support"""
        from config import conf
        
-        now = datetime.datetime.now()
-        
-        # Get timezone info
-        try:
-            offset = -time.timezone if not time.daylight else -time.altzone
-            hours = offset // 3600
-            minutes = (offset % 3600) // 60
-            timezone_name = f"UTC{hours:+03d}:{minutes:02d}" if minutes else f"UTC{hours:+03d}"
-        except Exception:
-            timezone_name = "UTC"
-        
-        # Chinese weekday mapping
-        weekday_map = {
-            'Monday': '星期一', 'Tuesday': '星期二', 'Wednesday': '星期三',
-            'Thursday': '星期四', 'Friday': '星期五', 'Saturday': '星期六', 'Sunday': '星期日'
-        }
-        weekday_zh = weekday_map.get(now.strftime("%A"), now.strftime("%A"))
+        def get_current_time():
+            """Get current time dynamically - called each time system prompt is accessed"""
+            now = datetime.datetime.now()
+            
+            # Get timezone info
+            try:
+                offset = -time.timezone if not time.daylight else -time.altzone
+                hours = offset // 3600
+                minutes = (offset % 3600) // 60
+                timezone_name = f"UTC{hours:+03d}:{minutes:02d}" if minutes else f"UTC{hours:+03d}"
+            except Exception:
+                timezone_name = "UTC"
+            
+            # Chinese weekday mapping
+            weekday_map = {
+                'Monday': '星期一', 'Tuesday': '星期二', 'Wednesday': '星期三',
+                'Thursday': '星期四', 'Friday': '星期五', 'Saturday': '星期六', 'Sunday': '星期日'
+            }
+            weekday_zh = weekday_map.get(now.strftime("%A"), now.strftime("%A"))
+            
+            return {
+                'time': now.strftime("%Y-%m-%d %H:%M:%S"),
+                'weekday': weekday_zh,
+                'timezone': timezone_name
+            }
        
        return {
            "model": conf().get("model", "unknown"),
            "workspace": workspace_root,
-            "channel": conf().get("channel_type", "unknown"),
-            "current_time": now.strftime("%Y-%m-%d %H:%M:%S"),
-            "weekday": weekday_zh,
-            "timezone": timezone_name
+            "channel": ", ".join(conf().get("channel_type")) if isinstance(conf().get("channel_type"), list) else conf().get("channel_type", "unknown"),
+            "_get_current_time": get_current_time  # Dynamic time function
        }
    
    def _migrate_config_to_env(self, workspace_root: str):
@@ -331,7 +484,7 @@ class AgentInitializer:
            "linkai_api_key": "LINKAI_API_KEY",
        }
        
-        env_file = os.path.expanduser("~/.cow/.env")
+        env_file = expand_path("~/.cow/.env")
        
        # Read existing env vars
        existing_env_vars = {}
@@ -373,3 +526,59 @@ class AgentInitializer:
                logger.info(f"[AgentInitializer] Migrated {len(keys_to_migrate)} API keys to .env: {list(keys_to_migrate.keys())}")
            except Exception as e:
                logger.warning(f"[AgentInitializer] Failed to migrate API keys: {e}")
+
+    def _start_daily_flush_timer(self):
+        """Start a background thread that flushes all agents' memory daily at 23:55."""
+        if getattr(self.agent_bridge, '_daily_flush_started', False):
+            return
+        self.agent_bridge._daily_flush_started = True
+
+        import threading
+
+        def _daily_flush_loop():
+            while True:
+                try:
+                    now = datetime.datetime.now()
+                    target = now.replace(hour=23, minute=55, second=0, microsecond=0)
+                    if target <= now:
+                        target += datetime.timedelta(days=1)
+                    wait_seconds = (target - now).total_seconds()
+                    logger.info(f"[DailyFlush] Next flush at {target.strftime('%Y-%m-%d %H:%M')} (in {wait_seconds/3600:.1f}h)")
+                    time.sleep(wait_seconds)
+
+                    self._flush_all_agents()
+                except Exception as e:
+                    logger.warning(f"[DailyFlush] Error in daily flush loop: {e}")
+                    time.sleep(3600)
+
+        t = threading.Thread(target=_daily_flush_loop, daemon=True)
+        t.start()
+
+    def _flush_all_agents(self):
+        """Flush memory for all active agent sessions."""
+        agents = []
+        if self.agent_bridge.default_agent:
+            agents.append(("default", self.agent_bridge.default_agent))
+        for sid, agent in self.agent_bridge.agents.items():
+            agents.append((sid, agent))
+
+        if not agents:
+            return
+
+        flushed = 0
+        for label, agent in agents:
+            try:
+                if not agent.memory_manager:
+                    continue
+                with agent.messages_lock:
+                    messages = list(agent.messages)
+                if not messages:
+                    continue
+                result = agent.memory_manager.flush_manager.create_daily_summary(messages)
+                if result:
+                    flushed += 1
+            except Exception as e:
+                logger.warning(f"[DailyFlush] Failed for session {label}: {e}")
+
+        if flushed:
+            logger.info(f"[DailyFlush] Flushed {flushed}/{len(agents)} agent session(s)")
--- a/bridge/bridge.py
+++ b/bridge/bridge.py
@@ -13,7 +13,7 @@ from voice.factory import create_voice
 class Bridge(object):
    def __init__(self):
        self.btype = {
-            "chat": const.CHATGPT,
+            "chat": const.OPENAI,
            "voice_to_text": conf().get("voice_to_text", "openai"),
            "text_to_voice": conf().get("text_to_voice", "google"),
            "translate": conf().get("translate", "baidu"),
@@ -24,6 +24,13 @@ class Bridge(object):
            self.btype["chat"] = bot_type
        else:
            model_type = conf().get("model") or const.GPT_41_MINI
+            
+            # Ensure model_type is string to prevent AttributeError when using startswith()
+            # This handles cases where numeric model names (e.g., "1") are parsed as integers from YAML
+            if not isinstance(model_type, str):
+                logger.warning(f"[Bridge] model_type is not a string: {model_type} (type: {type(model_type).__name__}), converting to string")
+                model_type = str(model_type)
+            
            if model_type in ["text-davinci-003"]:
                self.btype["chat"] = const.OPEN_AI
            if conf().get("use_azure_chatgpt", False):
@@ -48,6 +55,11 @@ class Bridge(object):

            if model_type in [const.MOONSHOT, "moonshot-v1-8k", "moonshot-v1-32k", "moonshot-v1-128k"]:
                self.btype["chat"] = const.MOONSHOT
+            if model_type and model_type.startswith("kimi"):
+                self.btype["chat"] = const.MOONSHOT
+
+            if model_type and model_type.startswith("doubao"):
+                self.btype["chat"] = const.DOUBAO

            if model_type in [const.MODELSCOPE]:
                self.btype["chat"] = const.MODELSCOPE
--- a/channel/channel.py
+++ b/channel/channel.py
@@ -13,12 +13,44 @@ class Channel(object):
    channel_type = ""
    NOT_SUPPORT_REPLYTYPE = [ReplyType.VOICE, ReplyType.IMAGE]

+    def __init__(self):
+        import threading
+        self._startup_event = threading.Event()
+        self._startup_error = None
+        self.cloud_mode = False  # set to True by ChannelManager when running with cloud client
+
    def startup(self):
        """
        init channel
        """
        raise NotImplementedError

+    def report_startup_success(self):
+        self._startup_error = None
+        self._startup_event.set()
+
+    def report_startup_error(self, error: str):
+        self._startup_error = error
+        self._startup_event.set()
+
+    def wait_startup(self, timeout: float = 3) -> (bool, str):
+        """
+        Wait for channel startup result.
+        Returns (success: bool, error_msg: str).
+        """
+        ready = self._startup_event.wait(timeout=timeout)
+        if not ready:
+            return True, ""
+        if self._startup_error:
+            return False, self._startup_error
+        return True, ""
+
+    def stop(self):
+        """
+        stop channel gracefully, called before restart
+        """
+        pass
+
    def handle_text(self, msg):
        """
        process received msg
@@ -51,11 +83,14 @@ class Channel(object):
                if context and "channel_type" not in context:
                    context["channel_type"] = self.channel_type

+                # Read on_event callback injected by the channel (e.g. web SSE)
+                on_event = context.get("on_event") if context else None
+
                # Use agent bridge to handle the query
                return Bridge().fetch_agent_reply(
                    query=query,
                    context=context,
-                    on_event=None,
+                    on_event=on_event,
                    clear_history=False
                )
            except Exception as e:
--- a/channel/channel_factory.py
+++ b/channel/channel_factory.py
@@ -12,16 +12,7 @@ def create_channel(channel_type) -> Channel:
    :return: channel instance
    """
    ch = Channel()
-    if channel_type == "wx":
-        from channel.wechat.wechat_channel import WechatChannel
-        ch = WechatChannel()
-    elif channel_type == "wxy":
-        from channel.wechat.wechaty_channel import WechatyChannel
-        ch = WechatyChannel()
-    elif channel_type == "wcf":
-        from channel.wechat.wcf_channel import WechatfChannel
-        ch = WechatfChannel()
-    elif channel_type == "terminal":
+    if channel_type == "terminal":
        from channel.terminal.terminal_channel import TerminalChannel
        ch = TerminalChannel()
    elif channel_type == 'web':
@@ -36,15 +27,18 @@ def create_channel(channel_type) -> Channel:
    elif channel_type == "wechatcom_app":
        from channel.wechatcom.wechatcomapp_channel import WechatComAppChannel
        ch = WechatComAppChannel()
-    elif channel_type == "wework":
-        from channel.wework.wework_channel import WeworkChannel
-        ch = WeworkChannel()
    elif channel_type == const.FEISHU:
        from channel.feishu.feishu_channel import FeiShuChanel
        ch = FeiShuChanel()
    elif channel_type == const.DINGTALK:
        from channel.dingtalk.dingtalk_channel import DingTalkChanel
        ch = DingTalkChanel()
+    elif channel_type == const.WECOM_BOT:
+        from channel.wecom_bot.wecom_bot_channel import WecomBotChannel
+        ch = WecomBotChannel()
+    elif channel_type == const.QQ:
+        from channel.qq.qq_channel import QQChannel
+        ch = QQChannel()
    else:
        raise RuntimeError
    ch.channel_type = channel_type
--- a/channel/chat_channel.py
+++ b/channel/chat_channel.py
@@ -24,11 +24,17 @@ handler_pool = ThreadPoolExecutor(max_workers=8)  # 处理消息的线程池
 class ChatChannel(Channel):
    name = None  # 登录的用户名
    user_id = None  # 登录的用户id
-    futures = {}  # 记录每个session_id提交到线程池的future对象, 用于重置会话时把没执行的future取消掉，正在执行的不会被取消
-    sessions = {}  # 用于控制并发，每个session_id同时只能有一个context在处理
-    lock = threading.Lock()  # 用于控制对sessions的访问

    def __init__(self):
+        super().__init__()
+        # Instance-level attributes so each channel subclass has its own
+        # independent session queue and lock. Previously these were class-level,
+        # which caused contexts from one channel (e.g. Feishu) to be consumed
+        # by another channel's consume() thread (e.g. Web), leading to errors
+        # like "No request_id found in context".
+        self.futures = {}
+        self.sessions = {}
+        self.lock = threading.Lock()
        _thread = threading.Thread(target=self.consume)
        _thread.setDaemon(True)
        _thread.start()
@@ -37,9 +43,8 @@ class ChatChannel(Channel):
    def _compose_context(self, ctype: ContextType, content, **kwargs):
        context = Context(ctype, content)
        context.kwargs = kwargs
-        # context首次传入时，origin_ctype是None,
-        # 引入的起因是：当输入语音时，会嵌套生成两个context，第一步语音转文本，第二步通过文本生成文字回复。
-        # origin_ctype用于第二步文本回复时，判断是否需要匹配前缀，如果是私聊的语音，就不需要匹配前缀
+        if "channel_type" not in context:
+            context["channel_type"] = self.channel_type
        if "origin_ctype" not in context:
            context["origin_ctype"] = ctype
        # context首次传入时，receiver是None，根据类型设置receiver
@@ -426,7 +431,7 @@ class ChatChannel(Channel):
            if session_id not in self.sessions:
                self.sessions[session_id] = [
                    Dequeue(),
-                    threading.BoundedSemaphore(conf().get("concurrency_in_session", 4)),
+                    threading.BoundedSemaphore(conf().get("concurrency_in_session", 1)),
                ]
            if context.type == ContextType.TEXT and context.content.startswith("#"):
                self.sessions[session_id][0].putleft(context)  # 优先处理管理命令
--- a/channel/chat_message.py
+++ b/channel/chat_message.py
@@ -1,5 +1,5 @@
 """
-本类表示聊天消息，用于对itchat和wechaty的消息进行统一的封装。
+Unified chat message class for different channel implementations.

 填好必填项(群聊6个，非群聊8个)，即可接入ChatChannel，并支持插件，参考TerminalChannel

--- a/channel/dingtalk/dingtalk_channel.py
+++ b/channel/dingtalk/dingtalk_channel.py
@@ -21,6 +21,7 @@ from dingtalk_stream.card_replier import CardReplier
 from bridge.context import Context, ContextType
 from bridge.reply import Reply, ReplyType
 from channel.chat_channel import ChatChannel
+from common.utils import expand_path
 from channel.dingtalk.dingtalk_message import DingTalkMessage
 from common.expired_dict import ExpiredDict
 from common.log import logger
@@ -89,13 +90,9 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
    dingtalk_client_secret = conf().get('dingtalk_client_secret')

    def setup_logger(self):
-        logger = logging.getLogger()
-        handler = logging.StreamHandler()
-        handler.setFormatter(
-            logging.Formatter('%(asctime)s %(name)-8s %(levelname)-8s %(message)s [%(filename)s:%(lineno)d]'))
-        logger.addHandler(handler)
-        logger.setLevel(logging.INFO)
-        return logger
+        # Suppress verbose logs from dingtalk_stream SDK
+        logging.getLogger("dingtalk_stream").setLevel(logging.WARNING)
+        return logging.getLogger("DingTalk")

    def __init__(self):
        super().__init__()
@@ -103,6 +100,9 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
        self.logger = self.setup_logger()
        # 历史消息id暂存，用于幂等控制
        self.receivedMsgs = ExpiredDict(conf().get("expires_in_seconds", 3600))
+        self._stream_client = None
+        self._running = False
+        self._event_loop = None
        logger.debug("[DingTalk] client_id={}, client_secret={} ".format(
            self.dingtalk_client_id, self.dingtalk_client_secret))
        # 无需群校验和前缀
@@ -115,12 +115,130 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
        # Robot code cache (extracted from incoming messages)
        self._robot_code = None

+    def _open_connection(self, client):
+        """
+        Open a DingTalk stream connection directly, bypassing SDK's internal error-swallowing.
+        Returns (connection_dict, error_str). On success error_str is empty; on failure
+        connection_dict is None and error_str contains a human-readable message.
+        """
+        try:
+            resp = requests.post(
+                "https://api.dingtalk.com/v1.0/gateway/connections/open",
+                headers={"Content-Type": "application/json", "Accept": "application/json"},
+                json={
+                    "clientId": client.credential.client_id,
+                    "clientSecret": client.credential.client_secret,
+                    "subscriptions": [{"type": "CALLBACK",
+                                       "topic": dingtalk_stream.chatbot.ChatbotMessage.TOPIC}],
+                    "ua": "dingtalk-sdk-python/cow",
+                    "localIp": "",
+                },
+                timeout=10,
+            )
+            body = resp.json()
+            if not resp.ok:
+                code = body.get("code", resp.status_code)
+                message = body.get("message", resp.reason)
+                return None, f"open connection failed: [{code}] {message}"
+            return body, ""
+        except Exception as e:
+            return None, f"open connection failed: {e}"
+
    def startup(self):
+        import asyncio
+        self.dingtalk_client_id = conf().get('dingtalk_client_id')
+        self.dingtalk_client_secret = conf().get('dingtalk_client_secret')
+        self._running = True
        credential = dingtalk_stream.Credential(self.dingtalk_client_id, self.dingtalk_client_secret)
        client = dingtalk_stream.DingTalkStreamClient(credential)
+        self._stream_client = client
        client.register_callback_handler(dingtalk_stream.chatbot.ChatbotMessage.TOPIC, self)
-        logger.info("[DingTalk] ✅ Stream connected, ready to receive messages")
-        client.start_forever()
+        logger.info("[DingTalk] ✅ Stream client initialized, ready to receive messages")
+
+        # Run the connection loop ourselves instead of delegating to client.start(),
+        # so we can get detailed error messages and respond to stop() quickly.
+        import urllib.parse as _urlparse
+        import websockets as _ws
+        import json as _json
+        client.pre_start()
+        _first_connect = True
+        while self._running:
+            # Open connection using our own request so we get detailed error info.
+            connection, err_msg = self._open_connection(client)
+
+            if connection is None:
+                if _first_connect:
+                    logger.warning(f"[DingTalk] {err_msg}")
+                    self.report_startup_error(err_msg)
+                    _first_connect = False
+                else:
+                    logger.warning(f"[DingTalk] {err_msg}, retrying in 10s...")
+
+                # Interruptible sleep: checks _running every 100ms.
+                for _ in range(100):
+                    if not self._running:
+                        break
+                    time.sleep(0.1)
+                continue
+
+            if _first_connect:
+                logger.info("[DingTalk] ✅ Connected to DingTalk stream")
+                self.report_startup_success()
+                _first_connect = False
+            else:
+                logger.info("[DingTalk] Reconnected to DingTalk stream")
+
+            # Run the WebSocket session in an asyncio loop.
+            uri = '%s?ticket=%s' % (
+                connection['endpoint'],
+                _urlparse.quote_plus(connection['ticket'])
+            )
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
+            self._event_loop = loop
+            try:
+                async def _session():
+                    async with _ws.connect(uri) as websocket:
+                        client.websocket = websocket
+                        async for raw_message in websocket:
+                            json_message = _json.loads(raw_message)
+                            result = await client.route_message(json_message)
+                            if result == dingtalk_stream.DingTalkStreamClient.TAG_DISCONNECT:
+                                break
+
+                loop.run_until_complete(_session())
+            except (KeyboardInterrupt, SystemExit):
+                logger.info("[DingTalk] Session loop received stop signal, exiting")
+                break
+            except Exception as e:
+                if not self._running:
+                    break
+                logger.warning(f"[DingTalk] Stream session error: {e}, reconnecting in 3s...")
+                for _ in range(30):
+                    if not self._running:
+                        break
+                    time.sleep(0.1)
+            finally:
+                self._event_loop = None
+                try:
+                    loop.close()
+                except Exception:
+                    pass
+
+        logger.info("[DingTalk] Startup loop exited")
+
+    def stop(self):
+        logger.info("[DingTalk] stop() called, setting _running=False")
+        self._running = False
+        loop = self._event_loop
+        if loop and not loop.is_closed():
+            try:
+                loop.call_soon_threadsafe(loop.stop)
+                logger.info("[DingTalk] Sent stop signal to event loop")
+            except Exception as e:
+                logger.warning(f"[DingTalk] Error stopping event loop: {e}")
+        self._stream_client = None
+        logger.info("[DingTalk] stop() completed")
    
    def get_access_token(self):
        """
@@ -276,7 +394,7 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
                
                # 保存到临时文件
                file_name = os.path.basename(file_path) or f"media_{uuid.uuid4()}"
-                workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+                workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
                tmp_dir = os.path.join(workspace_root, "tmp")
                os.makedirs(tmp_dir, exist_ok=True)
                temp_file = os.path.join(tmp_dir, file_name)
@@ -457,23 +575,21 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
    async def process(self, callback: dingtalk_stream.CallbackMessage):
        try:
            incoming_message = dingtalk_stream.ChatbotMessage.from_dict(callback.data)
-            
+
            # 缓存 robot_code，用于后续图片下载
            if hasattr(incoming_message, 'robot_code'):
                self._robot_code_cache = incoming_message.robot_code
-            
-            # Debug: 打印完整的 event 数据
-            logger.debug(f"[DingTalk] ===== Incoming Message Debug =====")
-            logger.debug(f"[DingTalk] callback.data keys: {callback.data.keys() if hasattr(callback.data, 'keys') else 'N/A'}")
-            logger.debug(f"[DingTalk] incoming_message attributes: {dir(incoming_message)}")
-            logger.debug(f"[DingTalk] robot_code: {getattr(incoming_message, 'robot_code', 'N/A')}")
-            logger.debug(f"[DingTalk] chatbot_corp_id: {getattr(incoming_message, 'chatbot_corp_id', 'N/A')}")
-            logger.debug(f"[DingTalk] chatbot_user_id: {getattr(incoming_message, 'chatbot_user_id', 'N/A')}")
-            logger.debug(f"[DingTalk] conversation_id: {getattr(incoming_message, 'conversation_id', 'N/A')}")
-            logger.debug(f"[DingTalk] Raw callback.data: {callback.data}")
-            logger.debug(f"[DingTalk] =====================================")
-            
-            image_download_handler = self  # 传入方法所在的类实例
+
+            # Filter out stale messages from before channel startup (offline backlog)
+            create_at = getattr(incoming_message, 'create_at', None)
+            if create_at:
+                msg_age_s = time.time() - int(create_at) / 1000
+                if msg_age_s > 60:
+                    logger.warning(f"[DingTalk] stale msg filtered (age={msg_age_s:.0f}s), "
+                                   f"msg_id={getattr(incoming_message, 'message_id', 'N/A')}")
+                    return AckMessage.STATUS_OK, 'OK'
+
+            image_download_handler = self
            dingtalk_msg = DingTalkMessage(incoming_message, image_download_handler)

            if dingtalk_msg.is_group:
@@ -482,8 +598,7 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
                self.handle_single(dingtalk_msg)
            return AckMessage.STATUS_OK, 'OK'
        except Exception as e:
-            logger.error(f"[DingTalk] process error: {e}")
-            logger.exception(e)  # 打印完整堆栈跟踪
+            logger.error(f"[DingTalk] process error: {e}", exc_info=True)
            return AckMessage.STATUS_SYSTEM_EXCEPTION, 'ERROR'

    @time_checker
@@ -607,7 +722,7 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):


    def send(self, reply: Reply, context: Context):
-        logger.info(f"[DingTalk] send() called with reply.type={reply.type}, content_length={len(str(reply.content))}")
+        logger.debug(f"[DingTalk] send() called with reply.type={reply.type}, content_length={len(str(reply.content))}")
        receiver = context["receiver"]
        
        # Check if msg exists (for scheduled tasks, msg might be None)
@@ -647,7 +762,7 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
            robot_code = msg.robot_code
            if robot_code and robot_code != self._robot_code:
                self._robot_code = robot_code
-                logger.info(f"[DingTalk] Cached robot_code: {robot_code}")
+                logger.debug(f"[DingTalk] Cached robot_code: {robot_code}")
        
        isgroup = msg.is_group
        incoming_message = msg.incoming_message
--- a/channel/dingtalk/dingtalk_message.py
+++ b/channel/dingtalk/dingtalk_message.py
@@ -9,6 +9,7 @@ from channel.chat_message import ChatMessage
 # -*- coding=utf-8 -*-
 from common.log import logger
 from common.tmp_dir import TmpDir
+from common.utils import expand_path
 from config import conf


@@ -49,7 +50,7 @@ class DingTalkMessage(ChatMessage):
                download_url = image_download_handler.get_image_download_url(download_code)
                
                # 下载到工作空间 tmp 目录
-                workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+                workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
                tmp_dir = os.path.join(workspace_root, "tmp")
                os.makedirs(tmp_dir, exist_ok=True)
                
@@ -67,7 +68,7 @@ class DingTalkMessage(ChatMessage):
                self.ctype = ContextType.TEXT
                
                # 下载到工作空间 tmp 目录
-                workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+                workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
                tmp_dir = os.path.join(workspace_root, "tmp")
                os.makedirs(tmp_dir, exist_ok=True)
                
--- a/channel/feishu/README.md
+++ b/channel/feishu/README.md
@@ -140,6 +140,23 @@ python3 app.py

 **解决**: 安装依赖 `pip install lark-oapi`

+### SSL证书验证失败
+
+```
+[Lark][ERROR] connect failed, err:[SSL:CERTIFICATE_VERIFY_FAILED] certificate verify failed: self signed certificate in certificate chain
+```
+
+**原因**: 网络环境中存在自签名证书或SSL中间人代理(如企业代理、VPN等)
+
+**解决**: 程序会自动检测SSL证书验证失败，并自动重试禁用证书验证的连接。无需手动配置。
+
+当遇到证书错误时，日志会显示：
+```
+[FeiShu] SSL certificate verification disabled due to certificate error. This may happen when using corporate proxy or self-signed certificates.
+```
+
+这是正常现象，程序会自动处理并继续运行。
+
 ### Webhook模式端口被占用

 ```
--- a/channel/feishu/feishu_channel.py
+++ b/channel/feishu/feishu_channel.py
@@ -11,8 +11,11 @@
@Date 2023/11/19
 """

+import importlib.util
 import json
+import logging
 import os
+import ssl
 import threading
 # -*- coding=utf-8 -*-
 import uuid
@@ -31,17 +34,25 @@ from common.log import logger
 from common.singleton import singleton
 from config import conf

+# Suppress verbose logs from Lark SDK
+logging.getLogger("Lark").setLevel(logging.WARNING)
+
 URL_VERIFICATION = "url_verification"

-# 尝试导入飞书SDK,如果未安装则websocket模式不可用
-try:
-    import lark_oapi as lark
+# Lazy-check for lark_oapi SDK availability without importing it at module level.
+# The full `import lark_oapi` pulls in 10k+ files and takes 4-10s, so we defer
+# the actual import to _startup_websocket() where it is needed.
+LARK_SDK_AVAILABLE = importlib.util.find_spec("lark_oapi") is not None
+lark = None  # will be populated on first use via _ensure_lark_imported()

-    LARK_SDK_AVAILABLE = True
-except ImportError:
-    LARK_SDK_AVAILABLE = False
-    logger.warning(
-        "[FeiShu] lark_oapi not installed, websocket mode is not available. Install with: pip install lark-oapi")
+
+def _ensure_lark_imported():
+    """Import lark_oapi on first use (takes 4-10s due to 10k+ source files)."""
+    global lark
+    if lark is None:
+        import lark_oapi as _lark
+        lark = _lark
+    return lark


@singleton
@@ -55,6 +66,10 @@ class FeiShuChanel(ChatChannel):
        super().__init__()
        # 历史消息id暂存，用于幂等控制
        self.receivedMsgs = ExpiredDict(60 * 60 * 7.1)
+        self._http_server = None
+        self._ws_client = None
+        self._ws_thread = None
+        self._bot_open_id = None  # cached bot open_id for @-mention matching
        logger.debug("[FeiShu] app_id={}, app_secret={}, verification_token={}, event_mode={}".format(
            self.feishu_app_id, self.feishu_app_secret, self.feishu_token, self.feishu_event_mode))
        # 无需群校验和前缀
@@ -67,11 +82,66 @@ class FeiShuChanel(ChatChannel):
            raise Exception("lark_oapi not installed")

    def startup(self):
+        self.feishu_app_id = conf().get('feishu_app_id')
+        self.feishu_app_secret = conf().get('feishu_app_secret')
+        self.feishu_token = conf().get('feishu_token')
+        self.feishu_event_mode = conf().get('feishu_event_mode', 'websocket')
+        self._fetch_bot_open_id()
        if self.feishu_event_mode == 'websocket':
            self._startup_websocket()
        else:
            self._startup_webhook()

+    def _fetch_bot_open_id(self):
+        """Fetch the bot's own open_id via API so we can match @-mentions without feishu_bot_name."""
+        try:
+            access_token = self.fetch_access_token()
+            if not access_token:
+                logger.warning("[FeiShu] Cannot fetch bot info: no access_token")
+                return
+            headers = {"Authorization": "Bearer " + access_token}
+            resp = requests.get("https://open.feishu.cn/open-apis/bot/v3/info/", headers=headers, timeout=5)
+            if resp.status_code == 200:
+                data = resp.json()
+                if data.get("code") == 0:
+                    self._bot_open_id = data.get("bot", {}).get("open_id")
+                    logger.info(f"[FeiShu] Bot open_id fetched: {self._bot_open_id}")
+                else:
+                    logger.warning(f"[FeiShu] Fetch bot info failed: code={data.get('code')}, msg={data.get('msg')}")
+        except Exception as e:
+            logger.warning(f"[FeiShu] Fetch bot open_id error: {e}")
+
+    def stop(self):
+        import ctypes
+        logger.info("[FeiShu] stop() called")
+        ws_client = self._ws_client
+        self._ws_client = None
+        ws_thread = self._ws_thread
+        self._ws_thread = None
+        # Interrupt the ws thread first so its blocking start() unblocks
+        if ws_thread and ws_thread.is_alive():
+            try:
+                tid = ws_thread.ident
+                if tid:
+                    res = ctypes.pythonapi.PyThreadState_SetAsyncExc(
+                        ctypes.c_ulong(tid), ctypes.py_object(SystemExit)
+                    )
+                    if res == 1:
+                        logger.info("[FeiShu] Interrupted ws thread via ctypes")
+                    elif res > 1:
+                        ctypes.pythonapi.PyThreadState_SetAsyncExc(ctypes.c_ulong(tid), None)
+            except Exception as e:
+                logger.warning(f"[FeiShu] Error interrupting ws thread: {e}")
+        # lark.ws.Client has no stop() method; thread interruption above is sufficient
+        if self._http_server:
+            try:
+                self._http_server.stop()
+                logger.info("[FeiShu] HTTP server stopped")
+            except Exception as e:
+                logger.warning(f"[FeiShu] Error stopping HTTP server: {e}")
+            self._http_server = None
+        logger.info("[FeiShu] stop() completed")
+
    def _startup_webhook(self):
        """启动HTTP服务器接收事件(webhook模式)"""
        logger.debug("[FeiShu] Starting in webhook mode...")
@@ -80,21 +150,33 @@ class FeiShuChanel(ChatChannel):
        )
        app = web.application(urls, globals(), autoreload=False)
        port = conf().get("feishu_port", 9891)
-        web.httpserver.runsimple(app.wsgifunc(), ("0.0.0.0", port))
+        func = web.httpserver.StaticMiddleware(app.wsgifunc())
+        func = web.httpserver.LogMiddleware(func)
+        server = web.httpserver.WSGIServer(("0.0.0.0", port), func)
+        self._http_server = server
+        try:
+            server.start()
+        except (KeyboardInterrupt, SystemExit):
+            server.stop()

    def _startup_websocket(self):
        """启动长连接接收事件(websocket模式)"""
+        _ensure_lark_imported()
        logger.debug("[FeiShu] Starting in websocket mode...")

        # 创建事件处理器
        def handle_message_event(data: lark.im.v1.P2ImMessageReceiveV1) -> None:
            """处理接收消息事件 v2.0"""
            try:
-                logger.debug(f"[FeiShu] websocket receive event: {lark.JSON.marshal(data, indent=2)}")
-
-                # 转换为标准的event格式
                event_dict = json.loads(lark.JSON.marshal(data))
                event = event_dict.get("event", {})
+                msg = event.get("message", {})
+
+                # Skip group messages that don't @-mention the bot (reduce log noise)
+                if msg.get("chat_type") == "group" and not msg.get("mentions") and msg.get("message_type") == "text":
+                    return
+
+                logger.debug(f"[FeiShu] websocket receive event: {lark.JSON.marshal(data, indent=2)}")

                # 处理消息
                self._handle_message_event(event)
@@ -107,29 +189,99 @@ class FeiShuChanel(ChatChannel):
            .register_p2_im_message_receive_v1(handle_message_event) \
            .build()

-        # 创建长连接客户端
-        ws_client = lark.ws.Client(
-            self.feishu_app_id,
-            self.feishu_app_secret,
-            event_handler=event_handler,
-            log_level=lark.LogLevel.DEBUG if conf().get("debug") else lark.LogLevel.INFO
-        )
+        def start_client_with_retry():
+            """Run ws client in this thread with its own event loop to avoid conflicts."""
+            import asyncio
+            import ssl as ssl_module
+            original_create_default_context = ssl_module.create_default_context

-        # 在新线程中启动客户端，避免阻塞主线程
-        def start_client():
+            def create_unverified_context(*args, **kwargs):
+                context = original_create_default_context(*args, **kwargs)
+                context.check_hostname = False
+                context.verify_mode = ssl.CERT_NONE
+                return context
+
+            # lark_oapi.ws.client captures the event loop at module-import time as a module-
+            # level global variable.  When a previous ws thread is force-killed via ctypes its
+            # loop may still be marked as "running", which causes the next ws_client.start()
+            # call (in this new thread) to raise "This event loop is already running".
+            # Fix: replace the module-level loop with a brand-new, idle loop before starting.
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
            try:
-                logger.debug("[FeiShu] Websocket client starting...")
-                ws_client.start()
-            except Exception as e:
-                logger.error(f"[FeiShu] Websocket client error: {e}", exc_info=True)
+                import lark_oapi.ws.client as _lark_ws_client_mod
+                _lark_ws_client_mod.loop = loop
+            except Exception:
+                pass

-        ws_thread = threading.Thread(target=start_client, daemon=True)
+            startup_error = None
+            for attempt in range(2):
+                try:
+                    if attempt == 1:
+                        logger.warning("[FeiShu] Retrying with SSL verification disabled...")
+                        ssl_module.create_default_context = create_unverified_context
+                        ssl_module._create_unverified_context = create_unverified_context
+
+                    ws_client = lark.ws.Client(
+                        self.feishu_app_id,
+                        self.feishu_app_secret,
+                        event_handler=event_handler,
+                        log_level=lark.LogLevel.WARNING
+                    )
+                    self._ws_client = ws_client
+                    logger.debug("[FeiShu] Websocket client starting...")
+                    ws_client.start()
+                    break
+
+                except (SystemExit, KeyboardInterrupt):
+                    logger.info("[FeiShu] Websocket thread received stop signal")
+                    break
+                except Exception as e:
+                    error_msg = str(e)
+                    is_ssl_error = ("CERTIFICATE_VERIFY_FAILED" in error_msg
+                                    or "certificate verify failed" in error_msg.lower())
+                    if is_ssl_error and attempt == 0:
+                        logger.warning(f"[FeiShu] SSL error: {error_msg}, retrying...")
+                        continue
+                    logger.error(f"[FeiShu] Websocket client error: {e}", exc_info=True)
+                    startup_error = error_msg
+                    ssl_module.create_default_context = original_create_default_context
+                    break
+            if startup_error:
+                self.report_startup_error(startup_error)
+            try:
+                loop.close()
+            except Exception:
+                pass
+            logger.info("[FeiShu] Websocket thread exited")
+
+        ws_thread = threading.Thread(target=start_client_with_retry, daemon=True)
+        self._ws_thread = ws_thread
        ws_thread.start()
-
-        # 保持主线程运行
-        logger.info("[FeiShu] ✅ Websocket connected, ready to receive messages")
+        logger.info("[FeiShu] ✅ Websocket thread started, ready to receive messages")
        ws_thread.join()

+    def _is_mention_bot(self, mentions: list) -> bool:
+        """Check whether any mention in the list refers to this bot.
+
+        Priority:
+        1. Match by open_id (obtained from /bot/v3/info at startup, no config needed)
+        2. Fallback to feishu_bot_name config for backward compatibility
+        3. If neither is available, assume the first mention is the bot (Feishu only
+           delivers group messages that @-mention the bot, so this is usually correct)
+        """
+        if self._bot_open_id:
+            return any(
+                m.get("id", {}).get("open_id") == self._bot_open_id
+                for m in mentions
+            )
+        bot_name = conf().get("feishu_bot_name")
+        if bot_name:
+            return any(m.get("name") == bot_name for m in mentions)
+        # Feishu event subscription only delivers messages that @-mention the bot,
+        # so reaching here means the bot was indeed mentioned.
+        return True
+
    def _handle_message_event(self, event: dict):
        """
        处理消息事件的核心逻辑
@@ -148,6 +300,15 @@ class FeiShuChanel(ChatChannel):
            return
        self.receivedMsgs[msg_id] = True

+        # Filter out stale messages from before channel startup (offline backlog)
+        import time as _time
+        create_time_ms = msg.get("create_time")
+        if create_time_ms:
+            msg_age_s = _time.time() - int(create_time_ms) / 1000
+            if msg_age_s > 60:
+                logger.warning(f"[FeiShu] stale msg filtered (age={msg_age_s:.0f}s), msg_id={msg_id}")
+                return
+
        is_group = False
        chat_type = msg.get("chat_type")

@@ -155,10 +316,9 @@ class FeiShuChanel(ChatChannel):
            if not msg.get("mentions") and msg.get("message_type") == "text":
                # 群聊中未@不响应
                return
-            if msg.get("mentions") and msg.get("mentions")[0].get("name") != conf().get("feishu_bot_name") and msg.get(
-                    "message_type") == "text":
-                # 不是@机器人，不响应
-                return
+            if msg.get("mentions") and msg.get("message_type") == "text":
+                if not self._is_mention_bot(msg.get("mentions")):
+                    return
            # 群聊
            is_group = True
            receive_id_type = "chat_id"
@@ -176,7 +336,7 @@ class FeiShuChanel(ChatChannel):
        # 处理文件缓存逻辑
        from channel.file_cache import get_file_cache
        file_cache = get_file_cache()
-        
+
        # 获取 session_id（用于缓存关联）
        if is_group:
            if conf().get("group_shared_session", True):
@@ -185,7 +345,7 @@ class FeiShuChanel(ChatChannel):
                session_id = feishu_msg.from_user_id + "_" + msg.get("chat_id")
        else:
            session_id = feishu_msg.from_user_id
-        
+
        # 如果是单张图片消息，缓存起来
        if feishu_msg.ctype == ContextType.IMAGE:
            if hasattr(feishu_msg, 'image_path') and feishu_msg.image_path:
@@ -193,7 +353,7 @@ class FeiShuChanel(ChatChannel):
                logger.info(f"[FeiShu] Image cached for session {session_id}, waiting for user query...")
            # 单张图片不直接处理，等待用户提问
            return
-        
+
        # 如果是文本消息，检查是否有缓存的文件
        if feishu_msg.ctype == ContextType.TEXT:
            cached_files = file_cache.get(session_id)
@@ -209,7 +369,7 @@ class FeiShuChanel(ChatChannel):
                        file_refs.append(f"[视频: {file_path}]")
                    else:
                        file_refs.append(f"[文件: {file_path}]")
-                
+
                feishu_msg.content = feishu_msg.content + "\n" + "\n".join(file_refs)
                logger.info(f"[FeiShu] Attached {len(cached_files)} cached file(s) to user query")
                # 清除缓存
@@ -258,26 +418,27 @@ class FeiShuChanel(ChatChannel):
                self._send(text_reply, context)
                import time
                time.sleep(0.3)  # 短暂延迟，确保文本先到达
-            
+
            # 判断是否为视频文件
            file_path = reply.content
            if file_path.startswith("file://"):
                file_path = file_path[7:]
-            
+
            is_video = file_path.lower().endswith(('.mp4', '.avi', '.mov', '.wmv', '.flv'))
-            
+
            if is_video:
                # 视频上传（包含duration信息）
                upload_data = self._upload_video_url(reply.content, access_token)
                if not upload_data or not upload_data.get('file_key'):
                    logger.warning("[FeiShu] upload video failed")
                    return
-                
+
                # 视频使用 media 类型（根据官方文档）
                # 错误码 230055 说明：上传 mp4 时必须使用 msg_type="media"
                msg_type = "media"
                reply_content = upload_data  # 完整的上传响应数据（包含file_key和duration）
-                logger.info(f"[FeiShu] Sending video: file_key={upload_data.get('file_key')}, duration={upload_data.get('duration')}ms")
+                logger.info(
+                    f"[FeiShu] Sending video: file_key={upload_data.get('file_key')}, duration={upload_data.get('duration')}ms")
                content_key = None  # 直接序列化整个对象
            else:
                # 其他文件使用 file 类型
@@ -288,14 +449,14 @@ class FeiShuChanel(ChatChannel):
                reply_content = file_key
                msg_type = "file"
                content_key = "file_key"
-        
+
        # Check if we can reply to an existing message (need msg_id)
        can_reply = is_group and msg and hasattr(msg, 'msg_id') and msg.msg_id
-        
+
        # Build content JSON
        content_json = json.dumps(reply_content) if content_key is None else json.dumps({content_key: reply_content})
        logger.debug(f"[FeiShu] Sending message: msg_type={msg_type}, content={content_json[:200]}")
-        
+
        if can_reply:
            # 群聊中回复已有消息
            url = f"https://open.feishu.cn/open-apis/im/v1/messages/{msg.msg_id}/reply"
@@ -320,7 +481,6 @@ class FeiShuChanel(ChatChannel):
        else:
            logger.error(f"[FeiShu] send message failed, code={res.get('code')}, msg={res.get('msg')}")

-
    def fetch_access_token(self) -> str:
        url = "https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal/"
        headers = {
@@ -342,35 +502,34 @@ class FeiShuChanel(ChatChannel):
        else:
            logger.error(f"[FeiShu] fetch token error, res={response}")

-
    def _upload_image_url(self, img_url, access_token):
        logger.debug(f"[FeiShu] start process image, img_url={img_url}")
-        
+
        # Check if it's a local file path (file:// protocol)
        if img_url.startswith("file://"):
            local_path = img_url[7:]  # Remove "file://" prefix
            logger.info(f"[FeiShu] uploading local file: {local_path}")
-            
+
            if not os.path.exists(local_path):
                logger.error(f"[FeiShu] local file not found: {local_path}")
                return None
-            
+
            # Upload directly from local file
            upload_url = "https://open.feishu.cn/open-apis/im/v1/images"
            data = {'image_type': 'message'}
            headers = {'Authorization': f'Bearer {access_token}'}
-            
+
            with open(local_path, "rb") as file:
                upload_response = requests.post(upload_url, files={"image": file}, data=data, headers=headers)
                logger.info(f"[FeiShu] upload file, res={upload_response.content}")
-                
+
                response_data = upload_response.json()
                if response_data.get("code") == 0:
                    return response_data.get("data").get("image_key")
                else:
                    logger.error(f"[FeiShu] upload failed: {response_data}")
                    return None
-        
+
        # Original logic for HTTP URLs
        response = requests.get(img_url)
        suffix = utils.get_path_suffix(img_url)
@@ -406,7 +565,7 @@ class FeiShuChanel(ChatChannel):
        """
        try:
            import subprocess
-            
+
            # 使用 ffprobe 获取视频时长
            cmd = [
                'ffprobe',
@@ -415,7 +574,7 @@ class FeiShuChanel(ChatChannel):
                '-of', 'default=noprint_wrappers=1:nokey=1',
                file_path
            ]
-            
+
            result = subprocess.run(cmd, capture_output=True, text=True, timeout=10)
            if result.returncode == 0:
                duration_seconds = float(result.stdout.strip())
@@ -444,7 +603,7 @@ class FeiShuChanel(ChatChannel):
        """
        local_path = None
        temp_file = None
-        
+
        try:
            # For file:// URLs (local files), upload directly
            if video_url.startswith("file://"):
@@ -459,65 +618,67 @@ class FeiShuChanel(ChatChannel):
                if response.status_code != 200:
                    logger.error(f"[FeiShu] download video failed, status={response.status_code}")
                    return None
-                
+
                # Save to temp file
                import uuid
                file_name = os.path.basename(video_url) or "video.mp4"
                temp_file = str(uuid.uuid4()) + "_" + file_name
-                
+
                with open(temp_file, "wb") as file:
                    file.write(response.content)
-                
+
                logger.info(f"[FeiShu] Video downloaded, size={len(response.content)} bytes")
                local_path = temp_file
-            
+
            # Get video duration
            duration = self._get_video_duration(local_path)
-            
+
            # Upload to Feishu
            file_name = os.path.basename(local_path)
            file_ext = os.path.splitext(file_name)[1].lower()
            file_type_map = {'.mp4': 'mp4'}
            file_type = file_type_map.get(file_ext, 'mp4')
-            
+
            upload_url = "https://open.feishu.cn/open-apis/im/v1/files"
            data = {
-                'file_type': file_type, 
+                'file_type': file_type,
                'file_name': file_name
            }
            # Add duration only if available (required for video/audio)
            if duration:
                data['duration'] = duration  # Must be int, not string
-            
+
            headers = {'Authorization': f'Bearer {access_token}'}
-            
+
            logger.info(f"[FeiShu] Uploading video: file_name={file_name}, duration={duration}ms")
-            
+
            with open(local_path, "rb") as file:
                upload_response = requests.post(
-                    upload_url, 
-                    files={"file": file}, 
-                    data=data, 
-                    headers=headers, 
+                    upload_url,
+                    files={"file": file},
+                    data=data,
+                    headers=headers,
                    timeout=(5, 60)
                )
-                logger.info(f"[FeiShu] upload video response, status={upload_response.status_code}, res={upload_response.content}")
-                
+                logger.info(
+                    f"[FeiShu] upload video response, status={upload_response.status_code}, res={upload_response.content}")
+
                response_data = upload_response.json()
                if response_data.get("code") == 0:
                    # Add duration to the response data (API doesn't return it)
                    upload_data = response_data.get("data")
                    upload_data['duration'] = duration  # Add our calculated duration
-                    logger.info(f"[FeiShu] Upload complete: file_key={upload_data.get('file_key')}, duration={duration}ms")
+                    logger.info(
+                        f"[FeiShu] Upload complete: file_key={upload_data.get('file_key')}, duration={duration}ms")
                    return upload_data
                else:
                    logger.error(f"[FeiShu] upload video failed: {response_data}")
                    return None
-        
+
        except Exception as e:
            logger.error(f"[FeiShu] upload video exception: {e}")
            return None
-        
+
        finally:
            # Clean up temp file
            if temp_file and os.path.exists(temp_file):
@@ -532,20 +693,20 @@ class FeiShuChanel(ChatChannel):
        Supports both local files (file://) and HTTP URLs
        """
        logger.debug(f"[FeiShu] start process file, file_url={file_url}")
-        
+
        # Check if it's a local file path (file:// protocol)
        if file_url.startswith("file://"):
            local_path = file_url[7:]  # Remove "file://" prefix
            logger.info(f"[FeiShu] uploading local file: {local_path}")
-            
+
            if not os.path.exists(local_path):
                logger.error(f"[FeiShu] local file not found: {local_path}")
                return None
-            
+
            # Get file info
            file_name = os.path.basename(local_path)
            file_ext = os.path.splitext(file_name)[1].lower()
-            
+
            # Determine file type for Feishu API
            # Feishu supports: opus, mp4, pdf, doc, xls, ppt, stream (other types)
            file_type_map = {
@@ -557,23 +718,24 @@ class FeiShuChanel(ChatChannel):
                '.ppt': 'ppt', '.pptx': 'ppt',
            }
            file_type = file_type_map.get(file_ext, 'stream')  # Default to stream for other types
-            
+
            # Upload file to Feishu
            upload_url = "https://open.feishu.cn/open-apis/im/v1/files"
            data = {'file_type': file_type, 'file_name': file_name}
            headers = {'Authorization': f'Bearer {access_token}'}
-            
+
            try:
                with open(local_path, "rb") as file:
                    upload_response = requests.post(
-                        upload_url, 
-                        files={"file": file}, 
-                        data=data, 
+                        upload_url,
+                        files={"file": file},
+                        data=data,
                        headers=headers,
                        timeout=(5, 30)  # 5s connect, 30s read timeout
                    )
-                    logger.info(f"[FeiShu] upload file response, status={upload_response.status_code}, res={upload_response.content}")
-                    
+                    logger.info(
+                        f"[FeiShu] upload file response, status={upload_response.status_code}, res={upload_response.content}")
+
                    response_data = upload_response.json()
                    if response_data.get("code") == 0:
                        return response_data.get("data").get("file_key")
@@ -583,22 +745,22 @@ class FeiShuChanel(ChatChannel):
            except Exception as e:
                logger.error(f"[FeiShu] upload file exception: {e}")
                return None
-        
+
        # For HTTP URLs, download first then upload
        try:
            response = requests.get(file_url, timeout=(5, 30))
            if response.status_code != 200:
                logger.error(f"[FeiShu] download file failed, status={response.status_code}")
                return None
-            
+
            # Save to temp file
            import uuid
            file_name = os.path.basename(file_url)
            temp_name = str(uuid.uuid4()) + "_" + file_name
-            
+
            with open(temp_name, "wb") as file:
                file.write(response.content)
-            
+
            # Upload
            file_ext = os.path.splitext(file_name)[1].lower()
            file_type_map = {
@@ -608,18 +770,18 @@ class FeiShuChanel(ChatChannel):
                '.ppt': 'ppt', '.pptx': 'ppt',
            }
            file_type = file_type_map.get(file_ext, 'stream')
-            
+
            upload_url = "https://open.feishu.cn/open-apis/im/v1/files"
            data = {'file_type': file_type, 'file_name': file_name}
            headers = {'Authorization': f'Bearer {access_token}'}
-            
+
            with open(temp_name, "rb") as file:
                upload_response = requests.post(upload_url, files={"file": file}, data=data, headers=headers)
                logger.info(f"[FeiShu] upload file, res={upload_response.content}")
-                
+
                response_data = upload_response.json()
                os.remove(temp_name)  # Clean up temp file
-                
+
                if response_data.get("code") == 0:
                    return response_data.get("data").get("file_key")
                else:
@@ -632,11 +794,13 @@ class FeiShuChanel(ChatChannel):
    def _compose_context(self, ctype: ContextType, content, **kwargs):
        context = Context(ctype, content)
        context.kwargs = kwargs
+        if "channel_type" not in context:
+            context["channel_type"] = self.channel_type
        if "origin_ctype" not in context:
            context["origin_ctype"] = ctype

        cmsg = context["msg"]
-        
+
        # Set session_id based on chat type
        if cmsg.is_group:
            # Group chat: check if group_shared_session is enabled
@@ -652,7 +816,7 @@ class FeiShuChanel(ChatChannel):
        else:
            # Private chat: use user_id only
            context["session_id"] = cmsg.from_user_id
-        
+
        context["receiver"] = cmsg.other_user_id

        if ctype == ContextType.TEXT:
--- a/channel/feishu/feishu_message.py
+++ b/channel/feishu/feishu_message.py
@@ -6,6 +6,7 @@ import requests
 from common.log import logger
 from common.tmp_dir import TmpDir
 from common import utils
+from common.utils import expand_path
 from config import conf


@@ -31,7 +32,7 @@ class FeishuMessage(ChatMessage):
            image_key = content.get("image_key")
            
            # 下载图片到工作空间临时目录
-            workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+            workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
            tmp_dir = os.path.join(workspace_root, "tmp")
            os.makedirs(tmp_dir, exist_ok=True)
            image_path = os.path.join(tmp_dir, f"{image_key}.png")
@@ -97,7 +98,7 @@ class FeishuMessage(ChatMessage):
            
            if image_keys:
                # 如果包含图片，下载并在文本中引用本地路径
-                workspace_root = os.path.expanduser(conf().get("agent_workspace", "~/cow"))
+                workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
                tmp_dir = os.path.join(workspace_root, "tmp")
                os.makedirs(tmp_dir, exist_ok=True)
                
--- a/channel/qq/init.py
+++ b/channel/qq/init.py
--- a/channel/qq/qq_channel.py
+++ b/channel/qq/qq_channel.py
@@ -0,0 +1,735 @@
+"""
+QQ Bot channel via WebSocket long connection.
+
+Supports:
+- Group chat (@bot), single chat (C2C), guild channel, guild DM
+- Text / image / file message send & receive
+- Heartbeat keep-alive and auto-reconnect with session resume
+"""
+
+import base64
+import json
+import os
+import threading
+import time
+
+import requests
+import websocket
+
+from bridge.context import Context, ContextType
+from bridge.reply import Reply, ReplyType
+from channel.chat_channel import ChatChannel, check_prefix
+from channel.qq.qq_message import QQMessage
+from common.expired_dict import ExpiredDict
+from common.log import logger
+from common.singleton import singleton
+from config import conf
+
+# Rich media file_type constants
+QQ_FILE_TYPE_IMAGE = 1
+QQ_FILE_TYPE_VIDEO = 2
+QQ_FILE_TYPE_VOICE = 3
+QQ_FILE_TYPE_FILE = 4
+
+QQ_API_BASE = "https://api.sgroup.qq.com"
+
+# Intents: GROUP_AND_C2C_EVENT(1<<25) | PUBLIC_GUILD_MESSAGES(1<<30)
+DEFAULT_INTENTS = (1 << 25) | (1 << 30)
+
+# OpCode constants
+OP_DISPATCH = 0
+OP_HEARTBEAT = 1
+OP_IDENTIFY = 2
+OP_RESUME = 6
+OP_RECONNECT = 7
+OP_INVALID_SESSION = 9
+OP_HELLO = 10
+OP_HEARTBEAT_ACK = 11
+
+# Resumable error codes
+RESUMABLE_CLOSE_CODES = {4008, 4009}
+
+
+@singleton
+class QQChannel(ChatChannel):
+
+    def __init__(self):
+        super().__init__()
+        self.app_id = ""
+        self.app_secret = ""
+
+        self._access_token = ""
+        self._token_expires_at = 0
+
+        self._ws = None
+        self._ws_thread = None
+        self._heartbeat_thread = None
+        self._connected = False
+        self._stop_event = threading.Event()
+        self._token_lock = threading.Lock()
+
+        self._session_id = None
+        self._last_seq = None
+        self._heartbeat_interval = 45000
+        self._can_resume = False
+
+        self.received_msgs = ExpiredDict(60 * 60 * 7.1)
+        self._msg_seq_counter = {}
+
+        conf()["group_name_white_list"] = ["ALL_GROUP"]
+        conf()["single_chat_prefix"] = [""]
+
+    # ------------------------------------------------------------------
+    # Lifecycle
+    # ------------------------------------------------------------------
+
+    def startup(self):
+        self.app_id = conf().get("qq_app_id", "")
+        self.app_secret = conf().get("qq_app_secret", "")
+
+        if not self.app_id or not self.app_secret:
+            err = "[QQ] qq_app_id and qq_app_secret are required"
+            logger.error(err)
+            self.report_startup_error(err)
+            return
+
+        self._refresh_access_token()
+        if not self._access_token:
+            err = "[QQ] Failed to get initial access_token"
+            logger.error(err)
+            self.report_startup_error(err)
+            return
+
+        self._stop_event.clear()
+        self._start_ws()
+
+    def stop(self):
+        logger.info("[QQ] stop() called")
+        self._stop_event.set()
+        if self._ws:
+            try:
+                self._ws.close()
+            except Exception:
+                pass
+        self._ws = None
+        self._connected = False
+
+    # ------------------------------------------------------------------
+    # Access Token
+    # ------------------------------------------------------------------
+
+    def _refresh_access_token(self):
+        try:
+            resp = requests.post(
+                "https://bots.qq.com/app/getAppAccessToken",
+                json={"appId": self.app_id, "clientSecret": self.app_secret},
+                timeout=10,
+            )
+            resp.raise_for_status()
+            data = resp.json()
+            self._access_token = data.get("access_token", "")
+            expires_in = int(data.get("expires_in", 7200))
+            self._token_expires_at = time.time() + expires_in - 60
+            logger.debug(f"[QQ] Access token refreshed, expires_in={expires_in}s")
+        except Exception as e:
+            logger.error(f"[QQ] Failed to refresh access_token: {e}")
+
+    def _get_access_token(self) -> str:
+        with self._token_lock:
+            if time.time() >= self._token_expires_at:
+                self._refresh_access_token()
+            return self._access_token
+
+    def _get_auth_headers(self) -> dict:
+        return {
+            "Authorization": f"QQBot {self._get_access_token()}",
+            "Content-Type": "application/json",
+        }
+
+    # ------------------------------------------------------------------
+    # WebSocket connection
+    # ------------------------------------------------------------------
+
+    def _get_ws_url(self) -> str:
+        try:
+            resp = requests.get(
+                f"{QQ_API_BASE}/gateway",
+                headers=self._get_auth_headers(),
+                timeout=10,
+            )
+            resp.raise_for_status()
+            url = resp.json().get("url", "")
+            logger.debug(f"[QQ] Gateway URL: {url}")
+            return url
+        except Exception as e:
+            logger.error(f"[QQ] Failed to get gateway URL: {e}")
+            return ""
+
+    def _start_ws(self):
+        ws_url = self._get_ws_url()
+        if not ws_url:
+            logger.error("[QQ] Cannot start WebSocket without gateway URL")
+            self.report_startup_error("Failed to get gateway URL")
+            return
+
+        def _on_open(ws):
+            logger.debug("[QQ] WebSocket connected, waiting for Hello...")
+
+        def _on_message(ws, raw):
+            try:
+                data = json.loads(raw)
+                self._handle_ws_message(data)
+            except Exception as e:
+                logger.error(f"[QQ] Failed to handle ws message: {e}", exc_info=True)
+
+        def _on_error(ws, error):
+            logger.error(f"[QQ] WebSocket error: {error}")
+
+        def _on_close(ws, close_status_code, close_msg):
+            logger.warning(f"[QQ] WebSocket closed: status={close_status_code}, msg={close_msg}")
+            self._connected = False
+            if not self._stop_event.is_set():
+                if close_status_code in RESUMABLE_CLOSE_CODES and self._session_id:
+                    self._can_resume = True
+                    logger.info("[QQ] Will attempt resume in 3s...")
+                    time.sleep(3)
+                else:
+                    self._can_resume = False
+                    logger.info("[QQ] Will reconnect in 5s...")
+                    time.sleep(5)
+                if not self._stop_event.is_set():
+                    self._start_ws()
+
+        self._ws = websocket.WebSocketApp(
+            ws_url,
+            on_open=_on_open,
+            on_message=_on_message,
+            on_error=_on_error,
+            on_close=_on_close,
+        )
+
+        def run_forever():
+            try:
+                self._ws.run_forever(ping_interval=0, reconnect=0)
+            except (SystemExit, KeyboardInterrupt):
+                logger.info("[QQ] WebSocket thread interrupted")
+            except Exception as e:
+                logger.error(f"[QQ] WebSocket run_forever error: {e}")
+
+        self._ws_thread = threading.Thread(target=run_forever, daemon=True)
+        self._ws_thread.start()
+        self._ws_thread.join()
+
+    def _ws_send(self, data: dict):
+        if self._ws:
+            self._ws.send(json.dumps(data, ensure_ascii=False))
+
+    # ------------------------------------------------------------------
+    # Identify & Resume & Heartbeat
+    # ------------------------------------------------------------------
+
+    def _send_identify(self):
+        self._ws_send({
+            "op": OP_IDENTIFY,
+            "d": {
+                "token": f"QQBot {self._get_access_token()}",
+                "intents": DEFAULT_INTENTS,
+                "shard": [0, 1],
+                "properties": {
+                    "$os": "linux",
+                    "$browser": "chatgpt-on-wechat",
+                    "$device": "chatgpt-on-wechat",
+                },
+            },
+        })
+        logger.debug(f"[QQ] Identify sent with intents={DEFAULT_INTENTS}")
+
+    def _send_resume(self):
+        self._ws_send({
+            "op": OP_RESUME,
+            "d": {
+                "token": f"QQBot {self._get_access_token()}",
+                "session_id": self._session_id,
+                "seq": self._last_seq,
+            },
+        })
+        logger.debug(f"[QQ] Resume sent: session_id={self._session_id}, seq={self._last_seq}")
+
+    def _start_heartbeat(self, interval_ms: int):
+        if self._heartbeat_thread and self._heartbeat_thread.is_alive():
+            return
+        self._heartbeat_interval = interval_ms
+        interval_sec = interval_ms / 1000.0
+
+        def heartbeat_loop():
+            while not self._stop_event.is_set() and self._connected:
+                try:
+                    self._ws_send({
+                        "op": OP_HEARTBEAT,
+                        "d": self._last_seq,
+                    })
+                except Exception as e:
+                    logger.warning(f"[QQ] Heartbeat send failed: {e}")
+                    break
+                self._stop_event.wait(interval_sec)
+
+        self._heartbeat_thread = threading.Thread(target=heartbeat_loop, daemon=True)
+        self._heartbeat_thread.start()
+
+    # ------------------------------------------------------------------
+    # Incoming message dispatch
+    # ------------------------------------------------------------------
+
+    def _handle_ws_message(self, data: dict):
+        op = data.get("op")
+        d = data.get("d")
+        t = data.get("t")
+        s = data.get("s")
+
+        if s is not None:
+            self._last_seq = s
+
+        if op == OP_HELLO:
+            heartbeat_interval = d.get("heartbeat_interval", 45000) if d else 45000
+            logger.debug(f"[QQ] Received Hello, heartbeat_interval={heartbeat_interval}ms")
+            self._heartbeat_interval = heartbeat_interval
+            if self._can_resume and self._session_id:
+                self._send_resume()
+            else:
+                self._send_identify()
+
+        elif op == OP_HEARTBEAT_ACK:
+            pass
+
+        elif op == OP_HEARTBEAT:
+            self._ws_send({"op": OP_HEARTBEAT, "d": self._last_seq})
+
+        elif op == OP_RECONNECT:
+            logger.warning("[QQ] Server requested reconnect")
+            self._can_resume = True
+            if self._ws:
+                self._ws.close()
+
+        elif op == OP_INVALID_SESSION:
+            logger.warning("[QQ] Invalid session, re-identifying...")
+            self._session_id = None
+            self._can_resume = False
+            time.sleep(2)
+            self._send_identify()
+
+        elif op == OP_DISPATCH:
+            if t == "READY":
+                self._session_id = d.get("session_id", "")
+                user = d.get("user", {})
+                bot_name = user.get('username', '')
+                logger.info(f"[QQ] ✅ Connected successfully (bot={bot_name})")
+                self._connected = True
+                self._can_resume = False
+                self._start_heartbeat(self._heartbeat_interval)
+                self.report_startup_success()
+
+            elif t == "RESUMED":
+                logger.info("[QQ] Session resumed successfully")
+                self._connected = True
+                self._can_resume = False
+                self._start_heartbeat(self._heartbeat_interval)
+
+            elif t in ("GROUP_AT_MESSAGE_CREATE", "C2C_MESSAGE_CREATE",
+                        "AT_MESSAGE_CREATE", "DIRECT_MESSAGE_CREATE"):
+                self._handle_msg_event(d, t)
+
+            elif t in ("GROUP_ADD_ROBOT", "FRIEND_ADD"):
+                logger.info(f"[QQ] Event: {t}")
+
+            else:
+                logger.debug(f"[QQ] Dispatch event: {t}")
+
+    # ------------------------------------------------------------------
+    # Message event handling
+    # ------------------------------------------------------------------
+
+    def _handle_msg_event(self, event_data: dict, event_type: str):
+        msg_id = event_data.get("id", "")
+        if self.received_msgs.get(msg_id):
+            logger.debug(f"[QQ] Duplicate msg filtered: {msg_id}")
+            return
+        self.received_msgs[msg_id] = True
+
+        try:
+            qq_msg = QQMessage(event_data, event_type)
+        except NotImplementedError as e:
+            logger.warning(f"[QQ] {e}")
+            return
+        except Exception as e:
+            logger.error(f"[QQ] Failed to parse message: {e}", exc_info=True)
+            return
+
+        is_group = qq_msg.is_group
+
+        from channel.file_cache import get_file_cache
+        file_cache = get_file_cache()
+
+        if is_group:
+            session_id = qq_msg.other_user_id
+        else:
+            session_id = qq_msg.from_user_id
+
+        if qq_msg.ctype == ContextType.IMAGE:
+            if hasattr(qq_msg, "image_path") and qq_msg.image_path:
+                file_cache.add(session_id, qq_msg.image_path, file_type="image")
+                logger.info(f"[QQ] Image cached for session {session_id}")
+            return
+
+        if qq_msg.ctype == ContextType.TEXT:
+            cached_files = file_cache.get(session_id)
+            if cached_files:
+                file_refs = []
+                for fi in cached_files:
+                    ftype = fi["type"]
+                    fpath = fi["path"]
+                    if ftype == "image":
+                        file_refs.append(f"[图片: {fpath}]")
+                    elif ftype == "video":
+                        file_refs.append(f"[视频: {fpath}]")
+                    else:
+                        file_refs.append(f"[文件: {fpath}]")
+                qq_msg.content = qq_msg.content + "\n" + "\n".join(file_refs)
+                logger.info(f"[QQ] Attached {len(cached_files)} cached file(s)")
+                file_cache.clear(session_id)
+
+        context = self._compose_context(
+            qq_msg.ctype,
+            qq_msg.content,
+            isgroup=is_group,
+            msg=qq_msg,
+            no_need_at=True,
+        )
+        if context:
+            self.produce(context)
+
+    # ------------------------------------------------------------------
+    # _compose_context
+    # ------------------------------------------------------------------
+
+    def _compose_context(self, ctype: ContextType, content, **kwargs):
+        context = Context(ctype, content)
+        context.kwargs = kwargs
+        if "channel_type" not in context:
+            context["channel_type"] = self.channel_type
+        if "origin_ctype" not in context:
+            context["origin_ctype"] = ctype
+
+        cmsg = context["msg"]
+
+        if cmsg.is_group:
+            context["session_id"] = cmsg.other_user_id
+        else:
+            context["session_id"] = cmsg.from_user_id
+
+        context["receiver"] = cmsg.other_user_id
+
+        if ctype == ContextType.TEXT:
+            img_match_prefix = check_prefix(content, conf().get("image_create_prefix"))
+            if img_match_prefix:
+                content = content.replace(img_match_prefix, "", 1)
+                context.type = ContextType.IMAGE_CREATE
+            else:
+                context.type = ContextType.TEXT
+            context.content = content.strip()
+
+        return context
+
+    # ------------------------------------------------------------------
+    # Send reply
+    # ------------------------------------------------------------------
+
+    def send(self, reply: Reply, context: Context):
+        msg = context.get("msg")
+        is_group = context.get("isgroup", False)
+        receiver = context.get("receiver", "")
+
+        if not msg:
+            # Active send (e.g. scheduled tasks), no original message to reply to
+            self._active_send_text(reply.content if reply.type == ReplyType.TEXT else str(reply.content),
+                                   receiver, is_group)
+            return
+
+        event_type = getattr(msg, "event_type", "")
+        msg_id = getattr(msg, "msg_id", "")
+
+        if reply.type == ReplyType.TEXT:
+            self._send_text(reply.content, msg, event_type, msg_id)
+        elif reply.type in (ReplyType.IMAGE_URL, ReplyType.IMAGE):
+            self._send_image(reply.content, msg, event_type, msg_id)
+        elif reply.type == ReplyType.FILE:
+            if hasattr(reply, "text_content") and reply.text_content:
+                self._send_text(reply.text_content, msg, event_type, msg_id)
+                time.sleep(0.3)
+            self._send_file(reply.content, msg, event_type, msg_id)
+        elif reply.type in (ReplyType.VIDEO, ReplyType.VIDEO_URL):
+            self._send_media(reply.content, msg, event_type, msg_id, QQ_FILE_TYPE_VIDEO)
+        else:
+            logger.warning(f"[QQ] Unsupported reply type: {reply.type}, falling back to text")
+            self._send_text(str(reply.content), msg, event_type, msg_id)
+
+    # ------------------------------------------------------------------
+    # Send helpers
+    # ------------------------------------------------------------------
+
+    def _get_next_msg_seq(self, msg_id: str) -> int:
+        seq = self._msg_seq_counter.get(msg_id, 1)
+        self._msg_seq_counter[msg_id] = seq + 1
+        return seq
+
+    def _build_msg_url_and_base_body(self, msg: QQMessage, event_type: str, msg_id: str):
+        """Build the API URL and base body dict for sending a message."""
+        if event_type == "GROUP_AT_MESSAGE_CREATE":
+            group_openid = msg._rawmsg.get("group_openid", "")
+            url = f"{QQ_API_BASE}/v2/groups/{group_openid}/messages"
+            body = {
+                "msg_id": msg_id,
+                "msg_seq": self._get_next_msg_seq(msg_id),
+            }
+            return url, body, "group", group_openid
+
+        elif event_type == "C2C_MESSAGE_CREATE":
+            user_openid = msg._rawmsg.get("author", {}).get("user_openid", "") or msg.from_user_id
+            url = f"{QQ_API_BASE}/v2/users/{user_openid}/messages"
+            body = {
+                "msg_id": msg_id,
+                "msg_seq": self._get_next_msg_seq(msg_id),
+            }
+            return url, body, "c2c", user_openid
+
+        elif event_type == "AT_MESSAGE_CREATE":
+            channel_id = msg._rawmsg.get("channel_id", "")
+            url = f"{QQ_API_BASE}/channels/{channel_id}/messages"
+            body = {"msg_id": msg_id}
+            return url, body, "channel", channel_id
+
+        elif event_type == "DIRECT_MESSAGE_CREATE":
+            guild_id = msg._rawmsg.get("guild_id", "")
+            url = f"{QQ_API_BASE}/dms/{guild_id}/messages"
+            body = {"msg_id": msg_id}
+            return url, body, "dm", guild_id
+
+        return None, None, None, None
+
+    def _post_message(self, url: str, body: dict, event_type: str):
+        try:
+            resp = requests.post(url, json=body, headers=self._get_auth_headers(), timeout=10)
+            if resp.status_code in (200, 201, 202, 204):
+                logger.info(f"[QQ] Message sent successfully: event_type={event_type}")
+            else:
+                logger.error(f"[QQ] Failed to send message: status={resp.status_code}, "
+                             f"body={resp.text}")
+        except Exception as e:
+            logger.error(f"[QQ] Send message error: {e}")
+
+    # ------------------------------------------------------------------
+    # Active send (no original message, e.g. scheduled tasks)
+    # ------------------------------------------------------------------
+
+    def _active_send_text(self, content: str, receiver: str, is_group: bool):
+        """Send text without an original message (active push). QQ limits active messages to 4/month per user."""
+        if not receiver:
+            logger.warning("[QQ] No receiver for active send")
+            return
+        if is_group:
+            url = f"{QQ_API_BASE}/v2/groups/{receiver}/messages"
+        else:
+            url = f"{QQ_API_BASE}/v2/users/{receiver}/messages"
+        body = {
+            "content": content,
+            "msg_type": 0,
+        }
+        event_label = "GROUP_ACTIVE" if is_group else "C2C_ACTIVE"
+        self._post_message(url, body, event_label)
+
+    # ------------------------------------------------------------------
+    # Send text
+    # ------------------------------------------------------------------
+
+    def _send_text(self, content: str, msg: QQMessage, event_type: str, msg_id: str):
+        url, body, _, _ = self._build_msg_url_and_base_body(msg, event_type, msg_id)
+        if not url:
+            logger.warning(f"[QQ] Cannot send reply for event_type: {event_type}")
+            return
+        body["content"] = content
+        body["msg_type"] = 0
+        self._post_message(url, body, event_type)
+
+    # ------------------------------------------------------------------
+    # Rich media upload & send (image / video / file)
+    # ------------------------------------------------------------------
+
+    def _upload_rich_media(self, file_url: str, file_type: int, msg: QQMessage,
+                           event_type: str) -> str:
+        """
+        Upload media via QQ rich media API and return file_info.
+        For group: POST /v2/groups/{group_openid}/files
+        For c2c:   POST /v2/users/{openid}/files
+        """
+        if event_type == "GROUP_AT_MESSAGE_CREATE":
+            group_openid = msg._rawmsg.get("group_openid", "")
+            upload_url = f"{QQ_API_BASE}/v2/groups/{group_openid}/files"
+        elif event_type == "C2C_MESSAGE_CREATE":
+            user_openid = (msg._rawmsg.get("author", {}).get("user_openid", "")
+                           or msg.from_user_id)
+            upload_url = f"{QQ_API_BASE}/v2/users/{user_openid}/files"
+        else:
+            logger.warning(f"[QQ] Rich media upload not supported for event_type: {event_type}")
+            return ""
+
+        upload_body = {
+            "file_type": file_type,
+            "url": file_url,
+            "srv_send_msg": False,
+        }
+
+        try:
+            resp = requests.post(
+                upload_url, json=upload_body,
+                headers=self._get_auth_headers(), timeout=30,
+            )
+            if resp.status_code in (200, 201):
+                data = resp.json()
+                file_info = data.get("file_info", "")
+                logger.info(f"[QQ] Rich media uploaded: file_type={file_type}, "
+                            f"file_uuid={data.get('file_uuid', '')}")
+                return file_info
+            else:
+                logger.error(f"[QQ] Rich media upload failed: status={resp.status_code}, "
+                             f"body={resp.text}")
+                return ""
+        except Exception as e:
+            logger.error(f"[QQ] Rich media upload error: {e}")
+            return ""
+
+    def _upload_rich_media_base64(self, file_path: str, file_type: int, msg: QQMessage,
+                                  event_type: str) -> str:
+        """Upload local file via base64 file_data field."""
+        if event_type == "GROUP_AT_MESSAGE_CREATE":
+            group_openid = msg._rawmsg.get("group_openid", "")
+            upload_url = f"{QQ_API_BASE}/v2/groups/{group_openid}/files"
+        elif event_type == "C2C_MESSAGE_CREATE":
+            user_openid = (msg._rawmsg.get("author", {}).get("user_openid", "")
+                           or msg.from_user_id)
+            upload_url = f"{QQ_API_BASE}/v2/users/{user_openid}/files"
+        else:
+            logger.warning(f"[QQ] Rich media upload not supported for event_type: {event_type}")
+            return ""
+
+        try:
+            with open(file_path, "rb") as f:
+                file_data = base64.b64encode(f.read()).decode("utf-8")
+        except Exception as e:
+            logger.error(f"[QQ] Failed to read file for upload: {e}")
+            return ""
+
+        upload_body = {
+            "file_type": file_type,
+            "file_data": file_data,
+            "srv_send_msg": False,
+        }
+
+        try:
+            resp = requests.post(
+                upload_url, json=upload_body,
+                headers=self._get_auth_headers(), timeout=30,
+            )
+            if resp.status_code in (200, 201):
+                data = resp.json()
+                file_info = data.get("file_info", "")
+                logger.info(f"[QQ] Rich media uploaded (base64): file_type={file_type}, "
+                            f"file_uuid={data.get('file_uuid', '')}")
+                return file_info
+            else:
+                logger.error(f"[QQ] Rich media upload (base64) failed: status={resp.status_code}, "
+                             f"body={resp.text}")
+                return ""
+        except Exception as e:
+            logger.error(f"[QQ] Rich media upload (base64) error: {e}")
+            return ""
+
+    def _send_media_msg(self, file_info: str, msg: QQMessage, event_type: str, msg_id: str):
+        """Send a message with msg_type=7 (rich media) using file_info."""
+        url, body, _, _ = self._build_msg_url_and_base_body(msg, event_type, msg_id)
+        if not url:
+            return
+        body["msg_type"] = 7
+        body["media"] = {"file_info": file_info}
+        self._post_message(url, body, event_type)
+
+    def _send_image(self, img_path_or_url: str, msg: QQMessage, event_type: str, msg_id: str):
+        """Send image reply. Supports URL and local file path."""
+        if event_type not in ("GROUP_AT_MESSAGE_CREATE", "C2C_MESSAGE_CREATE"):
+            self._send_text(str(img_path_or_url), msg, event_type, msg_id)
+            return
+
+        if img_path_or_url.startswith("file://"):
+            img_path_or_url = img_path_or_url[7:]
+
+        if img_path_or_url.startswith(("http://", "https://")):
+            file_info = self._upload_rich_media(
+                img_path_or_url, QQ_FILE_TYPE_IMAGE, msg, event_type)
+        elif os.path.exists(img_path_or_url):
+            file_info = self._upload_rich_media_base64(
+                img_path_or_url, QQ_FILE_TYPE_IMAGE, msg, event_type)
+        else:
+            logger.error(f"[QQ] Image not found: {img_path_or_url}")
+            self._send_text("[Image send failed]", msg, event_type, msg_id)
+            return
+
+        if file_info:
+            self._send_media_msg(file_info, msg, event_type, msg_id)
+        else:
+            self._send_text("[Image upload failed]", msg, event_type, msg_id)
+
+    def _send_file(self, file_path_or_url: str, msg: QQMessage, event_type: str, msg_id: str):
+        """Send file reply."""
+        if event_type not in ("GROUP_AT_MESSAGE_CREATE", "C2C_MESSAGE_CREATE"):
+            self._send_text(str(file_path_or_url), msg, event_type, msg_id)
+            return
+
+        if file_path_or_url.startswith("file://"):
+            file_path_or_url = file_path_or_url[7:]
+
+        if file_path_or_url.startswith(("http://", "https://")):
+            file_info = self._upload_rich_media(
+                file_path_or_url, QQ_FILE_TYPE_FILE, msg, event_type)
+        elif os.path.exists(file_path_or_url):
+            file_info = self._upload_rich_media_base64(
+                file_path_or_url, QQ_FILE_TYPE_FILE, msg, event_type)
+        else:
+            logger.error(f"[QQ] File not found: {file_path_or_url}")
+            self._send_text("[File send failed]", msg, event_type, msg_id)
+            return
+
+        if file_info:
+            self._send_media_msg(file_info, msg, event_type, msg_id)
+        else:
+            self._send_text("[File upload failed]", msg, event_type, msg_id)
+
+    def _send_media(self, path_or_url: str, msg: QQMessage, event_type: str,
+                    msg_id: str, file_type: int):
+        """Generic media send for video/voice etc."""
+        if event_type not in ("GROUP_AT_MESSAGE_CREATE", "C2C_MESSAGE_CREATE"):
+            self._send_text(str(path_or_url), msg, event_type, msg_id)
+            return
+
+        if path_or_url.startswith("file://"):
+            path_or_url = path_or_url[7:]
+
+        if path_or_url.startswith(("http://", "https://")):
+            file_info = self._upload_rich_media(path_or_url, file_type, msg, event_type)
+        elif os.path.exists(path_or_url):
+            file_info = self._upload_rich_media_base64(path_or_url, file_type, msg, event_type)
+        else:
+            logger.error(f"[QQ] Media not found: {path_or_url}")
+            return
+
+        if file_info:
+            self._send_media_msg(file_info, msg, event_type, msg_id)
+        else:
+            logger.error(f"[QQ] Media upload failed: {path_or_url}")
--- a/channel/qq/qq_message.py
+++ b/channel/qq/qq_message.py
@@ -0,0 +1,123 @@
+import os
+import requests
+
+from bridge.context import ContextType
+from channel.chat_message import ChatMessage
+from common.log import logger
+from common.utils import expand_path
+from config import conf
+
+
+def _get_tmp_dir() -> str:
+    """Return the workspace tmp directory (absolute path), creating it if needed."""
+    ws_root = expand_path(conf().get("agent_workspace", "~/cow"))
+    tmp_dir = os.path.join(ws_root, "tmp")
+    os.makedirs(tmp_dir, exist_ok=True)
+    return tmp_dir
+
+
+class QQMessage(ChatMessage):
+    """Message wrapper for QQ Bot (websocket long-connection mode)."""
+
+    def __init__(self, event_data: dict, event_type: str):
+        super().__init__(event_data)
+        self.msg_id = event_data.get("id", "")
+        self.create_time = event_data.get("timestamp", "")
+        self.is_group = event_type in ("GROUP_AT_MESSAGE_CREATE",)
+        self.event_type = event_type
+
+        author = event_data.get("author", {})
+        from_user_id = author.get("member_openid", "") or author.get("id", "")
+        group_openid = event_data.get("group_openid", "")
+
+        content = event_data.get("content", "").strip()
+
+        attachments = event_data.get("attachments", [])
+        has_image = any(
+            a.get("content_type", "").startswith("image/") for a in attachments
+        ) if attachments else False
+
+        if has_image and not content:
+            self.ctype = ContextType.IMAGE
+            img_attachment = next(
+                a for a in attachments if a.get("content_type", "").startswith("image/")
+            )
+            img_url = img_attachment.get("url", "")
+            if img_url and not img_url.startswith("http"):
+                img_url = "https://" + img_url
+            tmp_dir = _get_tmp_dir()
+            image_path = os.path.join(tmp_dir, f"qq_{self.msg_id}.png")
+            try:
+                resp = requests.get(img_url, timeout=30)
+                resp.raise_for_status()
+                with open(image_path, "wb") as f:
+                    f.write(resp.content)
+                self.content = image_path
+                self.image_path = image_path
+                logger.info(f"[QQ] Image downloaded: {image_path}")
+            except Exception as e:
+                logger.error(f"[QQ] Failed to download image: {e}")
+                self.content = "[Image download failed]"
+                self.image_path = None
+        elif has_image and content:
+            self.ctype = ContextType.TEXT
+            image_paths = []
+            tmp_dir = _get_tmp_dir()
+            for idx, att in enumerate(attachments):
+                if not att.get("content_type", "").startswith("image/"):
+                    continue
+                img_url = att.get("url", "")
+                if img_url and not img_url.startswith("http"):
+                    img_url = "https://" + img_url
+                img_path = os.path.join(tmp_dir, f"qq_{self.msg_id}_{idx}.png")
+                try:
+                    resp = requests.get(img_url, timeout=30)
+                    resp.raise_for_status()
+                    with open(img_path, "wb") as f:
+                        f.write(resp.content)
+                    image_paths.append(img_path)
+                except Exception as e:
+                    logger.error(f"[QQ] Failed to download mixed image: {e}")
+            content_parts = [content]
+            for p in image_paths:
+                content_parts.append(f"[图片: {p}]")
+            self.content = "\n".join(content_parts)
+        else:
+            self.ctype = ContextType.TEXT
+            self.content = content
+
+        if event_type == "GROUP_AT_MESSAGE_CREATE":
+            self.from_user_id = from_user_id
+            self.to_user_id = ""
+            self.other_user_id = group_openid
+            self.actual_user_id = from_user_id
+            self.actual_user_nickname = from_user_id
+
+        elif event_type == "C2C_MESSAGE_CREATE":
+            user_openid = author.get("user_openid", "") or from_user_id
+            self.from_user_id = user_openid
+            self.to_user_id = ""
+            self.other_user_id = user_openid
+            self.actual_user_id = user_openid
+
+        elif event_type == "AT_MESSAGE_CREATE":
+            self.from_user_id = from_user_id
+            self.to_user_id = ""
+            channel_id = event_data.get("channel_id", "")
+            self.other_user_id = channel_id
+            self.actual_user_id = from_user_id
+            self.actual_user_nickname = author.get("username", from_user_id)
+
+        elif event_type == "DIRECT_MESSAGE_CREATE":
+            self.from_user_id = from_user_id
+            self.to_user_id = ""
+            guild_id = event_data.get("guild_id", "")
+            self.other_user_id = f"dm_{guild_id}_{from_user_id}"
+            self.actual_user_id = from_user_id
+            self.actual_user_nickname = author.get("username", from_user_id)
+
+        else:
+            raise NotImplementedError(f"Unsupported QQ event type: {event_type}")
+
+        logger.debug(f"[QQ] Message parsed: type={event_type}, ctype={self.ctype}, "
+                     f"from={self.from_user_id}, content_len={len(self.content)}")
--- a/channel/web/chat.html
+++ b/channel/web/chat.html
--- a/channel/web/static/css/console.css
+++ b/channel/web/static/css/console.css
@@ -0,0 +1,448 @@
+/* =====================================================================
+   CowAgent Console Styles
+   ===================================================================== */
+
+/* Animations */
+@keyframes pulseDot {
+    0%, 80%, 100% { transform: scale(0.6); opacity: 0.4; }
+    40% { transform: scale(1); opacity: 1; }
+}
+
+/* Scrollbar */
+* { scrollbar-width: thin; scrollbar-color: #94a3b8 transparent; }
+::-webkit-scrollbar { width: 6px; height: 6px; }
+::-webkit-scrollbar-track { background: transparent; }
+::-webkit-scrollbar-thumb { background: #94a3b8; border-radius: 3px; }
+::-webkit-scrollbar-thumb:hover { background: #64748b; }
+.dark ::-webkit-scrollbar-thumb { background: #475569; }
+.dark ::-webkit-scrollbar-thumb:hover { background: #64748b; }
+
+/* Sidebar */
+.sidebar-item.active {
+    background: rgba(255, 255, 255, 0.08);
+    color: #FFFFFF;
+}
+.sidebar-item.active .item-icon { color: #4ABE6E; }
+
+/* Menu Groups */
+.menu-group-items { max-height: 0; overflow: hidden; transition: max-height 0.25s ease-out; }
+.menu-group.open .menu-group-items { max-height: 500px; transition: max-height 0.35s ease-in; }
+.menu-group .chevron { transition: transform 0.25s ease; }
+.menu-group.open .chevron { transform: rotate(90deg); }
+
+/* View Switching */
+.view { display: none; height: 100%; }
+.view.active { display: flex; flex-direction: column; }
+
+/* Markdown Content */
+.msg-content p { margin: 0.5em 0; line-height: 1.7; }
+.msg-content p:first-child { margin-top: 0; }
+.msg-content p:last-child { margin-bottom: 0; }
+.msg-content h1, .msg-content h2, .msg-content h3,
+.msg-content h4, .msg-content h5, .msg-content h6 {
+    margin-top: 1.2em; margin-bottom: 0.6em; font-weight: 600; line-height: 1.3;
+}
+.msg-content h1 { font-size: 1.4em; }
+.msg-content h2 { font-size: 1.25em; }
+.msg-content h3 { font-size: 1.1em; }
+.msg-content ul, .msg-content ol { margin: 0.5em 0; padding-left: 1.8em; }
+.msg-content li { margin: 0.25em 0; }
+.msg-content pre {
+    border-radius: 8px; overflow-x: auto; margin: 0.8em 0;
+    background: #f1f5f9; padding: 1em;
+}
+.dark .msg-content pre { background: #111111; }
+.msg-content code {
+    font-family: 'JetBrains Mono', 'Fira Code', Consolas, monospace;
+    font-size: 0.875em;
+}
+.msg-content :not(pre) > code {
+    background: rgba(74, 190, 110, 0.1); color: #1C6B3B;
+    padding: 2px 6px; border-radius: 4px;
+}
+.dark .msg-content :not(pre) > code {
+    background: rgba(74, 190, 110, 0.15); color: #74E9A4;
+}
+.msg-content pre code { background: transparent; padding: 0; color: inherit; }
+.msg-content blockquote {
+    border-left: 3px solid #4ABE6E; padding: 0.5em 1em;
+    margin: 0.8em 0; background: rgba(74, 190, 110, 0.05); border-radius: 0 6px 6px 0;
+}
+.dark .msg-content blockquote { background: rgba(74, 190, 110, 0.08); }
+.msg-content table { border-collapse: collapse; width: 100%; margin: 0.8em 0; }
+.msg-content th, .msg-content td {
+    border: 1px solid #e2e8f0; padding: 8px 12px; text-align: left;
+}
+.dark .msg-content th, .dark .msg-content td { border-color: rgba(255,255,255,0.1); }
+.msg-content th { background: #f1f5f9; font-weight: 600; }
+.dark .msg-content th { background: #111111; }
+.msg-content img { max-width: 100%; height: auto; border-radius: 8px; margin: 0.5em 0; }
+.msg-content a { color: #35A85B; text-decoration: underline; }
+.msg-content a:hover { color: #228547; }
+.msg-content hr { border: none; height: 1px; background: #e2e8f0; margin: 1.2em 0; }
+.dark .msg-content hr { background: rgba(255,255,255,0.1); }
+
+/* SSE Streaming cursor */
+@keyframes blink { 0%, 100% { opacity: 1; } 50% { opacity: 0; } }
+.sse-streaming::after {
+    content: '▋';
+    display: inline-block;
+    margin-left: 2px;
+    color: #4ABE6E;
+    animation: blink 0.9s step-end infinite;
+    font-size: 0.85em;
+    vertical-align: middle;
+}
+
+/* Agent steps (thinking summaries + tool indicators) */
+.agent-steps:empty { display: none; }
+.agent-steps:not(:empty) {
+    margin-bottom: 0.625rem;
+    padding-bottom: 0.5rem;
+    border-bottom: 1px dashed rgba(0, 0, 0, 0.08);
+}
+.dark .agent-steps:not(:empty) { border-bottom-color: rgba(255, 255, 255, 0.08); }
+
+.agent-step {
+    font-size: 0.75rem;
+    line-height: 1.4;
+    color: #94a3b8;
+    margin-bottom: 0.25rem;
+}
+.agent-step:last-child { margin-bottom: 0; }
+
+/* Thinking step - collapsible */
+.agent-thinking-step .thinking-header {
+    display: flex;
+    align-items: center;
+    gap: 0.375rem;
+    cursor: pointer;
+    user-select: none;
+}
+.agent-thinking-step .thinking-header.no-toggle { cursor: default; }
+.agent-thinking-step .thinking-header:not(.no-toggle):hover { color: #64748b; }
+.dark .agent-thinking-step .thinking-header:not(.no-toggle):hover { color: #cbd5e1; }
+.agent-thinking-step .thinking-header i:first-child { font-size: 0.625rem; margin-top: 1px; }
+.agent-thinking-step .thinking-chevron {
+    font-size: 0.5rem;
+    margin-left: auto;
+    transition: transform 0.2s ease;
+    opacity: 0.5;
+}
+.agent-thinking-step.expanded .thinking-chevron { transform: rotate(90deg); }
+.agent-thinking-step .thinking-full {
+    display: none;
+    margin-top: 0.375rem;
+    margin-left: 1rem;
+    padding: 0.5rem;
+    background: rgba(0, 0, 0, 0.02);
+    border-radius: 6px;
+    border: 1px solid rgba(0, 0, 0, 0.04);
+    font-size: 0.75rem;
+    line-height: 1.5;
+    color: #94a3b8;
+    max-height: 200px;
+    overflow-y: auto;
+}
+.dark .agent-thinking-step .thinking-full {
+    background: rgba(255, 255, 255, 0.02);
+    border-color: rgba(255, 255, 255, 0.04);
+}
+.agent-thinking-step.expanded .thinking-full { display: block; }
+.agent-thinking-step .thinking-full p { margin: 0.25em 0; }
+.agent-thinking-step .thinking-full p:first-child { margin-top: 0; }
+.agent-thinking-step .thinking-full p:last-child { margin-bottom: 0; }
+
+/* Tool step - collapsible */
+.agent-tool-step .tool-header {
+    display: flex;
+    align-items: center;
+    gap: 0.375rem;
+    cursor: pointer;
+    user-select: none;
+    padding: 1px 0;
+    border-radius: 4px;
+}
+.agent-tool-step .tool-header:hover { color: #64748b; }
+.dark .agent-tool-step .tool-header:hover { color: #cbd5e1; }
+.agent-tool-step .tool-icon { font-size: 0.625rem; }
+.agent-tool-step .tool-chevron {
+    font-size: 0.5rem;
+    margin-left: auto;
+    transition: transform 0.2s ease;
+    opacity: 0.5;
+}
+.agent-tool-step.expanded .tool-chevron { transform: rotate(90deg); }
+.agent-tool-step .tool-time {
+    font-size: 0.65rem;
+    opacity: 0.6;
+    margin-left: 0.25rem;
+}
+
+/* Tool detail panel */
+.agent-tool-step .tool-detail {
+    display: none;
+    margin-top: 0.375rem;
+    margin-left: 1rem;
+    padding: 0.5rem;
+    background: rgba(0, 0, 0, 0.02);
+    border-radius: 6px;
+    border: 1px solid rgba(0, 0, 0, 0.04);
+}
+.dark .agent-tool-step .tool-detail {
+    background: rgba(255, 255, 255, 0.02);
+    border-color: rgba(255, 255, 255, 0.04);
+}
+.agent-tool-step.expanded .tool-detail { display: block; }
+.tool-detail-section { margin-bottom: 0.375rem; }
+.tool-detail-section:last-child { margin-bottom: 0; }
+.tool-detail-label {
+    font-size: 0.625rem;
+    font-weight: 600;
+    text-transform: uppercase;
+    letter-spacing: 0.05em;
+    opacity: 0.6;
+    margin-bottom: 0.125rem;
+}
+.tool-detail-content {
+    font-family: 'JetBrains Mono', 'Fira Code', Consolas, monospace;
+    font-size: 0.7rem;
+    line-height: 1.5;
+    white-space: pre-wrap;
+    word-break: break-all;
+    max-height: 200px;
+    overflow-y: auto;
+    margin: 0;
+    padding: 0.25rem 0;
+    background: transparent;
+    color: inherit;
+}
+.tool-error-text { color: #f87171; }
+
+/* Tool failed state */
+.agent-tool-step.tool-failed .tool-name { color: #f87171; }
+
+/* Config form controls */
+#view-config input[type="text"],
+#view-config input[type="number"],
+#view-config input[type="password"] {
+    height: 40px;
+    transition: border-color 0.2s ease, box-shadow 0.2s ease;
+}
+#view-config input:focus {
+    border-color: #4ABE6E;
+    box-shadow: 0 0 0 3px rgba(74, 190, 110, 0.12);
+}
+#view-config input[type="text"]:hover,
+#view-config input[type="number"]:hover,
+#view-config input[type="password"]:hover {
+    border-color: #94a3b8;
+}
+.dark #view-config input[type="text"]:hover,
+.dark #view-config input[type="number"]:hover,
+.dark #view-config input[type="password"]:hover {
+    border-color: #64748b;
+}
+
+/* Custom dropdown */
+.cfg-dropdown {
+    position: relative;
+    outline: none;
+}
+.cfg-dropdown-selected {
+    display: flex;
+    align-items: center;
+    justify-content: space-between;
+    height: 40px;
+    padding: 0 0.75rem;
+    border-radius: 0.5rem;
+    border: 1px solid #e2e8f0;
+    background: #f8fafc;
+    font-size: 0.875rem;
+    color: #1e293b;
+    cursor: pointer;
+    transition: border-color 0.2s ease, box-shadow 0.2s ease;
+    user-select: none;
+}
+.dark .cfg-dropdown-selected {
+    border-color: #475569;
+    background: rgba(255, 255, 255, 0.05);
+    color: #f1f5f9;
+}
+.cfg-dropdown-selected:hover { border-color: #94a3b8; }
+.dark .cfg-dropdown-selected:hover { border-color: #64748b; }
+.cfg-dropdown.open .cfg-dropdown-selected,
+.cfg-dropdown:focus .cfg-dropdown-selected {
+    border-color: #4ABE6E;
+    box-shadow: 0 0 0 3px rgba(74, 190, 110, 0.12);
+}
+.cfg-dropdown-arrow {
+    font-size: 0.625rem;
+    color: #94a3b8;
+    transition: transform 0.2s ease;
+    flex-shrink: 0;
+    margin-left: 0.5rem;
+}
+.cfg-dropdown.open .cfg-dropdown-arrow { transform: rotate(180deg); }
+.cfg-dropdown-menu {
+    display: none;
+    position: absolute;
+    top: calc(100% + 4px);
+    left: 0;
+    right: 0;
+    z-index: 50;
+    max-height: 240px;
+    overflow-y: auto;
+    border-radius: 0.5rem;
+    border: 1px solid #e2e8f0;
+    background: #ffffff;
+    box-shadow: 0 10px 25px -5px rgba(0, 0, 0, 0.1), 0 4px 10px -5px rgba(0, 0, 0, 0.04);
+    padding: 4px;
+}
+.dark .cfg-dropdown-menu {
+    border-color: #334155;
+    background: #1e1e1e;
+    box-shadow: 0 10px 25px -5px rgba(0, 0, 0, 0.4);
+}
+.cfg-dropdown.open .cfg-dropdown-menu { display: block; }
+.cfg-dropdown-item {
+    display: flex;
+    align-items: center;
+    padding: 8px 10px;
+    border-radius: 6px;
+    font-size: 0.875rem;
+    color: #334155;
+    cursor: pointer;
+    transition: background 0.15s ease;
+    white-space: nowrap;
+    overflow: hidden;
+    text-overflow: ellipsis;
+}
+.dark .cfg-dropdown-item { color: #cbd5e1; }
+.cfg-dropdown-item:hover { background: #f1f5f9; }
+.dark .cfg-dropdown-item:hover { background: rgba(255, 255, 255, 0.08); }
+.cfg-dropdown-item.active {
+    background: rgba(74, 190, 110, 0.1);
+    color: #228547;
+    font-weight: 500;
+}
+.dark .cfg-dropdown-item.active {
+    background: rgba(74, 190, 110, 0.15);
+    color: #74E9A4;
+}
+
+/* API Key masking via CSS (avoids browser password prompts) */
+.cfg-key-masked {
+    -webkit-text-security: disc;
+    text-security: disc;
+}
+
+/* Chat Input */
+#chat-input {
+    resize: none; height: 42px; max-height: 180px;
+    overflow-y: hidden;
+    transition: border-color 0.2s ease;
+}
+
+/* Attachment Preview Bar */
+.attachment-preview {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 8px;
+    padding: 8px 0;
+}
+.attachment-preview.hidden { display: none; }
+
+.att-thumb {
+    position: relative;
+    width: 64px; height: 64px;
+    border-radius: 8px;
+    overflow: hidden;
+    border: 1px solid #e2e8f0;
+    flex-shrink: 0;
+}
+.dark .att-thumb { border-color: rgba(255,255,255,0.1); }
+.att-thumb img {
+    width: 100%; height: 100%;
+    object-fit: cover;
+}
+
+.att-chip {
+    position: relative;
+    display: flex;
+    align-items: center;
+    gap: 6px;
+    padding: 6px 28px 6px 10px;
+    border-radius: 8px;
+    background: #f1f5f9;
+    border: 1px solid #e2e8f0;
+    font-size: 12px;
+    color: #475569;
+    max-width: 180px;
+}
+.dark .att-chip { background: rgba(255,255,255,0.05); border-color: rgba(255,255,255,0.1); color: #94a3b8; }
+.att-uploading { opacity: 0.6; pointer-events: none; }
+.att-name {
+    overflow: hidden;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+}
+
+.att-remove {
+    position: absolute;
+    top: -4px; right: -4px;
+    width: 18px; height: 18px;
+    border-radius: 50%;
+    background: #ef4444;
+    color: #fff;
+    border: none;
+    font-size: 12px;
+    line-height: 18px;
+    text-align: center;
+    cursor: pointer;
+    padding: 0;
+    opacity: 0;
+    transition: opacity 0.15s;
+}
+.att-thumb:hover .att-remove,
+.att-chip:hover .att-remove { opacity: 1; }
+
+/* Drag-over highlight */
+.drag-over {
+    background: rgba(74, 190, 110, 0.08) !important;
+    border-color: #4ABE6E !important;
+}
+
+/* User message attachments */
+.user-msg-attachments {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 6px;
+    margin-bottom: 6px;
+}
+.user-msg-image {
+    max-width: 200px;
+    max-height: 160px;
+    border-radius: 8px;
+    object-fit: cover;
+    cursor: pointer;
+}
+.user-msg-image:hover { opacity: 0.9; }
+.user-msg-file {
+    display: flex;
+    align-items: center;
+    gap: 6px;
+    padding: 4px 10px;
+    border-radius: 6px;
+    background: rgba(255,255,255,0.15);
+    font-size: 12px;
+}
+
+/* Placeholder Cards */
+.placeholder-card {
+    transition: transform 0.2s ease, box-shadow 0.2s ease;
+}
+.placeholder-card:hover {
+    transform: translateY(-2px);
+    box-shadow: 0 8px 25px -5px rgba(0, 0, 0, 0.1);
+}
--- a/channel/web/static/js/console.js
+++ b/channel/web/static/js/console.js
--- a/channel/web/web_channel.py
+++ b/channel/web/web_channel.py
--- a/channel/wechat/wcf_channel.py
+++ b/channel/wechat/wcf_channel.py
@@ -1,179 +0,0 @@
-# encoding:utf-8
-
-"""
-wechat channel
-"""
-
-import io
-import json
-import os
-import threading
-import time
-from queue import Empty
-from typing import Any
-
-from bridge.context import *
-from bridge.reply import *
-from channel.chat_channel import ChatChannel
-from channel.wechat.wcf_message import WechatfMessage
-from common.log import logger
-from common.singleton import singleton
-from common.utils import *
-from config import conf, get_appdata_dir
-from wcferry import Wcf, WxMsg
-
-
-@singleton
-class WechatfChannel(ChatChannel):
-    NOT_SUPPORT_REPLYTYPE = []
-
-    def __init__(self):
-        super().__init__()
-        self.NOT_SUPPORT_REPLYTYPE = []
-        # 使用字典存储最近消息，用于去重
-        self.received_msgs = {}
-        # 初始化wcferry客户端
-        self.wcf = Wcf()
-        self.wxid = None  # 登录后会被设置为当前登录用户的wxid
-
-    def startup(self):
-        """
-        启动通道
-        """
-        try:
-            # wcferry会自动唤起微信并登录
-            self.wxid = self.wcf.get_self_wxid()
-            self.name = self.wcf.get_user_info().get("name")
-            logger.info(f"微信登录成功，当前用户ID: {self.wxid}, 用户名：{self.name}")
-            self.contact_cache = ContactCache(self.wcf)
-            self.contact_cache.update()
-            # 启动消息接收
-            self.wcf.enable_receiving_msg()
-            # 创建消息处理线程
-            t = threading.Thread(target=self._process_messages, name="WeChatThread", daemon=True)
-            t.start()
-
-
-        except Exception as e:
-            logger.error(f"微信通道启动失败: {e}")
-            raise e
-
-    def _process_messages(self):
-        """
-        处理消息队列
-        """
-        while True:
-            try:
-                msg = self.wcf.get_msg()
-                if msg:
-                    self._handle_message(msg)
-            except Empty:
-                continue
-            except Exception as e:
-                logger.error(f"处理消息失败: {e}")
-                continue
-
-    def _handle_message(self, msg: WxMsg):
-        """
-        处理单条消息
-        """
-        try:
-            # 构造消息对象
-            cmsg = WechatfMessage(self, msg)
-            # 消息去重
-            if cmsg.msg_id in self.received_msgs:
-                return
-            self.received_msgs[cmsg.msg_id] = time.time()
-            # 清理过期消息ID
-            self._clean_expired_msgs()
-
-            logger.debug(f"收到消息: {msg}")
-            context = self._compose_context(cmsg.ctype, cmsg.content,
-                                            isgroup=cmsg.is_group,
-                                            msg=cmsg)
-            if context:
-                self.produce(context)
-        except Exception as e:
-            logger.error(f"处理消息失败: {e}")
-
-    def _clean_expired_msgs(self, expire_time: float = 60):
-        """
-        清理过期的消息ID
-        """
-        now = time.time()
-        for msg_id in list(self.received_msgs.keys()):
-            if now - self.received_msgs[msg_id] > expire_time:
-                del self.received_msgs[msg_id]
-
-    def send(self, reply: Reply, context: Context):
-        """
-        发送消息
-        """
-        receiver = context["receiver"]
-        if not receiver:
-            logger.error("receiver is empty")
-            return
-
-        try:
-            if reply.type == ReplyType.TEXT:
-                # 处理@信息
-                at_list = []
-                if context.get("isgroup"):
-                    if context["msg"].actual_user_id:
-                        at_list = [context["msg"].actual_user_id]
-                at_str = ",".join(at_list) if at_list else ""
-                self.wcf.send_text(reply.content, receiver, at_str)
-
-            elif reply.type == ReplyType.ERROR or reply.type == ReplyType.INFO:
-                self.wcf.send_text(reply.content, receiver)
-            else:
-                logger.error(f"暂不支持的消息类型: {reply.type}")
-
-        except Exception as e:
-            logger.error(f"发送消息失败: {e}")
-
-    def close(self):
-        """
-        关闭通道
-        """
-        try:
-            self.wcf.cleanup()
-        except Exception as e:
-            logger.error(f"关闭通道失败: {e}")
-
-
-class ContactCache:
-    def __init__(self, wcf):
-        """
-        wcf: 一个 wcfferry.client.Wcf 实例
-        """
-        self.wcf = wcf
-        self._contact_map = {}  # 形如 {wxid: {完整联系人信息}}
-
-    def update(self):
-        """
-        更新缓存：调用 get_contacts()，
-        再把 wcf.contacts 构建成 {wxid: {完整信息}} 的字典
-        """
-        self.wcf.get_contacts()
-        self._contact_map.clear()
-        for item in self.wcf.contacts:
-            wxid = item.get('wxid')
-            if wxid:  # 确保有 wxid 字段
-                self._contact_map[wxid] = item
-
-    def get_contact(self, wxid: str) -> dict:
-        """
-        返回该 wxid 对应的完整联系人 dict，
-        如果没找到就返回 None
-        """
-        return self._contact_map.get(wxid)
-
-    def get_name_by_wxid(self, wxid: str) -> str:
-        """
-        通过wxid，获取成员/群名称
-        """
-        contact = self.get_contact(wxid)
-        if contact:
-            return contact.get('name', '')
-        return ''
--- a/channel/wechat/wcf_message.py
+++ b/channel/wechat/wcf_message.py
@@ -1,58 +0,0 @@
-# encoding:utf-8
-
-"""
-wechat channel message
-"""
-
-from bridge.context import ContextType
-from channel.chat_message import ChatMessage
-from common.log import logger
-from wcferry import WxMsg
-
-
-class WechatfMessage(ChatMessage):
-    """
-    微信消息封装类
-    """
-
-    def __init__(self, channel, wcf_msg: WxMsg, is_group=False):
-        """
-        初始化消息对象
-        :param wcf_msg: wcferry消息对象
-        :param is_group: 是否是群消息
-        """
-        super().__init__(wcf_msg)
-        self.msg_id = wcf_msg.id
-        self.create_time = wcf_msg.ts  # 使用消息时间戳
-        self.is_group = is_group or wcf_msg._is_group
-        self.wxid = channel.wxid
-        self.name = channel.name
-
-        # 解析消息类型
-        if wcf_msg.is_text():
-            self.ctype = ContextType.TEXT
-            self.content = wcf_msg.content
-        else:
-            raise NotImplementedError(f"Unsupported message type: {wcf_msg.type}")
-
-        # 设置发送者和接收者信息
-        self.from_user_id = self.wxid if wcf_msg.sender == self.wxid else wcf_msg.sender
-        self.from_user_nickname = self.name if wcf_msg.sender == self.wxid else channel.contact_cache.get_name_by_wxid(wcf_msg.sender)
-        self.to_user_id = self.wxid
-        self.to_user_nickname = self.name
-        self.other_user_id = wcf_msg.sender
-        self.other_user_nickname = channel.contact_cache.get_name_by_wxid(wcf_msg.sender)
-
-        # 群消息特殊处理
-        if self.is_group:
-            self.other_user_id = wcf_msg.roomid
-            self.other_user_nickname = channel.contact_cache.get_name_by_wxid(wcf_msg.roomid)
-            self.actual_user_id = wcf_msg.sender
-            self.actual_user_nickname = channel.wcf.get_alias_in_chatroom(wcf_msg.sender, wcf_msg.roomid)
-            if not self.actual_user_nickname:  # 群聊获取不到企微号成员昵称，这里尝试从联系人缓存去获取
-                self.actual_user_nickname = channel.contact_cache.get_name_by_wxid(wcf_msg.sender)
-            self.room_id = wcf_msg.roomid
-            self.is_at = wcf_msg.is_at(self.wxid)  # 是否被@当前登录用户
-
-        # 判断是否是自己发送的消息
-        self.my_msg = wcf_msg.from_self()
--- a/channel/wechat/wechat_channel.py
+++ b/channel/wechat/wechat_channel.py
@@ -1,309 +0,0 @@
-# encoding:utf-8
-
-"""
-wechat channel
-"""
-
-import io
-import json
-import os
-import threading
-import time
-import requests
-
-from bridge.context import *
-from bridge.reply import *
-from channel.chat_channel import ChatChannel
-from channel import chat_channel
-from channel.wechat.wechat_message import *
-from common.expired_dict import ExpiredDict
-from common.log import logger
-from common.singleton import singleton
-from common.time_check import time_checker
-from common.utils import convert_webp_to_png, remove_markdown_symbol
-from config import conf, get_appdata_dir
-from lib import itchat
-from lib.itchat.content import *
-
-
-@itchat.msg_register([TEXT, VOICE, PICTURE, NOTE, ATTACHMENT, SHARING])
-def handler_single_msg(msg):
-    try:
-        cmsg = WechatMessage(msg, False)
-    except NotImplementedError as e:
-        logger.debug("[WX]single message {} skipped: {}".format(msg["MsgId"], e))
-        return None
-    WechatChannel().handle_single(cmsg)
-    return None
-
-
-@itchat.msg_register([TEXT, VOICE, PICTURE, NOTE, ATTACHMENT, SHARING], isGroupChat=True)
-def handler_group_msg(msg):
-    try:
-        cmsg = WechatMessage(msg, True)
-    except NotImplementedError as e:
-        logger.debug("[WX]group message {} skipped: {}".format(msg["MsgId"], e))
-        return None
-    WechatChannel().handle_group(cmsg)
-    return None
-
-
-def _check(func):
-    def wrapper(self, cmsg: ChatMessage):
-        msgId = cmsg.msg_id
-        if msgId in self.receivedMsgs:
-            logger.info("Wechat message {} already received, ignore".format(msgId))
-            return
-        self.receivedMsgs[msgId] = True
-        create_time = cmsg.create_time  # 消息时间戳
-        if conf().get("hot_reload") == True and int(create_time) < int(time.time()) - 60:  # 跳过1分钟前的历史消息
-            logger.debug("[WX]history message {} skipped".format(msgId))
-            return
-        if cmsg.my_msg and not cmsg.is_group:
-            logger.debug("[WX]my message {} skipped".format(msgId))
-            return
-        return func(self, cmsg)
-
-    return wrapper
-
-
-# 可用的二维码生成接口
-# https://api.qrserver.com/v1/create-qr-code/?size=400×400&data=https://www.abc.com
-# https://api.isoyu.com/qr/?m=1&e=L&p=20&url=https://www.abc.com
-def qrCallback(uuid, status, qrcode):
-    # logger.debug("qrCallback: {} {}".format(uuid,status))
-    if status == "0":
-        try:
-            from PIL import Image
-
-            img = Image.open(io.BytesIO(qrcode))
-            _thread = threading.Thread(target=img.show, args=("QRCode",))
-            _thread.setDaemon(True)
-            _thread.start()
-        except Exception as e:
-            pass
-
-        import qrcode
-
-        url = f"https://login.weixin.qq.com/l/{uuid}"
-
-        qr_api1 = "https://api.isoyu.com/qr/?m=1&e=L&p=20&url={}".format(url)
-        qr_api2 = "https://api.qrserver.com/v1/create-qr-code/?size=400×400&data={}".format(url)
-        qr_api3 = "https://api.pwmqr.com/qrcode/create/?url={}".format(url)
-        qr_api4 = "https://my.tv.sohu.com/user/a/wvideo/getQRCode.do?text={}".format(url)
-        print("You can also scan QRCode in any website below:")
-        print(qr_api3)
-        print(qr_api4)
-        print(qr_api2)
-        print(qr_api1)
-        _send_qr_code([qr_api3, qr_api4, qr_api2, qr_api1])
-        qr = qrcode.QRCode(border=1)
-        qr.add_data(url)
-        qr.make(fit=True)
-        try:
-            qr.print_ascii(invert=True)
-        except UnicodeEncodeError:
-            print("ASCII QR code printing failed due to encoding issues.")
-
-
-@singleton
-class WechatChannel(ChatChannel):
-    NOT_SUPPORT_REPLYTYPE = []
-
-    def __init__(self):
-        super().__init__()
-        self.receivedMsgs = ExpiredDict(conf().get("expires_in_seconds", 3600))
-        self.auto_login_times = 0
-
-    def startup(self):
-        try:
-            time.sleep(3)
-            logger.error("""[WechatChannel] 当前channel暂不可用，目前支持的channel有:
-                1. terminal: 终端
-                2. wechatmp: 个人公众号
-                3. wechatmp_service: 企业公众号
-                4. wechatcom_app: 企微自建应用
-                5. dingtalk: 钉钉
-                6. feishu: 飞书
-                7. web: 网页
-                8. wcf: wechat (需Windows环境，参考 https://github.com/zhayujie/chatgpt-on-wechat/pull/2562 )
-            可修改 config.json 配置文件的 channel_type 字段进行切换""")
-
-            # itchat.instance.receivingRetryCount = 600  # 修改断线超时时间
-            # # login by scan QRCode
-            # hotReload = conf().get("hot_reload", False)
-            # status_path = os.path.join(get_appdata_dir(), "itchat.pkl")
-            # itchat.auto_login(
-            #     enableCmdQR=2,
-            #     hotReload=hotReload,
-            #     statusStorageDir=status_path,
-            #     qrCallback=qrCallback,
-            #     exitCallback=self.exitCallback,
-            #     loginCallback=self.loginCallback
-            # )
-            # self.user_id = itchat.instance.storageClass.userName
-            # self.name = itchat.instance.storageClass.nickName
-            # logger.info("Wechat login success, user_id: {}, nickname: {}".format(self.user_id, self.name))
-            # # start message listener
-            # itchat.run()
-        except Exception as e:
-            logger.exception(e)
-
-    def exitCallback(self):
-        try:
-            from common.linkai_client import chat_client
-            if chat_client.client_id and conf().get("use_linkai"):
-                _send_logout()
-                time.sleep(2)
-                self.auto_login_times += 1
-                if self.auto_login_times < 100:
-                    chat_channel.handler_pool._shutdown = False
-                    self.startup()
-        except Exception as e:
-            pass
-
-    def loginCallback(self):
-        logger.debug("Login success")
-        _send_login_success()
-
-    # handle_* 系列函数处理收到的消息后构造Context，然后传入produce函数中处理Context和发送回复
-    # Context包含了消息的所有信息，包括以下属性
-    #   type 消息类型, 包括TEXT、VOICE、IMAGE_CREATE
-    #   content 消息内容，如果是TEXT类型，content就是文本内容，如果是VOICE类型，content就是语音文件名，如果是IMAGE_CREATE类型，content就是图片生成命令
-    #   kwargs 附加参数字典，包含以下的key：
-    #        session_id: 会话id
-    #        isgroup: 是否是群聊
-    #        receiver: 需要回复的对象
-    #        msg: ChatMessage消息对象
-    #        origin_ctype: 原始消息类型，语音转文字后，私聊时如果匹配前缀失败，会根据初始消息是否是语音来放宽触发规则
-    #        desire_rtype: 希望回复类型，默认是文本回复，设置为ReplyType.VOICE是语音回复
-    @time_checker
-    @_check
-    def handle_single(self, cmsg: ChatMessage):
-        # filter system message
-        if cmsg.other_user_id in ["weixin"]:
-            return
-        if cmsg.ctype == ContextType.VOICE:
-            if conf().get("speech_recognition") != True:
-                return
-            logger.debug("[WX]receive voice msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.IMAGE:
-            logger.debug("[WX]receive image msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.PATPAT:
-            logger.debug("[WX]receive patpat msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.TEXT:
-            logger.debug("[WX]receive text msg: {}, cmsg={}".format(json.dumps(cmsg._rawmsg, ensure_ascii=False), cmsg))
-        else:
-            logger.debug("[WX]receive msg: {}, cmsg={}".format(cmsg.content, cmsg))
-        context = self._compose_context(cmsg.ctype, cmsg.content, isgroup=False, msg=cmsg)
-        if context:
-            self.produce(context)
-
-    @time_checker
-    @_check
-    def handle_group(self, cmsg: ChatMessage):
-        if cmsg.ctype == ContextType.VOICE:
-            if conf().get("group_speech_recognition") != True:
-                return
-            logger.debug("[WX]receive voice for group msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.IMAGE:
-            logger.debug("[WX]receive image for group msg: {}".format(cmsg.content))
-        elif cmsg.ctype in [ContextType.JOIN_GROUP, ContextType.PATPAT, ContextType.ACCEPT_FRIEND, ContextType.EXIT_GROUP]:
-            logger.debug("[WX]receive note msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.TEXT:
-            # logger.debug("[WX]receive group msg: {}, cmsg={}".format(json.dumps(cmsg._rawmsg, ensure_ascii=False), cmsg))
-            pass
-        elif cmsg.ctype == ContextType.FILE:
-            logger.debug(f"[WX]receive attachment msg, file_name={cmsg.content}")
-        else:
-            logger.debug("[WX]receive group msg: {}".format(cmsg.content))
-        context = self._compose_context(cmsg.ctype, cmsg.content, isgroup=True, msg=cmsg, no_need_at=conf().get("no_need_at", False))
-        if context:
-            self.produce(context)
-
-    # 统一的发送函数，每个Channel自行实现，根据reply的type字段发送不同类型的消息
-    def send(self, reply: Reply, context: Context):
-        receiver = context["receiver"]
-        if reply.type == ReplyType.TEXT:
-            reply.content = remove_markdown_symbol(reply.content)
-            itchat.send(reply.content, toUserName=receiver)
-            logger.info("[WX] sendMsg={}, receiver={}".format(reply, receiver))
-        elif reply.type == ReplyType.ERROR or reply.type == ReplyType.INFO:
-            reply.content = remove_markdown_symbol(reply.content)
-            itchat.send(reply.content, toUserName=receiver)
-            logger.info("[WX] sendMsg={}, receiver={}".format(reply, receiver))
-        elif reply.type == ReplyType.VOICE:
-            itchat.send_file(reply.content, toUserName=receiver)
-            logger.info("[WX] sendFile={}, receiver={}".format(reply.content, receiver))
-        elif reply.type == ReplyType.IMAGE_URL:  # 从网络下载图片
-            img_url = reply.content
-            logger.debug(f"[WX] start download image, img_url={img_url}")
-            pic_res = requests.get(img_url, stream=True)
-            image_storage = io.BytesIO()
-            size = 0
-            for block in pic_res.iter_content(1024):
-                size += len(block)
-                image_storage.write(block)
-            logger.info(f"[WX] download image success, size={size}, img_url={img_url}")
-            image_storage.seek(0)
-            if ".webp" in img_url:
-                try:
-                    image_storage = convert_webp_to_png(image_storage)
-                except Exception as e:
-                    logger.error(f"Failed to convert image: {e}")
-                    return
-            itchat.send_image(image_storage, toUserName=receiver)
-            logger.info("[WX] sendImage url={}, receiver={}".format(img_url, receiver))
-        elif reply.type == ReplyType.IMAGE:  # 从文件读取图片
-            image_storage = reply.content
-            image_storage.seek(0)
-            itchat.send_image(image_storage, toUserName=receiver)
-            logger.info("[WX] sendImage, receiver={}".format(receiver))
-        elif reply.type == ReplyType.FILE:  # 新增文件回复类型
-            file_storage = reply.content
-            itchat.send_file(file_storage, toUserName=receiver)
-            logger.info("[WX] sendFile, receiver={}".format(receiver))
-        elif reply.type == ReplyType.VIDEO:  # 新增视频回复类型
-            video_storage = reply.content
-            itchat.send_video(video_storage, toUserName=receiver)
-            logger.info("[WX] sendFile, receiver={}".format(receiver))
-        elif reply.type == ReplyType.VIDEO_URL:  # 新增视频URL回复类型
-            video_url = reply.content
-            logger.debug(f"[WX] start download video, video_url={video_url}")
-            video_res = requests.get(video_url, stream=True)
-            video_storage = io.BytesIO()
-            size = 0
-            for block in video_res.iter_content(1024):
-                size += len(block)
-                video_storage.write(block)
-            logger.info(f"[WX] download video success, size={size}, video_url={video_url}")
-            video_storage.seek(0)
-            itchat.send_video(video_storage, toUserName=receiver)
-            logger.info("[WX] sendVideo url={}, receiver={}".format(video_url, receiver))
-
-def _send_login_success():
-    try:
-        from common.linkai_client import chat_client
-        if chat_client.client_id:
-            chat_client.send_login_success()
-    except Exception as e:
-        pass
-
-
-def _send_logout():
-    try:
-        from common.linkai_client import chat_client
-        if chat_client.client_id:
-            chat_client.send_logout()
-    except Exception as e:
-        pass
-
-
-def _send_qr_code(qrcode_list: list):
-    try:
-        from common.linkai_client import chat_client
-        if chat_client.client_id:
-            chat_client.send_qrcode(qrcode_list)
-    except Exception as e:
-        pass
-
--- a/channel/wechat/wechat_message.py
+++ b/channel/wechat/wechat_message.py
@@ -1,124 +0,0 @@
-import re
-
-from bridge.context import ContextType
-from channel.chat_message import ChatMessage
-from common.log import logger
-from common.tmp_dir import TmpDir
-from lib import itchat
-from lib.itchat.content import *
-
-class WechatMessage(ChatMessage):
-    def __init__(self, itchat_msg, is_group=False):
-        super().__init__(itchat_msg)
-        self.msg_id = itchat_msg["MsgId"]
-        self.create_time = itchat_msg["CreateTime"]
-        self.is_group = is_group
-
-        notes_join_group = ["加入群聊", "加入了群聊", "invited", "joined"]  # 可通过添加对应语言的加入群聊通知中的关键词适配更多
-        notes_bot_join_group = ["邀请你", "invited you", "You've joined", "你通过扫描"]
-        notes_exit_group = ["移出了群聊", "removed"]  # 可通过添加对应语言的踢出群聊通知中的关键词适配更多
-        notes_patpat = ["拍了拍我", "tickled my", "tickled me"] # 可通过添加对应语言的拍一拍通知中的关键词适配更多
-
-        if itchat_msg["Type"] == TEXT:
-            self.ctype = ContextType.TEXT
-            self.content = itchat_msg["Text"]
-        elif itchat_msg["Type"] == VOICE:
-            self.ctype = ContextType.VOICE
-            self.content = TmpDir().path() + itchat_msg["FileName"]  # content直接存临时目录路径
-            self._prepare_fn = lambda: itchat_msg.download(self.content)
-        elif itchat_msg["Type"] == PICTURE and itchat_msg["MsgType"] == 3:
-            self.ctype = ContextType.IMAGE
-            self.content = TmpDir().path() + itchat_msg["FileName"]  # content直接存临时目录路径
-            self._prepare_fn = lambda: itchat_msg.download(self.content)
-        elif itchat_msg["Type"] == NOTE and itchat_msg["MsgType"] == 10000:
-            if is_group:
-                if any(note_bot_join_group in itchat_msg["Content"] for note_bot_join_group in notes_bot_join_group):  # 邀请机器人加入群聊
-                    logger.warn("机器人加入群聊消息，不处理~")
-                    pass
-                elif any(note_join_group in itchat_msg["Content"] for note_join_group in notes_join_group): # 若有任何在notes_join_group列表中的字符串出现在NOTE中
-                # 这里只能得到nickname， actual_user_id还是机器人的id
-                    if "加入群聊" not in itchat_msg["Content"]:
-                        self.ctype = ContextType.JOIN_GROUP
-                        self.content = itchat_msg["Content"]
-                        if "invited" in itchat_msg["Content"]: # 匹配英文信息
-                            self.actual_user_nickname = re.findall(r'invited\s+(.+?)\s+to\s+the\s+group\s+chat', itchat_msg["Content"])[0]
-                        elif "joined" in itchat_msg["Content"]: # 匹配通过二维码加入的英文信息
-                            self.actual_user_nickname = re.findall(r'"(.*?)" joined the group chat via the QR Code shared by', itchat_msg["Content"])[0]
-                        elif "加入了群聊" in itchat_msg["Content"]:
-                            self.actual_user_nickname = re.findall(r"\"(.*?)\"", itchat_msg["Content"])[-1]
-                    elif "加入群聊" in itchat_msg["Content"]:
-                        self.ctype = ContextType.JOIN_GROUP
-                        self.content = itchat_msg["Content"]
-                        self.actual_user_nickname = re.findall(r"\"(.*?)\"", itchat_msg["Content"])[0]
-
-                elif any(note_exit_group in itchat_msg["Content"] for note_exit_group in notes_exit_group):  # 若有任何在notes_exit_group列表中的字符串出现在NOTE中
-                    self.ctype = ContextType.EXIT_GROUP
-                    self.content = itchat_msg["Content"]
-                    self.actual_user_nickname = re.findall(r"\"(.*?)\"", itchat_msg["Content"])[0]
-
-                elif any(note_patpat in itchat_msg["Content"] for note_patpat in notes_patpat):  # 若有任何在notes_patpat列表中的字符串出现在NOTE中:
-                    self.ctype = ContextType.PATPAT
-                    self.content = itchat_msg["Content"]
-                    if "拍了拍我" in itchat_msg["Content"]:  # 识别中文
-                        self.actual_user_nickname = re.findall(r"\"(.*?)\"", itchat_msg["Content"])[0]
-                    elif "tickled my" in itchat_msg["Content"] or "tickled me" in itchat_msg["Content"]:
-                        self.actual_user_nickname = re.findall(r'^(.*?)(?:tickled my|tickled me)', itchat_msg["Content"])[0]
-                else:
-                    raise NotImplementedError("Unsupported note message: " + itchat_msg["Content"])
-                    
-            elif "你已添加了" in itchat_msg["Content"]:  #通过好友请求
-                self.ctype = ContextType.ACCEPT_FRIEND
-                self.content = itchat_msg["Content"]
-            elif any(note_patpat in itchat_msg["Content"] for note_patpat in notes_patpat):  # 若有任何在notes_patpat列表中的字符串出现在NOTE中:
-                self.ctype = ContextType.PATPAT
-                self.content = itchat_msg["Content"]
-            else:
-                raise NotImplementedError("Unsupported note message: " + itchat_msg["Content"])
-        elif itchat_msg["Type"] == ATTACHMENT:
-            self.ctype = ContextType.FILE
-            self.content = TmpDir().path() + itchat_msg["FileName"]  # content直接存临时目录路径
-            self._prepare_fn = lambda: itchat_msg.download(self.content)
-        elif itchat_msg["Type"] == SHARING:
-            self.ctype = ContextType.SHARING
-            self.content = itchat_msg.get("Url")
-
-        else:
-            raise NotImplementedError("Unsupported message type: Type:{} MsgType:{}".format(itchat_msg["Type"], itchat_msg["MsgType"]))
-
-        self.from_user_id = itchat_msg["FromUserName"]
-        self.to_user_id = itchat_msg["ToUserName"]
-
-        user_id = itchat.instance.storageClass.userName
-        nickname = itchat.instance.storageClass.nickName
-
-        # 虽然from_user_id和to_user_id用的少，但是为了保持一致性，还是要填充一下
-        # 以下很繁琐，一句话总结：能填的都填了。
-        if self.from_user_id == user_id:
-            self.from_user_nickname = nickname
-        if self.to_user_id == user_id:
-            self.to_user_nickname = nickname
-        try:  # 陌生人时候, User字段可能不存在
-            # my_msg 为True是表示是自己发送的消息
-            self.my_msg = itchat_msg["ToUserName"] == itchat_msg["User"]["UserName"] and \
-                          itchat_msg["ToUserName"] != itchat_msg["FromUserName"]
-            self.other_user_id = itchat_msg["User"]["UserName"]
-            self.other_user_nickname = itchat_msg["User"]["NickName"]
-            if self.other_user_id == self.from_user_id:
-                self.from_user_nickname = self.other_user_nickname
-            if self.other_user_id == self.to_user_id:
-                self.to_user_nickname = self.other_user_nickname
-            if itchat_msg["User"].get("Self"):
-                # 自身的展示名，当设置了群昵称时，该字段表示群昵称
-                self.self_display_name = itchat_msg["User"].get("Self").get("DisplayName")
-        except KeyError as e:  # 处理偶尔没有对方信息的情况
-            logger.warn("[WX]get other_user_id failed: " + str(e))
-            if self.from_user_id == user_id:
-                self.other_user_id = self.to_user_id
-            else:
-                self.other_user_id = self.from_user_id
-
-        if self.is_group:
-            self.is_at = itchat_msg["IsAt"]
-            self.actual_user_id = itchat_msg["ActualUserName"]
-            if self.ctype not in [ContextType.JOIN_GROUP, ContextType.PATPAT, ContextType.EXIT_GROUP]:
-                self.actual_user_nickname = itchat_msg["ActualNickName"]
--- a/channel/wechat/wechaty_channel.py
+++ b/channel/wechat/wechaty_channel.py
@@ -1,129 +0,0 @@
-# encoding:utf-8
-
-"""
-wechaty channel
-Python Wechaty - https://github.com/wechaty/python-wechaty
-"""
-import asyncio
-import base64
-import os
-import time
-
-from wechaty import Contact, Wechaty
-from wechaty.user import Message
-from wechaty_puppet import FileBox
-
-from bridge.context import *
-from bridge.context import Context
-from bridge.reply import *
-from channel.chat_channel import ChatChannel
-from channel.wechat.wechaty_message import WechatyMessage
-from common.log import logger
-from common.singleton import singleton
-from config import conf
-
-try:
-    from voice.audio_convert import any_to_sil
-except Exception as e:
-    pass
-
-
-@singleton
-class WechatyChannel(ChatChannel):
-    NOT_SUPPORT_REPLYTYPE = []
-
-    def __init__(self):
-        super().__init__()
-
-    def startup(self):
-        config = conf()
-        token = config.get("wechaty_puppet_service_token")
-        os.environ["WECHATY_PUPPET_SERVICE_TOKEN"] = token
-        asyncio.run(self.main())
-
-    async def main(self):
-        loop = asyncio.get_event_loop()
-        # 将asyncio的loop传入处理线程
-        self.handler_pool._initializer = lambda: asyncio.set_event_loop(loop)
-        self.bot = Wechaty()
-        self.bot.on("login", self.on_login)
-        self.bot.on("message", self.on_message)
-        await self.bot.start()
-
-    async def on_login(self, contact: Contact):
-        self.user_id = contact.contact_id
-        self.name = contact.name
-        logger.info("[WX] login user={}".format(contact))
-
-    # 统一的发送函数，每个Channel自行实现，根据reply的type字段发送不同类型的消息
-    def send(self, reply: Reply, context: Context):
-        receiver_id = context["receiver"]
-        loop = asyncio.get_event_loop()
-        if context["isgroup"]:
-            receiver = asyncio.run_coroutine_threadsafe(self.bot.Room.find(receiver_id), loop).result()
-        else:
-            receiver = asyncio.run_coroutine_threadsafe(self.bot.Contact.find(receiver_id), loop).result()
-        msg = None
-        if reply.type == ReplyType.TEXT:
-            msg = reply.content
-            asyncio.run_coroutine_threadsafe(receiver.say(msg), loop).result()
-            logger.info("[WX] sendMsg={}, receiver={}".format(reply, receiver))
-        elif reply.type == ReplyType.ERROR or reply.type == ReplyType.INFO:
-            msg = reply.content
-            asyncio.run_coroutine_threadsafe(receiver.say(msg), loop).result()
-            logger.info("[WX] sendMsg={}, receiver={}".format(reply, receiver))
-        elif reply.type == ReplyType.VOICE:
-            voiceLength = None
-            file_path = reply.content
-            sil_file = os.path.splitext(file_path)[0] + ".sil"
-            voiceLength = int(any_to_sil(file_path, sil_file))
-            if voiceLength >= 60000:
-                voiceLength = 60000
-                logger.info("[WX] voice too long, length={}, set to 60s".format(voiceLength))
-            # 发送语音
-            t = int(time.time())
-            msg = FileBox.from_file(sil_file, name=str(t) + ".sil")
-            if voiceLength is not None:
-                msg.metadata["voiceLength"] = voiceLength
-            asyncio.run_coroutine_threadsafe(receiver.say(msg), loop).result()
-            try:
-                os.remove(file_path)
-                if sil_file != file_path:
-                    os.remove(sil_file)
-            except Exception as e:
-                pass
-            logger.info("[WX] sendVoice={}, receiver={}".format(reply.content, receiver))
-        elif reply.type == ReplyType.IMAGE_URL:  # 从网络下载图片
-            img_url = reply.content
-            t = int(time.time())
-            msg = FileBox.from_url(url=img_url, name=str(t) + ".png")
-            asyncio.run_coroutine_threadsafe(receiver.say(msg), loop).result()
-            logger.info("[WX] sendImage url={}, receiver={}".format(img_url, receiver))
-        elif reply.type == ReplyType.IMAGE:  # 从文件读取图片
-            image_storage = reply.content
-            image_storage.seek(0)
-            t = int(time.time())
-            msg = FileBox.from_base64(base64.b64encode(image_storage.read()), str(t) + ".png")
-            asyncio.run_coroutine_threadsafe(receiver.say(msg), loop).result()
-            logger.info("[WX] sendImage, receiver={}".format(receiver))
-
-    async def on_message(self, msg: Message):
-        """
-        listen for message event
-        """
-        try:
-            cmsg = await WechatyMessage(msg)
-        except NotImplementedError as e:
-            logger.debug("[WX] {}".format(e))
-            return
-        except Exception as e:
-            logger.exception("[WX] {}".format(e))
-            return
-        logger.debug("[WX] message:{}".format(cmsg))
-        room = msg.room()  # 获取消息来自的群聊. 如果消息不是来自群聊, 则返回None
-        isgroup = room is not None
-        ctype = cmsg.ctype
-        context = self._compose_context(ctype, cmsg.content, isgroup=isgroup, msg=cmsg)
-        if context:
-            logger.info("[WX] receiveMsg={}, context={}".format(cmsg, context))
-            self.produce(context)
--- a/channel/wechat/wechaty_message.py
+++ b/channel/wechat/wechaty_message.py
@@ -1,89 +0,0 @@
-import asyncio
-import re
-
-from wechaty import MessageType
-from wechaty.user import Message
-
-from bridge.context import ContextType
-from channel.chat_message import ChatMessage
-from common.log import logger
-from common.tmp_dir import TmpDir
-
-
-class aobject(object):
-    """Inheriting this class allows you to define an async __init__.
-
-    So you can create objects by doing something like `await MyClass(params)`
-    """
-
-    async def __new__(cls, *a, **kw):
-        instance = super().__new__(cls)
-        await instance.__init__(*a, **kw)
-        return instance
-
-    async def __init__(self):
-        pass
-
-
-class WechatyMessage(ChatMessage, aobject):
-    async def __init__(self, wechaty_msg: Message):
-        super().__init__(wechaty_msg)
-
-        room = wechaty_msg.room()
-
-        self.msg_id = wechaty_msg.message_id
-        self.create_time = wechaty_msg.payload.timestamp
-        self.is_group = room is not None
-
-        if wechaty_msg.type() == MessageType.MESSAGE_TYPE_TEXT:
-            self.ctype = ContextType.TEXT
-            self.content = wechaty_msg.text()
-        elif wechaty_msg.type() == MessageType.MESSAGE_TYPE_AUDIO:
-            self.ctype = ContextType.VOICE
-            voice_file = await wechaty_msg.to_file_box()
-            self.content = TmpDir().path() + voice_file.name  # content直接存临时目录路径
-
-            def func():
-                loop = asyncio.get_event_loop()
-                asyncio.run_coroutine_threadsafe(voice_file.to_file(self.content), loop).result()
-
-            self._prepare_fn = func
-
-        else:
-            raise NotImplementedError("Unsupported message type: {}".format(wechaty_msg.type()))
-
-        from_contact = wechaty_msg.talker()  # 获取消息的发送者
-        self.from_user_id = from_contact.contact_id
-        self.from_user_nickname = from_contact.name
-
-        # group中的from和to，wechaty跟itchat含义不一样
-        # wecahty: from是消息实际发送者, to:所在群
-        # itchat: 如果是你发送群消息，from和to是你自己和所在群，如果是别人发群消息，from和to是所在群和你自己
-        # 但这个差别不影响逻辑，group中只使用到：1.用from来判断是否是自己发的，2.actual_user_id来判断实际发送用户
-
-        if self.is_group:
-            self.to_user_id = room.room_id
-            self.to_user_nickname = await room.topic()
-        else:
-            to_contact = wechaty_msg.to()
-            self.to_user_id = to_contact.contact_id
-            self.to_user_nickname = to_contact.name
-
-        if self.is_group or wechaty_msg.is_self():  # 如果是群消息，other_user设置为群，如果是私聊消息，而且自己发的，就设置成对方。
-            self.other_user_id = self.to_user_id
-            self.other_user_nickname = self.to_user_nickname
-        else:
-            self.other_user_id = self.from_user_id
-            self.other_user_nickname = self.from_user_nickname
-
-        if self.is_group:  # wechaty群聊中，实际发送用户就是from_user
-            self.is_at = await wechaty_msg.mention_self()
-            if not self.is_at:  # 有时候复制粘贴的消息，不算做@，但是内容里面会有@xxx，这里做一下兼容
-                name = wechaty_msg.wechaty.user_self().name
-                pattern = f"@{re.escape(name)}(\u2005|\u0020)"
-                if re.search(pattern, self.content):
-                    logger.debug(f"wechaty message {self.msg_id} include at")
-                    self.is_at = True
-
-            self.actual_user_id = self.from_user_id
-            self.actual_user_nickname = self.from_user_nickname
--- a/channel/wechatcom/wechatcomapp_channel.py
+++ b/channel/wechatcom/wechatcomapp_channel.py
@@ -36,6 +36,7 @@ class WechatComAppChannel(ChatChannel):
        self.agent_id = conf().get("wechatcomapp_agent_id")
        self.token = conf().get("wechatcomapp_token")
        self.aes_key = conf().get("wechatcomapp_aes_key")
+        self._http_server = None
        logger.info(
            "[wechatcom] Initializing WeCom app channel, corp_id: {}, agent_id: {}".format(self.corp_id, self.agent_id)
        )
@@ -51,13 +52,24 @@ class WechatComAppChannel(ChatChannel):
        logger.info("[wechatcom] 📡 Listening on http://0.0.0.0:{}/wxcomapp/".format(port))
        logger.info("[wechatcom] 🤖 Ready to receive messages")
        
-        # Suppress web.py's default server startup message
-        old_stdout = sys.stdout
-        sys.stdout = io.StringIO()
+        # Build WSGI app with middleware (same as runsimple but without print)
+        func = web.httpserver.StaticMiddleware(app.wsgifunc())
+        func = web.httpserver.LogMiddleware(func)
+        server = web.httpserver.WSGIServer(("0.0.0.0", port), func)
+        self._http_server = server
        try:
-            web.httpserver.runsimple(app.wsgifunc(), ("0.0.0.0", port))
-        finally:
-            sys.stdout = old_stdout
+            server.start()
+        except (KeyboardInterrupt, SystemExit):
+            server.stop()
+
+    def stop(self):
+        if self._http_server:
+            try:
+                self._http_server.stop()
+                logger.info("[wechatcom] HTTP server stopped")
+            except Exception as e:
+                logger.warning(f"[wechatcom] Error stopping HTTP server: {e}")
+            self._http_server = None

    def send(self, reply: Reply, context: Context):
        receiver = context["receiver"]
--- a/channel/wechatmp/README.md
+++ b/channel/wechatmp/README.md
@@ -1,6 +1,6 @@
 # 微信公众号channel

-鉴于个人微信号在服务器上通过itchat登录有封号风险，这里新增了微信公众号channel，提供无风险的服务。
+微信公众号channel，提供稳定的服务。
 目前支持订阅号和服务号两种类型的公众号，它们都支持文本交互，语音和图片输入。其中个人主体的微信订阅号由于无法通过微信认证，存在回复时间限制，每天的图片和声音回复次数也有限制。

 ## 使用方法（订阅号，服务号类似）
--- a/channel/wechatmp/wechatmp_channel.py
+++ b/channel/wechatmp/wechatmp_channel.py
@@ -41,6 +41,7 @@ class WechatMPChannel(ChatChannel):
        super().__init__()
        self.passive_reply = passive_reply
        self.NOT_SUPPORT_REPLYTYPE = []
+        self._http_server = None
        appid = conf().get("wechatmp_app_id")
        secret = conf().get("wechatmp_app_secret")
        token = conf().get("wechatmp_token")
@@ -69,7 +70,23 @@ class WechatMPChannel(ChatChannel):
            urls = ("/wx", "channel.wechatmp.active_reply.Query")
        app = web.application(urls, globals(), autoreload=False)
        port = conf().get("wechatmp_port", 8080)
-        web.httpserver.runsimple(app.wsgifunc(), ("0.0.0.0", port))
+        func = web.httpserver.StaticMiddleware(app.wsgifunc())
+        func = web.httpserver.LogMiddleware(func)
+        server = web.httpserver.WSGIServer(("0.0.0.0", port), func)
+        self._http_server = server
+        try:
+            server.start()
+        except (KeyboardInterrupt, SystemExit):
+            server.stop()
+
+    def stop(self):
+        if self._http_server:
+            try:
+                self._http_server.stop()
+                logger.info("[wechatmp] HTTP server stopped")
+            except Exception as e:
+                logger.warning(f"[wechatmp] Error stopping HTTP server: {e}")
+            self._http_server = None

    def start_loop(self, loop):
        asyncio.set_event_loop(loop)
--- a/channel/wecom_bot/init.py
+++ b/channel/wecom_bot/init.py
--- a/channel/wecom_bot/wecom_bot_channel.py
+++ b/channel/wecom_bot/wecom_bot_channel.py
@@ -0,0 +1,767 @@
+"""
+WeCom (企业微信) AI Bot channel via WebSocket long connection.
+
+Supports:
+- Single chat and group chat (text / image / file input & output)
+- Scheduled task push via aibot_send_msg
+- Heartbeat keep-alive and auto-reconnect
+"""
+
+import base64
+import hashlib
+import json
+import math
+import os
+import threading
+import time
+import uuid
+
+import requests
+import websocket
+
+from bridge.context import Context, ContextType
+from bridge.reply import Reply, ReplyType
+from channel.chat_channel import ChatChannel, check_prefix
+from channel.wecom_bot.wecom_bot_message import WecomBotMessage
+from common.expired_dict import ExpiredDict
+from common.log import logger
+from common.singleton import singleton
+from config import conf
+
+WECOM_WS_URL = "wss://openws.work.weixin.qq.com"
+HEARTBEAT_INTERVAL = 30
+MEDIA_CHUNK_SIZE = 512 * 1024  # 512KB per chunk (before base64 encoding)
+
+
+@singleton
+class WecomBotChannel(ChatChannel):
+
+    def __init__(self):
+        super().__init__()
+        self.bot_id = ""
+        self.bot_secret = ""
+        self.received_msgs = ExpiredDict(60 * 60 * 7.1)
+        self._ws = None
+        self._ws_thread = None
+        self._heartbeat_thread = None
+        self._connected = False
+        self._stop_event = threading.Event()
+        self._pending_responses = {}  # req_id -> (threading.Event, result_holder)
+        self._pending_lock = threading.Lock()
+        self._stream_states = {}  # req_id -> {"stream_id": str, "content": str}
+
+        conf()["group_name_white_list"] = ["ALL_GROUP"]
+        conf()["single_chat_prefix"] = [""]
+
+    # ------------------------------------------------------------------
+    # Lifecycle
+    # ------------------------------------------------------------------
+
+    def startup(self):
+        self.bot_id = conf().get("wecom_bot_id", "")
+        self.bot_secret = conf().get("wecom_bot_secret", "")
+
+        if not self.bot_id or not self.bot_secret:
+            err = "[WecomBot] wecom_bot_id and wecom_bot_secret are required"
+            logger.error(err)
+            self.report_startup_error(err)
+            return
+
+        self._stop_event.clear()
+        self._start_ws()
+
+    def stop(self):
+        logger.info("[WecomBot] stop() called")
+        self._stop_event.set()
+        if self._ws:
+            try:
+                self._ws.close()
+            except Exception:
+                pass
+        self._ws = None
+        self._connected = False
+
+    # ------------------------------------------------------------------
+    # WebSocket connection
+    # ------------------------------------------------------------------
+
+    def _start_ws(self):
+        def _on_open(ws):
+            logger.info("[WecomBot] WebSocket connected, sending subscribe...")
+            self._send_subscribe()
+
+        def _on_message(ws, raw):
+            try:
+                data = json.loads(raw)
+                self._handle_ws_message(data)
+            except Exception as e:
+                logger.error(f"[WecomBot] Failed to handle ws message: {e}", exc_info=True)
+
+        def _on_error(ws, error):
+            logger.error(f"[WecomBot] WebSocket error: {error}")
+
+        def _on_close(ws, close_status_code, close_msg):
+            logger.warning(f"[WecomBot] WebSocket closed: status={close_status_code}, msg={close_msg}")
+            self._connected = False
+            if not self._stop_event.is_set():
+                logger.info("[WecomBot] Will reconnect in 5s...")
+                time.sleep(5)
+                if not self._stop_event.is_set():
+                    self._start_ws()
+
+        self._ws = websocket.WebSocketApp(
+            WECOM_WS_URL,
+            on_open=_on_open,
+            on_message=_on_message,
+            on_error=_on_error,
+            on_close=_on_close,
+        )
+
+        def run_forever():
+            try:
+                self._ws.run_forever(ping_interval=0, reconnect=0)
+            except (SystemExit, KeyboardInterrupt):
+                logger.info("[WecomBot] WebSocket thread interrupted")
+            except Exception as e:
+                logger.error(f"[WecomBot] WebSocket run_forever error: {e}")
+
+        self._ws_thread = threading.Thread(target=run_forever, daemon=True)
+        self._ws_thread.start()
+        self._ws_thread.join()
+
+    def _ws_send(self, data: dict):
+        if self._ws:
+            self._ws.send(json.dumps(data, ensure_ascii=False))
+
+    def _gen_req_id(self) -> str:
+        return uuid.uuid4().hex[:16]
+
+    # ------------------------------------------------------------------
+    # Subscribe & heartbeat
+    # ------------------------------------------------------------------
+
+    def _send_subscribe(self):
+        self._ws_send({
+            "cmd": "aibot_subscribe",
+            "headers": {"req_id": self._gen_req_id()},
+            "body": {
+                "bot_id": self.bot_id,
+                "secret": self.bot_secret,
+            },
+        })
+
+    def _start_heartbeat(self):
+        if self._heartbeat_thread and self._heartbeat_thread.is_alive():
+            return
+
+        def heartbeat_loop():
+            while not self._stop_event.is_set() and self._connected:
+                try:
+                    self._ws_send({
+                        "cmd": "ping",
+                        "headers": {"req_id": self._gen_req_id()},
+                    })
+                except Exception as e:
+                    logger.warning(f"[WecomBot] Heartbeat send failed: {e}")
+                    break
+                self._stop_event.wait(HEARTBEAT_INTERVAL)
+
+        self._heartbeat_thread = threading.Thread(target=heartbeat_loop, daemon=True)
+        self._heartbeat_thread.start()
+
+    # ------------------------------------------------------------------
+    # Incoming message dispatch
+    # ------------------------------------------------------------------
+
+    def _send_and_wait(self, data: dict, timeout: float = 15) -> dict:
+        """Send a ws message and wait for the matching response by req_id."""
+        req_id = data.get("headers", {}).get("req_id", "")
+        event = threading.Event()
+        holder = {"data": None}
+        with self._pending_lock:
+            self._pending_responses[req_id] = (event, holder)
+        self._ws_send(data)
+        event.wait(timeout=timeout)
+        with self._pending_lock:
+            self._pending_responses.pop(req_id, None)
+        return holder["data"] or {}
+
+    def _handle_ws_message(self, data: dict):
+        cmd = data.get("cmd", "")
+        errcode = data.get("errcode")
+        req_id = data.get("headers", {}).get("req_id", "")
+
+        # Check if this is a response to a pending request
+        if req_id:
+            with self._pending_lock:
+                pending = self._pending_responses.get(req_id)
+            if pending:
+                event, holder = pending
+                holder["data"] = data
+                event.set()
+                return
+
+        # Subscribe response (only handle once before connected)
+        if errcode is not None and cmd == "":
+            if not self._connected:
+                if errcode == 0:
+                    logger.info("[WecomBot] ✅ Subscribe success")
+                    self._connected = True
+                    self._start_heartbeat()
+                    self.report_startup_success()
+                else:
+                    errmsg = data.get("errmsg", "unknown error")
+                    logger.error(f"[WecomBot] Subscribe failed: errcode={errcode}, errmsg={errmsg}")
+                    self.report_startup_error(errmsg)
+            return
+
+        if cmd == "aibot_msg_callback":
+            self._handle_msg_callback(data)
+        elif cmd == "aibot_event_callback":
+            self._handle_event_callback(data)
+        elif cmd == "":
+            if errcode and errcode != 0:
+                logger.warning(f"[WecomBot] Response error: {data}")
+
+    # ------------------------------------------------------------------
+    # Message callback
+    # ------------------------------------------------------------------
+
+    def _handle_msg_callback(self, data: dict):
+        body = data.get("body", {})
+        req_id = data.get("headers", {}).get("req_id", "")
+        msg_id = body.get("msgid", "")
+
+        if self.received_msgs.get(msg_id):
+            logger.debug(f"[WecomBot] Duplicate msg filtered: {msg_id}")
+            return
+        self.received_msgs[msg_id] = True
+
+        chattype = body.get("chattype", "single")
+        is_group = chattype == "group"
+
+        try:
+            wecom_msg = WecomBotMessage(body, is_group=is_group)
+        except NotImplementedError as e:
+            logger.warning(f"[WecomBot] {e}")
+            return
+        except Exception as e:
+            logger.error(f"[WecomBot] Failed to parse message: {e}", exc_info=True)
+            return
+
+        wecom_msg.req_id = req_id
+
+        # File cache logic (same pattern as feishu)
+        from channel.file_cache import get_file_cache
+        file_cache = get_file_cache()
+
+        if is_group:
+            if conf().get("group_shared_session", True):
+                session_id = body.get("chatid", "")
+            else:
+                session_id = wecom_msg.from_user_id + "_" + body.get("chatid", "")
+        else:
+            session_id = wecom_msg.from_user_id
+
+        if wecom_msg.ctype == ContextType.IMAGE:
+            if hasattr(wecom_msg, "image_path") and wecom_msg.image_path:
+                file_cache.add(session_id, wecom_msg.image_path, file_type="image")
+                logger.info(f"[WecomBot] Image cached for session {session_id}")
+            return
+
+        if wecom_msg.ctype == ContextType.FILE:
+            wecom_msg.prepare()
+            file_cache.add(session_id, wecom_msg.content, file_type="file")
+            logger.info(f"[WecomBot] File cached for session {session_id}: {wecom_msg.content}")
+            return
+
+        if wecom_msg.ctype == ContextType.TEXT:
+            cached_files = file_cache.get(session_id)
+            if cached_files:
+                file_refs = []
+                for fi in cached_files:
+                    ftype = fi["type"]
+                    fpath = fi["path"]
+                    if ftype == "image":
+                        file_refs.append(f"[图片: {fpath}]")
+                    elif ftype == "video":
+                        file_refs.append(f"[视频: {fpath}]")
+                    else:
+                        file_refs.append(f"[文件: {fpath}]")
+                wecom_msg.content = wecom_msg.content + "\n" + "\n".join(file_refs)
+                logger.info(f"[WecomBot] Attached {len(cached_files)} cached file(s)")
+                file_cache.clear(session_id)
+
+        context = self._compose_context(
+            wecom_msg.ctype,
+            wecom_msg.content,
+            isgroup=is_group,
+            msg=wecom_msg,
+            no_need_at=True,
+        )
+        if context:
+            if req_id:
+                context["on_event"] = self._make_stream_callback(req_id)
+            self.produce(context)
+
+    # ------------------------------------------------------------------
+    # Event callback
+    # ------------------------------------------------------------------
+
+    def _handle_event_callback(self, data: dict):
+        body = data.get("body", {})
+        event = body.get("event", {})
+        event_type = event.get("eventtype", "")
+
+        if event_type == "enter_chat":
+            logger.info(f"[WecomBot] User entered chat: {body.get('from', {}).get('userid')}")
+        elif event_type == "disconnected_event":
+            logger.warning("[WecomBot] Received disconnected_event, another connection took over")
+        else:
+            logger.debug(f"[WecomBot] Event: {event_type}")
+
+    # ------------------------------------------------------------------
+    # Stream callback (for agent on_event)
+    # ------------------------------------------------------------------
+
+    def _make_stream_callback(self, req_id: str):
+        """Build an on_event callback that pushes agent stream deltas to wecom via stream message.
+
+        All intermediate segments (thinking before tool calls) and the final answer
+        are accumulated into a single stream message, separated by '---'.
+        """
+        stream_id = uuid.uuid4().hex[:16]
+        self._stream_states[req_id] = {
+            "stream_id": stream_id,
+            "committed": "",  # finalized content from previous segments
+            "current": "",    # current segment being streamed
+        }
+
+        def _push_stream(state: dict):
+            """Push current stream content to wecom."""
+            self._ws_send({
+                "cmd": "aibot_respond_msg",
+                "headers": {"req_id": req_id},
+                "body": {
+                    "msgtype": "stream",
+                    "stream": {
+                        "id": state["stream_id"],
+                        "finish": False,
+                        "content": state["committed"] + state["current"],
+                    },
+                },
+            })
+
+        def on_event(event: dict):
+            event_type = event.get("type")
+            data = event.get("data", {})
+            state = self._stream_states.get(req_id)
+            if not state:
+                return
+
+            if event_type == "turn_start":
+                state["current"] = ""
+
+            elif event_type == "message_update":
+                delta = data.get("delta", "")
+                if delta:
+                    state["current"] += delta
+                    _push_stream(state)
+
+            elif event_type == "message_end":
+                tool_calls = data.get("tool_calls", [])
+                if tool_calls:
+                    if state["current"].strip():
+                        state["committed"] += state["current"].strip() + "\n\n---\n\n"
+                        state["current"] = ""
+                else:
+                    state["committed"] += state["current"]
+                    state["current"] = ""
+
+        return on_event
+
+    # ------------------------------------------------------------------
+    # _compose_context (same pattern as feishu)
+    # ------------------------------------------------------------------
+
+    def _compose_context(self, ctype: ContextType, content, **kwargs):
+        context = Context(ctype, content)
+        context.kwargs = kwargs
+        if "channel_type" not in context:
+            context["channel_type"] = self.channel_type
+        if "origin_ctype" not in context:
+            context["origin_ctype"] = ctype
+
+        cmsg = context["msg"]
+
+        if cmsg.is_group:
+            if conf().get("group_shared_session", True):
+                context["session_id"] = cmsg.other_user_id
+            else:
+                context["session_id"] = f"{cmsg.from_user_id}:{cmsg.other_user_id}"
+        else:
+            context["session_id"] = cmsg.from_user_id
+
+        context["receiver"] = cmsg.other_user_id
+
+        if ctype == ContextType.TEXT:
+            img_match_prefix = check_prefix(content, conf().get("image_create_prefix"))
+            if img_match_prefix:
+                content = content.replace(img_match_prefix, "", 1)
+                context.type = ContextType.IMAGE_CREATE
+            else:
+                context.type = ContextType.TEXT
+            context.content = content.strip()
+
+        return context
+
+    # ------------------------------------------------------------------
+    # Send reply
+    # ------------------------------------------------------------------
+
+    def send(self, reply: Reply, context: Context):
+        msg = context.get("msg")
+        is_group = context.get("isgroup", False)
+        receiver = context.get("receiver", "")
+
+        # Determine req_id for responding or use send_msg for scheduled push
+        req_id = getattr(msg, "req_id", None) if msg else None
+
+        if reply.type == ReplyType.TEXT:
+            self._send_text(reply.content, receiver, is_group, req_id)
+        elif reply.type in (ReplyType.IMAGE_URL, ReplyType.IMAGE):
+            self._send_image(reply.content, receiver, is_group, req_id)
+        elif reply.type == ReplyType.FILE:
+            if hasattr(reply, "text_content") and reply.text_content:
+                self._send_text(reply.text_content, receiver, is_group, req_id)
+                time.sleep(0.3)
+            self._send_file(reply.content, receiver, is_group, req_id)
+        elif reply.type == ReplyType.VIDEO or reply.type == ReplyType.VIDEO_URL:
+            self._send_file(reply.content, receiver, is_group, req_id, media_type="video")
+        else:
+            logger.warning(f"[WecomBot] Unsupported reply type: {reply.type}, falling back to text")
+            self._send_text(str(reply.content), receiver, is_group, req_id)
+
+    # ------------------------------------------------------------------
+    # Respond message (via websocket)
+    # ------------------------------------------------------------------
+
+    def _send_text(self, content: str, receiver: str, is_group: bool, req_id: str = None):
+        """Send text/markdown reply. Reuses stream state if available (streaming mode)."""
+        if req_id:
+            state = self._stream_states.pop(req_id, None)
+            if state:
+                final_content = state["committed"]
+                stream_id = state["stream_id"]
+            else:
+                final_content = content
+                stream_id = uuid.uuid4().hex[:16]
+            self._ws_send({
+                "cmd": "aibot_respond_msg",
+                "headers": {"req_id": req_id},
+                "body": {
+                    "msgtype": "stream",
+                    "stream": {
+                        "id": stream_id,
+                        "finish": True,
+                        "content": final_content,
+                    },
+                },
+            })
+        else:
+            self._active_send_markdown(content, receiver, is_group)
+
+    def _send_image(self, img_path_or_url: str, receiver: str, is_group: bool, req_id: str = None):
+        """Send image reply. Converts to JPG/PNG and compresses if >2MB."""
+        local_path = img_path_or_url
+        if local_path.startswith("file://"):
+            local_path = local_path[7:]
+
+        if local_path.startswith(("http://", "https://")):
+            try:
+                resp = requests.get(local_path, timeout=30)
+                resp.raise_for_status()
+                ct = resp.headers.get("Content-Type", "")
+                if "jpeg" in ct or "jpg" in ct:
+                    ext = ".jpg"
+                elif "webp" in ct:
+                    ext = ".webp"
+                elif "gif" in ct:
+                    ext = ".gif"
+                else:
+                    ext = ".png"
+                tmp_path = f"/tmp/wecom_img_{uuid.uuid4().hex[:8]}{ext}"
+                with open(tmp_path, "wb") as f:
+                    f.write(resp.content)
+                logger.info(f"[WecomBot] Image downloaded: size={len(resp.content)}, "
+                            f"content-type={ct}, path={tmp_path}")
+                local_path = tmp_path
+            except Exception as e:
+                logger.error(f"[WecomBot] Failed to download image for sending: {e}")
+                self._send_text("[Image send failed]", receiver, is_group, req_id)
+                return
+
+        if not os.path.exists(local_path):
+            logger.error(f"[WecomBot] Image file not found: {local_path}")
+            return
+
+        max_image_size = 2 * 1024 * 1024  # 2MB limit for image upload
+        local_path = self._ensure_image_format(local_path)
+        if not local_path:
+            self._send_text("[Image format conversion failed]", receiver, is_group, req_id)
+            return
+
+        if os.path.getsize(local_path) > max_image_size:
+            local_path = self._compress_image(local_path, max_image_size)
+            if not local_path:
+                self._send_text("[Image too large]", receiver, is_group, req_id)
+                return
+
+        file_size = os.path.getsize(local_path)
+        logger.info(f"[WecomBot] Uploading image: path={local_path}, size={file_size} bytes")
+        media_id = self._upload_media(local_path, "image")
+        if not media_id:
+            logger.error("[WecomBot] Failed to upload image")
+            self._send_text("[Image upload failed]", receiver, is_group, req_id)
+            return
+
+        if req_id:
+            self._ws_send({
+                "cmd": "aibot_respond_msg",
+                "headers": {"req_id": req_id},
+                "body": {
+                    "msgtype": "image",
+                    "image": {"media_id": media_id},
+                },
+            })
+        else:
+            self._ws_send({
+                "cmd": "aibot_send_msg",
+                "headers": {"req_id": self._gen_req_id()},
+                "body": {
+                    "chatid": receiver,
+                    "chat_type": 2 if is_group else 1,
+                    "msgtype": "image",
+                    "image": {"media_id": media_id},
+                },
+            })
+
+    @staticmethod
+    def _ensure_image_format(file_path: str) -> str:
+        """Ensure image is JPG or PNG (the only formats wecom supports). Convert if needed."""
+        try:
+            from PIL import Image
+            img = Image.open(file_path)
+            fmt = (img.format or "").upper()
+            if fmt in ("JPEG", "PNG"):
+                # Already a supported format, but make sure the filename extension matches
+                ext = os.path.splitext(file_path)[1].lower()
+                if fmt == "JPEG" and ext in (".jpg", ".jpeg"):
+                    return file_path
+                if fmt == "PNG" and ext == ".png":
+                    return file_path
+                # Extension doesn't match — rename/copy with correct extension
+                correct_ext = ".jpg" if fmt == "JPEG" else ".png"
+                out_path = f"/tmp/wecom_fmt_{uuid.uuid4().hex[:8]}{correct_ext}"
+                img.save(out_path, fmt)
+                logger.info(f"[WecomBot] Image renamed: {file_path} -> {out_path} ({fmt})")
+                return out_path
+
+            # Unsupported format (WebP, GIF, BMP, etc.) — convert to PNG
+            if img.mode == "RGBA":
+                out_path = f"/tmp/wecom_fmt_{uuid.uuid4().hex[:8]}.png"
+                img.save(out_path, "PNG")
+            else:
+                out_path = f"/tmp/wecom_fmt_{uuid.uuid4().hex[:8]}.jpg"
+                img.convert("RGB").save(out_path, "JPEG", quality=90)
+            logger.info(f"[WecomBot] Image converted from {fmt} -> {out_path}")
+            return out_path
+        except Exception as e:
+            logger.error(f"[WecomBot] Image format check failed: {e}")
+            return file_path
+
+    @staticmethod
+    def _compress_image(file_path: str, max_bytes: int) -> str:
+        """Compress image to fit within max_bytes. Returns new path or empty string."""
+        try:
+            from PIL import Image
+            img = Image.open(file_path)
+            if img.mode == "RGBA":
+                img = img.convert("RGB")
+
+            out_path = f"/tmp/wecom_compressed_{uuid.uuid4().hex[:8]}.jpg"
+            quality = 85
+            while quality >= 30:
+                img.save(out_path, "JPEG", quality=quality, optimize=True)
+                if os.path.getsize(out_path) <= max_bytes:
+                    logger.info(f"[WecomBot] Image compressed: quality={quality}, "
+                                f"size={os.path.getsize(out_path)} bytes")
+                    return out_path
+                quality -= 10
+
+            # Still too large — resize
+            ratio = (max_bytes / os.path.getsize(out_path)) ** 0.5
+            new_size = (int(img.width * ratio), int(img.height * ratio))
+            img = img.resize(new_size, Image.LANCZOS)
+            img.save(out_path, "JPEG", quality=70, optimize=True)
+            if os.path.getsize(out_path) <= max_bytes:
+                logger.info(f"[WecomBot] Image compressed with resize: {new_size}, "
+                            f"size={os.path.getsize(out_path)} bytes")
+                return out_path
+
+            logger.error(f"[WecomBot] Cannot compress image below {max_bytes} bytes")
+            return ""
+        except Exception as e:
+            logger.error(f"[WecomBot] Image compression failed: {e}")
+            return ""
+
+    def _send_file(self, file_path: str, receiver: str, is_group: bool,
+                   req_id: str = None, media_type: str = "file"):
+        """Send file/video reply by uploading media first."""
+        local_path = file_path
+        if local_path.startswith("file://"):
+            local_path = local_path[7:]
+
+        if local_path.startswith(("http://", "https://")):
+            try:
+                resp = requests.get(local_path, timeout=60)
+                resp.raise_for_status()
+                ext = os.path.splitext(local_path)[1] or ".bin"
+                tmp_path = f"/tmp/wecom_file_{uuid.uuid4().hex[:8]}{ext}"
+                with open(tmp_path, "wb") as f:
+                    f.write(resp.content)
+                local_path = tmp_path
+            except Exception as e:
+                logger.error(f"[WecomBot] Failed to download file for sending: {e}")
+                return
+
+        if not os.path.exists(local_path):
+            logger.error(f"[WecomBot] File not found: {local_path}")
+            return
+
+        media_id = self._upload_media(local_path, media_type)
+        if not media_id:
+            logger.error(f"[WecomBot] Failed to upload {media_type}")
+            return
+
+        if req_id:
+            self._ws_send({
+                "cmd": "aibot_respond_msg",
+                "headers": {"req_id": req_id},
+                "body": {
+                    "msgtype": media_type,
+                    media_type: {"media_id": media_id},
+                },
+            })
+        else:
+            self._ws_send({
+                "cmd": "aibot_send_msg",
+                "headers": {"req_id": self._gen_req_id()},
+                "body": {
+                    "chatid": receiver,
+                    "chat_type": 2 if is_group else 1,
+                    "msgtype": media_type,
+                    media_type: {"media_id": media_id},
+                },
+            })
+
+    def _active_send_markdown(self, content: str, receiver: str, is_group: bool):
+        """Proactively send markdown message (for scheduled tasks, no req_id)."""
+        self._ws_send({
+            "cmd": "aibot_send_msg",
+            "headers": {"req_id": self._gen_req_id()},
+            "body": {
+                "chatid": receiver,
+                "chat_type": 2 if is_group else 1,
+                "msgtype": "markdown",
+                "markdown": {"content": content},
+            },
+        })
+
+    # ------------------------------------------------------------------
+    # Media upload (chunked)
+    # ------------------------------------------------------------------
+
+    def _upload_media(self, file_path: str, media_type: str = "file") -> str:
+        """
+        Upload a local file to wecom bot via chunked upload protocol.
+        Returns media_id on success, empty string on failure.
+        """
+        if not os.path.exists(file_path):
+            logger.error(f"[WecomBot] Upload file not found: {file_path}")
+            return ""
+
+        file_size = os.path.getsize(file_path)
+        if file_size < 5:
+            logger.error(f"[WecomBot] File too small: {file_size} bytes")
+            return ""
+
+        filename = os.path.basename(file_path)
+        total_chunks = math.ceil(file_size / MEDIA_CHUNK_SIZE)
+        if total_chunks > 100:
+            logger.error(f"[WecomBot] Too many chunks: {total_chunks} > 100")
+            return ""
+
+        file_md5 = hashlib.md5()
+        with open(file_path, "rb") as f:
+            for block in iter(lambda: f.read(8192), b""):
+                file_md5.update(block)
+        md5_hex = file_md5.hexdigest()
+
+        # 1. Init upload
+        init_resp = self._send_and_wait({
+            "cmd": "aibot_upload_media_init",
+            "headers": {"req_id": self._gen_req_id()},
+            "body": {
+                "type": media_type,
+                "filename": filename,
+                "total_size": file_size,
+                "total_chunks": total_chunks,
+                "md5": md5_hex,
+            },
+        }, timeout=15)
+
+        if init_resp.get("errcode") != 0:
+            logger.error(f"[WecomBot] Upload init failed: {init_resp}")
+            return ""
+
+        upload_id = init_resp.get("body", {}).get("upload_id")
+        if not upload_id:
+            logger.error("[WecomBot] Failed to get upload_id")
+            return ""
+
+        # 2. Upload chunks
+        with open(file_path, "rb") as f:
+            for idx in range(total_chunks):
+                chunk = f.read(MEDIA_CHUNK_SIZE)
+                b64_data = base64.b64encode(chunk).decode("utf-8")
+                chunk_resp = self._send_and_wait({
+                    "cmd": "aibot_upload_media_chunk",
+                    "headers": {"req_id": self._gen_req_id()},
+                    "body": {
+                        "upload_id": upload_id,
+                        "chunk_index": idx,
+                        "base64_data": b64_data,
+                    },
+                }, timeout=30)
+                if chunk_resp.get("errcode") != 0:
+                    logger.error(f"[WecomBot] Chunk {idx} upload failed: {chunk_resp}")
+                    return ""
+
+        # 3. Finish upload
+        finish_resp = self._send_and_wait({
+            "cmd": "aibot_upload_media_finish",
+            "headers": {"req_id": self._gen_req_id()},
+            "body": {"upload_id": upload_id},
+        }, timeout=30)
+
+        if finish_resp.get("errcode") != 0:
+            logger.error(f"[WecomBot] Upload finish failed: {finish_resp}")
+            return ""
+
+        media_id = finish_resp.get("body", {}).get("media_id", "")
+        if media_id:
+            logger.info(f"[WecomBot] Media uploaded: media_id={media_id}")
+        else:
+            logger.error("[WecomBot] Failed to get media_id from finish response")
+        return media_id
--- a/channel/wecom_bot/wecom_bot_message.py
+++ b/channel/wecom_bot/wecom_bot_message.py
@@ -0,0 +1,216 @@
+import os
+import re
+import base64
+import requests
+
+from bridge.context import ContextType
+from channel.chat_message import ChatMessage
+from common.log import logger
+from common.utils import expand_path
+from config import conf
+from Crypto.Cipher import AES
+
+
+MAGIC_SIGNATURES = [
+    (b"%PDF", ".pdf"),
+    (b"\x89PNG\r\n\x1a\n", ".png"),
+    (b"\xff\xd8\xff", ".jpg"),
+    (b"GIF87a", ".gif"),
+    (b"GIF89a", ".gif"),
+    (b"RIFF", ".webp"),  # RIFF....WEBP, further checked below
+    (b"PK\x03\x04", ".zip"),  # zip / docx / xlsx / pptx
+    (b"\x1f\x8b", ".gz"),
+    (b"Rar!\x1a\x07", ".rar"),
+    (b"7z\xbc\xaf\x27\x1c", ".7z"),
+    (b"\x00\x00\x00", ".mp4"),  # ftyp box, further checked below
+    (b"#!AMR", ".amr"),
+]
+
+OFFICE_ZIP_MARKERS = {
+    b"word/": ".docx",
+    b"xl/": ".xlsx",
+    b"ppt/": ".pptx",
+}
+
+
+def _guess_ext_from_bytes(data: bytes) -> str:
+    """Guess file extension from file content magic bytes."""
+    if not data or len(data) < 8:
+        return ""
+    for sig, ext in MAGIC_SIGNATURES:
+        if data[:len(sig)] == sig:
+            if ext == ".webp" and data[8:12] != b"WEBP":
+                continue
+            if ext == ".mp4":
+                if b"ftyp" not in data[4:12]:
+                    continue
+            if ext == ".zip":
+                for marker, office_ext in OFFICE_ZIP_MARKERS.items():
+                    if marker in data[:2000]:
+                        return office_ext
+                return ".zip"
+            return ext
+    return ""
+
+
+def _decrypt_media(url: str, aeskey: str) -> bytes:
+    """
+    Download and decrypt AES-256-CBC encrypted media from wecom bot.
+    Returns decrypted bytes.
+    """
+    resp = requests.get(url, timeout=30)
+    resp.raise_for_status()
+    encrypted = resp.content
+
+    key = base64.b64decode(aeskey + "=" * (-len(aeskey) % 4))
+    if len(key) != 32:
+        raise ValueError(f"Invalid AES key length: {len(key)}, expected 32")
+
+    iv = key[:16]
+    cipher = AES.new(key, AES.MODE_CBC, iv)
+    decrypted = cipher.decrypt(encrypted)
+
+    pad_len = decrypted[-1]
+    if pad_len > 32:
+        raise ValueError(f"Invalid PKCS7 padding length: {pad_len}")
+    return decrypted[:-pad_len]
+
+
+def _get_tmp_dir() -> str:
+    """Return the workspace tmp directory (absolute path), creating it if needed."""
+    ws_root = expand_path(conf().get("agent_workspace", "~/cow"))
+    tmp_dir = os.path.join(ws_root, "tmp")
+    os.makedirs(tmp_dir, exist_ok=True)
+    return tmp_dir
+
+
+class WecomBotMessage(ChatMessage):
+    """Message wrapper for wecom bot (websocket long-connection mode)."""
+
+    def __init__(self, msg_body: dict, is_group: bool = False):
+        super().__init__(msg_body)
+        self.msg_id = msg_body.get("msgid")
+        self.create_time = msg_body.get("create_time")
+        self.is_group = is_group
+
+        msg_type = msg_body.get("msgtype")
+        from_userid = msg_body.get("from", {}).get("userid", "")
+        chat_id = msg_body.get("chatid", "")
+        bot_id = msg_body.get("aibotid", "")
+
+        if msg_type == "text":
+            self.ctype = ContextType.TEXT
+            content = msg_body.get("text", {}).get("content", "")
+            if is_group:
+                content = re.sub(r"@\S+\s*", "", content).strip()
+            self.content = content
+
+        elif msg_type == "voice":
+            self.ctype = ContextType.TEXT
+            self.content = msg_body.get("voice", {}).get("content", "")
+
+        elif msg_type == "image":
+            self.ctype = ContextType.IMAGE
+            image_info = msg_body.get("image", {})
+            image_url = image_info.get("url", "")
+            aeskey = image_info.get("aeskey", "")
+            tmp_dir = _get_tmp_dir()
+            image_path = os.path.join(tmp_dir, f"wecom_{self.msg_id}.png")
+
+            try:
+                data = _decrypt_media(image_url, aeskey)
+                with open(image_path, "wb") as f:
+                    f.write(data)
+                self.content = image_path
+                self.image_path = image_path
+                logger.info(f"[WecomBot] Image downloaded: {image_path}")
+            except Exception as e:
+                logger.error(f"[WecomBot] Failed to download image: {e}")
+                self.content = "[Image download failed]"
+                self.image_path = None
+
+        elif msg_type == "mixed":
+            self.ctype = ContextType.TEXT
+            text_parts = []
+            image_paths = []
+            mixed_items = msg_body.get("mixed", {}).get("msg_item", [])
+            tmp_dir = _get_tmp_dir()
+
+            for idx, item in enumerate(mixed_items):
+                item_type = item.get("msgtype")
+                if item_type == "text":
+                    txt = item.get("text", {}).get("content", "")
+                    if is_group:
+                        txt = re.sub(r"@\S+\s*", "", txt).strip()
+                    if txt:
+                        text_parts.append(txt)
+                elif item_type == "image":
+                    img_info = item.get("image", {})
+                    img_url = img_info.get("url", "")
+                    img_aeskey = img_info.get("aeskey", "")
+                    img_path = os.path.join(tmp_dir, f"wecom_{self.msg_id}_{idx}.png")
+                    try:
+                        img_data = _decrypt_media(img_url, img_aeskey)
+                        with open(img_path, "wb") as f:
+                            f.write(img_data)
+                        image_paths.append(img_path)
+                    except Exception as e:
+                        logger.error(f"[WecomBot] Failed to download mixed image: {e}")
+
+            content_parts = text_parts[:]
+            for p in image_paths:
+                content_parts.append(f"[图片: {p}]")
+            self.content = "\n".join(content_parts) if content_parts else "[Mixed message]"
+
+        elif msg_type == "file":
+            self.ctype = ContextType.FILE
+            file_info = msg_body.get("file", {})
+            file_url = file_info.get("url", "")
+            aeskey = file_info.get("aeskey", "")
+            tmp_dir = _get_tmp_dir()
+            base_path = os.path.join(tmp_dir, f"wecom_{self.msg_id}")
+            self.content = base_path
+
+            def _download_file():
+                try:
+                    data = _decrypt_media(file_url, aeskey)
+                    ext = _guess_ext_from_bytes(data)
+                    final_path = base_path + ext
+                    with open(final_path, "wb") as f:
+                        f.write(data)
+                    self.content = final_path
+                    logger.info(f"[WecomBot] File downloaded: {final_path}")
+                except Exception as e:
+                    logger.error(f"[WecomBot] Failed to download file: {e}")
+            self._prepare_fn = _download_file
+
+        elif msg_type == "video":
+            self.ctype = ContextType.FILE
+            video_info = msg_body.get("video", {})
+            video_url = video_info.get("url", "")
+            aeskey = video_info.get("aeskey", "")
+            tmp_dir = _get_tmp_dir()
+            self.content = os.path.join(tmp_dir, f"wecom_{self.msg_id}.mp4")
+
+            def _download_video():
+                try:
+                    data = _decrypt_media(video_url, aeskey)
+                    with open(self.content, "wb") as f:
+                        f.write(data)
+                    logger.info(f"[WecomBot] Video downloaded: {self.content}")
+                except Exception as e:
+                    logger.error(f"[WecomBot] Failed to download video: {e}")
+            self._prepare_fn = _download_video
+
+        else:
+            raise NotImplementedError(f"Unsupported message type: {msg_type}")
+
+        self.from_user_id = from_userid
+        self.to_user_id = bot_id
+        if is_group:
+            self.other_user_id = chat_id
+            self.actual_user_id = from_userid
+            self.actual_user_nickname = from_userid
+        else:
+            self.other_user_id = from_userid
+            self.actual_user_id = from_userid
--- a/channel/wework/run.py
+++ b/channel/wework/run.py
@@ -1,17 +0,0 @@
-import os
-import time
-os.environ['ntwork_LOG'] = "ERROR"
-import ntwork
-
-wework = ntwork.WeWork()
-
-
-def forever():
-    try:
-        while True:
-            time.sleep(0.1)
-    except KeyboardInterrupt:
-        ntwork.exit_()
-        os._exit(0)
-
-
--- a/channel/wework/wework_channel.py
+++ b/channel/wework/wework_channel.py
@@ -1,326 +0,0 @@
-import io
-import os
-import random
-import tempfile
-import threading
-os.environ['ntwork_LOG'] = "ERROR"
-import ntwork
-import requests
-import uuid
-
-from bridge.context import *
-from bridge.reply import *
-from channel.chat_channel import ChatChannel
-from channel.wework.wework_message import *
-from channel.wework.wework_message import WeworkMessage
-from common.singleton import singleton
-from common.log import logger
-from common.time_check import time_checker
-from common.utils import compress_imgfile, fsize
-from config import conf
-from channel.wework.run import wework
-from channel.wework import run
-from PIL import Image
-
-
-def get_wxid_by_name(room_members, group_wxid, name):
-    if group_wxid in room_members:
-        for member in room_members[group_wxid]['member_list']:
-            if member['room_nickname'] == name or member['username'] == name:
-                return member['user_id']
-    return None  # 如果没有找到对应的group_wxid或name，则返回None
-
-
-def download_and_compress_image(url, filename, quality=30):
-    # 确定保存图片的目录
-    directory = os.path.join(os.getcwd(), "tmp")
-    # 如果目录不存在，则创建目录
-    if not os.path.exists(directory):
-        os.makedirs(directory)
-
-    # 下载图片
-    pic_res = requests.get(url, stream=True)
-    image_storage = io.BytesIO()
-    for block in pic_res.iter_content(1024):
-        image_storage.write(block)
-
-    # 检查图片大小并可能进行压缩
-    sz = fsize(image_storage)
-    if sz >= 10 * 1024 * 1024:  # 如果图片大于 10 MB
-        logger.info("[wework] image too large, ready to compress, sz={}".format(sz))
-        image_storage = compress_imgfile(image_storage, 10 * 1024 * 1024 - 1)
-        logger.info("[wework] image compressed, sz={}".format(fsize(image_storage)))
-
-    # 将内存缓冲区的指针重置到起始位置
-    image_storage.seek(0)
-
-    # 读取并保存图片
-    image = Image.open(image_storage)
-    image_path = os.path.join(directory, f"{filename}.png")
-    image.save(image_path, "png")
-
-    return image_path
-
-
-def download_video(url, filename):
-    # 确定保存视频的目录
-    directory = os.path.join(os.getcwd(), "tmp")
-    # 如果目录不存在，则创建目录
-    if not os.path.exists(directory):
-        os.makedirs(directory)
-
-    # 下载视频
-    response = requests.get(url, stream=True)
-    total_size = 0
-
-    video_path = os.path.join(directory, f"{filename}.mp4")
-
-    with open(video_path, 'wb') as f:
-        for block in response.iter_content(1024):
-            total_size += len(block)
-
-            # 如果视频的总大小超过30MB (30 * 1024 * 1024 bytes)，则停止下载并返回
-            if total_size > 30 * 1024 * 1024:
-                logger.info("[WX] Video is larger than 30MB, skipping...")
-                return None
-
-            f.write(block)
-
-    return video_path
-
-
-def create_message(wework_instance, message, is_group):
-    logger.debug(f"正在为{'群聊' if is_group else '单聊'}创建 WeworkMessage")
-    cmsg = WeworkMessage(message, wework=wework_instance, is_group=is_group)
-    logger.debug(f"cmsg:{cmsg}")
-    return cmsg
-
-
-def handle_message(cmsg, is_group):
-    logger.debug(f"准备用 WeworkChannel 处理{'群聊' if is_group else '单聊'}消息")
-    if is_group:
-        WeworkChannel().handle_group(cmsg)
-    else:
-        WeworkChannel().handle_single(cmsg)
-    logger.debug(f"已用 WeworkChannel 处理完{'群聊' if is_group else '单聊'}消息")
-
-
-def _check(func):
-    def wrapper(self, cmsg: ChatMessage):
-        msgId = cmsg.msg_id
-        create_time = cmsg.create_time  # 消息时间戳
-        if create_time is None:
-            return func(self, cmsg)
-        if int(create_time) < int(time.time()) - 60:  # 跳过1分钟前的历史消息
-            logger.debug("[WX]history message {} skipped".format(msgId))
-            return
-        return func(self, cmsg)
-
-    return wrapper
-
-
-@wework.msg_register(
-    [ntwork.MT_RECV_TEXT_MSG, ntwork.MT_RECV_IMAGE_MSG, 11072, ntwork.MT_RECV_LINK_CARD_MSG,ntwork.MT_RECV_FILE_MSG, ntwork.MT_RECV_VOICE_MSG])
-def all_msg_handler(wework_instance: ntwork.WeWork, message):
-    logger.debug(f"收到消息: {message}")
-    if 'data' in message:
-        # 首先查找conversation_id，如果没有找到，则查找room_conversation_id
-        conversation_id = message['data'].get('conversation_id', message['data'].get('room_conversation_id'))
-        if conversation_id is not None:
-            is_group = "R:" in conversation_id
-            try:
-                cmsg = create_message(wework_instance=wework_instance, message=message, is_group=is_group)
-            except NotImplementedError as e:
-                logger.error(f"[WX]{message.get('MsgId', 'unknown')} 跳过: {e}")
-                return None
-            delay = random.randint(1, 2)
-            timer = threading.Timer(delay, handle_message, args=(cmsg, is_group))
-            timer.start()
-        else:
-            logger.debug("消息数据中无 conversation_id")
-            return None
-    return None
-
-
-def accept_friend_with_retries(wework_instance, user_id, corp_id):
-    result = wework_instance.accept_friend(user_id, corp_id)
-    logger.debug(f'result:{result}')
-
-
-# @wework.msg_register(ntwork.MT_RECV_FRIEND_MSG)
-# def friend(wework_instance: ntwork.WeWork, message):
-#     data = message["data"]
-#     user_id = data["user_id"]
-#     corp_id = data["corp_id"]
-#     logger.info(f"接收到好友请求，消息内容：{data}")
-#     delay = random.randint(1, 180)
-#     threading.Timer(delay, accept_friend_with_retries, args=(wework_instance, user_id, corp_id)).start()
-#
-#     return None
-
-
-def get_with_retry(get_func, max_retries=5, delay=5):
-    retries = 0
-    result = None
-    while retries < max_retries:
-        result = get_func()
-        if result:
-            break
-        logger.warning(f"获取数据失败，重试第{retries + 1}次······")
-        retries += 1
-        time.sleep(delay)  # 等待一段时间后重试
-    return result
-
-
-@singleton
-class WeworkChannel(ChatChannel):
-    NOT_SUPPORT_REPLYTYPE = []
-
-    def __init__(self):
-        super().__init__()
-
-    def startup(self):
-        smart = conf().get("wework_smart", True)
-        wework.open(smart)
-        logger.info("等待登录······")
-        wework.wait_login()
-        login_info = wework.get_login_info()
-        self.user_id = login_info['user_id']
-        self.name = login_info['nickname']
-        logger.info(f"登录信息:>>>user_id:{self.user_id}>>>>>>>>name:{self.name}")
-        logger.info("静默延迟60s，等待客户端刷新数据，请勿进行任何操作······")
-        time.sleep(60)
-        contacts = get_with_retry(wework.get_external_contacts)
-        rooms = get_with_retry(wework.get_rooms)
-        directory = os.path.join(os.getcwd(), "tmp")
-        if not contacts or not rooms:
-            logger.error("获取contacts或rooms失败，程序退出")
-            ntwork.exit_()
-            os.exit(0)
-        if not os.path.exists(directory):
-            os.makedirs(directory)
-        # 将contacts保存到json文件中
-        with open(os.path.join(directory, 'wework_contacts.json'), 'w', encoding='utf-8') as f:
-            json.dump(contacts, f, ensure_ascii=False, indent=4)
-        with open(os.path.join(directory, 'wework_rooms.json'), 'w', encoding='utf-8') as f:
-            json.dump(rooms, f, ensure_ascii=False, indent=4)
-        # 创建一个空字典来保存结果
-        result = {}
-
-        # 遍历列表中的每个字典
-        for room in rooms['room_list']:
-            # 获取聊天室ID
-            room_wxid = room['conversation_id']
-
-            # 获取聊天室成员
-            room_members = wework.get_room_members(room_wxid)
-
-            # 将聊天室成员保存到结果字典中
-            result[room_wxid] = room_members
-
-        # 将结果保存到json文件中
-        with open(os.path.join(directory, 'wework_room_members.json'), 'w', encoding='utf-8') as f:
-            json.dump(result, f, ensure_ascii=False, indent=4)
-        logger.info("wework程序初始化完成········")
-        run.forever()
-
-    @time_checker
-    @_check
-    def handle_single(self, cmsg: ChatMessage):
-        if cmsg.from_user_id == cmsg.to_user_id:
-            # ignore self reply
-            return
-        if cmsg.ctype == ContextType.VOICE:
-            if not conf().get("speech_recognition"):
-                return
-            logger.debug("[WX]receive voice msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.IMAGE:
-            logger.debug("[WX]receive image msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.PATPAT:
-            logger.debug("[WX]receive patpat msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.TEXT:
-            logger.debug("[WX]receive text msg: {}, cmsg={}".format(json.dumps(cmsg._rawmsg, ensure_ascii=False), cmsg))
-        else:
-            logger.debug("[WX]receive msg: {}, cmsg={}".format(cmsg.content, cmsg))
-        context = self._compose_context(cmsg.ctype, cmsg.content, isgroup=False, msg=cmsg)
-        if context:
-            self.produce(context)
-
-    @time_checker
-    @_check
-    def handle_group(self, cmsg: ChatMessage):
-        if cmsg.ctype == ContextType.VOICE:
-            if not conf().get("speech_recognition"):
-                return
-            logger.debug("[WX]receive voice for group msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.IMAGE:
-            logger.debug("[WX]receive image for group msg: {}".format(cmsg.content))
-        elif cmsg.ctype in [ContextType.JOIN_GROUP, ContextType.PATPAT]:
-            logger.debug("[WX]receive note msg: {}".format(cmsg.content))
-        elif cmsg.ctype == ContextType.TEXT:
-            pass
-        else:
-            logger.debug("[WX]receive group msg: {}".format(cmsg.content))
-        context = self._compose_context(cmsg.ctype, cmsg.content, isgroup=True, msg=cmsg)
-        if context:
-            self.produce(context)
-
-    # 统一的发送函数，每个Channel自行实现，根据reply的type字段发送不同类型的消息
-    def send(self, reply: Reply, context: Context):
-        logger.debug(f"context: {context}")
-        receiver = context["receiver"]
-        actual_user_id = context["msg"].actual_user_id
-        if reply.type == ReplyType.TEXT or reply.type == ReplyType.TEXT_:
-            match = re.search(r"^@(.*?)\n", reply.content)
-            logger.debug(f"match: {match}")
-            if match:
-                new_content = re.sub(r"^@(.*?)\n", "\n", reply.content)
-                at_list = [actual_user_id]
-                logger.debug(f"new_content: {new_content}")
-                wework.send_room_at_msg(receiver, new_content, at_list)
-            else:
-                wework.send_text(receiver, reply.content)
-            logger.info("[WX] sendMsg={}, receiver={}".format(reply, receiver))
-        elif reply.type == ReplyType.ERROR or reply.type == ReplyType.INFO:
-            wework.send_text(receiver, reply.content)
-            logger.info("[WX] sendMsg={}, receiver={}".format(reply, receiver))
-        elif reply.type == ReplyType.IMAGE:  # 从文件读取图片
-            image_storage = reply.content
-            image_storage.seek(0)
-            # Read data from image_storage
-            data = image_storage.read()
-            # Create a temporary file
-            with tempfile.NamedTemporaryFile(delete=False) as temp:
-                temp_path = temp.name
-                temp.write(data)
-            # Send the image
-            wework.send_image(receiver, temp_path)
-            logger.info("[WX] sendImage, receiver={}".format(receiver))
-            # Remove the temporary file
-            os.remove(temp_path)
-        elif reply.type == ReplyType.IMAGE_URL:  # 从网络下载图片
-            img_url = reply.content
-            filename = str(uuid.uuid4())
-
-            # 调用你的函数，下载图片并保存为本地文件
-            image_path = download_and_compress_image(img_url, filename)
-
-            wework.send_image(receiver, file_path=image_path)
-            logger.info("[WX] sendImage url={}, receiver={}".format(img_url, receiver))
-        elif reply.type == ReplyType.VIDEO_URL:
-            video_url = reply.content
-            filename = str(uuid.uuid4())
-            video_path = download_video(video_url, filename)
-
-            if video_path is None:
-                # 如果视频太大，下载可能会被跳过，此时 video_path 将为 None
-                wework.send_text(receiver, "抱歉，视频太大了！！！")
-            else:
-                wework.send_video(receiver, video_path)
-            logger.info("[WX] sendVideo, receiver={}".format(receiver))
-        elif reply.type == ReplyType.VOICE:
-            current_dir = os.getcwd()
-            voice_file = reply.content.split("/")[-1]
-            reply.content = os.path.join(current_dir, "tmp", voice_file)
-            wework.send_file(receiver, reply.content)
-            logger.info("[WX] sendFile={}, receiver={}".format(reply.content, receiver))
--- a/channel/wework/wework_message.py
+++ b/channel/wework/wework_message.py
@@ -1,227 +0,0 @@
-import datetime
-import json
-import os
-import re
-import time
-import pilk
-
-from bridge.context import ContextType
-from channel.chat_message import ChatMessage
-from common.log import logger
-from ntwork.const import send_type
-
-
-def get_with_retry(get_func, max_retries=5, delay=5):
-    retries = 0
-    result = None
-    while retries < max_retries:
-        result = get_func()
-        if result:
-            break
-        logger.warning(f"获取数据失败，重试第{retries + 1}次······")
-        retries += 1
-        time.sleep(delay)  # 等待一段时间后重试
-    return result
-
-
-def get_room_info(wework, conversation_id):
-    logger.debug(f"传入的 conversation_id: {conversation_id}")
-    rooms = wework.get_rooms()
-    if not rooms or 'room_list' not in rooms:
-        logger.error(f"获取群聊信息失败: {rooms}")
-        return None
-    time.sleep(1)
-    logger.debug(f"获取到的群聊信息: {rooms}")
-    for room in rooms['room_list']:
-        if room['conversation_id'] == conversation_id:
-            return room
-    return None
-
-
-def cdn_download(wework, message, file_name):
-    data = message["data"]
-    aes_key = data["cdn"]["aes_key"]
-    file_size = data["cdn"]["size"]
-
-    # 获取当前工作目录，然后与文件名拼接得到保存路径
-    current_dir = os.getcwd()
-    save_path = os.path.join(current_dir, "tmp", file_name)
-
-    # 下载保存图片到本地
-    if "url" in data["cdn"].keys() and "auth_key" in data["cdn"].keys():
-        url = data["cdn"]["url"]
-        auth_key = data["cdn"]["auth_key"]
-        # result = wework.wx_cdn_download(url, auth_key, aes_key, file_size, save_path)  # ntwork库本身接口有问题，缺失了aes_key这个参数
-        """
-        下载wx类型的cdn文件，以https开头
-        """
-        data = {
-            'url': url,
-            'auth_key': auth_key,
-            'aes_key': aes_key,
-            'size': file_size,
-            'save_path': save_path
-        }
-        result = wework._WeWork__send_sync(send_type.MT_WXCDN_DOWNLOAD_MSG, data)  # 直接用wx_cdn_download的接口内部实现来调用
-    elif "file_id" in data["cdn"].keys():
-        if message["type"] == 11042:
-            file_type = 2
-        elif message["type"] == 11045:
-            file_type = 5
-        file_id = data["cdn"]["file_id"]
-        result = wework.c2c_cdn_download(file_id, aes_key, file_size, file_type, save_path)
-    else:
-        logger.error(f"something is wrong, data: {data}")
-        return
-
-    # 输出下载结果
-    logger.debug(f"result: {result}")
-
-
-def c2c_download_and_convert(wework, message, file_name):
-    data = message["data"]
-    aes_key = data["cdn"]["aes_key"]
-    file_size = data["cdn"]["size"]
-    file_type = 5
-    file_id = data["cdn"]["file_id"]
-
-    current_dir = os.getcwd()
-    save_path = os.path.join(current_dir, "tmp", file_name)
-    result = wework.c2c_cdn_download(file_id, aes_key, file_size, file_type, save_path)
-    logger.debug(result)
-
-    # 在下载完SILK文件之后，立即将其转换为WAV文件
-    base_name, _ = os.path.splitext(save_path)
-    wav_file = base_name + ".wav"
-    pilk.silk_to_wav(save_path, wav_file, rate=24000)
-
-    # 删除SILK文件
-    try:
-        os.remove(save_path)
-    except Exception as e:
-        pass
-
-
-class WeworkMessage(ChatMessage):
-    def __init__(self, wework_msg, wework, is_group=False):
-        try:
-            super().__init__(wework_msg)
-            self.msg_id = wework_msg['data'].get('conversation_id', wework_msg['data'].get('room_conversation_id'))
-            # 使用.get()防止 'send_time' 键不存在时抛出错误
-            self.create_time = wework_msg['data'].get("send_time")
-            self.is_group = is_group
-            self.wework = wework
-
-            if wework_msg["type"] == 11041:  # 文本消息类型
-                if any(substring in wework_msg['data']['content'] for substring in ("该消息类型暂不能展示", "不支持的消息类型")):
-                    return
-                self.ctype = ContextType.TEXT
-                self.content = wework_msg['data']['content']
-            elif wework_msg["type"] == 11044:  # 语音消息类型，需要缓存文件
-                file_name = datetime.datetime.now().strftime('%Y%m%d%H%M%S') + ".silk"
-                base_name, _ = os.path.splitext(file_name)
-                file_name_2 = base_name + ".wav"
-                current_dir = os.getcwd()
-                self.ctype = ContextType.VOICE
-                self.content = os.path.join(current_dir, "tmp", file_name_2)
-                self._prepare_fn = lambda: c2c_download_and_convert(wework, wework_msg, file_name)
-            elif wework_msg["type"] == 11042:  # 图片消息类型，需要下载文件
-                file_name = datetime.datetime.now().strftime('%Y%m%d%H%M%S') + ".jpg"
-                current_dir = os.getcwd()
-                self.ctype = ContextType.IMAGE
-                self.content = os.path.join(current_dir, "tmp", file_name)
-                self._prepare_fn = lambda: cdn_download(wework, wework_msg, file_name)
-            elif wework_msg["type"] == 11045:  # 文件消息
-                print("文件消息")
-                print(wework_msg)
-                file_name = datetime.datetime.now().strftime('%Y%m%d%H%M%S')
-                file_name = file_name + wework_msg['data']['cdn']['file_name']
-                current_dir = os.getcwd()
-                self.ctype = ContextType.FILE
-                self.content = os.path.join(current_dir, "tmp", file_name)
-                self._prepare_fn = lambda: cdn_download(wework, wework_msg, file_name)
-            elif wework_msg["type"] == 11047:  # 链接消息
-                self.ctype = ContextType.SHARING
-                self.content = wework_msg['data']['url']
-            elif wework_msg["type"] == 11072:  # 新成员入群通知
-                self.ctype = ContextType.JOIN_GROUP
-                member_list = wework_msg['data']['member_list']
-                self.actual_user_nickname = member_list[0]['name']
-                self.actual_user_id = member_list[0]['user_id']
-                self.content = f"{self.actual_user_nickname}加入了群聊！"
-                directory = os.path.join(os.getcwd(), "tmp")
-                rooms = get_with_retry(wework.get_rooms)
-                if not rooms:
-                    logger.error("更新群信息失败···")
-                else:
-                    result = {}
-                    for room in rooms['room_list']:
-                        # 获取聊天室ID
-                        room_wxid = room['conversation_id']
-
-                        # 获取聊天室成员
-                        room_members = wework.get_room_members(room_wxid)
-
-                        # 将聊天室成员保存到结果字典中
-                        result[room_wxid] = room_members
-                    with open(os.path.join(directory, 'wework_room_members.json'), 'w', encoding='utf-8') as f:
-                        json.dump(result, f, ensure_ascii=False, indent=4)
-                    logger.info("有新成员加入，已自动更新群成员列表缓存！")
-            else:
-                raise NotImplementedError(
-                    "Unsupported message type: Type:{} MsgType:{}".format(wework_msg["type"], wework_msg["MsgType"]))
-
-            data = wework_msg['data']
-            login_info = self.wework.get_login_info()
-            logger.debug(f"login_info: {login_info}")
-            nickname = f"{login_info['username']}({login_info['nickname']})" if login_info['nickname'] else login_info['username']
-            user_id = login_info['user_id']
-
-            sender_id = data.get('sender')
-            conversation_id = data.get('conversation_id')
-            sender_name = data.get("sender_name")
-
-            self.from_user_id = user_id if sender_id == user_id else conversation_id
-            self.from_user_nickname = nickname if sender_id == user_id else sender_name
-            self.to_user_id = user_id
-            self.to_user_nickname = nickname
-            self.other_user_nickname = sender_name
-            self.other_user_id = conversation_id
-
-            if self.is_group:
-                conversation_id = data.get('conversation_id') or data.get('room_conversation_id')
-                self.other_user_id = conversation_id
-                if conversation_id:
-                    room_info = get_room_info(wework=wework, conversation_id=conversation_id)
-                    self.other_user_nickname = room_info.get('nickname', None) if room_info else None
-                    self.from_user_nickname = room_info.get('nickname', None) if room_info else None
-                    at_list = data.get('at_list', [])
-                    tmp_list = []
-                    for at in at_list:
-                        tmp_list.append(at['nickname'])
-                    at_list = tmp_list
-                    logger.debug(f"at_list: {at_list}")
-                    logger.debug(f"nickname: {nickname}")
-                    self.is_at = False
-                    if nickname in at_list or login_info['nickname'] in at_list or login_info['username'] in at_list:
-                        self.is_at = True
-                    self.at_list = at_list
-
-                    # 检查消息内容是否包含@用户名。处理复制粘贴的消息，这类消息可能不会触发@通知，但内容中可能包含 "@用户名"。
-                    content = data.get('content', '')
-                    name = nickname
-                    pattern = f"@{re.escape(name)}(\u2005|\u0020)"
-                    if re.search(pattern, content):
-                        logger.debug(f"Wechaty message {self.msg_id} includes at")
-                        self.is_at = True
-
-                    if not self.actual_user_id:
-                        self.actual_user_id = data.get("sender")
-                    self.actual_user_nickname = sender_name if self.ctype != ContextType.JOIN_GROUP else self.actual_user_nickname
-                else:
-                    logger.error("群聊消息中没有找到 conversation_id 或 room_conversation_id")
-
-            logger.debug(f"WeworkMessage has been successfully instantiated with message id: {self.msg_id}")
-        except Exception as e:
-            logger.error(f"在 WeworkMessage 的初始化过程中出现错误：{e}")
-            raise e
--- a/common/cloud_client.py
+++ b/common/cloud_client.py
@@ -0,0 +1,687 @@
+"""
+Cloud management client for connecting to the LinkAI control console.
+
+Handles remote configuration sync, message push, and skill management
+via the LinkAI socket protocol.
+"""
+
+from bridge.context import Context, ContextType
+from bridge.reply import Reply, ReplyType
+from common.log import logger
+from linkai import LinkAIClient, PushMsg
+from config import conf, pconf, plugin_config, available_setting, write_plugin_config, get_root
+from plugins import PluginManager
+import threading
+import time
+import json
+import os
+
+
+chat_client: LinkAIClient
+
+
+CHANNEL_ACTIONS = {"channel_create", "channel_update", "channel_delete"}
+
+# channelType -> config key mapping for app credentials
+CREDENTIAL_MAP = {
+    "feishu":            ("feishu_app_id",          "feishu_app_secret"),
+    "dingtalk":          ("dingtalk_client_id",      "dingtalk_client_secret"),
+    "wecom_bot":         ("wecom_bot_id",            "wecom_bot_secret"),
+    "qq":                ("qq_app_id",               "qq_app_secret"),
+    "wechatmp":          ("wechatmp_app_id",         "wechatmp_app_secret"),
+    "wechatmp_service":  ("wechatmp_app_id",         "wechatmp_app_secret"),
+    "wechatcom_app":     ("wechatcomapp_agent_id",   "wechatcomapp_secret"),
+}
+
+
+class CloudClient(LinkAIClient):
+    def __init__(self, api_key: str, channel, host: str = ""):
+        super().__init__(api_key, host)
+        self.channel = channel
+        self.client_type = channel.channel_type
+        self.channel_mgr = None
+        self._skill_service = None
+        self._memory_service = None
+        self._chat_service = None
+
+    @property
+    def skill_service(self):
+        """Lazy-init SkillService so it is available once SkillManager exists."""
+        if self._skill_service is None:
+            try:
+                from agent.skills.manager import SkillManager
+                from agent.skills.service import SkillService
+                from config import conf
+                from common.utils import expand_path
+                workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
+                manager = SkillManager(custom_dir=os.path.join(workspace_root, "skills"))
+                self._skill_service = SkillService(manager)
+                logger.debug("[CloudClient] SkillService initialised")
+            except Exception as e:
+                logger.error(f"[CloudClient] Failed to init SkillService: {e}")
+        return self._skill_service
+
+    @property
+    def memory_service(self):
+        """Lazy-init MemoryService."""
+        if self._memory_service is None:
+            try:
+                from agent.memory.service import MemoryService
+                from config import conf
+                from common.utils import expand_path
+                workspace_root = expand_path(conf().get("agent_workspace", "~/cow"))
+                self._memory_service = MemoryService(workspace_root)
+                logger.debug("[CloudClient] MemoryService initialised")
+            except Exception as e:
+                logger.error(f"[CloudClient] Failed to init MemoryService: {e}")
+        return self._memory_service
+
+    @property
+    def chat_service(self):
+        """Lazy-init ChatService (requires AgentBridge via Bridge singleton)."""
+        if self._chat_service is None:
+            try:
+                from agent.chat.service import ChatService
+                from bridge.bridge import Bridge
+                agent_bridge = Bridge().get_agent_bridge()
+                self._chat_service = ChatService(agent_bridge)
+                logger.debug("[CloudClient] ChatService initialised")
+            except Exception as e:
+                logger.error(f"[CloudClient] Failed to init ChatService: {e}")
+        return self._chat_service
+
+    # ------------------------------------------------------------------
+    # message push callback
+    # ------------------------------------------------------------------
+    def on_message(self, push_msg: PushMsg):
+        session_id = push_msg.session_id
+        msg_content = push_msg.msg_content
+        logger.info(f"receive msg push, session_id={session_id}, msg_content={msg_content}")
+        context = Context()
+        context.type = ContextType.TEXT
+        context["receiver"] = session_id
+        context["isgroup"] = push_msg.is_group
+        self.channel.send(Reply(ReplyType.TEXT, content=msg_content), context)
+
+    # ------------------------------------------------------------------
+    # config callback
+    # ------------------------------------------------------------------
+    def on_config(self, config: dict):
+        if not self.client_id:
+            return
+        logger.info(f"[CloudClient] Loading remote config: {config}")
+
+        action = config.get("action")
+        if action in CHANNEL_ACTIONS:
+            self._dispatch_channel_action(action, config.get("data", {}))
+            return
+
+        if config.get("enabled") != "Y":
+            return
+
+        local_config = conf()
+        need_restart_channel = False
+
+        for key in config.keys():
+            if key in available_setting and config.get(key) is not None:
+                local_config[key] = config.get(key)
+
+        # Voice settings
+        reply_voice_mode = config.get("reply_voice_mode")
+        if reply_voice_mode:
+            if reply_voice_mode == "voice_reply_voice":
+                local_config["voice_reply_voice"] = True
+                local_config["always_reply_voice"] = False
+            elif reply_voice_mode == "always_reply_voice":
+                local_config["always_reply_voice"] = True
+                local_config["voice_reply_voice"] = True
+            elif reply_voice_mode == "no_reply_voice":
+                local_config["always_reply_voice"] = False
+                local_config["voice_reply_voice"] = False
+
+        # Model configuration
+        if config.get("model"):
+            local_config["model"] = config.get("model")
+
+        # Channel configuration (legacy single-channel path)
+        if config.get("channelType"):
+            if local_config.get("channel_type") != config.get("channelType"):
+                local_config["channel_type"] = config.get("channelType")
+                need_restart_channel = True
+
+        # Channel-specific app credentials (legacy single-channel path)
+        current_channel_type = local_config.get("channel_type", "")
+        if self._set_channel_credentials(local_config, current_channel_type,
+                                         config.get("app_id"), config.get("app_secret")):
+            need_restart_channel = True
+
+        if config.get("admin_password"):
+            if not pconf("Godcmd"):
+                write_plugin_config({"Godcmd": {"password": config.get("admin_password"), "admin_users": []}})
+            else:
+                pconf("Godcmd")["password"] = config.get("admin_password")
+            PluginManager().instances["GODCMD"].reload()
+
+        if config.get("group_app_map") and pconf("linkai"):
+            local_group_map = {}
+            for mapping in config.get("group_app_map"):
+                local_group_map[mapping.get("group_name")] = mapping.get("app_code")
+            pconf("linkai")["group_app_map"] = local_group_map
+            PluginManager().instances["LINKAI"].reload()
+
+        if config.get("text_to_image") and config.get("text_to_image") == "midjourney" and pconf("linkai"):
+            if pconf("linkai")["midjourney"]:
+                pconf("linkai")["midjourney"]["enabled"] = True
+                pconf("linkai")["midjourney"]["use_image_create_prefix"] = True
+        elif config.get("text_to_image") and config.get("text_to_image") in ["dall-e-2", "dall-e-3"]:
+            if pconf("linkai")["midjourney"]:
+                pconf("linkai")["midjourney"]["use_image_create_prefix"] = False
+
+        self._save_config_to_file(local_config)
+
+        if need_restart_channel:
+            self._restart_channel(local_config.get("channel_type", ""))
+
+    # ------------------------------------------------------------------
+    # channel CRUD operations
+    # ------------------------------------------------------------------
+    def _dispatch_channel_action(self, action: str, data: dict):
+        channel_type = data.get("channelType")
+        if not channel_type:
+            logger.warning(f"[CloudClient] Channel action '{action}' missing channelType, data={data}")
+            return
+        logger.info(f"[CloudClient] Channel action: {action}, channelType={channel_type}")
+
+        if action == "channel_create":
+            self._handle_channel_create(channel_type, data)
+        elif action == "channel_update":
+            self._handle_channel_update(channel_type, data)
+        elif action == "channel_delete":
+            self._handle_channel_delete(channel_type, data)
+
+    def _handle_channel_create(self, channel_type: str, data: dict):
+        local_config = conf()
+        self._set_channel_credentials(local_config, channel_type,
+                                      data.get("appId"), data.get("appSecret"))
+        self._add_channel_type(local_config, channel_type)
+        self._save_config_to_file(local_config)
+
+        if self.channel_mgr:
+            threading.Thread(
+                target=self._do_add_channel, args=(channel_type,), daemon=True
+            ).start()
+
+    def _handle_channel_update(self, channel_type: str, data: dict):
+        local_config = conf()
+        enabled = data.get("enabled", "Y")
+
+        self._set_channel_credentials(local_config, channel_type,
+                                      data.get("appId"), data.get("appSecret"))
+        if enabled == "N":
+            self._remove_channel_type(local_config, channel_type)
+        else:
+            # Ensure channel_type is persisted even if this channel was not
+            # previously listed (e.g. update used as implicit create).
+            self._add_channel_type(local_config, channel_type)
+        self._save_config_to_file(local_config)
+
+        if not self.channel_mgr:
+            return
+
+        if enabled == "N":
+            threading.Thread(
+                target=self._do_remove_channel, args=(channel_type,), daemon=True
+            ).start()
+        else:
+            threading.Thread(
+                target=self._do_restart_channel, args=(self.channel_mgr, channel_type), daemon=True
+            ).start()
+
+    def _handle_channel_delete(self, channel_type: str, data: dict):
+        local_config = conf()
+        self._clear_channel_credentials(local_config, channel_type)
+        self._remove_channel_type(local_config, channel_type)
+        self._save_config_to_file(local_config)
+
+        if self.channel_mgr:
+            threading.Thread(
+                target=self._do_remove_channel, args=(channel_type,), daemon=True
+            ).start()
+
+    # ------------------------------------------------------------------
+    # channel credentials helpers
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _set_channel_credentials(local_config: dict, channel_type: str,
+                                 app_id, app_secret) -> bool:
+        """
+        Write app_id / app_secret into the correct config keys for *channel_type*.
+        Also syncs the values to environment variables (upper-cased key) so that
+        skills that rely on env-based checks (e.g. has_env_var) work immediately.
+        Returns True if any value actually changed.
+        """
+        cred = CREDENTIAL_MAP.get(channel_type)
+        if not cred:
+            return False
+        id_key, secret_key = cred
+        changed = False
+        if app_id is not None and local_config.get(id_key) != app_id:
+            local_config[id_key] = app_id
+            os.environ[id_key.upper()] = str(app_id)
+            changed = True
+        if app_secret is not None and local_config.get(secret_key) != app_secret:
+            local_config[secret_key] = app_secret
+            os.environ[secret_key.upper()] = str(app_secret)
+            changed = True
+        if changed:
+            logger.info(f"[CloudClient] Synced {channel_type} credentials to conf and env")
+        return changed
+
+    @staticmethod
+    def _clear_channel_credentials(local_config: dict, channel_type: str):
+        cred = CREDENTIAL_MAP.get(channel_type)
+        if not cred:
+            return
+        id_key, secret_key = cred
+        local_config.pop(id_key, None)
+        local_config.pop(secret_key, None)
+        os.environ.pop(id_key.upper(), None)
+        os.environ.pop(secret_key.upper(), None)
+
+    # ------------------------------------------------------------------
+    # channel_type list helpers
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _parse_channel_types(local_config: dict) -> list:
+        raw = local_config.get("channel_type", "")
+        if isinstance(raw, list):
+            return [ch.strip() for ch in raw if ch.strip()]
+        if isinstance(raw, str):
+            return [ch.strip() for ch in raw.split(",") if ch.strip()]
+        return []
+
+    @staticmethod
+    def _add_channel_type(local_config: dict, channel_type: str):
+        types = CloudClient._parse_channel_types(local_config)
+        if channel_type not in types:
+            types.append(channel_type)
+            local_config["channel_type"] = ", ".join(types)
+
+    @staticmethod
+    def _remove_channel_type(local_config: dict, channel_type: str):
+        types = CloudClient._parse_channel_types(local_config)
+        if channel_type in types:
+            types.remove(channel_type)
+            local_config["channel_type"] = ", ".join(types)
+
+    # ------------------------------------------------------------------
+    # channel manager thread helpers
+    # ------------------------------------------------------------------
+    def _do_add_channel(self, channel_type: str):
+        try:
+            self.channel_mgr.add_channel(channel_type)
+            logger.info(f"[CloudClient] Channel '{channel_type}' added successfully")
+        except Exception as e:
+            logger.error(f"[CloudClient] Failed to add channel '{channel_type}': {e}")
+            self.send_channel_status(channel_type, "error", str(e))
+            return
+        self._report_channel_startup(channel_type)
+
+    def _do_remove_channel(self, channel_type: str):
+        try:
+            self.channel_mgr.remove_channel(channel_type)
+            logger.info(f"[CloudClient] Channel '{channel_type}' removed successfully")
+        except Exception as e:
+            logger.error(f"[CloudClient] Failed to remove channel '{channel_type}': {e}")
+
+    def _report_channel_startup(self, channel_type: str):
+        """Wait for channel startup result and report to cloud."""
+        ch = self.channel_mgr.get_channel(channel_type)
+        if not ch:
+            self.send_channel_status(channel_type, "error", "channel instance not found")
+            return
+        success, error = ch.wait_startup(timeout=3)
+        if success:
+            logger.info(f"[CloudClient] Channel '{channel_type}' connected, reporting status")
+            self.send_channel_status(channel_type, "connected")
+        else:
+            logger.warning(f"[CloudClient] Channel '{channel_type}' startup failed: {error}")
+            self.send_channel_status(channel_type, "error", error)
+
+    # ------------------------------------------------------------------
+    # skill callback
+    # ------------------------------------------------------------------
+    def on_skill(self, data: dict) -> dict:
+        """
+        Handle SKILL messages from the cloud console.
+        Delegates to SkillService.dispatch for the actual operations.
+
+        :param data: message data with 'action', 'clientId', 'payload'
+        :return: response dict
+        """
+        action = data.get("action", "")
+        payload = data.get("payload")
+        logger.info(f"[CloudClient] on_skill: action={action}")
+
+        svc = self.skill_service
+        if svc is None:
+            return {"action": action, "code": 500, "message": "SkillService not available", "payload": None}
+
+        return svc.dispatch(action, payload)
+
+    # ------------------------------------------------------------------
+    # memory callback
+    # ------------------------------------------------------------------
+    def on_memory(self, data: dict) -> dict:
+        """
+        Handle MEMORY messages from the cloud console.
+        Delegates to MemoryService.dispatch for the actual operations.
+
+        :param data: message data with 'action', 'clientId', 'payload'
+        :return: response dict
+        """
+        action = data.get("action", "")
+        payload = data.get("payload")
+        logger.info(f"[CloudClient] on_memory: action={action}")
+
+        svc = self.memory_service
+        if svc is None:
+            return {"action": action, "code": 500, "message": "MemoryService not available", "payload": None}
+
+        return svc.dispatch(action, payload)
+
+    # ------------------------------------------------------------------
+    # chat callback
+    # ------------------------------------------------------------------
+    def on_chat(self, data: dict, send_chunk_fn):
+        """
+        Handle CHAT messages from the cloud console.
+        Runs the agent in streaming mode and sends chunks back via send_chunk_fn.
+
+        :param data: message data with 'action' and 'payload' (query, session_id)
+        :param send_chunk_fn: callable(chunk_data: dict) to send one streaming chunk
+        """
+        payload = data.get("payload", {})
+        query = payload.get("query", "")
+        session_id = payload.get("session_id", "cloud_console")
+        channel_type = payload.get("channel_type", "")
+        if not session_id.startswith("session_"):
+            session_id = f"session_{session_id}"
+        logger.info(f"[CloudClient] on_chat: session={session_id}, channel={channel_type}, query={query[:80]}")
+
+        svc = self.chat_service
+        if svc is None:
+            raise RuntimeError("ChatService not available")
+
+        svc.run(query=query, session_id=session_id, channel_type=channel_type, send_chunk_fn=send_chunk_fn)
+
+    # ------------------------------------------------------------------
+    # history callback
+    # ------------------------------------------------------------------
+    def on_history(self, data: dict) -> dict:
+        """
+        Handle HISTORY messages from the cloud console.
+        Returns paginated conversation history for a session.
+
+        :param data: message data with 'action' and 'payload' (session_id, page, page_size)
+        :return: response dict
+        """
+        action = data.get("action", "query")
+        payload = data.get("payload", {})
+        logger.info(f"[CloudClient] on_history: action={action}")
+
+        if action == "query":
+            return self._query_history(payload)
+
+        return {"action": action, "code": 404, "message": f"unknown action: {action}", "payload": None}
+
+    def _query_history(self, payload: dict) -> dict:
+        """Query paginated conversation history using ConversationStore."""
+        session_id = payload.get("session_id", "")
+        page = int(payload.get("page", 1))
+        page_size = int(payload.get("page_size", 20))
+
+        if not session_id:
+            return {
+                "action": "query",
+                "payload": {"status": "error", "message": "session_id required"},
+            }
+
+        # Web channel stores sessions with a "session_" prefix
+        if not session_id.startswith("session_"):
+            session_id = f"session_{session_id}"
+        logger.info(f"[CloudClient] history query: session={session_id}, page={page}, page_size={page_size}")
+
+        try:
+            from agent.memory.conversation_store import get_conversation_store
+            store = get_conversation_store()
+            result = store.load_history_page(
+                session_id=session_id,
+                page=page,
+                page_size=page_size,
+            )
+            return {
+                "action": "query",
+                "payload": {"status": "success", **result},
+            }
+        except Exception as e:
+            logger.error(f"[CloudClient] History query error: {e}")
+            return {
+                "action": "query",
+                "payload": {"status": "error", "message": str(e)},
+            }
+
+    # ------------------------------------------------------------------
+    # channel restart helpers
+    # ------------------------------------------------------------------
+    def _restart_channel(self, new_channel_type: str):
+        """
+        Restart the channel via ChannelManager when channel type changes.
+        """
+        if self.channel_mgr:
+            logger.info(f"[CloudClient] Restarting channel to '{new_channel_type}'...")
+            threading.Thread(target=self._do_restart_channel, args=(self.channel_mgr, new_channel_type), daemon=True).start()
+        else:
+            logger.warning("[CloudClient] ChannelManager not available, please restart the application manually")
+
+    def _do_restart_channel(self, mgr, new_channel_type: str):
+        """
+        Perform the channel restart in a separate thread to avoid blocking the config callback.
+        """
+        try:
+            mgr.restart(new_channel_type)
+            if mgr.channel:
+                self.channel = mgr.channel
+                self.client_type = mgr.channel.channel_type
+                logger.info(f"[CloudClient] Channel reference updated to '{new_channel_type}'")
+        except Exception as e:
+            logger.error(f"[CloudClient] Channel restart failed: {e}")
+            self.send_channel_status(new_channel_type, "error", str(e))
+            return
+        self._report_channel_startup(new_channel_type)
+
+    # ------------------------------------------------------------------
+    # config persistence
+    # ------------------------------------------------------------------
+    def _save_config_to_file(self, local_config: dict):
+        """
+        Save configuration to config.json file.
+        """
+        try:
+            config_path = os.path.join(get_root(), "config.json")
+            if not os.path.exists(config_path):
+                logger.warning(f"[CloudClient] config.json not found at {config_path}, skip saving")
+                return
+
+            with open(config_path, "r", encoding="utf-8") as f:
+                file_config = json.load(f)
+
+            file_config.update(dict(local_config))
+
+            with open(config_path, "w", encoding="utf-8") as f:
+                json.dump(file_config, f, indent=4, ensure_ascii=False)
+
+            logger.info("[CloudClient] Configuration saved to config.json successfully")
+        except Exception as e:
+            logger.error(f"[CloudClient] Failed to save configuration to config.json: {e}")
+
+
+def get_root_domain(host: str = "") -> str:
+    """Extract root domain from a hostname.
+
+    If *host* is empty, reads CLOUD_HOST env var / cloud_host config.
+    """
+    if not host:
+        host = os.environ.get("CLOUD_HOST") or conf().get("cloud_host", "")
+    if not host:
+        return ""
+    host = host.strip().rstrip("/")
+    if "://" in host:
+        host = host.split("://", 1)[1]
+    host = host.split("/", 1)[0].split(":")[0]
+    parts = host.split(".")
+    if len(parts) >= 2:
+        return ".".join(parts[-2:])
+    return host
+
+
+def get_deployment_id() -> str:
+    """Return cloud deployment id from env var or config."""
+    return os.environ.get("CLOUD_DEPLOYMENT_ID") or conf().get("cloud_deployment_id", "")
+
+
+def get_website_base_url() -> str:
+    """Return the public URL prefix that maps to the workspace websites/ dir.
+
+    Returns empty string when cloud deployment is not configured.
+    """
+    deployment_id = get_deployment_id()
+    if not deployment_id:
+        return ""
+
+    websites_domain = os.environ.get("CLOUD_WEBSITES_DOMAIN") or conf().get("cloud_websites_domain", "")
+    if websites_domain:
+        websites_domain = websites_domain.strip().rstrip("/")
+        return f"https://{websites_domain}/{deployment_id}"
+
+    domain = get_root_domain()
+    if not domain:
+        return ""
+    return f"https://app.{domain}/{deployment_id}"
+
+
+def build_website_prompt(workspace_dir: str) -> list:
+    """Build system prompt lines for cloud website/file sharing rules.
+
+    Returns an empty list when cloud deployment is not configured,
+    so callers can safely do ``lines.extend(build_website_prompt(...))``.
+    """
+    base_url = get_website_base_url()
+    if not base_url:
+        return []
+
+    return [
+        "**文件分享与网页生成规则** (非常重要 — 当前为云部署模式):",
+        "",
+        f"云端已为工作空间的 `websites/` 目录配置好公网路由映射，访问地址前缀为: `{base_url}`",
+        "",
+        "1. **网页/网站**: 编写网页、H5页面等前端代码时，**必须**将文件放到 `websites/` 目录中",
+        f"   - 例如: `websites/index.html` → `{base_url}/index.html`",
+        f"   - 例如: `websites/my-app/index.html` → `{base_url}/my-app/index.html`",
+        "",
+        "2. **生成文件分享** (PPT、PDF、图片、音视频等): 当你为用户生成了需要下载或查看的文件时，**可以**将文件保存到 `websites/` 目录中",
+        f"   - 例如: 生成的PPT保存到 `websites/files/report.pptx` → 下载链接为 `{base_url}/files/report.pptx`",
+        "   - 你仍然可以同时使用 `send` 工具发送文件（在飞书、钉钉等IM渠道中有效），但**必须同时在回复文本中提供下载链接**作为兜底，因为部分渠道（如网页端）无法通过 send 接收本地文件",
+        "",
+        "3. **必须发送链接**: 无论是网页还是文件，生成后**必须将完整的访问/下载链接直接写在回复文本中发送给用户**",
+        "",
+        "4. **文件名和路径尽量使用英文/拼音/数字等**，不要使用中文，避免链接无法访问",
+        "",
+        "5. 建议为每个独立项目在 `websites/` 下创建子目录，保持结构清晰",
+        "",
+    ]
+
+def start(channel, channel_mgr=None):
+    if not get_deployment_id():
+        return
+
+    global chat_client
+    chat_client = CloudClient(api_key=conf().get("linkai_api_key"), host=conf().get("cloud_host", ""), channel=channel)
+    chat_client.channel_mgr = channel_mgr
+    chat_client.config = _build_config()
+    chat_client.start()
+    time.sleep(1.5)
+    if chat_client.client_id:
+        logger.info("[CloudClient] Console: https://link-ai.tech/console/clients")
+        if channel_mgr:
+            channel_mgr.cloud_mode = True
+            threading.Thread(target=_report_existing_channels, args=(chat_client, channel_mgr), daemon=True).start()
+
+
+def _report_existing_channels(client: CloudClient, mgr):
+    """Report status for all channels that were started before cloud client connected."""
+    try:
+        for name, ch in list(mgr._channels.items()):
+            if name == "web":
+                continue
+            ch.cloud_mode = True
+            client._report_channel_startup(name)
+    except Exception as e:
+        logger.warning(f"[CloudClient] Failed to report existing channel status: {e}")
+
+
+def _build_config():
+    local_conf = conf()
+    config = {
+        "linkai_app_code": local_conf.get("linkai_app_code"),
+        "single_chat_prefix": local_conf.get("single_chat_prefix"),
+        "single_chat_reply_prefix": local_conf.get("single_chat_reply_prefix"),
+        "single_chat_reply_suffix": local_conf.get("single_chat_reply_suffix"),
+        "group_chat_prefix": local_conf.get("group_chat_prefix"),
+        "group_chat_reply_prefix": local_conf.get("group_chat_reply_prefix"),
+        "group_chat_reply_suffix": local_conf.get("group_chat_reply_suffix"),
+        "group_name_white_list": local_conf.get("group_name_white_list"),
+        "nick_name_black_list": local_conf.get("nick_name_black_list"),
+        "speech_recognition": "Y" if local_conf.get("speech_recognition") else "N",
+        "text_to_image": local_conf.get("text_to_image"),
+        "image_create_prefix": local_conf.get("image_create_prefix"),
+        "model": local_conf.get("model"),
+        "agent_max_context_turns": local_conf.get("agent_max_context_turns"),
+        "agent_max_context_tokens": local_conf.get("agent_max_context_tokens"),
+        "agent_max_steps": local_conf.get("agent_max_steps"),
+        "channelType": local_conf.get("channel_type"),
+    }
+
+    if local_conf.get("always_reply_voice"):
+        config["reply_voice_mode"] = "always_reply_voice"
+    elif local_conf.get("voice_reply_voice"):
+        config["reply_voice_mode"] = "voice_reply_voice"
+
+    if pconf("linkai"):
+        config["group_app_map"] = pconf("linkai").get("group_app_map")
+
+    if plugin_config.get("Godcmd"):
+        config["admin_password"] = plugin_config.get("Godcmd").get("password")
+
+    # Add channel-specific app credentials
+    current_channel_type = local_conf.get("channel_type", "")
+    if current_channel_type == "feishu":
+        config["app_id"] = local_conf.get("feishu_app_id")
+        config["app_secret"] = local_conf.get("feishu_app_secret")
+    elif current_channel_type == "dingtalk":
+        config["app_id"] = local_conf.get("dingtalk_client_id")
+        config["app_secret"] = local_conf.get("dingtalk_client_secret")
+    elif current_channel_type in ("wechatmp", "wechatmp_service"):
+        config["app_id"] = local_conf.get("wechatmp_app_id")
+        config["app_secret"] = local_conf.get("wechatmp_app_secret")
+    elif current_channel_type == "wecom_bot":
+        config["app_id"] = local_conf.get("wecom_bot_id")
+        config["app_secret"] = local_conf.get("wecom_bot_secret")
+    elif current_channel_type == "qq":
+        config["app_id"] = local_conf.get("qq_app_id")
+        config["app_secret"] = local_conf.get("qq_app_secret")
+    elif current_channel_type == "wechatcom_app":
+        config["app_id"] = local_conf.get("wechatcomapp_agent_id")
+        config["app_secret"] = local_conf.get("wechatcomapp_secret")
+
+    return config
--- a/common/const.py
+++ b/common/const.py
@@ -1,6 +1,7 @@
 # 厂商类型
 OPEN_AI = "openAI"
-CHATGPT = "chatGPT"
+OPENAI = "openai"
+CHATGPT = "chatGPT"  # legacy alias for OPENAI, kept for backward compatibility
 BAIDU = "baidu"
 XUNFEI = "xunfei"
 CHATGPTONAZURE = "chatGPTOnAzure"
@@ -9,9 +10,10 @@ CLAUDEAPI= "claudeAPI"
 QWEN = "qwen"  # 旧版千问接入
 QWEN_DASHSCOPE = "dashscope"  # 新版千问接入(百炼)
 GEMINI = "gemini" 
-ZHIPU_AI = "glm-4"
+ZHIPU_AI = "zhipu"  
 MOONSHOT = "moonshot"
 MiniMax = "minimax"
+DEEPSEEK = "deepseek"
 MODELSCOPE = "modelscope"

 # 模型列表
@@ -25,8 +27,10 @@ CLAUDE_35_SONNET = "claude-3-5-sonnet-latest"  # 带 latest 标签的模型名
 CLAUDE_35_SONNET_1022 = "claude-3-5-sonnet-20241022"  # 带具体日期的模型名称，会固定为该日期发布的模型
 CLAUDE_35_SONNET_0620 = "claude-3-5-sonnet-20240620"
 CLAUDE_4_OPUS = "claude-opus-4-0"
-CLAUDE_4_SONNET = "claude-sonnet-4-0"    # Claude Sonnet 4.0 - Agent推荐模型
+CLAUDE_4_6_OPUS = "claude-opus-4-6"      # Claude Opus 4.6 - Agent推荐模型
+CLAUDE_4_SONNET = "claude-sonnet-4-0"    # Claude Sonnet 4.0
 CLAUDE_4_5_SONNET = "claude-sonnet-4-5"  # Claude Sonnet 4.5 - Agent推荐模型
+CLAUDE_4_6_SONNET = "claude-sonnet-4-6"  # Claude Sonnet 4.6 - Agent推荐模型

 # Gemini (Google)
 GEMINI_PRO = "gemini-1.0-pro"
@@ -34,10 +38,12 @@ GEMINI_15_flash = "gemini-1.5-flash"
 GEMINI_15_PRO = "gemini-1.5-pro"
 GEMINI_20_flash_exp = "gemini-2.0-flash-exp"  # exp结尾为实验模型，会逐步不再支持
 GEMINI_20_FLASH = "gemini-2.0-flash"  # 正式版模型
-GEMINI_25_FLASH_PRE = "gemini-2.5-flash-preview-05-20"  # preview为预览版模型，主要是新能力体验
+GEMINI_25_FLASH_PRE = "gemini-2.5-flash-preview-05-20"
 GEMINI_25_PRO_PRE = "gemini-2.5-pro-preview-05-06"
 GEMINI_3_FLASH_PRE = "gemini-3-flash-preview"  # Gemini 3 Flash Preview - Agent推荐模型
-GEMINI_3_PRO_PRE = "gemini-3-pro-preview"  # Gemini 3 Pro Preview - Agent推荐模型
+GEMINI_3_PRO_PRE = "gemini-3-pro-preview"  # Gemini 3 Pro Preview
+GEMINI_31_PRO_PRE = "gemini-3.1-pro-preview"  # Gemini 3.1 Pro Preview - Agent推荐模型
+GEMINI_31_FLASH_LITE_PRE = "gemini-3.1-flash-lite-preview"  # Gemini 3.1 Flash Lite Preview - Agent推荐模型

 # OpenAI
 GPT35 = "gpt-3.5-turbo"
@@ -62,6 +68,9 @@ GPT_41_NANO = "gpt-4.1-nano"
 GPT_5 = "gpt-5"
 GPT_5_MINI = "gpt-5-mini"
 GPT_5_NANO = "gpt-5-nano"
+GPT_54 = "gpt-5.4"  # GPT-5.4 - Agent recommended model
+GPT_54_MINI = "gpt-5.4-mini"
+GPT_54_NANO = "gpt-5.4-nano"
 O1 = "o1-preview"
 O1_MINI = "o1-mini"
 WHISPER_1 = "whisper-1"
@@ -79,15 +88,18 @@ QWEN_PLUS = "qwen-plus"
 QWEN_MAX = "qwen-max"
 QWEN_LONG = "qwen-long"
 QWEN3_MAX = "qwen3-max"  # Qwen3 Max - Agent推荐模型
+QWEN35_PLUS = "qwen3.5-plus"  # Qwen3.5 Plus - Omni model (MultiModalConversation)
 QWQ_PLUS = "qwq-plus"

 # MiniMax
+MINIMAX_M2_5 = "MiniMax-M2.5"  # MiniMax M2.5 - Latest
 MINIMAX_M2_1 = "MiniMax-M2.1"  # MiniMax M2.1 - Agent推荐模型
 MINIMAX_M2_1_LIGHTNING = "MiniMax-M2.1-lightning"  # MiniMax M2.1 极速版
 MINIMAX_M2 = "MiniMax-M2"  # MiniMax M2
 MINIMAX_ABAB6_5 = "abab6.5-chat"  # MiniMax abab6.5

 # GLM (智谱AI)
+GLM_5 = "glm-5"  # 智谱 GLM-5 - Latest
 GLM_4 = "glm-4"
 GLM_4_PLUS = "glm-4-plus"
 GLM_4_flash = "glm-4-flash"
@@ -100,6 +112,15 @@ GLM_4_7 = "glm-4.7"  # 智谱 GLM-4.7 - Agent推荐模型

 # Kimi (Moonshot)
 MOONSHOT = "moonshot"
+KIMI_K2 = "kimi-k2"
+KIMI_K2_5 = "kimi-k2.5"
+
+# Doubao (Volcengine Ark)
+DOUBAO = "doubao"
+DOUBAO_SEED_2_CODE = "doubao-seed-2-0-code-preview-260215"
+DOUBAO_SEED_2_PRO = "doubao-seed-2-0-pro-260215"
+DOUBAO_SEED_2_LITE = "doubao-seed-2-0-lite-260215"
+DOUBAO_SEED_2_MINI = "doubao-seed-2-0-mini-260215"

 # 其他模型
 WEN_XIN = "wenxin"
@@ -120,12 +141,12 @@ MODELSCOPE_MODEL_LIST = ["LLM-Research/c4ai-command-r-plus-08-2024","mistralai/M

 MODEL_LIST = [
              # Claude
-              CLAUDE3, CLAUDE_4_OPUS, CLAUDE_4_5_SONNET, CLAUDE_4_SONNET, CLAUDE_3_OPUS, CLAUDE_3_OPUS_0229, 
+              CLAUDE3, CLAUDE_4_6_SONNET, CLAUDE_4_6_OPUS, CLAUDE_4_OPUS, CLAUDE_4_5_SONNET, CLAUDE_4_SONNET, CLAUDE_3_OPUS, CLAUDE_3_OPUS_0229, 
              CLAUDE_35_SONNET, CLAUDE_35_SONNET_1022, CLAUDE_35_SONNET_0620, CLAUDE_3_SONNET, CLAUDE_3_HAIKU, 
              "claude", "claude-3-haiku", "claude-3-sonnet", "claude-3-opus", "claude-3.5-sonnet",
              
              # Gemini
-              GEMINI_3_PRO_PRE, GEMINI_3_FLASH_PRE, GEMINI_25_PRO_PRE, GEMINI_25_FLASH_PRE, 
+              GEMINI_31_FLASH_LITE_PRE, GEMINI_31_PRO_PRE, GEMINI_3_PRO_PRE, GEMINI_3_FLASH_PRE, GEMINI_25_PRO_PRE, GEMINI_25_FLASH_PRE,
              GEMINI_20_FLASH, GEMINI_20_flash_exp, GEMINI_15_PRO, GEMINI_15_flash, GEMINI_PRO, GEMINI,
              
              # OpenAI
@@ -135,24 +156,29 @@ MODEL_LIST = [
              GPT_4o, GPT_4O_0806, GPT_4o_MINI,
              GPT_41, GPT_41_MINI, GPT_41_NANO,
              GPT_5, GPT_5_MINI, GPT_5_NANO,
+              GPT_54, GPT_54_MINI, GPT_54_NANO,
              O1, O1_MINI,
              
              # DeepSeek
              DEEPSEEK_CHAT, DEEPSEEK_REASONER,
              
              # Qwen
-              QWEN, QWEN_TURBO, QWEN_PLUS, QWEN_MAX, QWEN_LONG, QWEN3_MAX,
+              QWEN, QWEN_TURBO, QWEN_PLUS, QWEN_MAX, QWEN_LONG, QWEN3_MAX, QWEN35_PLUS,
              
              # MiniMax
-              MiniMax, MINIMAX_M2_1, MINIMAX_M2_1_LIGHTNING, MINIMAX_M2, MINIMAX_ABAB6_5,
-              
+              MiniMax, MINIMAX_M2_5, MINIMAX_M2_1, MINIMAX_M2_1_LIGHTNING, MINIMAX_M2, MINIMAX_ABAB6_5,
+
              # GLM
-              ZHIPU_AI, GLM_4, GLM_4_PLUS, GLM_4_flash, GLM_4_LONG, GLM_4_ALLTOOLS, 
+              ZHIPU_AI, GLM_5, GLM_4, GLM_4_PLUS, GLM_4_flash, GLM_4_LONG, GLM_4_ALLTOOLS,
              GLM_4_0520, GLM_4_AIR, GLM_4_AIRX, GLM_4_7,
-              
+
              # Kimi
              MOONSHOT, "moonshot-v1-8k", "moonshot-v1-32k", "moonshot-v1-128k",
-              
+              KIMI_K2, KIMI_K2_5,
+
+              # Doubao
+              DOUBAO, DOUBAO_SEED_2_CODE, DOUBAO_SEED_2_PRO, DOUBAO_SEED_2_LITE, DOUBAO_SEED_2_MINI,
+
              # 其他模型
              WEN_XIN, WEN_XIN_4, XUNFEI,
              LINKAI_35, LINKAI_4_TURBO, LINKAI_4o,
@@ -163,3 +189,5 @@ MODEL_LIST = MODEL_LIST + GITEE_AI_MODEL_LIST + MODELSCOPE_MODEL_LIST
 # channel
 FEISHU = "feishu"
 DINGTALK = "dingtalk"
+WECOM_BOT = "wecom_bot"
+QQ = "qq"
--- a/common/linkai_client.py
+++ b/common/linkai_client.py
@@ -1,110 +0,0 @@
-from bridge.context import Context, ContextType
-from bridge.reply import Reply, ReplyType
-from common.log import logger
-from linkai import LinkAIClient, PushMsg
-from config import conf, pconf, plugin_config, available_setting, write_plugin_config
-from plugins import PluginManager
-import time
-
-
-chat_client: LinkAIClient
-
-
-class ChatClient(LinkAIClient):
-    def __init__(self, api_key, host, channel):
-        super().__init__(api_key, host)
-        self.channel = channel
-        self.client_type = channel.channel_type
-
-    def on_message(self, push_msg: PushMsg):
-        session_id = push_msg.session_id
-        msg_content = push_msg.msg_content
-        logger.info(f"receive msg push, session_id={session_id}, msg_content={msg_content}")
-        context = Context()
-        context.type = ContextType.TEXT
-        context["receiver"] = session_id
-        context["isgroup"] = push_msg.is_group
-        self.channel.send(Reply(ReplyType.TEXT, content=msg_content), context)
-
-    def on_config(self, config: dict):
-        if not self.client_id:
-            return
-        logger.info(f"[LinkAI] 从客户端管理加载远程配置: {config}")
-        if config.get("enabled") != "Y":
-            return
-
-        local_config = conf()
-        for key in config.keys():
-            if key in available_setting and config.get(key) is not None:
-                local_config[key] = config.get(key)
-        # 语音配置
-        reply_voice_mode = config.get("reply_voice_mode")
-        if reply_voice_mode:
-            if reply_voice_mode == "voice_reply_voice":
-                local_config["voice_reply_voice"] = True
-                local_config["always_reply_voice"] = False
-            elif reply_voice_mode == "always_reply_voice":
-                local_config["always_reply_voice"] = True
-                local_config["voice_reply_voice"] = True
-            elif reply_voice_mode == "no_reply_voice":
-                local_config["always_reply_voice"] = False
-                local_config["voice_reply_voice"] = False
-
-        if config.get("admin_password"):
-            if not pconf("Godcmd"):
-                write_plugin_config({"Godcmd": {"password": config.get("admin_password"), "admin_users": []} })
-            else:
-                pconf("Godcmd")["password"] = config.get("admin_password")
-            PluginManager().instances["GODCMD"].reload()
-
-        if config.get("group_app_map") and pconf("linkai"):
-            local_group_map = {}
-            for mapping in config.get("group_app_map"):
-                local_group_map[mapping.get("group_name")] = mapping.get("app_code")
-            pconf("linkai")["group_app_map"] = local_group_map
-            PluginManager().instances["LINKAI"].reload()
-
-        if config.get("text_to_image") and config.get("text_to_image") == "midjourney" and pconf("linkai"):
-            if pconf("linkai")["midjourney"]:
-                pconf("linkai")["midjourney"]["enabled"] = True
-                pconf("linkai")["midjourney"]["use_image_create_prefix"] = True
-        elif config.get("text_to_image") and config.get("text_to_image") in ["dall-e-2", "dall-e-3"]:
-            if pconf("linkai")["midjourney"]:
-                pconf("linkai")["midjourney"]["use_image_create_prefix"] = False
-
-
-def start(channel):
-    global chat_client
-    chat_client = ChatClient(api_key=conf().get("linkai_api_key"), host="", channel=channel)
-    chat_client.config = _build_config()
-    chat_client.start()
-    time.sleep(1.5)
-    if chat_client.client_id:
-        logger.info("[LinkAI] 可前往控制台进行线上登录和配置：https://link-ai.tech/console/clients")
-
-
-def _build_config():
-    local_conf = conf()
-    config = {
-        "linkai_app_code": local_conf.get("linkai_app_code"),
-        "single_chat_prefix": local_conf.get("single_chat_prefix"),
-        "single_chat_reply_prefix": local_conf.get("single_chat_reply_prefix"),
-        "single_chat_reply_suffix": local_conf.get("single_chat_reply_suffix"),
-        "group_chat_prefix": local_conf.get("group_chat_prefix"),
-        "group_chat_reply_prefix": local_conf.get("group_chat_reply_prefix"),
-        "group_chat_reply_suffix": local_conf.get("group_chat_reply_suffix"),
-        "group_name_white_list": local_conf.get("group_name_white_list"),
-        "nick_name_black_list": local_conf.get("nick_name_black_list"),
-        "speech_recognition": "Y" if local_conf.get("speech_recognition") else "N",
-        "text_to_image": local_conf.get("text_to_image"),
-        "image_create_prefix": local_conf.get("image_create_prefix")
-    }
-    if local_conf.get("always_reply_voice"):
-        config["reply_voice_mode"] = "always_reply_voice"
-    elif local_conf.get("voice_reply_voice"):
-        config["reply_voice_mode"] = "voice_reply_voice"
-    if pconf("linkai"):
-        config["group_app_map"] = pconf("linkai").get("group_app_map")
-    if plugin_config.get("Godcmd"):
-        config["admin_password"] = plugin_config.get("Godcmd").get("password")
-    return config
--- a/common/package_manager.py
+++ b/common/package_manager.py
@@ -28,7 +28,7 @@ def check_dulwich():
        except ImportError:
            try:
                install("dulwich")
-            except:
+            except Exception:
                needwait = True
    try:
        import dulwich
--- a/common/utils.py
+++ b/common/utils.py
@@ -2,7 +2,6 @@ import io
 import os
 import re
 from urllib.parse import urlparse
-from PIL import Image
 from common.log import logger

 def fsize(file):
@@ -23,6 +22,7 @@ def fsize(file):
 def compress_imgfile(file, max_size):
    if fsize(file) <= max_size:
        return file
+    from PIL import Image
    file.seek(0)
    img = Image.open(file)
    rgb_image = img.convert("RGB")
@@ -76,3 +76,42 @@ def remove_markdown_symbol(text: str):
    if not text:
        return text
    return re.sub(r'\*\*(.*?)\*\*', r'\1', text)
+
+
+def expand_path(path: str) -> str:
+    """
+    Expand user path with proper Windows support.
+    
+    On Windows, os.path.expanduser('~') may not work properly in some shells (like PowerShell).
+    This function provides a more robust path expansion.
+    
+    Args:
+        path: Path string that may contain ~
+        
+    Returns:
+        Expanded absolute path
+    """
+    if not path:
+        return path
+    
+    # Try standard expansion first
+    expanded = os.path.expanduser(path)
+    
+    # If expansion didn't work (path still starts with ~), use HOME or USERPROFILE
+    if expanded.startswith('~'):
+        import platform
+        if platform.system() == 'Windows':
+            # On Windows, try USERPROFILE first, then HOME
+            home = os.environ.get('USERPROFILE') or os.environ.get('HOME')
+        else:
+            # On Unix-like systems, use HOME
+            home = os.environ.get('HOME')
+        
+        if home:
+            # Replace ~ with home directory
+            if path == '~':
+                expanded = home
+            elif path.startswith('~/') or path.startswith('~\\'):
+                expanded = os.path.join(home, path[2:])
+    
+    return expanded
--- a/config-template.json
+++ b/config-template.json
@@ -1,15 +1,17 @@
 {
  "channel_type": "web",
-  "model": "claude-sonnet-4-5",
+  "model": "MiniMax-M2.5",
+  "minimax_api_key": "",
+  "zhipu_ai_api_key": "",
+  "ark_api_key": "",
+  "moonshot_api_key": "",
+  "dashscope_api_key": "",
  "claude_api_key": "",
  "claude_api_base": "https://api.anthropic.com/v1",
  "open_ai_api_key": "",
  "open_ai_api_base": "https://api.openai.com/v1",
  "gemini_api_key": "",
  "gemini_api_base": "https://generativelanguage.googleapis.com",
-  "zhipu_ai_api_key": "",
-  "minimax_api_key": "",
-  "dashscope_api_key": "",
  "voice_to_text": "openai",
  "text_to_voice": "openai",
  "voice_reply_voice": false,
@@ -18,11 +20,12 @@
  "use_linkai": false,
  "linkai_api_key": "",
  "linkai_app_code": "",
-  "feishu_bot_name": "",
  "feishu_app_id": "",
  "feishu_app_secret": "",
  "dingtalk_client_id": "",
  "dingtalk_client_secret":"",
+  "wecom_bot_id": "",
+  "wecom_bot_secret": "",
  "agent": true,
  "agent_max_context_tokens": 40000,
  "agent_max_context_turns": 20,
--- a/config.py
+++ b/config.py
@@ -20,7 +20,7 @@ available_setting = {
    "proxy": "",  # openai使用的代理
    # chatgpt模型， 当use_azure_chatgpt为true时，其名称为Azure上model deployment名称
    "model": "gpt-3.5-turbo",  # 可选择: gpt-4o, pt-4o-mini, gpt-4-turbo, claude-3-sonnet, wenxin, moonshot, qwen-turbo, xunfei, glm-4, minimax, gemini等模型，全部可选模型详见common/const.py文件
-    "bot_type": "",  # 可选配置，使用兼容openai格式的三方服务时候，需填"chatGPT"。bot具体名称详见common/const.py文件列出的bot_type，如不填根据model名称判断，
+    "bot_type": "",  # 可选配置，使用兼容openai格式的三方服务时候，需填"openai"（历史值"chatGPT"仍兼容）。bot具体名称详见common/const.py文件，如不填根据model名称判断
    "use_azure_chatgpt": False,  # 是否使用azure的chatgpt
    "azure_deployment_id": "",  # azure 模型部署名称
    "azure_api_version": "",  # azure api版本
@@ -37,7 +37,7 @@ available_setting = {
    "group_name_white_list": ["ChatGPT测试群", "ChatGPT测试群2"],  # 开启自动回复的群名称列表
    "group_name_keyword_white_list": [],  # 开启自动回复的群名称关键词列表
    "group_chat_in_one_session": ["ChatGPT测试群"],  # 支持会话上下文共享的群名称
-    "group_shared_session": True,  # 群聊是否共享会话上下文（所有成员共享），默认为True。False时每个用户在群内有独立会话
+    "group_shared_session": False,  # 群聊是否共享会话上下文（所有成员共享）。False时每个用户在群内有独立会话
    "nick_name_black_list": [],  # 用户昵称黑名单
    "group_welcome_msg": "",  # 配置新人进群固定欢迎语，不配置则使用随机风格欢迎
    "trigger_by_self": False,  # 是否允许机器人触发
@@ -95,8 +95,6 @@ available_setting = {
    "dashscope_api_key": "",
    # Google Gemini Api Key
    "gemini_api_key": "",
-    # wework的通用配置
-    "wework_smart": True,  # 配置wework是否使用已登录的企业微信，False为多开
    # 语音设置
    "speech_recognition": True,  # 是否开启语音识别
    "group_speech_recognition": False,  # 是否开启群组语音识别
@@ -118,7 +116,7 @@ available_setting = {
    # elevenlabs 语音api配置
    "xi_api_key": "",  # 获取ap的方法可以参考https://docs.elevenlabs.io/api-reference/quick-start/authentication
    "xi_voice_id": "",  # ElevenLabs提供了9种英式、美式等英语发音id，分别是“Adam/Antoni/Arnold/Bella/Domi/Elli/Josh/Rachel/Sam”
-    # 服务时间限制，目前支持itchat
+    # 服务时间限制
    "chat_time_module": False,  # 是否开启服务时间限制
    "chat_start_time": "00:00",  # 服务开始时间
    "chat_stop_time": "24:00",  # 服务结束时间
@@ -127,10 +125,6 @@ available_setting = {
    # baidu翻译api的配置
    "baidu_translate_app_id": "",  # 百度翻译api的appid
    "baidu_translate_app_key": "",  # 百度翻译api的秘钥
-    # itchat的配置
-    "hot_reload": False,  # 是否开启热重载
-    # wechaty的配置
-    "wechaty_puppet_service_token": "",  # wechaty的token
    # wechatmp的配置
    "wechatmp_token": "",  # 微信公众平台的Token
    "wechatmp_port": 8080,  # 微信公众平台的端口,需要端口转发到80或443
@@ -156,11 +150,14 @@ available_setting = {
    "dingtalk_client_id": "",  # 钉钉机器人Client ID 
    "dingtalk_client_secret": "",  # 钉钉机器人Client Secret
    "dingtalk_card_enabled": False,
-    
+    # 企微智能机器人配置(长连接模式)
+    "wecom_bot_id": "",  # 企微智能机器人BotID
+    "wecom_bot_secret": "",  # 企微智能机器人长连接Secret
    # chatgpt指令自定义触发词
    "clear_memory_commands": ["#清除记忆"],  # 重置会话指令，必须以#开头
    # channel配置
-    "channel_type": "",  # 通道类型，支持：{wx,wxy,terminal,wechatmp,wechatmp_service,wechatcom_app,dingtalk}
+    "channel_type": "",  # 通道类型，支持多渠道同时运行。单个: "feishu"，多个: "feishu, dingtalk" 或 ["feishu", "dingtalk"]。可选值: web,feishu,dingtalk,wecom_bot,wechatmp,wechatmp_service,wechatcom_app
+    "web_console": True,  # 是否自动启动Web控制台（默认启动）。设为False可禁用
    "subscribe_msg": "",  # 订阅消息, 支持: wechatmp, wechatmp_service, wechatcom_app
    "debug": False,  # 是否开启debug模式，开启后会打印更多日志
    "appdata_dir": "",  # 数据目录
@@ -174,7 +171,10 @@ available_setting = {
    "zhipu_ai_api_key": "",
    "zhipu_ai_api_base": "https://open.bigmodel.cn/api/paas/v4",
    "moonshot_api_key": "",
-    "moonshot_base_url": "https://api.moonshot.cn/v1/chat/completions",
+    "moonshot_base_url": "https://api.moonshot.cn/v1",
+    # 豆包(火山方舟) 平台配置
+    "ark_api_key": "",
+    "ark_base_url": "https://ark.cn-beijing.volces.com/api/v3",
    #魔搭社区 平台配置
    "modelscope_api_key": "",
    "modelscope_base_url": "https://api-inference.modelscope.cn/v1/chat/completions",
@@ -183,6 +183,8 @@ available_setting = {
    "linkai_api_key": "",
    "linkai_app_code": "",
    "linkai_api_base": "https://api.link-ai.tech",  # linkAI服务地址
+    "cloud_host": "client.link-ai.tech",
+    "cloud_deployment_id": "",
    "minimax_api_key": "",
    "Minimax_group_id": "",
    "Minimax_base_url": "",
@@ -245,9 +247,9 @@ class Config(dict):
                self.user_datas = pickle.load(f)
                logger.debug("[Config] User datas loaded.")
        except FileNotFoundError as e:
-            logger.info("[Config] User datas file not found, ignore.")
+            logger.debug("[Config] User datas file not found, ignore.")
        except Exception as e:
-            logger.info("[Config] User datas error: {}".format(e))
+            logger.warning("[Config] User datas error: {}".format(e))
            self.user_datas = {}

    def save_user_datas(self):
@@ -319,7 +321,7 @@ def load_config():
            logger.info("[INIT] override config by environ args: {}={}".format(name, value))
            try:
                config[name] = eval(value)
-            except:
+            except Exception:
                if value == "false":
                    config[name] = False
                elif value == "true":
@@ -350,6 +352,48 @@ def load_config():
    logger.info("[INIT] Debug: {}".format(config.get("debug", False)))
    logger.info("[INIT] ========================================")

+    # Sync selected config values to environment variables so that
+    # subprocesses (e.g. shell skill scripts) can access them directly.
+    # Existing env vars are NOT overwritten (env takes precedence).
+    _CONFIG_TO_ENV = {
+        "open_ai_api_key": "OPENAI_API_KEY",
+        "open_ai_api_base": "OPENAI_API_BASE",
+        "linkai_api_key": "LINKAI_API_KEY",
+        "linkai_api_base": "LINKAI_API_BASE",
+        "claude_api_key": "CLAUDE_API_KEY",
+        "claude_api_base": "CLAUDE_API_BASE",
+        "gemini_api_key": "GEMINI_API_KEY",
+        "gemini_api_base": "GEMINI_API_BASE",
+        "minimax_api_key": "MINIMAX_API_KEY",
+        "minimax_api_base": "MINIMAX_API_BASE",
+        "zhipu_ai_api_key": "ZHIPU_AI_API_KEY",
+        "zhipu_ai_api_base": "ZHIPU_AI_API_BASE",
+        "moonshot_api_key": "MOONSHOT_API_KEY",
+        "moonshot_api_base": "MOONSHOT_API_BASE",
+        "ark_api_key": "ARK_API_KEY",
+        "ark_api_base": "ARK_API_BASE",
+        # Channel credentials (used by skills that check env vars)
+        "feishu_app_id": "FEISHU_APP_ID",
+        "feishu_app_secret": "FEISHU_APP_SECRET",
+        "dingtalk_client_id": "DINGTALK_CLIENT_ID",
+        "dingtalk_client_secret": "DINGTALK_CLIENT_SECRET",
+        "wechatmp_app_id": "WECHATMP_APP_ID",
+        "wechatmp_app_secret": "WECHATMP_APP_SECRET",
+        "wechatcomapp_agent_id": "WECHATCOMAPP_AGENT_ID",
+        "wechatcomapp_secret": "WECHATCOMAPP_SECRET",
+        "qq_app_id": "QQ_APP_ID",
+        "qq_app_secret": "QQ_APP_SECRET"
+    }
+    injected = 0
+    for conf_key, env_key in _CONFIG_TO_ENV.items():
+        if env_key not in os.environ:
+            val = config.get(conf_key, "")
+            if val:
+                os.environ[env_key] = str(val)
+                injected += 1
+    if injected:
+        logger.info("[INIT] Synced {} config values to environment variables".format(injected))
+
    config.load_user_datas()


--- a/docker/Dockerfile.latest
+++ b/docker/Dockerfile.latest
@@ -25,11 +25,11 @@ WORKDIR ${BUILD_PREFIX}
 ADD docker/entrypoint.sh /entrypoint.sh

 RUN chmod +x /entrypoint.sh \
-    && mkdir -p /home/noroot \
-    && groupadd -r noroot \
-    && useradd -r -g noroot -s /bin/bash -d /home/noroot noroot \
-    && chown -R noroot:noroot /home/noroot ${BUILD_PREFIX} /usr/local/lib
+    && mkdir -p /home/agent/cow \
+    && groupadd -r agent \
+    && useradd -r -g agent -s /bin/bash -d /home/agent agent \
+    && chown -R agent:agent /home/agent ${BUILD_PREFIX} /usr/local/lib

-USER noroot
+USER agent

 ENTRYPOINT ["/entrypoint.sh"]
--- a/docker/docker-compose.yml
+++ b/docker/docker-compose.yml
@@ -5,22 +5,39 @@ services:
    container_name: chatgpt-on-wechat
    security_opt:
      - seccomp:unconfined
+    ports:
+      - "9899:9899"
    environment:
      CHANNEL_TYPE: 'web'
-      OPEN_AI_API_KEY: 'YOUR API KEY'
-      MODEL: ''
-      PROXY: ''
-      SINGLE_CHAT_PREFIX: '["bot", "@bot"]'
-      SINGLE_CHAT_REPLY_PREFIX: '"[bot] "'
-      GROUP_CHAT_PREFIX: '["@bot"]'
-      GROUP_NAME_WHITE_LIST: '["ChatGPT测试群", "ChatGPT测试群2"]'
-      IMAGE_CREATE_PREFIX: '["画", "看", "找"]'
-      CONVERSATION_MAX_TOKENS: 1000
-      SPEECH_RECOGNITION: 'False'
-      CHARACTER_DESC: '你是基于大语言模型的AI智能助手，旨在回答并解决人们的任何问题，并且可以使用多种语言与人交流。'
-      EXPIRES_IN_SECONDS: 3600
-      USE_GLOBAL_PLUGIN_CONFIG: 'True'
+      MODEL: 'MiniMax-M2.5'
+      MINIMAX_API_KEY: ''
+      ZHIPU_AI_API_KEY: ''
+      ARK_API_KEY: ''
+      MOONSHOT_API_KEY: ''
+      DASHSCOPE_API_KEY: ''
+      CLAUDE_API_KEY: ''
+      CLAUDE_API_BASE: 'https://api.anthropic.com/v1'
+      OPEN_AI_API_KEY: ''
+      OPEN_AI_API_BASE: 'https://api.openai.com/v1'
+      GEMINI_API_KEY: ''
+      GEMINI_API_BASE: 'https://generativelanguage.googleapis.com'
+      VOICE_TO_TEXT: 'openai'
+      TEXT_TO_VOICE: 'openai'
+      VOICE_REPLY_VOICE: 'False'
+      SPEECH_RECOGNITION: 'True'
+      GROUP_SPEECH_RECOGNITION: 'False'
      USE_LINKAI: 'False'
-      AGENT: 'True'
      LINKAI_API_KEY: ''
      LINKAI_APP_CODE: ''
+      FEISHU_APP_ID: ''
+      FEISHU_APP_SECRET: ''
+      DINGTALK_CLIENT_ID: ''
+      DINGTALK_CLIENT_SECRET: ''
+      WECOM_BOT_ID: ''
+      WECOM_BOT_SECRET: ''
+      AGENT: 'True'
+      AGENT_MAX_CONTEXT_TOKENS: 40000
+      AGENT_MAX_CONTEXT_TURNS: 20
+      AGENT_MAX_STEPS: 15
+    volumes:
+      - ./cow:/home/agent/cow
--- a/docs/agent.md
+++ b/docs/agent.md
@@ -8,7 +8,7 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent
 - **工具系统**：内置实现10+种工具，包括文件读写、bash终端、浏览器、定时任务、记忆管理等，通过Agent管理你的计算机或服务器
 - **长期记忆**：自动将对话记忆持久化至本地文件和数据库中，包括全局记忆和天级记忆，支持关键词及向量检索
 - **Skills系统**：新增Skill运行引擎，内置多种技能，并支持通过自然语言对话完成自定义Skills开发
- **多渠道和多模型支持**：支持在Web、飞书、钉钉、企微等多渠道与Agent交互，支持Claude、Gemini、OpenAI、GLM、MiniMax、Qwen 等多种国内外主流模型
+- **多渠道和多模型支持**：支持在Web、飞书、钉钉、企微等多渠道与Agent交互，支持Claude、Gemini、OpenAI、GLM、MiniMax、Qwen、Kimi、Doubao 等多种国内外主流模型
 - **安全和成本**：通过秘钥管理工具、提示词控制、系统权限等手段控制Agent的访问安全；通过最大记忆轮次、最大上下文token、工具执行步数对token成本进行限制


@@ -23,7 +23,7 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent

 在后续的长期对话中，Agent会在需要的时候智能记录或检索记忆，并对自身设定、用户偏好、记忆文件等进行不断更新，总结和记录经验和教训，真正实现自主思考和不断成长。

-<img width="800" src="https://cdn.link-ai.tech/doc/20260203000455.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260203000455.png" />



@@ -37,14 +37,14 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent

 针对操作系统的终端和文件的访问能力，是最基础和核心的工具，其他很多工具或技能都是基于基础工具进行扩展。用户可通过手机端与Agent交互，操作个人电脑或服务器上的资源：

-<img width="800" src="https://cdn.link-ai.tech/doc/20260202181130.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260202181130.png" />

 #### 1.2 编程能力

 基于编程能力和系统访问能力，Agent可以实现从信息搜索、图片等素材生成、编码、测试、部署、Nginx配置修改、发布的 Vibecoding 全流程，通过手机端简单的一句命令完成应用的快速demo：


-<img width="800" src="https://cdn.link-ai.tech/doc/20260203121008.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260203121008.png" />



@@ -53,7 +53,7 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent
 基于 scheduler 工具实现动态定时任务，支持 **一次性任务、固定时间间隔、Cron表达式** 三种形式，任务触发可选择**固定消息发送** 或 **Agent动态任务** 执行两种模式，有很高灵活性：


-<img width="800" src="https://cdn.link-ai.tech/doc/20260202195402.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260202195402.png" />

 同时你也可以通过自然语言快速查看和管理已有的定时任务。

@@ -62,7 +62,7 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent

 技能所需要的秘钥存储在环境变量文件中，由 `env_config` 工具进行管理，你可以通过对话的方式更新秘钥，工具内置了安全保护和脱敏策略，会严格保护秘钥安全：

-<img width="800" src="https://cdn.link-ai.tech/doc/20260202234939.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260202234939.png" />

 ### 3. 技能系统

@@ -70,22 +70,22 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent

 - **内置技能：** 在项目的`skills`目录下，包含技能创造器、网络搜索、图像识别（openai-image-vision）、LinkAI智能体、网页抓取等。内置Skill根据依赖条件 (API Key、系统命令等) 自动判断是否启用。通过技能创造器可以快速创建自定义技能。

- **自定义技能：** 由用户通过对话创建，存放在工作空间中 (`~/cow/skills/`)，基于自定义技能可以实现任何
+- **自定义技能：** 由用户通过对话创建，存放在工作空间中 (`~/cow/skills/`)，基于自定义技能可以实现任何复杂的业务流程和第三方系统对接。


 #### 3.1 创建技能

 通过 `skill-creator` 技能可以通过对话的方式快速创建技能。你可以在与Agent的写作中让他对将某个工作流程固化为技能，或者把任意接口文档和示例发送给Agent，让他直接完成对接：

-<img width="800" src="https://cdn.link-ai.tech/doc/20260202202247.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260202202247.png" />


 #### 3.2 搜索和图像识别

- **搜索技能：** 系统内置实现了 `bocha-search`(博查搜索)的Skill，依赖环境变量 `BOCHA_SEARCH_API_KEY`，可在[控制台]()进行创建，并发送给Agent完成配置
+- **搜索技能：** 系统内置实现了 `bocha-search`(博查搜索)的Skill，依赖环境变量 `BOCHA_SEARCH_API_KEY`，可在[控制台](https://open.bochaai.com/)进行创建，并发送给Agent完成配置
 - **图像识别技能：** 实现了 `openai-image-vision` 插件，可使用 gpt-4.1-mini、gpt-4.1 等图像识别模型。依赖秘钥 `OPENAI_API_KEY`，可通过config.json或env_config工具进行维护。

-<img width="800" src="https://cdn.link-ai.tech/doc/20260202213219.png">
+<img width="800" src="https://cdn.link-ai.tech/doc/20260202213219.png" />


 #### 3.3 三方知识库和插件
@@ -113,7 +113,7 @@ Cow项目从简单的聊天机器人全面升级为超级智能助理 **CowAgent

 Agent可根据智能体的名称和描述进行决策，并通过 app_code 调用接口访问对应的应用/工作流，通过该技能，可以灵活访问LinkAI平台上的智能体、知识库、插件等能力，实现效果如下：

-<img width="750" src="https://cdn.link-ai.tech/doc/20260202234350.png">
+<img width="750" src="https://cdn.link-ai.tech/doc/20260202234350.png" />

 注：需通过 `env_config` 配置 `LINKAI_API_KEY`，或在config.json中添加 `linkai_api_key` 配置。

@@ -127,7 +127,7 @@ Agent可根据智能体的名称和描述进行决策，并通过 app_code 调
 在命令行中执行：

 ```bash
-bash <(curl -sS https://cdn.link-ai.tech/code/cow/run.sh)
+bash <(curl -fsSL https://cdn.link-ai.tech/code/cow/run.sh)
 ```

 详细说明及后续程序管理参考：[项目启动脚本](https://github.com/zhayujie/chatgpt-on-wechat/wiki/CowAgentQuickStart)
@@ -137,11 +137,14 @@ bash <(curl -sS https://cdn.link-ai.tech/code/cow/run.sh)

 Agent模式推荐使用以下模型，可根据效果及成本综合选择：

- **Claude**: `claude-sonnet-4-5`、`claude-sonnet-4-0`
- **Gemini**: `gemini-3-flash-preview`、`gemini-3-pro-preview`
- **GLM**: `glm-4.7`
- **MiniMax**: `MiniMax-M2.1`
- **Qwen**: `qwen3-max`
+- **MiniMax**: `MiniMax-M2.5`
+- **GLM**: `glm-5`
+- **Kimi**: `kimi-k2.5`
+- **Doubao**: `doubao-seed-2-0-code-preview-260215`
+- **Qwen**: `qwen3.5-plus`
+- **Claude**: `claude-sonnet-4-6`
+- **Gemini**: `gemini-3.1-flash-lite-preview`
+- **OpenAI**: `gpt-5.4`

 详细模型配置方式参考 [README.md 模型说明](../README.md#模型说明)

@@ -176,5 +179,7 @@ Agent支持在多种渠道中使用，只需修改 `config.json` 中的 `channel
 - **飞书接入**：[飞书接入文档](https://docs.link-ai.tech/cow/multi-platform/feishu)
 - **钉钉接入**：[钉钉接入文档](https://docs.link-ai.tech/cow/multi-platform/dingtalk)
 - **企业微信应用接入**：[企微应用文档](https://docs.link-ai.tech/cow/multi-platform/wechat-com)
+- **企微智能机器人**：[企微智能机器人文档](https://docs.link-ai.tech/cow/multi-platform/wecom-bot)
+- **QQ机器人**：[QQ机器人文档](https://docs.link-ai.tech/cow/multi-platform/qq)

 更多渠道配置参考：[通道说明](../README.md#通道说明)
--- a/docs/channels/dingtalk.mdx
+++ b/docs/channels/dingtalk.mdx
@@ -0,0 +1,56 @@
+---
+title: 钉钉
+description: 将 CowAgent 接入钉钉应用
+---
+
+通过钉钉开放平台创建智能机器人应用，将 CowAgent 接入钉钉。
+
+## 一、创建应用
+
+1. 进入 [钉钉开发者后台](https://open-dev.dingtalk.com/fe/app#/corp/app)，登录后点击 **创建应用**，填写应用相关信息：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-create-app.png" width="800"/>
+
+2. 点击添加应用能力，选择 **机器人** 能力，点击 **添加**：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-add-bot.png" width="800"/>
+
+3. 配置机器人信息后点击 **发布**。发布后，点击 "**点击调试**"，会自动创建测试群聊，可在客户端查看：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-config-bot.png" width="600"/>
+
+4. 点击 **版本管理与发布**，创建新版本发布：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-publish-bot.png" width="700"/>
+
+## 二、项目配置
+
+1. 点击 **凭证与基础信息**，获取 `Client ID` 和 `Client Secret`：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-get-secret.png" width="700"/>
+
+2. 将以下配置加入项目根目录的 `config.json` 文件：
+
+```json
+{
+  "channel_type": "dingtalk",
+  "dingtalk_client_id": "YOUR_CLIENT_ID",
+  "dingtalk_client_secret": "YOUR_CLIENT_SECRET"
+}
+```
+
+3. 安装依赖：
+
+```bash
+pip3 install dingtalk_stream
+```
+
+4. 启动项目后，在钉钉开发者后台点击 **事件订阅**，点击 **已完成接入，验证连接通道**，显示 **连接接入成功** 即表示配置完成：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-event-sub.png" width="700"/>
+
+## 三、使用
+
+与机器人私聊或将机器人拉入企业群中均可开启对话：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-hosting-demo.png" width="650"/>
--- a/docs/channels/feishu.mdx
+++ b/docs/channels/feishu.mdx
@@ -0,0 +1,69 @@
+---
+title: 飞书
+description: 将 CowAgent 接入飞书应用
+---
+
+通过自建应用将 CowAgent 接入飞书，需要是飞书企业用户且具有企业管理权限。
+
+## 一、创建企业自建应用
+
+### 1. 创建应用
+
+进入 [飞书开发平台](https://open.feishu.cn/app/)，点击 **创建企业自建应用**，填写必要信息后点击 **创建**：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/feishu-hosting-create-app.jpg" width="500"/>
+
+### 2. 添加机器人能力
+
+在 **添加应用能力** 菜单中，为应用添加 **机器人** 能力：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/feishu-hosting-add-bot.jpg" width="800"/>
+
+### 3. 配置应用权限
+
+点击 **权限管理**，复制以下权限配置，粘贴到 **权限配置** 下方的输入框内，全选筛选出来的权限，点击 **批量开通** 并确认：
+
+```
+im:message,im:message.group_at_msg,im:message.group_at_msg:readonly,im:message.p2p_msg,im:message.p2p_msg:readonly,im:message:send_as_bot,im:resource
+```
+
+<img src="https://cdn.link-ai.tech/doc/feishu-hosting-add-auth2.png" width="800"/>
+
+## 二、项目配置
+
+1. 在 **凭证与基础信息** 中获取 `App ID` 和 `App Secret`：
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/feishu-hosting-appid-secret.jpg" width="800"/>
+
+2. 将以下配置加入项目根目录的 `config.json` 文件：
+
+```json
+{
+  "channel_type": "feishu",
+  "feishu_app_id": "YOUR_APP_ID",
+  "feishu_app_secret": "YOUR_APP_SECRET",
+  "feishu_bot_name": "YOUR_BOT_NAME"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `feishu_app_id` | 飞书机器人应用 App ID |
+| `feishu_app_secret` | 飞书机器人 App Secret |
+| `feishu_bot_name` | 飞书机器人名称（创建应用时设置），群聊中使用依赖此配置 |
+
+配置完成后启动项目。
+
+## 三、配置事件订阅
+
+1. 成功运行项目后，在飞书开放平台点击 **事件与回调**，选择 **长连接** 方式，点击保存：
+
+<img src="https://cdn.link-ai.tech/doc/202601311731183.png" width="600"/>
+
+2. 点击下方的 **添加事件**，搜索 "接收消息"，选择 "**接收消息v2.0**"，确认添加。
+
+3. 点击 **版本管理与发布**，创建版本并申请 **线上发布**，在飞书客户端查看审批消息并审核通过：
+
+<img src="https://cdn.link-ai.tech/doc/202601311807356.png" width="600"/>
+
+完成后在飞书中搜索机器人名称，即可开始对话。
--- a/docs/channels/qq.mdx
+++ b/docs/channels/qq.mdx
@@ -0,0 +1,88 @@
+---
+title: QQ 机器人
+description: 将 CowAgent 接入 QQ 机器人（WebSocket 长连接模式）
+---
+
+> 通过 QQ 开放平台的机器人接口接入 CowAgent，支持 QQ 单聊、QQ 群聊（@机器人）、频道消息和频道私信，无需公网 IP，使用 WebSocket 长连接模式。
+
+<Note>
+  QQ 机器人通过 QQ 开放平台创建，使用 WebSocket 长连接接收消息，通过 OpenAPI 发送消息，无需公网 IP 和域名。
+</Note>
+
+## 一、创建 QQ 机器人
+
+> 进入[QQ 开放平台](https://q.qq.com)，QQ扫码登录，如果未注册开放平台账号，请先完成[账号注册](https://q.qq.com/#/register)。
+
+1.在 [QQ开放平台-机器人列表页](https://q.qq.com/#/apps)，点击创建机器人:
+
+<img src="https://cdn.link-ai.tech/doc/20260317162900.png" width="800"/>
+
+2.填写机器人名称、头像等基本信息，完成创建：
+
+<img src="https://cdn.link-ai.tech/doc/20260317163005.png" width="800"/>
+
+3.点击进入机器人配置页面，选择**开发管理**菜单，完成以下步骤：
+
+  - 复制并记录 **AppID**（机器人ID）
+  - 生成并记录 **AppSecret**（机器人秘钥）
+
+<img src="https://cdn.link-ai.tech/doc/20260317164955.png" width="800"/>
+
+## 二、配置和运行
+
+### 方式一：Web 控制台接入
+
+启动 Cow项目后打开 Web 控制台 (本地链接为: http://127.0.0.1:9899/ )，选择 **通道** 菜单，点击 **接入通道**，选择 **QQ 机器人**，填写上一步保存的 AppID 和 AppSecret，点击接入即可。
+
+<img src="https://cdn.link-ai.tech/doc/20260317165425.png" width="800"/>
+
+### 方式二：配置文件接入
+
+在 `config.json` 中添加以下配置：
+
+```json
+{
+  "channel_type": "qq",
+  "qq_app_id": "YOUR_APP_ID",
+  "qq_app_secret": "YOUR_APP_SECRET"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `qq_app_id` | QQ 机器人的 AppID，在开放平台开发管理中获取 |
+| `qq_app_secret` | QQ 机器人的 AppSecret，在开放平台开发管理中获取 |
+
+配置完成后启动程序，日志显示 `[QQ] ✅ Connected successfully` 即表示连接成功。
+
+
+## 三、使用
+
+在 QQ开放平台 - 管理 - **使用范围和人员** 菜单中，使用QQ客户端扫描 "添加到群和消息列表" 的二维码，即可开始与QQ机器人的聊天：
+
+<img src="https://cdn.link-ai.tech/doc/20260317165947.png" width="800"/>
+
+对话效果：
+<img src="https://cdn.link-ai.tech/doc/20260317171508.png" width="800"/>
+
+## 四、功能说明
+
+> 注意：若需在群聊及频道中使用QQ机器人，需完成发布上架审核并在使用范围配置权限使用范围。
+
+| 功能 | 支持情况 |
+| --- | --- |
+| QQ 单聊 | ✅ |
+| QQ 群聊（@机器人） | ✅ |
+| 频道消息（@机器人） | ✅ |
+| 频道私信 | ✅ |
+| 文本消息 | ✅ 收发 |
+| 图片消息 | ✅ 收发（群聊和单聊） |
+| 文件消息 | ✅ 发送（群聊和单聊） |
+| 定时任务 | ✅ 主动推送（每月每用户限 4 条） |
+
+
+## 五、注意事项
+
+- **被动消息限制**：QQ 单聊被动消息有效期为 60 分钟，每条消息最多回复 5 次；QQ 群聊被动消息有效期为 5 分钟。
+- **主动消息限制**：单聊和群聊每月主动消息上限为 4 条，在使用定时任务功能时需要注意这个限制
+- **事件权限**：默认订阅 `GROUP_AND_C2C_EVENT`（QQ群/单聊）和 `PUBLIC_GUILD_MESSAGES`（频道公域消息），如需其他事件类型请在开放平台申请权限。
--- a/docs/channels/web.mdx
+++ b/docs/channels/web.mdx
@@ -0,0 +1,75 @@
+---
+title: Web 控制台
+description: 通过 Web 控制台使用 CowAgent
+---
+
+Web 控制台是 CowAgent 的默认通道，启动后会自动运行，通过浏览器即可与 Agent 对话，并支持在线管理模型、技能、记忆、通道等配置。
+
+## 配置
+
+```json
+{
+  "channel_type": "web",
+  "web_port": 9899
+}
+```
+
+| 参数 | 说明 | 默认值 |
+| --- | --- | --- |
+| `channel_type` | 设为 `web` | `web` |
+| `web_port` | Web 服务监听端口 | `9899` |
+
+## 访问地址
+
+启动项目后访问：
+
+- 本地运行：`http://localhost:9899`
+- 服务器运行：`http://<server-ip>:9899`
+
+<Note>
+  请确保服务器防火墙和安全组已放行对应端口。
+</Note>
+
+## 功能介绍
+
+### 对话界面
+
+支持流式输出，可实时展示 Agent 的思考过程（Reasoning）和工具调用过程（Tool Calls），更直观地观察 Agent 的决策过程：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227180120.png" />
+
+### 模型管理
+
+支持在线管理模型配置，无需手动编辑配置文件：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227173811.png" />
+
+### 技能管理
+
+支持在线查看和管理 Agent 技能（Skills）：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227173403.png" />
+
+### 记忆管理
+
+支持在线查看和管理 Agent 记忆：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227173349.png" />
+
+### 通道管理
+
+支持在线管理接入通道，支持实时连接/断开操作：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227173331.png" />
+
+### 定时任务
+
+支持在线查看和管理定时任务，包括一次性任务、固定间隔、Cron 表达式等多种调度方式的可视化管理：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227173704.png" />
+
+### 日志
+
+支持在线实时查看 Agent 运行日志，便于监控运行状态和排查问题：
+
+<img width="850" src="https://cdn.link-ai.tech/doc/20260227173514.png" />
--- a/docs/channels/wechatmp.mdx
+++ b/docs/channels/wechatmp.mdx
@@ -0,0 +1,72 @@
+---
+title: 微信公众号
+description: 将 CowAgent 接入微信公众号
+---
+
+CowAgent 支持接入个人订阅号和企业服务号两种公众号类型。
+
+| 类型 | 要求 | 特点 |
+| --- | --- | --- |
+| **个人订阅号** | 个人可申请 | 收到消息时会回复一条提示，回复生成后需用户主动发消息获取 |
+| **企业服务号** | 企业申请，需通过微信认证开通客服接口 | 回复生成后可主动推送给用户 |
+
+<Note>
+  公众号仅支持服务器和 Docker 部署，不支持本地运行。需额外安装扩展依赖：`pip3 install -r requirements-optional.txt`
+</Note>
+
+## 一、个人订阅号
+
+在 `config.json` 中添加以下配置：
+
+```json
+{
+  "channel_type": "wechatmp",
+  "single_chat_prefix": [""],
+  "wechatmp_app_id": "wx73f9******d1e48",
+  "wechatmp_app_secret": "YOUR_APP_SECRET",
+  "wechatmp_aes_key": "",
+  "wechatmp_token": "YOUR_TOKEN",
+  "wechatmp_port": 80
+}
+```
+
+### 配置步骤
+
+这些配置需要和 [微信公众号后台](https://mp.weixin.qq.com/advanced/advanced?action=dev&t=advanced/dev) 中的保持一致，进入页面后，在左侧菜单选择 **设置与开发 → 基本配置 → 服务器配置**，按下图进行配置：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103506.png" width="480"/>
+
+1. 在公众平台启用开发者密码（对应配置 `wechatmp_app_secret`），并将服务器 IP 填入白名单
+2. 按上图填写 `config.json` 中与公众号相关的配置，要与公众号后台的配置一致
+3. 启动程序，启动后会监听 80 端口（若无权限监听，则在启动命令前加上 `sudo`；若 80 端口已被占用，则关闭该占用进程）
+4. 在公众号后台 **启用服务器配置** 并提交，保存成功则表示已成功配置。注意 **"服务器地址(URL)"** 需要配置为 `http://{HOST}/wx` 的格式，其中 `{HOST}` 可以是服务器的 IP 或域名
+
+随后关注公众号并发送消息即可看到以下效果：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103522.png" width="720"/>
+
+由于受订阅号限制，回复内容较短的情况下（15s 内），可以立即完成回复，但耗时较长的回复则会先回复一句 "正在思考中"，后续需要用户输入任意文字主动获取答案，而服务号则可以通过客服接口解决这一问题。
+
+<Tip>
+  **语音识别**：可利用微信自带的语音识别功能，需要在公众号管理页面的 "设置与开发 → 接口权限" 页面开启 "接收语音识别结果"。
+</Tip>
+
+## 二、企业服务号
+
+企业服务号与上述个人订阅号的接入过程基本相同，差异如下：
+
+1. 在公众平台申请企业服务号并完成微信认证，在接口权限中确认已获得 **客服接口** 的权限
+2. 在 `config.json` 中设置 `"channel_type": "wechatmp_service"`，其他配置与上述订阅号相同
+3. 交互效果上，即使是较长耗时的回复，也可以主动推送给用户，无需用户手动获取
+
+```json
+{
+  "channel_type": "wechatmp_service",
+  "single_chat_prefix": [""],
+  "wechatmp_app_id": "YOUR_APP_ID",
+  "wechatmp_app_secret": "YOUR_APP_SECRET",
+  "wechatmp_aes_key": "",
+  "wechatmp_token": "YOUR_TOKEN",
+  "wechatmp_port": 80
+}
+```
--- a/docs/channels/wecom-bot.mdx
+++ b/docs/channels/wecom-bot.mdx
@@ -0,0 +1,73 @@
+---
+title: 企微智能机器人
+description: 将 CowAgent 接入企业微信智能机器人（长连接模式）
+---
+
+> 通过企业微信智能机器人接入CowAgent，支持企业内部单聊和内部群聊，无需公网 IP，使用 WebSocket 长连接模式，支持Markdown渲染和流式输出。
+
+<Note>
+  智能机器人与企业微信自建应用是两种不同的接入方式。智能机器人使用 WebSocket 长连接，无需服务器公网 IP 和域名，配置更简单。
+</Note>
+
+## 一、创建智能机器人
+
+1. 打开企业微信客户端，进入工作台，点击**智能机器人**：
+
+<img src="https://cdn.link-ai.tech/doc/20260316180959.png" width="800"/>
+
+2. 点击创建机器人 - 手动创建：
+
+<img src="https://cdn.link-ai.tech/doc/20260316181118.png" width="800"/>
+
+3. 右侧窗口拖到最下方，选择**API模式创建**：
+
+<img src="https://cdn.link-ai.tech/doc/20260316181215.png" width="800"/>
+
+4. 设置机器人名称、头像、可见范围，并选择**长连接模式**，记录下 **Bot ID** 和 **Secret** 信息后点击保存。
+
+## 二、配置和运行
+
+### 方式一：Web 控制台接入
+
+启动Cow项目后打开 Web 控制台 (本地链接为: http://127.0.0.1:9899/ )，选择 **通道** 菜单，点击 **接入通道**，选择 **企微智能机器人**，填写上一步保存的 Bot ID 和 Secret，点击接入即可。
+
+<img src="https://cdn.link-ai.tech/doc/20260316181711.png" width="800"/>
+
+### 方式二：配置文件接入
+
+在 `config.json` 中添加以下配置：
+
+```json
+{
+  "channel_type": "wecom_bot",
+  "wecom_bot_id": "YOUR_BOT_ID",
+  "wecom_bot_secret": "YOUR_SECRET"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `wecom_bot_id` | 智能机器人的 BotID |
+| `wecom_bot_secret` | 智能机器人的 Secret |
+
+配置完成后启动程序，日志显示 `[WecomBot] Subscribe success` 即表示连接成功。
+
+## 三、功能说明
+
+| 功能 | 支持情况 |
+| --- | --- |
+| 单聊 | ✅ |
+| 群聊（@机器人） | ✅ |
+| 文本消息 | ✅ 收发 |
+| 图片消息 | ✅ 收发 |
+| 文件消息 | ✅ 收发 |
+| 流式回复 | ✅ |
+| 定时任务主动推送 | ✅ |
+
+## 四、使用
+
+在企业微信中搜索创建的机器人名称，即可开始单聊对话。
+
+如需在企微内部群聊中使用，将机器人添加到群中，@机器人发送消息即可。
+
+<img src="https://cdn.link-ai.tech/doc/20260316182902.png" width="800"/>
--- a/docs/channels/wecom.mdx
+++ b/docs/channels/wecom.mdx
@@ -0,0 +1,98 @@
+---
+title: 企微自建应用
+description: 将 CowAgent 接入企业微信自建应用
+---
+
+通过企业微信自建应用接入 CowAgent，支持企业内部人员单聊使用。
+
+<Note>
+  企业微信只能使用 Docker 部署或服务器 Python 部署，不支持本地运行模式。
+</Note>
+
+## 一、准备
+
+需要的资源：
+
+1. 一台服务器（有公网 IP）
+2. 注册一个企业微信（个人也可注册，但无法认证）
+3. 认证企业微信还需要对应主体备案的域名
+
+## 二、创建企业微信应用
+
+1. 在 [企业微信管理后台](https://work.weixin.qq.com/wework_admin/frame#profile) 点击 **我的企业**，在最下方获取 **企业ID**（后续填写到 `wechatcom_corp_id` 字段中）。
+
+2. 切换到 **应用管理**，点击创建应用：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103156.png" width="480"/>
+
+3. 进入应用创建页面，记录 `AgentId` 和 `Secret`：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103218.png" width="580"/>
+
+4. 点击 **设置API接收**，配置应用接口：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103211.png" width="520"/>
+
+- URL 格式为 `http://ip:port/wxcomapp`（认证企业需使用备案域名）
+- 随机获取 `Token` 和 `EncodingAESKey` 并保存
+
+<Note>
+  此时保存 API 接收配置会失败，因为程序还未启动，等项目运行后再回来保存。
+</Note>
+
+## 三、配置和运行
+
+在 `config.json` 中添加以下配置（各参数与企业微信后台的对应关系见上方截图）：
+
+```json
+{
+  "channel_type": "wechatcom_app",
+  "single_chat_prefix": [""],
+  "wechatcom_corp_id": "YOUR_CORP_ID",
+  "wechatcomapp_token": "YOUR_TOKEN",
+  "wechatcomapp_secret": "YOUR_SECRET",
+  "wechatcomapp_agent_id": "YOUR_AGENT_ID",
+  "wechatcomapp_aes_key": "YOUR_AES_KEY",
+  "wechatcomapp_port": 9898
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `wechatcom_corp_id` | 企业 ID |
+| `wechatcomapp_token` | API 接收配置中的 Token |
+| `wechatcomapp_secret` | 应用的 Secret |
+| `wechatcomapp_agent_id` | 应用的 AgentId |
+| `wechatcomapp_aes_key` | API 接收配置中的 EncodingAESKey |
+| `wechatcomapp_port` | 监听端口，默认 9898 |
+
+配置完成后启动程序。当后台日志显示 `http://0.0.0.0:9898/` 时说明程序运行成功，需要将该端口对外开放（如在云服务器安全组中放行）。
+
+程序启动后，回到企业微信后台保存 **消息服务器配置**，保存成功后还需将服务器 IP 添加到 **企业可信IP** 中，否则无法收发消息：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103224.png" width="520"/>
+
+<Warning>
+  如遇到 URL 配置回调不通过或配置失败：
+  1. 确保服务器防火墙关闭且安全组放行监听端口
+  2. 仔细检查 Token、Secret Key 等参数配置是否一致，URL 格式是否正确
+  3. 认证企业微信需要配置与主体一致的备案域名
+</Warning>
+
+## 四、使用
+
+在企业微信中搜索刚创建的应用名称，即可直接对话：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103228.png" width="720"/>
+
+如需让外部个人微信用户使用，可在 **我的企业 → 微信插件** 中分享邀请关注二维码，个人微信扫码关注后即可与应用对话：
+
+<img src="https://cdn.link-ai.tech/doc/20260228103232.png" width="520"/>
+
+## 常见问题
+
+需要确保已安装以下依赖：
+
+```bash
+pip install websocket-client pycryptodome
+```
--- a/docs/docs.json
+++ b/docs/docs.json
@@ -0,0 +1,333 @@
+{
+  "$schema": "https://mintlify.com/docs.json",
+  "name": "CowAgent",
+  "description": "CowAgent - AI Super Assistant powered by LLMs, with autonomous task planning, long-term memory, skills system, and multi-channel deployment.",
+  "theme": "mint",
+  "appearance": {
+    "default": "light"
+  },
+  "colors": {
+    "primary": "#35A85B",
+    "light": "#4ABE6E",
+    "dark": "#228547"
+  },
+  "logo": {
+    "light": "/images/logo.jpg",
+    "dark": "/images/logo.jpg"
+  },
+  "favicon": "/images/favicon.ico",
+  "navbar": {
+    "links": [
+      {
+        "label": "官网",
+        "href": "https://cowagent.ai/"
+      },
+      {
+        "label": "GitHub",
+        "href": "https://github.com/zhayujie/chatgpt-on-wechat"
+      }
+    ]
+  },
+  "footer": {
+    "socials": {
+      "github": "https://github.com/zhayujie/chatgpt-on-wechat"
+    }
+  },
+  "navigation": {
+    "languages": [
+      {
+        "language": "zh",
+        "default": true,
+        "tabs": [
+          {
+            "tab": "项目介绍",
+            "groups": [
+              {
+                "group": "概览",
+                "pages": [
+                  "intro/index",
+                  "intro/architecture",
+                  "intro/features"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "快速开始",
+            "groups": [
+              {
+                "group": "安装部署",
+                "pages": [
+                  "guide/quick-start",
+                  "guide/manual-install",
+                  "guide/upgrade"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "模型",
+            "groups": [
+              {
+                "group": "模型配置",
+                "pages": [
+                  "models/index",
+                  "models/minimax",
+                  "models/glm",
+                  "models/qwen",
+                  "models/kimi",
+                  "models/doubao",
+                  "models/claude",
+                  "models/gemini",
+                  "models/openai",
+                  "models/deepseek",
+                  "models/linkai",
+                  "models/coding-plan"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "工具",
+            "groups": [
+              {
+                "group": "工具系统",
+                "pages": [
+                  "tools/index"
+                ]
+              },
+              {
+                "group": "内置工具",
+                "pages": [
+                  "tools/read",
+                  "tools/write",
+                  "tools/edit",
+                  "tools/ls",
+                  "tools/bash",
+                  "tools/send",
+                  "tools/memory",
+                  "tools/env-config"
+                ]
+              },
+              {
+                "group": "可选工具",
+                "pages": [
+                  "tools/web-search",
+                  "tools/scheduler"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "技能",
+            "groups": [
+              {
+                "group": "技能系统",
+                "pages": [
+                  "skills/index",
+                  "skills/skill-creator"
+                ]
+              },
+              {
+                "group": "内置技能",
+                "pages": [
+                  "skills/image-vision",
+                  "skills/linkai-agent",
+                  "skills/web-fetch"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "记忆",
+            "groups": [
+              {
+                "group": "记忆系统",
+                "pages": [
+                  "memory"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "通道",
+            "groups": [
+              {
+                "group": "接入渠道",
+                "pages": [
+                  "channels/web",
+                  "channels/feishu",
+                  "channels/dingtalk",
+                  "channels/wecom-bot",
+                  "channels/qq",
+                  "channels/wecom",
+                  "channels/wechatmp"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "版本",
+            "groups": [
+              {
+                "group": "发布记录",
+                "pages": [
+                  "releases/overview",
+                  "releases/v2.0.3",
+                  "releases/v2.0.2",
+                  "releases/v2.0.1",
+                  "releases/v2.0.0"
+                ]
+              }
+            ]
+          }
+        ]
+      },
+      {
+        "language": "en",
+        "tabs": [
+          {
+            "tab": "Introduction",
+            "groups": [
+              {
+                "group": "Overview",
+                "pages": [
+                  "en/intro/index",
+                  "en/intro/architecture",
+                  "en/intro/features"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Get Started",
+            "groups": [
+              {
+                "group": "Installation",
+                "pages": [
+                  "en/guide/quick-start",
+                  "en/guide/manual-install"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Models",
+            "groups": [
+              {
+                "group": "Model Configuration",
+                "pages": [
+                  "en/models/index",
+                  "en/models/minimax",
+                  "en/models/glm",
+                  "en/models/qwen",
+                  "en/models/kimi",
+                  "en/models/doubao",
+                  "en/models/claude",
+                  "en/models/gemini",
+                  "en/models/openai",
+                  "en/models/deepseek",
+                  "en/models/linkai",
+                  "en/models/coding-plan"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Tools",
+            "groups": [
+              {
+                "group": "Tools System",
+                "pages": [
+                  "en/tools/index"
+                ]
+              },
+              {
+                "group": "Built-in Tools",
+                "pages": [
+                  "en/tools/read",
+                  "en/tools/write",
+                  "en/tools/edit",
+                  "en/tools/ls",
+                  "en/tools/bash",
+                  "en/tools/send",
+                  "en/tools/memory",
+                  "en/tools/env-config"
+                ]
+              },
+              {
+                "group": "Optional Tools",
+                "pages": [
+                  "en/tools/web-search",
+                  "en/tools/scheduler"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Skills",
+            "groups": [
+              {
+                "group": "Skills System",
+                "pages": [
+                  "en/skills/index",
+                  "en/skills/skill-creator"
+                ]
+              },
+              {
+                "group": "Built-in Skills",
+                "pages": [
+                  "en/skills/image-vision",
+                  "en/skills/linkai-agent",
+                  "en/skills/web-fetch"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Memory",
+            "groups": [
+              {
+                "group": "Memory System",
+                "pages": [
+                  "en/memory"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Channels",
+            "groups": [
+              {
+                "group": "Platforms",
+                "pages": [
+                  "en/channels/web",
+                  "en/channels/feishu",
+                  "en/channels/dingtalk",
+                  "en/channels/wecom-bot",
+                  "en/channels/qq",
+                  "en/channels/wecom",
+                  "en/channels/wechatmp"
+                ]
+              }
+            ]
+          },
+          {
+            "tab": "Releases",
+            "groups": [
+              {
+                "group": "Release Notes",
+                "pages": [
+                  "en/releases/overview",
+                  "en/releases/v2.0.2",
+                  "en/releases/v2.0.1",
+                  "en/releases/v2.0.0"
+                ]
+              }
+            ]
+          }
+        ]
+      }
+    ]
+  }
+}
--- a/docs/en/README.md
+++ b/docs/en/README.md
@@ -0,0 +1,205 @@
+<p align="center"><img src="https://github.com/user-attachments/assets/eca9a9ec-8534-4615-9e0f-96c5ac1d10a3" alt="CowAgent" width="550" /></p>
+
+<p align="center">
+  <a href="https://github.com/zhayujie/chatgpt-on-wechat/releases/latest"><img src="https://img.shields.io/github/v/release/zhayujie/chatgpt-on-wechat" alt="Latest release"></a>
+  <a href="https://github.com/zhayujie/chatgpt-on-wechat/blob/master/LICENSE"><img src="https://img.shields.io/github/license/zhayujie/chatgpt-on-wechat" alt="License: MIT"></a>
+  <a href="https://github.com/zhayujie/chatgpt-on-wechat"><img src="https://img.shields.io/github/stars/zhayujie/chatgpt-on-wechat?style=flat-square" alt="Stars"></a> <br/>
+  [<a href="https://github.com/zhayujie/chatgpt-on-wechat/blob/master/README.md">中文</a>] | [English]
+</p>
+
+**CowAgent** is an AI super assistant powered by LLMs, capable of autonomous task planning, operating computers and external resources, creating and executing Skills, and continuously growing with long-term memory. It supports flexible model switching, handles text, voice, images, and files, and can be integrated into Web, Feishu, DingTalk, WeCom Bot, WeCom App, and WeChat Official Account — running 7×24 hours on your personal computer or server.
+
+<p align="center">
+  <a href="https://cowagent.ai/">🌐 Website</a> &nbsp;·&nbsp;
+  <a href="https://docs.cowagent.ai/en/intro/index">📖 Docs</a> &nbsp;·&nbsp;
+  <a href="https://docs.cowagent.ai/en/guide/quick-start">🚀 Quick Start</a> &nbsp;·&nbsp;
+  <a href="https://link-ai.tech/cowagent/create">☁️ Try Online</a>
+</p>
+
+## Introduction
+
+> CowAgent is both an out-of-the-box AI super assistant and a highly extensible Agent framework. You can extend it with new model interfaces, channels, built-in tools, and the Skills system to flexibly implement various customization needs.
+
+- ✅ **Autonomous Task Planning**: Understands complex tasks and autonomously plans execution, continuously thinking and invoking tools until goals are achieved. Supports accessing files, terminal, browser, schedulers, and other system resources via tools.
+- ✅ **Long-term Memory**: Automatically persists conversation memory to local files and databases, including core memory and daily memory, with keyword and vector retrieval support.
+- ✅ **Skills System**: Implements a Skills creation and execution engine with multiple built-in skills, and supports custom Skills development through natural language conversation.
+- ✅ **Multimodal Messages**: Supports parsing, processing, generating, and sending text, images, voice, files, and other message types.
+- ✅ **Multiple Model Support**: Supports OpenAI, Claude, Gemini, DeepSeek, MiniMax, GLM, Qwen, Kimi, Doubao, and other mainstream model providers.
+- ✅ **Multi-platform Deployment**: Runs on local computers or servers, integrable into Web, Feishu, DingTalk, WeChat Official Account, and WeCom applications.
+- ✅ **Knowledge Base**: Integrates enterprise knowledge base capabilities via the [LinkAI](https://link-ai.tech) platform.
+
+## Disclaimer
+
+1. This project follows the [MIT License](/LICENSE) and is intended for technical research and learning. Users must comply with local laws, regulations, policies, and corporate bylaws. Any illegal or rights-infringing use is prohibited.
+2. Agent mode consumes more tokens than normal chat mode. Choose models based on effectiveness and cost. Agent has access to the host OS — please deploy in trusted environments.
+3. CowAgent focuses on open-source development and does not participate in, authorize, or issue any cryptocurrency.
+
+## Demo
+
+Try online (no deployment needed): [CowAgent](https://link-ai.tech/cowagent/create)
+
+## Changelog
+
+> **2026.02.27:** [v2.0.2](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.2) — Web console overhaul (streaming chat, model/skill/memory/channel/scheduler/log management), multi-channel concurrent running, session persistence, new models including Gemini 3.1 Pro / Claude 4.6 Sonnet / Qwen3.5 Plus.
+
+> **2026.02.13:** [v2.0.1](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.1) — Built-in Web Search tool, smart context trimming, runtime info dynamic update, Windows compatibility, fixes for scheduler memory loss, Feishu connection issues, and more.
+
+> **2026.02.03:** [v2.0.0](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.0) — Full upgrade to AI super assistant with multi-step task planning, long-term memory, built-in tools, Skills framework, new models, and optimized channels.
+
+> **2025.05.23:** [v1.7.6](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.6) — Web channel optimization, AgentMesh multi-agent plugin, Baidu TTS, claude-4-sonnet/opus support.
+
+> **2025.04.11:** [v1.7.5](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.5) — wechatferry protocol, DeepSeek model, Tencent Cloud voice, ModelScope and Gitee-AI support.
+
+> **2024.12.13:** [v1.7.4](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.4) — Gemini 2.0 model, Web channel, memory leak fix.
+
+Full changelog: [Release Notes](https://docs.cowagent.ai/en/releases/overview)
+
+<br/>
+
+## 🚀 Quick Start
+
+The project provides a one-click script for installation, configuration, startup, and management:
+
+```bash
+bash <(curl -fsSL https://cdn.link-ai.tech/code/cow/run.sh)
+```
+
+After running, the Web service starts by default. Access `http://localhost:9899/chat` to chat.
+
+Script usage: [One-click Install](https://docs.cowagent.ai/en/guide/quick-start)
+
+### Manual Installation
+
+**1. Clone the project**
+
+```bash
+git clone https://github.com/zhayujie/chatgpt-on-wechat
+cd chatgpt-on-wechat/
+```
+
+**2. Install dependencies**
+
+```bash
+pip3 install -r requirements.txt
+pip3 install -r requirements-optional.txt   # optional but recommended
+```
+
+**3. Configure**
+
+```bash
+cp config-template.json config.json
+```
+
+Fill in your model API key and channel type in `config.json`. See the [configuration docs](https://docs.cowagent.ai/en/guide/manual-install) for details.
+
+**4. Run**
+
+```bash
+python3 app.py
+```
+
+For server background run:
+
+```bash
+nohup python3 app.py & tail -f nohup.out
+```
+
+### Docker Deployment
+
+```bash
+curl -O https://cdn.link-ai.tech/code/cow/docker-compose.yml
+# Edit docker-compose.yml with your config
+sudo docker compose up -d
+sudo docker logs -f chatgpt-on-wechat
+```
+
+<br/>
+
+## Models
+
+Supports mainstream model providers. Recommended models for Agent mode:
+
+| Provider | Recommended Model |
+| --- | --- |
+| MiniMax | `MiniMax-M2.5` |
+| GLM | `glm-5` |
+| Kimi | `kimi-k2.5` |
+| Doubao | `doubao-seed-2-0-code-preview-260215` |
+| Qwen | `qwen3.5-plus` |
+| Claude | `claude-sonnet-4-6` |
+| Gemini | `gemini-3.1-pro-preview` |
+| OpenAI | `gpt-5.4` |
+| DeepSeek | `deepseek-chat` |
+
+For detailed configuration of each model, see the [Models documentation](https://docs.cowagent.ai/en/models/index).
+
+### Coding Plan
+
+Coding Plan is a monthly subscription package offered by various providers, ideal for high-frequency Agent usage. All providers can be accessed via OpenAI-compatible mode:
+
+```json
+{
+  "bot_type": "openai",
+  "model": "MODEL_NAME",
+  "open_ai_api_base": "PROVIDER_CODING_PLAN_API_BASE",
+  "open_ai_api_key": "YOUR_API_KEY"
+}
+```
+
+- `bot_type`: Must be `openai`
+- `model`: Model name supported by the provider
+- `open_ai_api_base`: Provider's Coding Plan API Base (different from standard pay-as-you-go)
+- `open_ai_api_key`: Provider's Coding Plan API Key
+
+> Note: Coding Plan API Base and API Key are usually separate from standard pay-as-you-go ones. Please obtain them from each provider's platform.
+
+Supported providers include Alibaba Cloud, MiniMax, Zhipu GLM, Kimi, Volcengine, and more. For detailed configuration of each provider, see the [Coding Plan documentation](https://docs.cowagent.ai/en/models/coding-plan).
+
+<br/>
+
+## Channels
+
+Supports multiple platforms. Set `channel_type` in `config.json` to switch:
+
+| Channel | `channel_type` | Docs |
+| --- | --- | --- |
+| Web (default) | `web` | [Web Channel](https://docs.cowagent.ai/en/channels/web) |
+| Feishu | `feishu` | [Feishu Setup](https://docs.cowagent.ai/en/channels/feishu) |
+| DingTalk | `dingtalk` | [DingTalk Setup](https://docs.cowagent.ai/en/channels/dingtalk) |
+| WeCom Bot | `wecom_bot` | [WeCom Bot Setup](https://docs.cowagent.ai/en/channels/wecom-bot) |
+| WeCom App | `wechatcom_app` | [WeCom Setup](https://docs.cowagent.ai/en/channels/wecom) |
+| WeChat MP | `wechatmp` / `wechatmp_service` | [WeChat MP Setup](https://docs.cowagent.ai/en/channels/wechatmp) |
+| Terminal | `terminal` | — |
+
+Multiple channels can be enabled simultaneously, separated by commas: `"channel_type": "feishu,dingtalk"`.
+
+<br/>
+
+## Enterprise Services
+
+<a href="https://link-ai.tech" target="_blank"><img width="720" src="https://cdn.link-ai.tech/image/link-ai-intro.jpg"></a>
+
+> [LinkAI](https://link-ai.tech/) is a one-stop AI agent platform for enterprises and developers, integrating multimodal LLMs, knowledge bases, Agent plugins, and workflows. Supports one-click integration with mainstream platforms, SaaS and private deployment.
+
+<br/>
+
+## 🔗 Related Projects
+
+- [bot-on-anything](https://github.com/zhayujie/bot-on-anything): Lightweight and highly extensible LLM application framework supporting Slack, Telegram, Discord, Gmail, and more.
+- [AgentMesh](https://github.com/MinimalFuture/AgentMesh): Open-source Multi-Agent framework for complex problem solving through agent team collaboration.
+
+## 🔎 FAQ
+
+FAQs: <https://github.com/zhayujie/chatgpt-on-wechat/wiki/FAQs>
+
+## 🛠️ Contributing
+
+Welcome to add new channels, referring to the [Feishu channel](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/channel/feishu/feishu_channel.py) as an example. Also welcome to contribute new Skills, referring to the [Skill Creator docs](https://github.com/zhayujie/chatgpt-on-wechat/blob/master/skills/skill-creator/SKILL.md).
+
+## ✉ Contact
+
+Welcome to submit PRs and Issues, and support the project with a 🌟 Star. For questions, check the [FAQ list](https://github.com/zhayujie/chatgpt-on-wechat/wiki/FAQs) or search [Issues](https://github.com/zhayujie/chatgpt-on-wechat/issues).
+
+## 🌟 Contributors
+
+![cow contributors](https://contrib.rocks/image?repo=zhayujie/chatgpt-on-wechat&max=1000)
--- a/docs/en/channels/dingtalk.mdx
+++ b/docs/en/channels/dingtalk.mdx
@@ -0,0 +1,58 @@
+---
+title: DingTalk
+description: Integrate CowAgent into DingTalk application
+---
+
+Integrate CowAgent into DingTalk by creating an intelligent robot app on the DingTalk Open Platform.
+
+## 1. Create App
+
+1. Go to [DingTalk Developer Console](https://open-dev.dingtalk.com/fe/app#/corp/app), log in and click **Create App**, fill in the app information:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-create-app.png" width="800"/>
+
+2. Click **Add App Capability**, select **Robot** capability and click **Add**:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-add-bot.png" width="800"/>
+
+3. Configure the robot information and click **Publish**. After publishing, click "**Debug**" to automatically create a test group chat, which can be viewed in the client:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-config-bot.png" width="600"/>
+
+4. Click **Version Management & Release**, create a new version and publish:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-publish-bot.png" width="700"/>
+
+## 2. Project Configuration
+
+1. Click **Credentials & Basic Info**, get the `Client ID` and `Client Secret`:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-get-secret.png" width="700"/>
+
+2. Add the following configuration to `config.json` in the project root:
+
+```json
+{
+  "channel_type": "dingtalk",
+  "dingtalk_client_id": "YOUR_CLIENT_ID",
+  "dingtalk_client_secret": "YOUR_CLIENT_SECRET"
+}
+```
+
+3. Install the dependency:
+
+```bash
+pip3 install dingtalk_stream
+```
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-app-config.png" width="700"/>
+
+4. After starting the project, go to the DingTalk Developer Console, click **Event Subscription**, then click **Connection verified, verify channel**. When "**Connection successful**" is displayed, the configuration is complete:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-event-sub.png" width="700"/>
+
+## 3. Usage
+
+Chat privately with the robot or add it to an enterprise group to start a conversation:
+
+<img src="https://img-1317903499.cos.ap-guangzhou.myqcloud.com/docs/dingtalk-hosting-demo.png" width="650"/>
--- a/Show More
+++ b/Show More
				`@@ -0,0 +1 @@`
				`from agent.tools.vision.vision import Vision`