chore: bump version to 4.9.4 in pyproject.toml and __init__.py

fix(web): correct UTC timestamp parsing in monitoring panel
Backend serializes monitoring timestamps as naive ISO strings without timezone designator. JavaScript's new Date() treats such strings as local time, causing displayed times to be off by the user's UTC offset. Add parseUTCTimestamp() utility that appends 'Z' to ensure correct UTC interpretation.
2026-06-02 12:05:54 +00:00 · 2026-03-26 00:16:21 +08:00 · 2026-03-26 00:05:44 +08:00 · 2026-03-25 23:44:34 +08:00 · 2026-03-25 23:34:35 +08:00 · 2026-03-25 21:10:59 +08:00
192 changed files with 16124 additions and 5162 deletions
--- a/.github/ISSUE_TEMPLATE/bug-report.yml
+++ b/.github/ISSUE_TEMPLATE/bug-report.yml
@@ -19,7 +19,7 @@ body:
  - type: textarea
    attributes:
      label: 复现步骤
-      description: 提供越多信息，我们会越快解决问题，建议多提供配置截图；**如果你不认真填写（只一两句话概括），我们会很生气并且立即关闭 issue 或两年后才回复你**
+      description: 提供越多信息，我们会越快解决问题，建议多提供配置截图；**如果涉及 Dify、n8n、Langflow 等外部平台，请提供应用的导出文件（如 Dify 应用的 DSL），我们将更快回复您。**
    validations:
      required: false
  - type: textarea
--- a/.github/ISSUE_TEMPLATE/bug-report_en.yml
+++ b/.github/ISSUE_TEMPLATE/bug-report_en.yml
@@ -19,7 +19,7 @@ body:
  - type: textarea
    attributes:
      label: Reproduction steps
-      description: How to reproduce this problem, the more detailed the better; the more information you provide, the faster we will solve the problem. 【注意】请务必认真填写此部分，若不提供完整信息（如只有一两句话的概括），我们将不会回复！
+      description: How to reproduce this problem, the more detailed the better; the more information you provide, the faster we will solve the problem.
    validations:
      required: false
  - type: textarea
--- a/README.md
+++ b/README.md
@@ -1,49 +1,69 @@
-
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

-<a href="https://hellogithub.com/repository/langbot-app/LangBot" target="_blank"><img src="https://abroad.hellogithub.com/v1/widgets/recommend.svg?rid=5ce8ae2aa4f74316bf393b57b952433c&claim_uid=gtmc6YWjMZkT21R" alt="Featured｜HelloGitHub" style="width: 250px; height: 54px;" width="250" height="54" /></a>
+<a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>使用 LangBot 快速构建、调试、部署即时通信机器人。</h3>
+<h3>Production-grade platform for building agentic IM bots.</h3>
+<h4>Quickly build, debug, and ship AI bots to Slack, Discord, Telegram, WeChat, and more.</h4>

-[English](README_EN.md) / 简体中文 / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+English / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
-[![QQ Group](https://img.shields.io/badge/%E7%A4%BE%E5%8C%BAQQ%E7%BE%A4-1030838208-blue)](https://qm.qq.com/q/DxZZcNxM1W)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
-[![star](https://gitcode.com/RockChinQ/LangBot/star/badge.svg)](https://gitcode.com/RockChinQ/LangBot)
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

-<a href="https://langbot.app">项目主页</a> ｜
-<a href="https://docs.langbot.app/zh/insight/features.html">规格特性</a> ｜
-<a href="https://docs.langbot.app/zh/insight/guide.html">部署文档</a> ｜
-<a href="https://docs.langbot.app/zh/tags/readme.html">API 集成</a> ｜
-<a href="https://space.langbot.app">插件市场</a> ｜
-<a href="https://langbot.featurebase.app/roadmap">路线图</a>
+<a href="https://langbot.app">Website</a> ｜
+<a href="https://docs.langbot.app/en/insight/features">Features</a> ｜
+<a href="https://docs.langbot.app/en/insight/guide">Docs</a> ｜
+<a href="https://docs.langbot.app/en/tags/readme">API</a> ｜
+<a href="https://space.langbot.app/cloud">Cloud</a> ｜
+<a href="https://space.langbot.app">Plugin Market</a> ｜
+<a href="https://langbot.featurebase.app/roadmap">Roadmap</a>

 </div>

 </p>

+---

-## 📦 开始使用
+## What is LangBot?

-#### 快速部署
+LangBot is an **open-source, production-grade platform** for building AI-powered instant messaging bots. It connects Large Language Models (LLMs) to any chat platform, enabling you to create intelligent agents that can converse, execute tasks, and integrate with your existing workflows.

-使用 `uvx` 一键启动（需要先安装 [uv](https://docs.astral.sh/uv/getting-started/installation/)）：
+### Key Capabilities
+
+- **AI Conversations & Agents** — Multi-turn dialogues, tool calling, multi-modal support, streaming output. Built-in RAG (knowledge base) with deep integration to [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org).
+- **Universal IM Platform Support** — One codebase for Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK.
+- **Production-Ready** — Access control, rate limiting, sensitive word filtering, comprehensive monitoring, and exception handling. Trusted by enterprises.
+- **Plugin Ecosystem** — Hundreds of plugins, event-driven architecture, component extensions, and [MCP protocol](https://modelcontextprotocol.io/) support.
+- **Web Management Panel** — Configure, manage, and monitor your bots through an intuitive browser interface. No YAML editing required.
+- **Multi-Pipeline Architecture** — Different bots for different scenarios, with comprehensive monitoring and exception handling.
+
+[→ Learn more about all features](https://docs.langbot.app/en/insight/features)
+
+---
+
+## Quick Start
+
+### ☁️ LangBot Cloud (Recommended)
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — Zero deployment, ready to use.
+
+### One-Line Launch

 ```bash
 uvx langbot
 ```

-访问 http://localhost:5300 即可开始使用。
+> Requires [uv](https://docs.astral.sh/uv/getting-started/installation/). Visit http://localhost:5300 — done.

-#### Docker Compose 部署
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -51,127 +71,102 @@ cd LangBot/docker
 docker compose up -d
 ```

-访问 http://localhost:5300 即可开始使用。
-
-详细文档[Docker 部署](https://docs.langbot.app/zh/deploy/langbot/docker.html)。
-
-#### 宝塔面板部署
-
-已上架宝塔面板，若您已安装宝塔面板，可以根据[文档](https://docs.langbot.app/zh/deploy/langbot/one-click/bt.html)使用。
-
-#### Zeabur 云部署
-
-社区贡献的 Zeabur 模板。
-
-[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/zh-CN/templates/ZKTBDH)
-
-#### Railway 云部署
+### One-Click Cloud Deploy

+[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### 手动部署
+**More options:** [Docker](https://docs.langbot.app/en/deploy/langbot/docker) · [Manual](https://docs.langbot.app/en/deploy/langbot/manual) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt) · [Kubernetes](./docker/README_K8S.md)

-直接使用发行版运行，查看文档[手动部署](https://docs.langbot.app/zh/deploy/langbot/manual.html)。
+---

-#### Kubernetes 部署
+## Supported Platforms

-参考 [Kubernetes 部署](./docker/README_K8S.md) 文档。
-
-## 😎 保持更新
-
-点击仓库右上角 Star 和 Watch 按钮，获取最新动态。
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ 特性
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-zh-rounded.png" />
-
-
- 💬 大模型对话、Agent：支持多种大模型，适配群聊和私聊；具有多轮对话、工具调用、多模态、流式输出能力，自带 RAG（知识库）实现，并深度适配 [Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org)等 LLMOps 平台。
- 🤖 多平台支持：目前支持 QQ、QQ频道、企业微信、个人微信、飞书、Discord、Telegram、KOOK、Slack、LINE 等平台。
- 🛠️ 高稳定性、功能完备：原生支持访问控制、限速、敏感词过滤等机制；配置简单，支持多种部署方式。
- 🧩 插件扩展、活跃社区：高稳定性、高安全性的生产级插件系统，支持事件驱动、组件扩展等插件机制；适配 Anthropic [MCP 协议](https://modelcontextprotocol.io/)；目前已有数百个插件。
- 😻 Web 管理面板：提供先进的 WebUI 管理面板，用最直观的方式配置、管理、监控机器人。
- 📊 生产级特性：支持多流水线配置，不同机器人用于不同应用场景。具有全面的监控和异常处理能力。已被多家企业采用。
-
-详细规格特性请访问[文档](https://docs.langbot.app/zh/insight/features.html)。
-
-或访问 demo 环境：https://demo.langbot.dev/  
-  - 登录信息：邮箱：`demo@langbot.app` 密码：`langbot123456`
-  - 注意：仅展示 WebUI 效果，公开环境，请不要在其中填入您的任何敏感信息。
-
-### 消息平台
-
-| 平台 | 状态 | 备注 |
-| --- | --- | --- |
-| QQ 个人号 | ✅ | QQ 个人号私聊、群聊 |
-| QQ 官方机器人 | ✅ | QQ 官方机器人，支持频道、私聊、群聊 |
-| 企业微信 | ✅ |  |
-| 企微对外客服 | ✅ |  |
-| 企微智能机器人 | ✅ |  |
-| 个人微信 | ✅ |  |
-| 微信公众号 | ✅ |  |
-| 飞书 | ✅ |  |
-| 钉钉 | ✅ |  |
-| KOOK | ✅ |  |
+| Platform | Status | Notes |
+|----------|--------|-------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
+| QQ | ✅ | Personal & Official API |
+| WeCom | ✅ | Enterprise WeChat, External CS, AI Bot |
+| WeChat | ✅ | Personal & Official Account |
+| Lark | ✅ |  |
+| DingTalk | ✅ |  |
+| KOOK | ✅ |  |
+| Satori | ✅ |  |

-### 大模型能力
+---

-| 模型 | 状态 | 备注 |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | 可接入任何 OpenAI 接口格式模型 |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [智谱AI](https://open.bigmodel.cn/) | ✅ |  |
-| [胜算云](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | 全球大模型都可调用（友情推荐） |
-| [优云智算](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | 大模型和 GPU 资源平台 |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | 大模型和 GPU 资源平台 |
-| [接口 AI](https://jiekou.ai/) | ✅ | 大模型聚合平台，专注全球大模型接入 |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | 大模型聚合平台 |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | LLMOps 平台 |
-| [Ollama](https://ollama.com/) | ✅ | 本地大模型运行平台 |
-| [LMStudio](https://lmstudio.ai/) | ✅ | 本地大模型运行平台 |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | 大模型接口聚合平台 |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | 大模型聚合平台 |
-| [小马算力](https://www.tokenpony.cn/453z1) | ✅ | 大模型聚合平台 |
-| [阿里云百炼](https://bailian.console.aliyun.com/) | ✅ | 大模型聚合平台, LLMOps 平台 |
-| [火山方舟](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | 大模型聚合平台, LLMOps 平台 |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | 大模型聚合平台 |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | 支持通过 MCP 协议获取工具 |
-| [百宝箱Tbox](https://www.tbox.cn/open) | ✅ | 蚂蚁百宝箱智能体平台，每月免费10亿大模型Token |
+## Supported LLMs & Integrations

-### TTS
+| Provider | Type | Status |
+|----------|------|--------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | Local LLM | ✅ |
+| [LM Studio](https://lmstudio.ai/) | Local LLM | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | Protocol | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | Gateway | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | Gateway | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | Gateway | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | Gateway | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | Gateway | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | GPU Platform | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | GPU Platform | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | GPU Platform | ✅ |
+| [接口 AI](https://jiekou.ai/) | Gateway | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | Gateway | ✅ |

-| 平台/模型 | 备注 |
-| --- | --- |
-| [FishAudio](https://fish.audio/zh-CN/discovery/) | [插件](https://github.com/the-lazy-me/NewChatVoice) |
-| [海豚 AI](https://www.ttson.cn/?source=thelazy) | [插件](https://github.com/the-lazy-me/NewChatVoice) |
-| [AzureTTS](https://portal.azure.com/) | [插件](https://github.com/Ingnaryk/LangBot_AzureTTS) |
+[→ View all integrations](https://docs.langbot.app/en/insight/features)

-### 文生图
+---

-| 平台/模型 | 备注 |
-| --- | --- |
-| 阿里云百炼 | [插件](https://github.com/Thetail001/LangBot_BailianTextToImagePlugin)
+## Why LangBot?

-## 😘 社区贡献
+| Use Case | How LangBot Helps |
+|----------|-------------------|
+| **Customer Support** | Deploy AI agents to Slack/Discord/Telegram that answer questions using your knowledge base |
+| **Internal Tools** | Connect n8n/Dify workflows to WeCom/DingTalk for automated business processes |
+| **Community Management** | Moderate QQ/Discord groups with AI-powered content filtering and interaction |
+| **Multi-Platform Presence** | One bot, all platforms. Manage from a single dashboard |

-感谢以下[代码贡献者](https://github.com/langbot-app/LangBot/graphs/contributors)和社区里其他成员对 LangBot 的贡献：
+---
+
+## Live Demo
+
+**Try it now:** https://demo.langbot.dev/
+- Email: `demo@langbot.app`
+- Password: `langbot123456`
+
+*Note: Public demo environment. Do not enter sensitive information.*
+
+---
+
+## Community
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Discord Community](https://discord.gg/wdNEHETs87)
+
+---
+
+## Star History
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## Contributors
+
+Thanks to all [contributors](https://github.com/langbot-app/LangBot/graphs/contributors) who have helped make LangBot better:

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
 </a>
-
-<!--
-## For Code Agents
-
-If you're a code agent based on LLM, you should read [AGENTS.md](AGENTS.md) to understand the project structure and development process before you start any development.
-->
--- a/README_CN.md
+++ b/README_CN.md
@@ -0,0 +1,195 @@
+<p align="center">
+<a href="https://langbot.app">
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
+</a>
+
+<div align="center">
+
+<a href="https://hellogithub.com/repository/langbot-app/LangBot" target="_blank"><img src="https://abroad.hellogithub.com/v1/widgets/recommend.svg?rid=5ce8ae2aa4f74316bf393b57b952433c&claim_uid=gtmc6YWjMZkT21R" alt="Featured｜HelloGitHub" style="width: 250px; height: 54px;" width="250" height="54" /></a>
+
+<h3>生产级 AI 即时通信机器人开发平台。</h3>
+<h4>快速构建、调试和部署 AI 机器人到微信、QQ、飞书、Slack、Discord、Telegram 等平台。</h4>
+
+[English](README.md) / 简体中文 / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
+[![QQ Group](https://img.shields.io/badge/%E7%A4%BE%E5%8C%BAQQ%E7%BE%A4-1030838208-blue)](https://qm.qq.com/q/DxZZcNxM1W)
+[![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
+[![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
+<img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)
+[![star](https://gitcode.com/RockChinQ/LangBot/star/badge.svg)](https://gitcode.com/RockChinQ/LangBot)
+
+<a href="https://langbot.app">官网</a> ｜
+<a href="https://docs.langbot.app/zh/insight/features.html">特性</a> ｜
+<a href="https://docs.langbot.app/zh/insight/guide.html">文档</a> ｜
+<a href="https://docs.langbot.app/zh/tags/readme.html">API</a> ｜
+<a href="https://space.langbot.app/cloud">Cloud</a> ｜
+<a href="https://space.langbot.app">插件市场</a> ｜
+<a href="https://langbot.featurebase.app/roadmap">路线图</a>
+
+</div>
+
+</p>
+
+---
+
+LangBot 是一个**开源的生产级平台**，用于构建 AI 驱动的即时通信机器人。它将大语言模型（LLM）连接到各种聊天平台，帮助你创建能够对话、执行任务、并集成到现有工作流程中的智能 Agent。
+
+### 核心能力
+
+- **AI 对话与 Agent** — 多轮对话、工具调用、多模态、流式输出。自带 RAG（知识库），深度集成 [Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org) 等 LLMOps 平台。
+- **全平台支持** — 一套代码，覆盖 QQ、微信、企业微信、飞书、钉钉、Discord、Telegram、Slack、LINE、KOOK 等平台。
+- **生产就绪** — 访问控制、限速、敏感词过滤、全面监控与异常处理，已被多家企业采用。
+- **插件生态** — 数百个插件，跨进程的事件驱动架构，组件扩展，适配 [MCP 协议](https://modelcontextprotocol.io/)。
+- **Web 管理面板** — 通过浏览器直观地配置、管理和监控机器人，无需手动编辑配置文件。
+- **多流水线架构** — 不同机器人用于不同场景，具备全面的监控和异常处理能力。
+
+[→ 了解更多功能特性](https://docs.langbot.app/zh/insight/features.html)
+
+---
+
+## 快速开始
+
+### ☁️ LangBot Cloud（推荐）
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — 免部署，开箱即用。
+
+### 一键启动
+
+```bash
+uvx langbot
+```
+
+> 需要安装 [uv](https://docs.astral.sh/uv/getting-started/installation/)。访问 http://localhost:5300 即可使用。
+
+### Docker Compose
+
+```bash
+git clone https://github.com/langbot-app/LangBot
+cd LangBot/docker
+docker compose up -d
+```
+
+### 一键云部署
+
+[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/zh-CN/templates/ZKTBDH)
+[![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)
+
+**更多方式：** [Docker](https://docs.langbot.app/zh/deploy/langbot/docker.html) · [手动部署](https://docs.langbot.app/zh/deploy/langbot/manual.html) · [宝塔面板](https://docs.langbot.app/zh/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)
+
+---
+
+## 支持的平台
+
+| 平台 | 状态 | 备注 |
+|------|------|------|
+| QQ | ✅ | 个人号、官方机器人（频道、私聊、群聊） |
+| 微信 | ✅ | 个人微信、微信公众号 |
+| 企业微信 | ✅ | 应用消息、对外客服、智能机器人 |
+| 飞书 | ✅ |  |
+| 钉钉 | ✅ |  |
+| Discord | ✅ |  |
+| Telegram | ✅ |  |
+| Slack | ✅ |  |
+| LINE | ✅ |  |
+| KOOK | ✅ |  |
+
+---
+
+## 支持的大模型与集成
+
+| 提供商 | 类型 | 状态 |
+|--------|------|------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [智谱AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | 本地 LLM | ✅ |
+| [LM Studio](https://lmstudio.ai/) | 本地 LLM | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | 协议 | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | 聚合平台 | ✅ |
+| [阿里云百炼](https://bailian.console.aliyun.com/) | 聚合平台 | ✅ |
+| [火山方舟](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | 聚合平台 | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | 聚合平台 | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | 聚合平台 | ✅ |
+| [胜算云](https://www.shengsuanyun.com/?from=CH_KYIPP758) | GPU 平台 | ✅ |
+| [优云智算](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | GPU 平台 | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | GPU 平台 | ✅ |
+| [接口 AI](https://jiekou.ai/) | 聚合平台 | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | 聚合平台 | ✅ |
+| [小马算力](https://www.tokenpony.cn/453z1) | 聚合平台 | ✅ |
+| [百宝箱Tbox](https://www.tbox.cn/open) | 智能体平台 | ✅ |
+
+[→ 查看完整集成列表](https://docs.langbot.app/zh/insight/features.html)
+
+### TTS（语音合成）
+
+| 平台/模型 | 备注 |
+|-----------|------|
+| [FishAudio](https://fish.audio/zh-CN/discovery/) | [插件](https://github.com/the-lazy-me/NewChatVoice) |
+| [海豚 AI](https://www.ttson.cn/?source=thelazy) | [插件](https://github.com/the-lazy-me/NewChatVoice) |
+| [AzureTTS](https://portal.azure.com/) | [插件](https://github.com/Ingnaryk/LangBot_AzureTTS) |
+
+### 文生图
+
+| 平台/模型 | 备注 |
+|-----------|------|
+| 阿里云百炼 | [插件](https://github.com/Thetail001/LangBot_BailianTextToImagePlugin) |
+
+---
+
+## 为什么选择 LangBot？
+
+| 使用场景 | LangBot 如何帮助 |
+|----------|------------------|
+| **客户服务** | 将 AI Agent 部署到微信/企微/钉钉/飞书，基于知识库自动回答用户问题 |
+| **内部工具** | 将 n8n/Dify 工作流接入企微/钉钉，实现业务流程自动化 |
+| **社群运营** | 在 QQ/Discord 群中使用 AI 驱动的内容审核与智能互动 |
+| **多平台触达** | 一个机器人，覆盖所有平台。通过统一面板集中管理 |
+
+---
+
+## 在线演示
+
+**立即体验：** https://demo.langbot.dev/
+- 邮箱：`demo@langbot.app`
+- 密码：`langbot123456`
+
+*注意：公开演示环境，请不要在其中填入任何敏感信息。*
+
+---
+
+## 社区
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+[![QQ Group](https://img.shields.io/badge/%E7%A4%BE%E5%8C%BAQQ%E7%BE%A4-1030838208-blue)](https://qm.qq.com/q/DxZZcNxM1W)
+
+- [Discord 社区](https://discord.gg/wdNEHETs87)
+- [QQ 社区群](https://qm.qq.com/q/DxZZcNxM1W)
+
+---
+
+## Star 趋势
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## 贡献者
+
+感谢所有[贡献者](https://github.com/langbot-app/LangBot/graphs/contributors)对 LangBot 的帮助：
+
+<a href="https://github.com/langbot-app/LangBot/graphs/contributors">
+  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
+</a>
+
+<!--
+## For Code Agents
+
+If you're a code agent based on LLM, you should read [AGENTS.md](AGENTS.md) to understand the project structure and development process before you start any development.
+-->
--- a/README_EN.md
+++ b/README_EN.md
@@ -1,151 +0,0 @@
-<p align="center">
-<a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
-</a>
-
-<div align="center">
-
-<a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>
-
-<h3>Quickly build, debug, and ship IM bots with LangBot.</h3>
-
-English / [简体中文](README.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
-
-[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
-[![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
-[![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
-<img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
-
-<a href="https://langbot.app">Home</a> ｜
-<a href="https://docs.langbot.app/en/insight/features.html">Features</a> ｜
-<a href="https://docs.langbot.app/en/insight/guide.html">Deployment</a> ｜
-<a href="https://docs.langbot.app/en/tags/readme.html">API Integration</a> ｜
-<a href="https://space.langbot.app">Plugin Market</a> ｜
-<a href="https://langbot.featurebase.app/roadmap">Roadmap</a>
-
-</div>
-
-</p>
-
-
-## 📦 Getting Started
-
-#### Quick Start
-
-Use `uvx` to start with one command (need to install [uv](https://docs.astral.sh/uv/getting-started/installation/)):
-
-```bash
-uvx langbot
-```
-
-Visit http://localhost:5300 to start using it.
-
-#### Docker Compose Deployment
-
-```bash
-git clone https://github.com/langbot-app/LangBot
-cd LangBot/docker
-docker compose up -d
-```
-
-Visit http://localhost:5300 to start using it.
-
-Detailed documentation [Docker Deployment](https://docs.langbot.app/en/deploy/langbot/docker.html).
-
-#### One-click Deployment on BTPanel
-
-LangBot has been listed on the BTPanel, if you have installed the BTPanel, you can use the [document](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) to use it.
-
-#### Zeabur Cloud Deployment
-
-Community contributed Zeabur template.
-
-[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Railway Cloud Deployment
-
-[![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)
-
-#### Other Deployment Methods
-
-Directly use the released version to run, see the [Manual Deployment](https://docs.langbot.app/en/deploy/langbot/manual.html) documentation.
-
-#### Kubernetes Deployment
-
-Refer to the [Kubernetes Deployment](./docker/README_K8S.md) documentation.
-
-## 😎 Stay Ahead
-
-Click the Star and Watch button in the upper right corner of the repository to get the latest updates.
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ Features
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 Chat with LLM / Agent: Supports multiple LLMs, adapt to group chats and private chats; Supports multi-round conversations, tool calls, multi-modal, and streaming output capabilities. Built-in RAG (knowledge base) implementation, and deeply integrates with [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org) etc. LLMOps platforms.
- 🤖 Multi-platform Support: Currently supports QQ, QQ Channel, WeCom, personal WeChat, Lark, DingTalk, Discord, Telegram, KOOK, Slack, LINE, etc.
- 🛠️ High Stability, Feature-rich: Native access control, rate limiting, sensitive word filtering, etc. mechanisms; Easy to use, supports multiple deployment methods.
- 🧩 Plugin Extension, Active Community: High stability, high security production-level plugin system; Support event-driven, component extension, etc. plugin mechanisms; Integrate Anthropic [MCP protocol](https://modelcontextprotocol.io/); Currently has hundreds of plugins.
- 😻 Web UI: Support management LangBot instance through the browser. No need to manually write configuration files.
- 📊 Production-grade Features: Supports multiple pipeline configurations, different bots can be used for different scenarios. Has comprehensive monitoring and exception handling capabilities.
-
-For more detailed specifications, please refer to the [documentation](https://docs.langbot.app/en/insight/features.html).
-
-Or visit the demo environment: https://demo.langbot.dev/
-  - Login information: Email: `demo@langbot.app` Password: `langbot123456`
-  - Note: For WebUI demo only, please do not fill in any sensitive information in the public environment.
-
-### Message Platform
-
-| Platform | Status | Remarks |
-| --- | --- | --- |
-| Discord | ✅ |  |
-| Telegram | ✅ |  |
-| Slack | ✅ |  |
-| LINE | ✅ |  |
-| Personal QQ | ✅ |  |
-| QQ Official API | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| Personal WeChat | ✅ |  |
-| Lark | ✅ |  |
-| DingTalk | ✅ |  |
-| KOOK | ✅ |  |
-
-### LLMs
-
-| LLM | Status | Remarks |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | Available for any OpenAI interface format model |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | LLM and GPU resource platform |
-| [Dify](https://dify.ai) | ✅ | LLMOps platform |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | LLM and GPU resource platform |
-| [接口 AI](https://jiekou.ai/) | ✅ | LLM aggregation platform, dedicated to global LLMs |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | LLM and GPU resource platform |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | LLM gateway(MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Ollama](https://ollama.com/) | ✅ | Local LLM running platform |
-| [LMStudio](https://lmstudio.ai/) | ✅ | Local LLM running platform |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | LLM interface gateway(MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | LLM gateway(MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | LLM gateway(MaaS), LLMOps platform |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | LLM gateway(MaaS), LLMOps platform |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | LLM gateway(MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | Support tool access through MCP protocol |
-
-## 🤝 Community Contribution
-
-Thank you for the following [code contributors](https://github.com/langbot-app/LangBot/graphs/contributors) and other members in the community for their contributions to LangBot:
-
-<a href="https://github.com/langbot-app/LangBot/graphs/contributors">
-  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
-</a>
--- a/README_ES.md
+++ b/README_ES.md
@@ -1,25 +1,27 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

 <a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>Cree, depure y despliegue bots de mensajería instantánea rápidamente con LangBot.</h3>
+<h3>Plataforma de grado de producción para construir bots de mensajería instantánea con agentes de IA.</h3>
+<h4>Construya, depure y despliegue bots de IA rápidamente en Slack, Discord, Telegram, WeChat y más.</h4>

-[English](README_EN.md) / [简体中文](README.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / Español / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+[English](README.md) / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / Español / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

 <a href="https://langbot.app">Inicio</a> ｜
 <a href="https://docs.langbot.app/en/insight/features.html">Características</a> ｜
-<a href="https://docs.langbot.app/en/insight/guide.html">Despliegue</a> ｜
-<a href="https://docs.langbot.app/en/tags/readme.html">Integración API</a> ｜
+<a href="https://docs.langbot.app/en/insight/guide.html">Documentación</a> ｜
+<a href="https://docs.langbot.app/en/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">Mercado de Plugins</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">Hoja de Ruta</a>

@@ -27,20 +29,40 @@

 </p>

+---

-## 📦 Comenzar
+## ¿Qué es LangBot?

-#### Inicio Rápido
+LangBot es una **plataforma de código abierto y grado de producción** para construir bots de mensajería instantánea impulsados por IA. Conecta modelos de lenguaje de gran escala (LLMs) con cualquier plataforma de chat, permitiéndole crear agentes inteligentes que pueden conversar, ejecutar tareas e integrarse con sus flujos de trabajo existentes.

-Use `uvx` para iniciar con un comando (necesita instalar [uv](https://docs.astral.sh/uv/getting-started/installation/)):
+### Capacidades Clave
+
+- **Conversaciones e Agentes IA** — Diálogos de múltiples turnos, llamadas a herramientas, soporte multimodal, salida en streaming. RAG (base de conocimientos) incorporado con integración profunda con [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org).
+- **Soporte Universal de Plataformas de MI** — Un solo código base para Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK.
+- **Listo para Producción** — Control de acceso, limitación de velocidad, filtrado de palabras sensibles, monitoreo completo y manejo de excepciones. De confianza para empresas.
+- **Ecosistema de Plugins** — Cientos de plugins, arquitectura basada en eventos, extensiones de componentes y soporte del [protocolo MCP](https://modelcontextprotocol.io/).
+- **Panel de Gestión Web** — Configure, gestione y monitoree sus bots a través de una interfaz de navegador intuitiva. Sin necesidad de editar YAML.
+- **Arquitectura Multi-Pipeline** — Diferentes bots para diferentes escenarios, con monitoreo completo y manejo de excepciones.
+
+[→ Conocer más sobre todas las funcionalidades](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Inicio Rápido
+
+### ☁️ LangBot Cloud (Recomendado)
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — Sin despliegue, listo para usar.
+
+### Lanzamiento en una línea

 ```bash
 uvx langbot
 ```

-Visite http://localhost:5300 para comenzar a usarlo.
+> Requiere [uv](https://docs.astral.sh/uv/getting-started/installation/). Visite http://localhost:5300 — listo.

-#### Despliegue con Docker Compose
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -48,103 +70,101 @@ cd LangBot/docker
 docker compose up -d
 ```

-Visite http://localhost:5300 para comenzar a usarlo.
-
-Documentación detallada [Despliegue con Docker](https://docs.langbot.app/en/deploy/langbot/docker.html).
-
-#### Despliegue con un clic en BTPanel
-
-LangBot ha sido listado en BTPanel. Si tiene BTPanel instalado, puede usar la [documentación](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) para usarlo.
-
-#### Despliegue en la Nube Zeabur
-
-Plantilla de Zeabur contribuida por la comunidad.
+### Despliegue en la Nube con un Clic

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Despliegue en la Nube Railway
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### Otros Métodos de Despliegue
+**Más opciones:** [Docker](https://docs.langbot.app/en/deploy/langbot/docker.html) · [Manual](https://docs.langbot.app/en/deploy/langbot/manual.html) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-Use directamente la versión publicada para ejecutar, consulte la documentación de [Despliegue Manual](https://docs.langbot.app/en/deploy/langbot/manual.html).
+---

-#### Despliegue en Kubernetes
+## Plataformas Soportadas

-Consulte la documentación de [Despliegue en Kubernetes](./docker/README_K8S.md).
-
-## 😎 Manténgase Actualizado
-
-Haga clic en los botones Star y Watch en la esquina superior derecha del repositorio para obtener las últimas actualizaciones.
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ Características
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 Chat con LLM / Agent: Compatible con múltiples LLMs, adaptado para chats grupales y privados; Admite conversaciones de múltiples rondas, llamadas a herramientas, capacidades multimodales y de salida en streaming. Implementación RAG (base de conocimientos) incorporada, e integración profunda con [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org) etc. LLMOps platforms.
- 🤖 Soporte Multiplataforma: Actualmente compatible con QQ, QQ Channel, WeCom, WeChat personal, Lark, DingTalk, Discord, Telegram, KOOK, Slack, LINE, etc.
- 🛠️ Alta Estabilidad, Rico en Funciones: Control de acceso nativo, limitación de velocidad, filtrado de palabras sensibles, etc.; Fácil de usar, admite múltiples métodos de despliegue.
- 🧩 Extensión de Plugin, Comunidad Activa: Sistema de plugin de alta estabilidad, alta seguridad de nivel de producción; Compatible con mecanismos de plugin impulsados por eventos, extensión de componentes, etc.; Integración del protocolo [MCP](https://modelcontextprotocol.io/) de Anthropic; Actualmente cuenta con cientos de plugins.
- 😻 Interfaz Web: Admite la gestión de instancias de LangBot a través del navegador. No es necesario escribir archivos de configuración manualmente.
- 📊 Características de Nivel de Producción: Compatible con múltiples configuraciones de pipeline, diferentes bots para diferentes escenarios. Cuenta con capacidades completas de monitoreo y manejo de excepciones.
-
-Para especificaciones más detalladas, consulte la [documentación](https://docs.langbot.app/en/insight/features.html).
-
-O visite el entorno de demostración: https://demo.langbot.dev/
-  - Información de inicio de sesión: Correo electrónico: `demo@langbot.app` Contraseña: `langbot123456`
-  - Nota: Solo para demostración de WebUI, por favor no ingrese información confidencial en el entorno público.
-
-### Plataformas de Mensajería
-
-| Plataforma | Estado | Observaciones |
-| --- | --- | --- |
+| Plataforma | Estado | Notas |
+|----------|--------|-------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| QQ Personal | ✅ |  |
-| QQ API Oficial | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| WeChat Personal | ✅ |  |
+| QQ | ✅ | Personal y API Oficial |
+| WeCom | ✅ | WeChat Empresarial, CS Externo, AI Bot |
+| WeChat | ✅ | Personal y Cuenta Oficial |
 | Lark | ✅ |  |
 | DingTalk | ✅ |  |
 | KOOK | ✅ |  |
+| Satori | ✅ |  |

-### LLMs
+---

-| LLM | Estado | Observaciones |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | Disponible para cualquier modelo con formato de interfaz OpenAI |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | Plataforma de recursos LLM y GPU |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | Plataforma de recursos LLM y GPU |
-| [接口 AI](https://jiekou.ai/) | ✅ | Plataforma de agregación LLM |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | Plataforma de recursos LLM y GPU |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | Gateway LLM (MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | Plataforma LLMOps |
-| [Ollama](https://ollama.com/) | ✅ | Plataforma de ejecución de LLM local |
-| [LMStudio](https://lmstudio.ai/) | ✅ | Plataforma de ejecución de LLM local |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | Gateway de interfaz LLM (MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | Gateway LLM (MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | Gateway LLM (MaaS), plataforma LLMOps |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | Gateway LLM (MaaS), plataforma LLMOps |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | Gateway LLM (MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | Compatible con acceso a herramientas a través del protocolo MCP |
+## LLMs e Integraciones Soportadas

-## 🤝 Contribución de la Comunidad
+| Proveedor | Tipo | Estado |
+|----------|------|--------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | LLM Local | ✅ |
+| [LM Studio](https://lmstudio.ai/) | LLM Local | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | Protocolo | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | Pasarela | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | Pasarela | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | Pasarela | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | Pasarela | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | Pasarela | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | Plataforma GPU | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | Plataforma GPU | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | Plataforma GPU | ✅ |
+| [接口 AI](https://jiekou.ai/) | Pasarela | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | Pasarela | ✅ |

-Gracias a los siguientes [contribuidores de código](https://github.com/langbot-app/LangBot/graphs/contributors) y otros miembros de la comunidad por sus contribuciones a LangBot:
+[→ Ver todas las integraciones](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## ¿Por qué LangBot?
+
+| Caso de Uso | Cómo Ayuda LangBot |
+|----------|-------------------|
+| **Atención al cliente** | Despliegue agentes de IA en Slack/Discord/Telegram que respondan preguntas usando su base de conocimientos |
+| **Herramientas internas** | Conecte flujos de trabajo de n8n/Dify a WeCom/DingTalk para procesos empresariales automatizados |
+| **Gestión de comunidades** | Modere grupos de QQ/Discord con filtrado de contenido e interacción impulsados por IA |
+| **Presencia multiplataforma** | Un solo bot, todas las plataformas. Gestione desde un único panel de control |
+
+---
+
+## Demo en Vivo
+
+**Pruébelo ahora:** https://demo.langbot.dev/
+- Correo electrónico: `demo@langbot.app`
+- Contraseña: `langbot123456`
+
+*Nota: Entorno de demostración público. No ingrese información confidencial.*
+
+---
+
+## Comunidad
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Comunidad de Discord](https://discord.gg/wdNEHETs87)
+
+---
+
+## Historial de Stars
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## Colaboradores
+
+Gracias a todos los [colaboradores](https://github.com/langbot-app/LangBot/graphs/contributors) que han ayudado a mejorar LangBot:

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
--- a/README_FR.md
+++ b/README_FR.md
@@ -1,25 +1,27 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

 <a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>Créez, déboguez et déployez rapidement des bots de messagerie instantanée avec LangBot.</h3>
+<h3>Plateforme de niveau production pour construire des bots de messagerie instantanée avec agents IA.</h3>
+<h4>Créez, déboguez et déployez rapidement des bots IA sur Slack, Discord, Telegram, WeChat et plus.</h4>

-[English](README_EN.md) / [简体中文](README.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / Français / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+[English](README.md) / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / Français / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

 <a href="https://langbot.app">Accueil</a> ｜
 <a href="https://docs.langbot.app/en/insight/features.html">Fonctionnalités</a> ｜
-<a href="https://docs.langbot.app/en/insight/guide.html">Déploiement</a> ｜
-<a href="https://docs.langbot.app/en/tags/readme.html">Intégration API</a> ｜
+<a href="https://docs.langbot.app/en/insight/guide.html">Documentation</a> ｜
+<a href="https://docs.langbot.app/en/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">Marché des Plugins</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">Feuille de Route</a>

@@ -27,19 +29,40 @@

 </p>

-## 📦 Commencer
+---

-#### Démarrage Rapide
+## Qu'est-ce que LangBot ?

-Utilisez `uvx` pour démarrer avec une commande (besoin d'installer [uv](https://docs.astral.sh/uv/getting-started/installation/)) :
+LangBot est une **plateforme open-source de niveau production** pour créer des bots de messagerie instantanée alimentés par l'IA. Elle connecte les grands modèles de langage (LLMs) à n'importe quelle plateforme de chat, vous permettant de créer des agents intelligents capables de converser, d'exécuter des tâches et de s'intégrer à vos workflows existants.
+
+### Capacités Clés
+
+- **Conversations IA & Agents** — Dialogues multi-tours, appels d'outils, support multimodal, sortie en streaming. RAG (base de connaissances) intégré avec intégration profonde de [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org).
+- **Support Universel des Plateformes de MI** — Un seul code pour Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK.
+- **Prêt pour la Production** — Contrôle d'accès, limitation de débit, filtrage de mots sensibles, surveillance complète et gestion des exceptions. Approuvé par les entreprises.
+- **Écosystème de Plugins** — Des centaines de plugins, architecture événementielle, extensions de composants, et support du [protocole MCP](https://modelcontextprotocol.io/).
+- **Panneau de Gestion Web** — Configurez, gérez et surveillez vos bots via une interface navigateur intuitive. Aucune édition de YAML requise.
+- **Architecture Multi-Pipeline** — Différents bots pour différents scénarios, avec surveillance complète et gestion des exceptions.
+
+[→ En savoir plus sur toutes les fonctionnalités](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Démarrage Rapide
+
+### ☁️ LangBot Cloud (Recommandé)
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — Sans déploiement, prêt à utiliser.
+
+### Lancement en une ligne

 ```bash
 uvx langbot
 ```

-Visitez http://localhost:5300 pour commencer à l'utiliser.
+> Nécessite [uv](https://docs.astral.sh/uv/getting-started/installation/). Visitez http://localhost:5300 — c'est prêt.

-#### Déploiement avec Docker Compose
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -47,103 +70,101 @@ cd LangBot/docker
 docker compose up -d
 ```

-Visitez http://localhost:5300 pour commencer à l'utiliser.
-
-Documentation détaillée [Déploiement Docker](https://docs.langbot.app/en/deploy/langbot/docker.html).
-
-#### Déploiement en un clic sur BTPanel
-
-LangBot a été répertorié sur BTPanel. Si vous avez installé BTPanel, vous pouvez utiliser la [documentation](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) pour l'utiliser.
-
-#### Déploiement Cloud Zeabur
-
-Modèle Zeabur contribué par la communauté.
+### Déploiement Cloud en un Clic

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Déploiement Cloud Railway
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### Autres Méthodes de Déploiement
+**Plus d'options :** [Docker](https://docs.langbot.app/en/deploy/langbot/docker.html) · [Manuel](https://docs.langbot.app/en/deploy/langbot/manual.html) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-Utilisez directement la version publiée pour exécuter, consultez la documentation de [Déploiement Manuel](https://docs.langbot.app/en/deploy/langbot/manual.html).
+---

-#### Déploiement Kubernetes
+## Plateformes Supportées

-Consultez la documentation de [Déploiement Kubernetes](./docker/README_K8S.md).
-
-## 😎 Restez à Jour
-
-Cliquez sur les boutons Star et Watch dans le coin supérieur droit du dépôt pour obtenir les dernières mises à jour.
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ Fonctionnalités
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 Chat avec LLM / Agent : Prend en charge plusieurs LLM, adapté aux chats de groupe et privés ; Prend en charge les conversations multi-tours, les appels d'outils, les capacités multimodales et de sortie en streaming. Implémentation RAG (base de connaissances) intégrée, et intégration profonde avec [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org) etc. LLMOps platforms.
- 🤖 Support Multi-plateforme : Actuellement compatible avec QQ, QQ Channel, WeCom, WeChat personnel, Lark, DingTalk, Discord, Telegram, KOOK, Slack, LINE, etc.
- 🛠️ Haute Stabilité, Riche en Fonctionnalités : Contrôle d'accès natif, limitation de débit, filtrage de mots sensibles, etc. ; Facile à utiliser, prend en charge plusieurs méthodes de déploiement.
- 🧩 Extension de Plugin, Communauté Active : Système de plugin de haute stabilité, haute sécurité de niveau production; Prend en charge les mécanismes de plugin pilotés par événements, l'extension de composants, etc. ; Intégration du protocole [MCP](https://modelcontextprotocol.io/) d'Anthropic ; Dispose actuellement de centaines de plugins.
- 😻 Interface Web : Prend en charge la gestion des instances LangBot via le navigateur. Pas besoin d'écrire manuellement les fichiers de configuration.
- 📊 Fonctionnalités de Niveau Production : Prend en charge plusieurs configurations de pipeline, différents bots pour différents scénarios. Dispose de capacités complètes de surveillance et de gestion des exceptions.
-
-Pour des spécifications plus détaillées, veuillez consulter la [documentation](https://docs.langbot.app/en/insight/features.html).
-
-Ou visitez l'environnement de démonstration : https://demo.langbot.dev/
-  - Informations de connexion : Email : `demo@langbot.app` Mot de passe : `langbot123456`
-  - Note : Pour la démonstration WebUI uniquement, veuillez ne pas entrer d'informations sensibles dans l'environnement public.
-
-### Plateformes de Messagerie
-
-| Plateforme | Statut | Remarques |
-| --- | --- | --- |
+| Plateforme | Statut | Notes |
+|----------|--------|-------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| QQ Personnel | ✅ |  |
-| API Officielle QQ | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| WeChat Personnel | ✅ |  |
+| QQ | ✅ | Personnel & API Officielle |
+| WeCom | ✅ | WeChat Entreprise, CS Externe, AI Bot |
+| WeChat | ✅ | Personnel & Compte Officiel |
 | Lark | ✅ |  |
 | DingTalk | ✅ |  |
 | KOOK | ✅ |  |
+| Satori | ✅ |  |

-### LLMs
+---

-| LLM | Statut | Remarques |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | Disponible pour tout modèle au format d'interface OpenAI |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | Plateforme de ressources LLM et GPU |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | Plateforme de ressources LLM et GPU |
-| [接口 AI](https://jiekou.ai/) | ✅ | Plateforme d'agrégation LLM |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | Plateforme de ressources LLM et GPU |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | Passerelle LLM (MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | Plateforme LLMOps |
-| [Ollama](https://ollama.com/) | ✅ | Plateforme d'exécution LLM locale |
-| [LMStudio](https://lmstudio.ai/) | ✅ | Plateforme d'exécution LLM locale |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | Passerelle d'interface LLM (MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | Passerelle LLM (MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | Passerelle LLM (MaaS), plateforme LLMOps |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | Passerelle LLM (MaaS), plateforme LLMOps |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | Passerelle LLM (MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | Prend en charge l'accès aux outils via le protocole MCP |
+## LLMs et Intégrations Supportés

-## 🤝 Contribution de la Communauté
+| Fournisseur | Type | Statut |
+|----------|------|--------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | LLM Local | ✅ |
+| [LM Studio](https://lmstudio.ai/) | LLM Local | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | Protocole | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | Passerelle | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | Passerelle | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | Passerelle | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | Passerelle | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | Passerelle | ✅ |
+| [接口 AI](https://jiekou.ai/) | Passerelle | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | Passerelle | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | Plateforme GPU | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | Plateforme GPU | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | Plateforme GPU | ✅ |

-Merci aux [contributeurs de code](https://github.com/langbot-app/LangBot/graphs/contributors) suivants et aux autres membres de la communauté pour leurs contributions à LangBot :
+[→ Voir toutes les intégrations](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Pourquoi LangBot ?
+
+| Cas d'Usage | Comment LangBot Aide |
+|----------|-------------------|
+| **Support Client** | Déployez des agents IA sur Slack/Discord/Telegram qui répondent aux questions en utilisant votre base de connaissances |
+| **Outils Internes** | Connectez les workflows n8n/Dify à WeCom/DingTalk pour automatiser vos processus métier |
+| **Gestion de Communauté** | Modérez les groupes QQ/Discord avec un filtrage de contenu et des interactions alimentés par l'IA |
+| **Présence Multi-plateforme** | Un seul bot, toutes les plateformes. Gérez tout depuis un tableau de bord unique |
+
+---
+
+## Démo en Ligne
+
+**Essayez maintenant :** https://demo.langbot.dev/
+- Email : `demo@langbot.app`
+- Mot de passe : `langbot123456`
+
+*Note : Environnement de démonstration public. Ne saisissez pas d'informations sensibles.*
+
+---
+
+## Communauté
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Communauté Discord](https://discord.gg/wdNEHETs87)
+
+---
+
+## Historique des Stars
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## Contributeurs
+
+Merci à tous les [contributeurs](https://github.com/langbot-app/LangBot/graphs/contributors) qui ont aidé à améliorer LangBot :

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
--- a/README_JP.md
+++ b/README_JP.md
@@ -1,25 +1,27 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

 <a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>LangBotでIMボットを素早く構築、デバッグ、デプロイ。</h3>
+<h3>AIエージェント搭載IMボットを構築するための本番グレードプラットフォーム。</h3>
+<h4>Slack、Discord、Telegram、WeChat などに AI ボットを素早く構築、デバッグ、デプロイ。</h4>

-[English](README_EN.md) / [简体中文](README.md) / [繁體中文](README_TW.md) / 日本語 / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+[English](README.md) / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / 日本語 / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

 <a href="https://langbot.app">ホーム</a> ｜
-<a href="https://docs.langbot.app/ja/insight/features.html">機能仕様</a> ｜
-<a href="https://docs.langbot.app/ja/insight/guide.html">デプロイ</a> ｜
-<a href="https://docs.langbot.app/ja/tags/readme.html">API統合</a> ｜
+<a href="https://docs.langbot.app/ja/insight/features.html">機能</a> ｜
+<a href="https://docs.langbot.app/ja/insight/guide.html">ドキュメント</a> ｜
+<a href="https://docs.langbot.app/ja/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">プラグインマーケット</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">ロードマップ</a>

@@ -27,19 +29,40 @@

 </p>

-## 📦 始め方
+---

-#### クイックスタート
+## LangBot とは？

-`uvx` を使用した迅速なデプロイ（[uv](https://docs.astral.sh/uv/getting-started/installation/) が必要です）：
+LangBot は、AI搭載のインスタントメッセージングボットを構築するための**オープンソースの本番グレードプラットフォーム**です。大規模言語モデル（LLM）をあらゆるチャットプラットフォームに接続し、会話、タスク実行、既存のワークフローとの統合が可能なインテリジェントエージェントを作成できます。
+
+### 主な機能
+
+- **AI対話とエージェント** — マルチターン対話、ツール呼び出し、マルチモーダル対応、ストリーミング出力。RAG（ナレッジベース）を内蔵し、[Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org) と深く統合。
+- **ユニバーサルIMプラットフォーム対応** — 単一のコードベースで Discord、Telegram、Slack、LINE、QQ、WeChat、WeCom、Lark、DingTalk、KOOK に対応。
+- **本番環境対応** — アクセス制御、レート制限、センシティブワードフィルタリング、包括的な監視、例外処理を搭載。エンタープライズの信頼に応える品質。
+- **プラグインエコシステム** — 数百のプラグイン、イベント駆動アーキテクチャ、コンポーネント拡張、[MCPプロトコル](https://modelcontextprotocol.io/)対応。
+- **Web管理パネル** — 直感的なブラウザインターフェースからボットの設定、管理、監視が可能。YAML編集は不要。
+- **マルチパイプラインアーキテクチャ** — 異なるシナリオに異なるボットを配置し、包括的な監視と例外処理を実現。
+
+[→ すべての機能について詳しく見る](https://docs.langbot.app/ja/insight/features.html)
+
+---
+
+## クイックスタート
+
+### ☁️ LangBot Cloud（推奨）
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — デプロイ不要、すぐに使えます。
+
+### ワンライン起動

 ```bash
 uvx langbot
 ```

-http://localhost:5300 にアクセスして使用を開始します。
+> [uv](https://docs.astral.sh/uv/getting-started/installation/) が必要です。http://localhost:5300 にアクセスして完了。

-#### Docker Compose デプロイ
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -47,103 +70,101 @@ cd LangBot/docker
 docker compose up -d
 ```

-http://localhost:5300 にアクセスして使用を開始します。
-
-詳細なドキュメントは[Dockerデプロイ](https://docs.langbot.app/en/deploy/langbot/docker.html)を参照してください。
-
-#### Panelでのワンクリックデプロイ
-
-LangBotはBTPanelにリストされています。BTPanelをインストールしている場合は、[ドキュメント](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html)を使用して使用できます。
-
-#### Zeaburクラウドデプロイ
-
-コミュニティが提供するZeaburテンプレート。
+### ワンクリッククラウドデプロイ

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Railwayクラウドデプロイ
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### その他のデプロイ方法
+**その他:** [Docker](https://docs.langbot.app/en/deploy/langbot/docker.html) · [手動デプロイ](https://docs.langbot.app/en/deploy/langbot/manual.html) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-リリースバージョンを直接使用して実行します。[手動デプロイ](https://docs.langbot.app/en/deploy/langbot/manual.html)のドキュメントを参照してください。
+---

-#### Kubernetes デプロイ
-
-[Kubernetes デプロイ](./docker/README_K8S.md) ドキュメントを参照してください。
-
-## 😎 最新情報を入手
-
-リポジトリの右上にある Star と Watch ボタンをクリックして、最新の更新を取得してください。
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ 機能
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 LLM / エージェントとのチャット: 複数のLLMをサポートし、グループチャットとプライベートチャットに対応。マルチラウンドの会話、ツールの呼び出し、マルチモーダル、ストリーミング出力機能をサポート、RAG（知識ベース）を組み込み、[Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org)などの LLMOps プラットフォームと深く統合。
- 🤖 多プラットフォーム対応: 現在、QQ、QQ チャンネル、WeChat、個人 WeChat、Lark、DingTalk、Discord、Telegram、KOOK、Slack、LINE など、複数のプラットフォームをサポートしています。
- 🛠️ 高い安定性、豊富な機能: ネイティブのアクセス制御、レート制限、敏感な単語のフィルタリングなどのメカニズムをサポート。使いやすく、複数のデプロイ方法をサポート。
- 🧩 プラグイン拡張、活発なコミュニティ: 高い安定性、高いセキュリティの生産レベルのプラグインシステム；イベント駆動、コンポーネント拡張などのプラグインメカニズムをサポート。適配 Anthropic [MCP プロトコル](https://modelcontextprotocol.io/)；豊富なエコシステム、現在数百のプラグインが存在。
- 😻 Web UI: ブラウザを通じてLangBotインスタンスを管理することをサポート。
- 📊 生産レベルの機能: 複数のパイプライン設定をサポートし、異なるボットを異なる用途に使用できます。包括的な監視と例外処理機能を備えています。
-
-詳細な仕様については、[ドキュメント](https://docs.langbot.app/en/insight/features.html)を参照してください。
-
-または、デモ環境にアクセスしてください: https://demo.langbot.dev/
-  - ログイン情報: メール: `demo@langbot.app` パスワード: `langbot123456`
-  - 注意: WebUI のデモンストレーションのみの場合、公開環境では機密情報を入力しないでください。
-
-### メッセージプラットフォーム
+## 対応プラットフォーム

 | プラットフォーム | ステータス | 備考 |
-| --- | --- | --- |
+|----------|--------|-------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| 個人QQ | ✅ |  |
-| QQ公式API | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| 個人WeChat | ✅ | |
+| QQ | ✅ | 個人 & 公式API |
+| WeCom | ✅ | 企業WeChat、外部CS、AIボット |
+| WeChat | ✅ | 個人 & 公式アカウント |
 | Lark | ✅ |  |
 | DingTalk | ✅ |  |
 | KOOK | ✅ |  |
+| Satori | ✅ |  |

-### LLMs
+---

-| LLM | ステータス | 備考 |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | 任意のOpenAIインターフェース形式モデルに対応 |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | 大模型とGPUリソースプラットフォーム |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | 大模型とGPUリソースプラットフォーム |
-| [接口 AI](https://jiekou.ai/) | ✅ | LLMゲートウェイ(MaaS) |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | LLMとGPUリソースプラットフォーム |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | LLMゲートウェイ(MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | LLMOpsプラットフォーム |
-| [Ollama](https://ollama.com/) | ✅ | ローカルLLM実行プラットフォーム |
-| [LMStudio](https://lmstudio.ai/) | ✅ | ローカルLLM実行プラットフォーム |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | LLMインターフェースゲートウェイ(MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | LLMゲートウェイ(MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | LLMゲートウェイ(MaaS), LLMOpsプラットフォーム |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | LLMゲートウェイ(MaaS), LLMOpsプラットフォーム |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | LLMゲートウェイ(MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | MCPプロトコルをサポート |
+## 対応LLMと統合

-## 🤝 コミュニティ貢献
+| プロバイダー | タイプ | ステータス |
+|----------|------|--------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | ローカルLLM | ✅ |
+| [LM Studio](https://lmstudio.ai/) | ローカルLLM | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | プロトコル | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | ゲートウェイ | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ゲートウェイ | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ゲートウェイ | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ゲートウェイ | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | ゲートウェイ | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | GPUプラットフォーム | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | GPUプラットフォーム | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | GPUプラットフォーム | ✅ |
+| [接口 AI](https://jiekou.ai/) | ゲートウェイ | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | ゲートウェイ | ✅ |

-LangBot への貢献に対して、以下の [コード貢献者](https://github.com/langbot-app/LangBot/graphs/contributors) とコミュニティの他のメンバーに感謝します。
+[→ すべての統合を表示](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## なぜ LangBot？
+
+| ユースケース | LangBot の活用方法 |
+|----------|-------------------|
+| **カスタマーサポート** | ナレッジベースを活用して質問に回答するAIエージェントをSlack/Discord/Telegramにデプロイ |
+| **社内ツール** | n8n/Difyのワークフローを WeCom/DingTalk に接続し、業務プロセスを自動化 |
+| **コミュニティ管理** | AI搭載のコンテンツフィルタリングとインタラクションでQQ/Discordグループをモデレーション |
+| **マルチプラットフォーム展開** | 1つのボットで全プラットフォームに対応。単一のダッシュボードから管理 |
+
+---
+
+## ライブデモ
+
+**今すぐ試す:** https://demo.langbot.dev/
+- メール: `demo@langbot.app`
+- パスワード: `langbot123456`
+
+*注意: 公開デモ環境です。機密情報を入力しないでください。*
+
+---
+
+## コミュニティ
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Discord コミュニティ](https://discord.gg/wdNEHETs87)
+
+---
+
+## Star 推移
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## コントリビューター
+
+LangBot をより良くするために貢献してくださったすべての[コントリビューター](https://github.com/langbot-app/LangBot/graphs/contributors)に感謝します:

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
--- a/README_KO.md
+++ b/README_KO.md
@@ -1,25 +1,27 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

 <a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>LangBot으로 IM 봇을 빠르게 구축, 디버그 및 배포하세요.</h3>
+<h3>AI 에이전트 IM 봇 구축을 위한 프로덕션 등급 플랫폼.</h3>
+<h4>Slack, Discord, Telegram, WeChat 등에 AI 봇을 빠르게 구축, 디버그 및 배포.</h4>

-[English](README_EN.md) / [简体中文](README.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / 한국어 / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+[English](README.md) / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / 한국어 / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

 <a href="https://langbot.app">홈</a> ｜
-<a href="https://docs.langbot.app/en/insight/features.html">기능 사양</a> ｜
-<a href="https://docs.langbot.app/en/insight/guide.html">배포</a> ｜
-<a href="https://docs.langbot.app/en/tags/readme.html">API 통합</a> ｜
+<a href="https://docs.langbot.app/en/insight/features.html">기능</a> ｜
+<a href="https://docs.langbot.app/en/insight/guide.html">문서</a> ｜
+<a href="https://docs.langbot.app/en/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">플러그인 마켓</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">로드맵</a>

@@ -27,19 +29,40 @@

 </p>

-## 📦 시작하기
+---

-#### 빠른 시작
+## LangBot이란?

-`uvx`를 사용하여 한 명령으로 시작하세요 ([uv](https://docs.astral.sh/uv/getting-started/installation/) 설치 필요):
+LangBot은 AI 기반 인스턴트 메시징 봇을 구축하기 위한 **오픈소스 프로덕션 등급 플랫폼**입니다. 대규모 언어 모델(LLM)을 모든 채팅 플랫폼에 연결하여 대화, 작업 실행, 기존 워크플로우와의 통합이 가능한 지능형 에이전트를 만들 수 있습니다.
+
+### 핵심 기능
+
+- **AI 대화 및 에이전트** — 멀티턴 대화, 도구 호출, 멀티모달 지원, 스트리밍 출력. 내장 RAG(지식 베이스)와 [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org) 심층 통합.
+- **유니버설 IM 플랫폼 지원** — 단일 코드베이스로 Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK 지원.
+- **프로덕션 레디** — 접근 제어, 속도 제한, 민감어 필터링, 종합 모니터링 및 예외 처리. 기업 환경에서 검증됨.
+- **플러그인 생태계** — 수백 개의 플러그인, 이벤트 기반 아키텍처, 컴포넌트 확장, [MCP 프로토콜](https://modelcontextprotocol.io/) 지원.
+- **웹 관리 패널** — 직관적인 브라우저 인터페이스로 봇을 구성, 관리 및 모니터링. YAML 편집 불필요.
+- **멀티 파이프라인 아키텍처** — 다양한 시나리오에 맞는 다양한 봇 구성, 종합 모니터링 및 예외 처리.
+
+[→ 모든 기능 자세히 보기](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## 빠른 시작
+
+### ☁️ LangBot Cloud (추천)
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — 배포 없이 바로 사용.
+
+### 원라인 실행

 ```bash
 uvx langbot
 ```

-http://localhost:5300을 방문하여 사용을 시작하세요.
+> [uv](https://docs.astral.sh/uv/getting-started/installation/) 설치 필요. http://localhost:5300 방문 — 완료.

-#### Docker Compose 배포
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -47,103 +70,101 @@ cd LangBot/docker
 docker compose up -d
 ```

-http://localhost:5300을 방문하여 사용을 시작하세요.
-
-자세한 문서는 [Docker 배포](https://docs.langbot.app/en/deploy/langbot/docker.html)를 참조하세요.
-
-#### BTPanel 원클릭 배포
-
-LangBot은 BTPanel에 등록되어 있습니다. BTPanel을 설치한 경우 [문서](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html)를 사용하여 사용할 수 있습니다.
-
-#### Zeabur 클라우드 배포
-
-커뮤니티에서 제공하는 Zeabur 템플릿입니다.
+### 원클릭 클라우드 배포

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Railway 클라우드 배포
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### 기타 배포 방법
+**더 많은 옵션:** [Docker](https://docs.langbot.app/en/deploy/langbot/docker.html) · [수동 배포](https://docs.langbot.app/en/deploy/langbot/manual.html) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-릴리스 버전을 직접 사용하여 실행하려면 [수동 배포](https://docs.langbot.app/en/deploy/langbot/manual.html) 문서를 참조하세요.
+---

-#### Kubernetes 배포
-
-[Kubernetes 배포](./docker/README_K8S.md) 문서를 참조하세요.
-
-## 😎 최신 정보 받기
-
-리포지토리 오른쪽 상단의 Star 및 Watch 버튼을 클릭하여 최신 업데이트를 받으세요.
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ 기능
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 LLM / Agent와 채팅: 여러 LLM을 지원하며 그룹 채팅 및 개인 채팅에 적응; 멀티 라운드 대화, 도구 호출, 멀티모달, 스트리밍 출력 기능을 지원합니다. 내장된 RAG(지식 베이스) 구현 및 [Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org)등의 LLMOps 플랫폼과 깊이 통합됩니다.
- 🤖 다중 플랫폼 지원: 현재 QQ, QQ Channel, WeCom, 개인 WeChat, Lark, DingTalk, Discord, Telegram, KOOK, Slack, LINE 등을 지원합니다.
- 🛠️ 높은 안정성, 풍부한 기능: 네이티브 액세스 제어, 속도 제한, 민감한 단어 필터링 등의 메커니즘; 사용하기 쉽고 여러 배포 방법을 지원합니다.
- 🧩 플러그인 확장, 활발한 커뮤니티: 고안정성, 고보안 생산 수준의 플러그인 시스템; 이벤트 기반, 컴포넌트 확장 등의 플러그인 메커니즘을 지원; Anthropic [MCP 프로토콜](https://modelcontextprotocol.io/) 통합; 현재 수백 개의 플러그인이 있습니다.
- 😻 웹 UI: 브라우저를 통해 LangBot 인스턴스 관리를 지원합니다. 구성 파일을 수동으로 작성할 필요가 없습니다.
- 📊 생산 수준의 기능: 여러 파이프라인 구성을 지원하며 다양한 시나리오에 대해 다른 봇을 사용할 수 있습니다. 포괄적인 모니터링 및 예외 처리 기능을 갖추고 있습니다.
-
-더 자세한 사양은 [문서](https://docs.langbot.app/en/insight/features.html)를 참조하세요.
-
-또는 데모 환경을 방문하세요: https://demo.langbot.dev/
-  - 로그인 정보: 이메일: `demo@langbot.app` 비밀번호: `langbot123456`
-  - 참고: WebUI 데모 전용이므로 공개 환경에서는 민감한 정보를 입력하지 마세요.
-
-### 메시징 플랫폼
+## 지원 플랫폼

 | 플랫폼 | 상태 | 비고 |
-| --- | --- | --- |
+|--------|------|------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| 개인 QQ | ✅ |  |
-| QQ 공식 API | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| 개인 WeChat | ✅ |  |
-| KOOK | ✅ |  |
+| QQ | ✅ | 개인 및 공식 API |
+| WeCom | ✅ | 기업 WeChat, 외부 CS, AI Bot |
+| WeChat | ✅ | 개인 및 공식 계정 |
 | Lark | ✅ |  |
 | DingTalk | ✅ |  |
+| KOOK | ✅ |  |
+| Satori | ✅ |  |

-### LLMs
+---

-| LLM | 상태 | 비고 |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | 모든 OpenAI 인터페이스 형식 모델에 사용 가능 |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | LLM 및 GPU 리소스 플랫폼 |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | LLM 및 GPU 리소스 플랫폼 |
-| [接口 AI](https://jiekou.ai/) | ✅ | LLM 집계 플랫폼 |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | LLM 및 GPU 리소스 플랫폼 |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | LLM 게이트웨이(MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | LLMOps 플랫폼 |
-| [Ollama](https://ollama.com/) | ✅ | 로컬 LLM 실행 플랫폼 |
-| [LMStudio](https://lmstudio.ai/) | ✅ | 로컬 LLM 실행 플랫폼 |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | LLM 인터페이스 게이트웨이(MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | LLM 게이트웨이(MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | LLM 게이트웨이(MaaS), LLMOps 플랫폼 |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | LLM 게이트웨이(MaaS), LLMOps 플랫폼 |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | LLM 게이트웨이(MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | MCP 프로토콜을 통한 도구 액세스 지원 |
+## 지원 LLM 및 통합

-## 🤝 커뮤니티 기여
+| 제공자 | 유형 | 상태 |
+|--------|------|------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | 로컬 LLM | ✅ |
+| [LM Studio](https://lmstudio.ai/) | 로컬 LLM | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | 프로토콜 | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | 게이트웨이 | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | 게이트웨이 | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | 게이트웨이 | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | 게이트웨이 | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | 게이트웨이 | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | GPU 플랫폼 | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | GPU 플랫폼 | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | GPU 플랫폼 | ✅ |
+| [接口 AI](https://jiekou.ai/) | 게이트웨이 | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | 게이트웨이 | ✅ |

-다음 [코드 기여자](https://github.com/langbot-app/LangBot/graphs/contributors) 및 커뮤니티의 다른 구성원들의 LangBot 기여에 감사드립니다:
+[→ 모든 통합 보기](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## 왜 LangBot인가?
+
+| 사용 사례 | LangBot 활용 방법 |
+|-----------|-------------------|
+| **고객 지원** | 지식 베이스를 활용하여 질문에 답변하는 AI 에이전트를 Slack/Discord/Telegram에 배포 |
+| **내부 도구** | n8n/Dify 워크플로우를 WeCom/DingTalk에 연결하여 비즈니스 프로세스 자동화 |
+| **커뮤니티 관리** | AI 기반 콘텐츠 필터링 및 상호작용으로 QQ/Discord 그룹 관리 |
+| **멀티 플랫폼** | 하나의 봇으로 모든 플랫폼 지원. 단일 대시보드에서 관리 |
+
+---
+
+## 라이브 데모
+
+**지금 체험:** https://demo.langbot.dev/
+- 이메일: `demo@langbot.app`
+- 비밀번호: `langbot123456`
+
+*참고: 공개 데모 환경입니다. 민감한 정보를 입력하지 마세요.*
+
+---
+
+## 커뮤니티
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Discord 커뮤니티](https://discord.gg/wdNEHETs87)
+
+---
+
+## Star 추이
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## 기여자
+
+LangBot을 더 나은 프로젝트로 만들어 주신 모든 [기여자](https://github.com/langbot-app/LangBot/graphs/contributors)분들께 감사드립니다:

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
--- a/README_RU.md
+++ b/README_RU.md
@@ -1,25 +1,27 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

 <a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>Быстро создавайте, отлаживайте и развертывайте IM-ботов с LangBot.</h3>
+<h3>Платформа производственного уровня для создания агентных IM-ботов.</h3>
+<h4>Быстро создавайте, отлаживайте и развертывайте ИИ-ботов в Slack, Discord, Telegram, WeChat и других платформах.</h4>

-[English](README_EN.md) / [简体中文](README.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / Русский / [Tiếng Việt](README_VI.md)
+[English](README.md) / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / Русский / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

 <a href="https://langbot.app">Главная</a> ｜
-<a href="https://docs.langbot.app/en/insight/features.html">Характеристики</a> ｜
-<a href="https://docs.langbot.app/en/insight/guide.html">Развертывание</a> ｜
-<a href="https://docs.langbot.app/en/tags/readme.html">Интеграция API</a> ｜
+<a href="https://docs.langbot.app/en/insight/features.html">Возможности</a> ｜
+<a href="https://docs.langbot.app/en/insight/guide.html">Документация</a> ｜
+<a href="https://docs.langbot.app/en/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">Магазин плагинов</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">Дорожная карта</a>

@@ -27,19 +29,40 @@

 </p>

-## 📦 Начало работы
+---

-#### Быстрый старт
+## Что такое LangBot?

-Используйте `uvx` для запуска одной командой (требуется установка [uv](https://docs.astral.sh/uv/getting-started/installation/)):
+LangBot — это **платформа с открытым исходным кодом производственного уровня** для создания ИИ-ботов в мессенджерах. Она связывает большие языковые модели (LLM) с любой чат-платформой, позволяя создавать интеллектуальных агентов, которые могут вести диалоги, выполнять задачи и интегрироваться с вашими существующими рабочими процессами.
+
+### Ключевые возможности
+
+- **ИИ-диалоги и агенты** — Многораундовые диалоги, вызов инструментов, мультимодальная поддержка, потоковый вывод. Встроенная реализация RAG (база знаний) с глубокой интеграцией в [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org).
+- **Универсальная поддержка IM-платформ** — Единая кодовая база для Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK.
+- **Готовность к продакшену** — Контроль доступа, ограничение скорости, фильтрация чувствительных слов, комплексный мониторинг и обработка исключений. Проверено в корпоративной среде.
+- **Экосистема плагинов** — Сотни плагинов, событийно-ориентированная архитектура, расширения компонентов и поддержка [протокола MCP](https://modelcontextprotocol.io/).
+- **Веб-панель управления** — Настраивайте, управляйте и мониторьте ваших ботов через интуитивный браузерный интерфейс. Ручное редактирование YAML не требуется.
+- **Мультиконвейерная архитектура** — Разные боты для разных сценариев с комплексным мониторингом и обработкой исключений.
+
+[→ Подробнее обо всех возможностях](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Быстрый старт
+
+### ☁️ LangBot Cloud (Рекомендуется)
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — Без развёртывания, готово к использованию.
+
+### Запуск одной командой

 ```bash
 uvx langbot
 ```

-Посетите http://localhost:5300, чтобы начать использование.
+> Требуется [uv](https://docs.astral.sh/uv/getting-started/installation/). Откройте http://localhost:5300 — готово.

-#### Развертывание с Docker Compose
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -47,103 +70,101 @@ cd LangBot/docker
 docker compose up -d
 ```

-Посетите http://localhost:5300, чтобы начать использование.
-
-Подробная документация [Развертывание Docker](https://docs.langbot.app/en/deploy/langbot/docker.html).
-
-#### Развертывание одним кликом на BTPanel
-
-LangBot добавлен в BTPanel. Если у вас установлен BTPanel, вы можете использовать [документацию](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) для его использования.
-
-#### Облачное развертывание Zeabur
-
-Шаблон Zeabur, предоставленный сообществом.
+### Облачное развертывание одним кликом

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Облачное развертывание Railway
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### Другие методы развертывания
+**Другие варианты:** [Docker](https://docs.langbot.app/en/deploy/langbot/docker.html) · [Ручная установка](https://docs.langbot.app/en/deploy/langbot/manual.html) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-Используйте выпущенную версию напрямую для запуска, см. документацию [Ручное развертывание](https://docs.langbot.app/en/deploy/langbot/manual.html).
+---

-#### Развертывание Kubernetes
-
-См. документацию [Развертывание Kubernetes](./docker/README_K8S.md).
-
-## 😎 Оставайтесь в курсе
-
-Нажмите кнопки Star и Watch в правом верхнем углу репозитория, чтобы получать последние обновления.
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ Функции
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 Чат с LLM / Agent: Поддержка нескольких LLM, адаптация к групповым и личным чатам; Поддержка многораундовых разговоров, вызовов инструментов, мультимодальных возможностей и потоковой передачи. Встроенная реализация RAG (база знаний) и глубокая интеграция с [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org) и др. LLMOps платформами.
- 🤖 Многоплатформенная поддержка: В настоящее время поддерживает QQ, QQ Channel, WeCom, личный WeChat, Lark, DingTalk, Discord, Telegram, KOOK, Slack, LINE и т.д.
- 🛠️ Высокая стабильность, богатство функций: Нативный контроль доступа, ограничение скорости, фильтрация чувствительных слов и т.д.; Простота в использовании, поддержка нескольких методов развертывания.
- 🧩 Расширение плагинов, активное сообщество: Высокая стабильность, высокая безопасность уровня производства; Поддержка механизмов плагинов, управляемых событиями, расширения компонентов и т.д.; Интеграция протокола [MCP](https://modelcontextprotocol.io/) от Anthropic; В настоящее время сотни плагинов.
- 😻 Веб-интерфейс: Поддержка управления экземплярами LangBot через браузер. Нет необходимости вручную писать конфигурационные файлы.
- 📊 Функции уровня производства: Поддержка нескольких конфигураций конвейера, разные боты для разных сценариев. Имеет комплексные возможности мониторинга и обработки исключений.
-
-Для более подробных спецификаций обратитесь к [документации](https://docs.langbot.app/en/insight/features.html).
-
-Или посетите демонстрационную среду: https://demo.langbot.dev/
-  - Информация для входа: Email: `demo@langbot.app` Пароль: `langbot123456`
-  - Примечание: Только для демонстрации WebUI, пожалуйста, не вводите конфиденциальную информацию в общедоступной среде.
-
-### Платформы обмена сообщениями
+## Поддерживаемые платформы

 | Платформа | Статус | Примечания |
-| --- | --- | --- |
+|-----------|--------|------------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| Личный QQ | ✅ |  |
-| Официальный API QQ | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| Личный WeChat | ✅ |  |
-| KOOK | ✅ |  |
+| QQ | ✅ | Личный и официальный API |
+| WeCom | ✅ | Корпоративный WeChat, внешний CS, AI-бот |
+| WeChat | ✅ | Личный и официальный аккаунт |
 | Lark | ✅ |  |
 | DingTalk | ✅ |  |
+| KOOK | ✅ |  |
+| Satori | ✅ |  |

-### LLMs
+---

-| LLM | Статус | Примечания |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | Доступна для любой модели формата интерфейса OpenAI |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | Платформа ресурсов LLM и GPU |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | Платформа ресурсов LLM и GPU |
-| [接口 AI](https://jiekou.ai/) | ✅ | Платформа агрегации LLM |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | Платформа ресурсов LLM и GPU |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | Шлюз LLM (MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | Платформа LLMOps |
-| [Ollama](https://ollama.com/) | ✅ | Платформа локального запуска LLM |
-| [LMStudio](https://lmstudio.ai/) | ✅ | Платформа локального запуска LLM |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | Шлюз интерфейса LLM (MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | Шлюз LLM (MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | Шлюз LLM (MaaS), платформа LLMOps |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | Шлюз LLM (MaaS), платформа LLMOps |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | Шлюз LLM (MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | Поддержка доступа к инструментам через протокол MCP |
+## Поддерживаемые LLM и интеграции

-## 🤝 Вклад сообщества
+| Провайдер | Тип | Статус |
+|-----------|-----|--------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | Локальный LLM | ✅ |
+| [LM Studio](https://lmstudio.ai/) | Локальный LLM | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | Протокол | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | Шлюз | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | Шлюз | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | Шлюз | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | Шлюз | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | Шлюз | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | Шлюз | ✅ |
+| [接口 AI](https://jiekou.ai/) | Шлюз | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | Платформа GPU | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | Платформа GPU | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | Платформа GPU | ✅ |

-Спасибо следующим [контрибьюторам кода](https://github.com/langbot-app/LangBot/graphs/contributors) и другим членам сообщества за их вклад в LangBot:
+[→ Смотреть все интеграции](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Почему LangBot?
+
+| Сценарий использования | Как помогает LangBot |
+|------------------------|----------------------|
+| **Поддержка клиентов** | Разверните ИИ-агентов в Slack/Discord/Telegram, которые отвечают на вопросы, используя вашу базу знаний |
+| **Внутренние инструменты** | Подключите рабочие процессы n8n/Dify к WeCom/DingTalk для автоматизации бизнес-процессов |
+| **Управление сообществом** | Модерируйте группы QQ/Discord с помощью ИИ-фильтрации контента и взаимодействия |
+| **Мультиплатформенное присутствие** | Один бот — все платформы. Управляйте из единой панели |
+
+---
+
+## Демо
+
+**Попробуйте прямо сейчас:** https://demo.langbot.dev/
+- Email: `demo@langbot.app`
+- Пароль: `langbot123456`
+
+*Примечание: Публичная демо-среда. Не вводите конфиденциальную информацию.*
+
+---
+
+## Сообщество
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Сообщество Discord](https://discord.gg/wdNEHETs87)
+
+---
+
+## История Stars
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## Участники
+
+Спасибо всем [участникам](https://github.com/langbot-app/LangBot/graphs/contributors), которые помогли сделать LangBot лучше:

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
--- a/README_TW.md
+++ b/README_TW.md
@@ -1,25 +1,29 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

-<div align="center"><a href="https://hellogithub.com/repository/langbot-app/LangBot" target="_blank"><img src="https://abroad.hellogithub.com/v1/widgets/recommend.svg?rid=5ce8ae2aa4f74316bf393b57b952433c&claim_uid=gtmc6YWjMZkT21R" alt="Featured｜HelloGitHub" style="width: 250px; height: 54px;" width="250" height="54" /></a>
+<div align="center">

-<h3>使用 LangBot 快速建構、除錯和部署 IM 機器人。</h3>
+<a href="https://hellogithub.com/repository/langbot-app/LangBot" target="_blank"><img src="https://abroad.hellogithub.com/v1/widgets/recommend.svg?rid=5ce8ae2aa4f74316bf393b57b952433c&claim_uid=gtmc6YWjMZkT21R" alt="Featured｜HelloGitHub" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-[English](README_EN.md) / [简体中文](README.md) / 繁體中文 / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)
+<h3>生產級 AI 即時通訊機器人開發平台。</h3>
+<h4>快速建構、除錯和部署 AI 機器人到微信、QQ、飛書、Slack、Discord、Telegram 等平台。</h4>
+
+[English](README.md) / [简体中文](README_CN.md) / 繁體中文 / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / [Tiếng Việt](README_VI.md)

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![QQ Group](https://img.shields.io/badge/%E7%A4%BE%E5%8C%BAQQ%E7%BE%A4-966235608-blue)](https://qm.qq.com/q/JLi38whHum)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)
 [![star](https://gitcode.com/RockChinQ/LangBot/star/badge.svg)](https://gitcode.com/RockChinQ/LangBot)

-<a href="https://langbot.app">主頁</a> ｜
-<a href="https://docs.langbot.app/zh/insight/features.html">規格特性</a> ｜
-<a href="https://docs.langbot.app/zh/insight/guide.html">部署文件</a> ｜
-<a href="https://docs.langbot.app/zh/tags/readme.html">API 整合</a> ｜
+<a href="https://langbot.app">官網</a> ｜
+<a href="https://docs.langbot.app/zh/insight/features.html">特性</a> ｜
+<a href="https://docs.langbot.app/zh/insight/guide.html">文件</a> ｜
+<a href="https://docs.langbot.app/zh/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">外掛市場</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">路線圖</a>

@@ -27,19 +31,40 @@

 </p>

-## 📦 開始使用
+---

-#### 快速部署
+## 什麼是 LangBot？

-使用 `uvx` 一鍵啟動（需要先安裝 [uv](https://docs.astral.sh/uv/getting-started/installation/) ）：
+LangBot 是一個**開源的生產級平台**，用於建構 AI 驅動的即時通訊機器人。它將大語言模型（LLM）連接到各種聊天平台，幫助你創建能夠對話、執行任務、並整合到現有工作流程中的智能 Agent。
+
+### 核心能力
+
+- **AI 對話與 Agent** — 多輪對話、工具調用、多模態、流式輸出。自帶 RAG（知識庫），深度整合 [Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org) 等 LLMOps 平台。
+- **全平台支援** — 一套程式碼，覆蓋 QQ、微信、企業微信、飛書、釘釘、Discord、Telegram、Slack、LINE、KOOK 等平台。
+- **生產就緒** — 存取控制、限速、敏感詞過濾、全面監控與異常處理，已被多家企業採用。
+- **外掛生態** — 數百個外掛，事件驅動架構，組件擴展，適配 [MCP 協議](https://modelcontextprotocol.io/)。
+- **Web 管理面板** — 透過瀏覽器直觀地配置、管理和監控機器人，無需手動編輯設定檔。
+- **多流水線架構** — 不同機器人用於不同場景，具備全面的監控和異常處理能力。
+
+[→ 了解更多功能特性](https://docs.langbot.app/zh/insight/features.html)
+
+---
+
+## 快速開始
+
+### ☁️ LangBot Cloud（推薦）
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — 免部署，開箱即用。
+
+### 一鍵啟動

 ```bash
 uvx langbot
 ```

-訪問 http://localhost:5300 即可開始使用。
+> 需要安裝 [uv](https://docs.astral.sh/uv/getting-started/installation/)。訪問 http://localhost:5300 即可使用。

-#### Docker Compose 部署
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -47,104 +72,63 @@ cd LangBot/docker
 docker compose up -d
 ```

-訪問 http://localhost:5300 即可開始使用。
-
-詳細文件[Docker 部署](https://docs.langbot.app/zh/deploy/langbot/docker.html)。
-
-#### 寶塔面板部署
-
-已上架寶塔面板，若您已安裝寶塔面板，可以根據[文件](https://docs.langbot.app/zh/deploy/langbot/one-click/bt.html)使用。
-
-#### Zeabur 雲端部署
-
-社群貢獻的 Zeabur 模板。
+### 一鍵雲端部署

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/zh-CN/templates/ZKTBDH)
-
-#### Railway 雲端部署
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### 手動部署
+**更多方式：** [Docker](https://docs.langbot.app/zh/deploy/langbot/docker.html) · [手動部署](https://docs.langbot.app/zh/deploy/langbot/manual.html) · [寶塔面板](https://docs.langbot.app/zh/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-直接使用發行版運行，查看文件[手動部署](https://docs.langbot.app/zh/deploy/langbot/manual.html)。
+---

-#### Kubernetes 部署
-
-參考 [Kubernetes 部署](./docker/README_K8S.md) 文件。
-
-## 😎 保持更新
-
-點擊倉庫右上角 Star 和 Watch 按鈕，獲取最新動態。
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ 特性
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 大模型對話、Agent：支援多種大模型，適配群聊和私聊；具有多輪對話、工具調用、多模態、流式輸出能力，自帶 RAG（知識庫）實現，並深度適配 [Dify](https://dify.ai)、[Coze](https://coze.com)、[n8n](https://n8n.io)、[Langflow](https://langflow.org)等 LLMOps 平台。
- 🤖 多平台支援：目前支援 QQ、QQ頻道、企業微信、個人微信、飛書、Discord、Telegram、KOOK、Slack、LINE 等平台。
- 🛠️ 高穩定性、功能完備：原生支援訪問控制、限速、敏感詞過濾等機制；配置簡單，支援多種部署方式。
- 🧩 外掛擴展、活躍社群：高穩定性、高安全性的生產級外掛系統；支援事件驅動、組件擴展等外掛機制；適配 Anthropic [MCP 協議](https://modelcontextprotocol.io/)；目前已有數百個外掛。
- 😻 Web 管理面板：提供先進的 WebUI 管理面板，用最直觀的方式配置、管理、監控機器人。
- 📊 生產級特性：支援多流水線配置，不同機器人用於不同應用場景。具有全面的監控和異常處理能力。
-
-詳細規格特性請訪問[文件](https://docs.langbot.app/zh/insight/features.html)。
-
-或訪問 demo 環境：https://demo.langbot.dev/  
-  - 登入資訊：郵箱：`demo@langbot.app` 密碼：`langbot123456`
-  - 注意：僅展示 WebUI 效果，公開環境，請不要在其中填入您的任何敏感資訊。
-
-### 訊息平台
+## 支援的平台

 | 平台 | 狀態 | 備註 |
-| --- | --- | --- |
+|------|------|------|
+| QQ | ✅ | 個人號、官方機器人（頻道、私聊、群聊） |
+| 微信 | ✅ | 個人微信、微信公眾號 |
+| 企業微信 | ✅ | 應用訊息、對外客服、智能機器人 |
+| 飛書 | ✅ |  |
+| 釘釘 | ✅ |  |
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| QQ 個人號 | ✅ | QQ 個人號私聊、群聊 |
-| QQ 官方機器人 | ✅ | QQ 官方機器人，支援頻道、私聊、群聊 |
-| 微信 | ✅ |  |
-| 企微對外客服 | ✅ |  |
-| 企微智能機器人 | ✅ |  |
-| 微信公眾號 | ✅ |  |
 | KOOK | ✅ |  |
-| Lark | ✅ |  |
-| DingTalk | ✅ |  |
+| Satori | ✅ |  |

-### 大模型能力
+---

-| 模型 | 狀態 | 備註 |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | 可接入任何 OpenAI 介面格式模型 |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [智譜AI](https://open.bigmodel.cn/) | ✅ |  |
-| [勝算雲](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | 大模型和 GPU 資源平台 |
-| [優雲智算](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | 大模型和 GPU 資源平台 |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | 大模型和 GPU 資源平台 |
-| [接口 AI](https://jiekou.ai/) | ✅ | 大模型聚合平台，專注全球大模型接入 |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | 大模型聚合平台 |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | LLMOps 平台 |
-| [Ollama](https://ollama.com/) | ✅ | 本地大模型運行平台 |
-| [LMStudio](https://lmstudio.ai/) | ✅ | 本地大模型運行平台 |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | 大模型介面聚合平台 |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | 大模型聚合平台 |
-| [阿里雲百煉](https://bailian.console.aliyun.com/) | ✅ | 大模型聚合平台, LLMOps 平台 |
-| [火山方舟](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | 大模型聚合平台, LLMOps 平台 |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | 大模型聚合平台 |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | 支援通過 MCP 協議獲取工具 |
+## 支援的大模型與整合

-### TTS
+| 提供商 | 類型 | 狀態 |
+|--------|------|------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [智譜AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | 本地 LLM | ✅ |
+| [LM Studio](https://lmstudio.ai/) | 本地 LLM | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | 協議 | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | 聚合平台 | ✅ |
+| [阿里雲百煉](https://bailian.console.aliyun.com/) | 聚合平台 | ✅ |
+| [火山方舟](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | 聚合平台 | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | 聚合平台 | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | 聚合平台 | ✅ |
+| [勝算雲](https://www.shengsuanyun.com/?from=CH_KYIPP758) | GPU 平台 | ✅ |
+| [優雲智算](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | GPU 平台 | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | GPU 平台 | ✅ |
+| [接口 AI](https://jiekou.ai/) | 聚合平台 | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | 聚合平台 | ✅ |
+
+### TTS（語音合成）

 | 平台/模型 | 備註 |
-| --- | --- |
+|-----------|------|
 | [FishAudio](https://fish.audio/zh-CN/discovery/) | [外掛](https://github.com/the-lazy-me/NewChatVoice) |
 | [海豚 AI](https://www.ttson.cn/?source=thelazy) | [外掛](https://github.com/the-lazy-me/NewChatVoice) |
 | [AzureTTS](https://portal.azure.com/) | [外掛](https://github.com/Ingnaryk/LangBot_AzureTTS) |
@@ -152,13 +136,54 @@ docker compose up -d
 ### 文生圖

 | 平台/模型 | 備註 |
-| --- | --- |
-| 阿里雲百煉 | [外掛](https://github.com/Thetail001/LangBot_BailianTextToImagePlugin)
+|-----------|------|
+| 阿里雲百煉 | [外掛](https://github.com/Thetail001/LangBot_BailianTextToImagePlugin) |

-## 😘 社群貢獻
+[→ 查看完整整合列表](https://docs.langbot.app/zh/insight/features.html)

-感謝以下[程式碼貢獻者](https://github.com/langbot-app/LangBot/graphs/contributors)和社群裡其他成員對 LangBot 的貢獻：
+---
+
+## 為什麼選擇 LangBot？
+
+| 使用場景 | LangBot 如何幫助 |
+|----------|------------------|
+| **客戶服務** | 將 AI Agent 部署到微信/企微/釘釘/飛書，基於知識庫自動回答使用者問題 |
+| **內部工具** | 將 n8n/Dify 工作流接入企微/釘釘，實現業務流程自動化 |
+| **社群運營** | 在 QQ/Discord 群中使用 AI 驅動的內容審核與智能互動 |
+| **多平台觸達** | 一個機器人，覆蓋所有平台。透過統一面板集中管理 |
+
+---
+
+## 線上演示
+
+**立即體驗：** https://demo.langbot.dev/
+- 信箱：`demo@langbot.app`
+- 密碼：`langbot123456`
+
+*注意：公開演示環境，請不要在其中填入任何敏感資訊。*
+
+---
+
+## 社群
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+[![QQ Group](https://img.shields.io/badge/%E7%A4%BE%E5%8C%BAQQ%E7%BE%A4-966235608-blue)](https://qm.qq.com/q/JLi38whHum)
+
+- [Discord 社群](https://discord.gg/wdNEHETs87)
+- [QQ 社群群](https://qm.qq.com/q/JLi38whHum)
+
+---
+
+## Star 趨勢
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## 貢獻者
+
+感謝所有[貢獻者](https://github.com/langbot-app/LangBot/graphs/contributors)對 LangBot 的幫助：

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
-</a> 
+</a>
--- a/README_VI.md
+++ b/README_VI.md
@@ -1,25 +1,27 @@
 <p align="center">
 <a href="https://langbot.app">
-<img width="130" src="https://docs.langbot.app/langbot-logo.png" alt="LangBot"/>
+<img width="130" src="res/logo-blue.png" alt="LangBot"/>
 </a>

 <div align="center">

 <a href="https://www.producthunt.com/products/langbot?utm_source=badge-follow&utm_medium=badge&utm_source=badge-langbot" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/follow.svg?product_id=1077185&theme=light" alt="LangBot - Production&#0045;grade&#0032;IM&#0032;bot&#0032;made&#0032;easy&#0046; | Product Hunt" style="width: 250px; height: 54px;" width="250" height="54" /></a>

-<h3>Xây dựng, gỡ lỗi và triển khai bot IM nhanh chóng với LangBot.</h3>
+<h3>Nền tảng cấp sản xuất để xây dựng bot IM với AI agent.</h3>
+<h4>Xây dựng, gỡ lỗi và triển khai bot AI nhanh chóng trên Slack, Discord, Telegram, WeChat và nhiều nền tảng khác.</h4>

-[English](README_EN.md) / [简体中文](README.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / Tiếng Việt
+[English](README.md) / [简体中文](README_CN.md) / [繁體中文](README_TW.md) / [日本語](README_JP.md) / [Español](README_ES.md) / [Français](README_FR.md) / [한국어](README_KO.md) / [Русский](README_RU.md) / Tiếng Việt

 [![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb)](https://discord.gg/wdNEHETs87)
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/langbot-app/LangBot)
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/langbot-app/LangBot)](https://github.com/langbot-app/LangBot/releases/latest)
 <img src="https://img.shields.io/badge/python-3.10 ~ 3.13 -blue.svg" alt="python">
+[![GitHub stars](https://img.shields.io/github/stars/langbot-app/LangBot?style=social)](https://github.com/langbot-app/LangBot/stargazers)

 <a href="https://langbot.app">Trang chủ</a> ｜
 <a href="https://docs.langbot.app/en/insight/features.html">Tính năng</a> ｜
-<a href="https://docs.langbot.app/en/insight/guide.html">Triển khai</a> ｜
-<a href="https://docs.langbot.app/en/tags/readme.html">Tích hợp API</a> ｜
+<a href="https://docs.langbot.app/en/insight/guide.html">Tài liệu</a> ｜
+<a href="https://docs.langbot.app/en/tags/readme.html">API</a> ｜
 <a href="https://space.langbot.app">Chợ Plugin</a> ｜
 <a href="https://langbot.featurebase.app/roadmap">Lộ trình</a>

@@ -27,19 +29,40 @@

 </p>

-## 📦 Bắt đầu
+---

-#### Khởi động Nhanh
+## LangBot là gì?

-Sử dụng `uvx` để khởi động bằng một lệnh (cần cài đặt [uv](https://docs.astral.sh/uv/getting-started/installation/)):
+LangBot là một **nền tảng mã nguồn mở, cấp sản xuất** để xây dựng bot nhắn tin tức thời được hỗ trợ bởi AI. Nó kết nối các Mô hình Ngôn ngữ Lớn (LLM) với bất kỳ nền tảng chat nào, cho phép bạn tạo các agent thông minh có thể trò chuyện, thực hiện tác vụ và tích hợp với quy trình làm việc hiện có của bạn.
+
+### Khả năng chính
+
+- **Hội thoại AI & Agent** — Đối thoại nhiều lượt, gọi công cụ, hỗ trợ đa phương thức, đầu ra streaming. RAG (cơ sở kiến thức) tích hợp sẵn với tích hợp sâu vào [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org).
+- **Hỗ trợ đa nền tảng IM** — Một mã nguồn cho Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK.
+- **Sẵn sàng cho sản xuất** — Kiểm soát truy cập, giới hạn tốc độ, lọc từ nhạy cảm, giám sát toàn diện và xử lý ngoại lệ. Được doanh nghiệp tin dùng.
+- **Hệ sinh thái Plugin** — Hàng trăm plugin, kiến trúc hướng sự kiện, mở rộng thành phần, và hỗ trợ [giao thức MCP](https://modelcontextprotocol.io/).
+- **Bảng quản lý Web** — Cấu hình, quản lý và giám sát bot thông qua giao diện trình duyệt trực quan. Không cần chỉnh sửa YAML.
+- **Kiến trúc đa Pipeline** — Các bot khác nhau cho các kịch bản khác nhau, với giám sát toàn diện và xử lý ngoại lệ.
+
+[→ Tìm hiểu thêm về tất cả tính năng](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Bắt đầu nhanh
+
+### ☁️ LangBot Cloud (Khuyên dùng)
+
+**[LangBot Cloud](https://space.langbot.app/cloud)** — Không cần triển khai, sẵn sàng sử dụng.
+
+### Khởi chạy một dòng

 ```bash
 uvx langbot
 ```

-Truy cập http://localhost:5300 để bắt đầu sử dụng.
+> Yêu cầu [uv](https://docs.astral.sh/uv/getting-started/installation/). Truy cập http://localhost:5300 — xong.

-#### Triển khai Docker Compose
+### Docker Compose

 ```bash
 git clone https://github.com/langbot-app/LangBot
@@ -47,103 +70,101 @@ cd LangBot/docker
 docker compose up -d
 ```

-Truy cập http://localhost:5300 để bắt đầu sử dụng.
-
-Tài liệu chi tiết [Triển khai Docker](https://docs.langbot.app/en/deploy/langbot/docker.html).
-
-#### Triển khai Một cú nhấp chuột trên BTPanel
-
-LangBot đã được liệt kê trên BTPanel. Nếu bạn đã cài đặt BTPanel, bạn có thể sử dụng [tài liệu](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) để sử dụng nó.
-
-#### Triển khai Cloud Zeabur
-
-Mẫu Zeabur được đóng góp bởi cộng đồng.
+### Triển khai đám mây một cú nhấp

 [![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/en-US/templates/ZKTBDH)
-
-#### Triển khai Cloud Railway
-
 [![Deploy on Railway](https://railway.com/button.svg)](https://railway.app/template/yRrAyL?referralCode=vogKPF)

-#### Các Phương pháp Triển khai Khác
+**Thêm tùy chọn:** [Docker](https://docs.langbot.app/en/deploy/langbot/docker.html) · [Thủ công](https://docs.langbot.app/en/deploy/langbot/manual.html) · [BTPanel](https://docs.langbot.app/en/deploy/langbot/one-click/bt.html) · [Kubernetes](./docker/README_K8S.md)

-Sử dụng trực tiếp phiên bản phát hành để chạy, xem tài liệu [Triển khai Thủ công](https://docs.langbot.app/en/deploy/langbot/manual.html).
+---

-#### Triển khai Kubernetes
-
-Tham khảo tài liệu [Triển khai Kubernetes](./docker/README_K8S.md).
-
-## 😎 Cập nhật Mới nhất
-
-Nhấp vào các nút Star và Watch ở góc trên bên phải của kho lưu trữ để nhận các bản cập nhật mới nhất.
-
-![star gif](https://docs.langbot.app/star.gif)
-
-## ✨ Tính năng
-
-<img width="500" src="https://docs.langbot.app/ui/bot-page-en-rounded.png" />
-
-
- 💬 Chat với LLM / Agent: Hỗ trợ nhiều LLM, thích ứng với chat nhóm và chat riêng tư; Hỗ trợ các cuộc trò chuyện nhiều vòng, gọi công cụ, khả năng đa phương thức và đầu ra streaming. Triển khai RAG (cơ sở kiến thức) tích hợp sẵn và tích hợp sâu với [Dify](https://dify.ai), [Coze](https://coze.com), [n8n](https://n8n.io), [Langflow](https://langflow.org) v.v. LLMOps platforms.
- 🤖 Hỗ trợ Đa nền tảng: Hiện hỗ trợ QQ, QQ Channel, WeCom, WeChat cá nhân, Lark, DingTalk, Discord, Telegram, KOOK, Slack, LINE, v.v.
- 🛠️ Độ ổn định Cao, Tính năng Phong phú: Kiểm soát truy cập gốc, giới hạn tốc độ, lọc từ nhạy cảm, v.v.; Dễ sử dụng, hỗ trợ nhiều phương pháp triển khai.
- 🧩 Mở rộng Plugin, Cộng đồng Hoạt động: Hỗ trợ các cơ chế plugin hướng sự kiện, mở rộng thành phần, v.v.; Tích hợp giao thức [MCP](https://modelcontextprotocol.io/) của Anthropic; Hiện có hàng trăng plugin.
- 😻 Giao diện Web: Hỗ trợ quản lý các phiên bản LangBot thông qua trình duyệt. Không cần viết tệp cấu hình thủ công.
- 📊 Tính năng Cấp sản xuất: Hỗ trợ nhiều cấu hình pipeline, các bot khác nhau cho các kịch bản khác nhau. Có khả năng giám sát toàn diện và xử lý ngoại lệ.
-
-Để biết thêm thông số kỹ thuật chi tiết, vui lòng tham khảo [tài liệu](https://docs.langbot.app/en/insight/features.html).
-
-Hoặc truy cập môi trường demo: https://demo.langbot.dev/
-  - Thông tin đăng nhập: Email: `demo@langbot.app` Mật khẩu: `langbot123456`
-  - Lưu ý: Chỉ dành cho demo WebUI, vui lòng không nhập bất kỳ thông tin nhạy cảm nào trong môi trường công cộng.
-
-### Nền tảng Nhắn tin
+## Nền tảng được hỗ trợ

 | Nền tảng | Trạng thái | Ghi chú |
-| --- | --- | --- |
+|----------|--------|-------|
 | Discord | ✅ |  |
 | Telegram | ✅ |  |
 | Slack | ✅ |  |
 | LINE | ✅ |  |
-| QQ Cá nhân | ✅ |  |
-| QQ API Chính thức | ✅ |  |
-| WeCom | ✅ |  |
-| WeComCS | ✅ |  |
-| WeCom AI Bot | ✅ |  |
-| WeChat Cá nhân | ✅ |  |
-| KOOK | ✅ |  |
+| QQ | ✅ | Cá nhân & API chính thức |
+| WeCom | ✅ | WeChat doanh nghiệp, CS bên ngoài, AI Bot |
+| WeChat | ✅ | Cá nhân & Tài khoản công khai |
 | Lark | ✅ |  |
 | DingTalk | ✅ |  |
+| KOOK | ✅ |  |
+| Satori | ✅ |  |

-### LLMs
+---

-| LLM | Trạng thái | Ghi chú |
-| --- | --- | --- |
-| [OpenAI](https://platform.openai.com/) | ✅ | Có sẵn cho bất kỳ mô hình định dạng giao diện OpenAI nào |
-| [DeepSeek](https://www.deepseek.com/) | ✅ |  |
-| [Moonshot](https://www.moonshot.cn/) | ✅ |  |
-| [Anthropic](https://www.anthropic.com/) | ✅ |  |
-| [xAI](https://x.ai/) | ✅ |  |
-| [Zhipu AI](https://open.bigmodel.cn/) | ✅ |  |
-| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | ✅ | Nền tảng tài nguyên LLM và GPU |
-| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | ✅ | Nền tảng tài nguyên LLM và GPU |
-| [接口 AI](https://jiekou.ai/) | ✅ | Nền tảng tổng hợp LLM |
-| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | ✅ | Nền tảng tài nguyên LLM và GPU |
-| [302.AI](https://share.302.ai/SuTG99) | ✅ | Cổng LLM (MaaS) |
-| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | ✅ | |
-| [Dify](https://dify.ai) | ✅ | Nền tảng LLMOps |
-| [Ollama](https://ollama.com/) | ✅ | Nền tảng chạy LLM cục bộ |
-| [LMStudio](https://lmstudio.ai/) | ✅ | Nền tảng chạy LLM cục bộ |
-| [GiteeAI](https://ai.gitee.com/) | ✅ | Cổng giao diện LLM (MaaS) |
-| [SiliconFlow](https://siliconflow.cn/) | ✅ | Cổng LLM (MaaS) |
-| [Aliyun Bailian](https://bailian.console.aliyun.com/) | ✅ | Cổng LLM (MaaS), nền tảng LLMOps |
-| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | ✅ | Cổng LLM (MaaS), nền tảng LLMOps |
-| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | ✅ | Cổng LLM (MaaS) |
-| [MCP](https://modelcontextprotocol.io/) | ✅ | Hỗ trợ truy cập công cụ qua giao thức MCP |
+## LLM và tích hợp được hỗ trợ

-## 🤝 Đóng góp Cộng đồng
+| Nhà cung cấp | Loại | Trạng thái |
+|----------|------|--------|
+| [OpenAI](https://platform.openai.com/) | LLM | ✅ |
+| [Anthropic](https://www.anthropic.com/) | LLM | ✅ |
+| [DeepSeek](https://www.deepseek.com/) | LLM | ✅ |
+| [Google Gemini](https://aistudio.google.com/prompts/new_chat) | LLM | ✅ |
+| [xAI](https://x.ai/) | LLM | ✅ |
+| [Moonshot](https://www.moonshot.cn/) | LLM | ✅ |
+| [Zhipu AI](https://open.bigmodel.cn/) | LLM | ✅ |
+| [Ollama](https://ollama.com/) | LLM cục bộ | ✅ |
+| [LM Studio](https://lmstudio.ai/) | LLM cục bộ | ✅ |
+| [Dify](https://dify.ai) | LLMOps | ✅ |
+| [MCP](https://modelcontextprotocol.io/) | Giao thức | ✅ |
+| [SiliconFlow](https://siliconflow.cn/) | Cổng | ✅ |
+| [Aliyun Bailian](https://bailian.console.aliyun.com/) | Cổng | ✅ |
+| [Volc Engine Ark](https://console.volcengine.com/ark/region:ark+cn-beijing/model?vendor=Bytedance&view=LIST_VIEW) | Cổng | ✅ |
+| [ModelScope](https://modelscope.cn/docs/model-service/API-Inference/intro) | Cổng | ✅ |
+| [GiteeAI](https://ai.gitee.com/) | Cổng | ✅ |
+| [CompShare](https://www.compshare.cn/?ytag=GPU_YY-gh_langbot) | Nền tảng GPU | ✅ |
+| [PPIO](https://ppinfra.com/user/register?invited_by=QJKFYD&utm_source=github_langbot) | Nền tảng GPU | ✅ |
+| [ShengSuanYun](https://www.shengsuanyun.com/?from=CH_KYIPP758) | Nền tảng GPU | ✅ |
+| [接口 AI](https://jiekou.ai/) | Cổng | ✅ |
+| [302.AI](https://share.302.ai/SuTG99) | Cổng | ✅ |

-Cảm ơn các [người đóng góp mã](https://github.com/langbot-app/LangBot/graphs/contributors) sau đây và các thành viên khác trong cộng đồng vì những đóng góp của họ cho LangBot:
+[→ Xem tất cả tích hợp](https://docs.langbot.app/en/insight/features.html)
+
+---
+
+## Tại sao chọn LangBot?
+
+| Trường hợp sử dụng | LangBot giúp như thế nào |
+|----------|-------------------|
+| **Hỗ trợ khách hàng** | Triển khai agent AI trên Slack/Discord/Telegram để trả lời câu hỏi bằng cơ sở kiến thức của bạn |
+| **Công cụ nội bộ** | Kết nối quy trình n8n/Dify với WeCom/DingTalk để tự động hóa quy trình kinh doanh |
+| **Quản lý cộng đồng** | Quản lý nhóm QQ/Discord với tính năng lọc nội dung và tương tác được hỗ trợ bởi AI |
+| **Đa nền tảng** | Một bot, tất cả nền tảng. Quản lý từ một bảng điều khiển duy nhất |
+
+---
+
+## Demo trực tuyến
+
+**Thử ngay:** https://demo.langbot.dev/
+- Email: `demo@langbot.app`
+- Mật khẩu: `langbot123456`
+
+*Lưu ý: Môi trường demo công khai. Không nhập thông tin nhạy cảm.*
+
+---
+
+## Cộng đồng
+
+[![Discord](https://img.shields.io/discord/1335141740050649118?logo=discord&label=Discord)](https://discord.gg/wdNEHETs87)
+
+- [Cộng đồng Discord](https://discord.gg/wdNEHETs87)
+
+---
+
+## Lịch sử Star
+
+[![Star History Chart](https://api.star-history.com/svg?repos=langbot-app/LangBot&type=Date)](https://star-history.com/#langbot-app/LangBot&Date)
+
+---
+
+## Người đóng góp
+
+Cảm ơn tất cả [người đóng góp](https://github.com/langbot-app/LangBot/graphs/contributors) đã giúp LangBot trở nên tốt hơn:

 <a href="https://github.com/langbot-app/LangBot/graphs/contributors">
  <img src="https://contrib.rocks/image?repo=langbot-app/LangBot" />
--- a/docs/service-api-openapi.json
+++ b/docs/service-api-openapi.json
@@ -9,7 +9,7 @@
      "url": "https://langbot.app"
    },
    "license": {
-      "name": "AGPL-3.0",
+      "name": "Apache-2.0",
      "url": "https://github.com/langbot-app/LangBot/blob/master/LICENSE"
    }
  },
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "langbot"
-version = "4.8.3"
+version = "4.9.4"
 description = "Production-grade platform for building agentic IM bots"
 readme = "README.md"
 license-files = ["LICENSE"]
@@ -61,16 +61,17 @@ dependencies = [
    "html2text>=2024.2.26",
    "langchain>=0.2.0",
    "langchain-text-splitters>=0.0.1",
-    "chromadb>=0.4.24",
+    "chromadb>=1.0.0,<2.0.0",
    "qdrant-client (>=1.15.1,<2.0.0)",
-    "pyseekdb==1.0.0b7",
-    "langbot-plugin==0.2.5",
+    "pyseekdb==1.1.0.post3",
+    "langbot-plugin==0.3.5",
    "asyncpg>=0.30.0",
    "line-bot-sdk>=3.19.0",
    "tboxsdk>=0.0.10",
    "boto3>=1.35.0",
    "pymilvus>=2.6.4",
    "pgvector>=0.4.1",
+    "botocore>=1.42.39",
 ]
 keywords = [
    "bot",
--- a/res/logo-blue.png
+++ b/res/logo-blue.png
--- a/src/langbot/init.py
+++ b/src/langbot/init.py
@@ -1,3 +1,3 @@
 """LangBot - Production-grade platform for building agentic IM bots"""

-__version__ = '4.8.3'
+__version__ = '4.9.4'
--- a/src/langbot/libs/openclaw_weixin_api/init.py
+++ b/src/langbot/libs/openclaw_weixin_api/init.py
@@ -0,0 +1,3 @@
+from .client import OpenClawWeixinClient as OpenClawWeixinClient
+from .types import ApiError as ApiError
+from .types import LoginResult as LoginResult
--- a/src/langbot/libs/openclaw_weixin_api/client.py
+++ b/src/langbot/libs/openclaw_weixin_api/client.py
@@ -0,0 +1,807 @@
+"""Async HTTP client for the OpenClaw WeChat API.
+
+Implements the iLink Bot API protocol.
+Reference: https://github.com/epiral/weixin-bot
+
+Endpoints: getUpdates (long-poll), sendMessage, getUploadUrl, getConfig, sendTyping.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import base64
+import io
+import logging
+import os
+import struct
+import typing
+import uuid
+from typing import Optional
+from urllib.parse import quote
+
+import aiohttp
+
+from .types import (
+    ApiError,
+    CDNMedia,
+    FileItem,
+    GetConfigResponse,
+    GetUpdatesResponse,
+    GetUploadUrlResponse,
+    ImageItem,
+    LoginResult,
+    MessageItem,
+    QRCodeResponse,
+    QRStatusResponse,
+    RefMessage,
+    TextItem,
+    VideoItem,
+    VoiceItem,
+    WeixinMessage,
+)
+
+logger = logging.getLogger('openclaw-weixin-sdk')
+
+DEFAULT_BASE_URL = 'https://ilinkai.weixin.qq.com'
+CDN_BASE_URL = 'https://novac2c.cdn.weixin.qq.com/c2c'
+
+CHANNEL_VERSION = '1.0.0'
+
+DEFAULT_API_TIMEOUT = 15
+DEFAULT_LONG_POLL_TIMEOUT = 40
+DEFAULT_CONFIG_TIMEOUT = 10
+DEFAULT_QR_POLL_TIMEOUT = 35
+
+SESSION_EXPIRED_ERRCODE = -14
+
+DEFAULT_BOT_TYPE = '3'
+
+# Maximum text length per message chunk (WeChat limit)
+MAX_TEXT_CHUNK_SIZE = 2000
+
+
+def _random_wechat_uin() -> str:
+    """Generate the X-WECHAT-UIN header: random uint32 -> decimal string -> base64."""
+    rand_bytes = os.urandom(4)
+    uint32_val = struct.unpack('>I', rand_bytes)[0]
+    return base64.b64encode(str(uint32_val).encode('utf-8')).decode('utf-8')
+
+
+def _build_base_info() -> dict:
+    """Build the base_info payload included in every API request."""
+    return {'channel_version': CHANNEL_VERSION}
+
+
+def _chunk_text(text: str, max_size: int = MAX_TEXT_CHUNK_SIZE) -> list[str]:
+    """Split long text into chunks that fit within WeChat's message size limit."""
+    if len(text) <= max_size:
+        return [text]
+    chunks = []
+    while text:
+        chunks.append(text[:max_size])
+        text = text[max_size:]
+    return chunks
+
+
+class OpenClawWeixinClient:
+    """Async client for the OpenClaw WeChat HTTP JSON API."""
+
+    def __init__(self, base_url: str, token: str):
+        self.base_url = base_url.rstrip('/')
+        self.token = token
+        self._session: Optional[aiohttp.ClientSession] = None
+
+    async def _get_session(self) -> aiohttp.ClientSession:
+        if self._session is None or self._session.closed:
+            self._session = aiohttp.ClientSession()
+        return self._session
+
+    async def close(self):
+        if self._session and not self._session.closed:
+            await self._session.close()
+
+    def _build_headers(self) -> dict[str, str]:
+        headers = {
+            'Content-Type': 'application/json',
+            'AuthorizationType': 'ilink_bot_token',
+            'X-WECHAT-UIN': _random_wechat_uin(),
+        }
+        if self.token:
+            headers['Authorization'] = f'Bearer {self.token}'
+        return headers
+
+    async def _post(self, endpoint: str, payload: dict, timeout: float = DEFAULT_API_TIMEOUT) -> dict:
+        """Make a POST request and return the JSON response.
+
+        Raises ApiError on HTTP errors or when the response contains a non-zero errcode.
+        """
+        payload['base_info'] = _build_base_info()
+
+        session = await self._get_session()
+        url = f'{self.base_url}/{endpoint}'
+        headers = self._build_headers()
+
+        async with session.post(
+            url, json=payload, headers=headers, timeout=aiohttp.ClientTimeout(total=timeout)
+        ) as resp:
+            if resp.status != 200:
+                text = await resp.text()
+                raise ApiError(
+                    f'OpenClaw API error {resp.status}: {text}',
+                    status=resp.status,
+                )
+            data = await resp.json(content_type=None)
+
+        # Check for application-level errors in the response body
+        errcode = data.get('errcode') or data.get('ret')
+        if errcode and errcode != 0:
+            raise ApiError(
+                data.get('errmsg') or f'API errcode {errcode}',
+                status=200,
+                code=errcode,
+                payload=data,
+            )
+
+        return data
+
+    async def get_updates(
+        self, get_updates_buf: str = '', timeout: float = DEFAULT_LONG_POLL_TIMEOUT
+    ) -> GetUpdatesResponse:
+        """Long-poll for new messages.
+
+        Note: This method does NOT raise ApiError for errcode responses —
+        it returns them in the GetUpdatesResponse so the caller can handle
+        session expiry and other errors with full context.
+        """
+        try:
+            # Bypass the errcode check in _post since get_updates needs
+            # to return error info (e.g. session expired) to the caller.
+            payload: dict = {'get_updates_buf': get_updates_buf}
+            payload['base_info'] = _build_base_info()
+
+            session = await self._get_session()
+            url = f'{self.base_url}/ilink/bot/getupdates'
+            headers = self._build_headers()
+
+            async with session.post(
+                url,
+                json=payload,
+                headers=headers,
+                timeout=aiohttp.ClientTimeout(total=timeout),
+            ) as resp:
+                if resp.status != 200:
+                    text = await resp.text()
+                    raise ApiError(
+                        f'OpenClaw API error {resp.status}: {text}',
+                        status=resp.status,
+                    )
+                data = await resp.json(content_type=None)
+
+        except (asyncio.TimeoutError, aiohttp.ServerTimeoutError):
+            return GetUpdatesResponse(ret=0, msgs=[], get_updates_buf=get_updates_buf)
+        except ApiError:
+            raise
+        except Exception as e:
+            if 'timeout' in str(e).lower():
+                return GetUpdatesResponse(ret=0, msgs=[], get_updates_buf=get_updates_buf)
+            raise
+
+        return _parse_get_updates_response(data)
+
+    async def send_message(
+        self,
+        to_user_id: str,
+        item_list: list[MessageItem],
+        context_token: str = '',
+    ) -> None:
+        """Send a message to a user."""
+        items_payload = [_message_item_to_dict(item) for item in item_list]
+
+        payload = {
+            'msg': {
+                'from_user_id': '',
+                'to_user_id': to_user_id,
+                'client_id': f'langbot-{uuid.uuid4().hex[:16]}',
+                'message_type': WeixinMessage.TYPE_BOT,
+                'message_state': WeixinMessage.STATE_FINISH,
+                'item_list': items_payload,
+                'context_token': context_token or None,
+            }
+        }
+        await self._post('ilink/bot/sendmessage', payload)
+
+    async def send_text(self, to_user_id: str, text: str, context_token: str = '') -> None:
+        """Send a plain text message, automatically chunking if too long."""
+        chunks = _chunk_text(text)
+        for chunk in chunks:
+            item = MessageItem(type=MessageItem.TEXT, text_item=TextItem(text=chunk))
+            await self.send_message(to_user_id, [item], context_token)
+
+    async def get_config(self, ilink_user_id: str, context_token: str = '') -> GetConfigResponse:
+        """Get bot config including typing_ticket."""
+        data = await self._post(
+            'ilink/bot/getconfig',
+            {'ilink_user_id': ilink_user_id, 'context_token': context_token or None},
+            timeout=DEFAULT_CONFIG_TIMEOUT,
+        )
+        return GetConfigResponse(
+            ret=data.get('ret'),
+            errmsg=data.get('errmsg'),
+            typing_ticket=data.get('typing_ticket'),
+        )
+
+    async def send_typing(self, ilink_user_id: str, typing_ticket: str, status: int = 1) -> None:
+        """Send typing indicator. status: 1=typing, 2=cancel."""
+        await self._post(
+            'ilink/bot/sendtyping',
+            {
+                'ilink_user_id': ilink_user_id,
+                'typing_ticket': typing_ticket,
+                'status': status,
+            },
+            timeout=DEFAULT_CONFIG_TIMEOUT,
+        )
+
+    async def stop_typing(self, ilink_user_id: str, typing_ticket: str) -> None:
+        """Cancel the typing indicator for a user."""
+        await self.send_typing(ilink_user_id, typing_ticket, status=2)
+
+    async def download_media(
+        self,
+        media: CDNMedia,
+    ) -> bytes:
+        """Download and decrypt a file from the WeChat CDN.
+
+        Args:
+            media: CDNMedia object with encrypt_query_param and aes_key.
+
+        Returns:
+            Decrypted file bytes.
+        """
+        from cryptography.hazmat.primitives.ciphers import Cipher, algorithms, modes
+        from cryptography.hazmat.primitives.padding import PKCS7
+
+        if not media.encrypt_query_param:
+            raise ApiError('CDN media has no encrypt_query_param', status=0)
+        if not media.aes_key:
+            raise ApiError('CDN media has no aes_key', status=0)
+
+        # Derive 16-byte AES key
+        # aes_key is base64-encoded; the decoded content may be:
+        #   - raw 16 bytes (direct AES key)
+        #   - 32-char hex string (decode hex to get 16 bytes)
+        raw = base64.b64decode(media.aes_key)
+        if len(raw) == 16:
+            aes_key = raw
+        elif len(raw) == 32:
+            # Hex-encoded 16-byte key
+            aes_key = bytes.fromhex(raw.decode('utf-8'))
+        else:
+            raise ApiError(f'Invalid AES key length: {len(raw)} (expected 16 or 32)', status=0)
+
+        # Download encrypted bytes from CDN
+        session = await self._get_session()
+        cdn_url = f'{CDN_BASE_URL}/download?encrypted_query_param={quote(media.encrypt_query_param, safe="")}'
+
+        async with session.get(cdn_url, timeout=aiohttp.ClientTimeout(total=120)) as resp:
+            if resp.status != 200:
+                text = await resp.text()
+                raise ApiError(f'CDN download failed: {resp.status} {text}', status=resp.status)
+            encrypted = await resp.read()
+
+        # Decrypt AES-128-ECB with PKCS7 padding
+        cipher = Cipher(algorithms.AES(aes_key), modes.ECB())
+        decryptor = cipher.decryptor()
+        padded = decryptor.update(encrypted) + decryptor.finalize()
+
+        unpadder = PKCS7(128).unpadder()
+        return unpadder.update(padded) + unpadder.finalize()
+
+    async def upload_media(
+        self,
+        file_bytes: bytes,
+        to_user_id: str,
+        media_type: int,
+    ) -> CDNMedia:
+        """Encrypt and upload media to WeChat CDN.
+
+        Args:
+            file_bytes: Raw file bytes to upload.
+            to_user_id: Recipient user ID.
+            media_type: 1=IMAGE, 2=VIDEO, 3=FILE, 4=VOICE.
+
+        Returns:
+            CDNMedia with encrypt_query_param and aes_key for use in sendMessage.
+        """
+        import hashlib
+
+        from cryptography.hazmat.primitives.ciphers import Cipher, algorithms, modes
+        from cryptography.hazmat.primitives.padding import PKCS7
+
+        # 1. Generate random 16-byte AES key
+        raw_key = os.urandom(16)
+        aes_key_hex = raw_key.hex()  # 32-char hex string
+
+        # 2. Encode key for CDNMedia: base64(hex_string) — same for all media types
+        # Matches official SDK: Buffer.from(aeskey_hex).toString("base64")
+        encoded_key = base64.b64encode(aes_key_hex.encode('utf-8')).decode('utf-8')
+
+        # 3. Encrypt file with AES-128-ECB + PKCS7
+        padder = PKCS7(128).padder()
+        padded = padder.update(file_bytes) + padder.finalize()
+        cipher = Cipher(algorithms.AES(raw_key), modes.ECB())
+        encryptor = cipher.encryptor()
+        encrypted = encryptor.update(padded) + encryptor.finalize()
+
+        # 4. Get upload URL
+        raw_md5 = hashlib.md5(file_bytes).hexdigest()
+        filekey = os.urandom(16).hex()  # 32-char hex, matches official SDK
+
+        upload_resp = await self.get_upload_url(
+            filekey=filekey,
+            media_type=media_type,
+            to_user_id=to_user_id,
+            rawsize=len(file_bytes),
+            rawfilemd5=raw_md5,
+            filesize=len(encrypted),
+            aeskey=aes_key_hex,  # hex string, as expected by the API
+        )
+
+        if not upload_resp.upload_param:
+            raise ApiError('Failed to get upload URL', status=0)
+
+        # 5. Upload to CDN
+        # upload_param is an opaque token from the server — pass it as-is
+        session = await self._get_session()
+        cdn_url = f'{CDN_BASE_URL}/upload?encrypted_query_param={quote(upload_resp.upload_param, safe="")}&filekey={quote(filekey, safe="")}'
+        logger.debug(
+            'CDN upload: url=%s raw_size=%d encrypted_size=%d md5=%s aeskey=%s',
+            cdn_url,
+            len(file_bytes),
+            len(encrypted),
+            raw_md5,
+            encoded_key,
+        )
+
+        async with session.post(
+            cdn_url,
+            data=encrypted,
+            headers={'Content-Type': 'application/octet-stream'},
+            timeout=aiohttp.ClientTimeout(total=120),
+        ) as resp:
+            if resp.status != 200:
+                text = await resp.text()
+                logger.error('CDN upload failed: status=%d url=%s body=%s', resp.status, cdn_url, text[:500])
+                raise ApiError(f'CDN upload failed: {resp.status} {text}', status=resp.status)
+            download_param = resp.headers.get('x-encrypted-param', '')
+
+        if not download_param:
+            raise ApiError('CDN upload succeeded but no x-encrypted-param returned', status=0)
+
+        return CDNMedia(
+            encrypt_query_param=download_param,
+            aes_key=encoded_key,
+            encrypt_type=1,
+        )
+
+    async def send_image(
+        self,
+        to_user_id: str,
+        image_bytes: bytes,
+        context_token: str = '',
+    ) -> None:
+        """Upload an image to CDN and send it."""
+        media = await self.upload_media(image_bytes, to_user_id, media_type=1)
+        item = MessageItem(
+            type=MessageItem.IMAGE,
+            image_item=ImageItem(
+                media=media,
+                aeskey=media.aes_key,
+            ),
+        )
+        await self.send_message(to_user_id, [item], context_token)
+
+    async def send_file(
+        self,
+        to_user_id: str,
+        file_bytes: bytes,
+        file_name: str,
+        context_token: str = '',
+    ) -> None:
+        """Upload a file to CDN and send it."""
+        import hashlib
+
+        media = await self.upload_media(file_bytes, to_user_id, media_type=3)
+        item = MessageItem(
+            type=MessageItem.FILE,
+            file_item=FileItem(
+                media=media,
+                file_name=file_name,
+                md5=hashlib.md5(file_bytes).hexdigest(),
+                len=str(len(file_bytes)),
+            ),
+        )
+        await self.send_message(to_user_id, [item], context_token)
+
+    async def send_voice(
+        self,
+        to_user_id: str,
+        voice_bytes: bytes,
+        playtime: int = 0,
+        context_token: str = '',
+    ) -> None:
+        """Upload a voice message to CDN and send it."""
+        media = await self.upload_media(voice_bytes, to_user_id, media_type=4)
+        item = MessageItem(
+            type=MessageItem.VOICE,
+            voice_item=VoiceItem(
+                media=media,
+                playtime=playtime,
+            ),
+        )
+        await self.send_message(to_user_id, [item], context_token)
+
+    async def get_upload_url(
+        self,
+        filekey: str,
+        media_type: int,
+        to_user_id: str,
+        rawsize: int,
+        rawfilemd5: str,
+        filesize: int,
+        thumb_rawsize: Optional[int] = None,
+        thumb_rawfilemd5: Optional[str] = None,
+        thumb_filesize: Optional[int] = None,
+        aeskey: Optional[str] = None,
+    ) -> GetUploadUrlResponse:
+        """Get a pre-signed CDN upload URL."""
+        payload: dict = {
+            'filekey': filekey,
+            'media_type': media_type,
+            'to_user_id': to_user_id,
+            'rawsize': rawsize,
+            'rawfilemd5': rawfilemd5,
+            'filesize': filesize,
+            'no_need_thumb': True,
+        }
+        if thumb_rawsize is not None:
+            payload['thumb_rawsize'] = thumb_rawsize
+        if thumb_rawfilemd5 is not None:
+            payload['thumb_rawfilemd5'] = thumb_rawfilemd5
+        if thumb_filesize is not None:
+            payload['thumb_filesize'] = thumb_filesize
+        if aeskey is not None:
+            payload['aeskey'] = aeskey
+
+        data = await self._post('ilink/bot/getuploadurl', payload)
+        logger.debug('get_upload_url response: %s', data)
+        return GetUploadUrlResponse(
+            upload_param=data.get('upload_param'),
+            thumb_upload_param=data.get('thumb_upload_param'),
+        )
+
+    # -----------------------------------------------------------------------
+    # QR Code Login
+    # -----------------------------------------------------------------------
+
+    async def fetch_qrcode(self, bot_type: str = DEFAULT_BOT_TYPE) -> QRCodeResponse:
+        """Fetch a QR code for WeChat login authorization (GET, no auth needed)."""
+        session = await self._get_session()
+        url = f'{self.base_url}/ilink/bot/get_bot_qrcode?bot_type={bot_type}'
+
+        async with session.get(url, timeout=aiohttp.ClientTimeout(total=DEFAULT_API_TIMEOUT)) as resp:
+            if resp.status != 200:
+                text = await resp.text()
+                raise ApiError(
+                    f'Failed to fetch QR code: {resp.status} {text}',
+                    status=resp.status,
+                )
+            data = await resp.json(content_type=None)
+
+        logger.debug(
+            'fetch_qrcode response: qrcode=%s, img=%s', data.get('qrcode'), bool(data.get('qrcode_img_content'))
+        )
+
+        return QRCodeResponse(
+            qrcode=data.get('qrcode'),
+            qrcode_img_content=data.get('qrcode_img_content'),
+        )
+
+    async def _fetch_qr_image_base64(self, url: str) -> str:
+        """Generate a QR code image from the URL and return a data URI string.
+
+        The qrcode_img_content URL points to an HTML page (not a raw image),
+        so we generate the QR code locally using the qrcode library.
+        """
+        import qrcode
+
+        qr = qrcode.QRCode(error_correction=qrcode.constants.ERROR_CORRECT_L)
+        qr.add_data(url)
+        qr.make(fit=True)
+        img = qr.make_image(fill_color='black', back_color='white')
+
+        buf = io.BytesIO()
+        img.save(buf, format='PNG')
+        b64 = base64.b64encode(buf.getvalue()).decode('utf-8')
+        return f'data:image/png;base64,{b64}'
+
+    async def poll_qrcode_status(self, qrcode: str) -> QRStatusResponse:
+        """Long-poll the QR code scan status (GET with iLink-App-ClientVersion header)."""
+        session = await self._get_session()
+        url = f'{self.base_url}/ilink/bot/get_qrcode_status?qrcode={quote(qrcode, safe="")}'
+        headers = {'iLink-App-ClientVersion': '1'}
+
+        try:
+            async with session.get(
+                url, headers=headers, timeout=aiohttp.ClientTimeout(total=DEFAULT_QR_POLL_TIMEOUT)
+            ) as resp:
+                if resp.status != 200:
+                    text = await resp.text()
+                    raise ApiError(
+                        f'Failed to poll QR status: {resp.status} {text}',
+                        status=resp.status,
+                    )
+                data = await resp.json(content_type=None)
+                logger.debug('QR status poll response: %s', data)
+        except (asyncio.TimeoutError, aiohttp.ServerTimeoutError):
+            return QRStatusResponse(status='wait')
+
+        return QRStatusResponse(
+            status=data.get('status'),
+            bot_token=data.get('bot_token'),
+            ilink_bot_id=data.get('ilink_bot_id'),
+            baseurl=data.get('baseurl'),
+            ilink_user_id=data.get('ilink_user_id'),
+        )
+
+    async def login(
+        self,
+        max_retries: int = 5,
+        poll_timeout_ms: int = 480_000,
+        on_qrcode: Optional[typing.Callable[[str, str], typing.Any]] = None,
+        on_status: Optional[typing.Callable[[str], typing.Any]] = None,
+    ) -> LoginResult:
+        """Complete QR code login flow with auto-retry on expiry.
+
+        Args:
+            max_retries: Max number of QR code refreshes on expiry.
+            poll_timeout_ms: Timeout per QR code in milliseconds.
+            on_qrcode: Callback(qr_image_base64, qr_url) called each time a
+                        new QR code is fetched. Use this to display the QR code.
+            on_status: Callback(status_str) called on each status poll change.
+
+        Returns:
+            LoginResult with token, base_url, and account_id.
+
+        Raises:
+            ApiError: On unrecoverable API errors.
+            Exception: If all retries are exhausted.
+        """
+        last_qr_base64: Optional[str] = None
+
+        for attempt in range(max_retries):
+            qr_resp = await self.fetch_qrcode()
+            if not qr_resp.qrcode or not qr_resp.qrcode_img_content:
+                raise ApiError('Failed to get QR code from server', status=0)
+
+            # Convert QR image to base64 and notify caller
+            last_qr_base64 = await self._fetch_qr_image_base64(qr_resp.qrcode_img_content)
+            if on_qrcode:
+                try:
+                    result = on_qrcode(last_qr_base64, qr_resp.qrcode_img_content)
+                    if asyncio.iscoroutine(result) or asyncio.isfuture(result):
+                        await result
+                except Exception as e:
+                    logger.warning('on_qrcode callback error: %s', e)
+
+            # Poll until confirmed / expired / timeout
+            loop = asyncio.get_running_loop()
+            deadline = loop.time() + poll_timeout_ms / 1000.0
+
+            while loop.time() < deadline:
+                try:
+                    status_resp = await self.poll_qrcode_status(qr_resp.qrcode)
+                except Exception as e:
+                    logger.error('Error polling QR status: %s', e)
+                    await asyncio.sleep(2)
+                    continue
+
+                if on_status:
+                    try:
+                        cb_result = on_status(status_resp.status or 'unknown')
+                        if asyncio.iscoroutine(cb_result) or asyncio.isfuture(cb_result):
+                            await cb_result
+                    except Exception as e:
+                        logger.warning('on_status callback error: %s', e)
+
+                if status_resp.status == 'confirmed' and status_resp.bot_token:
+                    new_base_url = status_resp.baseurl or self.base_url
+                    # Update this client instance as well
+                    self.token = status_resp.bot_token
+                    self.base_url = new_base_url.rstrip('/')
+                    return LoginResult(
+                        token=status_resp.bot_token,
+                        base_url=new_base_url,
+                        account_id=status_resp.ilink_bot_id or '',
+                        qr_image_base64=last_qr_base64,
+                    )
+
+                if status_resp.status == 'expired':
+                    break  # retry with a new QR code
+
+                await asyncio.sleep(1)
+            else:
+                # While-loop ended without break → poll timeout, treat as expired
+                pass
+
+            remaining = max_retries - attempt - 1
+            if remaining > 0:
+                logger.info('QR code expired, refreshing... (%d retries left)', remaining)
+            else:
+                raise ApiError('QR code login failed: max retries exceeded', status=0)
+
+        # Should not reach here, but just in case
+        raise ApiError('QR code login failed', status=0)
+
+
+# ---------------------------------------------------------------------------
+# Parsing helpers
+# ---------------------------------------------------------------------------
+
+
+def _parse_cdn_media(data: Optional[dict]) -> Optional[CDNMedia]:
+    if not data:
+        return None
+    return CDNMedia(
+        encrypt_query_param=data.get('encrypt_query_param'),
+        aes_key=data.get('aes_key'),
+        encrypt_type=data.get('encrypt_type'),
+    )
+
+
+def _parse_message_item(data: dict) -> MessageItem:
+    item = MessageItem(
+        type=data.get('type'),
+        create_time_ms=data.get('create_time_ms'),
+        update_time_ms=data.get('update_time_ms'),
+        is_completed=data.get('is_completed'),
+        msg_id=data.get('msg_id'),
+    )
+
+    if data.get('text_item'):
+        item.text_item = TextItem(text=data['text_item'].get('text'))
+
+    if data.get('image_item'):
+        img = data['image_item']
+        item.image_item = ImageItem(
+            media=_parse_cdn_media(img.get('media')),
+            thumb_media=_parse_cdn_media(img.get('thumb_media')),
+            aeskey=img.get('aeskey'),
+            url=img.get('url'),
+            mid_size=img.get('mid_size'),
+        )
+
+    if data.get('voice_item'):
+        v = data['voice_item']
+        item.voice_item = VoiceItem(
+            media=_parse_cdn_media(v.get('media')),
+            encode_type=v.get('encode_type'),
+            playtime=v.get('playtime'),
+            text=v.get('text'),
+        )
+
+    if data.get('file_item'):
+        f = data['file_item']
+        item.file_item = FileItem(
+            media=_parse_cdn_media(f.get('media')),
+            file_name=f.get('file_name'),
+            md5=f.get('md5'),
+            len=f.get('len'),
+        )
+
+    if data.get('video_item'):
+        vid = data['video_item']
+        item.video_item = VideoItem(
+            media=_parse_cdn_media(vid.get('media')),
+            video_size=vid.get('video_size'),
+            play_length=vid.get('play_length'),
+            video_md5=vid.get('video_md5'),
+            thumb_media=_parse_cdn_media(vid.get('thumb_media')),
+        )
+
+    if data.get('ref_msg'):
+        ref = data['ref_msg']
+        item.ref_msg = RefMessage(
+            title=ref.get('title'),
+            message_item=_parse_message_item(ref['message_item']) if ref.get('message_item') else None,
+        )
+
+    return item
+
+
+def _parse_weixin_message(data: dict) -> WeixinMessage:
+    msg = WeixinMessage(
+        seq=data.get('seq'),
+        message_id=data.get('message_id'),
+        from_user_id=data.get('from_user_id'),
+        to_user_id=data.get('to_user_id'),
+        client_id=data.get('client_id'),
+        create_time_ms=data.get('create_time_ms'),
+        session_id=data.get('session_id'),
+        group_id=data.get('group_id'),
+        message_type=data.get('message_type'),
+        message_state=data.get('message_state'),
+        context_token=data.get('context_token'),
+    )
+    if data.get('item_list'):
+        msg.item_list = [_parse_message_item(item) for item in data['item_list']]
+    return msg
+
+
+def _parse_get_updates_response(data: dict) -> GetUpdatesResponse:
+    resp = GetUpdatesResponse(
+        ret=data.get('ret'),
+        errcode=data.get('errcode'),
+        errmsg=data.get('errmsg'),
+        get_updates_buf=data.get('get_updates_buf'),
+        longpolling_timeout_ms=data.get('longpolling_timeout_ms'),
+    )
+    if data.get('msgs'):
+        resp.msgs = [_parse_weixin_message(m) for m in data['msgs']]
+    return resp
+
+
+def _cdn_media_to_dict(media: Optional[CDNMedia]) -> Optional[dict]:
+    if not media:
+        return None
+    d: dict = {}
+    if media.encrypt_query_param is not None:
+        d['encrypt_query_param'] = media.encrypt_query_param
+    if media.aes_key is not None:
+        d['aes_key'] = media.aes_key
+    if media.encrypt_type is not None:
+        d['encrypt_type'] = media.encrypt_type
+    return d or None
+
+
+def _message_item_to_dict(item: MessageItem) -> dict:
+    d: dict = {'type': item.type}
+
+    if item.text_item:
+        d['text_item'] = {'text': item.text_item.text}
+
+    if item.image_item:
+        img_d: dict = {}
+        if item.image_item.media:
+            img_d['media'] = _cdn_media_to_dict(item.image_item.media)
+        if item.image_item.mid_size is not None:
+            img_d['mid_size'] = item.image_item.mid_size
+        d['image_item'] = img_d
+
+    if item.voice_item:
+        voice_d: dict = {}
+        if item.voice_item.media:
+            voice_d['media'] = _cdn_media_to_dict(item.voice_item.media)
+        if item.voice_item.playtime is not None:
+            voice_d['playtime'] = item.voice_item.playtime
+        d['voice_item'] = voice_d
+
+    if item.file_item:
+        file_d: dict = {}
+        if item.file_item.media:
+            file_d['media'] = _cdn_media_to_dict(item.file_item.media)
+        if item.file_item.file_name:
+            file_d['file_name'] = item.file_item.file_name
+        if item.file_item.len:
+            file_d['len'] = item.file_item.len
+        d['file_item'] = file_d
+
+    if item.video_item:
+        vid_d: dict = {}
+        if item.video_item.media:
+            vid_d['media'] = _cdn_media_to_dict(item.video_item.media)
+        if item.video_item.video_size is not None:
+            vid_d['video_size'] = item.video_item.video_size
+        d['video_item'] = vid_d
+
+    return d
--- a/src/langbot/libs/openclaw_weixin_api/types.py
+++ b/src/langbot/libs/openclaw_weixin_api/types.py
@@ -0,0 +1,200 @@
+"""Type definitions for the OpenClaw WeChat API, mirroring the upstream protocol."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any, Optional
+
+SESSION_EXPIRED_ERRCODE = -14
+
+
+class ApiError(Exception):
+    """Structured error raised by the OpenClaw WeChat API."""
+
+    def __init__(
+        self,
+        message: str,
+        *,
+        status: int = 0,
+        code: int | None = None,
+        payload: Any = None,
+    ):
+        super().__init__(message)
+        self.status = status
+        self.code = code
+        self.payload = payload
+
+    @property
+    def is_session_expired(self) -> bool:
+        return self.code == SESSION_EXPIRED_ERRCODE
+
+
+@dataclass
+class CDNMedia:
+    encrypt_query_param: Optional[str] = None
+    aes_key: Optional[str] = None
+    encrypt_type: Optional[int] = None
+
+
+@dataclass
+class TextItem:
+    text: Optional[str] = None
+
+
+@dataclass
+class ImageItem:
+    media: Optional[CDNMedia] = None
+    thumb_media: Optional[CDNMedia] = None
+    aeskey: Optional[str] = None
+    url: Optional[str] = None
+    mid_size: Optional[int] = None
+    thumb_size: Optional[int] = None
+    thumb_height: Optional[int] = None
+    thumb_width: Optional[int] = None
+    hd_size: Optional[int] = None
+    _downloaded_bytes: Optional[bytes] = field(default=None, repr=False)
+
+
+@dataclass
+class VoiceItem:
+    media: Optional[CDNMedia] = None
+    encode_type: Optional[int] = None
+    bits_per_sample: Optional[int] = None
+    sample_rate: Optional[int] = None
+    playtime: Optional[int] = None
+    text: Optional[str] = None
+    _downloaded_bytes: Optional[bytes] = field(default=None, repr=False)
+
+
+@dataclass
+class FileItem:
+    media: Optional[CDNMedia] = None
+    file_name: Optional[str] = None
+    md5: Optional[str] = None
+    len: Optional[str] = None
+    _downloaded_bytes: Optional[bytes] = field(default=None, repr=False)
+
+
+@dataclass
+class VideoItem:
+    media: Optional[CDNMedia] = None
+    video_size: Optional[int] = None
+    play_length: Optional[int] = None
+    video_md5: Optional[str] = None
+    thumb_media: Optional[CDNMedia] = None
+    thumb_size: Optional[int] = None
+    thumb_height: Optional[int] = None
+    thumb_width: Optional[int] = None
+    _downloaded_bytes: Optional[bytes] = field(default=None, repr=False)
+
+
+@dataclass
+class RefMessage:
+    message_item: Optional[MessageItem] = None
+    title: Optional[str] = None
+
+
+@dataclass
+class MessageItem:
+    """A single content item inside a WeixinMessage."""
+
+    # Item types
+    NONE = 0
+    TEXT = 1
+    IMAGE = 2
+    VOICE = 3
+    FILE = 4
+    VIDEO = 5
+
+    type: Optional[int] = None
+    create_time_ms: Optional[int] = None
+    update_time_ms: Optional[int] = None
+    is_completed: Optional[bool] = None
+    msg_id: Optional[str] = None
+    ref_msg: Optional[RefMessage] = None
+    text_item: Optional[TextItem] = None
+    image_item: Optional[ImageItem] = None
+    voice_item: Optional[VoiceItem] = None
+    file_item: Optional[FileItem] = None
+    video_item: Optional[VideoItem] = None
+
+
+@dataclass
+class WeixinMessage:
+    """Unified message from getUpdates or for sendMessage."""
+
+    # Message types
+    TYPE_USER = 1
+    TYPE_BOT = 2
+
+    # Message states
+    STATE_NEW = 0
+    STATE_GENERATING = 1
+    STATE_FINISH = 2
+
+    seq: Optional[int] = None
+    message_id: Optional[int] = None
+    from_user_id: Optional[str] = None
+    to_user_id: Optional[str] = None
+    client_id: Optional[str] = None
+    create_time_ms: Optional[int] = None
+    update_time_ms: Optional[int] = None
+    delete_time_ms: Optional[int] = None
+    session_id: Optional[str] = None
+    group_id: Optional[str] = None
+    message_type: Optional[int] = None
+    message_state: Optional[int] = None
+    item_list: Optional[list[MessageItem]] = None
+    context_token: Optional[str] = None
+
+
+@dataclass
+class GetUpdatesResponse:
+    ret: Optional[int] = None
+    errcode: Optional[int] = None
+    errmsg: Optional[str] = None
+    msgs: list[WeixinMessage] = field(default_factory=list)
+    get_updates_buf: Optional[str] = None
+    longpolling_timeout_ms: Optional[int] = None
+
+
+@dataclass
+class GetConfigResponse:
+    ret: Optional[int] = None
+    errmsg: Optional[str] = None
+    typing_ticket: Optional[str] = None
+
+
+@dataclass
+class GetUploadUrlResponse:
+    upload_param: Optional[str] = None
+    thumb_upload_param: Optional[str] = None
+
+
+@dataclass
+class QRCodeResponse:
+    """Response from get_bot_qrcode endpoint."""
+
+    qrcode: Optional[str] = None
+    qrcode_img_content: Optional[str] = None
+
+
+@dataclass
+class QRStatusResponse:
+    """Response from get_qrcode_status endpoint."""
+
+    status: Optional[str] = None  # "wait" | "scaned" | "confirmed" | "expired"
+    bot_token: Optional[str] = None
+    ilink_bot_id: Optional[str] = None
+    baseurl: Optional[str] = None
+    ilink_user_id: Optional[str] = None
+
+
+@dataclass
+class LoginResult:
+    """Result returned by the login flow."""
+
+    token: str
+    base_url: str
+    account_id: str
+    qr_image_base64: Optional[str] = None  # data URI of the last QR code shown
--- a/src/langbot/libs/wechatpad_api/util/http_util.py
+++ b/src/langbot/libs/wechatpad_api/util/http_util.py
@@ -1,5 +1,5 @@
 import requests
-import aiohttp
+from langbot.pkg.utils import httpclient


 def post_json(base_url, token, data=None):
@@ -63,16 +63,16 @@ async def async_request(
    """
    headers = {'Content-Type': 'application/json'}
    url = f'{base_url}?key={token_key}'
-    async with aiohttp.ClientSession() as session:
-        async with session.request(
-            method=method, url=url, params=params, headers=headers, data=data, json=json
-        ) as response:
-            response.raise_for_status()  # 如果状态码不是200，抛出异常
-            result = await response.json()
-            # print(result)
-            return result
-            # if result.get('Code') == 200:
-            #
-            #     return await result
-            # else:
-            #     raise RuntimeError("请求失败",response.text)
+    session = httpclient.get_session()
+    async with session.request(
+        method=method, url=url, params=params, headers=headers, data=data, json=json
+    ) as response:
+        response.raise_for_status()  # 如果状态码不是200，抛出异常
+        result = await response.json()
+        # print(result)
+        return result
+        # if result.get('Code') == 200:
+        #
+        #     return await result
+        # else:
+        #     raise RuntimeError("请求失败",response.text)
--- a/src/langbot/libs/wecom_ai_bot_api/api.py
+++ b/src/langbot/libs/wecom_ai_bot_api/api.py
@@ -199,6 +199,253 @@ class StreamSessionManager:
                self._msg_index.pop(msg_id, None)


+async def download_encrypted_file(download_url: str, encoding_aes_key: str, logger: EventLogger) -> Optional[str]:
+    """Download an AES-encrypted file from WeChat Work and return as data URI.
+
+    Args:
+        download_url: The encrypted file download URL.
+        encoding_aes_key: The AES key used for decryption (base64-encoded, without trailing '=').
+        logger: Logger instance.
+
+    Returns:
+        A data URI string (e.g. 'data:image/jpeg;base64,...') or None on failure.
+    """
+    if not download_url:
+        return None
+    async with httpx.AsyncClient() as client:
+        response = await client.get(download_url)
+        if response.status_code != 200:
+            await logger.error(f'failed to get file: {response.text}')
+            return None
+        encrypted_bytes = response.content
+
+    aes_key = base64.b64decode(encoding_aes_key + '=')
+    iv = aes_key[:16]
+
+    cipher = AES.new(aes_key, AES.MODE_CBC, iv)
+    decrypted = cipher.decrypt(encrypted_bytes)
+
+    pad_len = decrypted[-1]
+    decrypted = decrypted[:-pad_len]
+
+    if decrypted.startswith(b'\xff\xd8'):
+        mime_type = 'image/jpeg'
+    elif decrypted.startswith(b'\x89PNG'):
+        mime_type = 'image/png'
+    elif decrypted.startswith((b'GIF87a', b'GIF89a')):
+        mime_type = 'image/gif'
+    elif decrypted.startswith(b'BM'):
+        mime_type = 'image/bmp'
+    elif decrypted.startswith(b'II*\x00') or decrypted.startswith(b'MM\x00*'):
+        mime_type = 'image/tiff'
+    else:
+        mime_type = 'application/octet-stream'
+
+    base64_str = base64.b64encode(decrypted).decode('utf-8')
+    return f'data:{mime_type};base64,{base64_str}'
+
+
+async def parse_wecom_bot_message(
+    msg_json: dict[str, Any], encoding_aes_key: str, logger: EventLogger
+) -> dict[str, Any]:
+    """Parse a decrypted WeChat Work AI Bot message JSON into a unified message dict.
+
+    This is the shared message parsing logic used by both webhook and WebSocket modes.
+
+    Args:
+        msg_json: The decrypted message JSON from WeChat Work.
+        encoding_aes_key: AES key for file decryption.
+        logger: Logger instance.
+
+    Returns:
+        A dict suitable for constructing a WecomBotEvent.
+    """
+    message_data: dict[str, Any] = {}
+
+    msg_type = msg_json.get('msgtype', '')
+    if msg_type:
+        message_data['msgtype'] = msg_type
+
+    if msg_json.get('chattype', '') == 'single':
+        message_data['type'] = 'single'
+    elif msg_json.get('chattype', '') == 'group':
+        message_data['type'] = 'group'
+
+    max_inline_file_size = 5 * 1024 * 1024
+
+    async def _safe_download(url: str):
+        if not url:
+            return None
+        return await download_encrypted_file(url, encoding_aes_key, logger)
+
+    if msg_type == 'text':
+        message_data['content'] = msg_json.get('text', {}).get('content')
+    elif msg_type == 'markdown':
+        message_data['content'] = msg_json.get('markdown', {}).get('content') or msg_json.get('text', {}).get(
+            'content', ''
+        )
+    elif msg_type == 'image':
+        picurl = msg_json.get('image', {}).get('url', '')
+        base64_data = await _safe_download(picurl)
+        if base64_data:
+            message_data['picurl'] = base64_data
+            message_data['images'] = [base64_data]
+    elif msg_type == 'voice':
+        voice_info = msg_json.get('voice', {}) or {}
+        download_url = voice_info.get('url')
+        message_data['voice'] = {
+            'url': download_url,
+            'md5sum': voice_info.get('md5sum') or voice_info.get('md5'),
+            'filesize': voice_info.get('filesize') or voice_info.get('size'),
+            'sdkfileid': voice_info.get('sdkfileid') or voice_info.get('fileid'),
+        }
+        if voice_info.get('content'):
+            message_data['content'] = voice_info.get('content')
+        if (message_data['voice'].get('filesize') or 0) <= max_inline_file_size:
+            voice_base64 = await _safe_download(download_url)
+            if voice_base64:
+                message_data['voice']['base64'] = voice_base64
+    elif msg_type == 'video':
+        video_info = msg_json.get('video', {}) or {}
+        download_url = video_info.get('url')
+        video_data = {
+            'url': download_url,
+            'filesize': video_info.get('filesize') or video_info.get('size'),
+            'sdkfileid': video_info.get('sdkfileid') or video_info.get('fileid'),
+            'md5sum': video_info.get('md5sum') or video_info.get('md5'),
+            'filename': video_info.get('filename') or video_info.get('name'),
+        }
+        if (video_data.get('filesize') or 0) <= max_inline_file_size:
+            video_base64 = await _safe_download(download_url)
+            if video_base64:
+                video_data['base64'] = video_base64
+        message_data['video'] = video_data
+    elif msg_type == 'file':
+        file_info = msg_json.get('file', {}) or {}
+        download_url = file_info.get('url') or file_info.get('fileurl')
+        file_data = {
+            'filename': file_info.get('filename') or file_info.get('name'),
+            'filesize': file_info.get('filesize') or file_info.get('size'),
+            'md5sum': file_info.get('md5sum') or file_info.get('md5'),
+            'sdkfileid': file_info.get('sdkfileid') or file_info.get('fileid'),
+            'download_url': download_url,
+            'extra': file_info,
+        }
+        if (file_data.get('filesize') or 0) <= max_inline_file_size:
+            file_base64 = await _safe_download(download_url)
+            if file_base64:
+                file_data['base64'] = file_base64
+        message_data['file'] = file_data
+    elif msg_type == 'link':
+        message_data['link'] = msg_json.get('link', {})
+        if not message_data.get('content'):
+            title = message_data['link'].get('title', '')
+            desc = message_data['link'].get('description') or message_data['link'].get('digest', '')
+            message_data['content'] = '\n'.join(filter(None, [title, desc]))
+    elif msg_type == 'mixed':
+        items = msg_json.get('mixed', {}).get('msg_item', [])
+        texts = []
+        images = []
+        files = []
+        voices = []
+        videos = []
+        links = []
+        for item in items:
+            item_type = item.get('msgtype')
+            if item_type == 'text':
+                texts.append(item.get('text', {}).get('content', ''))
+            elif item_type == 'image':
+                img_url = item.get('image', {}).get('url')
+                base64_data = await _safe_download(img_url)
+                if base64_data:
+                    images.append(base64_data)
+            elif item_type == 'file':
+                file_info = item.get('file', {}) or {}
+                download_url = file_info.get('url') or file_info.get('fileurl')
+                file_data = {
+                    'filename': file_info.get('filename') or file_info.get('name'),
+                    'filesize': file_info.get('filesize') or file_info.get('size'),
+                    'md5sum': file_info.get('md5sum') or file_info.get('md5'),
+                    'sdkfileid': file_info.get('sdkfileid') or file_info.get('fileid'),
+                    'download_url': download_url,
+                    'extra': file_info,
+                }
+                if (file_data.get('filesize') or 0) <= max_inline_file_size:
+                    file_base64 = await _safe_download(download_url)
+                    if file_base64:
+                        file_data['base64'] = file_base64
+                files.append(file_data)
+            elif item_type == 'voice':
+                voice_info = item.get('voice', {}) or {}
+                download_url = voice_info.get('url')
+                voice_data = {
+                    'url': download_url,
+                    'md5sum': voice_info.get('md5sum') or voice_info.get('md5'),
+                    'filesize': voice_info.get('filesize') or voice_info.get('size'),
+                    'sdkfileid': voice_info.get('sdkfileid') or voice_info.get('fileid'),
+                }
+                if voice_info.get('content'):
+                    texts.append(voice_info.get('content'))
+                if (voice_data.get('filesize') or 0) <= max_inline_file_size:
+                    voice_base64 = await _safe_download(download_url)
+                    if voice_base64:
+                        voice_data['base64'] = voice_base64
+                voices.append(voice_data)
+            elif item_type == 'video':
+                video_info = item.get('video', {}) or {}
+                download_url = video_info.get('url')
+                video_data = {
+                    'url': download_url,
+                    'filesize': video_info.get('filesize') or video_info.get('size'),
+                    'sdkfileid': video_info.get('sdkfileid') or video_info.get('fileid'),
+                    'md5sum': video_info.get('md5sum') or video_info.get('md5'),
+                    'filename': video_info.get('filename') or video_info.get('name'),
+                }
+                if (video_data.get('filesize') or 0) <= max_inline_file_size:
+                    video_base64 = await _safe_download(download_url)
+                    if video_base64:
+                        video_data['base64'] = video_base64
+                videos.append(video_data)
+            elif item_type == 'link':
+                links.append(item.get('link', {}))
+
+        if texts:
+            message_data['content'] = ' '.join(texts)
+        if images:
+            message_data['images'] = images
+            message_data['picurl'] = images[0]
+        if files:
+            message_data['files'] = files
+            message_data['file'] = files[0]
+        if voices:
+            message_data['voices'] = voices
+            message_data['voice'] = voices[0]
+        if videos:
+            message_data['videos'] = videos
+            message_data['video'] = videos[0]
+        if links:
+            message_data['link'] = links[0]
+        if items:
+            message_data['attachments'] = items
+    else:
+        message_data['raw_msg'] = msg_json
+
+    from_info = msg_json.get('from', {})
+    message_data['userid'] = from_info.get('userid', '')
+    message_data['username'] = from_info.get('alias', '') or from_info.get('name', '') or from_info.get('userid', '')
+
+    if msg_json.get('chattype', '') == 'group':
+        message_data['chatid'] = msg_json.get('chatid', '')
+        message_data['chatname'] = msg_json.get('chatname', '') or msg_json.get('chatid', '')
+
+    message_data['msgid'] = msg_json.get('msgid', '')
+
+    if msg_json.get('aibotid'):
+        message_data['aibotid'] = msg_json.get('aibotid', '')
+
+    return message_data
+
+
 class WecomBotClient:
    def __init__(self, Token: str, EnCodingAESKey: str, Corpid: str, logger: EventLogger, unified_mode: bool = False):
        """企业微信智能机器人客户端。
@@ -455,196 +702,7 @@ class WecomBotClient:
        return await self._handle_post_initial_response(msg_json, nonce)

    async def get_message(self, msg_json):
-        message_data = {}
-
-        msg_type = msg_json.get('msgtype', '')
-        if msg_type:
-            message_data['msgtype'] = msg_type
-
-        if msg_json.get('chattype', '') == 'single':
-            message_data['type'] = 'single'
-        elif msg_json.get('chattype', '') == 'group':
-            message_data['type'] = 'group'
-
-        max_inline_file_size = 5 * 1024 * 1024  # avoid decoding very large payloads by default
-
-        async def _safe_download(url: str):
-            if not url:
-                return None
-            return await self.download_url_to_base64(url, self.EnCodingAESKey)
-
-        if msg_type == 'text':
-            message_data['content'] = msg_json.get('text', {}).get('content')
-        elif msg_type == 'markdown':
-            message_data['content'] = msg_json.get('markdown', {}).get('content') or msg_json.get('text', {}).get(
-                'content', ''
-            )
-        elif msg_type == 'image':
-            picurl = msg_json.get('image', {}).get('url', '')
-            base64_data = await _safe_download(picurl)
-            if base64_data:
-                message_data['picurl'] = base64_data
-                message_data['images'] = [base64_data]
-        elif msg_type == 'voice':
-            voice_info = msg_json.get('voice', {}) or {}
-            download_url = voice_info.get('url')
-            message_data['voice'] = {
-                'url': download_url,
-                'md5sum': voice_info.get('md5sum') or voice_info.get('md5'),
-                'filesize': voice_info.get('filesize') or voice_info.get('size'),
-                'sdkfileid': voice_info.get('sdkfileid') or voice_info.get('fileid'),
-            }
-            # 企业微信智能转写文本（如果已有）直接复用，避免重复转写
-            if voice_info.get('content'):
-                message_data['content'] = voice_info.get('content')
-            if (message_data['voice'].get('filesize') or 0) <= max_inline_file_size:
-                voice_base64 = await _safe_download(download_url)
-                if voice_base64:
-                    message_data['voice']['base64'] = voice_base64
-        elif msg_type == 'video':
-            video_info = msg_json.get('video', {}) or {}
-            download_url = video_info.get('url')
-            video_data = {
-                'url': download_url,
-                'filesize': video_info.get('filesize') or video_info.get('size'),
-                'sdkfileid': video_info.get('sdkfileid') or video_info.get('fileid'),
-                'md5sum': video_info.get('md5sum') or video_info.get('md5'),
-                'filename': video_info.get('filename') or video_info.get('name'),
-            }
-            if (video_data.get('filesize') or 0) <= max_inline_file_size:
-                video_base64 = await _safe_download(download_url)
-                if video_base64:
-                    video_data['base64'] = video_base64
-            message_data['video'] = video_data
-        elif msg_type == 'file':
-            file_info = msg_json.get('file', {}) or {}
-            download_url = file_info.get('url') or file_info.get('fileurl')
-            file_data = {
-                'filename': file_info.get('filename') or file_info.get('name'),
-                'filesize': file_info.get('filesize') or file_info.get('size'),
-                'md5sum': file_info.get('md5sum') or file_info.get('md5'),
-                'sdkfileid': file_info.get('sdkfileid') or file_info.get('fileid'),
-                'download_url': download_url,
-                'extra': file_info,
-            }
-            if (file_data.get('filesize') or 0) <= max_inline_file_size:
-                file_base64 = await _safe_download(download_url)
-                if file_base64:
-                    file_data['base64'] = file_base64
-            message_data['file'] = file_data
-        elif msg_type == 'link':
-            message_data['link'] = msg_json.get('link', {})
-            if not message_data.get('content'):
-                title = message_data['link'].get('title', '')
-                desc = message_data['link'].get('description') or message_data['link'].get('digest', '')
-                message_data['content'] = '\n'.join(filter(None, [title, desc]))
-        elif msg_type == 'mixed':
-            items = msg_json.get('mixed', {}).get('msg_item', [])
-            texts = []
-            images = []
-            files = []
-            voices = []
-            videos = []
-            links = []
-            for item in items:
-                item_type = item.get('msgtype')
-                if item_type == 'text':
-                    texts.append(item.get('text', {}).get('content', ''))
-                elif item_type == 'image':
-                    img_url = item.get('image', {}).get('url')
-                    base64_data = await _safe_download(img_url)
-                    if base64_data:
-                        images.append(base64_data)
-                elif item_type == 'file':
-                    file_info = item.get('file', {}) or {}
-                    download_url = file_info.get('url') or file_info.get('fileurl')
-                    file_data = {
-                        'filename': file_info.get('filename') or file_info.get('name'),
-                        'filesize': file_info.get('filesize') or file_info.get('size'),
-                        'md5sum': file_info.get('md5sum') or file_info.get('md5'),
-                        'sdkfileid': file_info.get('sdkfileid') or file_info.get('fileid'),
-                        'download_url': download_url,
-                        'extra': file_info,
-                    }
-                    if (file_data.get('filesize') or 0) <= max_inline_file_size:
-                        file_base64 = await _safe_download(download_url)
-                        if file_base64:
-                            file_data['base64'] = file_base64
-                    files.append(file_data)
-                elif item_type == 'voice':
-                    voice_info = item.get('voice', {}) or {}
-                    download_url = voice_info.get('url')
-                    voice_data = {
-                        'url': download_url,
-                        'md5sum': voice_info.get('md5sum') or voice_info.get('md5'),
-                        'filesize': voice_info.get('filesize') or voice_info.get('size'),
-                        'sdkfileid': voice_info.get('sdkfileid') or voice_info.get('fileid'),
-                    }
-                    if voice_info.get('content'):
-                        texts.append(voice_info.get('content'))
-                    if (voice_data.get('filesize') or 0) <= max_inline_file_size:
-                        voice_base64 = await _safe_download(download_url)
-                        if voice_base64:
-                            voice_data['base64'] = voice_base64
-                    voices.append(voice_data)
-                elif item_type == 'video':
-                    video_info = item.get('video', {}) or {}
-                    download_url = video_info.get('url')
-                    video_data = {
-                        'url': download_url,
-                        'filesize': video_info.get('filesize') or video_info.get('size'),
-                        'sdkfileid': video_info.get('sdkfileid') or video_info.get('fileid'),
-                        'md5sum': video_info.get('md5sum') or video_info.get('md5'),
-                        'filename': video_info.get('filename') or video_info.get('name'),
-                    }
-                    if (video_data.get('filesize') or 0) <= max_inline_file_size:
-                        video_base64 = await _safe_download(download_url)
-                        if video_base64:
-                            video_data['base64'] = video_base64
-                    videos.append(video_data)
-                elif item_type == 'link':
-                    links.append(item.get('link', {}))
-
-            if texts:
-                message_data['content'] = ' '.join(texts)  # 拼接所有 text
-            if images:
-                message_data['images'] = images
-                message_data['picurl'] = images[0]  # 只保留第一个 image
-            if files:
-                message_data['files'] = files
-                message_data['file'] = files[0]
-            if voices:
-                message_data['voices'] = voices
-                message_data['voice'] = voices[0]
-            if videos:
-                message_data['videos'] = videos
-                message_data['video'] = videos[0]
-            if links:
-                message_data['link'] = links[0]
-            if items:
-                message_data['attachments'] = items
-        else:
-            message_data['raw_msg'] = msg_json
-
-        # Extract user information
-        from_info = msg_json.get('from', {})
-        message_data['userid'] = from_info.get('userid', '')
-        message_data['username'] = (
-            from_info.get('alias', '') or from_info.get('name', '') or from_info.get('userid', '')
-        )
-
-        # Extract chat/group information
-        if msg_json.get('chattype', '') == 'group':
-            message_data['chatid'] = msg_json.get('chatid', '')
-            # Try to get group name if available
-            message_data['chatname'] = msg_json.get('chatname', '') or msg_json.get('chatid', '')
-
-        message_data['msgid'] = msg_json.get('msgid', '')
-
-        if msg_json.get('aibotid'):
-            message_data['aibotid'] = msg_json.get('aibotid', '')
-
-        return message_data
+        return await parse_wecom_bot_message(msg_json, self.EnCodingAESKey, self.logger)

    async def _handle_message(self, event: wecombotevent.WecomBotEvent):
        """
@@ -712,39 +770,7 @@ class WecomBotClient:
        return decorator

    async def download_url_to_base64(self, download_url, encoding_aes_key):
-        async with httpx.AsyncClient() as client:
-            response = await client.get(download_url)
-            if response.status_code != 200:
-                await self.logger.error(f'failed to get file: {response.text}')
-                return None
-
-            encrypted_bytes = response.content
-
-        aes_key = base64.b64decode(encoding_aes_key + '=')  # base64 补齐
-        iv = aes_key[:16]
-
-        cipher = AES.new(aes_key, AES.MODE_CBC, iv)
-        decrypted = cipher.decrypt(encrypted_bytes)
-
-        pad_len = decrypted[-1]
-        decrypted = decrypted[:-pad_len]
-
-        if decrypted.startswith(b'\xff\xd8'):  # JPEG
-            mime_type = 'image/jpeg'
-        elif decrypted.startswith(b'\x89PNG'):  # PNG
-            mime_type = 'image/png'
-        elif decrypted.startswith((b'GIF87a', b'GIF89a')):  # GIF
-            mime_type = 'image/gif'
-        elif decrypted.startswith(b'BM'):  # BMP
-            mime_type = 'image/bmp'
-        elif decrypted.startswith(b'II*\x00') or decrypted.startswith(b'MM\x00*'):  # TIFF
-            mime_type = 'image/tiff'
-        else:
-            mime_type = 'application/octet-stream'
-
-        # 转 base64
-        base64_str = base64.b64encode(decrypted).decode('utf-8')
-        return f'data:{mime_type};base64,{base64_str}'
+        return await download_encrypted_file(download_url, encoding_aes_key, self.logger)

    async def run_task(self, host: str, port: int, *args, **kwargs):
        """
--- a/src/langbot/libs/wecom_ai_bot_api/ws_client.py
+++ b/src/langbot/libs/wecom_ai_bot_api/ws_client.py
@@ -0,0 +1,596 @@
+"""WeChat Work AI Bot WebSocket long connection client.
+
+Implements the WebSocket protocol for receiving messages and sending replies
+via a persistent connection to wss://openws.work.weixin.qq.com, as an
+alternative to the HTTP callback (webhook) mode.
+
+Protocol reference: https://developer.work.weixin.qq.com/document/path/101463
+Official Node.js SDK: https://github.com/WecomTeam/aibot-node-sdk
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import secrets
+import time
+import traceback
+from typing import Any, Callable, Optional
+
+import aiohttp
+
+from langbot.libs.wecom_ai_bot_api import wecombotevent
+from langbot.libs.wecom_ai_bot_api.api import parse_wecom_bot_message
+from langbot.pkg.platform.logger import EventLogger
+
+DEFAULT_WS_URL = 'wss://openws.work.weixin.qq.com'
+
+# WebSocket frame command constants
+CMD_SUBSCRIBE = 'aibot_subscribe'
+CMD_HEARTBEAT = 'ping'
+CMD_MSG_CALLBACK = 'aibot_msg_callback'
+CMD_EVENT_CALLBACK = 'aibot_event_callback'
+CMD_RESPOND_MSG = 'aibot_respond_msg'
+CMD_RESPOND_WELCOME = 'aibot_respond_welcome_msg'
+CMD_RESPOND_UPDATE = 'aibot_respond_update_msg'
+CMD_SEND_MSG = 'aibot_send_msg'
+
+
+def _generate_req_id(prefix: str) -> str:
+    """Generate a unique request ID in the format: {prefix}_{timestamp}_{random}."""
+    ts = int(time.time() * 1000)
+    rand = secrets.token_hex(4)
+    return f'{prefix}_{ts}_{rand}'
+
+
+class WecomBotWsClient:
+    """WeChat Work AI Bot WebSocket long connection client.
+
+    Provides message receiving, streaming reply, proactive message sending,
+    and event callback handling over a persistent WebSocket connection.
+    """
+
+    def __init__(
+        self,
+        bot_id: str,
+        secret: str,
+        logger: EventLogger,
+        encoding_aes_key: str = '',
+        ws_url: str = DEFAULT_WS_URL,
+        heartbeat_interval: float = 30.0,
+        max_reconnect_attempts: int = -1,
+        reconnect_base_delay: float = 1.0,
+        reconnect_max_delay: float = 30.0,
+    ):
+        self.bot_id = bot_id
+        self.secret = secret
+        self.logger = logger
+        self.encoding_aes_key = encoding_aes_key
+        self.ws_url = ws_url
+        self.heartbeat_interval = heartbeat_interval
+        self.max_reconnect_attempts = max_reconnect_attempts
+        self.reconnect_base_delay = reconnect_base_delay
+        self.reconnect_max_delay = reconnect_max_delay
+
+        self._ws: Optional[aiohttp.ClientWebSocketResponse] = None
+        self._session: Optional[aiohttp.ClientSession] = None
+        self._running = False
+        self._heartbeat_task: Optional[asyncio.Task] = None
+        self._missed_pong_count = 0
+        self._max_missed_pong = 2
+        self._reconnect_attempts = 0
+
+        # Message handler registry (same pattern as WecomBotClient)
+        self._message_handlers: dict[str, list[Callable]] = {}
+        # Message deduplication
+        self._msg_id_map: dict[str, int] = {}
+
+        # Pending ACK futures: req_id -> Future[dict]
+        self._pending_acks: dict[str, asyncio.Future] = {}
+        # Per-req_id serial reply queues
+        self._reply_queues: dict[str, asyncio.Queue] = {}
+        self._reply_workers: dict[str, asyncio.Task] = {}
+        self._reply_ack_timeout = 5.0
+
+        # Stream ID tracking for WebSocket mode
+        self._stream_ids: dict[str, str] = {}  # msg_id -> req_id|stream_id
+        # Dedup: skip sending when content hasn't changed
+        self._stream_last_content: dict[str, str] = {}  # msg_id -> last content sent
+
+    # ── Public API ──────────────────────────────────────────────────
+
+    async def connect(self):
+        """Connect to WebSocket server with automatic reconnection.
+
+        This method blocks until disconnect() is called or max reconnect
+        attempts are exhausted.
+        """
+        self._running = True
+        self._reconnect_attempts = 0
+
+        while self._running:
+            try:
+                await self._connect_once()
+            except Exception:
+                if not self._running:
+                    break
+                await self.logger.error(f'WebSocket connection error: {traceback.format_exc()}')
+
+            if not self._running:
+                break
+
+            # Reconnect with exponential backoff
+            if self.max_reconnect_attempts != -1 and self._reconnect_attempts >= self.max_reconnect_attempts:
+                await self.logger.error(f'Max reconnect attempts reached ({self.max_reconnect_attempts}), giving up')
+                break
+
+            self._reconnect_attempts += 1
+            delay = min(
+                self.reconnect_base_delay * (2 ** (self._reconnect_attempts - 1)),
+                self.reconnect_max_delay,
+            )
+            await self.logger.info(f'Reconnecting in {delay:.1f}s (attempt {self._reconnect_attempts})...')
+            await asyncio.sleep(delay)
+
+    async def disconnect(self):
+        """Gracefully disconnect from the WebSocket server."""
+        self._running = False
+        if self._heartbeat_task and not self._heartbeat_task.done():
+            self._heartbeat_task.cancel()
+        for task in self._reply_workers.values():
+            if not task.done():
+                task.cancel()
+        if self._ws and not self._ws.closed:
+            await self._ws.close()
+        self._ws = None
+        if self._session and not self._session.closed:
+            await self._session.close()
+        self._session = None
+
+    def on_message(self, msg_type: str) -> Callable:
+        """Decorator to register a message handler.
+
+        Same interface as WecomBotClient.on_message for compatibility.
+
+        Args:
+            msg_type: 'single', 'group', or specific message type.
+        """
+
+        def decorator(func: Callable[[wecombotevent.WecomBotEvent], Any]):
+            if msg_type not in self._message_handlers:
+                self._message_handlers[msg_type] = []
+            self._message_handlers[msg_type].append(func)
+            return func
+
+        return decorator
+
+    async def reply_stream(
+        self,
+        req_id: str,
+        stream_id: str,
+        content: str,
+        finish: bool = False,
+    ) -> Optional[dict]:
+        """Send a streaming reply frame.
+
+        Args:
+            req_id: The req_id from the original message frame (must be passed through).
+            stream_id: The stream ID for this streaming session.
+            content: The content to send (supports Markdown).
+            finish: Whether this is the final chunk.
+
+        Returns:
+            The ACK frame dict, or None on failure.
+        """
+        body = {
+            'msgtype': 'stream',
+            'stream': {
+                'id': stream_id,
+                'finish': finish,
+                'content': content,
+            },
+        }
+        return await self._send_reply(req_id, body)
+
+    async def reply_text(self, req_id: str, content: str) -> Optional[dict]:
+        """Send a non-streaming text reply.
+
+        Args:
+            req_id: The req_id from the original message frame.
+            content: The text content to reply.
+
+        Returns:
+            The ACK frame dict, or None on failure.
+        """
+        body = {
+            'msgtype': 'markdown',
+            'markdown': {
+                'content': content,
+            },
+        }
+        return await self._send_reply(req_id, body)
+
+    async def send_message(self, chat_id: str, content: str, msgtype: str = 'markdown') -> Optional[dict]:
+        """Proactively send a message to a specified chat.
+
+        Args:
+            chat_id: The chat ID (userid for single chat, chatid for group chat).
+            content: The message content.
+            msgtype: Message type, 'markdown' by default.
+
+        Returns:
+            The ACK frame dict, or None on failure.
+        """
+        req_id = _generate_req_id(CMD_SEND_MSG)
+        body: dict[str, Any] = {
+            'chatid': chat_id,
+            'msgtype': msgtype,
+        }
+        if msgtype == 'markdown':
+            body['markdown'] = {'content': content}
+        elif msgtype == 'text':
+            body['text'] = {'content': content}
+        return await self._send_reply(req_id, body, cmd=CMD_SEND_MSG)
+
+    async def push_stream_chunk(self, msg_id: str, content: str, is_final: bool = False) -> bool:
+        """Push a streaming chunk for a given message ID.
+
+        Compatible interface with WecomBotClient.push_stream_chunk.
+
+        Args:
+            msg_id: The original message ID.
+            content: The cumulative content from the pipeline.
+            is_final: Whether this is the final chunk.
+
+        Returns:
+            True if the stream session exists and chunk was sent.
+        """
+        key = self._stream_ids.get(msg_id)
+        if not key:
+            return False
+        req_id, stream_id = key.split('|', 1)
+        try:
+            # Skip sending if content hasn't changed (e.g. during tool call argument streaming)
+            if not is_final and content == self._stream_last_content.get(msg_id):
+                return True
+            await self.reply_stream(req_id, stream_id, content, finish=is_final)
+            self._stream_last_content[msg_id] = content
+            if is_final:
+                self._stream_ids.pop(msg_id, None)
+                self._stream_last_content.pop(msg_id, None)
+            return True
+        except Exception:
+            await self.logger.error(f'Failed to push stream chunk: {traceback.format_exc()}')
+            return False
+
+    async def set_message(self, msg_id: str, content: str):
+        """Fallback: send content as a final stream chunk or direct reply.
+
+        Compatible interface with WecomBotClient.set_message.
+        """
+        handled = await self.push_stream_chunk(msg_id, content, is_final=True)
+        if not handled:
+            await self.logger.warning(f'No active stream for msg_id={msg_id}, message dropped')
+
+    # ── Connection lifecycle ────────────────────────────────────────
+
+    async def _connect_once(self):
+        """Establish a single WebSocket connection, authenticate, and listen."""
+        await self.logger.info(f'Connecting to {self.ws_url}...')
+
+        self._session = aiohttp.ClientSession()
+        try:
+            self._ws = await self._session.ws_connect(self.ws_url)
+            self._missed_pong_count = 0
+            self._reconnect_attempts = 0
+            await self.logger.info('WebSocket connected, sending auth...')
+
+            await self._send_auth()
+
+            # Wait for auth response
+            auth_ok = await self._wait_for_auth()
+            if not auth_ok:
+                await self.logger.error('Authentication failed')
+                return
+
+            await self.logger.info('Authenticated successfully')
+
+            # Start heartbeat
+            self._heartbeat_task = asyncio.create_task(self._heartbeat_loop())
+
+            try:
+                await self._listen_loop()
+            finally:
+                if self._heartbeat_task and not self._heartbeat_task.done():
+                    self._heartbeat_task.cancel()
+                self._clear_pending_acks('Connection closed')
+        finally:
+            if self._ws and not self._ws.closed:
+                await self._ws.close()
+            self._ws = None
+            if self._session and not self._session.closed:
+                await self._session.close()
+            self._session = None
+
+    async def _send_auth(self):
+        """Send the authentication frame."""
+        frame = {
+            'cmd': CMD_SUBSCRIBE,
+            'headers': {'req_id': _generate_req_id(CMD_SUBSCRIBE)},
+            'body': {
+                'bot_id': self.bot_id,
+                'secret': self.secret,
+            },
+        }
+        await self._send_frame(frame)
+
+    async def _wait_for_auth(self) -> bool:
+        """Wait for and validate the authentication response."""
+        try:
+            msg = await asyncio.wait_for(self._ws.receive(), timeout=10.0)
+            if msg.type in (aiohttp.WSMsgType.TEXT,):
+                frame = json.loads(msg.data)
+                req_id = frame.get('headers', {}).get('req_id', '')
+                if req_id.startswith(CMD_SUBSCRIBE) and frame.get('errcode') == 0:
+                    return True
+                await self.logger.error(f'Auth response: errcode={frame.get("errcode")}, errmsg={frame.get("errmsg")}')
+                return False
+            elif msg.type in (aiohttp.WSMsgType.ERROR, aiohttp.WSMsgType.CLOSED, aiohttp.WSMsgType.CLOSING):
+                await self.logger.error(f'WebSocket closed during auth: {msg.type}')
+                return False
+            await self.logger.error(f'Unexpected message type during auth: {msg.type}')
+            return False
+        except asyncio.TimeoutError:
+            await self.logger.error('Auth response timeout')
+            return False
+
+    async def _heartbeat_loop(self):
+        """Periodically send heartbeat pings."""
+        try:
+            while self._running and self._ws and not self._ws.closed:
+                await asyncio.sleep(self.heartbeat_interval)
+                if not self._running or not self._ws or self._ws.closed:
+                    break
+
+                if self._missed_pong_count >= self._max_missed_pong:
+                    await self.logger.warning(
+                        f'No heartbeat ack for {self._missed_pong_count} consecutive pings, connection considered dead'
+                    )
+                    await self._ws.close()
+                    break
+
+                self._missed_pong_count += 1
+                frame = {
+                    'cmd': CMD_HEARTBEAT,
+                    'headers': {'req_id': _generate_req_id(CMD_HEARTBEAT)},
+                }
+                try:
+                    await self._send_frame(frame)
+                except Exception:
+                    break
+        except asyncio.CancelledError:
+            pass
+
+    async def _listen_loop(self):
+        """Listen for incoming WebSocket frames and dispatch them."""
+        async for msg in self._ws:
+            if not self._running:
+                break
+            if msg.type == aiohttp.WSMsgType.TEXT:
+                try:
+                    frame = json.loads(msg.data)
+                    await self._handle_frame(frame)
+                except json.JSONDecodeError:
+                    await self.logger.error(f'Failed to parse WebSocket message: {str(msg.data)[:200]}')
+                except Exception:
+                    await self.logger.error(f'Error handling frame: {traceback.format_exc()}')
+            elif msg.type == aiohttp.WSMsgType.BINARY:
+                try:
+                    frame = json.loads(msg.data)
+                    await self._handle_frame(frame)
+                except Exception:
+                    await self.logger.error(f'Error handling binary frame: {traceback.format_exc()}')
+            elif msg.type in (aiohttp.WSMsgType.ERROR, aiohttp.WSMsgType.CLOSED, aiohttp.WSMsgType.CLOSING):
+                await self.logger.warning(f'WebSocket connection closed: {msg.type}')
+                break
+
+    # ── Frame handling ──────────────────────────────────────────────
+
+    async def _handle_frame(self, frame: dict):
+        """Route an incoming frame to the appropriate handler."""
+        cmd = frame.get('cmd', '')
+
+        # Message push
+        if cmd == CMD_MSG_CALLBACK:
+            asyncio.create_task(self._handle_message_callback(frame))
+            return
+
+        # Event push
+        if cmd == CMD_EVENT_CALLBACK:
+            asyncio.create_task(self._handle_event_callback(frame))
+            return
+
+        # No cmd → response/ACK frame, dispatch by req_id prefix
+        req_id = frame.get('headers', {}).get('req_id', '')
+
+        # Check pending ACKs first
+        if req_id in self._pending_acks:
+            future = self._pending_acks.pop(req_id)
+            if not future.done():
+                future.set_result(frame)
+            return
+
+        # Heartbeat response
+        if req_id.startswith(CMD_HEARTBEAT):
+            if frame.get('errcode') == 0:
+                self._missed_pong_count = 0
+            return
+
+        # Unknown frame
+        await self.logger.warning(f'Unknown frame: {json.dumps(frame, ensure_ascii=False)[:200]}')
+
+    async def _handle_message_callback(self, frame: dict):
+        """Handle an incoming message callback frame."""
+        try:
+            body = frame.get('body', {})
+            req_id = frame.get('headers', {}).get('req_id', '')
+
+            # Parse message using shared logic
+            message_data = await parse_wecom_bot_message(body, self.encoding_aes_key, self.logger)
+            if not message_data:
+                return
+
+            # Generate stream_id for this message and store the mapping
+            stream_id = _generate_req_id('stream')
+            msg_id = message_data.get('msgid', '')
+            if msg_id:
+                self._stream_ids[msg_id] = f'{req_id}|{stream_id}'
+            message_data['stream_id'] = stream_id
+            message_data['req_id'] = req_id
+
+            event = wecombotevent.WecomBotEvent(message_data)
+            await self._dispatch_event(event)
+        except Exception:
+            await self.logger.error(f'Error in message callback: {traceback.format_exc()}')
+
+    async def _handle_event_callback(self, frame: dict):
+        """Handle an incoming event callback frame (enter_chat, template_card_event, etc.)."""
+        try:
+            body = frame.get('body', {})
+            req_id = frame.get('headers', {}).get('req_id', '')
+
+            event_info = body.get('event', {})
+            event_type = event_info.get('eventtype', '')
+
+            message_data = {
+                'msgtype': 'event',
+                'type': body.get('chattype', 'single'),
+                'event': event_info,
+                'eventtype': event_type,
+                'msgid': body.get('msgid', ''),
+                'aibotid': body.get('aibotid', ''),
+                'req_id': req_id,
+            }
+
+            from_info = body.get('from', {})
+            message_data['userid'] = from_info.get('userid', '')
+            message_data['username'] = from_info.get('alias', '') or from_info.get('userid', '')
+
+            if body.get('chatid'):
+                message_data['chatid'] = body.get('chatid', '')
+
+            event = wecombotevent.WecomBotEvent(message_data)
+
+            # Dispatch to event-specific handlers
+            if event_type in self._message_handlers:
+                for handler in self._message_handlers[event_type]:
+                    await handler(event)
+
+            # Also dispatch to generic 'event' handlers
+            if 'event' in self._message_handlers:
+                for handler in self._message_handlers['event']:
+                    await handler(event)
+
+        except Exception:
+            await self.logger.error(f'Error in event callback: {traceback.format_exc()}')
+
+    async def _dispatch_event(self, event: wecombotevent.WecomBotEvent):
+        """Dispatch a message event to registered handlers with deduplication."""
+        try:
+            message_id = event.message_id
+            if message_id in self._msg_id_map:
+                self._msg_id_map[message_id] += 1
+                return
+            self._msg_id_map[message_id] = 1
+
+            msg_type = event.type
+            if msg_type in self._message_handlers:
+                for handler in self._message_handlers[msg_type]:
+                    await handler(event)
+        except Exception:
+            await self.logger.error(f'Error dispatching event: {traceback.format_exc()}')
+
+    # ── Reply sending with serial queue ─────────────────────────────
+
+    async def _send_reply(
+        self,
+        req_id: str,
+        body: dict,
+        cmd: str = CMD_RESPOND_MSG,
+    ) -> Optional[dict]:
+        """Send a reply frame and wait for ACK.
+
+        Replies with the same req_id are serialized to maintain ordering.
+        """
+        if not self._ws or self._ws.closed:
+            return None
+
+        frame = {
+            'cmd': cmd,
+            'headers': {'req_id': req_id},
+            'body': body,
+        }
+
+        # Ensure serial delivery per req_id
+        if req_id not in self._reply_queues:
+            self._reply_queues[req_id] = asyncio.Queue()
+            self._reply_workers[req_id] = asyncio.create_task(self._reply_queue_worker(req_id))
+
+        future: asyncio.Future = asyncio.get_event_loop().create_future()
+        await self._reply_queues[req_id].put((frame, future))
+        return await future
+
+    async def _reply_queue_worker(self, req_id: str):
+        """Process reply queue items serially for a given req_id."""
+        queue = self._reply_queues[req_id]
+        try:
+            while self._running:
+                try:
+                    frame, future = await asyncio.wait_for(queue.get(), timeout=60.0)
+                except asyncio.TimeoutError:
+                    # Queue idle, clean up worker
+                    break
+
+                try:
+                    ack = await self._send_and_wait_ack(frame)
+                    if not future.done():
+                        future.set_result(ack)
+                except Exception as e:
+                    if not future.done():
+                        future.set_exception(e)
+        except asyncio.CancelledError:
+            pass
+        finally:
+            self._reply_queues.pop(req_id, None)
+            self._reply_workers.pop(req_id, None)
+
+    async def _send_and_wait_ack(self, frame: dict) -> Optional[dict]:
+        """Send a frame and wait for the corresponding ACK."""
+        req_id = frame['headers']['req_id']
+        ack_future: asyncio.Future = asyncio.get_event_loop().create_future()
+        self._pending_acks[req_id] = ack_future
+
+        try:
+            await self._send_frame(frame)
+            result = await asyncio.wait_for(ack_future, timeout=self._reply_ack_timeout)
+            if result.get('errcode', 0) != 0:
+                await self.logger.warning(
+                    f'Reply ACK error: errcode={result.get("errcode")}, errmsg={result.get("errmsg")}'
+                )
+            return result
+        except asyncio.TimeoutError:
+            self._pending_acks.pop(req_id, None)
+            await self.logger.warning(f'Reply ACK timeout ({self._reply_ack_timeout}s) for req_id={req_id}')
+            return None
+
+    async def _send_frame(self, frame: dict):
+        """Send a JSON frame over the WebSocket connection."""
+        if self._ws and not self._ws.closed:
+            await self._ws.send_str(json.dumps(frame, ensure_ascii=False))
+
+    def _clear_pending_acks(self, reason: str):
+        """Reject all pending ACK futures on disconnection."""
+        for req_id, future in self._pending_acks.items():
+            if not future.done():
+                future.set_exception(ConnectionError(reason))
+        self._pending_acks.clear()
--- a/src/langbot/libs/wecom_api/api.py
+++ b/src/langbot/libs/wecom_api/api.py
@@ -4,6 +4,7 @@ import base64
 import binascii
 import httpx
 import traceback
+from urllib.parse import quote
 from quart import Quart
 import xml.etree.ElementTree as ET
 from typing import Callable, Dict, Any
@@ -67,6 +68,31 @@ class WecomClient:
                await self.logger.error(f'获取accesstoken失败:{response.json()}')
                raise Exception(f'未获取access token: {data}')

+    async def get_user_info(self, userid: str) -> dict:
+        """
+        Get user information by user ID using the application secret.
+
+        Args:
+            userid: The user ID to look up.
+
+        Returns:
+            dict: User information including 'name' field.
+        """
+        if not await self.check_access_token():
+            self.access_token = await self.get_access_token(self.secret)
+
+        url = self.base_url + '/user/get?access_token=' + self.access_token + '&userid=' + quote(userid)
+        async with httpx.AsyncClient() as client:
+            response = await client.get(url)
+            data = response.json()
+            if data.get('errcode') == 40014 or data.get('errcode') == 42001:
+                self.access_token = await self.get_access_token(self.secret)
+                return await self.get_user_info(userid)
+            if data.get('errcode', 0) != 0:
+                await self.logger.error(f'获取用户信息失败:{data}')
+                return {}
+            return data
+
    async def get_users(self):
        if not self.check_access_token_for_contacts():
            self.access_token_for_contacts = await self.get_access_token(self.secret_for_contacts)
--- a/src/langbot/libs/wecom_customer_service_api/api.py
+++ b/src/langbot/libs/wecom_customer_service_api/api.py
@@ -10,6 +10,7 @@ from typing import Callable
 from .wecomcsevent import WecomCSEvent
 import langbot_plugin.api.entities.builtin.platform.message as platform_message
 import aiofiles
+import time


 class WecomCSClient:
@@ -34,6 +35,10 @@ class WecomCSClient:
        self.unified_mode = unified_mode
        self.app = Quart(__name__)

+        # Customer info cache: {external_userid: (info_dict, timestamp)}
+        self._customer_cache: dict[str, tuple[dict, float]] = {}
+        self._cache_ttl = 60  # Cache TTL in seconds (1 minute)
+
        # 只有在非统一模式下才注册独立路由
        if not self.unified_mode:
            self.app.add_url_rule(
@@ -378,3 +383,53 @@ class WecomCSClient:
    async def get_media_id(self, image: platform_message.Image):
        media_id = await self.upload_to_work(image=image)
        return media_id
+
+    async def get_customer_info(self, external_userid: str) -> dict | None:
+        """
+        Get customer information by external_userid with caching.
+
+        Uses a 1-minute cache to avoid repeated API calls for the same user.
+
+        Args:
+            external_userid: The external user ID of the customer.
+
+        Returns:
+            Customer info dict with 'nickname', 'avatar', etc., or None if not found.
+        """
+        # Check cache first
+        current_time = time.time()
+        if external_userid in self._customer_cache:
+            cached_info, cached_time = self._customer_cache[external_userid]
+            if current_time - cached_time < self._cache_ttl:
+                return cached_info
+
+        # Cache miss or expired, fetch from API
+        if not await self.check_access_token():
+            self.access_token = await self.get_access_token(self.secret)
+
+        url = f'{self.base_url}/kf/customer/batchget?access_token={self.access_token}'
+
+        payload = {
+            'external_userid_list': [external_userid],
+        }
+
+        async with httpx.AsyncClient() as client:
+            response = await client.post(url, json=payload)
+            data = response.json()
+
+            if data.get('errcode') in [40014, 42001]:
+                self.access_token = await self.get_access_token(self.secret)
+                return await self.get_customer_info(external_userid)
+
+            if data.get('errcode', 0) != 0:
+                if self.logger:
+                    await self.logger.warning(f'Failed to get customer info: {data}')
+                return None
+
+            customer_list = data.get('customer_list', [])
+            if customer_list:
+                customer_info = customer_list[0]
+                # Store in cache
+                self._customer_cache[external_userid] = (customer_info, current_time)
+                return customer_info
+            return None
--- a/src/langbot/pkg/api/http/controller/groups/files.py
+++ b/src/langbot/pkg/api/http/controller/groups/files.py
@@ -13,9 +13,9 @@ from .. import group
@group.group_class('files', '/api/v1/files')
 class FilesRouterGroup(group.RouterGroup):
    async def initialize(self) -> None:
-        @self.route('/image/<image_key>', methods=['GET'], auth_type=group.AuthType.NONE)
+        @self.route('/image/<path:image_key>', methods=['GET'], auth_type=group.AuthType.NONE)
        async def _(image_key: str) -> quart.Response:
-            if '/' in image_key or '\\' in image_key:
+            if '..' in image_key or '\\' in image_key:
                return quart.Response(status=404)

            if not await self.ap.storage_mgr.storage_provider.exists(image_key):
--- a/src/langbot/pkg/api/http/controller/groups/knowledge/base.py
+++ b/src/langbot/pkg/api/http/controller/groups/knowledge/base.py
@@ -13,7 +13,10 @@ class KnowledgeBaseRouterGroup(group.RouterGroup):

            elif quart.request.method == 'POST':
                json_data = await quart.request.json
-                knowledge_base_uuid = await self.ap.knowledge_service.create_knowledge_base(json_data)
+                try:
+                    knowledge_base_uuid = await self.ap.knowledge_service.create_knowledge_base(json_data)
+                except ValueError as e:
+                    return self.http_status(400, -1, str(e))
                return self.success(data={'uuid': knowledge_base_uuid})

            return self.http_status(405, -1, 'Method not allowed')
@@ -39,7 +42,7 @@ class KnowledgeBaseRouterGroup(group.RouterGroup):
            elif quart.request.method == 'PUT':
                json_data = await quart.request.json
                await self.ap.knowledge_service.update_knowledge_base(knowledge_base_uuid, json_data)
-                return self.success({})
+                return self.success(data={'uuid': knowledge_base_uuid})

            elif quart.request.method == 'DELETE':
                await self.ap.knowledge_service.delete_knowledge_base(knowledge_base_uuid)
@@ -65,8 +68,12 @@ class KnowledgeBaseRouterGroup(group.RouterGroup):
                if not file_id:
                    return self.http_status(400, -1, 'File ID is required')

+                parser_plugin_id = json_data.get('parser_plugin_id')
+
                # 调用服务层方法将文件与知识库关联
-                task_id = await self.ap.knowledge_service.store_file(knowledge_base_uuid, file_id)
+                task_id = await self.ap.knowledge_service.store_file(
+                    knowledge_base_uuid, file_id, parser_plugin_id=parser_plugin_id
+                )
                return self.success(
                    {
                        'task_id': task_id,
@@ -90,5 +97,13 @@ class KnowledgeBaseRouterGroup(group.RouterGroup):
        async def retrieve_knowledge_base(knowledge_base_uuid: str) -> str:
            json_data = await quart.request.json
            query = json_data.get('query')
-            results = await self.ap.knowledge_service.retrieve_knowledge_base(knowledge_base_uuid, query)
+
+            if not query or not query.strip():
+                return self.http_status(400, -1, 'Query is required and cannot be empty')
+
+            # Extract retrieval_settings to allow dynamic control over Knowledge Engine behavior (e.g. top_k, filters)
+            retrieval_settings = json_data.get('retrieval_settings', {})
+            results = await self.ap.knowledge_service.retrieve_knowledge_base(
+                knowledge_base_uuid, query, retrieval_settings
+            )
            return self.success(data={'results': results})
--- a/src/langbot/pkg/api/http/controller/groups/knowledge/engines.py
+++ b/src/langbot/pkg/api/http/controller/groups/knowledge/engines.py
@@ -0,0 +1,45 @@
+import quart
+from urllib.parse import unquote
+from ... import group
+
+
+@group.group_class('knowledge_engines', '/api/v1/knowledge/engines')
+class KnowledgeEnginesRouterGroup(group.RouterGroup):
+    async def initialize(self) -> None:
+        @self.route('', methods=['GET'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY)
+        async def list_knowledge_engines() -> quart.Response:
+            """List all available Knowledge Engines from plugins.
+
+            Returns a list of Knowledge Engines with their capabilities and configuration schemas.
+            This is used by the frontend to render the knowledge base creation wizard.
+            """
+            engines = await self.ap.knowledge_service.list_knowledge_engines()
+            return self.success(data={'engines': engines})
+
+        @self.route(
+            '/<path:plugin_id>/creation-schema', methods=['GET'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY
+        )
+        async def get_engine_creation_schema(plugin_id: str) -> quart.Response:
+            """Get creation settings schema for a specific Knowledge Engine.
+
+            plugin_id is in 'author/name' format, captured via <path:> converter.
+            """
+            plugin_id = unquote(plugin_id)
+            if '/' not in plugin_id:
+                return self.http_status(400, -1, 'Invalid plugin_id format. Expected author/name.')
+            schema = await self.ap.knowledge_service.get_engine_creation_schema(plugin_id)
+            return self.success(data={'schema': schema})
+
+        @self.route(
+            '/<path:plugin_id>/retrieval-schema', methods=['GET'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY
+        )
+        async def get_engine_retrieval_schema(plugin_id: str) -> quart.Response:
+            """Get retrieval settings schema for a specific Knowledge Engine.
+
+            plugin_id is in 'author/name' format, captured via <path:> converter.
+            """
+            plugin_id = unquote(plugin_id)
+            if '/' not in plugin_id:
+                return self.http_status(400, -1, 'Invalid plugin_id format. Expected author/name.')
+            schema = await self.ap.knowledge_service.get_engine_retrieval_schema(plugin_id)
+            return self.success(data={'schema': schema})
--- a/src/langbot/pkg/api/http/controller/groups/knowledge/external.py
+++ b/src/langbot/pkg/api/http/controller/groups/knowledge/external.py
@@ -1,61 +0,0 @@
-import quart
-from ... import group
-
-
-@group.group_class('external_knowledge_base', '/api/v1/knowledge/external-bases')
-class ExternalKnowledgeBaseRouterGroup(group.RouterGroup):
-    async def initialize(self) -> None:
-        @self.route('/retrievers', methods=['GET'])
-        async def list_knowledge_retrievers() -> quart.Response:
-            """List all available knowledge retrievers from plugins."""
-            retrievers = await self.ap.plugin_connector.list_knowledge_retrievers()
-            return self.success(data={'retrievers': retrievers})
-
-        @self.route('', methods=['POST', 'GET'])
-        async def handle_external_knowledge_bases() -> quart.Response:
-            if quart.request.method == 'GET':
-                external_kbs = await self.ap.external_kb_service.get_external_knowledge_bases()
-                return self.success(data={'bases': external_kbs})
-
-            elif quart.request.method == 'POST':
-                json_data = await quart.request.json
-                kb_uuid = await self.ap.external_kb_service.create_external_knowledge_base(json_data)
-                return self.success(data={'uuid': kb_uuid})
-
-            return self.http_status(405, -1, 'Method not allowed')
-
-        @self.route(
-            '/<kb_uuid>',
-            methods=['GET', 'DELETE', 'PUT'],
-        )
-        async def handle_specific_external_knowledge_base(kb_uuid: str) -> quart.Response:
-            if quart.request.method == 'GET':
-                external_kb = await self.ap.external_kb_service.get_external_knowledge_base(kb_uuid)
-
-                if external_kb is None:
-                    return self.http_status(404, -1, 'external knowledge base not found')
-
-                return self.success(
-                    data={
-                        'base': external_kb,
-                    }
-                )
-
-            elif quart.request.method == 'PUT':
-                json_data = await quart.request.json
-                await self.ap.external_kb_service.update_external_knowledge_base(kb_uuid, json_data)
-                return self.success({})
-
-            elif quart.request.method == 'DELETE':
-                await self.ap.external_kb_service.delete_external_knowledge_base(kb_uuid)
-                return self.success({})
-
-        @self.route(
-            '/<kb_uuid>/retrieve',
-            methods=['POST'],
-        )
-        async def retrieve_external_knowledge_base(kb_uuid: str) -> str:
-            json_data = await quart.request.json
-            query = json_data.get('query')
-            results = await self.ap.external_kb_service.retrieve_external_knowledge_base(kb_uuid, query)
-            return self.success(data={'results': results})
--- a/src/langbot/pkg/api/http/controller/groups/knowledge/migration.py
+++ b/src/langbot/pkg/api/http/controller/groups/knowledge/migration.py
@@ -0,0 +1,372 @@
+import asyncio
+import json
+
+import httpx
+import quart
+import sqlalchemy
+
+from ... import group
+from ......core import taskmgr
+from ......entity.persistence import metadata as persistence_metadata
+from langbot_plugin.runtime.plugin.mgr import PluginInstallSource
+
+LANGRAG_PLUGIN_AUTHOR = 'langbot-team'
+LANGRAG_PLUGIN_NAME = 'LangRAG'
+LANGRAG_PLUGIN_ID = f'{LANGRAG_PLUGIN_AUTHOR}/{LANGRAG_PLUGIN_NAME}'
+DEFAULT_SPACE_URL = 'https://space.langbot.app'
+
+# Old Retriever plugin_name -> New Connector plugin_name
+EXTERNAL_PLUGIN_NAME_MAPPING = {
+    'DifyDatasetsRetriever': 'DifyDatasetsConnector',
+    'RAGFlowRetriever': 'RAGFlowConnector',
+    'FastGPTRetriever': 'FastGPTConnector',
+}
+
+# Per-plugin: which old retriever_config fields belong to creation_settings.
+# Remaining fields go to retrieval_settings.
+# None means ALL fields go to creation_settings (no retrieval_schema).
+EXTERNAL_PLUGIN_CREATION_FIELDS: dict[str, set[str] | None] = {
+    'langbot-team/DifyDatasetsConnector': {'api_base_url', 'dify_apikey', 'dataset_id'},
+    'langbot-team/RAGFlowConnector': {'api_base_url', 'api_key', 'dataset_ids'},
+    'langbot-team/FastGPTConnector': None,  # all fields -> creation_settings
+}
+
+
+@group.group_class('knowledge/migration', '/api/v1/knowledge/migration')
+class KnowledgeMigrationRouterGroup(group.RouterGroup):
+    async def _get_migration_flag(self) -> bool:
+        """Check if rag_plugin_migration_needed flag is set."""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.select(persistence_metadata.Metadata).where(
+                persistence_metadata.Metadata.key == 'rag_plugin_migration_needed'
+            )
+        )
+        row = result.first()
+        return row is not None and row.value == 'true'
+
+    async def _set_migration_flag(self, value: str):
+        """Set rag_plugin_migration_needed flag."""
+        await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.update(persistence_metadata.Metadata)
+            .where(persistence_metadata.Metadata.key == 'rag_plugin_migration_needed')
+            .values(value=value)
+        )
+
+    async def _table_exists(self, table_name: str) -> bool:
+        """Check if a table exists."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT EXISTS (SELECT FROM information_schema.tables WHERE table_name = :table_name);'
+                ).bindparams(table_name=table_name)
+            )
+            return result.scalar()
+        else:
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text("SELECT name FROM sqlite_master WHERE type='table' AND name=:table_name;").bindparams(
+                    table_name=table_name
+                )
+            )
+            return result.first() is not None
+
+    async def _install_plugin_from_marketplace(
+        self, plugin_id: str, task_context: taskmgr.TaskContext, space_url: str
+    ) -> None:
+        """Install a single plugin from the marketplace."""
+        p_author, p_name = plugin_id.split('/', 1)
+        self.ap.logger.info(f'RAG migration: installing plugin {plugin_id} from marketplace...')
+        task_context.trace(f'Installing plugin {plugin_id} from marketplace...')
+
+        async with httpx.AsyncClient(trust_env=True, timeout=15) as client:
+            resp = await client.get(f'{space_url}/api/v1/marketplace/plugins/{p_author}/{p_name}')
+            resp.raise_for_status()
+            p_data = resp.json().get('data', {}).get('plugin', {})
+            p_version = p_data.get('latest_version')
+            if not p_version:
+                raise Exception(f'Could not determine latest version for {plugin_id}')
+
+        await self.ap.plugin_connector.install_plugin(
+            PluginInstallSource.MARKETPLACE,
+            {
+                'plugin_author': p_author,
+                'plugin_name': p_name,
+                'plugin_version': p_version,
+            },
+            task_context=task_context,
+        )
+        self.ap.logger.info(f'RAG migration: plugin {plugin_id} install request sent.')
+
+    async def _execute_rag_migration(self, task_context: taskmgr.TaskContext, install_plugin: bool = True):
+        """Execute RAG migration: install required plugins and restore backup data."""
+        warnings = []
+
+        # Collect all plugins we need: LangRAG (always) + connector plugins (from external KBs)
+        needed_plugins: dict[str, str] = {
+            LANGRAG_PLUGIN_ID: LANGRAG_PLUGIN_NAME,
+        }
+
+        has_external = await self._table_exists('external_knowledge_bases')
+        if has_external:
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text('SELECT DISTINCT plugin_author, plugin_name FROM external_knowledge_bases;')
+            )
+            for row in result.fetchall():
+                plugin_author = row[0] or ''
+                plugin_name = row[1] or ''
+                mapped_name = EXTERNAL_PLUGIN_NAME_MAPPING.get(plugin_name, plugin_name)
+                plugin_id = f'{plugin_author}/{mapped_name}'
+                if plugin_id not in needed_plugins:
+                    needed_plugins[plugin_id] = mapped_name
+
+        self.ap.logger.info(f'RAG migration: plugins needed: {list(needed_plugins.keys())}')
+
+        if install_plugin:
+            # Step 1: Install all required plugins from marketplace
+            task_context.trace('Installing required plugins...', action='install-plugin')
+            space_url = self.ap.instance_config.data.get('space', {}).get('url', DEFAULT_SPACE_URL).rstrip('/')
+
+            for plugin_id in needed_plugins:
+                try:
+                    await self._install_plugin_from_marketplace(plugin_id, task_context, space_url)
+                except Exception as e:
+                    self.ap.logger.warning(f'RAG migration: plugin {plugin_id} install returned: {e}')
+                    task_context.trace(f'Plugin install note ({plugin_id}): {e}')
+
+            # Step 2: Wait for all plugins to become available as knowledge engines
+            task_context.trace(
+                f'Waiting for plugins to become available: {list(needed_plugins.keys())}...',
+                action='wait-plugin',
+            )
+            max_retries = 30
+            engine_id_set: set[str] = set()
+            for i in range(max_retries):
+                try:
+                    engines = await self.ap.plugin_connector.list_knowledge_engines()
+                    engine_id_set = {e.get('plugin_id') for e in engines}
+                except Exception:
+                    pass
+                if all(pid in engine_id_set for pid in needed_plugins):
+                    self.ap.logger.info(f'RAG migration: all plugins ready: {engine_id_set}')
+                    task_context.trace('All required plugins are ready.')
+                    break
+                if i == max_retries - 1:
+                    still_missing = [pid for pid in needed_plugins if pid not in engine_id_set]
+                    warning = f'Plugin(s) {still_missing} did not become available after {max_retries} retries'
+                    self.ap.logger.warning(f'RAG migration: {warning}')
+                    warnings.append(warning)
+                    task_context.trace(warning)
+                await asyncio.sleep(2)
+        else:
+            try:
+                engines = await self.ap.plugin_connector.list_knowledge_engines()
+                engine_id_set = {e.get('plugin_id') for e in engines}
+            except Exception:
+                engine_id_set = set()
+
+        # Step 3: Restore internal knowledge bases from backup
+        task_context.trace('Restoring internal knowledge bases...', action='restore-internal')
+        if await self._table_exists('knowledge_bases_backup'):
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text('SELECT * FROM knowledge_bases_backup;')
+            )
+            rows = result.fetchall()
+            columns = result.keys()
+
+            for row in rows:
+                row_dict = dict(zip(columns, row))
+                kb_uuid = row_dict.get('uuid')
+                name = row_dict.get('name', 'Untitled')
+                description = row_dict.get('description', '')
+                emoji = row_dict.get('emoji', '\U0001f4da')
+                embedding_model_uuid = row_dict.get('embedding_model_uuid', '')
+                top_k = row_dict.get('top_k', 5)
+                created_at = row_dict.get('created_at')
+                updated_at = row_dict.get('updated_at')
+
+                creation_settings = json.dumps({'embedding_model_uuid': embedding_model_uuid})
+                retrieval_settings = json.dumps({'top_k': top_k})
+
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'INSERT INTO knowledge_bases '
+                        '(uuid, name, description, emoji, created_at, updated_at, '
+                        'knowledge_engine_plugin_id, collection_id, creation_settings, retrieval_settings) '
+                        'VALUES (:uuid, :name, :description, :emoji, :created_at, :updated_at, '
+                        ':plugin_id, :collection_id, :creation_settings, :retrieval_settings);'
+                    ).bindparams(
+                        uuid=kb_uuid,
+                        name=name,
+                        description=description,
+                        emoji=emoji,
+                        created_at=created_at,
+                        updated_at=updated_at,
+                        plugin_id=LANGRAG_PLUGIN_ID,
+                        collection_id=kb_uuid,
+                        creation_settings=creation_settings,
+                        retrieval_settings=retrieval_settings,
+                    )
+                )
+
+                try:
+                    config = {'embedding_model_uuid': embedding_model_uuid}
+                    await self.ap.plugin_connector.rag_on_kb_create(LANGRAG_PLUGIN_ID, kb_uuid, config)
+                    task_context.trace(f'Restored internal KB: {name} ({kb_uuid})')
+                except Exception as e:
+                    warning = f'Failed to notify plugin for KB {name} ({kb_uuid}): {e}'
+                    warnings.append(warning)
+                    task_context.trace(warning)
+
+            await self.ap.rag_mgr.load_knowledge_bases_from_db()
+
+        # Step 4: Restore external knowledge bases
+        task_context.trace('Restoring external knowledge bases...', action='restore-external')
+        if has_external:
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text('SELECT * FROM external_knowledge_bases;')
+            )
+            rows = result.fetchall()
+            columns = result.keys()
+
+            self.ap.logger.info(
+                f'RAG migration: {len(rows)} external KB(s) to restore. Available engines: {engine_id_set}'
+            )
+            task_context.trace(f'Found {len(rows)} external KB(s). Available engines: {engine_id_set}')
+
+            for row in rows:
+                row_dict = dict(zip(columns, row))
+                kb_uuid = row_dict.get('uuid')
+                name = row_dict.get('name', 'Untitled')
+                description = row_dict.get('description', '')
+                emoji = row_dict.get('emoji', '\U0001f517')
+                plugin_author = row_dict.get('plugin_author', '')
+                plugin_name = row_dict.get('plugin_name', '')
+                retriever_config = row_dict.get('retriever_config', {})
+                created_at = row_dict.get('created_at')
+
+                mapped_plugin_name = EXTERNAL_PLUGIN_NAME_MAPPING.get(plugin_name, plugin_name)
+                external_plugin_id = f'{plugin_author}/{mapped_plugin_name}'
+
+                self.ap.logger.info(
+                    f'RAG migration: processing external KB "{name}" ({kb_uuid}), '
+                    f'plugin: {plugin_author}/{plugin_name} -> {external_plugin_id}'
+                )
+
+                if isinstance(retriever_config, str):
+                    try:
+                        retriever_config = json.loads(retriever_config)
+                    except (json.JSONDecodeError, TypeError):
+                        retriever_config = {}
+
+                creation_fields = EXTERNAL_PLUGIN_CREATION_FIELDS.get(external_plugin_id)
+                if creation_fields is None:
+                    creation_settings_dict = retriever_config
+                    retrieval_settings_dict = {}
+                else:
+                    creation_settings_dict = {k: v for k, v in retriever_config.items() if k in creation_fields}
+                    retrieval_settings_dict = {k: v for k, v in retriever_config.items() if k not in creation_fields}
+
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'INSERT INTO knowledge_bases '
+                        '(uuid, name, description, emoji, created_at, updated_at, '
+                        'knowledge_engine_plugin_id, collection_id, creation_settings, retrieval_settings) '
+                        'VALUES (:uuid, :name, :description, :emoji, :created_at, :updated_at, '
+                        ':plugin_id, :collection_id, :creation_settings, :retrieval_settings);'
+                    ).bindparams(
+                        uuid=kb_uuid,
+                        name=name,
+                        description=description,
+                        emoji=emoji,
+                        created_at=created_at,
+                        updated_at=created_at,
+                        plugin_id=external_plugin_id,
+                        collection_id=kb_uuid,
+                        creation_settings=json.dumps(creation_settings_dict),
+                        retrieval_settings=json.dumps(retrieval_settings_dict),
+                    )
+                )
+
+                if external_plugin_id not in engine_id_set:
+                    warning = (
+                        f'External KB "{name}" ({kb_uuid}) record saved, but plugin {external_plugin_id} '
+                        f'is not installed yet. Install the connector plugin to use it.'
+                    )
+                    warnings.append(warning)
+                    task_context.trace(warning)
+                else:
+                    try:
+                        await self.ap.plugin_connector.rag_on_kb_create(
+                            external_plugin_id, kb_uuid, creation_settings_dict
+                        )
+                        task_context.trace(f'Restored external KB: {name} ({kb_uuid})')
+                    except Exception as e:
+                        warning = f'Failed to notify plugin for external KB {name} ({kb_uuid}): {e}'
+                        warnings.append(warning)
+                        task_context.trace(warning)
+
+            await self.ap.rag_mgr.load_knowledge_bases_from_db()
+
+        # Step 5: Clear migration flag
+        await self._set_migration_flag('false')
+        task_context.trace('RAG migration completed.', action='done')
+
+        if warnings:
+            task_context.trace(f'Completed with {len(warnings)} warning(s).')
+
+    async def initialize(self) -> None:
+        @self.route('/status', methods=['GET'], auth_type=group.AuthType.USER_TOKEN)
+        async def _() -> str:
+            needed = await self._get_migration_flag()
+
+            internal_kb_count = 0
+            external_kb_count = 0
+
+            if needed:
+                if await self._table_exists('knowledge_bases_backup'):
+                    result = await self.ap.persistence_mgr.execute_async(
+                        sqlalchemy.text('SELECT COUNT(*) FROM knowledge_bases_backup;')
+                    )
+                    internal_kb_count = result.scalar() or 0
+
+                if await self._table_exists('external_knowledge_bases'):
+                    result = await self.ap.persistence_mgr.execute_async(
+                        sqlalchemy.text('SELECT COUNT(*) FROM external_knowledge_bases;')
+                    )
+                    external_kb_count = result.scalar() or 0
+
+            return self.success(
+                data={
+                    'needed': needed,
+                    'internal_kb_count': internal_kb_count,
+                    'external_kb_count': external_kb_count,
+                }
+            )
+
+        @self.route('/execute', methods=['POST'], auth_type=group.AuthType.USER_TOKEN)
+        async def _() -> str:
+            needed = await self._get_migration_flag()
+            if not needed:
+                return self.http_status(400, -1, 'RAG migration is not needed')
+
+            data = await quart.request.get_json(silent=True) or {}
+            install_plugin = data.get('install_plugin', True)
+
+            ctx = taskmgr.TaskContext.new()
+            wrapper = self.ap.task_mgr.create_user_task(
+                self._execute_rag_migration(task_context=ctx, install_plugin=install_plugin),
+                kind='rag-migration',
+                name='rag-migration-execute',
+                label='Migrating knowledge bases to plugin architecture',
+                context=ctx,
+            )
+
+            return self.success(data={'task_id': wrapper.id})
+
+        @self.route('/dismiss', methods=['POST'], auth_type=group.AuthType.USER_TOKEN)
+        async def _() -> str:
+            needed = await self._get_migration_flag()
+            if not needed:
+                return self.http_status(400, -1, 'RAG migration is not needed')
+
+            await self._set_migration_flag('false')
+            return self.success()
--- a/src/langbot/pkg/api/http/controller/groups/knowledge/parsers.py
+++ b/src/langbot/pkg/api/http/controller/groups/knowledge/parsers.py
@@ -0,0 +1,16 @@
+import quart
+from ... import group
+
+
+@group.group_class('parsers', '/api/v1/knowledge/parsers')
+class ParsersRouterGroup(group.RouterGroup):
+    async def initialize(self) -> None:
+        @self.route('', methods=['GET'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY)
+        async def list_parsers() -> quart.Response:
+            """List all available parsers from plugins.
+
+            Optional query parameter `mime_type` to filter parsers by supported MIME type.
+            """
+            mime_type = quart.request.args.get('mime_type')
+            parsers = await self.ap.knowledge_service.list_parsers(mime_type)
+            return self.success(data={'parsers': parsers})
--- a/src/langbot/pkg/api/http/controller/groups/monitoring.py
+++ b/src/langbot/pkg/api/http/controller/groups/monitoring.py
@@ -52,6 +52,7 @@ class MonitoringRouterGroup(group.RouterGroup):
            # Parse query parameters
            bot_ids = quart.request.args.getlist('botId')
            pipeline_ids = quart.request.args.getlist('pipelineId')
+            session_ids = quart.request.args.getlist('sessionId')
            start_time_str = quart.request.args.get('startTime')
            end_time_str = quart.request.args.get('endTime')
            limit = int(quart.request.args.get('limit', 100))
@@ -64,6 +65,7 @@ class MonitoringRouterGroup(group.RouterGroup):
            messages, total = await self.ap.monitoring_service.get_messages(
                bot_ids=bot_ids if bot_ids else None,
                pipeline_ids=pipeline_ids if pipeline_ids else None,
+                session_ids=session_ids if session_ids else None,
                start_time=start_time,
                end_time=end_time,
                limit=limit,
@@ -323,3 +325,164 @@ class MonitoringRouterGroup(group.RouterGroup):
                return self.error(message=f'Message {message_id} not found', code=404)

            return self.success(data=details)
+
+        @self.route('/export', methods=['GET'], auth_type=group.AuthType.USER_TOKEN)
+        async def export_data() -> tuple[str, int]:
+            """Export monitoring data as CSV"""
+            # Parse query parameters
+            export_type = quart.request.args.get('type', 'messages')
+            bot_ids = quart.request.args.getlist('botId')
+            pipeline_ids = quart.request.args.getlist('pipelineId')
+            start_time_str = quart.request.args.get('startTime')
+            end_time_str = quart.request.args.get('endTime')
+            limit = int(quart.request.args.get('limit', 100000))
+
+            # Parse datetime
+            start_time = parse_iso_datetime(start_time_str)
+            end_time = parse_iso_datetime(end_time_str)
+
+            # Get data based on export type
+            if export_type == 'messages':
+                data = await self.ap.monitoring_service.export_messages(
+                    bot_ids=bot_ids if bot_ids else None,
+                    pipeline_ids=pipeline_ids if pipeline_ids else None,
+                    start_time=start_time,
+                    end_time=end_time,
+                    limit=limit,
+                )
+                headers = [
+                    'id',
+                    'timestamp',
+                    'bot_id',
+                    'bot_name',
+                    'pipeline_id',
+                    'pipeline_name',
+                    'runner_name',
+                    'message_content',
+                    'message_text',
+                    'session_id',
+                    'status',
+                    'level',
+                    'platform',
+                    'user_id',
+                ]
+            elif export_type == 'llm-calls':
+                data = await self.ap.monitoring_service.export_llm_calls(
+                    bot_ids=bot_ids if bot_ids else None,
+                    pipeline_ids=pipeline_ids if pipeline_ids else None,
+                    start_time=start_time,
+                    end_time=end_time,
+                    limit=limit,
+                )
+                headers = [
+                    'id',
+                    'timestamp',
+                    'model_name',
+                    'input_tokens',
+                    'output_tokens',
+                    'total_tokens',
+                    'duration_ms',
+                    'cost',
+                    'status',
+                    'bot_id',
+                    'bot_name',
+                    'pipeline_id',
+                    'pipeline_name',
+                    'session_id',
+                    'message_id',
+                    'error_message',
+                ]
+            elif export_type == 'embedding-calls':
+                data = await self.ap.monitoring_service.export_embedding_calls(
+                    start_time=start_time,
+                    end_time=end_time,
+                    limit=limit,
+                )
+                headers = [
+                    'id',
+                    'timestamp',
+                    'model_name',
+                    'prompt_tokens',
+                    'total_tokens',
+                    'duration_ms',
+                    'input_count',
+                    'status',
+                    'error_message',
+                    'knowledge_base_id',
+                    'query_text',
+                    'session_id',
+                    'message_id',
+                    'call_type',
+                ]
+            elif export_type == 'errors':
+                data = await self.ap.monitoring_service.export_errors(
+                    bot_ids=bot_ids if bot_ids else None,
+                    pipeline_ids=pipeline_ids if pipeline_ids else None,
+                    start_time=start_time,
+                    end_time=end_time,
+                    limit=limit,
+                )
+                headers = [
+                    'id',
+                    'timestamp',
+                    'error_type',
+                    'error_message',
+                    'bot_id',
+                    'bot_name',
+                    'pipeline_id',
+                    'pipeline_name',
+                    'session_id',
+                    'message_id',
+                    'stack_trace',
+                ]
+            elif export_type == 'sessions':
+                data = await self.ap.monitoring_service.export_sessions(
+                    bot_ids=bot_ids if bot_ids else None,
+                    pipeline_ids=pipeline_ids if pipeline_ids else None,
+                    start_time=start_time,
+                    end_time=end_time,
+                    limit=limit,
+                )
+                headers = [
+                    'session_id',
+                    'bot_id',
+                    'bot_name',
+                    'pipeline_id',
+                    'pipeline_name',
+                    'message_count',
+                    'start_time',
+                    'last_activity',
+                    'is_active',
+                    'platform',
+                    'user_id',
+                ]
+            else:
+                return self.error(message=f'Invalid export type: {export_type}', code=400)
+
+            # Generate CSV content with UTF-8 BOM for Excel compatibility
+            import io
+
+            output = io.StringIO()
+            # Write UTF-8 BOM for Excel
+            output.write('\ufeff')
+            # Write header
+            output.write(','.join(headers) + '\n')
+
+            # Escape and write each row
+            for row in data:
+                escaped_values = []
+                for header in headers:
+                    value = row.get(header, '')
+                    escaped_values.append(self.ap.monitoring_service._escape_csv_field(value))
+                output.write(','.join(escaped_values) + '\n')
+
+            csv_content = output.getvalue()
+
+            # Return as file download
+            response = await quart.make_response(csv_content)
+            response.headers['Content-Type'] = 'text/csv; charset=utf-8'
+            response.headers['Content-Disposition'] = (
+                f'attachment; filename="monitoring-{export_type}-{int(datetime.datetime.now().timestamp())}.csv"'
+            )
+
+            return response, 200
--- a/src/langbot/pkg/api/http/controller/groups/pipelines/pipelines.py
+++ b/src/langbot/pkg/api/http/controller/groups/pipelines/pipelines.py
@@ -68,7 +68,7 @@ class PipelinesRouterGroup(group.RouterGroup):
                    return self.http_status(404, -1, 'pipeline not found')

                # Only include plugins with pipeline-related components (Command, EventListener, Tool)
-                # Plugins that only have KnowledgeRetriever components are not suitable for pipeline extensions
+                # Plugins that only have KnowledgeEngine components are not suitable for pipeline extensions
                pipeline_component_kinds = ['Command', 'EventListener', 'Tool']
                plugins = await self.ap.plugin_connector.list_plugins(component_kinds=pipeline_component_kinds)
                mcp_servers = await self.ap.mcp_service.get_mcp_servers(contain_runtime_info=True)
--- a/src/langbot/pkg/api/http/controller/groups/plugins.py
+++ b/src/langbot/pkg/api/http/controller/groups/plugins.py
@@ -14,6 +14,18 @@ from langbot_plugin.runtime.plugin.mgr import PluginInstallSource

@group.group_class('plugins', '/api/v1/plugins')
 class PluginsRouterGroup(group.RouterGroup):
+    async def _check_extensions_limit(self) -> str | None:
+        """Check if extensions limit is reached. Returns error response if limit exceeded, None otherwise."""
+        limitation = self.ap.instance_config.data.get('system', {}).get('limitation', {})
+        max_extensions = limitation.get('max_extensions', -1)
+        if max_extensions >= 0:
+            plugins = await self.ap.plugin_connector.list_plugins()
+            mcp_servers = await self.ap.mcp_service.get_mcp_servers()
+            total_extensions = len(plugins) + len(mcp_servers)
+            if total_extensions >= max_extensions:
+                return self.http_status(400, -1, f'Maximum number of extensions ({max_extensions}) reached')
+        return None
+
    async def initialize(self) -> None:
        @self.route('', methods=['GET'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY)
        async def _() -> str:
@@ -239,6 +251,10 @@ class PluginsRouterGroup(group.RouterGroup):
        @self.route('/install/github', methods=['POST'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY)
        async def _() -> str:
            """Install plugin from GitHub release asset"""
+            limit_error = await self._check_extensions_limit()
+            if limit_error is not None:
+                return limit_error
+
            data = await quart.request.json
            asset_url = data.get('asset_url', '')
            owner = data.get('owner', '')
@@ -273,6 +289,10 @@ class PluginsRouterGroup(group.RouterGroup):
            auth_type=group.AuthType.USER_TOKEN_OR_API_KEY,
        )
        async def _() -> str:
+            limit_error = await self._check_extensions_limit()
+            if limit_error is not None:
+                return limit_error
+
            data = await quart.request.json

            ctx = taskmgr.TaskContext.new()
@@ -288,6 +308,10 @@ class PluginsRouterGroup(group.RouterGroup):

        @self.route('/install/local', methods=['POST'], auth_type=group.AuthType.USER_TOKEN_OR_API_KEY)
        async def _() -> str:
+            limit_error = await self._check_extensions_limit()
+            if limit_error is not None:
+                return limit_error
+
            file = (await quart.request.files).get('file')
            if file is None:
                return self.http_status(400, -1, 'file is required')
--- a/src/langbot/pkg/api/http/controller/groups/survey.py
+++ b/src/langbot/pkg/api/http/controller/groups/survey.py
@@ -0,0 +1,47 @@
+import quart
+
+from .. import group
+
+
+@group.group_class('survey', '/api/v1/survey')
+class SurveyRouterGroup(group.RouterGroup):
+    async def initialize(self) -> None:
+        @self.route('/pending', methods=['GET'], auth_type=group.AuthType.USER_TOKEN)
+        async def _get_pending() -> str:
+            """Get pending survey for the frontend to display."""
+            survey = self.ap.survey.get_pending_survey() if self.ap.survey else None
+            return self.success(data={'survey': survey})
+
+        @self.route('/respond', methods=['POST'], auth_type=group.AuthType.USER_TOKEN)
+        async def _respond() -> str:
+            """Submit survey response."""
+            json_data = await quart.request.json
+            survey_id = json_data.get('survey_id')
+            answers = json_data.get('answers', {})
+            completed = json_data.get('completed', True)
+
+            if not survey_id:
+                return self.fail(1, 'survey_id required')
+
+            if self.ap.survey:
+                ok = await self.ap.survey.submit_response(survey_id, answers, completed)
+                if ok:
+                    return self.success()
+                return self.fail(2, 'Failed to submit response')
+            return self.fail(3, 'Survey not available')
+
+        @self.route('/dismiss', methods=['POST'], auth_type=group.AuthType.USER_TOKEN)
+        async def _dismiss() -> str:
+            """Dismiss survey."""
+            json_data = await quart.request.json
+            survey_id = json_data.get('survey_id')
+
+            if not survey_id:
+                return self.fail(1, 'survey_id required')
+
+            if self.ap.survey:
+                ok = await self.ap.survey.dismiss_survey(survey_id)
+                if ok:
+                    return self.success()
+                return self.fail(2, 'Failed to dismiss')
+            return self.fail(3, 'Survey not available')
--- a/src/langbot/pkg/api/http/controller/groups/system.py
+++ b/src/langbot/pkg/api/http/controller/groups/system.py
@@ -13,6 +13,7 @@ class SystemRouterGroup(group.RouterGroup):
                data={
                    'version': constants.semantic_version,
                    'debug': constants.debug_mode,
+                    'edition': constants.edition,
                    'enable_marketplace': self.ap.instance_config.data.get('plugin', {}).get(
                        'enable_marketplace', True
                    ),
@@ -25,6 +26,7 @@ class SystemRouterGroup(group.RouterGroup):
                    'disable_models_service': self.ap.instance_config.data.get('space', {}).get(
                        'disable_models_service', False
                    ),
+                    'limitation': self.ap.instance_config.data.get('system', {}).get('limitation', {}),
                }
            )

--- a/src/langbot/pkg/api/http/service/bot.py
+++ b/src/langbot/pkg/api/http/service/bot.py
@@ -70,12 +70,17 @@ class BotService:
            'lark',
        ]:
            webhook_prefix = self.ap.instance_config.data['api'].get('webhook_prefix', 'http://127.0.0.1:5300')
+            extra_webhook_prefix = self.ap.instance_config.data['api'].get('extra_webhook_prefix', '')
            webhook_url = f'/bots/{bot_uuid}'
            adapter_runtime_values['webhook_url'] = webhook_url
            adapter_runtime_values['webhook_full_url'] = f'{webhook_prefix}{webhook_url}'
+            adapter_runtime_values['extra_webhook_full_url'] = (
+                f'{extra_webhook_prefix}{webhook_url}' if extra_webhook_prefix else ''
+            )
        else:
            adapter_runtime_values['webhook_url'] = None
            adapter_runtime_values['webhook_full_url'] = None
+            adapter_runtime_values['extra_webhook_full_url'] = None

        persistence_bot['adapter_runtime_values'] = adapter_runtime_values

@@ -83,6 +88,14 @@ class BotService:

    async def create_bot(self, bot_data: dict) -> str:
        """Create bot"""
+        # Check limitation
+        limitation = self.ap.instance_config.data.get('system', {}).get('limitation', {})
+        max_bots = limitation.get('max_bots', -1)
+        if max_bots >= 0:
+            existing_bots = await self.get_bots()
+            if len(existing_bots) >= max_bots:
+                raise ValueError(f'Maximum number of bots ({max_bots}) reached')
+
        # TODO: 检查配置信息格式
        bot_data['uuid'] = str(uuid.uuid4())

--- a/src/langbot/pkg/api/http/service/external_kb.py
+++ b/src/langbot/pkg/api/http/service/external_kb.py
@@ -1,80 +0,0 @@
-from __future__ import annotations
-
-from ....core import app
-import sqlalchemy
-from langbot.pkg.entity.persistence import rag as persistence_rag
-import uuid
-
-
-class ExternalKBService:
-    """External KB service"""
-
-    ap: app.Application
-
-    def __init__(self, ap: app.Application) -> None:
-        self.ap = ap
-
-    # External Knowledge Base methods
-    async def get_external_knowledge_bases(self) -> list[dict]:
-        result = await self.ap.persistence_mgr.execute_async(sqlalchemy.select(persistence_rag.ExternalKnowledgeBase))
-        external_kbs = result.all()
-        return [
-            self.ap.persistence_mgr.serialize_model(persistence_rag.ExternalKnowledgeBase, external_kb)
-            for external_kb in external_kbs
-        ]
-
-    async def get_external_knowledge_base(self, kb_uuid: str) -> dict | None:
-        result = await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.select(persistence_rag.ExternalKnowledgeBase).where(
-                persistence_rag.ExternalKnowledgeBase.uuid == kb_uuid
-            )
-        )
-        external_kb = result.first()
-        if external_kb is None:
-            return None
-        return self.ap.persistence_mgr.serialize_model(persistence_rag.ExternalKnowledgeBase, external_kb)
-
-    async def create_external_knowledge_base(self, kb_data: dict) -> str:
-        kb_data['uuid'] = str(uuid.uuid4())
-        await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.insert(persistence_rag.ExternalKnowledgeBase).values(kb_data)
-        )
-
-        kb = await self.get_external_knowledge_base(kb_data['uuid'])
-
-        await self.ap.rag_mgr.load_external_knowledge_base(kb)
-
-        return kb_data['uuid']
-
-    async def retrieve_external_knowledge_base(self, kb_uuid: str, query: str) -> list[dict]:
-        """Retrieve external knowledge base"""
-        runtime_kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_uuid)
-        if runtime_kb is None:
-            raise Exception('Knowledge base not found')
-        return [
-            result.model_dump() for result in await runtime_kb.retrieve(query, 5)
-        ]  # top_k is just a placeholder for external knowledge base
-
-    async def update_external_knowledge_base(self, kb_uuid: str, kb_data: dict) -> None:
-        if 'uuid' in kb_data:
-            del kb_data['uuid']
-
-        await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.update(persistence_rag.ExternalKnowledgeBase)
-            .values(kb_data)
-            .where(persistence_rag.ExternalKnowledgeBase.uuid == kb_uuid)
-        )
-        await self.ap.rag_mgr.remove_knowledge_base_from_runtime(kb_uuid)
-
-        kb = await self.get_external_knowledge_base(kb_uuid)
-
-        await self.ap.rag_mgr.load_external_knowledge_base(kb)
-
-    async def delete_external_knowledge_base(self, kb_uuid: str) -> None:
-        await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.delete(persistence_rag.ExternalKnowledgeBase).where(
-                persistence_rag.ExternalKnowledgeBase.uuid == kb_uuid
-            )
-        )
-
-        await self.ap.rag_mgr.delete_knowledge_base(kb_uuid)
--- a/src/langbot/pkg/api/http/service/knowledge.py
+++ b/src/langbot/pkg/api/http/service/knowledge.py
@@ -1,6 +1,5 @@
 from __future__ import annotations

-import uuid
 import sqlalchemy

 from ....core import app
@@ -17,64 +16,77 @@ class KnowledgeService:

    async def get_knowledge_bases(self) -> list[dict]:
        """获取所有知识库"""
-        result = await self.ap.persistence_mgr.execute_async(sqlalchemy.select(persistence_rag.KnowledgeBase))
-        knowledge_bases = result.all()
-        return [
-            self.ap.persistence_mgr.serialize_model(persistence_rag.KnowledgeBase, knowledge_base)
-            for knowledge_base in knowledge_bases
-        ]
+        return await self.ap.rag_mgr.get_all_knowledge_base_details()

    async def get_knowledge_base(self, kb_uuid: str) -> dict | None:
        """获取知识库"""
-        result = await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.select(persistence_rag.KnowledgeBase).where(persistence_rag.KnowledgeBase.uuid == kb_uuid)
-        )
-        knowledge_base = result.first()
-        if knowledge_base is None:
-            return None
-        return self.ap.persistence_mgr.serialize_model(persistence_rag.KnowledgeBase, knowledge_base)
+        return await self.ap.rag_mgr.get_knowledge_base_details(kb_uuid)

    async def create_knowledge_base(self, kb_data: dict) -> str:
        """创建知识库"""
-        kb_data['uuid'] = str(uuid.uuid4())
-        await self.ap.persistence_mgr.execute_async(sqlalchemy.insert(persistence_rag.KnowledgeBase).values(kb_data))
+        # In new architecture, we delegate entirely to RAGManager which uses plugins.
+        # Legacy internal KB creation is removed.

-        kb = await self.get_knowledge_base(kb_data['uuid'])
+        knowledge_engine_plugin_id = kb_data.get('knowledge_engine_plugin_id')
+        if not knowledge_engine_plugin_id:
+            raise ValueError('knowledge_engine_plugin_id is required')

-        await self.ap.rag_mgr.load_knowledge_base(kb)
-
-        return kb_data['uuid']
+        kb = await self.ap.rag_mgr.create_knowledge_base(
+            name=kb_data.get('name', 'Untitled'),
+            knowledge_engine_plugin_id=knowledge_engine_plugin_id,
+            creation_settings=kb_data.get('creation_settings', {}),
+            retrieval_settings=kb_data.get('retrieval_settings', {}),
+            description=kb_data.get('description', ''),
+        )
+        return kb.uuid

    async def update_knowledge_base(self, kb_uuid: str, kb_data: dict) -> None:
        """更新知识库"""
-        if 'uuid' in kb_data:
-            del kb_data['uuid']
+        # Filter to only mutable fields
+        filtered_data = {k: v for k, v in kb_data.items() if k in persistence_rag.KnowledgeBase.MUTABLE_FIELDS}

-        if 'embedding_model_uuid' in kb_data:
-            del kb_data['embedding_model_uuid']
+        if not filtered_data:
+            return

        await self.ap.persistence_mgr.execute_async(
            sqlalchemy.update(persistence_rag.KnowledgeBase)
-            .values(kb_data)
+            .values(filtered_data)
            .where(persistence_rag.KnowledgeBase.uuid == kb_uuid)
        )
        await self.ap.rag_mgr.remove_knowledge_base_from_runtime(kb_uuid)

        kb = await self.get_knowledge_base(kb_uuid)
+        if kb is None:
+            raise Exception('Knowledge base not found after update')

        await self.ap.rag_mgr.load_knowledge_base(kb)

-    async def store_file(self, kb_uuid: str, file_id: str) -> int:
+    async def _check_doc_capability(self, kb_uuid: str, operation: str) -> None:
+        """Check if the KB's Knowledge Engine supports document operations.
+
+        Args:
+            kb_uuid: Knowledge base UUID.
+            operation: Human-readable operation name for error messages.
+
+        Raises:
+            Exception: If the KB does not support doc_ingestion.
+        """
+        kb_info = await self.ap.rag_mgr.get_knowledge_base_details(kb_uuid)
+        if not kb_info:
+            raise Exception('Knowledge base not found')
+        capabilities = kb_info.get('knowledge_engine', {}).get('capabilities', [])
+        if 'doc_ingestion' not in capabilities:
+            raise Exception(f'This knowledge base does not support {operation}')
+
+    async def store_file(self, kb_uuid: str, file_id: str, parser_plugin_id: str | None = None) -> str:
        """存储文件"""
-        # await self.ap.persistence_mgr.execute_async(sqlalchemy.insert(persistence_rag.File).values(kb_id=kb_uuid, file_id=file_id))
-        # await self.ap.rag_mgr.store_file(file_id)
        runtime_kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_uuid)
        if runtime_kb is None:
            raise Exception('Knowledge base not found')
-        # Only internal KBs support file storage
-        if runtime_kb.get_type() != 'internal':
-            raise Exception('Only internal knowledge bases support file storage')
-        result = await runtime_kb.store_file(file_id)
+
+        await self._check_doc_capability(kb_uuid, 'document upload')
+
+        result = await runtime_kb.store_file(file_id, parser_plugin_id=parser_plugin_id)

        # Update the KB's updated_at timestamp
        await self.ap.persistence_mgr.execute_async(
@@ -85,14 +97,18 @@ class KnowledgeService:

        return result

-    async def retrieve_knowledge_base(self, kb_uuid: str, query: str) -> list[dict]:
+    async def retrieve_knowledge_base(
+        self, kb_uuid: str, query: str, retrieval_settings: dict | None = None
+    ) -> list[dict]:
        """检索知识库"""
        runtime_kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_uuid)
        if runtime_kb is None:
            raise Exception('Knowledge base not found')
-        return [
-            result.model_dump() for result in await runtime_kb.retrieve(query, runtime_kb.knowledge_base_entity.top_k)
-        ]
+
+        # Pass retrieval_settings
+        results = await runtime_kb.retrieve(query, settings=retrieval_settings)
+
+        return [result.model_dump() for result in results]

    async def get_files_by_knowledge_base(self, kb_uuid: str) -> list[dict]:
        """获取知识库文件"""
@@ -107,9 +123,9 @@ class KnowledgeService:
        runtime_kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_uuid)
        if runtime_kb is None:
            raise Exception('Knowledge base not found')
-        # Only internal KBs support file deletion
-        if runtime_kb.get_type() != 'internal':
-            raise Exception('Only internal knowledge bases support file deletion')
+
+        await self._check_doc_capability(kb_uuid, 'document deletion')
+
        await runtime_kb.delete_file(file_id)

        # Update the KB's updated_at timestamp
@@ -121,13 +137,14 @@ class KnowledgeService:

    async def delete_knowledge_base(self, kb_uuid: str) -> None:
        """删除知识库"""
-        await self.ap.rag_mgr.delete_knowledge_base(kb_uuid)
-
+        # Delete from DB first to commit the deletion, then clean up runtime/plugin (best-effort)
        await self.ap.persistence_mgr.execute_async(
            sqlalchemy.delete(persistence_rag.KnowledgeBase).where(persistence_rag.KnowledgeBase.uuid == kb_uuid)
        )

        # delete files
+        # NOTE: Chunk cleanup is for legacy (pre-plugin) KBs that stored chunks locally.
+        # For plugin-based Knowledge Engines, the Chunk table is not populated, so this is a no-op.
        files = await self.ap.persistence_mgr.execute_async(
            sqlalchemy.select(persistence_rag.File).where(persistence_rag.File.kb_id == kb_uuid)
        )
@@ -140,3 +157,53 @@ class KnowledgeService:
            await self.ap.persistence_mgr.execute_async(
                sqlalchemy.delete(persistence_rag.File).where(persistence_rag.File.uuid == file.uuid)
            )
+
+        # Remove from runtime and notify plugin (best-effort, DB is already cleaned up)
+        await self.ap.rag_mgr.delete_knowledge_base(kb_uuid)
+
+    # ================= Knowledge Engine Discovery =================
+
+    async def list_knowledge_engines(self) -> list[dict]:
+        """List all available Knowledge Engines from plugins."""
+        engines = []
+
+        if not self.ap.plugin_connector.is_enable_plugin:
+            return engines
+
+        # Get KnowledgeEngine plugins
+        try:
+            knowledge_engines = await self.ap.plugin_connector.list_knowledge_engines()
+            engines.extend(knowledge_engines)
+        except Exception as e:
+            self.ap.logger.warning(f'Failed to list Knowledge Engines from plugins: {e}')
+
+        return engines
+
+    async def list_parsers(self, mime_type: str | None = None) -> list[dict]:
+        """List available parsers, optionally filtered by MIME type."""
+        if not self.ap.plugin_connector.is_enable_plugin:
+            return []
+        try:
+            parsers = await self.ap.plugin_connector.list_parsers()
+            if mime_type:
+                parsers = [p for p in parsers if mime_type in p.get('supported_mime_types', [])]
+            return parsers
+        except Exception as e:
+            self.ap.logger.warning(f'Failed to list parsers: {e}')
+            return []
+
+    async def get_engine_creation_schema(self, plugin_id: str) -> dict:
+        """Get creation settings schema for a specific Knowledge Engine."""
+        try:
+            return await self.ap.plugin_connector.get_rag_creation_schema(plugin_id)
+        except Exception as e:
+            self.ap.logger.warning(f'Failed to get creation schema for {plugin_id}: {e}')
+            return {}
+
+    async def get_engine_retrieval_schema(self, plugin_id: str) -> dict:
+        """Get retrieval settings schema for a specific Knowledge Engine."""
+        try:
+            return await self.ap.plugin_connector.get_rag_retrieval_schema(plugin_id)
+        except Exception as e:
+            self.ap.logger.warning(f'Failed to get retrieval schema for {plugin_id}: {e}')
+            return {}
--- a/src/langbot/pkg/api/http/service/mcp.py
+++ b/src/langbot/pkg/api/http/service/mcp.py
@@ -38,6 +38,16 @@ class MCPService:
        return serialized_servers

    async def create_mcp_server(self, server_data: dict) -> str:
+        # Check limitation (extensions = MCP servers + plugins)
+        limitation = self.ap.instance_config.data.get('system', {}).get('limitation', {})
+        max_extensions = limitation.get('max_extensions', -1)
+        if max_extensions >= 0:
+            existing_mcp_servers = await self.get_mcp_servers()
+            plugins = await self.ap.plugin_connector.list_plugins()
+            total_extensions = len(existing_mcp_servers) + len(plugins)
+            if total_extensions >= max_extensions:
+                raise ValueError(f'Maximum number of extensions ({max_extensions}) reached')
+
        server_data['uuid'] = str(uuid.uuid4())
        await self.ap.persistence_mgr.execute_async(sqlalchemy.insert(persistence_mcp.MCPServer).values(server_data))

--- a/src/langbot/pkg/api/http/service/model.py
+++ b/src/langbot/pkg/api/http/service/model.py
@@ -105,11 +105,16 @@ class LLMModelsService:
                )
            )
            pipeline = result.first()
-            if pipeline is not None and pipeline.config['ai']['local-agent']['model'] == '':
-                pipeline_config = pipeline.config
-                pipeline_config['ai']['local-agent']['model'] = model_data['uuid']
-                pipeline_data = {'config': pipeline_config}
-                await self.ap.pipeline_service.update_pipeline(pipeline.uuid, pipeline_data)
+            if pipeline is not None:
+                model_config = pipeline.config.get('ai', {}).get('local-agent', {}).get('model', {})
+                if not model_config.get('primary', ''):
+                    pipeline_config = pipeline.config
+                    pipeline_config['ai']['local-agent']['model'] = {
+                        'primary': model_data['uuid'],
+                        'fallbacks': [],
+                    }
+                    pipeline_data = {'config': pipeline_config}
+                    await self.ap.pipeline_service.update_pipeline(pipeline.uuid, pipeline_data)

        return model_data['uuid']

--- a/src/langbot/pkg/api/http/service/monitoring.py
+++ b/src/langbot/pkg/api/http/service/monitoring.py
@@ -30,8 +30,10 @@ class MonitoringService:
        level: str = 'info',
        platform: str | None = None,
        user_id: str | None = None,
+        user_name: str | None = None,
        runner_name: str | None = None,
        variables: str | None = None,
+        role: str = 'user',
    ) -> str:
        """Record a message"""
        message_id = str(uuid.uuid4())
@@ -48,8 +50,10 @@ class MonitoringService:
            'level': level,
            'platform': platform,
            'user_id': user_id,
+            'user_name': user_name,
            'runner_name': runner_name,
            'variables': variables,
+            'role': role,
        }

        await self.ap.persistence_mgr.execute_async(
@@ -150,6 +154,7 @@ class MonitoringService:
        pipeline_name: str,
        platform: str | None = None,
        user_id: str | None = None,
+        user_name: str | None = None,
    ) -> None:
        """Record a new session"""
        session_data = {
@@ -164,6 +169,7 @@ class MonitoringService:
            'is_active': True,
            'platform': platform,
            'user_id': user_id,
+            'user_name': user_name,
        }

        await self.ap.persistence_mgr.execute_async(
@@ -355,6 +361,7 @@ class MonitoringService:
        self,
        bot_ids: list[str] | None = None,
        pipeline_ids: list[str] | None = None,
+        session_ids: list[str] | None = None,
        start_time: datetime.datetime | None = None,
        end_time: datetime.datetime | None = None,
        limit: int = 100,
@@ -367,6 +374,8 @@ class MonitoringService:
            conditions.append(persistence_monitoring.MonitoringMessage.bot_id.in_(bot_ids))
        if pipeline_ids:
            conditions.append(persistence_monitoring.MonitoringMessage.pipeline_id.in_(pipeline_ids))
+        if session_ids:
+            conditions.append(persistence_monitoring.MonitoringMessage.session_id.in_(session_ids))
        if start_time:
            conditions.append(persistence_monitoring.MonitoringMessage.timestamp >= start_time)
        if end_time:
@@ -794,3 +803,332 @@ class MonitoringService:
            },
            'errors': errors,
        }
+
+    # ========== Export Methods ==========
+
+    def _escape_csv_field(self, field: str | None) -> str:
+        """Escape a field for CSV output"""
+        if field is None:
+            return ''
+        # Convert non-string types to string first
+        if not isinstance(field, str):
+            field = str(field)
+        # Replace common escape sequences
+        field = field.replace('\r\n', '\n').replace('\r', '\n')
+        # If field contains comma, double quote, or newline, wrap in quotes
+        if ',' in field or '"' in field or '\n' in field:
+            # Escape double quotes by doubling them
+            field = '"' + field.replace('"', '""') + '"'
+        return field
+
+    def _format_timestamp(self, dt: datetime.datetime) -> str:
+        """Format datetime to ISO format string"""
+        return dt.strftime('%Y-%m-%d %H:%M:%S')
+
+    def _extract_message_text(self, message_content: str) -> str:
+        """Extract plain text from message chain JSON"""
+        if not message_content:
+            return ''
+
+        try:
+            import json
+
+            message_chain = json.loads(message_content)
+            if not isinstance(message_chain, list):
+                return message_content
+
+            text_parts = []
+            for component in message_chain:
+                if not isinstance(component, dict):
+                    continue
+                component_type = component.get('type')
+                if component_type == 'Plain':
+                    text = component.get('text', '')
+                    text_parts.append(text)
+                elif component_type == 'At':
+                    display = component.get('display', '')
+                    target = component.get('target', '')
+                    if display:
+                        text_parts.append(f'@{display}')
+                    elif target:
+                        text_parts.append(f'@{target}')
+                elif component_type == 'AtAll':
+                    text_parts.append('@All')
+                elif component_type == 'Image':
+                    text_parts.append('[Image]')
+                elif component_type == 'File':
+                    name = component.get('name', 'File')
+                    text_parts.append(f'[File: {name}]')
+                elif component_type == 'Voice':
+                    length = component.get('length', 0)
+                    text_parts.append(f'[Voice {length}s]')
+                elif component_type == 'Quote':
+                    # Quote content is in 'origin' field
+                    origin = component.get('origin', [])
+                    if isinstance(origin, list):
+                        for item in origin:
+                            if isinstance(item, dict) and item.get('type') == 'Plain':
+                                text_parts.append(f'> {item.get("text", "")}')
+                elif component_type == 'Source':
+                    # Skip Source component
+                    continue
+                else:
+                    # Other unknown types
+                    text_parts.append(f'[{component_type}]')
+
+            return ''.join(text_parts)
+        except (json.JSONDecodeError, TypeError, KeyError):
+            # If not valid JSON, return as-is
+            return message_content
+
+    async def export_messages(
+        self,
+        bot_ids: list[str] | None = None,
+        pipeline_ids: list[str] | None = None,
+        start_time: datetime.datetime | None = None,
+        end_time: datetime.datetime | None = None,
+        limit: int = 100000,
+    ) -> list[dict]:
+        """Export messages as list of dictionaries for CSV conversion"""
+        conditions = []
+
+        if bot_ids:
+            conditions.append(persistence_monitoring.MonitoringMessage.bot_id.in_(bot_ids))
+        if pipeline_ids:
+            conditions.append(persistence_monitoring.MonitoringMessage.pipeline_id.in_(pipeline_ids))
+        if start_time:
+            conditions.append(persistence_monitoring.MonitoringMessage.timestamp >= start_time)
+        if end_time:
+            conditions.append(persistence_monitoring.MonitoringMessage.timestamp <= end_time)
+
+        query = sqlalchemy.select(persistence_monitoring.MonitoringMessage).order_by(
+            persistence_monitoring.MonitoringMessage.timestamp.desc()
+        )
+        if conditions:
+            query = query.where(sqlalchemy.and_(*conditions))
+
+        query = query.limit(limit)
+
+        result = await self.ap.persistence_mgr.execute_async(query)
+        rows = result.all()
+
+        return [
+            {
+                'id': row[0].id if isinstance(row, tuple) else row.id,
+                'timestamp': self._format_timestamp(row[0].timestamp if isinstance(row, tuple) else row.timestamp),
+                'bot_id': row[0].bot_id if isinstance(row, tuple) else row.bot_id,
+                'bot_name': row[0].bot_name if isinstance(row, tuple) else row.bot_name,
+                'pipeline_id': row[0].pipeline_id if isinstance(row, tuple) else row.pipeline_id,
+                'pipeline_name': row[0].pipeline_name if isinstance(row, tuple) else row.pipeline_name,
+                'runner_name': row[0].runner_name if isinstance(row, tuple) else row.runner_name,
+                'message_content': row[0].message_content if isinstance(row, tuple) else row.message_content,
+                'message_text': self._extract_message_text(
+                    row[0].message_content if isinstance(row, tuple) else row.message_content
+                ),
+                'session_id': row[0].session_id if isinstance(row, tuple) else row.session_id,
+                'status': row[0].status if isinstance(row, tuple) else row.status,
+                'level': row[0].level if isinstance(row, tuple) else row.level,
+                'platform': row[0].platform if isinstance(row, tuple) else row.platform,
+                'user_id': row[0].user_id if isinstance(row, tuple) else row.user_id,
+            }
+            for row in rows
+        ]
+
+    async def export_llm_calls(
+        self,
+        bot_ids: list[str] | None = None,
+        pipeline_ids: list[str] | None = None,
+        start_time: datetime.datetime | None = None,
+        end_time: datetime.datetime | None = None,
+        limit: int = 100000,
+    ) -> list[dict]:
+        """Export LLM calls as list of dictionaries for CSV conversion"""
+        conditions = []
+
+        if bot_ids:
+            conditions.append(persistence_monitoring.MonitoringLLMCall.bot_id.in_(bot_ids))
+        if pipeline_ids:
+            conditions.append(persistence_monitoring.MonitoringLLMCall.pipeline_id.in_(pipeline_ids))
+        if start_time:
+            conditions.append(persistence_monitoring.MonitoringLLMCall.timestamp >= start_time)
+        if end_time:
+            conditions.append(persistence_monitoring.MonitoringLLMCall.timestamp <= end_time)
+
+        query = sqlalchemy.select(persistence_monitoring.MonitoringLLMCall).order_by(
+            persistence_monitoring.MonitoringLLMCall.timestamp.desc()
+        )
+        if conditions:
+            query = query.where(sqlalchemy.and_(*conditions))
+
+        query = query.limit(limit)
+
+        result = await self.ap.persistence_mgr.execute_async(query)
+        rows = result.all()
+
+        return [
+            {
+                'id': row[0].id if isinstance(row, tuple) else row.id,
+                'timestamp': self._format_timestamp(row[0].timestamp if isinstance(row, tuple) else row.timestamp),
+                'model_name': row[0].model_name if isinstance(row, tuple) else row.model_name,
+                'input_tokens': row[0].input_tokens if isinstance(row, tuple) else row.input_tokens,
+                'output_tokens': row[0].output_tokens if isinstance(row, tuple) else row.output_tokens,
+                'total_tokens': row[0].total_tokens if isinstance(row, tuple) else row.total_tokens,
+                'duration_ms': row[0].duration if isinstance(row, tuple) else row.duration,
+                'cost': row[0].cost if isinstance(row, tuple) else row.cost,
+                'status': row[0].status if isinstance(row, tuple) else row.status,
+                'bot_id': row[0].bot_id if isinstance(row, tuple) else row.bot_id,
+                'bot_name': row[0].bot_name if isinstance(row, tuple) else row.bot_name,
+                'pipeline_id': row[0].pipeline_id if isinstance(row, tuple) else row.pipeline_id,
+                'pipeline_name': row[0].pipeline_name if isinstance(row, tuple) else row.pipeline_name,
+                'session_id': row[0].session_id if isinstance(row, tuple) else row.session_id,
+                'message_id': row[0].message_id if isinstance(row, tuple) else row.message_id,
+                'error_message': row[0].error_message if isinstance(row, tuple) else row.error_message,
+            }
+            for row in rows
+        ]
+
+    async def export_embedding_calls(
+        self,
+        start_time: datetime.datetime | None = None,
+        end_time: datetime.datetime | None = None,
+        knowledge_base_id: str | None = None,
+        limit: int = 100000,
+    ) -> list[dict]:
+        """Export embedding calls as list of dictionaries for CSV conversion"""
+        conditions = []
+
+        if start_time:
+            conditions.append(persistence_monitoring.MonitoringEmbeddingCall.timestamp >= start_time)
+        if end_time:
+            conditions.append(persistence_monitoring.MonitoringEmbeddingCall.timestamp <= end_time)
+        if knowledge_base_id:
+            conditions.append(persistence_monitoring.MonitoringEmbeddingCall.knowledge_base_id == knowledge_base_id)
+
+        query = sqlalchemy.select(persistence_monitoring.MonitoringEmbeddingCall).order_by(
+            persistence_monitoring.MonitoringEmbeddingCall.timestamp.desc()
+        )
+        if conditions:
+            query = query.where(sqlalchemy.and_(*conditions))
+
+        query = query.limit(limit)
+
+        result = await self.ap.persistence_mgr.execute_async(query)
+        rows = result.all()
+
+        return [
+            {
+                'id': row[0].id if isinstance(row, tuple) else row.id,
+                'timestamp': self._format_timestamp(row[0].timestamp if isinstance(row, tuple) else row.timestamp),
+                'model_name': row[0].model_name if isinstance(row, tuple) else row.model_name,
+                'prompt_tokens': row[0].prompt_tokens if isinstance(row, tuple) else row.prompt_tokens,
+                'total_tokens': row[0].total_tokens if isinstance(row, tuple) else row.total_tokens,
+                'duration_ms': row[0].duration if isinstance(row, tuple) else row.duration,
+                'input_count': row[0].input_count if isinstance(row, tuple) else row.input_count,
+                'status': row[0].status if isinstance(row, tuple) else row.status,
+                'error_message': row[0].error_message if isinstance(row, tuple) else row.error_message,
+                'knowledge_base_id': row[0].knowledge_base_id if isinstance(row, tuple) else row.knowledge_base_id,
+                'query_text': row[0].query_text if isinstance(row, tuple) else row.query_text,
+                'session_id': row[0].session_id if isinstance(row, tuple) else row.session_id,
+                'message_id': row[0].message_id if isinstance(row, tuple) else row.message_id,
+                'call_type': row[0].call_type if isinstance(row, tuple) else row.call_type,
+            }
+            for row in rows
+        ]
+
+    async def export_errors(
+        self,
+        bot_ids: list[str] | None = None,
+        pipeline_ids: list[str] | None = None,
+        start_time: datetime.datetime | None = None,
+        end_time: datetime.datetime | None = None,
+        limit: int = 100000,
+    ) -> list[dict]:
+        """Export errors as list of dictionaries for CSV conversion"""
+        conditions = []
+
+        if bot_ids:
+            conditions.append(persistence_monitoring.MonitoringError.bot_id.in_(bot_ids))
+        if pipeline_ids:
+            conditions.append(persistence_monitoring.MonitoringError.pipeline_id.in_(pipeline_ids))
+        if start_time:
+            conditions.append(persistence_monitoring.MonitoringError.timestamp >= start_time)
+        if end_time:
+            conditions.append(persistence_monitoring.MonitoringError.timestamp <= end_time)
+
+        query = sqlalchemy.select(persistence_monitoring.MonitoringError).order_by(
+            persistence_monitoring.MonitoringError.timestamp.desc()
+        )
+        if conditions:
+            query = query.where(sqlalchemy.and_(*conditions))
+
+        query = query.limit(limit)
+
+        result = await self.ap.persistence_mgr.execute_async(query)
+        rows = result.all()
+
+        return [
+            {
+                'id': row[0].id if isinstance(row, tuple) else row.id,
+                'timestamp': self._format_timestamp(row[0].timestamp if isinstance(row, tuple) else row.timestamp),
+                'error_type': row[0].error_type if isinstance(row, tuple) else row.error_type,
+                'error_message': row[0].error_message if isinstance(row, tuple) else row.error_message,
+                'bot_id': row[0].bot_id if isinstance(row, tuple) else row.bot_id,
+                'bot_name': row[0].bot_name if isinstance(row, tuple) else row.bot_name,
+                'pipeline_id': row[0].pipeline_id if isinstance(row, tuple) else row.pipeline_id,
+                'pipeline_name': row[0].pipeline_name if isinstance(row, tuple) else row.pipeline_name,
+                'session_id': row[0].session_id if isinstance(row, tuple) else row.session_id,
+                'message_id': row[0].message_id if isinstance(row, tuple) else row.message_id,
+                'stack_trace': row[0].stack_trace if isinstance(row, tuple) else row.stack_trace,
+            }
+            for row in rows
+        ]
+
+    async def export_sessions(
+        self,
+        bot_ids: list[str] | None = None,
+        pipeline_ids: list[str] | None = None,
+        start_time: datetime.datetime | None = None,
+        end_time: datetime.datetime | None = None,
+        limit: int = 100000,
+    ) -> list[dict]:
+        """Export sessions as list of dictionaries for CSV conversion"""
+        conditions = []
+
+        if bot_ids:
+            conditions.append(persistence_monitoring.MonitoringSession.bot_id.in_(bot_ids))
+        if pipeline_ids:
+            conditions.append(persistence_monitoring.MonitoringSession.pipeline_id.in_(pipeline_ids))
+        if start_time:
+            conditions.append(persistence_monitoring.MonitoringSession.start_time >= start_time)
+        if end_time:
+            conditions.append(persistence_monitoring.MonitoringSession.start_time <= end_time)
+
+        query = sqlalchemy.select(persistence_monitoring.MonitoringSession).order_by(
+            persistence_monitoring.MonitoringSession.last_activity.desc()
+        )
+        if conditions:
+            query = query.where(sqlalchemy.and_(*conditions))
+
+        query = query.limit(limit)
+
+        result = await self.ap.persistence_mgr.execute_async(query)
+        rows = result.all()
+
+        return [
+            {
+                'session_id': row[0].session_id if isinstance(row, tuple) else row.session_id,
+                'bot_id': row[0].bot_id if isinstance(row, tuple) else row.bot_id,
+                'bot_name': row[0].bot_name if isinstance(row, tuple) else row.bot_name,
+                'pipeline_id': row[0].pipeline_id if isinstance(row, tuple) else row.pipeline_id,
+                'pipeline_name': row[0].pipeline_name if isinstance(row, tuple) else row.pipeline_name,
+                'message_count': row[0].message_count if isinstance(row, tuple) else row.message_count,
+                'start_time': self._format_timestamp(row[0].start_time if isinstance(row, tuple) else row.start_time),
+                'last_activity': self._format_timestamp(
+                    row[0].last_activity if isinstance(row, tuple) else row.last_activity
+                ),
+                'is_active': str(row[0].is_active if isinstance(row, tuple) else row.is_active),
+                'platform': row[0].platform if isinstance(row, tuple) else row.platform,
+                'user_id': row[0].user_id if isinstance(row, tuple) else row.user_id,
+            }
+            for row in rows
+        ]
--- a/src/langbot/pkg/api/http/service/pipeline.py
+++ b/src/langbot/pkg/api/http/service/pipeline.py
@@ -76,6 +76,14 @@ class PipelineService:
    async def create_pipeline(self, pipeline_data: dict, default: bool = False) -> str:
        from ....utils import paths as path_utils

+        # Check limitation
+        limitation = self.ap.instance_config.data.get('system', {}).get('limitation', {})
+        max_pipelines = limitation.get('max_pipelines', -1)
+        if max_pipelines >= 0:
+            existing_pipelines = await self.get_pipelines()
+            if len(existing_pipelines) >= max_pipelines:
+                raise ValueError(f'Maximum number of pipelines ({max_pipelines}) reached')
+
        pipeline_data['uuid'] = str(uuid.uuid4())
        pipeline_data['for_version'] = self.ap.ver_mgr.get_current_version()
        pipeline_data['stages'] = default_stage_order.copy()
@@ -153,6 +161,14 @@ class PipelineService:

    async def copy_pipeline(self, pipeline_uuid: str) -> str:
        """Copy a pipeline with all its configurations"""
+        # Check limitation
+        limitation = self.ap.instance_config.data.get('system', {}).get('limitation', {})
+        max_pipelines = limitation.get('max_pipelines', -1)
+        if max_pipelines >= 0:
+            existing_pipelines = await self.get_pipelines()
+            if len(existing_pipelines) >= max_pipelines:
+                raise ValueError(f'Maximum number of pipelines ({max_pipelines}) reached')
+
        # Get the original pipeline
        result = await self.ap.persistence_mgr.execute_async(
            sqlalchemy.select(persistence_pipeline.LegacyPipeline).where(
--- a/src/langbot/pkg/api/http/service/space.py
+++ b/src/langbot/pkg/api/http/service/space.py
@@ -1,6 +1,6 @@
 from __future__ import annotations

-import aiohttp
+from langbot.pkg.utils import httpclient
 import typing
 import datetime
 import time
@@ -99,49 +99,49 @@ class SpaceService:
        space_config = self._get_space_config()
        space_url = space_config['url']

-        async with aiohttp.ClientSession() as session:
-            async with session.post(
-                f'{space_url}/api/v1/accounts/oauth/token',
-                json={'code': code, 'instance_id': constants.instance_id},
-            ) as response:
-                if response.status != 200:
-                    raise ValueError(f'Failed to exchange OAuth code: {await response.text()}')
-                data = await response.json()
-                if data.get('code') != 0:
-                    raise ValueError(f'Failed to exchange OAuth code: {data.get("msg")}')
-                return data.get('data', {})
+        session = httpclient.get_session()
+        async with session.post(
+            f'{space_url}/api/v1/accounts/oauth/token',
+            json={'code': code, 'instance_id': constants.instance_id},
+        ) as response:
+            if response.status != 200:
+                raise ValueError(f'Failed to exchange OAuth code: {await response.text()}')
+            data = await response.json()
+            if data.get('code') != 0:
+                raise ValueError(f'Failed to exchange OAuth code: {data.get("msg")}')
+            return data.get('data', {})

    async def refresh_token(self, refresh_token: str) -> typing.Dict:
        """Refresh Space access token"""
        space_config = self._get_space_config()
        space_url = space_config['url']

-        async with aiohttp.ClientSession() as session:
-            async with session.post(
-                f'{space_url}/api/v1/accounts/token/refresh', json={'refresh_token': refresh_token}
-            ) as response:
-                if response.status != 200:
-                    raise ValueError(f'Failed to refresh token: {await response.text()}')
-                data = await response.json()
-                if data.get('code') != 0:
-                    raise ValueError(f'Failed to refresh token: {data.get("msg")}')
-                return data.get('data', {})
+        session = httpclient.get_session()
+        async with session.post(
+            f'{space_url}/api/v1/accounts/token/refresh', json={'refresh_token': refresh_token}
+        ) as response:
+            if response.status != 200:
+                raise ValueError(f'Failed to refresh token: {await response.text()}')
+            data = await response.json()
+            if data.get('code') != 0:
+                raise ValueError(f'Failed to refresh token: {data.get("msg")}')
+            return data.get('data', {})

    async def get_user_info_raw(self, access_token: str) -> typing.Dict:
        """Get user info from Space using access token (no validation)"""
        space_config = self._get_space_config()
        space_url = space_config['url']

-        async with aiohttp.ClientSession() as session:
-            async with session.get(
-                f'{space_url}/api/v1/accounts/me', headers={'Authorization': f'Bearer {access_token}'}
-            ) as response:
-                if response.status != 200:
-                    raise ValueError(f'Failed to get user info: {await response.text()}')
-                data = await response.json()
-                if data.get('code') != 0:
-                    raise ValueError(f'Failed to get user info: {data.get("msg")}')
-                return data.get('data', {})
+        session = httpclient.get_session()
+        async with session.get(
+            f'{space_url}/api/v1/accounts/me', headers={'Authorization': f'Bearer {access_token}'}
+        ) as response:
+            if response.status != 200:
+                raise ValueError(f'Failed to get user info: {await response.text()}')
+            data = await response.json()
+            if data.get('code') != 0:
+                raise ValueError(f'Failed to get user info: {data.get("msg")}')
+            return data.get('data', {})

    # === API calls with token validation ===

@@ -178,12 +178,12 @@ class SpaceService:
        space_config = self._get_space_config()
        space_url = space_config['url']

-        async with aiohttp.ClientSession() as session:
-            async with session.get(f'{space_url}/api/v1/models') as response:
-                if response.status != 200:
-                    raise ValueError(f'Failed to get models: {await response.text()}')
-                data = await response.json()
-                if data.get('code') != 0:
-                    raise ValueError(f'Failed to get models: {data.get("msg")}')
-                models_data = data.get('data', {}).get('models', [])
-                return [SpaceModel.model_validate(model_dict) for model_dict in models_data]
+        session = httpclient.get_session()
+        async with session.get(f'{space_url}/api/v1/models') as response:
+            if response.status != 200:
+                raise ValueError(f'Failed to get models: {await response.text()}')
+            data = await response.json()
+            if data.get('code') != 0:
+                raise ValueError(f'Failed to get models: {data.get("msg")}')
+            models_data = data.get('data', {}).get('models', [])
+            return [SpaceModel.model_validate(model_dict) for model_dict in models_data]
--- a/src/langbot/pkg/core/app.py
+++ b/src/langbot/pkg/core/app.py
@@ -9,12 +9,14 @@ from ..platform import botmgr as im_mgr
 from ..platform.webhook_pusher import WebhookPusher
 from ..provider.session import sessionmgr as llm_session_mgr
 from ..provider.modelmgr import modelmgr as llm_model_mgr
+
 from langbot.pkg.provider.tools import toolmgr as llm_tool_mgr
 from ..config import manager as config_mgr
 from ..command import cmdmgr
 from ..plugin import connector as plugin_connector
 from ..pipeline import pool
 from ..pipeline import controller, pipelinemgr
+from ..pipeline import aggregator as message_aggregator
 from ..utils import version as version_mgr, proxy as proxy_mgr
 from ..persistence import mgr as persistencemgr
 from ..api.http.controller import main as http_controller
@@ -28,16 +30,18 @@ from ..api.http.service import knowledge as knowledge_service
 from ..api.http.service import mcp as mcp_service
 from ..api.http.service import apikey as apikey_service
 from ..api.http.service import webhook as webhook_service
-from ..api.http.service import external_kb as external_kb_service
 from ..api.http.service import monitoring as monitoring_service
+
 from ..discover import engine as discover_engine
 from ..storage import mgr as storagemgr
 from ..utils import logcache
 from . import taskmgr
 from . import entities as core_entities
 from ..rag.knowledge import kbmgr as rag_mgr
+from ..rag.service import RAGRuntimeService
 from ..vector import mgr as vectordb_mgr
 from ..telemetry import telemetry as telemetry_module
+from ..survey import manager as survey_module


 class Application:
@@ -61,6 +65,7 @@ class Application:
    model_mgr: llm_model_mgr.ModelManager = None

    rag_mgr: rag_mgr.RAGManager = None
+    rag_runtime_service: RAGRuntimeService = None

    # TODO move to pipeline
    tool_mgr: llm_tool_mgr.ToolManager = None
@@ -96,6 +101,8 @@ class Application:

    query_pool: pool.QueryPool = None

+    msg_aggregator: message_aggregator.MessageAggregator = None
+
    ctrl: controller.Controller = None

    pipeline_mgr: pipelinemgr.PipelineManager = None
@@ -134,8 +141,6 @@ class Application:

    knowledge_service: knowledge_service.KnowledgeService = None

-    external_kb_service: external_kb_service.ExternalKBService = None
-
    mcp_service: mcp_service.MCPService = None

    apikey_service: apikey_service.ApiKeyService = None
@@ -144,6 +149,8 @@ class Application:

    telemetry: telemetry_module.TelemetryManager = None

+    survey: survey_module.SurveyManager = None
+
    monitoring_service: monitoring_service.MonitoringService = None

    def __init__(self):
--- a/src/langbot/pkg/core/bootutils/deps.py
+++ b/src/langbot/pkg/core/bootutils/deps.py
@@ -1,3 +1,4 @@
+import importlib.util
 import pip
 import os
 from ...utils import pkgmgr
@@ -49,9 +50,10 @@ async def check_deps() -> list[str]:

    missing_deps = []
    for dep in required_deps:
-        try:
-            __import__(dep)
-        except ImportError:
+        # Use find_spec instead of __import__ to avoid actually loading
+        # all modules into memory. find_spec only checks if the module
+        # can be found, without executing module-level code.
+        if importlib.util.find_spec(dep) is None:
            missing_deps.append(dep)
    return missing_deps

--- a/src/langbot/pkg/core/stages/build_app.py
+++ b/src/langbot/pkg/core/stages/build_app.py
@@ -5,12 +5,14 @@ import asyncio
 from .. import stage, app
 from ...utils import version, proxy
 from ...pipeline import pool, controller, pipelinemgr
+from ...pipeline import aggregator as message_aggregator
 from ...plugin import connector as plugin_connector
 from ...command import cmdmgr
 from ...provider.session import sessionmgr as llm_session_mgr
 from ...provider.modelmgr import modelmgr as llm_model_mgr
 from ...provider.tools import toolmgr as llm_tool_mgr
 from ...rag.knowledge import kbmgr as rag_mgr
+from ...rag.service import RAGRuntimeService
 from ...platform import botmgr as im_mgr
 from ...platform.webhook_pusher import WebhookPusher
 from ...persistence import mgr as persistencemgr
@@ -25,7 +27,6 @@ from ...api.http.service import knowledge as knowledge_service
 from ...api.http.service import mcp as mcp_service
 from ...api.http.service import apikey as apikey_service
 from ...api.http.service import webhook as webhook_service
-from ...api.http.service import external_kb as external_kb_service
 from ...api.http.service import monitoring as monitoring_service
 from ...discover import engine as discover_engine
 from ...storage import mgr as storagemgr
@@ -33,6 +34,7 @@ from ...utils import logcache
 from ...vector import mgr as vectordb_mgr
 from .. import taskmgr
 from ...telemetry import telemetry as telemetry_module
+from ...survey import manager as survey_module


@stage.stage_class('BuildAppStage')
@@ -71,9 +73,6 @@ class BuildAppStage(stage.BootingStage):
        knowledge_service_inst = knowledge_service.KnowledgeService(ap)
        ap.knowledge_service = knowledge_service_inst

-        external_kb_service_inst = external_kb_service.ExternalKBService(ap)
-        ap.external_kb_service = external_kb_service_inst
-
        mcp_service_inst = mcp_service.MCPService(ap)
        ap.mcp_service = mcp_service_inst

@@ -109,6 +108,11 @@ class BuildAppStage(stage.BootingStage):
        await telemetry_inst.initialize()
        ap.telemetry = telemetry_inst

+        # Survey manager
+        survey_inst = survey_module.SurveyManager(ap)
+        await survey_inst.initialize()
+        ap.survey = survey_inst
+
        cmd_mgr_inst = cmdmgr.CommandManager(ap)
        await cmd_mgr_inst.initialize()
        ap.cmd_mgr = cmd_mgr_inst
@@ -137,10 +141,17 @@ class BuildAppStage(stage.BootingStage):
        await pipeline_mgr.initialize()
        ap.pipeline_mgr = pipeline_mgr

+        # Initialize message aggregator (after pipeline_mgr, as it needs pipeline config)
+        msg_aggregator_inst = message_aggregator.MessageAggregator(ap)
+        ap.msg_aggregator = msg_aggregator_inst
+
        rag_mgr_inst = rag_mgr.RAGManager(ap)
        await rag_mgr_inst.initialize()
        ap.rag_mgr = rag_mgr_inst

+        # Initialize RAG Runtime Service for plugins
+        ap.rag_runtime_service = RAGRuntimeService(ap)
+
        # 初始化向量数据库管理器
        vectordb_mgr_inst = vectordb_mgr.VectorDBManager(ap)
        await vectordb_mgr_inst.initialize()
--- a/src/langbot/pkg/core/stages/load_config.py
+++ b/src/langbot/pkg/core/stages/load_config.py
@@ -74,20 +74,26 @@ def _apply_env_overrides_to_config(cfg: dict) -> dict:
        current = cfg

        for i, key in enumerate(keys):
-            if not isinstance(current, dict) or key not in current:
+            if not isinstance(current, dict):
                break

            if i == len(keys) - 1:
-                # At the final key - check if it's a scalar value
-                if isinstance(current[key], (dict, list)):
-                    # Skip dict and list types
-                    pass
+                # At the final key
+                if key in current:
+                    if isinstance(current[key], (dict, list)):
+                        # Skip dict and list types
+                        pass
+                    else:
+                        # Valid scalar value - convert and set it
+                        converted_value = convert_value(env_value, current[key])
+                        current[key] = converted_value
                else:
-                    # Valid scalar value - convert and set it
-                    converted_value = convert_value(env_value, current[key])
-                    current[key] = converted_value
+                    # Key doesn't exist yet - create it as string
+                    current[key] = env_value
            else:
-                # Navigate deeper
+                # Navigate deeper - create intermediate dict if needed
+                if key not in current:
+                    current[key] = {}
                current = current[key]

    return cfg
@@ -146,18 +152,54 @@ class LoadConfigStage(stage.BootingStage):
        await ap.instance_config.dump_config()

        # load or generate instance id
-        ap.instance_id = await config.load_json_config(
-            'data/labels/instance_id.json',
-            template_data={
-                'instance_id': f'instance_{str(uuid.uuid4())}',
-                'instance_create_ts': int(time.time()),
-            },
-            completion=False,
-        )
+        # Priority:
+        # 1. system.instance_id from config.yaml (can be set via SYSTEM__INSTANCE_ID env var)
+        # 2. data/labels/instance_id.json (if file exists)
+        # 3. Generate new and save to file
+        config_instance_id = ap.instance_config.data.get('system', {}).get('instance_id', '')

-        constants.instance_id = ap.instance_id.data['instance_id']
+        if config_instance_id:
+            # Use the instance_id from config.yaml
+            constants.instance_id = config_instance_id
+            # Still load/create the file for backward compat, but don't use its value
+            ap.instance_id = await config.load_json_config(
+                'data/labels/instance_id.json',
+                template_data={
+                    'instance_id': f'instance_{str(uuid.uuid4())}',
+                    'instance_create_ts': int(time.time()),
+                },
+                completion=False,
+            )
+        else:
+            # Try loading file-based instance id
+            instance_id_path = os.path.join('data', 'labels', 'instance_id.json')
+            if os.path.exists(instance_id_path):
+                # File exists, read it
+                ap.instance_id = await config.load_json_config(
+                    'data/labels/instance_id.json',
+                    template_data={
+                        'instance_id': '',
+                        'instance_create_ts': 0,
+                    },
+                    completion=False,
+                )
+                constants.instance_id = ap.instance_id.data['instance_id']
+            else:
+                # Neither config nor file, generate new and save to file
+                new_id = f'instance_{str(uuid.uuid4())}'
+                ap.instance_id = await config.load_json_config(
+                    'data/labels/instance_id.json',
+                    template_data={
+                        'instance_id': new_id,
+                        'instance_create_ts': int(time.time()),
+                    },
+                    completion=False,
+                )
+                constants.instance_id = new_id
+        constants.edition = ap.instance_config.data.get('system', {}).get('edition', 'community')

        print(f'LangBot instance id: {constants.instance_id}')
+        print(f'LangBot edition: {constants.edition}')

        await ap.instance_id.dump_config()

--- a/src/langbot/pkg/entity/persistence/monitoring.py
+++ b/src/langbot/pkg/entity/persistence/monitoring.py
@@ -20,8 +20,10 @@ class MonitoringMessage(Base):
    level = sqlalchemy.Column(sqlalchemy.String(50), nullable=False)  # info, warning, error, debug
    platform = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
    user_id = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
+    user_name = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)  # User display name
    runner_name = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)  # Runner name for this query
    variables = sqlalchemy.Column(sqlalchemy.Text, nullable=True)  # Query variables as JSON string
+    role = sqlalchemy.Column(sqlalchemy.String(50), nullable=True, default='user')  # user, assistant


 class MonitoringLLMCall(Base):
@@ -63,6 +65,7 @@ class MonitoringSession(Base):
    is_active = sqlalchemy.Column(sqlalchemy.Boolean, nullable=False, default=True, index=True)
    platform = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
    user_id = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
+    user_name = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)  # User display name


 class MonitoringError(Base):
--- a/src/langbot/pkg/entity/persistence/rag.py
+++ b/src/langbot/pkg/entity/persistence/rag.py
@@ -10,8 +10,21 @@ class KnowledgeBase(Base):
    emoji = sqlalchemy.Column(sqlalchemy.String(10), nullable=True, default='📚')
    created_at = sqlalchemy.Column(sqlalchemy.DateTime, default=sqlalchemy.func.now())
    updated_at = sqlalchemy.Column(sqlalchemy.DateTime, default=sqlalchemy.func.now(), onupdate=sqlalchemy.func.now())
-    embedding_model_uuid = sqlalchemy.Column(sqlalchemy.String, default='')
-    top_k = sqlalchemy.Column(sqlalchemy.Integer, default=5)
+    # New fields for plugin-based RAG
+    knowledge_engine_plugin_id = sqlalchemy.Column(sqlalchemy.String, nullable=True)
+    collection_id = sqlalchemy.Column(sqlalchemy.String, nullable=True)
+    creation_settings = sqlalchemy.Column(sqlalchemy.JSON, nullable=True, default=None)
+    retrieval_settings = sqlalchemy.Column(sqlalchemy.JSON, nullable=True, default=None)
+
+    # Field sets for different operations
+    MUTABLE_FIELDS = {'name', 'description', 'retrieval_settings'}
+    """Fields that can be updated after creation."""
+
+    CREATE_FIELDS = MUTABLE_FIELDS | {'uuid', 'knowledge_engine_plugin_id', 'collection_id', 'creation_settings'}
+    """Fields used when creating a new knowledge base."""
+
+    ALL_DB_FIELDS = CREATE_FIELDS | {'emoji', 'created_at', 'updated_at'}
+    """All fields stored in database (for loading from DB row)."""


 class File(Base):
@@ -29,16 +42,3 @@ class Chunk(Base):
    uuid = sqlalchemy.Column(sqlalchemy.String(255), primary_key=True, unique=True)
    file_id = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
    text = sqlalchemy.Column(sqlalchemy.Text)
-
-
-class ExternalKnowledgeBase(Base):
-    __tablename__ = 'external_knowledge_bases'
-    uuid = sqlalchemy.Column(sqlalchemy.String(255), primary_key=True, unique=True)
-    name = sqlalchemy.Column(sqlalchemy.String, index=True)
-    description = sqlalchemy.Column(sqlalchemy.Text)
-    emoji = sqlalchemy.Column(sqlalchemy.String(10), nullable=True, default='🔗')
-    plugin_author = sqlalchemy.Column(sqlalchemy.String, nullable=False)
-    plugin_name = sqlalchemy.Column(sqlalchemy.String, nullable=False)
-    retriever_name = sqlalchemy.Column(sqlalchemy.String, nullable=False)
-    retriever_config = sqlalchemy.Column(sqlalchemy.JSON, nullable=False, default={})
-    created_at = sqlalchemy.Column(sqlalchemy.DateTime, default=sqlalchemy.func.now())
--- a/src/langbot/pkg/persistence/migrations/dbm019_monitoring_message_role.py
+++ b/src/langbot/pkg/persistence/migrations/dbm019_monitoring_message_role.py
@@ -0,0 +1,24 @@
+import sqlalchemy
+from .. import migration
+
+
+@migration.migration_class(19)
+class DBMigrateMonitoringMessageRole(migration.DBMigration):
+    """Add role column to monitoring_messages table"""
+
+    async def upgrade(self):
+        """Upgrade"""
+        try:
+            sql_text = sqlalchemy.text("ALTER TABLE monitoring_messages ADD COLUMN role VARCHAR(50) DEFAULT 'user'")
+            await self.ap.persistence_mgr.execute_async(sql_text)
+        except Exception:
+            # Column may already exist
+            pass
+
+    async def downgrade(self):
+        """Downgrade"""
+        try:
+            sql_text = sqlalchemy.text('ALTER TABLE monitoring_messages DROP COLUMN role')
+            await self.ap.persistence_mgr.execute_async(sql_text)
+        except Exception:
+            pass
--- a/src/langbot/pkg/persistence/migrations/dbm020_knowledge_engine_plugin_architecture.py
+++ b/src/langbot/pkg/persistence/migrations/dbm020_knowledge_engine_plugin_architecture.py
@@ -0,0 +1,161 @@
+import sqlalchemy
+from .. import migration
+
+
+@migration.migration_class(20)
+class DBMigrateKnowledgeEnginePluginArchitecture(migration.DBMigration):
+    """Migrate to unified Knowledge Engine plugin architecture.
+
+    Changes:
+    - Backup existing knowledge_bases data to knowledge_bases_backup
+    - Clear knowledge_bases table and add new plugin architecture columns
+    - Drop old columns (PostgreSQL only; SQLite leaves them unmapped)
+    - Preserve external_knowledge_bases table as-is for future migration
+    - Set rag_plugin_migration_needed flag in metadata if old data exists
+    """
+
+    async def upgrade(self):
+        """Upgrade"""
+        has_internal_data = await self._backup_knowledge_bases()
+        has_external_data = await self._check_external_knowledge_bases()
+        await self._clear_knowledge_bases()
+        await self._add_columns_to_knowledge_bases()
+        await self._drop_old_columns()
+        if has_internal_data or has_external_data:
+            await self._set_migration_flag()
+
+    async def _get_table_columns(self, table_name: str) -> list[str]:
+        """Get column names from a table (works for both SQLite and PostgreSQL)."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT column_name FROM information_schema.columns WHERE table_name = :table_name;'
+                ).bindparams(table_name=table_name)
+            )
+            return [row[0] for row in result.fetchall()]
+        else:
+            # SQLite PRAGMA does not support bind parameters; validate identifier.
+            if not table_name.isidentifier():
+                raise ValueError(f'Invalid table name: {table_name}')
+            result = await self.ap.persistence_mgr.execute_async(sqlalchemy.text(f'PRAGMA table_info({table_name});'))
+            return [row[1] for row in result.fetchall()]
+
+    async def _table_exists(self, table_name: str) -> bool:
+        """Check if a table exists."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT EXISTS (SELECT FROM information_schema.tables WHERE table_name = :table_name);'
+                ).bindparams(table_name=table_name)
+            )
+            return result.scalar()
+        else:
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text("SELECT name FROM sqlite_master WHERE type='table' AND name=:table_name;").bindparams(
+                    table_name=table_name
+                )
+            )
+            return result.first() is not None
+
+    async def _backup_knowledge_bases(self) -> bool:
+        """Backup knowledge_bases data. Returns True if data was backed up."""
+        result = await self.ap.persistence_mgr.execute_async(sqlalchemy.text('SELECT COUNT(*) FROM knowledge_bases;'))
+        count = result.scalar()
+        if count == 0:
+            return False
+
+        # Drop backup table if it already exists (from a previous failed migration)
+        if await self._table_exists('knowledge_bases_backup'):
+            await self.ap.persistence_mgr.execute_async(sqlalchemy.text('DROP TABLE knowledge_bases_backup;'))
+
+        await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('CREATE TABLE knowledge_bases_backup AS SELECT * FROM knowledge_bases;')
+        )
+        self.ap.logger.info(
+            'Backed up %d knowledge base(s) to knowledge_bases_backup table.',
+            count,
+        )
+        return True
+
+    async def _check_external_knowledge_bases(self) -> bool:
+        """Check if external_knowledge_bases table exists and has data.
+
+        The table is preserved as-is (not dropped) for future migration.
+        """
+        if not await self._table_exists('external_knowledge_bases'):
+            return False
+
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT COUNT(*) FROM external_knowledge_bases;')
+        )
+        count = result.scalar()
+        if count > 0:
+            self.ap.logger.info(
+                'Found %d external knowledge base(s) in external_knowledge_bases table. '
+                'Table preserved for future migration.',
+                count,
+            )
+        return count > 0
+
+    async def _clear_knowledge_bases(self):
+        """Clear all rows from knowledge_bases table (preserve table structure)."""
+        await self.ap.persistence_mgr.execute_async(sqlalchemy.text('DELETE FROM knowledge_bases;'))
+
+    async def _add_columns_to_knowledge_bases(self):
+        """Add new RAG plugin architecture columns to knowledge_bases table."""
+        columns = await self._get_table_columns('knowledge_bases')
+
+        new_columns = {
+            'knowledge_engine_plugin_id': 'VARCHAR',
+            'collection_id': 'VARCHAR',
+            'creation_settings': 'TEXT',  # JSON stored as TEXT for SQLite compatibility
+            'retrieval_settings': 'TEXT',
+        }
+
+        for col_name, col_type in new_columns.items():
+            if col_name not in columns:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(f'ALTER TABLE knowledge_bases ADD COLUMN {col_name} {col_type};')
+                )
+
+    async def _drop_old_columns(self):
+        """Drop embedding_model_uuid and top_k columns (PostgreSQL only).
+
+        SQLite does not support DROP COLUMN in older versions, so we leave the
+        columns in place — the SQLAlchemy entity simply won't map them.
+        """
+        if self.ap.persistence_mgr.db.name != 'postgresql':
+            return
+
+        columns = await self._get_table_columns('knowledge_bases')
+
+        if 'embedding_model_uuid' in columns:
+            await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text('ALTER TABLE knowledge_bases DROP COLUMN embedding_model_uuid;')
+            )
+
+        if 'top_k' in columns:
+            await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text('ALTER TABLE knowledge_bases DROP COLUMN top_k;')
+            )
+
+    async def _set_migration_flag(self):
+        """Set rag_plugin_migration_needed flag in metadata table."""
+        # Check if the key already exists
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text("SELECT value FROM metadata WHERE key = 'rag_plugin_migration_needed';")
+        )
+        row = result.first()
+        if row is not None:
+            await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text("UPDATE metadata SET value = 'true' WHERE key = 'rag_plugin_migration_needed';")
+            )
+        else:
+            await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text("INSERT INTO metadata (key, value) VALUES ('rag_plugin_migration_needed', 'true');")
+            )
+        self.ap.logger.info('Set rag_plugin_migration_needed=true in metadata.')
+
+    async def downgrade(self):
+        """Downgrade"""
+        pass
--- a/src/langbot/pkg/persistence/migrations/dbm021_merge_exception_handling.py
+++ b/src/langbot/pkg/persistence/migrations/dbm021_merge_exception_handling.py
@@ -0,0 +1,74 @@
+from .. import migration
+
+import sqlalchemy
+import json
+
+
+@migration.migration_class(21)
+class DBMigrateMergeExceptionHandling(migration.DBMigration):
+    """Merge hide-exception and block-failed-request-output into a single exception-handling select option,
+    and add failure-hint field.
+
+    Conversion logic:
+    - block-failed-request-output=true  ->  exception-handling: hide
+    - hide-exception=true               ->  exception-handling: show-hint
+    - hide-exception=false              ->  exception-handling: show-error
+    """
+
+    async def upgrade(self):
+        """Upgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT uuid, config FROM legacy_pipelines')
+        )
+        pipelines = result.fetchall()
+
+        current_version = self.ap.ver_mgr.get_current_version()
+
+        for pipeline_row in pipelines:
+            uuid = pipeline_row[0]
+            config = json.loads(pipeline_row[1]) if isinstance(pipeline_row[1], str) else pipeline_row[1]
+
+            if 'output' not in config:
+                config['output'] = {}
+            if 'misc' not in config['output']:
+                config['output']['misc'] = {}
+
+            misc = config['output']['misc']
+
+            # Determine new exception-handling value from legacy fields
+            hide_exception = misc.get('hide-exception', True)
+            block_failed = misc.get('block-failed-request-output', False)
+
+            if block_failed:
+                exception_handling = 'hide'
+            elif hide_exception:
+                exception_handling = 'show-hint'
+            else:
+                exception_handling = 'show-error'
+
+            misc['exception-handling'] = exception_handling
+
+            # Add failure-hint with default value
+            misc['failure-hint'] = 'Request failed.'
+
+            # Remove legacy fields
+            misc.pop('hide-exception', None)
+
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config::jsonb, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+
+    async def downgrade(self):
+        """Downgrade"""
+        pass
--- a/src/langbot/pkg/persistence/migrations/dbm022_monitoring_user_name.py
+++ b/src/langbot/pkg/persistence/migrations/dbm022_monitoring_user_name.py
@@ -0,0 +1,73 @@
+import sqlalchemy
+from .. import migration
+
+
+@migration.migration_class(22)
+class DBMigrateMonitoringUserId(migration.DBMigration):
+    """Add user_id and user_name columns to monitoring_sessions table
+
+    This migration adds the missing user_id column and also ensures user_name
+    column exists (in case migration 21 failed or was skipped).
+    """
+
+    async def _table_exists(self, table_name: str) -> bool:
+        """Check if a table exists (works for both SQLite and PostgreSQL)."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT EXISTS (SELECT FROM information_schema.tables WHERE table_name = :table_name);'
+                ).bindparams(table_name=table_name)
+            )
+            return bool(result.scalar())
+        else:
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text("SELECT name FROM sqlite_master WHERE type='table' AND name=:table_name;").bindparams(
+                    table_name=table_name
+                )
+            )
+            return result.first() is not None
+
+    async def _get_table_columns(self, table_name: str) -> list[str]:
+        """Get column names from a table (works for both SQLite and PostgreSQL)."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT column_name FROM information_schema.columns WHERE table_name = :table_name;'
+                ).bindparams(table_name=table_name)
+            )
+            return [row[0] for row in result.fetchall()]
+        else:
+            if not table_name.isidentifier():
+                raise ValueError(f'Invalid table name: {table_name}')
+            result = await self.ap.persistence_mgr.execute_async(sqlalchemy.text(f'PRAGMA table_info({table_name});'))
+            return [row[1] for row in result.fetchall()]
+
+    async def _add_column_if_not_exists(self, table_name: str, column_name: str, column_type: str):
+        """Add a column to a table if it does not already exist."""
+        columns = await self._get_table_columns(table_name)
+        if column_name in columns:
+            self.ap.logger.debug('%s column already exists in %s.', column_name, table_name)
+            return
+        await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text(f'ALTER TABLE {table_name} ADD COLUMN {column_name} {column_type};')
+        )
+        self.ap.logger.info('Added %s column to %s table.', column_name, table_name)
+
+    async def upgrade(self):
+        # Check if monitoring_sessions table exists
+        if not await self._table_exists('monitoring_sessions'):
+            self.ap.logger.warning('monitoring_sessions table does not exist, skipping migration.')
+            return
+
+        # Add user_id column to monitoring_sessions table
+        await self._add_column_if_not_exists('monitoring_sessions', 'user_id', 'VARCHAR(255)')
+
+        # Add user_name column to monitoring_sessions table (in case migration 21 failed)
+        await self._add_column_if_not_exists('monitoring_sessions', 'user_name', 'VARCHAR(255)')
+
+        # Add user_name column to monitoring_messages table (in case migration 21 failed)
+        if await self._table_exists('monitoring_messages'):
+            await self._add_column_if_not_exists('monitoring_messages', 'user_name', 'VARCHAR(255)')
+
+    async def downgrade(self):
+        pass
--- a/src/langbot/pkg/persistence/migrations/dbm023_model_fallback_config.py
+++ b/src/langbot/pkg/persistence/migrations/dbm023_model_fallback_config.py
@@ -0,0 +1,102 @@
+from .. import migration
+
+import sqlalchemy
+import json
+
+
+@migration.migration_class(23)
+class DBMigrateModelFallbackConfig(migration.DBMigration):
+    """Convert model field from plain UUID string to object with primary/fallbacks"""
+
+    async def upgrade(self):
+        """Upgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT uuid, config FROM legacy_pipelines')
+        )
+        pipelines = result.fetchall()
+
+        current_version = self.ap.ver_mgr.get_current_version()
+
+        for pipeline_row in pipelines:
+            uuid = pipeline_row[0]
+            config = json.loads(pipeline_row[1]) if isinstance(pipeline_row[1], str) else pipeline_row[1]
+
+            if 'ai' not in config or 'local-agent' not in config['ai']:
+                continue
+
+            local_agent = config['ai']['local-agent']
+            changed = False
+
+            # Convert model from string to object
+            model_value = local_agent.get('model', '')
+            if isinstance(model_value, str):
+                local_agent['model'] = {
+                    'primary': model_value,
+                    'fallbacks': [],
+                }
+                changed = True
+
+            # Remove leftover fallback-models field if present
+            if 'fallback-models' in local_agent:
+                del local_agent['fallback-models']
+                changed = True
+
+            if not changed:
+                continue
+
+            # Update using raw SQL with compatibility for both SQLite and PostgreSQL
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config::jsonb, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+
+    async def downgrade(self):
+        """Downgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT uuid, config FROM legacy_pipelines')
+        )
+        pipelines = result.fetchall()
+
+        current_version = self.ap.ver_mgr.get_current_version()
+
+        for pipeline_row in pipelines:
+            uuid = pipeline_row[0]
+            config = json.loads(pipeline_row[1]) if isinstance(pipeline_row[1], str) else pipeline_row[1]
+
+            if 'ai' not in config or 'local-agent' not in config['ai']:
+                continue
+
+            local_agent = config['ai']['local-agent']
+
+            # Convert model from object back to string
+            model_value = local_agent.get('model', '')
+            if isinstance(model_value, dict):
+                local_agent['model'] = model_value.get('primary', '')
+            else:
+                continue
+
+            # Update using raw SQL with compatibility for both SQLite and PostgreSQL
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config::jsonb, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
--- a/src/langbot/pkg/persistence/migrations/dbm024_wecombot_websocket_mode.py
+++ b/src/langbot/pkg/persistence/migrations/dbm024_wecombot_websocket_mode.py
@@ -0,0 +1,49 @@
+from .. import migration
+
+import sqlalchemy
+import json
+
+
+@migration.migration_class(24)
+class DBMigrateWecomBotWebSocketMode(migration.DBMigration):
+    """Add enable-webhook field to existing wecombot adapter configs.
+
+    Existing wecombot bots were all using webhook mode, so we set
+    enable-webhook=true to preserve their behavior after the new
+    WebSocket long connection mode is introduced as default.
+    """
+
+    async def upgrade(self):
+        """Upgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text("SELECT uuid, adapter_config FROM bots WHERE adapter = 'wecombot'")
+        )
+        bots = result.fetchall()
+
+        for bot_row in bots:
+            bot_uuid = bot_row[0]
+            adapter_config = json.loads(bot_row[1]) if isinstance(bot_row[1], str) else bot_row[1]
+
+            if 'enable-webhook' in adapter_config:
+                continue
+
+            # Determine mode based on existing config: if webhook fields are present, keep webhook mode
+            has_webhook_config = bool(
+                adapter_config.get('Token') and adapter_config.get('EncodingAESKey') and adapter_config.get('Corpid')
+            )
+            adapter_config['enable-webhook'] = has_webhook_config
+
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text('UPDATE bots SET adapter_config = :config::jsonb WHERE uuid = :uuid'),
+                    {'config': json.dumps(adapter_config), 'uuid': bot_uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text('UPDATE bots SET adapter_config = :config WHERE uuid = :uuid'),
+                    {'config': json.dumps(adapter_config), 'uuid': bot_uuid},
+                )
+
+    async def downgrade(self):
+        """Downgrade"""
+        pass
--- a/src/langbot/pkg/pipeline/aggregator.py
+++ b/src/langbot/pkg/pipeline/aggregator.py
@@ -0,0 +1,289 @@
+"""Message Aggregator Module
+
+This module provides message aggregation/debounce functionality.
+When users send multiple messages consecutively, the aggregator will wait
+for a configurable delay period and merge them into a single message
+before processing.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import time
+import typing
+from dataclasses import dataclass, field
+
+import langbot_plugin.api.entities.builtin.platform.message as platform_message
+import langbot_plugin.api.entities.builtin.platform.events as platform_events
+import langbot_plugin.api.entities.builtin.provider.session as provider_session
+import langbot_plugin.api.definition.abstract.platform.adapter as abstract_platform_adapter
+
+if typing.TYPE_CHECKING:
+    from ..core import app
+
+# Maximum number of messages to buffer before forcing a flush
+MAX_BUFFER_MESSAGES = 10
+
+
+@dataclass
+class PendingMessage:
+    """A pending message waiting to be aggregated"""
+
+    bot_uuid: str
+    launcher_type: provider_session.LauncherTypes
+    launcher_id: typing.Union[int, str]
+    sender_id: typing.Union[int, str]
+    message_event: platform_events.MessageEvent
+    message_chain: platform_message.MessageChain
+    adapter: abstract_platform_adapter.AbstractMessagePlatformAdapter
+    pipeline_uuid: typing.Optional[str]
+    timestamp: float = field(default_factory=time.time)
+
+
+@dataclass
+class SessionBuffer:
+    """Buffer for a single session's pending messages"""
+
+    session_id: str
+    messages: list[PendingMessage] = field(default_factory=list)
+    timer_task: typing.Optional[asyncio.Task] = None
+    last_message_time: float = field(default_factory=time.time)
+
+
+class MessageAggregator:
+    """Message aggregator that buffers and merges consecutive messages
+
+    This class implements a debounce mechanism for incoming messages.
+    When a message arrives, it starts a timer. If more messages arrive
+    before the timer expires, they are buffered. When the timer expires,
+    all buffered messages are merged and sent to the query pool.
+    """
+
+    ap: app.Application
+
+    buffers: dict[str, SessionBuffer]
+    """Session ID -> SessionBuffer mapping"""
+
+    lock: asyncio.Lock
+    """Lock for thread-safe buffer operations"""
+
+    def __init__(self, ap: app.Application):
+        self.ap = ap
+        self.buffers = {}
+        self.lock = asyncio.Lock()
+
+    def _get_session_id(
+        self,
+        bot_uuid: str,
+        launcher_type: provider_session.LauncherTypes,
+        launcher_id: typing.Union[int, str],
+    ) -> str:
+        """Generate a unique session ID"""
+        return f'{bot_uuid}:{launcher_type.value}:{launcher_id}'
+
+    async def _get_aggregation_config(self, pipeline_uuid: typing.Optional[str]) -> tuple[bool, float]:
+        """Get aggregation configuration for a pipeline
+
+        Returns:
+            tuple: (enabled, delay_seconds)
+        """
+        default_enabled = False
+        default_delay = 1.5
+
+        if pipeline_uuid is None:
+            return default_enabled, default_delay
+
+        # Get pipeline from pipeline manager
+        pipeline = await self.ap.pipeline_mgr.get_pipeline_by_uuid(pipeline_uuid)
+        if pipeline is None:
+            return default_enabled, default_delay
+
+        config = pipeline.pipeline_entity.config or {}
+        trigger_config = config.get('trigger', {})
+        aggregation_config = trigger_config.get('message-aggregation', {})
+
+        enabled = aggregation_config.get('enabled', default_enabled)
+
+        delay_raw = aggregation_config.get('delay', default_delay)
+        try:
+            delay = float(delay_raw)
+        except (TypeError, ValueError):
+            delay = default_delay
+
+        # Clamp delay to valid range
+        delay = max(1.0, min(10.0, delay))
+
+        return enabled, delay
+
+    async def add_message(
+        self,
+        bot_uuid: str,
+        launcher_type: provider_session.LauncherTypes,
+        launcher_id: typing.Union[int, str],
+        sender_id: typing.Union[int, str],
+        message_event: platform_events.MessageEvent,
+        message_chain: platform_message.MessageChain,
+        adapter: abstract_platform_adapter.AbstractMessagePlatformAdapter,
+        pipeline_uuid: typing.Optional[str] = None,
+    ) -> None:
+        """Add a message to the aggregation buffer
+
+        If aggregation is disabled for the pipeline, the message is sent
+        directly to the query pool. Otherwise, it's buffered and will be
+        merged with other messages from the same session.
+        """
+        enabled, delay = await self._get_aggregation_config(pipeline_uuid)
+
+        if not enabled:
+            # Aggregation disabled, send directly to query pool
+            await self.ap.query_pool.add_query(
+                bot_uuid=bot_uuid,
+                launcher_type=launcher_type,
+                launcher_id=launcher_id,
+                sender_id=sender_id,
+                message_event=message_event,
+                message_chain=message_chain,
+                adapter=adapter,
+                pipeline_uuid=pipeline_uuid,
+            )
+            return
+
+        session_id = self._get_session_id(bot_uuid, launcher_type, launcher_id)
+
+        pending_msg = PendingMessage(
+            bot_uuid=bot_uuid,
+            launcher_type=launcher_type,
+            launcher_id=launcher_id,
+            sender_id=sender_id,
+            message_event=message_event,
+            message_chain=message_chain,
+            adapter=adapter,
+            pipeline_uuid=pipeline_uuid,
+        )
+
+        force_flush = False
+        async with self.lock:
+            if session_id in self.buffers:
+                buffer = self.buffers[session_id]
+                # Cancel existing timer (just cancel, don't await inside lock)
+                if buffer.timer_task and not buffer.timer_task.done():
+                    buffer.timer_task.cancel()
+                buffer.messages.append(pending_msg)
+            else:
+                buffer = SessionBuffer(
+                    session_id=session_id,
+                    messages=[pending_msg],
+                )
+                self.buffers[session_id] = buffer
+
+            buffer.last_message_time = time.time()
+
+            # Check if buffer reached max capacity
+            if len(buffer.messages) >= MAX_BUFFER_MESSAGES:
+                force_flush = True
+            else:
+                # Start new timer
+                buffer.timer_task = asyncio.create_task(self._delayed_flush(session_id, delay))
+
+        if force_flush:
+            await self._flush_buffer(session_id)
+
+    async def _delayed_flush(self, session_id: str, delay: float) -> None:
+        """Wait for delay then flush the buffer"""
+        try:
+            await asyncio.sleep(delay)
+            await self._flush_buffer(session_id)
+        except asyncio.CancelledError:
+            # Timer was cancelled, new message arrived
+            pass
+
+    async def _flush_buffer(self, session_id: str) -> None:
+        """Flush the buffer for a session, merging all messages"""
+        async with self.lock:
+            buffer = self.buffers.pop(session_id, None)
+
+        if buffer is None or not buffer.messages:
+            return
+
+        if len(buffer.messages) == 1:
+            # Only one message, no need to merge
+            msg = buffer.messages[0]
+            await self.ap.query_pool.add_query(
+                bot_uuid=msg.bot_uuid,
+                launcher_type=msg.launcher_type,
+                launcher_id=msg.launcher_id,
+                sender_id=msg.sender_id,
+                message_event=msg.message_event,
+                message_chain=msg.message_chain,
+                adapter=msg.adapter,
+                pipeline_uuid=msg.pipeline_uuid,
+            )
+            return
+
+        # Merge multiple messages
+        merged_msg = self._merge_messages(buffer.messages)
+        await self.ap.query_pool.add_query(
+            bot_uuid=merged_msg.bot_uuid,
+            launcher_type=merged_msg.launcher_type,
+            launcher_id=merged_msg.launcher_id,
+            sender_id=merged_msg.sender_id,
+            message_event=merged_msg.message_event,
+            message_chain=merged_msg.message_chain,
+            adapter=merged_msg.adapter,
+            pipeline_uuid=merged_msg.pipeline_uuid,
+        )
+
+    def _merge_messages(self, messages: list[PendingMessage]) -> PendingMessage:
+        """Merge multiple messages into one
+
+        The merged message uses the first message as base and combines
+        all message chains with newline separators.
+        The original message_event is kept unmodified to preserve
+        message metadata (message_id, etc.) for reply/quote.
+        """
+        if len(messages) == 1:
+            return messages[0]
+
+        base_msg = messages[0]
+
+        # Build merged message chain
+        merged_chain = platform_message.MessageChain([])
+
+        for i, msg in enumerate(messages):
+            if i > 0:
+                # Add newline separator between messages
+                merged_chain.append(platform_message.Plain(text='\n'))
+
+            # Copy all components from this message
+            for component in msg.message_chain:
+                merged_chain.append(component)
+
+        # Keep message_event unmodified (preserves original message_id and
+        # metadata for reply/quote), only pass merged chain separately
+        return PendingMessage(
+            bot_uuid=base_msg.bot_uuid,
+            launcher_type=base_msg.launcher_type,
+            launcher_id=base_msg.launcher_id,
+            sender_id=base_msg.sender_id,
+            message_event=base_msg.message_event,
+            message_chain=merged_chain,
+            adapter=base_msg.adapter,
+            pipeline_uuid=base_msg.pipeline_uuid,
+        )
+
+    async def flush_all(self) -> None:
+        """Flush all pending buffers immediately
+
+        This is useful during shutdown to ensure no messages are lost.
+        """
+        # Snapshot session IDs and cancel all timers under lock
+        async with self.lock:
+            session_ids = list(self.buffers.keys())
+            for sid in session_ids:
+                buffer = self.buffers.get(sid)
+                if buffer and buffer.timer_task and not buffer.timer_task.done():
+                    buffer.timer_task.cancel()
+
+        # Flush each buffer outside the lock
+        for session_id in session_ids:
+            await self._flush_buffer(session_id)
--- a/src/langbot/pkg/pipeline/cntfilter/filters/baiduexamine.py
+++ b/src/langbot/pkg/pipeline/cntfilter/filters/baiduexamine.py
@@ -1,10 +1,9 @@
 from __future__ import annotations

-import aiohttp
-
 from .. import entities
 from .. import filter as filter_model
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
+from langbot.pkg.utils import httpclient

 BAIDU_EXAMINE_URL = 'https://aip.baidubce.com/rest/2.0/solution/v1/text_censor/v2/user_defined?access_token={}'
 BAIDU_EXAMINE_TOKEN_URL = 'https://aip.baidubce.com/oauth/2.0/token'
@@ -15,50 +14,50 @@ class BaiduCloudExamine(filter_model.ContentFilter):
    """百度云内容审核"""

    async def _get_token(self) -> str:
-        async with aiohttp.ClientSession() as session:
-            async with session.post(
-                BAIDU_EXAMINE_TOKEN_URL,
-                params={
-                    'grant_type': 'client_credentials',
-                    'client_id': self.ap.pipeline_cfg.data['baidu-cloud-examine']['api-key'],
-                    'client_secret': self.ap.pipeline_cfg.data['baidu-cloud-examine']['api-secret'],
-                },
-            ) as resp:
-                return (await resp.json())['access_token']
+        session = httpclient.get_session()
+        async with session.post(
+            BAIDU_EXAMINE_TOKEN_URL,
+            params={
+                'grant_type': 'client_credentials',
+                'client_id': self.ap.pipeline_cfg.data['baidu-cloud-examine']['api-key'],
+                'client_secret': self.ap.pipeline_cfg.data['baidu-cloud-examine']['api-secret'],
+            },
+        ) as resp:
+            return (await resp.json())['access_token']

    async def process(self, query: pipeline_query.Query, message: str) -> entities.FilterResult:
-        async with aiohttp.ClientSession() as session:
-            async with session.post(
-                BAIDU_EXAMINE_URL.format(await self._get_token()),
-                headers={
-                    'Content-Type': 'application/x-www-form-urlencoded',
-                    'Accept': 'application/json',
-                },
-                data=f'text={message}'.encode('utf-8'),
-            ) as resp:
-                result = await resp.json()
+        session = httpclient.get_session()
+        async with session.post(
+            BAIDU_EXAMINE_URL.format(await self._get_token()),
+            headers={
+                'Content-Type': 'application/x-www-form-urlencoded',
+                'Accept': 'application/json',
+            },
+            data=f'text={message}'.encode('utf-8'),
+        ) as resp:
+            result = await resp.json()

-                if 'error_code' in result:
+            if 'error_code' in result:
+                return entities.FilterResult(
+                    level=entities.ResultLevel.BLOCK,
+                    replacement=message,
+                    user_notice='',
+                    console_notice=f'百度云判定出错，错误信息：{result["error_msg"]}',
+                )
+            else:
+                conclusion = result['conclusion']
+
+                if conclusion in ('合规'):
+                    return entities.FilterResult(
+                        level=entities.ResultLevel.PASS,
+                        replacement=message,
+                        user_notice='',
+                        console_notice=f'百度云判定结果：{conclusion}',
+                    )
+                else:
                    return entities.FilterResult(
                        level=entities.ResultLevel.BLOCK,
                        replacement=message,
-                        user_notice='',
-                        console_notice=f'百度云判定出错，错误信息：{result["error_msg"]}',
+                        user_notice='消息中存在不合适的内容, 请修改',
+                        console_notice=f'百度云判定结果：{conclusion}',
                    )
-                else:
-                    conclusion = result['conclusion']
-
-                    if conclusion in ('合规'):
-                        return entities.FilterResult(
-                            level=entities.ResultLevel.PASS,
-                            replacement=message,
-                            user_notice='',
-                            console_notice=f'百度云判定结果：{conclusion}',
-                        )
-                    else:
-                        return entities.FilterResult(
-                            level=entities.ResultLevel.BLOCK,
-                            replacement=message,
-                            user_notice='消息中存在不合适的内容, 请修改',
-                            console_notice=f'百度云判定结果：{conclusion}',
-                        )
--- a/src/langbot/pkg/pipeline/config_coercion.py
+++ b/src/langbot/pkg/pipeline/config_coercion.py
@@ -0,0 +1,105 @@
+from __future__ import annotations
+
+import logging
+
+logger = logging.getLogger(__name__)
+
+# metadata type -> coercion function
+_COERCE_MAP = {
+    'integer': lambda v: int(v),
+    'number': lambda v: float(v),
+    'float': lambda v: float(v),
+}
+
+
+def _coerce_bool(v):
+    if isinstance(v, bool):
+        return v
+    if isinstance(v, str):
+        if v.lower() == 'true':
+            return True
+        if v.lower() == 'false':
+            return False
+        raise ValueError(f'Cannot convert string {v!r} to bool')
+    return bool(v)
+
+
+def _coerce_value(value, expected_type: str):
+    """Convert a single value to the expected type.
+
+    Returns the converted value, or the original value if no conversion needed.
+    """
+    if value is None:
+        return value
+
+    if expected_type == 'boolean':
+        if isinstance(value, bool):
+            return value
+        return _coerce_bool(value)
+
+    coerce_fn = _COERCE_MAP.get(expected_type)
+    if coerce_fn is None:
+        return value
+
+    # Already the correct type
+    if expected_type == 'integer' and isinstance(value, int) and not isinstance(value, bool):
+        return value
+    if expected_type in ('number', 'float') and isinstance(value, (int, float)) and not isinstance(value, bool):
+        return float(value)
+
+    return coerce_fn(value)
+
+
+def coerce_pipeline_config(
+    config: dict,
+    *metadata_list: dict,
+) -> None:
+    """Coerce pipeline config values according to metadata type definitions.
+
+    Walks each metadata dict (trigger, safety, ai, output) and converts
+    config values in-place so that strings coming from the JSON column are
+    cast to their declared types (integer, number/float, boolean).
+
+    Args:
+        config: The pipeline config dict to modify in-place.
+        *metadata_list: Metadata dicts loaded from the YAML templates.
+    """
+    for meta in metadata_list:
+        section_name = meta.get('name')
+        if not section_name or section_name not in config:
+            continue
+
+        section = config[section_name]
+        if not isinstance(section, dict):
+            continue
+
+        for stage_def in meta.get('stages', []):
+            stage_name = stage_def.get('name')
+            if not stage_name or stage_name not in section:
+                continue
+
+            stage_config = section[stage_name]
+            if not isinstance(stage_config, dict):
+                continue
+
+            for field_def in stage_def.get('config', []):
+                field_name = field_def.get('name')
+                field_type = field_def.get('type')
+                if not field_name or not field_type or field_name not in stage_config:
+                    continue
+
+                old_value = stage_config[field_name]
+                try:
+                    new_value = _coerce_value(old_value, field_type)
+                    if new_value is not old_value:
+                        stage_config[field_name] = new_value
+                except (ValueError, TypeError) as e:
+                    logger.warning(
+                        'Failed to coerce config %s.%s.%s (%r) to %s: %s',
+                        section_name,
+                        stage_name,
+                        field_name,
+                        old_value,
+                        field_type,
+                        e,
+                    )
--- a/src/langbot/pkg/pipeline/monitoring_helper.py
+++ b/src/langbot/pkg/pipeline/monitoring_helper.py
@@ -34,6 +34,15 @@ class MonitoringHelper:
            # Check if session exists, if not, record session start
            session_id = f'{query.launcher_type}_{query.launcher_id}'

+            # Get sender name from message event
+            sender_name = None
+            if hasattr(query, 'message_event'):
+                if hasattr(query.message_event, 'sender'):
+                    if hasattr(query.message_event.sender, 'nickname'):
+                        sender_name = query.message_event.sender.nickname
+                    elif hasattr(query.message_event.sender, 'member_name'):
+                        sender_name = query.message_event.sender.member_name
+
            # Try to record message
            # Use JSON serialization to preserve message chain structure (including image URLs, etc.)
            if hasattr(query, 'message_chain') and hasattr(query.message_chain, 'model_dump'):
@@ -57,6 +66,7 @@ class MonitoringHelper:
                if hasattr(query.launcher_type, 'value')
                else str(query.launcher_type),
                user_id=query.sender_id,
+                user_name=sender_name,
                runner_name=runner_name,
                variables=None,  # Will be updated in record_query_success
            )
@@ -80,6 +90,7 @@ class MonitoringHelper:
                    if hasattr(query.launcher_type, 'value')
                    else str(query.launcher_type),
                    user_id=query.sender_id,
+                    user_name=sender_name,
                )

            return message_id
@@ -114,6 +125,70 @@ class MonitoringHelper:
        except Exception as e:
            ap.logger.error(f'Failed to record query success: {e}')

+    @staticmethod
+    async def record_query_response(
+        ap: app.Application,
+        query: pipeline_query.Query,
+        bot_id: str,
+        bot_name: str,
+        pipeline_id: str,
+        pipeline_name: str,
+        runner_name: str | None = None,
+    ):
+        """Record bot response message to monitoring"""
+        try:
+            session_id = f'{query.launcher_type}_{query.launcher_id}'
+
+            # Get sender name from message event
+            sender_name = None
+            if hasattr(query, 'message_event'):
+                if hasattr(query.message_event, 'sender'):
+                    if hasattr(query.message_event.sender, 'nickname'):
+                        sender_name = query.message_event.sender.nickname
+                    elif hasattr(query.message_event.sender, 'member_name'):
+                        sender_name = query.message_event.sender.member_name
+
+            # Extract response content from resp_message_chain
+            if hasattr(query, 'resp_message_chain') and query.resp_message_chain:
+                # Serialize the last response message chain
+                last_resp = query.resp_message_chain[-1]
+                if hasattr(last_resp, 'model_dump'):
+                    message_content = json.dumps(last_resp.model_dump(), ensure_ascii=False)
+                else:
+                    message_content = str(last_resp)
+            elif hasattr(query, 'resp_messages') and query.resp_messages:
+                last_resp = query.resp_messages[-1]
+                if hasattr(last_resp, 'get_content_platform_message_chain'):
+                    chain = last_resp.get_content_platform_message_chain()
+                    if hasattr(chain, 'model_dump'):
+                        message_content = json.dumps(chain.model_dump(), ensure_ascii=False)
+                    else:
+                        message_content = str(chain)
+                else:
+                    message_content = str(last_resp)
+            else:
+                return  # No response to record
+
+            await ap.monitoring_service.record_message(
+                bot_id=bot_id,
+                bot_name=bot_name,
+                pipeline_id=pipeline_id,
+                pipeline_name=pipeline_name,
+                message_content=message_content,
+                session_id=session_id,
+                status='success',
+                level='info',
+                platform=query.launcher_type.value
+                if hasattr(query.launcher_type, 'value')
+                else str(query.launcher_type),
+                user_id=query.sender_id,
+                user_name=sender_name,
+                runner_name=runner_name,
+                role='assistant',
+            )
+        except Exception as e:
+            ap.logger.error(f'Failed to record query response: {e}')
+
    @staticmethod
    async def record_query_error(
        ap: app.Application,
@@ -129,6 +204,15 @@ class MonitoringHelper:
        try:
            session_id = f'{query.launcher_type}_{query.launcher_id}'

+            # Get sender name from message event
+            sender_name = None
+            if hasattr(query, 'message_event'):
+                if hasattr(query.message_event, 'sender'):
+                    if hasattr(query.message_event.sender, 'nickname'):
+                        sender_name = query.message_event.sender.nickname
+                    elif hasattr(query.message_event.sender, 'member_name'):
+                        sender_name = query.message_event.sender.member_name
+
            # Record error message
            message_id = await ap.monitoring_service.record_message(
                bot_id=bot_id,
@@ -143,6 +227,7 @@ class MonitoringHelper:
                if hasattr(query.launcher_type, 'value')
                else str(query.launcher_type),
                user_id=query.sender_id,
+                user_name=sender_name,
                runner_name=runner_name,
            )

--- a/src/langbot/pkg/pipeline/pipelinemgr.py
+++ b/src/langbot/pkg/pipeline/pipelinemgr.py
@@ -13,6 +13,7 @@ import langbot_plugin.api.entities.builtin.platform.message as platform_message
 import langbot_plugin.api.entities.builtin.platform.events as platform_events
 import langbot_plugin.api.entities.events as events
 from ..utils import importutil
+from .config_coercion import coerce_pipeline_config

 import langbot_plugin.api.entities.builtin.provider.session as provider_session
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
@@ -339,6 +340,20 @@ class RuntimePipeline:
                except Exception as e:
                    self.ap.logger.error(f'Failed to record query success: {e}')

+                # Record bot response message
+                try:
+                    await monitoring_helper.MonitoringHelper.record_query_response(
+                        ap=self.ap,
+                        query=query,
+                        bot_id=query.bot_uuid or 'unknown',
+                        bot_name=bot_name,
+                        pipeline_id=self.pipeline_entity.uuid,
+                        pipeline_name=pipeline_name,
+                        runner_name=runner_name,
+                    )
+                except Exception as e:
+                    self.ap.logger.error(f'Failed to record query response: {e}')
+
        except Exception as e:
            inst_name = query.current_stage_name if query.current_stage_name else 'unknown'
            self.ap.logger.error(f'Error processing query {query.query_id} stage={inst_name} : {e}')
@@ -369,8 +384,6 @@ class RuntimePipeline:
 class PipelineManager:
    """流水线管理器"""

-    # ====== 4.0 ======
-
    ap: app.Application

    pipelines: list[RuntimePipeline]
@@ -408,6 +421,14 @@ class PipelineManager:
        elif isinstance(pipeline_entity, dict):
            pipeline_entity = persistence_pipeline.LegacyPipeline(**pipeline_entity)

+        coerce_pipeline_config(
+            pipeline_entity.config,
+            getattr(self.ap, 'pipeline_config_meta_trigger', {'name': 'trigger', 'stages': []}),
+            getattr(self.ap, 'pipeline_config_meta_safety', {'name': 'safety', 'stages': []}),
+            getattr(self.ap, 'pipeline_config_meta_ai', {'name': 'ai', 'stages': []}),
+            getattr(self.ap, 'pipeline_config_meta_output', {'name': 'output', 'stages': []}),
+        )
+
        # initialize stage containers according to pipeline_entity.stages
        stage_containers: list[StageInstContainer] = []
        for stage_name in pipeline_entity.stages:
--- a/src/langbot/pkg/pipeline/preproc/preproc.py
+++ b/src/langbot/pkg/pipeline/preproc/preproc.py
@@ -36,17 +36,36 @@ class PreProcessor(stage.PipelineStage):
        session = await self.ap.sess_mgr.get_session(query)

        # When not local-agent, llm_model is None
-        try:
-            llm_model = (
-                await self.ap.model_mgr.get_model_by_uuid(query.pipeline_config['ai']['local-agent']['model'])
-                if selected_runner == 'local-agent'
-                else None
-            )
-        except ValueError:
-            self.ap.logger.warning(
-                f'LLM model {query.pipeline_config["ai"]["local-agent"]["model"] + " "}not found or not configured'
-            )
-            llm_model = None
+        llm_model = None
+        if selected_runner == 'local-agent':
+            # Read model config — new format is { primary: str, fallbacks: [str] },
+            # but handle legacy plain string for backward compatibility
+            model_config = query.pipeline_config['ai']['local-agent'].get('model', {})
+            if isinstance(model_config, str):
+                # Legacy format: plain UUID string
+                primary_uuid = model_config
+                fallback_uuids = []
+            else:
+                primary_uuid = model_config.get('primary', '')
+                fallback_uuids = model_config.get('fallbacks', [])
+
+            if primary_uuid:
+                try:
+                    llm_model = await self.ap.model_mgr.get_model_by_uuid(primary_uuid)
+                except ValueError:
+                    self.ap.logger.warning(f'LLM model {primary_uuid} not found or not configured')
+
+            # Resolve fallback model UUIDs
+            if fallback_uuids:
+                valid_fallbacks = []
+                for fb_uuid in fallback_uuids:
+                    try:
+                        await self.ap.model_mgr.get_model_by_uuid(fb_uuid)
+                        valid_fallbacks.append(fb_uuid)
+                    except ValueError:
+                        self.ap.logger.warning(f'Fallback model {fb_uuid} not found, skipping')
+                if valid_fallbacks:
+                    query.variables['_fallback_model_uuids'] = valid_fallbacks

        conversation = await self.ap.sess_mgr.get_conversation(
            query,
@@ -61,20 +80,28 @@ class PreProcessor(stage.PipelineStage):
        query.prompt = conversation.prompt.copy()
        query.messages = conversation.messages.copy()

-        if selected_runner == 'local-agent' and llm_model:
+        if selected_runner == 'local-agent':
            query.use_funcs = []
-            query.use_llm_model_uuid = llm_model.model_entity.uuid
+            if llm_model:
+                query.use_llm_model_uuid = llm_model.model_entity.uuid

-            if llm_model.model_entity.abilities.__contains__('func_call'):
-                # Get bound plugins and MCP servers for filtering tools
+                if llm_model.model_entity.abilities.__contains__('func_call'):
+                    # Get bound plugins and MCP servers for filtering tools
+                    bound_plugins = query.variables.get('_pipeline_bound_plugins', None)
+                    bound_mcp_servers = query.variables.get('_pipeline_bound_mcp_servers', None)
+                    query.use_funcs = await self.ap.tool_mgr.get_all_tools(bound_plugins, bound_mcp_servers)
+
+                    self.ap.logger.debug(f'Bound plugins: {bound_plugins}')
+                    self.ap.logger.debug(f'Bound MCP servers: {bound_mcp_servers}')
+                    self.ap.logger.debug(f'Use funcs: {query.use_funcs}')
+
+            # If primary model doesn't support func_call but fallback models exist,
+            # load tools anyway since fallback models may support them
+            if not query.use_funcs and query.variables.get('_fallback_model_uuids'):
                bound_plugins = query.variables.get('_pipeline_bound_plugins', None)
                bound_mcp_servers = query.variables.get('_pipeline_bound_mcp_servers', None)
                query.use_funcs = await self.ap.tool_mgr.get_all_tools(bound_plugins, bound_mcp_servers)

-                self.ap.logger.debug(f'Bound plugins: {bound_plugins}')
-                self.ap.logger.debug(f'Bound MCP servers: {bound_mcp_servers}')
-                self.ap.logger.debug(f'Use funcs: {query.use_funcs}')
-
        sender_name = ''

        if isinstance(query.message_event, platform_events.GroupMessage):
@@ -149,6 +176,16 @@ class PreProcessor(stage.PipelineStage):
        query.variables['user_message_text'] = plain_text

        query.user_message = provider_message.Message(role='user', content=content_list)
+
+        # Extract knowledge base UUIDs into query variables so plugins can modify them
+        # during PromptPreProcessing before the runner performs retrieval.
+        kb_uuids = query.pipeline_config['ai']['local-agent'].get('knowledge-bases', [])
+        if not kb_uuids:
+            old_kb_uuid = query.pipeline_config['ai']['local-agent'].get('knowledge-base', '')
+            if old_kb_uuid and old_kb_uuid != '__none__':
+                kb_uuids = [old_kb_uuid]
+        query.variables['_knowledge_base_uuids'] = list(kb_uuids)
+
        # =========== 触发事件 PromptPreProcessing

        event = events.PromptPreProcessing(
--- a/src/langbot/pkg/pipeline/process/handlers/chat.py
+++ b/src/langbot/pkg/pipeline/process/handlers/chat.py
@@ -12,7 +12,7 @@ from ... import entities
 from ....provider import runner as runner_module

 import langbot_plugin.api.entities.events as events
-from ....utils import importutil, constants
+from ....utils import importutil, constants, runner as runner_utils
 from ....provider import runners
 import langbot_plugin.api.entities.builtin.provider.session as provider_session
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
@@ -149,12 +149,19 @@ class ChatMessageHandler(handler.MessageHandler):
                self.ap.logger.error(f'Conversation({query.query_id}) Request Failed: {error_info}')
                traceback.print_exc()

-                hide_exception_info = query.pipeline_config['output']['misc']['hide-exception']
+                exception_handling = query.pipeline_config['output']['misc'].get('exception-handling', 'show-hint')
+
+                if exception_handling == 'show-error':
+                    user_notice = f'{e}'
+                elif exception_handling == 'show-hint':
+                    user_notice = query.pipeline_config['output']['misc'].get('failure-hint', 'Request failed.')
+                else:  # hide
+                    user_notice = None

                yield entities.StageProcessResult(
                    result_type=entities.ResultType.INTERRUPT,
                    new_query=query,
-                    user_notice='请求失败' if hide_exception_info else f'{e}',
+                    user_notice=user_notice,
                    error_notice=f'{e}',
                    debug_notice=traceback.format_exc(),
                )
@@ -185,10 +192,15 @@ class ChatMessageHandler(handler.MessageHandler):

                    pipeline_plugins = query.variables.get('_pipeline_bound_plugins', None)

+                    runner_category = runner_utils.get_runner_category_from_runner(
+                        runner_name, runner, query.pipeline_config
+                    )
+
                    payload = {
                        'query_id': query.query_id,
                        'adapter': adapter_name,
                        'runner': runner_name,
+                        'runner_category': runner_category,
                        'duration_ms': duration_ms,
                        'model_name': model_name,
                        'version': constants.semantic_version,
@@ -200,6 +212,11 @@ class ChatMessageHandler(handler.MessageHandler):

                    # Send telemetry asynchronously and do not block pipeline via app's telemetry manager
                    await self.ap.telemetry.start_send_task(payload)
+
+                    # Trigger survey event on first successful non-WebSocket response
+                    if not locals().get('error_info') and adapter_name and 'WebSocket' not in adapter_name:
+                        if self.ap.survey:
+                            await self.ap.survey.trigger_event('first_bot_response_success')
                except Exception as ex:
                    # Ensure telemetry issues do not affect normal flow
                    self.ap.logger.warning(f'Failed to send telemetry: {ex}')
--- a/src/langbot/pkg/platform/botmgr.py
+++ b/src/langbot/pkg/platform/botmgr.py
@@ -82,7 +82,7 @@ class RuntimeBot:
                    if custom_launcher_id:
                        launcher_id = custom_launcher_id

-                await self.ap.query_pool.add_query(
+                await self.ap.msg_aggregator.add_message(
                    bot_uuid=self.bot_entity.uuid,
                    launcher_type=provider_session.LauncherTypes.PERSON,
                    launcher_id=launcher_id,
@@ -125,7 +125,7 @@ class RuntimeBot:
                    if custom_launcher_id:
                        launcher_id = custom_launcher_id

-                await self.ap.query_pool.add_query(
+                await self.ap.msg_aggregator.add_message(
                    bot_uuid=self.bot_entity.uuid,
                    launcher_type=provider_session.LauncherTypes.GROUP,
                    launcher_id=launcher_id,
@@ -282,6 +282,8 @@ class PlatformManager:
        return runtime_bot

    async def get_bot_by_uuid(self, bot_uuid: str) -> RuntimeBot | None:
+        if self.websocket_proxy_bot and self.websocket_proxy_bot.bot_entity.uuid == bot_uuid:
+            return self.websocket_proxy_bot
        for bot in self.bots:
            if bot.bot_entity.uuid == bot_uuid:
                return bot
--- a/src/langbot/pkg/platform/sources/aiocqhttp.py
+++ b/src/langbot/pkg/platform/sources/aiocqhttp.py
@@ -375,6 +375,18 @@ class AiocqhttpAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter)
            self.bot = aiocqhttp.CQHttp()

    async def send_message(self, target_type: str, target_id: str, message: platform_message.MessageChain):
+        # Check if message contains a Forward component
+        forward_msg = message.get_first(platform_message.Forward)
+        if forward_msg:
+            if target_type == 'group':
+                # Send as merged forward message via OneBot API
+                await self._send_forward_message(int(target_id), forward_msg)
+                return
+            else:
+                await self.logger.warning(
+                    f'Forward message is only supported for group targets, got target_type={target_type}. Falling through to normal send.'
+                )
+
        aiocq_msg = (await AiocqhttpMessageConverter.yiri2target(message))[0]

        if target_type == 'group':
@@ -382,6 +394,90 @@ class AiocqhttpAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter)
        elif target_type == 'person':
            await self.bot.send_private_msg(user_id=int(target_id), message=aiocq_msg)

+    async def _send_forward_message(self, group_id: int, forward: platform_message.Forward):
+        """Send a merged forward message to a group using NapCat extended API."""
+        messages = []
+
+        for node in forward.node_list:
+            # Build content for each node
+            content = []
+            if node.message_chain:
+                for component in node.message_chain:
+                    if isinstance(component, platform_message.Plain):
+                        if component.text:
+                            content.append({'type': 'text', 'data': {'text': component.text}})
+                    elif isinstance(component, platform_message.Image):
+                        img_data = {}
+                        if component.base64:
+                            b64 = component.base64
+                            if b64.startswith('data:'):
+                                b64 = b64.split(',', 1)[-1] if ',' in b64 else b64
+                            img_data['file'] = f'base64://{b64}'
+                        elif component.url:
+                            img_data['file'] = component.url
+                        elif component.path:
+                            img_data['file'] = str(component.path)
+
+                        if img_data:
+                            content.append({'type': 'image', 'data': img_data})
+
+            if not content:
+                continue
+
+            # Build node data - use user_id and nickname format for NapCat
+            user_id = str(node.sender_id) if node.sender_id else str(self.bot_account_id or '10000')
+            node_data = {
+                'type': 'node',
+                'data': {
+                    'user_id': user_id,
+                    'nickname': node.sender_name or '未知',
+                    'content': content,
+                },
+            }
+
+            messages.append(node_data)
+
+        if not messages:
+            return
+
+        # Build the full message payload for NapCat's send_forward_msg API
+        # This matches the format used by GiveMeSetuPlugin
+        bot_id = str(self.bot_account_id) if self.bot_account_id else '10000'
+        payload = {
+            'group_id': group_id,
+            'user_id': bot_id,  # Required by NapCat for display
+            'messages': messages,
+        }
+
+        # Add display settings if available
+        if forward.display:
+            if forward.display.title:
+                payload['news'] = [{'text': forward.display.title}]
+            if forward.display.brief:
+                payload['prompt'] = forward.display.brief
+            if forward.display.summary:
+                payload['summary'] = forward.display.summary
+            if forward.display.source:
+                payload['source'] = forward.display.source
+
+        try:
+            # Use send_forward_msg (NapCat extended API) instead of send_group_forward_msg
+            await self.logger.info(
+                f'Sending forward message to group {group_id} with {len(messages)} nodes, payload keys: {list(payload.keys())}'
+            )
+            result = await self.bot.call_action('send_forward_msg', **payload)
+            await self.logger.info(f'Forward message sent to group {group_id}, result: {result}')
+        except Exception as e:
+            await self.logger.error(f'Failed to send forward message to group {group_id}: {e}')
+            # Fallback: try standard OneBot API with integer group_id
+            try:
+                await self.logger.info('Trying fallback API send_group_forward_msg')
+                await self.bot.call_action('send_group_forward_msg', group_id=group_id, messages=messages)
+                await self.logger.info(f'Forward message sent via fallback API to group {group_id}')
+            except Exception as e2:
+                await self.logger.error(f'Fallback also failed: {e2}')
+                raise
+
    async def reply_message(
        self,
        message_source: platform_events.MessageEvent,
--- a/src/langbot/pkg/platform/sources/discord.py
+++ b/src/langbot/pkg/platform/sources/discord.py
@@ -14,7 +14,7 @@ import io
 import asyncio
 from enum import Enum

-import aiohttp
+from langbot.pkg.utils import httpclient
 import pydantic

 import langbot_plugin.api.definition.abstract.platform.adapter as abstract_platform_adapter
@@ -622,23 +622,23 @@ class DiscordMessageConverter(abstract_platform_adapter.AbstractMessageConverter
                    image_bytes = base64.b64decode(base64_data)
                elif ele.url:
                    # 从URL下载图片
-                    async with aiohttp.ClientSession() as session:
-                        async with session.get(ele.url) as response:
-                            image_bytes = await response.read()
-                            # 从URL或Content-Type推断文件类型
-                            content_type = response.headers.get('Content-Type', '')
-                            if 'jpeg' in content_type or 'jpg' in content_type:
-                                filename = f'{uuid.uuid4()}.jpg'
-                            elif 'gif' in content_type:
-                                filename = f'{uuid.uuid4()}.gif'
-                            elif 'webp' in content_type:
-                                filename = f'{uuid.uuid4()}.webp'
-                            elif ele.url.lower().endswith(('.jpg', '.jpeg')):
-                                filename = f'{uuid.uuid4()}.jpg'
-                            elif ele.url.lower().endswith('.gif'):
-                                filename = f'{uuid.uuid4()}.gif'
-                            elif ele.url.lower().endswith('.webp'):
-                                filename = f'{uuid.uuid4()}.webp'
+                    session = httpclient.get_session()
+                    async with session.get(ele.url) as response:
+                        image_bytes = await response.read()
+                        # 从URL或Content-Type推断文件类型
+                        content_type = response.headers.get('Content-Type', '')
+                        if 'jpeg' in content_type or 'jpg' in content_type:
+                            filename = f'{uuid.uuid4()}.jpg'
+                        elif 'gif' in content_type:
+                            filename = f'{uuid.uuid4()}.gif'
+                        elif 'webp' in content_type:
+                            filename = f'{uuid.uuid4()}.webp'
+                        elif ele.url.lower().endswith(('.jpg', '.jpeg')):
+                            filename = f'{uuid.uuid4()}.jpg'
+                        elif ele.url.lower().endswith('.gif'):
+                            filename = f'{uuid.uuid4()}.gif'
+                        elif ele.url.lower().endswith('.webp'):
+                            filename = f'{uuid.uuid4()}.webp'
                elif ele.path:
                    # 从文件路径读取图片
                    # 确保路径没有空字节
@@ -702,9 +702,9 @@ class DiscordMessageConverter(abstract_platform_adapter.AbstractMessageConverter
                    file_base64 = ele.base64.split(',')[-1]
                    file_bytes = base64.b64decode(file_base64)
                elif ele.url:
-                    async with aiohttp.ClientSession() as session:
-                        async with session.get(ele.url) as response:
-                            file_bytes = await response.read()
+                    session = httpclient.get_session()
+                    async with session.get(ele.url) as response:
+                        file_bytes = await response.read()
                if file_bytes:
                    files.append(discord.File(fp=io.BytesIO(file_bytes), filename=filename))
            elif isinstance(ele, platform_message.File):
@@ -717,9 +717,9 @@ class DiscordMessageConverter(abstract_platform_adapter.AbstractMessageConverter
                    else:
                        file_bytes = base64.b64decode(ele.base64)
                elif ele.url:
-                    async with aiohttp.ClientSession() as session:
-                        async with session.get(ele.url) as response:
-                            file_bytes = await response.read()
+                    session = httpclient.get_session()
+                    async with session.get(ele.url) as response:
+                        file_bytes = await response.read()
                if file_bytes:
                    files.append(discord.File(fp=io.BytesIO(file_bytes), filename=filename))
            elif isinstance(ele, platform_message.Forward):
@@ -775,12 +775,12 @@ class DiscordMessageConverter(abstract_platform_adapter.AbstractMessageConverter

        # attachments
        for attachment in message.attachments:
-            async with aiohttp.ClientSession(trust_env=True) as session:
-                async with session.get(attachment.url) as response:
-                    image_data = await response.read()
-                    image_base64 = base64.b64encode(image_data).decode('utf-8')
-                    image_format = response.headers['Content-Type']
-                    element_list.append(platform_message.Image(base64=f'data:{image_format};base64,{image_base64}'))
+            session = httpclient.get_session(trust_env=True)
+            async with session.get(attachment.url) as response:
+                image_data = await response.read()
+                image_base64 = base64.b64encode(image_data).decode('utf-8')
+                image_format = response.headers['Content-Type']
+                element_list.append(platform_message.Image(base64=f'data:{image_format};base64,{image_base64}'))

        return platform_message.MessageChain(element_list)

--- a/src/langbot/pkg/platform/sources/kook.py
+++ b/src/langbot/pkg/platform/sources/kook.py
@@ -9,6 +9,8 @@ import traceback
 import time

 import aiohttp
+
+from langbot.pkg.utils import httpclient
 import websockets
 import pydantic

@@ -120,16 +122,16 @@ class KookMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
            if content:
                # Download image and convert to base64
                try:
-                    async with aiohttp.ClientSession() as session:
-                        async with session.get(content) as response:
-                            if response.status == 200:
-                                image_bytes = await response.read()
-                                image_base64 = base64.b64encode(image_bytes).decode('utf-8')
-                                # Detect image format
-                                content_type = response.headers.get('Content-Type', 'image/png')
-                                components.append(
-                                    platform_message.Image(base64=f'data:{content_type};base64,{image_base64}')
-                                )
+                    session = httpclient.get_session()
+                    async with session.get(content) as response:
+                        if response.status == 200:
+                            image_bytes = await response.read()
+                            image_base64 = base64.b64encode(image_bytes).decode('utf-8')
+                            # Detect image format
+                            content_type = response.headers.get('Content-Type', 'image/png')
+                            components.append(
+                                platform_message.Image(base64=f'data:{content_type};base64,{image_base64}')
+                            )
                except Exception:
                    # If download fails, just add as plain text
                    components.append(platform_message.Plain(text=f'[Image: {content}]'))
@@ -295,17 +297,17 @@ class KookAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
            'Authorization': f'Bot {self.config["token"]}',
        }

-        async with aiohttp.ClientSession() as session:
-            async with session.get(base_url, params=params, headers=headers) as response:
-                if response.status == 200:
-                    data = await response.json()
-                    if data.get('code') == 0:
-                        gateway_url = data['data']['url']
-                        return gateway_url
-                    else:
-                        raise Exception(f'Failed to get gateway URL: {data.get("message")}')
+        session = httpclient.get_session()
+        async with session.get(base_url, params=params, headers=headers) as response:
+            if response.status == 200:
+                data = await response.json()
+                if data.get('code') == 0:
+                    gateway_url = data['data']['url']
+                    return gateway_url
                else:
-                    raise Exception(f'Failed to get gateway URL: HTTP {response.status}')
+                    raise Exception(f'Failed to get gateway URL: {data.get("message")}')
+            else:
+                raise Exception(f'Failed to get gateway URL: HTTP {response.status}')

    async def _get_bot_user_info(self) -> dict:
        """Get bot's own user information from KOOK API"""
@@ -315,17 +317,17 @@ class KookAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
            'Authorization': f'Bot {self.config["token"]}',
        }

-        async with aiohttp.ClientSession() as session:
-            async with session.get(base_url, headers=headers) as response:
-                if response.status == 200:
-                    data = await response.json()
-                    if data.get('code') == 0:
-                        user_info = data['data']
-                        return user_info
-                    else:
-                        raise Exception(f'Failed to get bot user info: {data.get("message")}')
+        session = httpclient.get_session()
+        async with session.get(base_url, headers=headers) as response:
+            if response.status == 200:
+                data = await response.json()
+                if data.get('code') == 0:
+                    user_info = data['data']
+                    return user_info
                else:
-                    raise Exception(f'Failed to get bot user info: HTTP {response.status}')
+                    raise Exception(f'Failed to get bot user info: {data.get("message")}')
+            else:
+                raise Exception(f'Failed to get bot user info: HTTP {response.status}')

    async def _handle_hello(self, data: dict):
        """Handle HELLO signal (signal 1)"""
@@ -510,7 +512,7 @@ class KookAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):

        try:
            if not self.http_session:
-                self.http_session = aiohttp.ClientSession()
+                self.http_session = httpclient.get_session()

            async with self.http_session.post(url, json=payload, headers=headers) as response:
                if response.status == 200:
@@ -576,7 +578,7 @@ class KookAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):

        try:
            if not self.http_session:
-                self.http_session = aiohttp.ClientSession()
+                self.http_session = httpclient.get_session()

            async with self.http_session.post(url, json=payload, headers=headers) as response:
                if response.status == 200:
@@ -624,7 +626,7 @@ class KookAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):

        try:
            # Create HTTP session
-            self.http_session = aiohttp.ClientSession()
+            self.http_session = httpclient.get_session()

            await self.logger.info('Starting KOOK adapter')

--- a/src/langbot/pkg/platform/sources/lark.py
+++ b/src/langbot/pkg/platform/sources/lark.py
@@ -1,7 +1,7 @@
 from __future__ import annotations

 import lark_oapi
-from lark_oapi.api.im.v1 import CreateImageRequest, CreateImageRequestBody
+from lark_oapi.api.im.v1 import CreateImageRequest, CreateImageRequestBody, CreateFileRequest, CreateFileRequestBody
 import traceback
 import typing
 import asyncio
@@ -17,7 +17,7 @@ import tempfile
 import os
 import mimetypes

-import aiohttp
+from langbot.pkg.utils import httpclient
 import lark_oapi.ws.exception
 import quart
 from lark_oapi.api.im.v1 import *
@@ -78,13 +78,13 @@ class LarkMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
                return None
        elif msg.url:
            try:
-                async with aiohttp.ClientSession() as session:
-                    async with session.get(msg.url) as response:
-                        if response.status == 200:
-                            image_bytes = await response.read()
-                        else:
-                            print(f'Failed to download image from {msg.url}: HTTP {response.status}')
-                            return None
+                session = httpclient.get_session()
+                async with session.get(msg.url) as response:
+                    if response.status == 200:
+                        image_bytes = await response.read()
+                    else:
+                        print(f'Failed to download image from {msg.url}: HTTP {response.status}')
+                        return None
            except Exception as e:
                print(f'Failed to download image from {msg.url}: {e}')
                traceback.print_exc()
@@ -141,6 +141,88 @@ class LarkMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
            traceback.print_exc()
            return None

+    @staticmethod
+    async def upload_file_to_lark(
+        file_bytes: bytes,
+        api_client: lark_oapi.Client,
+        file_type: str,
+        file_name: str = 'file',
+        duration: typing.Optional[int] = None,
+    ) -> typing.Optional[str]:
+        """Upload a file to Lark and return the file_key, or None if upload fails.
+
+        Args:
+            file_bytes: Raw file bytes.
+            api_client: Lark API client.
+            file_type: Lark file type, e.g. 'opus', 'mp4', 'pdf', 'doc', etc.
+            file_name: Display name for the file.
+            duration: Duration in milliseconds (for audio files).
+        """
+        try:
+            with tempfile.NamedTemporaryFile(delete=False) as temp_file:
+                temp_file.write(file_bytes)
+                temp_file_path = temp_file.name
+
+            try:
+                body_builder = (
+                    CreateFileRequestBody.builder()
+                    .file_type(file_type)
+                    .file_name(file_name)
+                    .file(open(temp_file_path, 'rb'))
+                )
+                if duration is not None:
+                    body_builder = body_builder.duration(duration)
+
+                request = CreateFileRequest.builder().request_body(body_builder.build()).build()
+
+                response = await api_client.im.v1.file.acreate(request)
+
+                if not response.success():
+                    print(
+                        f'client.im.v1.file.create failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}'
+                    )
+                    return None
+
+                return response.data.file_key
+            finally:
+                os.unlink(temp_file_path)
+        except Exception as e:
+            print(f'Failed to upload file to Lark: {e}')
+            traceback.print_exc()
+            return None
+
+    @staticmethod
+    async def _get_media_bytes(
+        msg: typing.Union[platform_message.Voice, platform_message.File],
+    ) -> typing.Optional[bytes]:
+        """Get bytes from a Voice or File message (base64, url, or path)."""
+        data = None
+
+        if msg.base64:
+            try:
+                base64_str = msg.base64
+                if ',' in base64_str:
+                    base64_str = base64_str.split(',', 1)[1]
+                data = base64.b64decode(base64_str)
+            except Exception:
+                pass
+        elif msg.url:
+            try:
+                session = httpclient.get_session()
+                async with session.get(msg.url) as resp:
+                    if resp.status == 200:
+                        data = await resp.read()
+            except Exception:
+                pass
+        elif msg.path:
+            try:
+                with open(msg.path, 'rb') as f:
+                    data = f.read()
+            except Exception:
+                pass
+
+        return data
+
    @staticmethod
    async def yiri2target(
        message_chain: platform_message.MessageChain, api_client: lark_oapi.Client
@@ -150,10 +232,10 @@ class LarkMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
        Returns:
            Tuple of (text_elements, image_keys):
            - text_elements: List of paragraphs for post message format
-            - image_keys: List of image_key strings for separate image messages
+            - media_items: List of dicts with 'msg_type' and 'content' for separate media messages
        """
        message_elements = []
-        image_keys = []
+        media_items = []
        pending_paragraph = []

        # Regex pattern to match Markdown image syntax: ![alt](url)
@@ -196,40 +278,77 @@ class LarkMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
                # Check for and extract Markdown images from text
                cleaned_text, extracted_urls = await process_text_with_images(text)

-                # Add cleaned text if not empty
+                # Split by blank lines to create separate paragraphs for Lark post format.
+                # Lark truncates md elements at the first \n\n, so we must use the
+                # post format's native paragraph structure instead.
                if cleaned_text:
-                    pending_paragraph.append({'tag': 'md', 'text': cleaned_text})
+                    segments = re.split(r'\n\s*\n', cleaned_text)
+                    for i, segment in enumerate(segments):
+                        segment = segment.strip()
+                        if not segment:
+                            continue
+                        if i > 0 and pending_paragraph:
+                            message_elements.append(pending_paragraph)
+                            pending_paragraph = []
+                        pending_paragraph.append({'tag': 'md', 'text': segment})

                # Process extracted image URLs
                for url in extracted_urls:
-                    # Create a temporary Image message to upload
                    temp_image = platform_message.Image(url=url)
                    image_key = await LarkMessageConverter.upload_image_to_lark(temp_image, api_client)
                    if image_key:
-                        image_keys.append(image_key)
+                        media_items.append({'msg_type': 'image', 'content': {'image_key': image_key}})

            elif isinstance(msg, platform_message.At):
                pending_paragraph.append({'tag': 'at', 'user_id': msg.target, 'style': []})
            elif isinstance(msg, platform_message.AtAll):
                pending_paragraph.append({'tag': 'at', 'user_id': 'all', 'style': []})
            elif isinstance(msg, platform_message.Image):
-                # Upload image and get image_key
                image_key = await LarkMessageConverter.upload_image_to_lark(msg, api_client)
                if image_key:
-                    # Store image_key for separate image message
-                    image_keys.append(image_key)
+                    media_items.append({'msg_type': 'image', 'content': {'image_key': image_key}})
+            elif isinstance(msg, platform_message.Voice):
+                data = await LarkMessageConverter._get_media_bytes(msg)
+                if data:
+                    duration = int(msg.length * 1000) if msg.length else None
+                    file_key = await LarkMessageConverter.upload_file_to_lark(
+                        data, api_client, file_type='opus', file_name='voice.opus', duration=duration
+                    )
+                    if file_key:
+                        media_items.append({'msg_type': 'audio', 'content': {'file_key': file_key}})
+            elif isinstance(msg, platform_message.File):
+                data = await LarkMessageConverter._get_media_bytes(msg)
+                if data:
+                    file_name = msg.name or 'file'
+                    # Guess file_type from extension
+                    ext = os.path.splitext(file_name)[1].lstrip('.').lower() if file_name else ''
+                    file_type_map = {
+                        'opus': 'opus',
+                        'mp4': 'mp4',
+                        'pdf': 'pdf',
+                        'doc': 'doc',
+                        'docx': 'doc',
+                        'xls': 'xls',
+                        'xlsx': 'xls',
+                        'ppt': 'ppt',
+                        'pptx': 'ppt',
+                    }
+                    file_type = file_type_map.get(ext, 'stream')
+                    file_key = await LarkMessageConverter.upload_file_to_lark(
+                        data, api_client, file_type=file_type, file_name=file_name
+                    )
+                    if file_key:
+                        media_items.append({'msg_type': 'file', 'content': {'file_key': file_key}})
            elif isinstance(msg, platform_message.Forward):
                for node in msg.node_list:
-                    sub_elements, sub_image_keys = await LarkMessageConverter.yiri2target(
-                        node.message_chain, api_client
-                    )
+                    sub_elements, sub_media = await LarkMessageConverter.yiri2target(node.message_chain, api_client)
                    message_elements.extend(sub_elements)
-                    image_keys.extend(sub_image_keys)
+                    media_items.extend(sub_media)

        if pending_paragraph:
            message_elements.append(pending_paragraph)

-        return message_elements, image_keys
+        return message_elements, media_items

    @staticmethod
    async def target2yiri(
@@ -456,6 +575,127 @@ class LarkMessageConverter(abstract_platform_adapter.AbstractMessageConverter):


 class LarkEventConverter(abstract_platform_adapter.AbstractEventConverter):
+    _processed_thread_quote_cache: typing.ClassVar[dict[str, float]] = {}
+    _processed_thread_quote_cache_max_size: typing.ClassVar[int] = 4096
+    _processed_thread_quote_cache_ttl_seconds: typing.ClassVar[int] = 86400
+
+    @classmethod
+    def _prune_processed_thread_quote_cache(cls, now: typing.Optional[float] = None) -> None:
+        if now is None:
+            now = time.time()
+
+        expire_before = now - cls._processed_thread_quote_cache_ttl_seconds
+        while cls._processed_thread_quote_cache:
+            oldest_key, oldest_ts = next(iter(cls._processed_thread_quote_cache.items()))
+            if oldest_ts >= expire_before:
+                break
+            cls._processed_thread_quote_cache.pop(oldest_key, None)
+
+        while len(cls._processed_thread_quote_cache) > cls._processed_thread_quote_cache_max_size:
+            oldest_key = next(iter(cls._processed_thread_quote_cache))
+            cls._processed_thread_quote_cache.pop(oldest_key, None)
+
+    @classmethod
+    def _mark_thread_quote_processed(cls, thread_id: str) -> None:
+        now = time.time()
+        cls._prune_processed_thread_quote_cache(now)
+        cls._processed_thread_quote_cache[thread_id] = now
+
+    @classmethod
+    def _extract_quote_message_id(cls, message: EventMessage) -> typing.Optional[str]:
+        """
+        Extract the message ID to quote from the given message.
+
+        Rules:
+        - First thread reply in a topic: return parent_id and mark topic as processed
+        - Follow-up thread replies in the same topic: return None
+        - Non-thread message: return parent_id if valid (non-empty, different from message_id)
+
+        Thread reply state is kept in a bounded TTL cache to avoid unbounded memory growth.
+        """
+        parent_id = getattr(message, 'parent_id', None)
+        if not parent_id:
+            return None
+
+        message_id = getattr(message, 'message_id', None)
+        if parent_id == message_id:
+            return None
+
+        thread_id = getattr(message, 'thread_id', None)
+        if thread_id:
+            cls._prune_processed_thread_quote_cache()
+            if thread_id in cls._processed_thread_quote_cache:
+                return None
+            cls._mark_thread_quote_processed(thread_id)
+
+        return parent_id
+
+    @staticmethod
+    def _build_event_message_from_message_item(message_item: Message) -> typing.Optional[EventMessage]:
+        """
+        Build EventMessage from SDK typed Message item.
+
+        Returns None if body or content is missing.
+        """
+        body = getattr(message_item, 'body', None)
+        if not body:
+            return None
+
+        content = getattr(body, 'content', None)
+        if not content:
+            return None
+
+        event_data = {
+            'message_id': message_item.message_id,
+            'message_type': message_item.msg_type,
+            'content': content,
+            'create_time': message_item.create_time,
+            'mentions': getattr(message_item, 'mentions', []) or [],
+        }
+
+        # Preserve thread-related fields
+        if hasattr(message_item, 'parent_id') and message_item.parent_id:
+            event_data['parent_id'] = message_item.parent_id
+        if hasattr(message_item, 'root_id') and message_item.root_id:
+            event_data['root_id'] = message_item.root_id
+        if hasattr(message_item, 'thread_id') and message_item.thread_id:
+            event_data['thread_id'] = message_item.thread_id
+        if hasattr(message_item, 'chat_id') and message_item.chat_id:
+            event_data['chat_id'] = message_item.chat_id
+
+        return EventMessage(event_data)
+
+    @staticmethod
+    async def _fetch_quoted_message(
+        quote_message_id: str,
+        api_client: lark_oapi.Client,
+    ) -> typing.Optional[platform_message.MessageChain]:
+        """
+        Fetch the quoted message and convert to MessageChain.
+
+        Returns None if:
+        - API call fails
+        - Response items is empty
+        - Message item normalization fails
+        """
+        request = GetMessageRequest.builder().message_id(quote_message_id).build()
+        response = await api_client.im.v1.message.aget(request)
+
+        if not response.success():
+            return None
+
+        items = getattr(response.data, 'items', None)
+        if not items:
+            return None
+
+        message_item = items[0]
+        event_message = LarkEventConverter._build_event_message_from_message_item(message_item)
+        if event_message is None:
+            return None
+
+        quote_chain = await LarkMessageConverter.target2yiri(event_message, api_client)
+        return quote_chain
+
    @staticmethod
    async def yiri2target(
        event: platform_events.MessageEvent,
@@ -468,6 +708,23 @@ class LarkEventConverter(abstract_platform_adapter.AbstractEventConverter):
    ) -> platform_events.Event:
        message_chain = await LarkMessageConverter.target2yiri(event.event.message, api_client)

+        # Check for quote/reply message
+        quote_message_id = LarkEventConverter._extract_quote_message_id(event.event.message)
+        if quote_message_id:
+            quote_chain = await LarkEventConverter._fetch_quoted_message(quote_message_id, api_client)
+            if quote_chain:
+                # Filter out Source component from quoted chain, keep only content
+                quote_origin = platform_message.MessageChain(
+                    [comp for comp in quote_chain if not isinstance(comp, platform_message.Source)]
+                )
+                if quote_origin:
+                    message_chain.append(
+                        platform_message.Quote(
+                            message_id=quote_message_id,
+                            origin=quote_origin,
+                        )
+                    )
+
        if event.event.message.chat_type == 'p2p':
            return platform_events.FriendMessage(
                sender=platform_entities.Friend(
@@ -651,6 +908,32 @@ class LarkAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
            self.request_tenant_access_token(tenant_key)
        return self.tenant_access_tokens.get(tenant_key)['token'] if self.tenant_access_tokens.get(tenant_key) else None

+    def get_launcher_id(self, event: platform_events.MessageEvent) -> str | None:
+        """
+        Get topic-scoped launcher_id for thread-aware session isolation.
+
+        For group thread messages, returns "{group_id}_{thread_id}"
+        to ensure conversation context stays stable per topic.
+
+        Returns None for non-thread messages or P2P messages.
+        """
+        source_event = getattr(event.source_platform_object, 'event', None)
+        if not source_event:
+            return None
+
+        message = getattr(source_event, 'message', None)
+        if not message:
+            return None
+
+        thread_id = getattr(message, 'thread_id', None)
+        if not thread_id:
+            return None
+
+        if isinstance(event, platform_events.GroupMessage):
+            return f'{event.group.id}_{thread_id}'
+
+        return None
+
    def build_api_client(self, config):
        app_id = config['app_id']
        app_secret = config['app_secret']
@@ -917,23 +1200,40 @@ class LarkAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
    ):
        # 不再需要了，因为message_id已经被包含到message_chain中
        # lark_event = await self.event_converter.yiri2target(message_source)
-        text_elements, image_keys = await self.message_converter.yiri2target(message, self.api_client)
+        text_elements, media_items = await self.message_converter.yiri2target(message, self.api_client)

        # Send text message if there are text elements
        if text_elements:
-            final_content = {
-                'zh_Hans': {
-                    'title': '',
-                    'content': text_elements,
-                },
-            }
+            # Determine msg_type based on content: use 'post' if at mentions
+            # are present (requires post paragraph structure), otherwise 'text'
+            needs_post = any(ele['tag'] == 'at' for paragraph in text_elements for ele in paragraph)
+
+            if needs_post:
+                msg_type = 'post'
+                final_content = json.dumps(
+                    {
+                        'zh_Hans': {
+                            'title': '',
+                            'content': text_elements,
+                        },
+                    }
+                )
+            else:
+                msg_type = 'text'
+                parts = []
+                for paragraph in text_elements:
+                    para_text = ''.join(ele.get('text', '') for ele in paragraph)
+                    if para_text:
+                        parts.append(para_text)
+                final_content = json.dumps({'text': '\n\n'.join(parts)})
+
            request: ReplyMessageRequest = (
                ReplyMessageRequest.builder()
                .message_id(message_source.message_chain.message_id)
                .request_body(
                    ReplyMessageRequestBody.builder()
-                    .content(json.dumps(final_content))
-                    .msg_type('post')
+                    .content(final_content)
+                    .msg_type(msg_type)
                    .reply_in_thread(False)
                    .uuid(str(uuid.uuid4()))
                    .build()
@@ -963,17 +1263,15 @@ class LarkAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
                    f'client.im.v1.message.reply failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}, resp: \n{json.dumps(json.loads(response.raw.content), indent=4, ensure_ascii=False)}'
                )

-        # Send image messages separately using msg_type='image'
-        for image_key in image_keys:
-            image_content = json.dumps({'image_key': image_key})
-
+        # Send media messages separately (image, audio, file, etc.)
+        for media in media_items:
            request: ReplyMessageRequest = (
                ReplyMessageRequest.builder()
                .message_id(message_source.message_chain.message_id)
                .request_body(
                    ReplyMessageRequestBody.builder()
-                    .content(image_content)
-                    .msg_type('image')
+                    .content(json.dumps(media['content']))
+                    .msg_type(media['msg_type'])
                    .reply_in_thread(False)
                    .uuid(str(uuid.uuid4()))
                    .build()
@@ -1000,7 +1298,7 @@ class LarkAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):

            if not response.success():
                raise Exception(
-                    f'client.im.v1.message.reply (image) failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}, resp: \n{json.dumps(json.loads(response.raw.content), indent=4, ensure_ascii=False)}'
+                    f'client.im.v1.message.reply ({media["msg_type"]}) failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}, resp: \n{json.dumps(json.loads(response.raw.content), indent=4, ensure_ascii=False)}'
                )

    async def reply_message_chunk(
@@ -1018,15 +1316,16 @@ class LarkAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        message_id = bot_message.resp_message_id
        msg_seq = bot_message.msg_sequence
        if msg_seq % 8 == 0 or is_final:
-            text_elements, image_keys = await self.message_converter.yiri2target(message, self.api_client)
+            text_elements, media_items = await self.message_converter.yiri2target(message, self.api_client)

            text_message = ''
            if text_elements:
-                for ele in text_elements[0]:
-                    if ele['tag'] == 'text':
-                        text_message += ele['text']
-                    elif ele['tag'] == 'md':
-                        text_message += ele['text']
+                parts = []
+                for paragraph in text_elements:
+                    para_text = ''.join(ele['text'] for ele in paragraph if ele['tag'] in ('text', 'md'))
+                    if para_text:
+                        parts.append(para_text)
+                text_message = '\n\n'.join(parts)

            # content = {
            #     'type': 'card_json',
@@ -1076,6 +1375,30 @@ class LarkAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
                )
                return

+            # Send media messages when streaming is done
+            if is_final and media_items:
+                for media in media_items:
+                    media_request: ReplyMessageRequest = (
+                        ReplyMessageRequest.builder()
+                        .message_id(message_source.message_chain.message_id)
+                        .request_body(
+                            ReplyMessageRequestBody.builder()
+                            .content(json.dumps(media['content']))
+                            .msg_type(media['msg_type'])
+                            .reply_in_thread(False)
+                            .uuid(str(uuid.uuid4()))
+                            .build()
+                        )
+                        .build()
+                    )
+                    media_response: ReplyMessageResponse = await self.api_client.im.v1.message.areply(
+                        media_request, req_opt
+                    )
+                    if not media_response.success():
+                        raise Exception(
+                            f'client.im.v1.message.reply ({media["msg_type"]}) failed, code: {media_response.code}, msg: {media_response.msg}, log_id: {media_response.get_log_id()}'
+                        )
+
    async def is_muted(self, group_id: int) -> bool:
        return False

--- a/src/langbot/pkg/platform/sources/legacy/gewechat.py
+++ b/src/langbot/pkg/platform/sources/legacy/gewechat.py
@@ -9,7 +9,7 @@ import copy
 import threading

 import quart
-import aiohttp
+from langbot.pkg.utils import httpclient

 import langbot_plugin.api.definition.abstract.platform.adapter as abstract_platform_adapter
 from ....core import app
@@ -639,14 +639,14 @@ class GeWeChatAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):

    async def run_async(self):
        if not self.config['token']:
-            async with aiohttp.ClientSession() as session:
-                async with session.post(
-                    f'{self.config["gewechat_url"]}/v2/api/tools/getTokenId',
-                    json={'app_id': self.config['app_id']},
-                ) as response:
-                    if response.status != 200:
-                        raise Exception(f'获取gewechat token失败: {await response.text()}')
-                    self.config['token'] = (await response.json())['data']
+            session = httpclient.get_session()
+            async with session.post(
+                f'{self.config["gewechat_url"]}/v2/api/tools/getTokenId',
+                json={'app_id': self.config['app_id']},
+            ) as response:
+                if response.status != 200:
+                    raise Exception(f'获取gewechat token失败: {await response.text()}')
+                self.config['token'] = (await response.json())['data']

        self.bot = gewechat_client.GewechatClient(f'{self.config["gewechat_url"]}/v2/api', self.config['token'])

--- a/src/langbot/pkg/platform/sources/openclaw_weixin.py
+++ b/src/langbot/pkg/platform/sources/openclaw_weixin.py
@@ -0,0 +1,577 @@
+"""OpenClaw WeChat adapter for LangBot.
+
+Uses the OpenClaw WeChat HTTP JSON API (long-poll getUpdates + sendMessage)
+to integrate personal WeChat accounts with LangBot.
+
+Reference: https://github.com/epiral/weixin-bot
+"""
+
+from __future__ import annotations
+
+import asyncio
+import base64
+import traceback
+import typing
+
+import pydantic
+import sqlalchemy
+
+from langbot.libs.openclaw_weixin_api.client import (
+    DEFAULT_BASE_URL,
+    SESSION_EXPIRED_ERRCODE,
+    OpenClawWeixinClient,
+)
+from langbot.libs.openclaw_weixin_api.types import (
+    MessageItem,
+    WeixinMessage,
+)
+from langbot.pkg.entity.persistence import bot as persistence_bot
+
+import langbot_plugin.api.definition.abstract.platform.adapter as abstract_platform_adapter
+import langbot_plugin.api.definition.abstract.platform.event_logger as abstract_platform_logger
+import langbot_plugin.api.entities.builtin.platform.entities as platform_entities
+import langbot_plugin.api.entities.builtin.platform.events as platform_events
+import langbot_plugin.api.entities.builtin.platform.message as platform_message
+
+
+class OpenClawWeixinMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
+    """Converts between LangBot MessageChain and OpenClaw WeChat message items."""
+
+    @staticmethod
+    async def yiri2target(message_chain: platform_message.MessageChain) -> list[dict]:
+        """Convert LangBot MessageChain to a list of OpenClaw message item dicts."""
+        items = []
+        for component in message_chain:
+            if isinstance(component, platform_message.Plain):
+                items.append({'type': MessageItem.TEXT, 'text_item': {'text': component.text}})
+            elif isinstance(component, platform_message.Image):
+                # OpenClaw WeChat only supports text messages without CDN upload.
+                # For images, we send a placeholder text with the URL if available.
+                if component.url:
+                    items.append(
+                        {
+                            'type': MessageItem.TEXT,
+                            'text_item': {'text': f'[Image: {component.url}]'},
+                        }
+                    )
+                elif component.base64:
+                    items.append(
+                        {
+                            'type': MessageItem.TEXT,
+                            'text_item': {'text': '[Image]'},
+                        }
+                    )
+            elif isinstance(component, platform_message.File):
+                if component.name:
+                    items.append(
+                        {
+                            'type': MessageItem.TEXT,
+                            'text_item': {'text': f'[File: {component.name}]'},
+                        }
+                    )
+            elif isinstance(component, platform_message.Forward):
+                for node in component.node_list:
+                    if node.message_chain:
+                        items.extend(await OpenClawWeixinMessageConverter.yiri2target(node.message_chain))
+        return items
+
+    @staticmethod
+    async def target2yiri(
+        msg: WeixinMessage,
+    ) -> platform_message.MessageChain:
+        """Convert an OpenClaw WeixinMessage to LangBot MessageChain."""
+        components: list[platform_message.MessageComponent] = []
+
+        if not msg.item_list:
+            return platform_message.MessageChain(components)
+
+        for item in msg.item_list:
+            if item.type == MessageItem.TEXT and item.text_item and item.text_item.text:
+                text = item.text_item.text
+
+                # Handle quoted messages
+                if item.ref_msg:
+                    ref_parts = []
+                    if item.ref_msg.title:
+                        ref_parts.append(item.ref_msg.title)
+                    if item.ref_msg.message_item:
+                        ref_item = item.ref_msg.message_item
+                        if ref_item.text_item and ref_item.text_item.text:
+                            ref_parts.append(ref_item.text_item.text)
+                    if ref_parts:
+                        components.append(
+                            platform_message.Quote(
+                                sender_id='',
+                                origin=platform_message.MessageChain(
+                                    [platform_message.Plain(text=' | '.join(ref_parts))]
+                                ),
+                            )
+                        )
+
+                components.append(platform_message.Plain(text=text))
+
+            elif item.type == MessageItem.IMAGE and item.image_item:
+                if hasattr(item.image_item, '_downloaded_bytes') and item.image_item._downloaded_bytes:
+                    b64 = base64.b64encode(item.image_item._downloaded_bytes).decode('utf-8')
+                    components.append(platform_message.Image(base64=f'data:image/jpeg;base64,{b64}'))
+                else:
+                    components.append(platform_message.Unknown(text='[Image]'))
+
+            elif item.type == MessageItem.VOICE and item.voice_item:
+                # Voice with speech-to-text: use the transcribed text
+                if item.voice_item.text:
+                    components.append(platform_message.Plain(text=item.voice_item.text))
+                else:
+                    components.append(platform_message.Unknown(text='[Voice]'))
+
+            # TODO: enable after full testing
+            # elif item.type == MessageItem.VOICE and item.voice_item:
+            #     if item.voice_item.text:
+            #         components.append(platform_message.Plain(text=item.voice_item.text))
+            #     elif hasattr(item.voice_item, '_downloaded_bytes') and item.voice_item._downloaded_bytes:
+            #         b64 = base64.b64encode(item.voice_item._downloaded_bytes).decode('utf-8')
+            #         components.append(
+            #             platform_message.Voice(
+            #                 base64=b64,
+            #                 length=item.voice_item.playtime or 0,
+            #             )
+            #         )
+            #     else:
+            #         components.append(
+            #             platform_message.Voice(
+            #                 length=item.voice_item.playtime or 0,
+            #             )
+            #         )
+
+            elif item.type == MessageItem.FILE and item.file_item:
+                components.append(platform_message.Unknown(text=f'[File: {item.file_item.file_name or ""}]'))
+
+            # TODO: enable after full testing
+            # elif item.type == MessageItem.FILE and item.file_item:
+            #     file_name = item.file_item.file_name or ''
+            #     file_size = int(item.file_item.len) if item.file_item.len else 0
+            #     if hasattr(item.file_item, '_downloaded_bytes') and item.file_item._downloaded_bytes:
+            #         b64 = base64.b64encode(item.file_item._downloaded_bytes).decode('utf-8')
+            #         components.append(
+            #             platform_message.File(
+            #                 name=file_name,
+            #                 size=file_size,
+            #                 base64=b64,
+            #             )
+            #         )
+            #     else:
+            #         components.append(
+            #             platform_message.File(
+            #                 name=file_name,
+            #                 size=file_size,
+            #             )
+            #         )
+
+            elif item.type == MessageItem.VIDEO and item.video_item:
+                components.append(platform_message.Unknown(text='[Video]'))
+
+            # TODO: enable after full testing
+            # elif item.type == MessageItem.VIDEO and item.video_item:
+            #     if hasattr(item.video_item, '_downloaded_bytes') and item.video_item._downloaded_bytes:
+            #         b64 = base64.b64encode(item.video_item._downloaded_bytes).decode('utf-8')
+            #         components.append(
+            #             platform_message.File(
+            #                 name='video.mp4',
+            #                 size=item.video_item.video_size or 0,
+            #                 base64=b64,
+            #             )
+            #         )
+            #     else:
+            #         components.append(
+            #             platform_message.File(
+            #                 name='video.mp4',
+            #                 size=item.video_item.video_size or 0,
+            #             )
+            #         )
+
+            else:
+                components.append(platform_message.Unknown(text='[Unknown message type]'))
+
+        return platform_message.MessageChain(components)
+
+
+class OpenClawWeixinEventConverter(abstract_platform_adapter.AbstractEventConverter):
+    """Converts OpenClaw WeChat messages to LangBot events."""
+
+    @staticmethod
+    async def yiri2target(event: platform_events.MessageEvent) -> dict:
+        return event.source_platform_object
+
+    @staticmethod
+    async def target2yiri(msg: WeixinMessage) -> typing.Optional[platform_events.MessageEvent]:
+        """Convert an inbound WeixinMessage to a LangBot event."""
+        if msg.message_type != WeixinMessage.TYPE_USER:
+            return None
+
+        from_user_id = msg.from_user_id or ''
+        if not from_user_id:
+            return None
+
+        message_chain = await OpenClawWeixinMessageConverter.target2yiri(msg)
+        if not message_chain:
+            return None
+
+        timestamp = (msg.create_time_ms or 0) / 1000.0
+
+        return platform_events.FriendMessage(
+            sender=platform_entities.Friend(
+                id=from_user_id,
+                nickname=from_user_id,
+                remark='',
+            ),
+            message_chain=message_chain,
+            time=timestamp,
+            source_platform_object=msg,
+        )
+
+
+class OpenClawWeixinAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
+    """LangBot adapter for OpenClaw WeChat (long-poll based)."""
+
+    name: str = 'openclaw-weixin'
+
+    client: OpenClawWeixinClient = pydantic.Field(exclude=True)
+
+    config: dict
+
+    message_converter: OpenClawWeixinMessageConverter = OpenClawWeixinMessageConverter()
+    event_converter: OpenClawWeixinEventConverter = OpenClawWeixinEventConverter()
+
+    # context_token cache: from_user_id -> context_token
+    _context_tokens: dict[str, str] = pydantic.PrivateAttr(default_factory=dict)
+
+    _polling: bool = pydantic.PrivateAttr(default=False)
+    _poll_task: typing.Optional[asyncio.Task] = pydantic.PrivateAttr(default=None)
+    _bot_uuid: typing.Optional[str] = pydantic.PrivateAttr(default=None)
+
+    listeners: typing.Dict[
+        typing.Type[platform_events.Event],
+        typing.Callable[[platform_events.Event, abstract_platform_adapter.AbstractMessagePlatformAdapter], None],
+    ] = {}
+
+    def __init__(self, config: dict, logger: abstract_platform_logger.AbstractEventLogger):
+        client = OpenClawWeixinClient(
+            base_url=config.get('base_url', DEFAULT_BASE_URL),
+            token=config.get('token', ''),
+        )
+        super().__init__(
+            config=config,
+            logger=logger,
+            client=client,
+            bot_account_id='',
+            listeners={},
+            name='openclaw-weixin',
+        )
+
+    def set_bot_uuid(self, bot_uuid: str):
+        """Called by BotManager to provide the bot's UUID for config persistence."""
+        self._bot_uuid = bot_uuid
+
+    async def _persist_config(self) -> None:
+        """Persist current self.config to the database so token survives restart."""
+        if not self._bot_uuid:
+            return
+        try:
+            ap = self.logger.ap
+            await ap.persistence_mgr.execute_async(
+                sqlalchemy.update(persistence_bot.Bot)
+                .where(persistence_bot.Bot.uuid == self._bot_uuid)
+                .values(adapter_config=self.config)
+            )
+        except Exception as e:
+            await self.logger.warning(f'Failed to persist adapter config: {e}')
+
+    async def _do_login(self) -> None:
+        """Run the QR code login flow via client.login() and update config."""
+        adapter_logger = self.logger
+
+        async def _on_qrcode(qr_base64: str, _qr_url: str):
+            await adapter_logger.info(
+                f'Please scan the QR code to login WeChat: {_qr_url}',
+                images=[platform_message.Image(base64=qr_base64)],
+            )
+
+        login_result = await self.client.login(
+            on_qrcode=_on_qrcode,
+        )
+
+        # client.login() already updates client.token and client.base_url
+        self.config['token'] = login_result.token
+        self.config['base_url'] = login_result.base_url
+        if login_result.account_id:
+            self.config['account_id'] = login_result.account_id
+
+        await self.logger.info(f'WeChat login successful! account_id={login_result.account_id}')
+
+        # Persist token to database so it survives restart
+        await self._persist_config()
+
+    async def send_message(
+        self,
+        target_type: str,
+        target_id: str,
+        message: platform_message.MessageChain,
+    ):
+        """Send a message to a user."""
+        context_token = self._context_tokens.get(target_id, '')
+
+        for component in message:
+            try:
+                if isinstance(component, platform_message.Plain):
+                    if component.text:
+                        await self.client.send_text(target_id, component.text, context_token)
+
+                elif isinstance(component, platform_message.Image):
+                    img_bytes, _ = await component.get_bytes()
+                    await self.client.send_image(target_id, img_bytes, context_token)
+
+                elif isinstance(component, platform_message.File):
+                    file_bytes = await self._get_component_bytes(component)
+                    if file_bytes:
+                        await self.client.send_file(target_id, file_bytes, component.name or 'file', context_token)
+
+                elif isinstance(component, platform_message.Voice):
+                    voice_bytes = await self._get_component_bytes(component)
+                    if voice_bytes:
+                        await self.client.send_voice(target_id, voice_bytes, component.length or 0, context_token)
+
+                elif isinstance(component, platform_message.Forward):
+                    for node in component.node_list:
+                        if node.message_chain:
+                            await self.send_message(target_type, target_id, node.message_chain)
+
+            except Exception:
+                await self.logger.error(
+                    f'Failed to send component {type(component).__name__}: {traceback.format_exc()}'
+                )
+
+    async def reply_message(
+        self,
+        message_source: platform_events.MessageEvent,
+        message: platform_message.MessageChain,
+        quote_origin: bool = False,
+    ):
+        """Reply to a received message."""
+        source_msg = message_source.source_platform_object
+        if isinstance(source_msg, WeixinMessage):
+            target_id = source_msg.from_user_id or ''
+            if target_id:
+                await self.send_message('friend', target_id, message)
+
+    async def is_muted(self, group_id: int) -> bool:
+        return False
+
+    @staticmethod
+    async def _get_component_bytes(component: platform_message.MessageComponent) -> typing.Optional[bytes]:
+        """Extract raw bytes from a File or Voice component."""
+        b64_val = getattr(component, 'base64', None)
+        url_val = getattr(component, 'url', None)
+        path_val = getattr(component, 'path', None)
+
+        if b64_val:
+            return base64.b64decode(b64_val)
+        elif url_val and url_val.startswith(('http://', 'https://')):
+            import aiohttp
+
+            async with aiohttp.ClientSession() as session:
+                async with session.get(url_val) as resp:
+                    if resp.status == 200:
+                        return await resp.read()
+        elif path_val:
+            import asyncio
+
+            with open(path_val, 'rb') as f:
+                return await asyncio.to_thread(f.read)
+        return None
+
+    def register_listener(
+        self,
+        event_type: typing.Type[platform_events.Event],
+        callback: typing.Callable[
+            [platform_events.Event, abstract_platform_adapter.AbstractMessagePlatformAdapter],
+            None,
+        ],
+    ):
+        self.listeners[event_type] = callback
+
+    def unregister_listener(
+        self,
+        event_type: typing.Type[platform_events.Event],
+        callback: typing.Callable[
+            [platform_events.Event, abstract_platform_adapter.AbstractMessagePlatformAdapter],
+            None,
+        ],
+    ):
+        self.listeners.pop(event_type, None)
+
+    async def run_async(self):
+        """Start the adapter. If no token is configured, trigger QR code login first."""
+        base_url = self.config.get('base_url', DEFAULT_BASE_URL)
+        token = self.config.get('token', '')
+
+        await self.logger.info('OpenClaw WeChat adapter starting...')
+
+        # QR code login flow when no token is provided
+        if not token:
+            await self.logger.info('No token configured, starting QR code login...')
+            try:
+                await self._do_login()
+            except Exception as e:
+                await self.logger.error(f'QR code login failed: {e}')
+                raise
+
+        # Rebuild client with the (possibly updated) config
+        self.client = OpenClawWeixinClient(
+            base_url=self.config.get('base_url', base_url),
+            token=self.config.get('token', token),
+        )
+        self.bot_account_id = self.config.get('account_id', 'openclaw-weixin')
+        self._polling = True
+
+        # Start the long-poll loop
+        self._poll_task = asyncio.create_task(self._poll_loop())
+        await self.logger.info('OpenClaw WeChat adapter running')
+
+        try:
+            await self._poll_task
+        except asyncio.CancelledError:
+            pass
+
+    async def _poll_loop(self):
+        """Long-poll loop: call getUpdates continuously.
+
+        Error handling follows the weixin-bot SDK pattern:
+        - Exponential backoff (1s -> 10s max) on failures
+        - Session expired (errcode -14) triggers automatic re-login
+        """
+        get_updates_buf = ''
+        poll_timeout = float(self.config.get('poll_timeout', 35))
+
+        backoff_delay = 1.0
+        max_backoff = 10.0
+
+        while self._polling:
+            try:
+                resp = await self.client.get_updates(
+                    get_updates_buf=get_updates_buf,
+                    timeout=poll_timeout + 5,
+                )
+
+                if resp.longpolling_timeout_ms and resp.longpolling_timeout_ms > 0:
+                    poll_timeout = resp.longpolling_timeout_ms / 1000.0
+
+                is_api_error = (resp.ret is not None and resp.ret != 0) or (
+                    resp.errcode is not None and resp.errcode != 0
+                )
+                if is_api_error:
+                    is_session_expired = resp.errcode == SESSION_EXPIRED_ERRCODE or resp.ret == SESSION_EXPIRED_ERRCODE
+
+                    if is_session_expired:
+                        await self.logger.error('OpenClaw WeChat session expired, attempting re-login...')
+                        try:
+                            await self._do_login()
+                            # Rebuild client with new credentials
+                            self.client = OpenClawWeixinClient(
+                                base_url=self.config.get('base_url', DEFAULT_BASE_URL),
+                                token=self.config.get('token', ''),
+                            )
+                            self._context_tokens.clear()
+                            get_updates_buf = ''
+                            backoff_delay = 1.0
+                            continue
+                        except Exception:
+                            await self.logger.error(f'Re-login failed: {traceback.format_exc()}')
+                            break
+
+                    await self.logger.error(
+                        f'OpenClaw getUpdates failed: ret={resp.ret} errcode={resp.errcode} errmsg={resp.errmsg}'
+                    )
+                    await asyncio.sleep(backoff_delay)
+                    backoff_delay = min(backoff_delay * 2, max_backoff)
+                    continue
+
+                backoff_delay = 1.0
+
+                if resp.get_updates_buf:
+                    get_updates_buf = resp.get_updates_buf
+
+                for msg in resp.msgs:
+                    try:
+                        await self._handle_inbound_message(msg)
+                    except Exception:
+                        await self.logger.error(f'Error handling message: {traceback.format_exc()}')
+
+            except asyncio.CancelledError:
+                break
+            except Exception:
+                await self.logger.error(f'OpenClaw poll error: {traceback.format_exc()}')
+                await asyncio.sleep(backoff_delay)
+                backoff_delay = min(backoff_delay * 2, max_backoff)
+
+    async def _handle_inbound_message(self, msg: WeixinMessage):
+        """Process a single inbound message from getUpdates."""
+        if msg.context_token and msg.from_user_id:
+            self._context_tokens[msg.from_user_id] = msg.context_token
+
+        # Download CDN media (files, images) before converting to LangBot events
+        await self._download_media_items(msg)
+
+        event = await OpenClawWeixinEventConverter.target2yiri(msg)
+        if event is None:
+            return
+
+        if type(event) in self.listeners:
+            await self.listeners[type(event)](event, self)
+
+    async def _download_media_items(self, msg: WeixinMessage):
+        """Download CDN media for image items in the message."""
+        if not msg.item_list:
+            return
+
+        for item in msg.item_list:
+            try:
+                if item.type == MessageItem.IMAGE and item.image_item:
+                    if (
+                        item.image_item.media
+                        and item.image_item.media.encrypt_query_param
+                        and item.image_item.media.aes_key
+                    ):
+                        img_bytes = await self.client.download_media(item.image_item.media)
+                        item.image_item._downloaded_bytes = img_bytes
+
+                # TODO: enable after full testing
+                # elif item.type == MessageItem.FILE and item.file_item and item.file_item.media:
+                #     if item.file_item.media.encrypt_query_param and item.file_item.media.aes_key:
+                #         file_bytes = await self.client.download_media(item.file_item.media)
+                #         item.file_item._downloaded_bytes = file_bytes
+                #
+                # elif item.type == MessageItem.VOICE and item.voice_item and item.voice_item.media:
+                #     if item.voice_item.media.encrypt_query_param and item.voice_item.media.aes_key:
+                #         voice_bytes = await self.client.download_media(item.voice_item.media)
+                #         item.voice_item._downloaded_bytes = voice_bytes
+                #
+                # elif item.type == MessageItem.VIDEO and item.video_item and item.video_item.media:
+                #     if item.video_item.media.encrypt_query_param and item.video_item.media.aes_key:
+                #         video_bytes = await self.client.download_media(item.video_item.media)
+                #         item.video_item._downloaded_bytes = video_bytes
+
+            except Exception:
+                await self.logger.warning(f'Failed to download CDN media: {traceback.format_exc()}')
+
+    async def kill(self) -> bool:
+        """Stop the adapter."""
+        self._polling = False
+        if self._poll_task and not self._poll_task.done():
+            self._poll_task.cancel()
+            try:
+                await self._poll_task
+            except asyncio.CancelledError:
+                pass
+        await self.client.close()
+        await self.logger.info('OpenClaw WeChat adapter stopped')
+        return True
--- a/src/langbot/pkg/platform/sources/openclaw_weixin.yaml
+++ b/src/langbot/pkg/platform/sources/openclaw_weixin.yaml
@@ -0,0 +1,57 @@
+apiVersion: v1
+kind: MessagePlatformAdapter
+metadata:
+  name: openclaw-weixin
+  label:
+    en_US: OpenClaw WeChat
+    zh_Hans: OpenClaw 微信
+  description:
+    en_US: OpenClaw WeChat adapter, supports personal WeChat via QR code login
+    zh_Hans: OpenClaw 微信适配器，通过扫码登录支持个人微信
+  icon: wechat.png
+spec:
+  config:
+    - name: base_url
+      label:
+        en_US: API Base URL
+        zh_Hans: API 基础地址
+      description:
+        en_US: The base URL of the OpenClaw WeChat backend API
+        zh_Hans: OpenClaw 微信后端 API 的基础地址
+      type: string
+      required: true
+      default: "https://ilinkai.weixin.qq.com"
+    - name: token
+      label:
+        en_US: Token
+        zh_Hans: 令牌
+      description:
+        en_US: Bearer token obtained after QR code login authorization. Leave empty to trigger QR code login on startup.
+        zh_Hans: 扫码登录授权后获取的 Bearer 令牌。留空则启动时自动触发扫码登录。
+      type: string
+      required: false
+      default: ""
+    - name: account_id
+      label:
+        en_US: Account ID
+        zh_Hans: 账号标识
+      description:
+        en_US: A label for this WeChat account (used for display purposes)
+        zh_Hans: 此微信账号的标识（用于显示）
+      type: string
+      required: false
+      default: "openclaw-weixin"
+    - name: poll_timeout
+      label:
+        en_US: Poll Timeout (seconds)
+        zh_Hans: 轮询超时（秒）
+      description:
+        en_US: Long-poll timeout for getUpdates, the server may hold the request up to this duration
+        zh_Hans: getUpdates 长轮询超时时间，服务端最多持有请求的时长
+      type: integer
+      required: false
+      default: 35
+execution:
+  python:
+    path: ./openclaw_weixin.py
+    attr: OpenClawWeixinAdapter
--- a/src/langbot/pkg/platform/sources/satori.png
+++ b/src/langbot/pkg/platform/sources/satori.png
--- a/src/langbot/pkg/platform/sources/satori.py
+++ b/src/langbot/pkg/platform/sources/satori.py
--- a/src/langbot/pkg/platform/sources/satori.yaml
+++ b/src/langbot/pkg/platform/sources/satori.yaml
@@ -0,0 +1,65 @@
+apiVersion: v1
+kind: MessagePlatformAdapter
+metadata:
+  name: satori
+  label:
+    en_US: Satori
+    zh_Hans: Satori
+  description:
+    en_US: SatoriAdapter
+    zh_Hans: 古明地觉协议适配器
+  icon: satori.png
+spec:
+  config:
+    - name: platform
+      label:
+        en_US: Platform
+        zh_Hans: 平台名称
+      type: string
+      required: true
+      default: "llonebot"
+      description:
+        en_US: The platform name (e.g., llonebot, discord, telegram)
+        zh_Hans: 平台名称（如 llonebot, discord, telegram）
+    - name: host
+      label:
+        en_US: Host
+        zh_Hans: 主机地址
+      type: string
+      required: true
+      default: "127.0.0.1"
+      description:
+        en_US: The host address of LLOneBot Satori server (e.g., 127.0.0.1, localhost, 192.168.1.100)
+        zh_Hans: LLOneBot Satori服务器的主机地址（如 127.0.0.1, localhost, 192.168.1.100）
+    - name: port
+      label:
+        en_US: Port
+        zh_Hans: 监听端口
+      type: integer
+      required: true
+      default: 5600
+    - name: satori_api_base_url
+      label:
+        en_US: Satori API Endpoint
+        zh_Hans: Satori API 终结点
+      type: string
+      required: true
+      default: "http://localhost:5600/v1"
+    - name: satori_endpoint
+      label:
+        en_US: Satori WebSocket Endpoint
+        zh_Hans: Satori WebSocket 终结点
+      type: string
+      required: true
+      default: "ws://localhost:5600/v1/events"
+    - name: token
+      label:
+        en_US: Token
+        zh_Hans: 令牌
+      type: string
+      required: true
+      default: ""
+execution:
+  python:
+    path: ./satori.py
+    attr: SatoriAdapter
--- a/src/langbot/pkg/platform/sources/telegram.py
+++ b/src/langbot/pkg/platform/sources/telegram.py
@@ -1,4 +1,5 @@
 from __future__ import annotations
+import time


 import telegram
@@ -9,9 +10,9 @@ import telegramify_markdown
 import typing
 import traceback
 import base64
-import aiohttp
 import pydantic

+from langbot.pkg.utils import httpclient
 import langbot_plugin.api.definition.abstract.platform.adapter as abstract_platform_adapter
 import langbot_plugin.api.entities.builtin.platform.message as platform_message
 import langbot_plugin.api.entities.builtin.platform.events as platform_events
@@ -33,14 +34,33 @@ class TelegramMessageConverter(abstract_platform_adapter.AbstractMessageConverte
                if component.base64:
                    photo_bytes = base64.b64decode(component.base64)
                elif component.url:
-                    async with aiohttp.ClientSession() as session:
-                        async with session.get(component.url) as response:
-                            photo_bytes = await response.read()
+                    session = httpclient.get_session()
+                    async with session.get(component.url) as response:
+                        photo_bytes = await response.read()
                elif component.path:
                    with open(component.path, 'rb') as f:
                        photo_bytes = f.read()

                components.append({'type': 'photo', 'photo': photo_bytes})
+            elif isinstance(component, platform_message.File):
+                file_bytes = None
+
+                if component.base64:
+                    # Strip data URI prefix if present (e.g. "data:application/pdf;base64,...")
+                    b64_data = component.base64
+                    if ';base64,' in b64_data:
+                        b64_data = b64_data.split(';base64,', 1)[1]
+                    file_bytes = base64.b64decode(b64_data)
+                elif component.url:
+                    session = httpclient.get_session()
+                    async with session.get(component.url) as response:
+                        file_bytes = await response.read()
+                elif component.path:
+                    with open(component.path, 'rb') as f:
+                        file_bytes = f.read()
+
+                file_name = getattr(component, 'name', None) or 'file'
+                components.append({'type': 'document', 'document': file_bytes, 'filename': file_name})
            elif isinstance(component, platform_message.Forward):
                for node in component.node_list:
                    components.extend(await TelegramMessageConverter.yiri2target(node.message_chain, bot))
@@ -74,10 +94,9 @@ class TelegramMessageConverter(abstract_platform_adapter.AbstractMessageConverte
            file_bytes = None
            file_format = ''

-            async with aiohttp.ClientSession(trust_env=True) as session:
-                async with session.get(file.file_path) as response:
-                    file_bytes = await response.read()
-                    file_format = 'image/jpeg'
+            async with httpclient.get_session(trust_env=True).get(file.file_path) as response:
+                file_bytes = await response.read()
+                file_format = 'image/jpeg'

            message_components.append(
                platform_message.Image(
@@ -94,9 +113,8 @@ class TelegramMessageConverter(abstract_platform_adapter.AbstractMessageConverte
            file_bytes = None
            file_format = message.voice.mime_type or 'audio/ogg'

-            async with aiohttp.ClientSession(trust_env=True) as session:
-                async with session.get(file.file_path) as response:
-                    file_bytes = await response.read()
+            async with httpclient.get_session(trust_env=True).get(file.file_path) as response:
+                file_bytes = await response.read()

            message_components.append(
                platform_message.Voice(
@@ -105,6 +123,27 @@ class TelegramMessageConverter(abstract_platform_adapter.AbstractMessageConverte
                )
            )

+        if message.document:
+            if message.caption:
+                message_components.extend(parse_message_text(message.caption))
+
+            file = await message.document.get_file()
+            file_name = message.document.file_name or 'document'
+            file_size = message.document.file_size or 0
+            file_format = message.document.mime_type or 'application/octet-stream'
+
+            file_bytes = None
+            async with httpclient.get_session(trust_env=True).get(file.file_path) as response:
+                file_bytes = await response.read()
+
+            message_components.append(
+                platform_message.File(
+                    name=file_name,
+                    size=file_size,
+                    base64=f'data:{file_format};base64,{base64.b64encode(file_bytes).decode("utf-8")}',
+                )
+            )
+
        return platform_message.MessageChain(message_components)


@@ -180,7 +219,10 @@ class TelegramAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        application = ApplicationBuilder().token(config['token']).build()
        bot = application.bot
        application.add_handler(
-            MessageHandler(filters.TEXT | (filters.COMMAND) | filters.PHOTO | filters.VOICE, telegram_callback)
+            MessageHandler(
+                filters.TEXT | (filters.COMMAND) | filters.PHOTO | filters.VOICE | filters.Document.ALL,
+                telegram_callback,
+            )
        )
        super().__init__(
            config=config,
@@ -194,7 +236,38 @@ class TelegramAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        )

    async def send_message(self, target_type: str, target_id: str, message: platform_message.MessageChain):
-        pass
+        components = await TelegramMessageConverter.yiri2target(message, self.bot)
+
+        chat_id_str, _, thread_id_str = str(target_id).partition('#')
+        chat_id: int | str = int(chat_id_str) if chat_id_str.lstrip('-').isdigit() else chat_id_str
+        message_thread_id = int(thread_id_str) if thread_id_str and thread_id_str.isdigit() else None
+
+        for component in components:
+            component_type = component.get('type')
+            args = {'chat_id': chat_id}
+            if message_thread_id is not None:
+                args['message_thread_id'] = message_thread_id
+
+            if component_type == 'text':
+                text = component.get('text', '')
+                if self.config['markdown_card'] is True:
+                    text = telegramify_markdown.markdownify(content=text)
+                    args['parse_mode'] = 'MarkdownV2'
+                args['text'] = text
+                await self.bot.send_message(**args)
+            elif component_type == 'photo':
+                photo = component.get('photo')
+                if photo is None:
+                    continue
+                args['photo'] = telegram.InputFile(photo)
+                await self.bot.send_photo(**args)
+            elif component_type == 'document':
+                doc = component.get('document')
+                if doc is None:
+                    continue
+                filename = component.get('filename', 'file')
+                args['document'] = telegram.InputFile(doc, filename=filename)
+                await self.bot.send_document(**args)

    async def reply_message(
        self,
@@ -228,6 +301,39 @@ class TelegramAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):

        await self.bot.send_message(**args)

+    def _process_markdown(self, text: str) -> str:
+        if self.config.get('markdown_card', False):
+            return telegramify_markdown.markdownify(content=text)
+        return text
+
+    def _build_message_args(self, chat_id: int, text: str, message_thread_id: int = None, **extra_args) -> dict:
+        args = {'chat_id': chat_id, 'text': self._process_markdown(text), **extra_args}
+        if message_thread_id:
+            args['message_thread_id'] = message_thread_id
+        if self.config.get('markdown_card', False):
+            args['parse_mode'] = 'MarkdownV2'
+        return args
+
+    async def create_message_card(self, message_id, event):
+        assert isinstance(event.source_platform_object, Update)
+        update = event.source_platform_object
+        chat_id = update.effective_chat.id
+        chat_type = update.effective_chat.type
+        message_thread_id = update.message.message_thread_id
+
+        if chat_type == 'private':
+            draft_id = int(time.time() * 1000)
+            self.msg_stream_id[message_id] = ('private', draft_id)
+
+            args = self._build_message_args(chat_id, 'Thinking...', message_thread_id, draft_id=draft_id)
+            await self.bot.send_message_draft(**args)
+        else:
+            args = self._build_message_args(chat_id, 'Thinking...', message_thread_id)
+            send_msg = await self.bot.send_message(**args)
+            self.msg_stream_id[message_id] = ('group', send_msg.message_id)
+
+        return True
+
    async def reply_message_chunk(
        self,
        message_source: platform_events.MessageEvent,
@@ -236,59 +342,47 @@ class TelegramAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        quote_origin: bool = False,
        is_final: bool = False,
    ):
+        message_id = bot_message.resp_message_id
        msg_seq = bot_message.msg_sequence
-        if (msg_seq - 1) % 8 == 0 or is_final:
-            assert isinstance(message_source.source_platform_object, Update)
-            components = await TelegramMessageConverter.yiri2target(message, self.bot)
-            args = {}
-            message_id = message_source.source_platform_object.message.id
+        assert isinstance(message_source.source_platform_object, Update)
+        update = message_source.source_platform_object
+        chat_id = update.effective_chat.id
+        message_thread_id = update.message.message_thread_id

-            component = components[0]
-            if message_id not in self.msg_stream_id:  # 当消息回复第一次时，发送新消息
-                # time.sleep(0.6)
-                if component['type'] == 'text':
-                    if self.config['markdown_card'] is True:
-                        content = telegramify_markdown.markdownify(
-                            content=component['text'],
-                        )
-                    else:
-                        content = component['text']
-                    args = {
-                        'chat_id': message_source.source_platform_object.effective_chat.id,
-                        'text': content,
-                    }
-                    if message_source.source_platform_object.message.message_thread_id:
-                        args['message_thread_id'] = message_source.source_platform_object.message.message_thread_id
+        if message_id not in self.msg_stream_id:
+            return

-                    if quote_origin:
-                        args['reply_to_message_id'] = message_source.source_platform_object.message.id
+        chat_mode, draft_id = self.msg_stream_id[message_id]
+        components = await TelegramMessageConverter.yiri2target(message, self.bot)

-                    if self.config['markdown_card'] is True:
-                        args['parse_mode'] = 'MarkdownV2'
-
-                send_msg = await self.bot.send_message(**args)
-                send_msg_id = send_msg.message_id
-                self.msg_stream_id[message_id] = send_msg_id
-            else:  # 存在消息的时候直接编辑消息1
-                if component['type'] == 'text':
-                    if self.config['markdown_card'] is True:
-                        content = telegramify_markdown.markdownify(
-                            content=component['text'],
-                        )
-                    else:
-                        content = component['text']
-                    args = {
-                        'message_id': self.msg_stream_id[message_id],
-                        'chat_id': message_source.source_platform_object.effective_chat.id,
-                        'text': content,
-                    }
-                    if self.config['markdown_card'] is True:
-                        args['parse_mode'] = 'MarkdownV2'
-
-                await self.bot.edit_message_text(**args)
+        if not components or components[0]['type'] != 'text':
            if is_final and bot_message.tool_calls is None:
-                # self.seq = 1  # 消息回复结束之后重置seq
-                self.msg_stream_id.pop(message_id)  # 消息回复结束之后删除流式消息id
+                self.msg_stream_id.pop(message_id)
+            return
+
+        content = components[0]['text']
+
+        if chat_mode == 'private':
+            args = self._build_message_args(chat_id, content, message_thread_id, draft_id=draft_id)
+            await self.bot.send_message_draft(**args)
+            if is_final and bot_message.tool_calls is None:
+                del args['draft_id']
+                await self.bot.send_message(**args)
+                self.msg_stream_id.pop(message_id)
+        else:
+            stream_id = draft_id
+            if (msg_seq - 1) % 8 == 0 or is_final:
+                args = {
+                    'message_id': stream_id,
+                    'chat_id': chat_id,
+                    'text': self._process_markdown(content),
+                }
+                if self.config.get('markdown_card', False):
+                    args['parse_mode'] = 'MarkdownV2'
+                await self.bot.edit_message_text(**args)
+
+            if is_final and bot_message.tool_calls is None:
+                self.msg_stream_id.pop(message_id)

    def get_launcher_id(self, event: platform_events.MessageEvent) -> str | None:
        if not isinstance(event.source_platform_object, Update):
--- a/src/langbot/pkg/platform/sources/websocket_adapter.py
+++ b/src/langbot/pkg/platform/sources/websocket_adapter.py
@@ -37,16 +37,24 @@ class WebSocketSession:
    id: str
    message_lists: dict[str, list[WebSocketMessage]] = {}
    """消息列表 {pipeline_uuid: [messages]}"""
+    stream_message_indexes: dict[str, dict[str, int]] = {}
+    """流式消息索引 {pipeline_uuid: {resp_message_id: message_index}}"""

    def __init__(self, id: str):
        self.id = id
        self.message_lists = {}
+        self.stream_message_indexes = {}

    def get_message_list(self, pipeline_uuid: str) -> list[WebSocketMessage]:
        if pipeline_uuid not in self.message_lists:
            self.message_lists[pipeline_uuid] = []
        return self.message_lists[pipeline_uuid]

+    def get_stream_message_indexes(self, pipeline_uuid: str) -> dict[str, int]:
+        if pipeline_uuid not in self.stream_message_indexes:
+            self.stream_message_indexes[pipeline_uuid] = {}
+        return self.stream_message_indexes[pipeline_uuid]
+

 class WebSocketAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
    """WebSocket适配器 - 支持双向实时通信"""
@@ -89,20 +97,46 @@ class WebSocketAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter)
        target_id: str,
        message: platform_message.MessageChain,
    ) -> dict:
-        """发送消息 - 这里用于主动推送消息到前端"""
-        message_data = {
-            'type': 'bot_message',
-            'target_type': target_type,
-            'target_id': target_id,
-            'content': str(message),
-            'message_chain': [component.__dict__ for component in message],
-            'timestamp': datetime.now().isoformat(),
-        }
+        """发送消息 - 这里用于主动推送消息到前端

-        # 推送到所有相关连接
-        await self.outbound_message_queue.put(message_data)
+        对于 WebSocket 适配器，我们需要将消息广播到正确的 pipeline 连接。
+        target_id 可能是 launcher_id（如 websocket_xxx）或 pipeline_uuid。
+        我们需要尝试两种方式来确保消息能够送达。
+        """
+        # 获取当前的 pipeline_uuid
+        pipeline_uuid = self.ap.platform_mgr.websocket_proxy_bot.bot_entity.use_pipeline_uuid
+        session_type = 'group' if target_type == 'group' else 'person'

-        return message_data
+        # 选择会话
+        session = self.websocket_group_session if session_type == 'group' else self.websocket_person_session
+
+        # 生成唯一消息ID
+        msg_id = len(session.get_message_list(pipeline_uuid)) + 1
+
+        message_data = WebSocketMessage(
+            id=msg_id,
+            role='assistant',
+            content=str(message),
+            message_chain=[component.__dict__ for component in message],
+            timestamp=datetime.now().isoformat(),
+            is_final=True,
+        )
+
+        # 保存到历史记录
+        session.get_message_list(pipeline_uuid).append(message_data)
+
+        # 直接广播到当前pipeline的连接
+        await ws_connection_manager.broadcast_to_pipeline(
+            pipeline_uuid,
+            {
+                'type': 'response',
+                'session_type': session_type,
+                'data': message_data.model_dump(),
+            },
+            session_type=session_type,
+        )
+
+        return message_data.model_dump()

    async def reply_message(
        self,
@@ -169,10 +203,16 @@ class WebSocketAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter)
        pipeline_uuid = self.ap.platform_mgr.websocket_proxy_bot.bot_entity.use_pipeline_uuid
        session_type = 'group' if isinstance(message_source, platform_events.GroupMessage) else 'person'
        message_list = session.get_message_list(pipeline_uuid)
+        stream_message_indexes = session.get_stream_message_indexes(pipeline_uuid)

-        # 检查是否是新的流式消息（通过bot_message对象判断）
-        # 如果列表为空，或者最后一条消息已经is_final=True，则创建新消息
-        if not message_list or message_list[-1].is_final:
+        # Streaming messages in LangBot have a stable resp_message_id during the same assistant reply.
+        # Use it as the primary key to avoid overwriting an old card from a previous reply.
+        resp_message_id = str(getattr(bot_message, 'resp_message_id', '') or '')
+        existing_index = stream_message_indexes.get(resp_message_id) if resp_message_id else None
+
+        message_is_final = is_final and bot_message.tool_calls is None
+
+        if existing_index is None or existing_index >= len(message_list):
            # 创建新消息
            msg_id = len(message_list) + 1
            message_data = WebSocketMessage(
@@ -181,27 +221,31 @@ class WebSocketAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter)
                content=str(message),
                message_chain=[component.__dict__ for component in message],
                timestamp=datetime.now().isoformat(),
-                is_final=is_final and bot_message.tool_calls is None,
+                is_final=message_is_final,
            )

-            # 只有在is_final时才保存到历史记录
-            if is_final and bot_message.tool_calls is None:
-                message_list.append(message_data)
+            # 立即添加到历史记录（即使is_final=False），以便后续块可以更新它
+            message_list.append(message_data)
+            if resp_message_id:
+                stream_message_indexes[resp_message_id] = len(message_list) - 1
        else:
-            # 更新最后一条消息
-            msg_id = message_list[-1].id
+            # 更新同一条流式消息
+            old_message = message_list[existing_index]
+            msg_id = old_message.id
            message_data = WebSocketMessage(
                id=msg_id,
                role='assistant',
                content=str(message),
                message_chain=[component.__dict__ for component in message],
-                timestamp=message_list[-1].timestamp,  # 保持原始时间戳
-                is_final=is_final and bot_message.tool_calls is None,
+                timestamp=old_message.timestamp,  # 保持原始时间戳
+                is_final=message_is_final,
            )

-            # 如果是final，更新历史记录中的最后一条
-            if is_final and bot_message.tool_calls is None:
-                message_list[-1] = message_data
+            # 更新历史记录中的对应消息
+            message_list[existing_index] = message_data
+
+        if message_is_final and resp_message_id:
+            stream_message_indexes.pop(resp_message_id, None)

        # 直接广播到所有该pipeline的连接，包含session_type信息
        await ws_connection_manager.broadcast_to_pipeline(
@@ -410,6 +454,10 @@ class WebSocketAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter)
        if session_type == 'person':
            if pipeline_uuid in self.websocket_person_session.message_lists:
                self.websocket_person_session.message_lists[pipeline_uuid] = []
+            if pipeline_uuid in self.websocket_person_session.stream_message_indexes:
+                self.websocket_person_session.stream_message_indexes[pipeline_uuid] = {}
        else:
            if pipeline_uuid in self.websocket_group_session.message_lists:
                self.websocket_group_session.message_lists[pipeline_uuid] = []
+            if pipeline_uuid in self.websocket_group_session.stream_message_indexes:
+                self.websocket_group_session.stream_message_indexes[pipeline_uuid] = {}
--- a/src/langbot/pkg/platform/sources/wechat.png
+++ b/src/langbot/pkg/platform/sources/wechat.png
--- a/src/langbot/pkg/platform/sources/wecom.py
+++ b/src/langbot/pkg/platform/sources/wecom.py
@@ -148,51 +148,54 @@ class WecomEventConverter(abstract_platform_adapter.AbstractEventConverter):
            pass

        if type(event) is platform_events.FriendMessage:
-            payload = {
-                'MsgType': 'text',
-                'Content': '',
-                'FromUserName': event.sender.id,
-                'ToUserName': bot_account_id,
-                'CreateTime': int(datetime.datetime.now().timestamp()),
-                'AgentID': event.sender.nickname,
-            }
-            wecom_event = WecomEvent.from_payload(payload=payload)
-            if not wecom_event:
-                raise ValueError('无法从 message_data 构造 WecomEvent 对象')
-
-            return wecom_event
+            return event.source_platform_object

    @staticmethod
-    async def target2yiri(event: WecomEvent):
+    async def target2yiri(event: WecomEvent, bot: WecomClient = None):
        """
        将 WecomEvent 转换为平台的 FriendMessage 对象。

        Args:
            event (WecomEvent): 企业微信事件。
+            bot (WecomClient): 企业微信客户端，用于获取用户信息。

        Returns:
            platform_events.FriendMessage: 转换后的 FriendMessage 对象。
        """
+        # Try to get the user's real name from the WeCom API
+        nickname = str(event.user_id)
+        if bot and event.user_id:
+            try:
+                user_info = await bot.get_user_info(event.user_id)
+                if user_info and user_info.get('name'):
+                    nickname = user_info.get('name')
+            except Exception:
+                pass  # Fall back to user_id as nickname
+
        # 转换消息链
        if event.type == 'text':
            yiri_chain = await WecomMessageConverter.target2yiri(event.message, event.message_id)
            friend = platform_entities.Friend(
                id=f'u{event.user_id}',
-                nickname=str(event.agent_id),
+                nickname=nickname,
                remark='',
            )

-            return platform_events.FriendMessage(sender=friend, message_chain=yiri_chain, time=event.timestamp)
+            return platform_events.FriendMessage(
+                sender=friend, message_chain=yiri_chain, time=event.timestamp, source_platform_object=event
+            )
        elif event.type == 'image':
            friend = platform_entities.Friend(
                id=f'u{event.user_id}',
-                nickname=str(event.agent_id),
+                nickname=nickname,
                remark='',
            )

            yiri_chain = await WecomMessageConverter.target2yiri_image(picurl=event.picurl, message_id=event.message_id)

-            return platform_events.FriendMessage(sender=friend, message_chain=yiri_chain, time=event.timestamp)
+            return platform_events.FriendMessage(
+                sender=friend, message_chain=yiri_chain, time=event.timestamp, source_platform_object=event
+            )


 class WecomAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
@@ -210,7 +213,6 @@ class WecomAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
            'secret',
            'token',
            'EncodingAESKey',
-            'contacts_secret',
        ]

        missing_keys = [key for key in required_keys if key not in config]
@@ -223,7 +225,7 @@ class WecomAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
            secret=config['secret'],
            token=config['token'],
            EncodingAESKey=config['EncodingAESKey'],
-            contacts_secret=config['contacts_secret'],
+            contacts_secret=config.get('contacts_secret', ''),  # Optional, kept for backward compatibility
            logger=logger,
            unified_mode=True,
            api_base_url=config.get('api_base_url', 'https://qyapi.weixin.qq.com/cgi-bin'),
@@ -248,18 +250,17 @@ class WecomAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
    ):
        Wecom_event = await WecomEventConverter.yiri2target(message_source, self.bot_account_id, self.bot)
        content_list = await WecomMessageConverter.yiri2target(message, self.bot)
-        fixed_user_id = Wecom_event.user_id
-        # 删掉开头的u
-        fixed_user_id = fixed_user_id[1:]
+        # user_id is the original FromUserName from WecomEvent
+        user_id = Wecom_event.user_id
        for content in content_list:
            if content['type'] == 'text':
-                await self.bot.send_private_msg(fixed_user_id, Wecom_event.agent_id, content['content'])
+                await self.bot.send_private_msg(user_id, Wecom_event.agent_id, content['content'])
            elif content['type'] == 'image':
-                await self.bot.send_image(fixed_user_id, Wecom_event.agent_id, content['media_id'])
+                await self.bot.send_image(user_id, Wecom_event.agent_id, content['media_id'])
            elif content['type'] == 'voice':
-                await self.bot.send_voice(fixed_user_id, Wecom_event.agent_id, content['media_id'])
+                await self.bot.send_voice(user_id, Wecom_event.agent_id, content['media_id'])
            elif content['type'] == 'file':
-                await self.bot.send_file(fixed_user_id, Wecom_event.agent_id, content['media_id'])
+                await self.bot.send_file(user_id, Wecom_event.agent_id, content['media_id'])

    async def send_message(self, target_type: str, target_id: str, message: platform_message.MessageChain):
        content_list = await WecomMessageConverter.yiri2target(message, self.bot)
@@ -287,7 +288,7 @@ class WecomAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        async def on_message(event: WecomEvent):
            self.bot_account_id = event.receiver_id
            try:
-                return await callback(await self.event_converter.target2yiri(event), self)
+                return await callback(await self.event_converter.target2yiri(event, self.bot), self)
            except Exception:
                await self.logger.error(f'Error in wecom callback: {traceback.format_exc()}')

--- a/src/langbot/pkg/platform/sources/wecom.yaml
+++ b/src/langbot/pkg/platform/sources/wecom.yaml
@@ -39,13 +39,6 @@ spec:
      type: string
      required: true
      default: ""
-    - name: contacts_secret
-      label:
-        en_US: Contacts Secret
-        zh_Hans: 通讯录密钥
-      type: string
-      required: true
-      default: ""
    - name: api_base_url
      label:
        en_US: API Base URL
--- a/src/langbot/pkg/platform/sources/wecombot.py
+++ b/src/langbot/pkg/platform/sources/wecombot.py
@@ -11,6 +11,7 @@ import langbot_plugin.api.entities.builtin.platform.entities as platform_entitie
 from ..logger import EventLogger
 from langbot.libs.wecom_ai_bot_api.wecombotevent import WecomBotEvent
 from langbot.libs.wecom_ai_bot_api.api import WecomBotClient
+from langbot.libs.wecom_ai_bot_api.ws_client import WecomBotWsClient


 class WecomBotMessageConverter(abstract_platform_adapter.AbstractMessageConverter):
@@ -23,14 +24,18 @@ class WecomBotMessageConverter(abstract_platform_adapter.AbstractMessageConverte
        return content

    @staticmethod
-    async def target2yiri(event: WecomBotEvent):
+    async def target2yiri(event: WecomBotEvent, bot_name: str = ''):
        yiri_msg_list = []
        if event.type == 'group':
            yiri_msg_list.append(platform_message.At(target=event.ai_bot_id))
+
        yiri_msg_list.append(platform_message.Source(id=event.message_id, time=datetime.datetime.now()))

        if event.content:
-            yiri_msg_list.append(platform_message.Plain(text=event.content))
+            content = event.content
+            if bot_name:
+                content = content.replace(f'@{bot_name}', '').strip()
+            yiri_msg_list.append(platform_message.Plain(text=content))

        images = []
        if event.images:
@@ -133,13 +138,15 @@ class WecomBotMessageConverter(abstract_platform_adapter.AbstractMessageConverte


 class WecomBotEventConverter(abstract_platform_adapter.AbstractEventConverter):
+    def __init__(self, bot_name: str = ''):
+        self.bot_name = bot_name
+
    @staticmethod
    async def yiri2target(event: platform_events.MessageEvent):
        return event.source_platform_object

-    @staticmethod
-    async def target2yiri(event: WecomBotEvent):
-        message_chain = await WecomBotMessageConverter.target2yiri(event)
+    async def target2yiri(self, event: WecomBotEvent):
+        message_chain = await WecomBotMessageConverter.target2yiri(event, bot_name=self.bot_name)
        if event.type == 'single':
            return platform_events.FriendMessage(
                sender=platform_entities.Friend(
@@ -176,34 +183,53 @@ class WecomBotEventConverter(abstract_platform_adapter.AbstractEventConverter):


 class WecomBotAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
-    bot: WecomBotClient
+    bot: typing.Union[WecomBotClient, WecomBotWsClient]
    bot_account_id: str
    message_converter: WecomBotMessageConverter = WecomBotMessageConverter()
-    event_converter: WecomBotEventConverter = WecomBotEventConverter()
+    event_converter: WecomBotEventConverter
    config: dict
    bot_uuid: str = None
+    _ws_mode: bool = False
+    bot_name: str = ''
+    listeners: dict = {}

    def __init__(self, config: dict, logger: EventLogger):
-        required_keys = ['Token', 'EncodingAESKey', 'Corpid', 'BotId']
-        missing_keys = [key for key in required_keys if key not in config]
-        if missing_keys:
-            raise Exception(f'WecomBot 缺少配置项: {missing_keys}')
+        enable_webhook = config.get('enable-webhook', False)
+        bot_name = config.get('robot_name', '')

-        bot = WecomBotClient(
-            Token=config['Token'],
-            EnCodingAESKey=config['EncodingAESKey'],
-            Corpid=config['Corpid'],
-            logger=logger,
-            unified_mode=True,
-        )
-        bot_account_id = config['BotId']
+        if not enable_webhook:
+            bot = WecomBotWsClient(
+                bot_id=config['BotId'],
+                secret=config['Secret'],
+                logger=logger,
+                encoding_aes_key=config.get('EncodingAESKey', ''),
+            )
+        else:
+            # Webhook callback mode
+            required_keys = ['Token', 'EncodingAESKey', 'Corpid']
+            missing_keys = [key for key in required_keys if key not in config or not config[key]]
+            if missing_keys:
+                raise Exception(f'WecomBot webhook mode missing config: {missing_keys}')

+            bot = WecomBotClient(
+                Token=config['Token'],
+                EnCodingAESKey=config['EncodingAESKey'],
+                Corpid=config['Corpid'],
+                logger=logger,
+                unified_mode=True,
+            )
+
+        bot_account_id = config.get('BotId', '')
+        event_converter = WecomBotEventConverter(bot_name=bot_name)
        super().__init__(
            config=config,
            logger=logger,
            bot=bot,
            bot_account_id=bot_account_id,
+            bot_name=bot_name,
+            event_converter=event_converter,
        )
+        self.listeners = {}

    async def reply_message(
        self,
@@ -212,7 +238,17 @@ class WecomBotAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        quote_origin: bool = False,
    ):
        content = await self.message_converter.yiri2target(message)
-        await self.bot.set_message(message_source.source_platform_object.message_id, content)
+        _ws_mode = not self.config.get('enable-webhook', False)
+
+        if _ws_mode:
+            event = message_source.source_platform_object
+            req_id = event.get('req_id', '')
+            if req_id:
+                await self.bot.reply_text(req_id, content)
+            else:
+                await self.bot.set_message(event.message_id, content)
+        else:
+            await self.bot.set_message(message_source.source_platform_object.message_id, content)

    async def reply_message_chunk(
        self,
@@ -222,31 +258,23 @@ class WecomBotAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        quote_origin: bool = False,
        is_final: bool = False,
    ):
-        """将流水线增量输出写入企业微信 stream 会话。
-
-        Args:
-            message_source: 流水线提供的原始消息事件。
-            bot_message: 当前片段对应的模型元信息（未使用）。
-            message: 需要回复的消息链。
-            quote_origin: 是否引用原消息（企业微信暂不支持）。
-            is_final: 标记当前片段是否为最终回复。
-
-        Returns:
-            dict: 包含 `stream` 键，标识写入是否成功。
-
-        Example:
-            在流水线 `reply_message_chunk` 调用中自动触发，无需手动调用。
-        """
-        # 转换为纯文本（智能机器人当前协议仅支持文本流）
        content = await self.message_converter.yiri2target(message)
        msg_id = message_source.source_platform_object.message_id
+        _ws_mode = not self.config.get('enable-webhook', False)

-        # 将片段推送到 WecomBotClient 中的队列，返回值用于判断是否走降级逻辑
-        success = await self.bot.push_stream_chunk(msg_id, content, is_final=is_final)
-        if not success and is_final:
-            # 未命中流式队列时使用旧有 set_message 兜底
-            await self.bot.set_message(msg_id, content)
-        return {'stream': success}
+        if _ws_mode:
+            success = await self.bot.push_stream_chunk(msg_id, content, is_final=is_final)
+            if not success and is_final:
+                event = message_source.source_platform_object
+                req_id = event.get('req_id', '')
+                if req_id:
+                    await self.bot.reply_text(req_id, content)
+            return {'stream': success}
+        else:
+            success = await self.bot.push_stream_chunk(msg_id, content, is_final=is_final)
+            if not success and is_final:
+                await self.bot.set_message(msg_id, content)
+            return {'stream': success}

    async def is_stream_output_supported(self) -> bool:
        """智能机器人侧默认开启流式能力。
@@ -259,7 +287,21 @@ class WecomBotAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        return True

    async def send_message(self, target_type, target_id, message):
-        pass
+        _ws_mode = not self.config.get('enable-webhook', False)
+        if _ws_mode:
+            content = await self.message_converter.yiri2target(message)
+            await self.bot.send_message(target_id, content)
+        else:
+            pass
+
+    async def on_message(self, event: WecomBotEvent):
+        try:
+            lb_event = await self.event_converter.target2yiri(event)
+            if lb_event:
+                await self.listeners[type(lb_event)](lb_event, self)
+        except Exception:
+            await self.logger.error(f'Error in wecombot callback: {traceback.format_exc()}')
+            print(traceback.format_exc())

    def register_listener(
        self,
@@ -268,18 +310,13 @@ class WecomBotAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
            [platform_events.Event, abstract_platform_adapter.AbstractMessagePlatformAdapter], None
        ],
    ):
-        async def on_message(event: WecomBotEvent):
-            try:
-                return await callback(await self.event_converter.target2yiri(event), self)
-            except Exception:
-                await self.logger.error(f'Error in wecombot callback: {traceback.format_exc()}')
-                print(traceback.format_exc())
+        self.listeners[event_type] = callback

        try:
            if event_type == platform_events.FriendMessage:
-                self.bot.on_message('single')(on_message)
+                self.bot.on_message('single')(self.on_message)
            elif event_type == platform_events.GroupMessage:
-                self.bot.on_message('group')(on_message)
+                self.bot.on_message('group')(self.on_message)
        except Exception:
            print(traceback.format_exc())

@@ -288,29 +325,28 @@ class WecomBotAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        self.bot_uuid = bot_uuid

    async def handle_unified_webhook(self, bot_uuid: str, path: str, request):
-        """处理统一 webhook 请求。
-
-        Args:
-            bot_uuid: Bot 的 UUID
-            path: 子路径（如果有的话）
-            request: Quart Request 对象
-
-        Returns:
-            响应数据
-        """
+        _ws_mode = not self.config.get('enable-webhook', False)
+        if _ws_mode:
+            return None
        return await self.bot.handle_unified_webhook(request)

    async def run_async(self):
-        # 统一 webhook 模式下，不启动独立的 Quart 应用
-        # 保持运行但不启动独立端口
+        _ws_mode = not self.config.get('enable-webhook', False)
+        if _ws_mode:
+            await self.bot.connect()
+        else:

-        async def keep_alive():
-            while True:
-                await asyncio.sleep(1)
+            async def keep_alive():
+                while True:
+                    await asyncio.sleep(1)

-        await keep_alive()
+            await keep_alive()

    async def kill(self) -> bool:
+        _ws_mode = not self.config.get('enable-webhook', False)
+        if _ws_mode:
+            await self.bot.disconnect()
+            return True
        return False

    async def unregister_listener(
--- a/src/langbot/pkg/platform/sources/wecombot.yaml
+++ b/src/langbot/pkg/platform/sources/wecombot.yaml
@@ -11,35 +11,71 @@ metadata:
  icon: wecombot.png
 spec:
  config:
+    - name: BotId
+      label:
+        en_US: BotId
+        zh_Hans: 机器人ID (BotId)
+      type: string
+      required: true
+      default: ""
+    - name: robot_name
+      label:
+        en_US: Robot Name
+        zh_Hans: 机器人名称
+      type: string
+      required: true
+      default: ""
+    - name: enable-webhook
+      label:
+        en_US: Enable Webhook Mode
+        zh_Hans: 启用Webhook模式
+      description:
+        en_US: If enabled, the bot will use webhook mode to receive messages. Otherwise, it will use WS long connection mode
+        zh_Hans: 如果启用，机器人将使用 Webhook 模式接收消息。否则，将使用 WS 长连接模式
+      type: boolean
+      required: true
+      default: false
+    - name: Secret
+      label:
+        en_US: Secret
+        zh_Hans: 机器人密钥 (Secret)
+      description:
+        en_US: Required for WebSocket long connection mode
+        zh_Hans: 使用 WS 长连接模式时必填
+      type: string
+      required: false
+      default: ""
    - name: Corpid
      label:
        en_US: Corpid
        zh_Hans: 企业ID
+      description:
+        en_US: Required for Webhook mode
+        zh_Hans: 使用 Webhook 模式时必填
      type: string
-      required: true
+      required: false
      default: ""
    - name: Token
      label:
        en_US: Token
        zh_Hans: 令牌 (Token)
+      description:
+        en_US: Required for Webhook mode
+        zh_Hans: 使用 Webhook 模式时必填
      type: string
-      required: true
+      required: false
      default: ""
    - name: EncodingAESKey
      label:
        en_US: EncodingAESKey
        zh_Hans: 消息加解密密钥 (EncodingAESKey)
-      type: string
-      required: true
-      default: ""
-    - name: BotId
-      label:
-        en_US: BotId
-        zh_Hans: 机器人ID
+      description:
+        en_US: Required for Webhook mode. Optional for WebSocket mode (used for file decryption)
+        zh_Hans: 使用 Webhook 模式时必填。WebSocket 模式下可选（用于文件解密）
      type: string
      required: false
      default: ""
 execution:
  python:
    path: ./wecombot.py
-    attr: WecomBotAdapter
+    attr: WecomBotAdapter
--- a/src/langbot/pkg/platform/sources/wecomcs.py
+++ b/src/langbot/pkg/platform/sources/wecomcs.py
@@ -81,22 +81,33 @@ class WecomEventConverter(abstract_platform_adapter.AbstractEventConverter):
            return event.source_platform_object

    @staticmethod
-    async def target2yiri(event: WecomCSEvent):
+    async def target2yiri(event: WecomCSEvent, bot: WecomCSClient = None):
        """
        将 WecomEvent 转换为平台的 FriendMessage 对象。

        Args:
            event (WecomEvent): 企业微信客服事件。
+            bot (WecomCSClient): 企业微信客服客户端，用于获取用户信息。

        Returns:
            platform_events.FriendMessage: 转换后的 FriendMessage 对象。
        """
+        # Try to get customer nickname from WeChat API
+        nickname = str(event.user_id)
+        if bot and event.user_id:
+            try:
+                customer_info = await bot.get_customer_info(event.user_id)
+                if customer_info and customer_info.get('nickname'):
+                    nickname = customer_info.get('nickname')
+            except Exception:
+                pass  # Fall back to user_id as nickname
+
        # 转换消息链
        if event.type == 'text':
            yiri_chain = await WecomMessageConverter.target2yiri(event.message, event.message_id)
            friend = platform_entities.Friend(
                id=f'u{event.user_id}',
-                nickname=str(event.user_id),
+                nickname=nickname,
                remark='',
            )

@@ -106,7 +117,7 @@ class WecomEventConverter(abstract_platform_adapter.AbstractEventConverter):
        elif event.type == 'image':
            friend = platform_entities.Friend(
                id=f'u{event.user_id}',
-                nickname=str(event.user_id),
+                nickname=nickname,
                remark='',
            )

@@ -187,7 +198,7 @@ class WecomCSAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
        async def on_message(event: WecomCSEvent):
            self.bot_account_id = event.receiver_id
            try:
-                return await callback(await self.event_converter.target2yiri(event), self)
+                return await callback(await self.event_converter.target2yiri(event, self.bot), self)
            except Exception:
                await self.logger.error(f'Error in wecomcs callback: {traceback.format_exc()}')

--- a/src/langbot/pkg/platform/webhook_pusher.py
+++ b/src/langbot/pkg/platform/webhook_pusher.py
@@ -3,6 +3,8 @@ from __future__ import annotations
 import asyncio
 import logging
 import aiohttp
+
+from langbot.pkg.utils import httpclient
 import uuid
 from typing import TYPE_CHECKING

@@ -119,23 +121,23 @@ class WebhookPusher:
            dict | None: The response JSON if successful, None otherwise
        """
        try:
-            async with aiohttp.ClientSession() as session:
-                async with session.post(
-                    url,
-                    json=payload,
-                    headers={'Content-Type': 'application/json'},
-                    timeout=aiohttp.ClientTimeout(total=15),
-                ) as response:
-                    if response.status >= 400:
-                        self.logger.warning(f'Webhook {url} returned status {response.status}')
+            session = httpclient.get_session()
+            async with session.post(
+                url,
+                json=payload,
+                headers={'Content-Type': 'application/json'},
+                timeout=aiohttp.ClientTimeout(total=15),
+            ) as response:
+                if response.status >= 400:
+                    self.logger.warning(f'Webhook {url} returned status {response.status}')
+                    return None
+                else:
+                    self.logger.debug(f'Successfully pushed to webhook {url}')
+                    try:
+                        return await response.json()
+                    except Exception as json_error:
+                        self.logger.debug(f'Failed to parse JSON response from webhook {url}: {json_error}')
                        return None
-                    else:
-                        self.logger.debug(f'Successfully pushed to webhook {url}')
-                        try:
-                            return await response.json()
-                        except Exception as json_error:
-                            self.logger.debug(f'Failed to parse JSON response from webhook {url}: {json_error}')
-                            return None
        except asyncio.TimeoutError:
            self.logger.warning(f'Timeout pushing to webhook {url}')
            return None
--- a/src/langbot/pkg/plugin/connector.py
+++ b/src/langbot/pkg/plugin/connector.py
@@ -7,7 +7,6 @@ import typing
 import os
 import sys
 import httpx
-import traceback
 import sqlalchemy
 from async_lru import alru_cache
 from langbot_plugin.api.entities.builtin.pipeline.query import provider_session
@@ -102,12 +101,6 @@ class PluginRuntimeConnector:
            self.handler_task = asyncio.create_task(self.handler.run())
            _ = await self.handler.ping()
            self.ap.logger.info('Connected to plugin runtime.')
-            # Sync polymorphic component instances after connection
-            try:
-                await self.sync_polymorphic_component_instances()
-            except Exception as e:
-                traceback.print_exc()
-                self.ap.logger.error(f'Failed to sync polymorphic component instances: {e}')
            await self.handler_task

        task: asyncio.Task | None = None
@@ -463,30 +456,18 @@ class PluginRuntimeConnector:

            yield cmd_ret

-    # KnowledgeRetriever methods
-    async def list_knowledge_retrievers(self, bound_plugins: list[str] | None = None) -> list[dict[str, Any]]:
-        """List all available KnowledgeRetriever components."""
-        if not self.is_enable_plugin:
-            return []
-
-        retrievers_data = await self.handler.list_knowledge_retrievers(include_plugins=bound_plugins)
-        return retrievers_data
-
    async def retrieve_knowledge(
        self,
        plugin_author: str,
        plugin_name: str,
        retriever_name: str,
-        instance_id: str,
        retrieval_context: dict[str, Any],
-    ) -> list[dict[str, Any]]:
-        """Retrieve knowledge using a KnowledgeRetriever instance."""
+    ) -> dict[str, Any]:
+        """Retrieve knowledge using a KnowledgeEngine instance."""
        if not self.is_enable_plugin:
-            return []
+            return {'results': []}

-        return await self.handler.retrieve_knowledge(
-            plugin_author, plugin_name, retriever_name, instance_id, retrieval_context
-        )
+        return await self.handler.retrieve_knowledge(plugin_author, plugin_name, retriever_name, retrieval_context)

    def dispose(self):
        # No need to consider the shutdown on Windows
@@ -500,41 +481,84 @@ class PluginRuntimeConnector:
            self.heartbeat_task.cancel()
            self.heartbeat_task = None

-    async def sync_polymorphic_component_instances(self) -> dict[str, Any]:
-        """Sync polymorphic component instances with runtime.
+    @staticmethod
+    def _parse_plugin_id(plugin_id: str) -> tuple[str, str]:
+        """Parse a plugin ID string into (author, name).

-        This collects all external knowledge bases from database and sends to runtime
-        to ensure instance integrity across restarts.
+        Args:
+            plugin_id: Plugin ID in 'author/name' format.
+
+        Returns:
+            Tuple of (plugin_author, plugin_name).
+
+        Raises:
+            ValueError: If plugin_id is not in the expected 'author/name' format.
+        """
+        if '/' not in plugin_id:
+            raise ValueError(
+                f"Invalid plugin_id format: '{plugin_id}'. Expected 'author/name' format (e.g. 'langbot/rag-engine')."
+            )
+        return plugin_id.split('/', 1)
+
+    async def call_rag_ingest(self, plugin_id: str, context_data: dict[str, Any]) -> dict[str, Any]:
+        """Call plugin to ingest document.
+
+        Args:
+            plugin_id: Target plugin ID (author/name).
+            context_data: IngestionContext data.
+        """
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.rag_ingest_document(plugin_author, plugin_name, context_data)
+
+    async def call_rag_delete_document(self, plugin_id: str, document_id: str, kb_id: str) -> bool:
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.rag_delete_document(plugin_author, plugin_name, document_id, kb_id)
+
+    async def get_rag_creation_schema(self, plugin_id: str) -> dict[str, Any]:
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.get_rag_creation_schema(plugin_author, plugin_name)
+
+    async def get_rag_retrieval_schema(self, plugin_id: str) -> dict[str, Any]:
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.get_rag_retrieval_schema(plugin_author, plugin_name)
+
+    async def rag_on_kb_create(self, plugin_id: str, kb_id: str, config: dict[str, Any]) -> dict[str, Any]:
+        """Notify plugin about KB creation."""
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.rag_on_kb_create(plugin_author, plugin_name, kb_id, config)
+
+    async def rag_on_kb_delete(self, plugin_id: str, kb_id: str) -> dict[str, Any]:
+        """Notify plugin about KB deletion."""
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.rag_on_kb_delete(plugin_author, plugin_name, kb_id)
+
+    async def call_rag_retrieve(self, plugin_id: str, retrieval_context: dict[str, Any]) -> dict[str, Any]:
+        """Call plugin to retrieve knowledge.
+
+        Args:
+            plugin_id: Target plugin ID (author/name).
+            retrieval_context: RetrievalContext data.
+        """
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.retrieve_knowledge(plugin_author, plugin_name, '', retrieval_context)
+
+    async def list_knowledge_engines(self) -> list[dict[str, Any]]:
+        """List all available Knowledge Engines from plugins.
+
+        Returns a list of Knowledge Engines with their capabilities and configuration schemas.
        """
        if not self.is_enable_plugin:
-            return {}
+            return []

-        # ===== external knowledge bases =====
+        return await self.handler.list_knowledge_engines()

-        external_kbs = await self.ap.external_kb_service.get_external_knowledge_bases()
+    async def list_parsers(self) -> list[dict[str, Any]]:
+        """List all available parsers from plugins."""
+        if not self.is_enable_plugin:
+            return []
+        return await self.handler.list_parsers()

-        # Build required_instances list
-        required_instances = []
-        for kb in external_kbs:
-            required_instances.append(
-                {
-                    'instance_id': kb['uuid'],
-                    'plugin_author': kb['plugin_author'],
-                    'plugin_name': kb['plugin_name'],
-                    'component_kind': 'KnowledgeRetriever',
-                    'component_name': kb['retriever_name'],
-                    'config': kb['retriever_config'],
-                }
-            )
-
-        self.ap.logger.info(f'Syncing {len(required_instances)} polymorphic component instances to runtime')
-
-        # Send to runtime
-        sync_result = await self.handler.sync_polymorphic_component_instances(required_instances)
-
-        self.ap.logger.info(
-            f'Sync complete: {len(sync_result.get("success_instances", []))} succeeded, '
-            f'{len(sync_result.get("failed_instances", []))} failed'
-        )
-
-        return sync_result
+    async def call_parser(self, plugin_id: str, context_data: dict[str, Any], file_bytes: bytes) -> dict[str, Any]:
+        """Call plugin to parse a document."""
+        plugin_author, plugin_name = self._parse_plugin_id(plugin_id)
+        return await self.handler.parse_document(plugin_author, plugin_name, context_data, file_bytes)
--- a/src/langbot/pkg/plugin/handler.py
+++ b/src/langbot/pkg/plugin/handler.py
@@ -26,6 +26,20 @@ from ..core import app
 from ..utils import constants


+def _make_rag_error_response(error: Exception, error_type: str, **extra_context) -> handler.ActionResponse:
+    """Create a clean error response for RAG operations.
+
+    Args:
+        error: The caught exception.
+        error_type: A category string like 'EmbeddingError', 'VectorStoreError'.
+        **extra_context: Additional context fields for the error message.
+    """
+    context_parts = [f'{k}={v}' for k, v in extra_context.items()]
+    context_str = f' [{", ".join(context_parts)}]' if context_parts else ''
+    message = f'[{error_type}/{type(error).__name__}]{context_str} {str(error)}'
+    return handler.ActionResponse.error(message=message)
+
+
 class RuntimeConnectionHandler(handler.Handler):
    """Runtime connection handler"""

@@ -279,6 +293,7 @@ class RuntimeConnectionHandler(handler.Handler):
            target_id = data['target_id']
            message_chain = data['message_chain']

+            # Use custom deserializer that properly handles Forward messages
            message_chain_obj = platform_message.MessageChain.model_validate(message_chain)

            bot = await self.ap.platform_mgr.get_bot_by_uuid(bot_uuid)
@@ -299,11 +314,11 @@ class RuntimeConnectionHandler(handler.Handler):

        @self.action(PluginToRuntimeAction.GET_LLM_MODELS)
        async def get_llm_models(data: dict[str, Any]) -> handler.ActionResponse:
-            """Get llm models"""
+            """Get llm models, returns list of UUID strings"""
            llm_models = await self.ap.llm_model_service.get_llm_models(include_secret=False)
            return handler.ActionResponse.success(
                data={
-                    'llm_models': llm_models,
+                    'llm_models': [m['uuid'] for m in llm_models],
                },
            )

@@ -322,7 +337,14 @@ class RuntimeConnectionHandler(handler.Handler):
                )

            messages_obj = [provider_message.Message.model_validate(message) for message in messages]
-            funcs_obj = [resource_tool.LLMTool.model_validate(func) for func in funcs]
+
+            # The func field is excluded during model_dump() in plugin side (marked as exclude=True),
+            # but it's a required field for LLMTool validation. We need to provide a placeholder
+            # function when reconstructing the LLMTool objects from serialized data.
+            async def _placeholder_func(**kwargs):
+                pass
+
+            funcs_obj = [resource_tool.LLMTool.model_validate({**func, 'func': _placeholder_func}) for func in funcs]

            result = await llm_model.provider.invoke_llm(
                query=None,
@@ -438,7 +460,7 @@ class RuntimeConnectionHandler(handler.Handler):
                },
            )

-        @self.action(RuntimeToLangBotAction.GET_CONFIG_FILE)
+        @self.action(PluginToRuntimeAction.GET_CONFIG_FILE)
        async def get_config_file(data: dict[str, Any]) -> handler.ActionResponse:
            """Get a config file by file key"""
            file_key = data['file_key']
@@ -457,6 +479,282 @@ class RuntimeConnectionHandler(handler.Handler):
                    message=f'Failed to load config file {file_key}: {e}',
                )

+        # ================= RAG Capability Handlers =================
+
+        @self.action(PluginToRuntimeAction.INVOKE_EMBEDDING)
+        async def invoke_embedding(data: dict[str, Any]) -> handler.ActionResponse:
+            embedding_model_uuid = data['embedding_model_uuid']
+            texts = data['texts']
+
+            embedding_model = await self.ap.model_mgr.get_embedding_model_by_uuid(embedding_model_uuid)
+            if embedding_model is None:
+                return handler.ActionResponse.error(
+                    message=f'Embedding model with embedding_model_uuid {embedding_model_uuid} not found',
+                )
+
+            try:
+                vectors = await embedding_model.provider.invoke_embedding(embedding_model, texts)
+                return handler.ActionResponse.success(data={'vectors': vectors})
+            except Exception as e:
+                return _make_rag_error_response(e, 'EmbeddingError', embedding_model_uuid=embedding_model_uuid)
+
+        @self.action(PluginToRuntimeAction.VECTOR_UPSERT)
+        async def vector_upsert(data: dict[str, Any]) -> handler.ActionResponse:
+            collection_id = data['collection_id']
+            vectors = data['vectors']
+            ids = data['ids']
+            metadata = data.get('metadata')
+            documents = data.get('documents')
+            if len(vectors) != len(ids):
+                return handler.ActionResponse.error(message='vectors and ids must have same length')
+            if metadata and len(metadata) != len(vectors):
+                return handler.ActionResponse.error(message='metadata must match vectors length')
+            if documents and len(documents) != len(vectors):
+                return handler.ActionResponse.error(message='documents must match vectors length')
+            try:
+                await self.ap.rag_runtime_service.vector_upsert(
+                    collection_id,
+                    vectors,
+                    ids,
+                    metadata,
+                    documents,
+                )
+                return handler.ActionResponse.success(data={})
+            except Exception as e:
+                return _make_rag_error_response(e, 'VectorStoreError', collection_id=collection_id)
+
+        @self.action(PluginToRuntimeAction.VECTOR_SEARCH)
+        async def vector_search(data: dict[str, Any]) -> handler.ActionResponse:
+            collection_id = data['collection_id']
+            query_vector = data['query_vector']
+            top_k = data['top_k']
+            filters = data.get('filters')
+            search_type = data.get('search_type', 'vector')
+            query_text = data.get('query_text', '')
+            vector_weight = data.get('vector_weight')
+            try:
+                results = await self.ap.rag_runtime_service.vector_search(
+                    collection_id,
+                    query_vector,
+                    top_k,
+                    filters,
+                    search_type,
+                    query_text,
+                    vector_weight=vector_weight,
+                )
+                return handler.ActionResponse.success(data={'results': results})
+            except Exception as e:
+                return _make_rag_error_response(e, 'VectorStoreError', collection_id=collection_id)
+
+        @self.action(PluginToRuntimeAction.VECTOR_DELETE)
+        async def vector_delete(data: dict[str, Any]) -> handler.ActionResponse:
+            collection_id = data['collection_id']
+            file_ids = data.get('file_ids')
+            filters = data.get('filters')
+            try:
+                count = await self.ap.rag_runtime_service.vector_delete(collection_id, file_ids, filters)
+                return handler.ActionResponse.success(data={'count': count})
+            except Exception as e:
+                return _make_rag_error_response(e, 'VectorStoreError', collection_id=collection_id)
+
+        @self.action(PluginToRuntimeAction.VECTOR_LIST)
+        async def vector_list(data: dict[str, Any]) -> handler.ActionResponse:
+            collection_id = data['collection_id']
+            filters = data.get('filters')
+            limit = data.get('limit', 20)
+            offset = data.get('offset', 0)
+            try:
+                items, total = await self.ap.rag_runtime_service.vector_list(collection_id, filters, limit, offset)
+                return handler.ActionResponse.success(data={'items': items, 'total': total})
+            except Exception as e:
+                return _make_rag_error_response(e, 'VectorStoreError', collection_id=collection_id)
+
+        @self.action(PluginToRuntimeAction.GET_KNOWLEDEGE_FILE_STREAM)
+        async def get_knowledge_file_stream(data: dict[str, Any]) -> handler.ActionResponse:
+            storage_path = data['storage_path']
+            try:
+                content_bytes = await self.ap.rag_runtime_service.get_file_stream(storage_path)
+                file_key = await self.send_file(content_bytes, '')
+                return handler.ActionResponse.success(data={'file_key': file_key})
+            except Exception as e:
+                return _make_rag_error_response(e, 'FileServiceError', storage_path=storage_path)
+
+        @self.action(PluginToRuntimeAction.LIST_PARSERS)
+        async def list_parsers(data: dict[str, Any]) -> handler.ActionResponse:
+            """Plugin requests host to list available parser plugins."""
+            mime_type = data.get('mime_type')
+            try:
+                parsers = await self.ap.knowledge_service.list_parsers(mime_type)
+                return handler.ActionResponse.success(data={'parsers': parsers})
+            except Exception as e:
+                return _make_rag_error_response(e, 'ParserDiscoveryError', mime_type=mime_type)
+
+        @self.action(PluginToRuntimeAction.INVOKE_PARSER)
+        async def invoke_parser(data: dict[str, Any]) -> handler.ActionResponse:
+            """Plugin requests host to invoke a parser plugin."""
+            plugin_author = data['plugin_author']
+            plugin_name = data['plugin_name']
+            storage_path = data['storage_path']
+            mime_type = data.get('mime_type', 'application/octet-stream')
+            filename = data.get('filename', '')
+            metadata = data.get('metadata', {})
+            try:
+                # Read file from storage
+                file_bytes = await self.ap.rag_runtime_service.get_file_stream(storage_path)
+                context_data = {
+                    'mime_type': mime_type,
+                    'filename': filename,
+                    'metadata': metadata,
+                }
+                result = await self.ap.plugin_connector.call_parser(
+                    f'{plugin_author}/{plugin_name}', context_data, file_bytes
+                )
+                return handler.ActionResponse.success(data=result)
+            except Exception as e:
+                return _make_rag_error_response(e, 'ParserError')
+
+        # ================= Knowledge Base Query APIs =================
+
+        @self.action(PluginToRuntimeAction.LIST_KNOWLEDGE_BASES)
+        async def list_knowledge_bases(data: dict[str, Any]) -> handler.ActionResponse:
+            """List all knowledge bases available in the LangBot instance (unrestricted)."""
+            knowledge_bases = []
+            for kb_uuid, kb in self.ap.rag_mgr.knowledge_bases.items():
+                knowledge_bases.append(
+                    {
+                        'uuid': kb.get_uuid(),
+                        'name': kb.get_name(),
+                        'description': kb.knowledge_base_entity.description or '',
+                    }
+                )
+            return handler.ActionResponse.success(data={'knowledge_bases': knowledge_bases})
+
+        @self.action(PluginToRuntimeAction.RETRIEVE_KNOWLEDGE)
+        async def retrieve_knowledge(data: dict[str, Any]) -> handler.ActionResponse:
+            """Retrieve documents from any knowledge base (unrestricted)."""
+            kb_id = data['kb_id']
+            query_text = data['query_text']
+            top_k = data.get('top_k', 5)
+            filters = data.get('filters', {})
+
+            kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_id)
+            if not kb:
+                return handler.ActionResponse.error(
+                    message=f'Knowledge base {kb_id} not found',
+                )
+
+            try:
+                entries = await kb.retrieve(
+                    query_text,
+                    settings={
+                        'top_k': top_k,
+                        'filters': filters,
+                    },
+                )
+                results = [entry.model_dump(mode='json') for entry in entries]
+                return handler.ActionResponse.success(data={'results': results})
+            except Exception as e:
+                return _make_rag_error_response(e, 'RetrievalError', kb_id=kb_id)
+
+        @self.action(PluginToRuntimeAction.LIST_PIPELINE_KNOWLEDGE_BASES)
+        async def list_pipeline_knowledge_bases(data: dict[str, Any]) -> handler.ActionResponse:
+            """List knowledge bases configured for the current query's pipeline."""
+            query_id = data['query_id']
+
+            if query_id not in self.ap.query_pool.cached_queries:
+                return handler.ActionResponse.error(
+                    message=f'Query with query_id {query_id} not found',
+                )
+
+            query = self.ap.query_pool.cached_queries[query_id]
+
+            kb_uuids = []
+            if query.pipeline_config:
+                local_agent_config = query.pipeline_config.get('ai', {}).get('local-agent', {})
+                kb_uuids = local_agent_config.get('knowledge-bases', [])
+                # Backward compatibility
+                if not kb_uuids:
+                    old_kb_uuid = local_agent_config.get('knowledge-base', '')
+                    if old_kb_uuid and old_kb_uuid != '__none__':
+                        kb_uuids = [old_kb_uuid]
+
+            knowledge_bases = []
+            for kb_uuid in kb_uuids:
+                kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_uuid)
+                if kb:
+                    knowledge_bases.append(
+                        {
+                            'uuid': kb.get_uuid(),
+                            'name': kb.get_name(),
+                            'description': kb.knowledge_base_entity.description or '',
+                        }
+                    )
+
+            return handler.ActionResponse.success(data={'knowledge_bases': knowledge_bases})
+
+        @self.action(PluginToRuntimeAction.RETRIEVE_KNOWLEDGE_BASE)
+        async def retrieve_knowledge_base(data: dict[str, Any]) -> handler.ActionResponse:
+            """Retrieve documents from a knowledge base within the pipeline's scope."""
+            query_id = data['query_id']
+            kb_id = data['kb_id']
+            query_text = data['query_text']
+            top_k = data.get('top_k', 5)
+            filters = data.get('filters', {})
+
+            if query_id not in self.ap.query_pool.cached_queries:
+                return handler.ActionResponse.error(
+                    message=f'Query with query_id {query_id} not found',
+                )
+
+            query = self.ap.query_pool.cached_queries[query_id]
+
+            # Validate kb_id is in pipeline's allowed list
+            allowed_kb_uuids = []
+            if query.pipeline_config:
+                local_agent_config = query.pipeline_config.get('ai', {}).get('local-agent', {})
+                allowed_kb_uuids = local_agent_config.get('knowledge-bases', [])
+                if not allowed_kb_uuids:
+                    old_kb_uuid = local_agent_config.get('knowledge-base', '')
+                    if old_kb_uuid and old_kb_uuid != '__none__':
+                        allowed_kb_uuids = [old_kb_uuid]
+
+            if kb_id not in allowed_kb_uuids:
+                return handler.ActionResponse.error(
+                    message=f'Knowledge base {kb_id} is not configured for this pipeline',
+                )
+
+            kb = await self.ap.rag_mgr.get_knowledge_base_by_uuid(kb_id)
+            if not kb:
+                return handler.ActionResponse.error(
+                    message=f'Knowledge base {kb_id} not found',
+                )
+
+            try:
+                session_name = f'{query.session.launcher_type.value}_{query.session.launcher_id}'
+                entries = await kb.retrieve(
+                    query_text,
+                    settings={
+                        'top_k': top_k,
+                        'filters': filters,
+                        'session_name': session_name,
+                        'bot_uuid': query.bot_uuid or '',
+                        'sender_id': str(query.sender_id),
+                    },
+                )
+                results = [entry.model_dump(mode='json') for entry in entries]
+                return handler.ActionResponse.success(data={'results': results})
+            except Exception as e:
+                return _make_rag_error_response(e, 'RetrievalError', kb_id=kb_id)
+
+        @self.action(CommonAction.PING)
+        async def ping(data: dict[str, Any]) -> handler.ActionResponse:
+            """Ping"""
+            return handler.ActionResponse.success(
+                data={
+                    'pong': 'pong',
+                },
+            )
+
    async def ping(self) -> dict[str, Any]:
        """Ping the runtime"""
        return await self.call_action(
@@ -716,26 +1014,13 @@ class RuntimeConnectionHandler(handler.Handler):
        async for ret in gen:
            yield ret

-    # KnowledgeRetriever methods
-    async def list_knowledge_retrievers(self, include_plugins: list[str] | None = None) -> list[dict[str, Any]]:
-        """List knowledge retrievers"""
-        result = await self.call_action(
-            LangBotToRuntimeAction.LIST_KNOWLEDGE_RETRIEVERS,
-            {
-                'include_plugins': include_plugins,
-            },
-            timeout=10,
-        )
-        return result['retrievers']
-
    async def retrieve_knowledge(
        self,
        plugin_author: str,
        plugin_name: str,
        retriever_name: str,
-        instance_id: str,
        retrieval_context: dict[str, Any],
-    ) -> list[dict[str, Any]]:
+    ) -> dict[str, Any]:
        """Retrieve knowledge"""
        result = await self.call_action(
            LangBotToRuntimeAction.RETRIEVE_KNOWLEDGE,
@@ -743,22 +1028,10 @@ class RuntimeConnectionHandler(handler.Handler):
                'plugin_author': plugin_author,
                'plugin_name': plugin_name,
                'retriever_name': retriever_name,
-                'instance_id': instance_id,
                'retrieval_context': retrieval_context,
            },
            timeout=30,
        )
-        return result['retrieval_results']
-
-    async def sync_polymorphic_component_instances(self, required_instances: list[dict[str, Any]]) -> dict[str, Any]:
-        """Sync polymorphic component instances with runtime"""
-        result = await self.call_action(
-            LangBotToRuntimeAction.SYNC_POLYMORPHIC_COMPONENT_INSTANCES,
-            {
-                'required_instances': required_instances,
-            },
-            timeout=30,
-        )
        return result

    async def get_debug_info(self) -> dict[str, Any]:
@@ -769,3 +1042,91 @@ class RuntimeConnectionHandler(handler.Handler):
            timeout=10,
        )
        return result
+
+    # ================= RAG Capability Callers (LangBot -> Runtime) =================
+
+    async def rag_ingest_document(
+        self, plugin_author: str, plugin_name: str, context_data: dict[str, Any]
+    ) -> dict[str, Any]:
+        """Send INGEST_DOCUMENT action to runtime."""
+        result = await self.call_action(
+            LangBotToRuntimeAction.RAG_INGEST_DOCUMENT,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name, 'context': context_data},
+            timeout=1200,  # Ingestion can be slow for large documents
+        )
+        return result
+
+    async def rag_delete_document(self, plugin_author: str, plugin_name: str, document_id: str, kb_id: str) -> bool:
+        result = await self.call_action(
+            LangBotToRuntimeAction.RAG_DELETE_DOCUMENT,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name, 'document_id': document_id, 'kb_id': kb_id},
+            timeout=30,
+        )
+        return result.get('success', False)
+
+    async def rag_on_kb_create(
+        self, plugin_author: str, plugin_name: str, kb_id: str, config: dict[str, Any]
+    ) -> dict[str, Any]:
+        """Notify plugin about KB creation."""
+        result = await self.call_action(
+            LangBotToRuntimeAction.RAG_ON_KB_CREATE,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name, 'kb_id': kb_id, 'config': config},
+            timeout=30,
+        )
+        return result
+
+    async def rag_on_kb_delete(self, plugin_author: str, plugin_name: str, kb_id: str) -> dict[str, Any]:
+        """Notify plugin about KB deletion."""
+        result = await self.call_action(
+            LangBotToRuntimeAction.RAG_ON_KB_DELETE,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name, 'kb_id': kb_id},
+            timeout=30,
+        )
+        return result
+
+    async def get_rag_creation_schema(self, plugin_author: str, plugin_name: str) -> dict[str, Any]:
+        return await self.call_action(
+            LangBotToRuntimeAction.GET_RAG_CREATION_SETTINGS_SCHEMA,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name},
+            timeout=10,
+        )
+
+    async def get_rag_retrieval_schema(self, plugin_author: str, plugin_name: str) -> dict[str, Any]:
+        return await self.call_action(
+            LangBotToRuntimeAction.GET_RAG_RETRIEVAL_SETTINGS_SCHEMA,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name},
+            timeout=10,
+        )
+
+    async def list_knowledge_engines(self) -> list[dict[str, Any]]:
+        """List all available Knowledge Engines from plugins."""
+        result = await self.call_action(LangBotToRuntimeAction.LIST_KNOWLEDGE_ENGINES, {}, timeout=60)
+        return result.get('engines', [])
+
+    # ================= Parser Capability Callers (LangBot -> Runtime) =================
+
+    async def list_parsers(self) -> list[dict[str, Any]]:
+        """List all available parsers from plugins."""
+        result = await self.call_action(LangBotToRuntimeAction.LIST_PARSERS, {}, timeout=60)
+        return result.get('parsers', [])
+
+    async def parse_document(
+        self, plugin_author: str, plugin_name: str, context_data: dict[str, Any], file_bytes: bytes
+    ) -> dict[str, Any]:
+        """Send PARSE_DOCUMENT action to runtime.
+
+        Sends file content via chunked FILE_CHUNK transfer, then invokes
+        the PARSE_DOCUMENT action with a file_key reference.
+        """
+        # Send file to runtime via chunked transfer
+        file_key = await self.send_file(file_bytes, '')
+
+        # Include file_key in context_data for the runtime to read
+        context_data['file_key'] = file_key
+
+        result = await self.call_action(
+            LangBotToRuntimeAction.PARSE_DOCUMENT,
+            {'plugin_author': plugin_author, 'plugin_name': plugin_name, 'context': context_data},
+            timeout=300,
+        )
+        return result
--- a/src/langbot/pkg/provider/modelmgr/requesters/anthropicmsgs.py
+++ b/src/langbot/pkg/provider/modelmgr/requesters/anthropicmsgs.py
@@ -288,10 +288,10 @@ class AnthropicMessages(requester.ProviderAPIRequester):
            think_started = False
            think_ended = False
            finish_reason = False
-            content = ''
            tool_name = ''
            tool_id = ''
            async for chunk in await self.client.messages.create(**args):
+                content = ''
                tool_call = {'id': None, 'function': {'name': None, 'arguments': None}, 'type': 'function'}
                if isinstance(
                    chunk, anthropic.types.raw_content_block_start_event.RawContentBlockStartEvent
--- a/src/langbot/pkg/provider/runners/difysvapi.py
+++ b/src/langbot/pkg/provider/runners/difysvapi.py
@@ -72,6 +72,28 @@ class DifyServiceAPIRunner(runner.RequestRunner):
                content = f'<think>\n{thinking_content}\n</think>\n{content}'.strip()
            return content, thinking_content

+    def _extract_dify_text_output(self, value: typing.Any) -> str:
+        """Extract text content from Dify output payload."""
+        if value is None:
+            return ''
+        if isinstance(value, dict):
+            content = value.get('content')
+            if isinstance(content, str):
+                return content
+            return json.dumps(value, ensure_ascii=False)
+        if isinstance(value, str):
+            text = value.strip()
+            if not text:
+                return ''
+            try:
+                parsed = json.loads(text)
+            except json.JSONDecodeError:
+                return value
+            if isinstance(parsed, dict) and isinstance(parsed.get('content'), str):
+                return parsed['content']
+            return value
+        return str(value)
+
    async def _preprocess_user_message(self, query: pipeline_query.Query) -> tuple[str, list[dict]]:
        """预处理用户消息，提取纯文本，并将图片/文件上传到 Dify 服务

@@ -192,7 +214,8 @@ class DifyServiceAPIRunner(runner.RequestRunner):
            if mode == 'workflow':
                if chunk['event'] == 'node_finished':
                    if chunk['data']['node_type'] == 'answer':
-                        content, _ = self._process_thinking_content(chunk['data']['outputs']['answer'])
+                        answer = self._extract_dify_text_output(chunk['data']['outputs'].get('answer'))
+                        content, _ = self._process_thinking_content(answer)

                        yield provider_message.Message(
                            role='assistant',
@@ -405,6 +428,7 @@ class DifyServiceAPIRunner(runner.RequestRunner):
            for f in upload_files
        ]

+        mode = 'basic'
        basic_mode_pending_chunk = ''

        inputs = {}
@@ -417,6 +441,7 @@ class DifyServiceAPIRunner(runner.RequestRunner):
        is_final = False
        think_start = False
        think_end = False
+        yielded_final = False

        remove_think = self.pipeline_config['output'].get('misc', '').get('remove-think')

@@ -430,11 +455,12 @@ class DifyServiceAPIRunner(runner.RequestRunner):
        ):
            self.ap.logger.debug('dify-chat-chunk: ' + str(chunk))

-            # if chunk['event'] == 'workflow_started':
-            #     mode = 'workflow'
-            # if mode == 'workflow':
-            # elif mode == 'basic':
-            # 因为都只是返回的 message也没有工具调用什么的，暂时不分类
+            if chunk['event'] == 'workflow_started':
+                mode = 'workflow'
+            elif chunk['event'] in ('node_started', 'node_finished', 'workflow_finished'):
+                # Some Dify deployments may omit workflow_started in streamed chunks.
+                mode = 'workflow'
+
            if chunk['event'] == 'message':
                message_idx += 1
                if remove_think:
@@ -457,14 +483,30 @@ class DifyServiceAPIRunner(runner.RequestRunner):

            if chunk['event'] == 'message_end':
                is_final = True
+            elif chunk['event'] == 'workflow_finished':
+                is_final = True
+                if chunk['data'].get('error'):
+                    raise errors.DifyAPIError(chunk['data']['error'])

-            if is_final or message_idx % 8 == 0:
+            if mode == 'workflow' and chunk['event'] == 'node_finished':
+                if chunk['data'].get('node_type') == 'answer':
+                    answer = self._extract_dify_text_output(chunk['data'].get('outputs', {}).get('answer'))
+                    if answer:
+                        basic_mode_pending_chunk = answer
+
+            if (
+                not yielded_final
+                and (is_final or message_idx % 8 == 0)
+                and (basic_mode_pending_chunk != '' or is_final)
+            ):
                # content, _ = self._process_thinking_content(basic_mode_pending_chunk)
                yield provider_message.MessageChunk(
                    role='assistant',
                    content=basic_mode_pending_chunk,
                    is_final=is_final,
                )
+                if is_final:
+                    yielded_final = True

        if chunk is None:
            raise errors.DifyAPIError('Dify API 没有返回任何响应，请检查网络连接和API配置')
--- a/src/langbot/pkg/provider/runners/localagent.py
+++ b/src/langbot/pkg/provider/runners/localagent.py
@@ -4,6 +4,7 @@ import json
 import copy
 import typing
 from .. import runner
+from ..modelmgr import requester as modelmgr_requester
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
 import langbot_plugin.api.entities.builtin.provider.message as provider_message
 import langbot_plugin.api.entities.builtin.rag.context as rag_context
@@ -26,29 +27,114 @@ Respond in the same language as the user's input.

@runner.runner_class('local-agent')
 class LocalAgentRunner(runner.RequestRunner):
-    """本地Agent请求运行器"""
+    """Local agent request runner"""

-    class ToolCallTracker:
-        """工具调用追踪器"""
+    async def _get_model_candidates(
+        self,
+        query: pipeline_query.Query,
+    ) -> list[modelmgr_requester.RuntimeLLMModel]:
+        """Build ordered list of models to try: primary model + fallback models."""
+        candidates = []

-        def __init__(self):
-            self.active_calls: dict[str, dict] = {}
-            self.completed_calls: list[provider_message.ToolCall] = []
+        # Primary model
+        if query.use_llm_model_uuid:
+            try:
+                primary = await self.ap.model_mgr.get_model_by_uuid(query.use_llm_model_uuid)
+                candidates.append(primary)
+            except ValueError:
+                self.ap.logger.warning(f'Primary model {query.use_llm_model_uuid} not found')
+
+        # Fallback models
+        fallback_uuids = (query.variables or {}).get('_fallback_model_uuids', [])
+        for fb_uuid in fallback_uuids:
+            try:
+                fb_model = await self.ap.model_mgr.get_model_by_uuid(fb_uuid)
+                candidates.append(fb_model)
+            except ValueError:
+                self.ap.logger.warning(f'Fallback model {fb_uuid} not found, skipping')
+
+        return candidates
+
+    async def _invoke_with_fallback(
+        self,
+        query: pipeline_query.Query,
+        candidates: list[modelmgr_requester.RuntimeLLMModel],
+        messages: list,
+        funcs: list,
+        remove_think: bool,
+    ) -> tuple[provider_message.Message, modelmgr_requester.RuntimeLLMModel]:
+        """Try non-streaming invocation with sequential fallback. Returns (message, model_used)."""
+        last_error = None
+        for model in candidates:
+            try:
+                msg = await model.provider.invoke_llm(
+                    query,
+                    model,
+                    messages,
+                    funcs if model.model_entity.abilities.__contains__('func_call') else [],
+                    extra_args=model.model_entity.extra_args,
+                    remove_think=remove_think,
+                )
+                return msg, model
+            except Exception as e:
+                last_error = e
+                self.ap.logger.warning(f'Model {model.model_entity.name} failed: {e}, trying next fallback...')
+        raise last_error or RuntimeError('No model candidates available')
+
+    async def _invoke_stream_with_fallback(
+        self,
+        query: pipeline_query.Query,
+        candidates: list[modelmgr_requester.RuntimeLLMModel],
+        messages: list,
+        funcs: list,
+        remove_think: bool,
+    ) -> tuple[typing.AsyncGenerator, modelmgr_requester.RuntimeLLMModel]:
+        """Try streaming invocation with sequential fallback. Returns (stream_generator, model_used).
+
+        Fallback is only possible before any chunks have been yielded to the client.
+        Once streaming starts, the model is committed.
+        """
+        last_error = None
+        for model in candidates:
+            try:
+                stream = model.provider.invoke_llm_stream(
+                    query,
+                    model,
+                    messages,
+                    funcs if model.model_entity.abilities.__contains__('func_call') else [],
+                    extra_args=model.model_entity.extra_args,
+                    remove_think=remove_think,
+                )
+                # Attempt to get the first chunk to verify the stream works
+                first_chunk = await stream.__anext__()
+
+                async def _chain_stream(first, rest):
+                    yield first
+                    async for chunk in rest:
+                        yield chunk
+
+                return _chain_stream(first_chunk, stream), model
+            except StopAsyncIteration:
+                # Empty stream — treat as success (model returned nothing)
+                async def _empty_stream():
+                    return
+                    yield  # make it a generator
+
+                return _empty_stream(), model
+            except Exception as e:
+                last_error = e
+                self.ap.logger.warning(f'Model {model.model_entity.name} stream failed: {e}, trying next fallback...')
+        raise last_error or RuntimeError('No model candidates available')

    async def run(
        self, query: pipeline_query.Query
    ) -> typing.AsyncGenerator[provider_message.Message | provider_message.MessageChunk, None]:
-        """运行请求"""
+        """Run request"""
        pending_tool_calls = []

-        # Get knowledge bases list (new field)
-        kb_uuids = query.pipeline_config['ai']['local-agent'].get('knowledge-bases', [])
-
-        # Fallback to old field for backward compatibility
-        if not kb_uuids:
-            old_kb_uuid = query.pipeline_config['ai']['local-agent'].get('knowledge-base', '')
-            if old_kb_uuid and old_kb_uuid != '__none__':
-                kb_uuids = [old_kb_uuid]
+        # Get knowledge bases list from query variables (set by PreProcessor,
+        # may have been modified by plugins during PromptPreProcessing)
+        kb_uuids = query.variables.get('_knowledge_base_uuids', [])

        user_message = copy.deepcopy(query.user_message)

@@ -74,15 +160,14 @@ class LocalAgentRunner(runner.RequestRunner):
                    self.ap.logger.warning(f'Knowledge base {kb_uuid} not found, skipping')
                    continue

-                # Get top_k based on KB type
-                if kb.get_type() == 'internal':
-                    top_k = kb.knowledge_base_entity.top_k
-                elif kb.get_type() == 'external':
-                    top_k = 5  # external kb's top_k is managed by plugin config
-                else:
-                    top_k = 5  # default fallback
-
-                result = await kb.retrieve(user_message_text, top_k)
+                result = await kb.retrieve(
+                    user_message_text,
+                    settings={
+                        'bot_uuid': query.bot_uuid or '',
+                        'sender_id': str(query.sender_id),
+                        'session_name': f'{query.session.launcher_type.value}_{query.session.launcher_id}',
+                    },
+                )

                if result:
                    all_results.extend(result)
@@ -97,9 +182,9 @@ class LocalAgentRunner(runner.RequestRunner):
                        if content.type == 'text' and content.text is not None:
                            texts.append(f'[{idx}] {content.text}')
                            idx += 1
-                rag_context = '\n\n'.join(texts)
+                rag_context_text = '\n\n'.join(texts)
                final_user_message_text = rag_combined_prompt_template.format(
-                    rag_context=rag_context, user_message=user_message_text
+                    rag_context=rag_context_text, user_message=user_message_text
                )

            else:
@@ -121,51 +206,51 @@ class LocalAgentRunner(runner.RequestRunner):

        remove_think = query.pipeline_config['output'].get('misc', '').get('remove-think')

-        use_llm_model = await self.ap.model_mgr.get_model_by_uuid(query.use_llm_model_uuid)
+        # Build ordered candidate list (primary + fallbacks)
+        candidates = await self._get_model_candidates(query)
+        if not candidates:
+            raise RuntimeError('No LLM model configured for local-agent runner')

        self.ap.logger.debug(
-            f'localagent req: query={query.query_id} req_messages={req_messages} use_llm_model={query.use_llm_model_uuid}'
+            f'localagent req: query={query.query_id} req_messages={req_messages} '
+            f'candidates={[m.model_entity.name for m in candidates]}'
        )

        if not is_stream:
-            # 非流式输出，直接请求
-
-            msg = await use_llm_model.provider.invoke_llm(
+            # Non-streaming: invoke with fallback
+            msg, use_llm_model = await self._invoke_with_fallback(
                query,
-                use_llm_model,
+                candidates,
                req_messages,
                query.use_funcs,
-                extra_args=use_llm_model.model_entity.extra_args,
-                remove_think=remove_think,
+                remove_think,
            )
            yield msg
            final_msg = msg
        else:
-            # 流式输出，需要处理工具调用
+            # Streaming: invoke with fallback
            tool_calls_map: dict[str, provider_message.ToolCall] = {}
            msg_idx = 0
-            accumulated_content = ''  # 从开始累积的所有内容
+            accumulated_content = ''
            last_role = 'assistant'
            msg_sequence = 1
-            async for msg in use_llm_model.provider.invoke_llm_stream(
+
+            stream_src, use_llm_model = await self._invoke_stream_with_fallback(
                query,
-                use_llm_model,
+                candidates,
                req_messages,
                query.use_funcs,
-                extra_args=use_llm_model.model_entity.extra_args,
-                remove_think=remove_think,
-            ):
+                remove_think,
+            )
+            async for msg in stream_src:
                msg_idx = msg_idx + 1

-                # 记录角色
                if msg.role:
                    last_role = msg.role

-                # 累积内容
                if msg.content:
                    accumulated_content += msg.content

-                # 处理工具调用
                if msg.tool_calls:
                    for tool_call in msg.tool_calls:
                        if tool_call.id not in tool_calls_map:
@@ -177,21 +262,18 @@ class LocalAgentRunner(runner.RequestRunner):
                                ),
                            )
                        if tool_call.function and tool_call.function.arguments:
-                            # 流式处理中，工具调用参数可能分多个chunk返回，需要追加而不是覆盖
                            tool_calls_map[tool_call.id].function.arguments += tool_call.function.arguments
-                # continue
-                # 每8个chunk或最后一个chunk时，输出所有累积的内容
+
                if msg_idx % 8 == 0 or msg.is_final:
                    msg_sequence += 1
                    yield provider_message.MessageChunk(
                        role=last_role,
-                        content=accumulated_content,  # 输出所有累积内容
+                        content=accumulated_content,
                        tool_calls=list(tool_calls_map.values()) if (tool_calls_map and msg.is_final) else None,
                        is_final=msg.is_final,
                        msg_sequence=msg_sequence,
                    )

-            # 创建最终消息用于后续处理
            final_msg = provider_message.MessageChunk(
                role=last_role,
                content=accumulated_content,
@@ -206,7 +288,8 @@ class LocalAgentRunner(runner.RequestRunner):

        req_messages.append(final_msg)

-        # 持续请求，只要还有待处理的工具调用就继续处理调用
+        # Once a model succeeds, commit to it for the tool call loop
+        # (no fallback mid-conversation — different models may interpret tool results differently)
        while pending_tool_calls:
            for tool_call in pending_tool_calls:
                try:
@@ -247,7 +330,6 @@ class LocalAgentRunner(runner.RequestRunner):

                    req_messages.append(msg)
                except Exception as e:
-                    # 工具调用出错，添加一个报错信息到 req_messages
                    err_msg = provider_message.Message(role='tool', content=f'err: {e}', tool_call_id=tool_call.id)

                    yield err_msg
@@ -255,39 +337,38 @@ class LocalAgentRunner(runner.RequestRunner):
                    req_messages.append(err_msg)

            self.ap.logger.debug(
-                f'localagent req: query={query.query_id} req_messages={req_messages} use_llm_model={query.use_llm_model_uuid}'
+                f'localagent req: query={query.query_id} req_messages={req_messages} '
+                f'use_llm_model={use_llm_model.model_entity.name}'
            )

            if is_stream:
                tool_calls_map = {}
                msg_idx = 0
-                accumulated_content = ''  # 从开始累积的所有内容
+                accumulated_content = ''
                last_role = 'assistant'
                msg_sequence = first_end_sequence

-                async for msg in use_llm_model.provider.invoke_llm_stream(
+                tool_stream_src = use_llm_model.provider.invoke_llm_stream(
                    query,
                    use_llm_model,
                    req_messages,
-                    query.use_funcs,
+                    query.use_funcs if use_llm_model.model_entity.abilities.__contains__('func_call') else [],
                    extra_args=use_llm_model.model_entity.extra_args,
                    remove_think=remove_think,
-                ):
+                )
+                async for msg in tool_stream_src:
                    msg_idx += 1

-                    # 记录角色
                    if msg.role:
                        last_role = msg.role

-                    # 第一次请求工具调用时的内容
+                    # Prepend first-round content on first chunk of tool-call round
                    if msg_idx == 1:
                        accumulated_content = first_content if first_content is not None else accumulated_content

-                    # 累积内容
                    if msg.content:
                        accumulated_content += msg.content

-                    # 处理工具调用
                    if msg.tool_calls:
                        for tool_call in msg.tool_calls:
                            if tool_call.id not in tool_calls_map:
@@ -299,15 +380,13 @@ class LocalAgentRunner(runner.RequestRunner):
                                    ),
                                )
                            if tool_call.function and tool_call.function.arguments:
-                                # 流式处理中，工具调用参数可能分多个chunk返回，需要追加而不是覆盖
                                tool_calls_map[tool_call.id].function.arguments += tool_call.function.arguments

-                    # 每8个chunk或最后一个chunk时，输出所有累积的内容
                    if msg_idx % 8 == 0 or msg.is_final:
                        msg_sequence += 1
                        yield provider_message.MessageChunk(
                            role=last_role,
-                            content=accumulated_content,  # 输出所有累积内容
+                            content=accumulated_content,
                            tool_calls=list(tool_calls_map.values()) if (tool_calls_map and msg.is_final) else None,
                            is_final=msg.is_final,
                            msg_sequence=msg_sequence,
@@ -320,12 +399,12 @@ class LocalAgentRunner(runner.RequestRunner):
                    msg_sequence=msg_sequence,
                )
            else:
-                # 处理完所有调用，再次请求
+                # Non-streaming: use committed model directly (no fallback in tool loop)
                msg = await use_llm_model.provider.invoke_llm(
                    query,
                    use_llm_model,
                    req_messages,
-                    query.use_funcs,
+                    query.use_funcs if use_llm_model.model_entity.abilities.__contains__('func_call') else [],
                    extra_args=use_llm_model.model_entity.extra_args,
                    remove_think=remove_think,
                )
--- a/src/langbot/pkg/provider/runners/n8nsvapi.py
+++ b/src/langbot/pkg/provider/runners/n8nsvapi.py
@@ -5,6 +5,8 @@ import json
 import uuid
 import aiohttp

+from langbot.pkg.utils import httpclient
+
 from .. import runner
 from ...core import app
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
@@ -217,50 +219,50 @@ class N8nServiceAPIRunner(runner.RequestRunner):
                self.ap.logger.debug('no auth')

            # 调用webhook
-            async with aiohttp.ClientSession() as session:
-                if is_stream:
-                    # 流式请求
-                    async with session.post(
-                        self.webhook_url, json=payload, headers=headers, auth=auth, timeout=self.timeout
-                    ) as response:
+            session = httpclient.get_session()
+            if is_stream:
+                # 流式请求
+                async with session.post(
+                    self.webhook_url, json=payload, headers=headers, auth=auth, timeout=self.timeout
+                ) as response:
+                    if response.status != 200:
+                        error_text = await response.text()
+                        self.ap.logger.error(f'n8n webhook call failed: {response.status}, {error_text}')
+                        raise Exception(f'n8n webhook call failed: {response.status}, {error_text}')
+
+                    # 处理流式响应
+                    async for chunk in self._process_stream_response(response):
+                        yield chunk
+            else:
+                async with session.post(
+                    self.webhook_url, json=payload, headers=headers, auth=auth, timeout=self.timeout
+                ) as response:
+                    try:
+                        async for chunk in self._process_stream_response(response):
+                            output_content = chunk.content if chunk.is_final else ''
+                    except:
+                        # 非流式请求（保持原有逻辑）
                        if response.status != 200:
                            error_text = await response.text()
                            self.ap.logger.error(f'n8n webhook call failed: {response.status}, {error_text}')
                            raise Exception(f'n8n webhook call failed: {response.status}, {error_text}')

-                        # 处理流式响应
-                        async for chunk in self._process_stream_response(response):
-                            yield chunk
-                else:
-                    async with session.post(
-                        self.webhook_url, json=payload, headers=headers, auth=auth, timeout=self.timeout
-                    ) as response:
-                        try:
-                            async for chunk in self._process_stream_response(response):
-                                output_content = chunk.content if chunk.is_final else ''
-                        except:
-                            # 非流式请求（保持原有逻辑）
-                            if response.status != 200:
-                                error_text = await response.text()
-                                self.ap.logger.error(f'n8n webhook call failed: {response.status}, {error_text}')
-                                raise Exception(f'n8n webhook call failed: {response.status}, {error_text}')
+                        # 解析响应
+                        response_data = await response.json()
+                        self.ap.logger.debug(f'n8n webhook response: {response_data}')

-                            # 解析响应
-                            response_data = await response.json()
-                            self.ap.logger.debug(f'n8n webhook response: {response_data}')
+                        # 从响应中提取输出
+                        if self.output_key in response_data:
+                            output_content = response_data[self.output_key]
+                        else:
+                            # 如果没有指定的输出键，则使用整个响应
+                            output_content = json.dumps(response_data, ensure_ascii=False)

-                            # 从响应中提取输出
-                            if self.output_key in response_data:
-                                output_content = response_data[self.output_key]
-                            else:
-                                # 如果没有指定的输出键，则使用整个响应
-                                output_content = json.dumps(response_data, ensure_ascii=False)
-
-                        # 返回消息
-                        yield provider_message.Message(
-                            role='assistant',
-                            content=output_content,
-                        )
+                    # 返回消息
+                    yield provider_message.Message(
+                        role='assistant',
+                        content=output_content,
+                    )
        except Exception as e:
            self.ap.logger.error(f'n8n webhook call exception: {str(e)}')
            raise N8nAPIError(f'n8n webhook call exception: {str(e)}')
--- a/src/langbot/pkg/rag/knowledge/base.py
+++ b/src/langbot/pkg/rag/knowledge/base.py
@@ -22,12 +22,12 @@ class KnowledgeBaseInterface(metaclass=abc.ABCMeta):
        pass

    @abc.abstractmethod
-    async def retrieve(self, query: str, top_k: int) -> list[rag_context.RetrievalResultEntry]:
+    async def retrieve(self, query: str, settings: dict | None = None) -> list[rag_context.RetrievalResultEntry]:
        """Retrieve relevant documents from the knowledge base

        Args:
            query: The query string
-            top_k: Number of top results to return
+            settings: Optional per-request retrieval settings overrides

        Returns:
            List of retrieve result entries
@@ -45,8 +45,8 @@ class KnowledgeBaseInterface(metaclass=abc.ABCMeta):
        pass

    @abc.abstractmethod
-    def get_type(self) -> str:
-        """Get the type of knowledge base (internal/external)"""
+    def get_knowledge_engine_plugin_id(self) -> str:
+        """Get the Knowledge Engine plugin ID"""
        pass

    @abc.abstractmethod
--- a/src/langbot/pkg/rag/knowledge/external.py
+++ b/src/langbot/pkg/rag/knowledge/external.py
@@ -1,85 +0,0 @@
-"""External knowledge base implementation"""
-
-from __future__ import annotations
-
-from langbot.pkg.core import app
-from langbot.pkg.entity.persistence import rag as persistence_rag
-from langbot_plugin.api.entities.builtin.rag import context as rag_context
-from .base import KnowledgeBaseInterface
-
-
-class ExternalKnowledgeBase(KnowledgeBaseInterface):
-    """External knowledge base that queries via HTTP API or plugin retriever"""
-
-    external_kb_entity: persistence_rag.ExternalKnowledgeBase
-
-    # Plugin retriever instance ID
-    retriever_instance_id: str | None
-
-    def __init__(self, ap: app.Application, external_kb_entity: persistence_rag.ExternalKnowledgeBase):
-        super().__init__(ap)
-        self.external_kb_entity = external_kb_entity
-        self.retriever_instance_id = None
-
-    async def initialize(self):
-        """Initialize the external knowledge base"""
-        # Use KB UUID as instance ID
-        # Instance creation is now handled by the unified sync mechanism
-        # when LangBot connects to runtime
-        self.retriever_instance_id = self.external_kb_entity.uuid
-
-        self.ap.logger.info(
-            f'Initialized external KB {self.external_kb_entity.uuid}, instance will be created by sync mechanism'
-        )
-
-    async def retrieve(self, query: str, top_k: int = 5) -> list[rag_context.RetrievalResultEntry]:
-        """Retrieve documents from external knowledge base via plugin retriever"""
-        if not self.retriever_instance_id:
-            self.ap.logger.error(f'No retriever instance for KB {self.external_kb_entity.uuid}')
-            return []
-
-        try:
-            results = await self.ap.plugin_connector.retrieve_knowledge(
-                self.external_kb_entity.plugin_author,
-                self.external_kb_entity.plugin_name,
-                self.external_kb_entity.retriever_name,
-                self.retriever_instance_id,
-                {'query': query},
-            )
-
-            # Convert plugin results to RetrievalResultEntry
-            retrieval_entries = []
-            for result in results:
-                retrieval_entries.append(rag_context.RetrievalResultEntry(**result))
-
-            return retrieval_entries
-        except Exception as e:
-            self.ap.logger.error(f'Plugin retriever error: {e}')
-            import traceback
-
-            traceback.print_exc()
-            return []
-
-    def get_uuid(self) -> str:
-        """Get the UUID of the external knowledge base"""
-        return self.external_kb_entity.uuid
-
-    def get_name(self) -> str:
-        """Get the name of the external knowledge base"""
-        return self.external_kb_entity.name
-
-    def get_type(self) -> str:
-        """Get the type of knowledge base"""
-        return 'external'
-
-    async def dispose(self):
-        """Clean up resources"""
-        # Trigger sync to immediately delete the instance from plugin process
-        # This ensures instance is cleaned up without waiting for next LangBot restart
-        try:
-            await self.ap.plugin_connector.sync_polymorphic_component_instances()
-            self.ap.logger.info(
-                f'Disposed external KB {self.external_kb_entity.uuid}, triggered sync to delete instance'
-            )
-        except Exception as e:
-            self.ap.logger.error(f'Failed to sync after disposing KB: {e}')
--- a/src/langbot/pkg/rag/knowledge/kbmgr.py
+++ b/src/langbot/pkg/rag/knowledge/kbmgr.py
@@ -1,18 +1,19 @@
 from __future__ import annotations
+import mimetypes
+import os.path
 import traceback
 import uuid
 import zipfile
 import io
-from .services import parser, chunker
+from typing import Any
 from langbot.pkg.core import app
-from langbot.pkg.rag.knowledge.services.embedder import Embedder
-from langbot.pkg.rag.knowledge.services.retriever import Retriever
 import sqlalchemy
+
+
 from langbot.pkg.entity.persistence import rag as persistence_rag
 from langbot.pkg.core import taskmgr
 from langbot_plugin.api.entities.builtin.rag import context as rag_context
 from .base import KnowledgeBaseInterface
-from .external import ExternalKnowledgeBase


 class RuntimeKnowledgeBase(KnowledgeBaseInterface):
@@ -20,28 +21,16 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):

    knowledge_base_entity: persistence_rag.KnowledgeBase

-    parser: parser.FileParser
-
-    chunker: chunker.Chunker
-
-    embedder: Embedder
-
-    retriever: Retriever
-
    def __init__(self, ap: app.Application, knowledge_base_entity: persistence_rag.KnowledgeBase):
        super().__init__(ap)
        self.knowledge_base_entity = knowledge_base_entity
-        self.parser = parser.FileParser(ap=self.ap)
-        self.chunker = chunker.Chunker(ap=self.ap)
-        self.embedder = Embedder(ap=self.ap)
-        self.retriever = Retriever(ap=self.ap)
-        # 传递kb_id给retriever
-        self.retriever.kb_id = knowledge_base_entity.uuid

    async def initialize(self):
        pass

-    async def _store_file_task(self, file: persistence_rag.File, task_context: taskmgr.TaskContext):
+    async def _store_file_task(
+        self, file: persistence_rag.File, task_context: taskmgr.TaskContext, parser_plugin_id: str | None = None
+    ):
        try:
            # set file status to processing
            await self.ap.persistence_mgr.execute_async(
@@ -50,31 +39,46 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
                .values(status='processing')
            )

-            task_context.set_current_action('Parsing file')
-            # parse file
-            text = await self.parser.parse(file.file_name, file.extension)
-            if not text:
-                raise Exception(f'No text extracted from file {file.file_name}')
+            task_context.set_current_action('Processing file')

-            task_context.set_current_action('Chunking file')
-            # chunk file
-            chunks_texts = await self.chunker.chunk(text)
-            if not chunks_texts:
-                raise Exception(f'No chunks extracted from file {file.file_name}')
+            # Get file size from storage
+            file_size = await self.ap.storage_mgr.storage_provider.size(file.file_name)

-            task_context.set_current_action('Embedding chunks')
+            # Detect MIME type from extension
+            mime_type, _ = mimetypes.guess_type(file.file_name)
+            if mime_type is None:
+                mime_type = 'application/octet-stream'

-            embedding_model = await self.ap.model_mgr.get_embedding_model_by_uuid(
-                self.knowledge_base_entity.embedding_model_uuid
-            )
-            # embed chunks
-            await self.embedder.embed_and_store(
-                kb_id=self.knowledge_base_entity.uuid,
-                file_id=file.uuid,
-                chunks=chunks_texts,
-                embedding_model=embedding_model,
+            # If a parser plugin is specified, call it before ingestion
+            parsed_content = None
+            if parser_plugin_id:
+                task_context.set_current_action('Parsing file')
+                file_bytes = await self.ap.storage_mgr.storage_provider.load(file.file_name)
+                parse_context = {
+                    'mime_type': mime_type,
+                    'filename': file.file_name,
+                    'metadata': {},
+                }
+                parsed_content = await self.ap.plugin_connector.call_parser(parser_plugin_id, parse_context, file_bytes)
+
+            # Call plugin to ingest document
+            result = await self._ingest_document(
+                {
+                    'document_id': file.uuid,
+                    'filename': file.file_name,
+                    'extension': file.extension,
+                    'file_size': file_size,
+                    'mime_type': mime_type,
+                },
+                file.file_name,  # storage path
+                parsed_content=parsed_content,
            )

+            # Check plugin result status
+            if result.get('status') == 'failed':
+                error_msg = result.get('error_message', 'Plugin ingestion returned failed status')
+                raise Exception(error_msg)
+
            # set file status to completed
            await self.ap.persistence_mgr.execute_async(
                sqlalchemy.update(persistence_rag.File)
@@ -97,16 +101,17 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
            # delete file from storage
            await self.ap.storage_mgr.storage_provider.delete(file.file_name)

-    async def store_file(self, file_id: str) -> str:
+    async def store_file(self, file_id: str, parser_plugin_id: str | None = None) -> str:
        # pre checking
        if not await self.ap.storage_mgr.storage_provider.exists(file_id):
            raise Exception(f'File {file_id} not found')

        file_name = file_id
-        extension = file_name.split('.')[-1].lower()
+        _, ext = os.path.splitext(file_name)
+        extension = ext.lstrip('.').lower() if ext else ''

        if extension == 'zip':
-            return await self._store_zip_file(file_id)
+            return await self._store_zip_file(file_id, parser_plugin_id=parser_plugin_id)

        file_uuid = str(uuid.uuid4())
        kb_id = self.knowledge_base_entity.uuid
@@ -126,7 +131,7 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
        # run background task asynchronously
        ctx = taskmgr.TaskContext.new()
        wrapper = self.ap.task_mgr.create_user_task(
-            self._store_file_task(file_obj, task_context=ctx),
+            self._store_file_task(file_obj, task_context=ctx, parser_plugin_id=parser_plugin_id),
            kind='knowledge-operation',
            name=f'knowledge-store-file-{file_id}',
            label=f'Store file {file_id}',
@@ -134,7 +139,7 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
        )
        return wrapper.id

-    async def _store_zip_file(self, zip_file_id: str) -> str:
+    async def _store_zip_file(self, zip_file_id: str, parser_plugin_id: str | None = None) -> str:
        """Handle ZIP file by extracting each document and storing them separately."""
        self.ap.logger.info(f'Processing ZIP file: {zip_file_id}')

@@ -150,7 +155,8 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
                if file_info.is_dir() or file_info.filename.startswith('.'):
                    continue

-                file_extension = file_info.filename.split('.')[-1].lower()
+                _, file_ext = os.path.splitext(file_info.filename)
+                file_extension = file_ext.lstrip('.').lower()
                if file_extension not in supported_extensions:
                    self.ap.logger.debug(f'Skipping unsupported file in ZIP: {file_info.filename}')
                    continue
@@ -159,18 +165,18 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
                    file_content = zip_ref.read(file_info.filename)

                    base_name = file_info.filename.replace('/', '_').replace('\\', '_')
-                    extension = base_name.split('.')[-1]
-                    file_name = base_name.split('.')[0]
+                    file_stem, file_ext = os.path.splitext(base_name)
+                    extension = file_ext.lstrip('.')

-                    if file_name.startswith('__MACOSX'):
+                    if file_stem.startswith('__MACOSX'):
                        continue

-                    extracted_file_id = file_name + '_' + str(uuid.uuid4())[:8] + '.' + extension
+                    extracted_file_id = file_stem + '_' + str(uuid.uuid4())[:8] + '.' + extension
                    # save file to storage

                    await self.ap.storage_mgr.storage_provider.save(extracted_file_id, file_content)

-                    task_id = await self.store_file(extracted_file_id)
+                    task_id = await self.store_file(extracted_file_id, parser_plugin_id=parser_plugin_id)
                    stored_file_tasks.append(task_id)

                    self.ap.logger.info(
@@ -189,21 +195,28 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):

        return stored_file_tasks[0] if stored_file_tasks else ''

-    async def retrieve(self, query: str, top_k: int) -> list[rag_context.RetrievalResultEntry]:
-        embedding_model = await self.ap.model_mgr.get_embedding_model_by_uuid(
-            self.knowledge_base_entity.embedding_model_uuid
-        )
-        return await self.retriever.retrieve(self.knowledge_base_entity.uuid, query, embedding_model, top_k)
+    async def retrieve(self, query: str, settings: dict | None = None) -> list[rag_context.RetrievalResultEntry]:
+        # Merge stored retrieval_settings with per-request overrides
+        stored = self.knowledge_base_entity.retrieval_settings or {}
+        merged = {**stored, **(settings or {})}
+        if 'top_k' not in merged:
+            merged['top_k'] = 5  # fallback default
+
+        response = await self._retrieve(query, merged)
+
+        results_data = response.get('results', [])
+        entries = []
+        for r in results_data:
+            if isinstance(r, dict):
+                entries.append(rag_context.RetrievalResultEntry(**r))
+            elif isinstance(r, rag_context.RetrievalResultEntry):
+                entries.append(r)
+        return entries

    async def delete_file(self, file_id: str):
-        # delete vector
-        await self.ap.vector_db_mgr.vector_db.delete_by_file_id(self.knowledge_base_entity.uuid, file_id)
-
-        # delete chunk
-        await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.delete(persistence_rag.Chunk).where(persistence_rag.Chunk.file_id == file_id)
-        )
+        await self._delete_document(file_id)

+        # Also cleanup DB record
        await self.ap.persistence_mgr.execute_async(
            sqlalchemy.delete(persistence_rag.File).where(persistence_rag.File.uuid == file_id)
        )
@@ -216,32 +229,295 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
        """Get the name of the knowledge base"""
        return self.knowledge_base_entity.name

-    def get_type(self) -> str:
-        """Get the type of knowledge base"""
-        return 'internal'
+    def get_knowledge_engine_plugin_id(self) -> str:
+        """Get the Knowledge Engine plugin ID"""
+        return self.knowledge_base_entity.knowledge_engine_plugin_id or ''

    async def dispose(self):
-        await self.ap.vector_db_mgr.vector_db.delete_collection(self.knowledge_base_entity.uuid)
+        """Dispose the knowledge base, notifying the plugin to cleanup."""
+        await self._on_kb_delete()
+
+    # ========== Plugin Communication Methods ==========
+
+    async def _on_kb_create(self) -> None:
+        """Notify plugin about KB creation."""
+        plugin_id = self.knowledge_base_entity.knowledge_engine_plugin_id
+        if not plugin_id:
+            return
+
+        try:
+            config = self.knowledge_base_entity.creation_settings or {}
+            self.ap.logger.info(
+                f'Calling RAG plugin {plugin_id}: on_knowledge_base_create(kb_id={self.knowledge_base_entity.uuid})'
+            )
+            await self.ap.plugin_connector.rag_on_kb_create(plugin_id, self.knowledge_base_entity.uuid, config)
+        except Exception as e:
+            self.ap.logger.error(f'Failed to notify plugin {plugin_id} on KB create: {e}')
+            raise
+
+    async def _on_kb_delete(self) -> None:
+        """Notify plugin about KB deletion."""
+        plugin_id = self.knowledge_base_entity.knowledge_engine_plugin_id
+        if not plugin_id:
+            return
+
+        try:
+            self.ap.logger.info(
+                f'Calling RAG plugin {plugin_id}: on_knowledge_base_delete(kb_id={self.knowledge_base_entity.uuid})'
+            )
+            await self.ap.plugin_connector.rag_on_kb_delete(plugin_id, self.knowledge_base_entity.uuid)
+        except Exception as e:
+            self.ap.logger.error(f'Failed to notify plugin {plugin_id} on KB delete: {e}')
+
+    async def _ingest_document(
+        self,
+        file_metadata: dict[str, Any],
+        storage_path: str,
+        parsed_content: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
+        """Call plugin to ingest document."""
+        kb = self.knowledge_base_entity
+        plugin_id = kb.knowledge_engine_plugin_id
+        if not plugin_id:
+            self.ap.logger.error(f'No RAG plugin ID configured for KB {kb.uuid}. Ingestion failed.')
+            raise ValueError('RAG Plugin ID required')
+
+        self.ap.logger.info(f'Calling RAG plugin {plugin_id}: ingest(doc={file_metadata.get("filename")})')
+
+        # Inject knowledge_base_id into file metadata as required by SDK schema
+        file_metadata['knowledge_base_id'] = kb.uuid
+
+        context_data = {
+            'file_object': {
+                'metadata': file_metadata,
+                'storage_path': storage_path,
+            },
+            'knowledge_base_id': kb.uuid,
+            'collection_id': kb.collection_id or kb.uuid,
+            'creation_settings': kb.creation_settings or {},
+            'parsed_content': parsed_content,
+        }
+
+        try:
+            result = await self.ap.plugin_connector.call_rag_ingest(plugin_id, context_data)
+            return result
+        except Exception as e:
+            self.ap.logger.error(f'Plugin ingestion failed: {e}')
+            raise
+
+    async def _retrieve(
+        self,
+        query: str,
+        settings: dict[str, Any],
+    ) -> dict[str, Any]:
+        """Call plugin to retrieve documents.
+
+        Raises:
+            ValueError: If no RAG plugin is configured for this KB.
+            Exception: If the plugin retrieval call fails.
+        """
+        kb = self.knowledge_base_entity
+        plugin_id = kb.knowledge_engine_plugin_id
+        if not plugin_id:
+            raise ValueError(f'No RAG plugin ID configured for KB {kb.uuid}. Retrieval failed.')
+
+        # Session context (e.g. session_name) stays in retrieval_settings
+        # for plugins that need it. Do NOT move them into filters, as filters
+        # are passed directly to vector_search by some plugins (e.g. LangRAG)
+        # and would cause empty results when the metadata field doesn't exist.
+        filters = settings.pop('filters', {})
+
+        retrieval_context = {
+            'query': query,
+            'knowledge_base_id': kb.uuid,
+            'collection_id': kb.collection_id or kb.uuid,
+            'retrieval_settings': settings,
+            'creation_settings': kb.creation_settings or {},
+            'filters': filters,
+        }
+
+        result = await self.ap.plugin_connector.call_rag_retrieve(
+            plugin_id,
+            retrieval_context,
+        )
+        return result
+
+    async def _delete_document(self, document_id: str) -> bool:
+        """Call plugin to delete document."""
+        kb = self.knowledge_base_entity
+        plugin_id = kb.knowledge_engine_plugin_id
+        if not plugin_id:
+            return False
+
+        self.ap.logger.info(f'Calling RAG plugin {plugin_id}: delete_document(doc_id={document_id})')
+
+        try:
+            return await self.ap.plugin_connector.call_rag_delete_document(plugin_id, document_id, kb.uuid)
+        except Exception as e:
+            self.ap.logger.error(f'Plugin document deletion failed: {e}')
+            return False


 class RAGManager:
    ap: app.Application

-    knowledge_bases: list[KnowledgeBaseInterface]
+    knowledge_bases: dict[str, KnowledgeBaseInterface]

    def __init__(self, ap: app.Application):
        self.ap = ap
-        self.knowledge_bases = []
+        self.knowledge_bases = {}

    async def initialize(self):
        await self.load_knowledge_bases_from_db()

+    async def get_all_knowledge_base_details(self) -> list[dict]:
+        """Get all knowledge bases with enriched Knowledge Engine details."""
+        # 1. Get raw KBs from DB
+        result = await self.ap.persistence_mgr.execute_async(sqlalchemy.select(persistence_rag.KnowledgeBase))
+        knowledge_bases = result.all()
+
+        # 2. Get all available Knowledge Engines for enrichment
+        engine_map = {}
+        if self.ap.plugin_connector.is_enable_plugin:
+            try:
+                engines = await self.ap.plugin_connector.list_knowledge_engines()
+                engine_map = {e['plugin_id']: e for e in engines}
+            except Exception as e:
+                self.ap.logger.warning(f'Failed to list Knowledge Engines: {e}')
+
+        # 3. Serialize and enrich
+        kb_list = []
+        for kb in knowledge_bases:
+            kb_dict = self.ap.persistence_mgr.serialize_model(persistence_rag.KnowledgeBase, kb)
+            self._enrich_kb_dict(kb_dict, engine_map)
+            kb_list.append(kb_dict)
+
+        return kb_list
+
+    async def get_knowledge_base_details(self, kb_uuid: str) -> dict | None:
+        """Get specific knowledge base with enriched Knowledge Engine details."""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.select(persistence_rag.KnowledgeBase).where(persistence_rag.KnowledgeBase.uuid == kb_uuid)
+        )
+        kb = result.first()
+        if not kb:
+            return None
+
+        kb_dict = self.ap.persistence_mgr.serialize_model(persistence_rag.KnowledgeBase, kb)
+
+        # Fetch engines
+        engine_map = {}
+        if self.ap.plugin_connector.is_enable_plugin:
+            try:
+                engines = await self.ap.plugin_connector.list_knowledge_engines()
+                engine_map = {e['plugin_id']: e for e in engines}
+            except Exception as e:
+                self.ap.logger.warning(f'Failed to list Knowledge Engines: {e}')
+
+        self._enrich_kb_dict(kb_dict, engine_map)
+        return kb_dict
+
+    @staticmethod
+    def _to_i18n_name(name) -> dict:
+        """Ensure name is always an I18nObject-compatible dict.
+
+        If *name* is already a dict (with ``en_US`` / ``zh_Hans`` keys) it is
+        returned as-is.  A plain string is wrapped into an I18nObject so the
+        frontend ``extractI18nObject`` helper never receives an unexpected type.
+        """
+        if isinstance(name, dict):
+            return name
+        return {'en_US': str(name), 'zh_Hans': str(name)}
+
+    def _enrich_kb_dict(self, kb_dict: dict, engine_map: dict) -> None:
+        """Helper to inject engine info into KB dict."""
+        plugin_id = kb_dict.get('knowledge_engine_plugin_id')
+
+        # Default fallback structure — name must be I18nObject for frontend compatibility
+        fallback_name = self._to_i18n_name(plugin_id or 'Internal (Legacy)')
+        fallback_info = {
+            'plugin_id': plugin_id,
+            'name': fallback_name,
+            'capabilities': [],
+        }
+
+        if not plugin_id:
+            kb_dict['knowledge_engine'] = fallback_info
+            return
+
+        engine_info = engine_map.get(plugin_id)
+        if engine_info:
+            kb_dict['knowledge_engine'] = {
+                'plugin_id': plugin_id,
+                'name': self._to_i18n_name(engine_info.get('name', plugin_id)),
+                'capabilities': engine_info.get('capabilities', []),
+            }
+        else:
+            kb_dict['knowledge_engine'] = fallback_info
+
+    async def create_knowledge_base(
+        self,
+        name: str,
+        knowledge_engine_plugin_id: str,
+        creation_settings: dict,
+        retrieval_settings: dict | None = None,
+        description: str = '',
+    ) -> persistence_rag.KnowledgeBase:
+        """Create a new knowledge base using a RAG plugin."""
+        # Validate that the Knowledge Engine plugin exists
+        if self.ap.plugin_connector.is_enable_plugin:
+            try:
+                engines = await self.ap.plugin_connector.list_knowledge_engines()
+                engine_ids = [e.get('plugin_id') for e in engines]
+                if knowledge_engine_plugin_id not in engine_ids:
+                    raise ValueError(f'Knowledge Engine plugin {knowledge_engine_plugin_id} not found')
+            except ValueError:
+                raise
+            except Exception as e:
+                self.ap.logger.warning(f'Failed to validate Knowledge Engine plugin existence: {e}')
+
+        kb_uuid = str(uuid.uuid4())
+        # Use UUID as collection ID by default for isolation
+        collection_id = kb_uuid
+
+        kb_data = {
+            'uuid': kb_uuid,
+            'name': name,
+            'description': description,
+            'knowledge_engine_plugin_id': knowledge_engine_plugin_id,
+            'collection_id': collection_id,
+            'creation_settings': creation_settings,
+            'retrieval_settings': retrieval_settings or {},
+        }
+
+        # Create Entity
+        kb = persistence_rag.KnowledgeBase(**kb_data)
+
+        # Persist
+        await self.ap.persistence_mgr.execute_async(sqlalchemy.insert(persistence_rag.KnowledgeBase).values(kb_data))
+
+        # Load into Runtime
+        runtime_kb = await self.load_knowledge_base(kb)
+
+        # Notify Plugin — rollback DB record and runtime entry on failure
+        try:
+            await runtime_kb._on_kb_create()
+        except Exception:
+            self.knowledge_bases.pop(kb_uuid, None)
+            await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.delete(persistence_rag.KnowledgeBase).where(persistence_rag.KnowledgeBase.uuid == kb_uuid)
+            )
+            raise
+
+        self.ap.logger.info(f'Created new Knowledge Base {name} ({kb_uuid}) using plugin {knowledge_engine_plugin_id}')
+        return kb
+
    async def load_knowledge_bases_from_db(self):
        self.ap.logger.info('Loading knowledge bases from db...')

-        self.knowledge_bases = []
+        self.knowledge_bases = {}

-        # Load internal knowledge bases
+        # Load knowledge bases
        result = await self.ap.persistence_mgr.execute_async(sqlalchemy.select(persistence_rag.KnowledgeBase))
        knowledge_bases = result.all()

@@ -253,86 +529,37 @@ class RAGManager:
                    f'Error loading knowledge base {knowledge_base.uuid}: {e}\n{traceback.format_exc()}'
                )

-        # Load external knowledge bases
-        external_result = await self.ap.persistence_mgr.execute_async(
-            sqlalchemy.select(persistence_rag.ExternalKnowledgeBase)
-        )
-        external_kbs = external_result.all()
-
-        for external_kb in external_kbs:
-            try:
-                # Don't trigger sync during batch loading - will sync once after LangBot connects to runtime
-                await self.load_external_knowledge_base(external_kb, trigger_sync=False)
-            except Exception as e:
-                self.ap.logger.error(
-                    f'Error loading external knowledge base {external_kb.uuid}: {e}\n{traceback.format_exc()}'
-                )
-
    async def load_knowledge_base(
        self,
        knowledge_base_entity: persistence_rag.KnowledgeBase | sqlalchemy.Row | dict,
    ) -> RuntimeKnowledgeBase:
        if isinstance(knowledge_base_entity, sqlalchemy.Row):
+            # Safe access to _mapping for SQLAlchemy 1.4+
            knowledge_base_entity = persistence_rag.KnowledgeBase(**knowledge_base_entity._mapping)
        elif isinstance(knowledge_base_entity, dict):
-            knowledge_base_entity = persistence_rag.KnowledgeBase(**knowledge_base_entity)
+            # Filter out non-database fields (like knowledge_engine which is computed)
+            filtered_dict = {
+                k: v for k, v in knowledge_base_entity.items() if k in persistence_rag.KnowledgeBase.ALL_DB_FIELDS
+            }
+            knowledge_base_entity = persistence_rag.KnowledgeBase(**filtered_dict)

        runtime_knowledge_base = RuntimeKnowledgeBase(ap=self.ap, knowledge_base_entity=knowledge_base_entity)

        await runtime_knowledge_base.initialize()

-        self.knowledge_bases.append(runtime_knowledge_base)
+        self.knowledge_bases[runtime_knowledge_base.get_uuid()] = runtime_knowledge_base

        return runtime_knowledge_base

-    async def load_external_knowledge_base(
-        self,
-        external_kb_entity: persistence_rag.ExternalKnowledgeBase | sqlalchemy.Row | dict,
-        trigger_sync: bool = True,
-    ) -> ExternalKnowledgeBase:
-        """Load external knowledge base into runtime
-
-        Args:
-            external_kb_entity: External KB entity to load
-            trigger_sync: Whether to trigger sync after loading (default True for manual creation, False for batch loading)
-        """
-        if isinstance(external_kb_entity, sqlalchemy.Row):
-            external_kb_entity = persistence_rag.ExternalKnowledgeBase(**external_kb_entity._mapping)
-        elif isinstance(external_kb_entity, dict):
-            external_kb_entity = persistence_rag.ExternalKnowledgeBase(**external_kb_entity)
-
-        external_kb = ExternalKnowledgeBase(ap=self.ap, external_kb_entity=external_kb_entity)
-
-        await external_kb.initialize()
-
-        self.knowledge_bases.append(external_kb)
-
-        # Trigger sync to create the instance immediately (for manual creation)
-        # Skip sync during batch loading from DB to avoid multiple sync calls
-        if trigger_sync:
-            try:
-                await self.ap.plugin_connector.sync_polymorphic_component_instances()
-                self.ap.logger.info(f'Triggered sync after loading external KB {external_kb_entity.uuid}')
-            except Exception as e:
-                self.ap.logger.error(f'Failed to sync after loading external KB: {e}')
-
-        return external_kb
-
    async def get_knowledge_base_by_uuid(self, kb_uuid: str) -> KnowledgeBaseInterface | None:
-        for kb in self.knowledge_bases:
-            if kb.get_uuid() == kb_uuid:
-                return kb
-        return None
+        return self.knowledge_bases.get(kb_uuid)

    async def remove_knowledge_base_from_runtime(self, kb_uuid: str):
-        for kb in self.knowledge_bases:
-            if kb.get_uuid() == kb_uuid:
-                self.knowledge_bases.remove(kb)
-                return
+        self.knowledge_bases.pop(kb_uuid, None)

    async def delete_knowledge_base(self, kb_uuid: str):
-        for kb in self.knowledge_bases:
-            if kb.get_uuid() == kb_uuid:
-                await kb.dispose()
-                self.knowledge_bases.remove(kb)
-                return
+        kb = self.knowledge_bases.pop(kb_uuid, None)
+        if kb is not None:
+            await kb.dispose()
+        else:
+            self.ap.logger.warning(f'Knowledge base {kb_uuid} not found in runtime, skipping plugin notification')
--- a/src/langbot/pkg/rag/knowledge/services/init.py
+++ b/src/langbot/pkg/rag/knowledge/services/init.py
--- a/src/langbot/pkg/rag/knowledge/services/base_service.py
+++ b/src/langbot/pkg/rag/knowledge/services/base_service.py
@@ -1,15 +0,0 @@
-# 封装异步操作
-import asyncio
-
-
-class BaseService:
-    def __init__(self):
-        pass
-
-    async def _run_sync(self, func, *args, **kwargs):
-        """
-        在单独的线程中运行同步函数。
-        如果第一个参数是 session，则在 to_thread 中获取新的 session。
-        """
-
-        return await asyncio.to_thread(func, *args, **kwargs)
--- a/src/langbot/pkg/rag/knowledge/services/chunker.py
+++ b/src/langbot/pkg/rag/knowledge/services/chunker.py
@@ -1,49 +0,0 @@
-from __future__ import annotations
-
-import json
-from typing import List
-from langbot.pkg.rag.knowledge.services import base_service
-from langbot.pkg.core import app
-from langchain_text_splitters import RecursiveCharacterTextSplitter
-
-
-class Chunker(base_service.BaseService):
-    """
-    A class for splitting long texts into smaller, overlapping chunks.
-    """
-
-    def __init__(self, ap: app.Application, chunk_size: int = 500, chunk_overlap: int = 50):
-        self.ap = ap
-        self.chunk_size = chunk_size
-        self.chunk_overlap = chunk_overlap
-        if self.chunk_overlap >= self.chunk_size:
-            self.ap.logger.warning(
-                'Chunk overlap is greater than or equal to chunk size. This may lead to empty or malformed chunks.'
-            )
-
-    def _split_text_sync(self, text: str) -> List[str]:
-        """
-        Synchronously splits a long text into chunks with specified overlap.
-        This is a CPU-bound operation, intended to be run in a separate thread.
-        """
-        if not text:
-            return []
-
-        text_splitter = RecursiveCharacterTextSplitter(
-            chunk_size=self.chunk_size,
-            chunk_overlap=self.chunk_overlap,
-            length_function=len,
-            is_separator_regex=False,
-        )
-        return text_splitter.split_text(text)
-
-    async def chunk(self, text: str) -> List[str]:
-        """
-        Asynchronously chunks a given text into smaller pieces.
-        """
-        self.ap.logger.info(f'Chunking text (length: {len(text)})...')
-        # Run the synchronous splitting logic in a separate thread
-        chunks = await self._run_sync(self._split_text_sync, text)
-        self.ap.logger.info(f'Text chunked into {len(chunks)} pieces.')
-        self.ap.logger.debug(f'Chunks: {json.dumps(chunks, indent=4, ensure_ascii=False)}')
-        return chunks
--- a/src/langbot/pkg/rag/knowledge/services/embedder.py
+++ b/src/langbot/pkg/rag/knowledge/services/embedder.py
@@ -1,55 +0,0 @@
-from __future__ import annotations
-import uuid
-from typing import List
-from langbot.pkg.rag.knowledge.services.base_service import BaseService
-from langbot.pkg.entity.persistence import rag as persistence_rag
-from langbot.pkg.core import app
-from langbot.pkg.provider.modelmgr.requester import RuntimeEmbeddingModel
-import sqlalchemy
-
-
-class Embedder(BaseService):
-    def __init__(self, ap: app.Application) -> None:
-        super().__init__()
-        self.ap = ap
-
-    async def embed_and_store(
-        self, kb_id: str, file_id: str, chunks: List[str], embedding_model: RuntimeEmbeddingModel
-    ) -> list[persistence_rag.Chunk]:
-        # save chunk to db
-        chunk_entities: list[persistence_rag.Chunk] = []
-        chunk_ids: list[str] = []
-
-        for chunk_text in chunks:
-            chunk_uuid = str(uuid.uuid4())
-            chunk_ids.append(chunk_uuid)
-            chunk_entity = persistence_rag.Chunk(uuid=chunk_uuid, file_id=file_id, text=chunk_text)
-            chunk_entities.append(chunk_entity)
-
-        chunk_dicts = [
-            self.ap.persistence_mgr.serialize_model(persistence_rag.Chunk, chunk) for chunk in chunk_entities
-        ]
-
-        await self.ap.persistence_mgr.execute_async(sqlalchemy.insert(persistence_rag.Chunk).values(chunk_dicts))
-
-        # get embeddings (batch size limit: 64 for OpenAI)
-        MAX_BATCH_SIZE = 64
-        embeddings_list: list[list[float]] = []
-
-        for i in range(0, len(chunks), MAX_BATCH_SIZE):
-            batch = chunks[i : i + MAX_BATCH_SIZE]
-            batch_embeddings = await embedding_model.provider.invoke_embedding(
-                model=embedding_model,
-                input_text=batch,
-                extra_args={},  # TODO: add extra args
-                knowledge_base_id=kb_id,
-                call_type='embedding',
-            )
-            embeddings_list.extend(batch_embeddings)
-
-        # save embeddings to vdb
-        await self.ap.vector_db_mgr.vector_db.add_embeddings(kb_id, chunk_ids, embeddings_list, chunk_dicts)
-
-        self.ap.logger.info(f'Successfully saved {len(chunk_entities)} embeddings to Knowledge Base.')
-
-        return chunk_entities
--- a/src/langbot/pkg/rag/knowledge/services/parser.py
+++ b/src/langbot/pkg/rag/knowledge/services/parser.py
@@ -1,291 +0,0 @@
-from __future__ import annotations
-
-import PyPDF2
-import io
-from docx import Document
-import chardet
-from typing import Union, Callable, Any
-import markdown
-from bs4 import BeautifulSoup
-import re
-import asyncio  # Import asyncio for async operations
-from langbot.pkg.core import app
-
-
-class FileParser:
-    """
-    A robust file parser class to extract text content from various document formats.
-    It supports TXT, PDF, DOCX, XLSX, CSV, Markdown, HTML, and EPUB files.
-    All core file reading operations are designed to be run synchronously in a thread pool
-    to avoid blocking the asyncio event loop.
-    """
-
-    def __init__(self, ap: app.Application):
-        self.ap = ap
-
-    async def _run_sync(self, sync_func: Callable, *args: Any, **kwargs: Any) -> Any:
-        """
-        Runs a synchronous function in a separate thread to prevent blocking the event loop.
-        This is a general utility method for wrapping blocking I/O operations.
-        """
-        try:
-            return await asyncio.to_thread(sync_func, *args, **kwargs)
-        except Exception as e:
-            self.ap.logger.error(f'Error running synchronous function {sync_func.__name__}: {e}')
-            raise
-
-    async def parse(self, file_name: str, extension: str) -> Union[str, None]:
-        """
-        Parses the file based on its extension and returns the extracted text content.
-        This is the main asynchronous entry point for parsing.
-
-        Args:
-            file_name (str): The name of the file to be parsed, get from ap.storage_mgr
-
-        Returns:
-            Union[str, None]: The extracted text content as a single string, or None if parsing fails.
-        """
-
-        file_extension = extension.lower()
-        parser_method = getattr(self, f'_parse_{file_extension}', None)
-
-        if parser_method is None:
-            self.ap.logger.error(f'Unsupported file format: {file_extension} for file {file_name}')
-            return None
-
-        try:
-            # Pass file_path to the specific parser methods
-            return await parser_method(file_name)
-        except Exception as e:
-            self.ap.logger.error(f'Failed to parse {file_extension} file {file_name}: {e}')
-            return None
-
-    # --- Helper for reading files with encoding detection ---
-    async def _read_file_content(self, file_name: str) -> Union[str, bytes]:
-        """
-        Reads a file with automatic encoding detection, ensuring the synchronous
-        file read operation runs in a separate thread.
-        """
-
-        # def _read_sync():
-        #     with open(file_path, 'rb') as file:
-        #         raw_data = file.read()
-        #         detected = chardet.detect(raw_data)
-        #         encoding = detected['encoding'] or 'utf-8'
-
-        #     if mode == 'r':
-        #         return raw_data.decode(encoding, errors='ignore')
-        #     return raw_data  # For binary mode
-
-        # return await self._run_sync(_read_sync)
-        file_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-        detected = chardet.detect(file_bytes)
-        encoding = detected['encoding'] or 'utf-8'
-
-        return file_bytes.decode(encoding, errors='ignore')
-
-    # --- Specific Parser Methods ---
-
-    async def _parse_txt(self, file_name: str) -> str:
-        """Parses a TXT file and returns its content."""
-        self.ap.logger.info(f'Parsing TXT file: {file_name}')
-        return await self._read_file_content(file_name)
-
-    async def _parse_pdf(self, file_name: str) -> str:
-        """Parses a PDF file and returns its text content."""
-        self.ap.logger.info(f'Parsing PDF file: {file_name}')
-
-        # def _parse_pdf_sync():
-        #     text_content = []
-        #     with open(file_name, 'rb') as file:
-        #         pdf_reader = PyPDF2.PdfReader(file)
-        #         for page in pdf_reader.pages:
-        #             text = page.extract_text()
-        #             if text:
-        #                 text_content.append(text)
-        #     return '\n'.join(text_content)
-
-        # return await self._run_sync(_parse_pdf_sync)
-
-        pdf_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-        def _parse_pdf_sync():
-            pdf_reader = PyPDF2.PdfReader(io.BytesIO(pdf_bytes))
-            text_content = []
-            for page in pdf_reader.pages:
-                text = page.extract_text()
-                if text:
-                    text_content.append(text)
-            return '\n'.join(text_content)
-
-        return await self._run_sync(_parse_pdf_sync)
-
-    async def _parse_docx(self, file_name: str) -> str:
-        """Parses a DOCX file and returns its text content."""
-        self.ap.logger.info(f'Parsing DOCX file: {file_name}')
-
-        docx_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-        def _parse_docx_sync():
-            doc = Document(io.BytesIO(docx_bytes))
-            text_content = [paragraph.text for paragraph in doc.paragraphs if paragraph.text.strip()]
-            return '\n'.join(text_content)
-
-        return await self._run_sync(_parse_docx_sync)
-
-    async def _parse_doc(self, file_name: str) -> str:
-        """Handles .doc files, explicitly stating lack of direct support."""
-        self.ap.logger.warning(f'Direct .doc parsing is not supported for {file_name}. Please convert to .docx first.')
-        raise NotImplementedError('Direct .doc parsing not supported. Please convert to .docx first.')
-
-    # async def _parse_xlsx(self, file_name: str) -> str:
-    #     """Parses an XLSX file, returning text from all sheets."""
-    #     self.ap.logger.info(f'Parsing XLSX file: {file_name}')
-
-    #     xlsx_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-    #     def _parse_xlsx_sync():
-    #         excel_file = pd.ExcelFile(io.BytesIO(xlsx_bytes))
-    #         all_sheet_content = []
-    #         for sheet_name in excel_file.sheet_names:
-    #             df = pd.read_excel(io.BytesIO(xlsx_bytes), sheet_name=sheet_name)
-    #             sheet_text = f'--- Sheet: {sheet_name} ---\n{df.to_string(index=False)}\n'
-    #             all_sheet_content.append(sheet_text)
-    #         return '\n'.join(all_sheet_content)
-
-    #     return await self._run_sync(_parse_xlsx_sync)
-
-    # async def _parse_csv(self, file_name: str) -> str:
-    #     """Parses a CSV file and returns its content as a string."""
-    #     self.ap.logger.info(f'Parsing CSV file: {file_name}')
-
-    #     csv_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-    #     def _parse_csv_sync():
-    #         # pd.read_csv can often detect encoding, but explicit detection is safer
-    #         # raw_data = self._read_file_content(
-    #         #     file_name, mode='rb'
-    #         # )  # Note: this will need to be await outside this sync function
-    #         # _ = raw_data
-    #         # For simplicity, we'll let pandas handle encoding internally after a raw read.
-    #         # A more robust solution might pass encoding directly to pd.read_csv after detection.
-    #         detected = chardet.detect(io.BytesIO(csv_bytes))
-    #         encoding = detected['encoding'] or 'utf-8'
-    #         df = pd.read_csv(io.BytesIO(csv_bytes), encoding=encoding)
-    #         return df.to_string(index=False)
-
-    #     return await self._run_sync(_parse_csv_sync)
-
-    async def _parse_md(self, file_name: str) -> str:
-        """Parses a Markdown file, converting it to structured plain text."""
-        self.ap.logger.info(f'Parsing Markdown file: {file_name}')
-
-        md_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-        def _parse_markdown_sync():
-            md_content = io.BytesIO(md_bytes).read().decode('utf-8', errors='ignore')
-            html_content = markdown.markdown(
-                md_content, extensions=['extra', 'codehilite', 'tables', 'toc', 'fenced_code']
-            )
-            soup = BeautifulSoup(html_content, 'html.parser')
-            text_parts = []
-            for element in soup.children:
-                if element.name in ['h1', 'h2', 'h3', 'h4', 'h5', 'h6']:
-                    level = int(element.name[1])
-                    text_parts.append('#' * level + ' ' + element.get_text().strip())
-                elif element.name == 'p':
-                    text = element.get_text().strip()
-                    if text:
-                        text_parts.append(text)
-                elif element.name in ['ul', 'ol']:
-                    for li in element.find_all('li'):
-                        text_parts.append(f'* {li.get_text().strip()}')
-                elif element.name == 'pre':
-                    code_block = element.get_text().strip()
-                    if code_block:
-                        text_parts.append(f'```\n{code_block}\n```')
-                elif element.name == 'table':
-                    table_str = self._extract_table_to_markdown_sync(element)  # Call sync helper
-                    if table_str:
-                        text_parts.append(table_str)
-                elif element.name:
-                    text = element.get_text(separator=' ', strip=True)
-                    if text:
-                        text_parts.append(text)
-            cleaned_text = re.sub(r'\n\s*\n', '\n\n', '\n'.join(text_parts))
-            return cleaned_text.strip()
-
-        return await self._run_sync(_parse_markdown_sync)
-
-    async def _parse_html(self, file_name: str) -> str:
-        """Parses an HTML file, extracting structured plain text."""
-        self.ap.logger.info(f'Parsing HTML file: {file_name}')
-
-        html_bytes = await self.ap.storage_mgr.storage_provider.load(file_name)
-
-        def _parse_html_sync():
-            html_content = io.BytesIO(html_bytes).read().decode('utf-8', errors='ignore')
-            soup = BeautifulSoup(html_content, 'html.parser')
-            for script_or_style in soup(['script', 'style']):
-                script_or_style.decompose()
-            text_parts = []
-            for element in soup.body.children if soup.body else soup.children:
-                if element.name in ['h1', 'h2', 'h3', 'h4', 'h5', 'h6']:
-                    level = int(element.name[1])
-                    text_parts.append('#' * level + ' ' + element.get_text().strip())
-                elif element.name == 'p':
-                    text = element.get_text().strip()
-                    if text:
-                        text_parts.append(text)
-                elif element.name in ['ul', 'ol']:
-                    for li in element.find_all('li'):
-                        text = li.get_text().strip()
-                        if text:
-                            text_parts.append(f'* {text}')
-                elif element.name == 'table':
-                    table_str = self._extract_table_to_markdown_sync(element)  # Call sync helper
-                    if table_str:
-                        text_parts.append(table_str)
-                elif element.name:
-                    text = element.get_text(separator=' ', strip=True)
-                    if text:
-                        text_parts.append(text)
-            cleaned_text = re.sub(r'\n\s*\n', '\n\n', '\n'.join(text_parts))
-            return cleaned_text.strip()
-
-        return await self._run_sync(_parse_html_sync)
-
-    def _add_toc_items_sync(self, toc_list: list, text_content: list, level: int):
-        """Recursively adds TOC items to text_content (synchronous helper)."""
-        indent = '  ' * level
-        for item in toc_list:
-            if isinstance(item, tuple):
-                chapter, subchapters = item
-                text_content.append(f'{indent}- {chapter.title}')
-                self._add_toc_items_sync(subchapters, text_content, level + 1)
-            else:
-                text_content.append(f'{indent}- {item.title}')
-
-    def _extract_table_to_markdown_sync(self, table_element: BeautifulSoup) -> str:
-        """Helper to convert a BeautifulSoup table element into a Markdown table string (synchronous)."""
-        headers = [th.get_text().strip() for th in table_element.find_all('th')]
-        rows = []
-        for tr in table_element.find_all('tr'):
-            cells = [td.get_text().strip() for td in tr.find_all('td')]
-            if cells:
-                rows.append(cells)
-
-        if not headers and not rows:
-            return ''
-
-        table_lines = []
-        if headers:
-            table_lines.append(' | '.join(headers))
-            table_lines.append(' | '.join(['---'] * len(headers)))
-
-        for row_cells in rows:
-            padded_cells = row_cells + [''] * (len(headers) - len(row_cells)) if headers else row_cells
-            table_lines.append(' | '.join(padded_cells))
-
-        return '\n'.join(table_lines)
--- a/src/langbot/pkg/rag/knowledge/services/retriever.py
+++ b/src/langbot/pkg/rag/knowledge/services/retriever.py
@@ -1,53 +0,0 @@
-from __future__ import annotations
-
-from . import base_service
-from ....core import app
-from ....provider.modelmgr.requester import RuntimeEmbeddingModel
-from langbot_plugin.api.entities.builtin.rag import context as rag_context
-from langbot_plugin.api.entities.builtin.provider.message import ContentElement
-
-
-class Retriever(base_service.BaseService):
-    def __init__(self, ap: app.Application):
-        super().__init__()
-        self.ap = ap
-
-    async def retrieve(
-        self, kb_id: str, query: str, embedding_model: RuntimeEmbeddingModel, k: int = 5
-    ) -> list[rag_context.RetrievalResultEntry]:
-        self.ap.logger.info(
-            f"Retrieving for query: '{query[:10]}' with k={k} using {embedding_model.model_entity.uuid}"
-        )
-
-        query_embedding: list[float] = await embedding_model.provider.invoke_embedding(
-            model=embedding_model,
-            input_text=[query],
-            extra_args={},  # TODO: add extra args
-            knowledge_base_id=kb_id,
-            query_text=query,
-            call_type='retrieve',
-        )
-
-        vector_results = await self.ap.vector_db_mgr.vector_db.search(kb_id, query_embedding[0], k)
-
-        # 'ids' shape mirrors the Chroma-style response contract for compatibility
-        matched_vector_ids = vector_results.get('ids', [[]])[0]
-        distances = vector_results.get('distances', [[]])[0]
-        vector_metadatas = vector_results.get('metadatas', [[]])[0]
-
-        if not matched_vector_ids:
-            self.ap.logger.info('No relevant chunks found in vector database.')
-            return []
-
-        result: list[rag_context.RetrievalResultEntry] = []
-
-        for i, id in enumerate(matched_vector_ids):
-            entry = rag_context.RetrievalResultEntry(
-                id=id,
-                content=[ContentElement.from_text(vector_metadatas[i].get('text', ''))],
-                metadata=vector_metadatas[i],
-                distance=distances[i],
-            )
-            result.append(entry)
-
-        return result
--- a/src/langbot/pkg/rag/service/init.py
+++ b/src/langbot/pkg/rag/service/init.py
@@ -0,0 +1 @@
+from .runtime import RAGRuntimeService as RAGRuntimeService
--- a/src/langbot/pkg/rag/service/runtime.py
+++ b/src/langbot/pkg/rag/service/runtime.py
@@ -0,0 +1,116 @@
+from __future__ import annotations
+
+import posixpath
+from typing import Any
+from langbot.pkg.core import app
+
+
+class RAGRuntimeService:
+    """Service to handle RAG-related requests from plugins (Runtime).
+
+    This service acts as the bridge between plugin RPC requests and
+    LangBot's infrastructure (embedding models, vector databases, file storage).
+    """
+
+    def __init__(self, ap: app.Application):
+        self.ap = ap
+
+    async def vector_upsert(
+        self,
+        collection_id: str,
+        vectors: list[list[float]],
+        ids: list[str],
+        metadata: list[dict[str, Any]] | None = None,
+        documents: list[str] | None = None,
+    ) -> None:
+        """Handle VECTOR_UPSERT action."""
+        metadatas = metadata if metadata else [{} for _ in vectors]
+        await self.ap.vector_db_mgr.upsert(
+            collection_name=collection_id,
+            vectors=vectors,
+            ids=ids,
+            metadata=metadatas,
+            documents=documents,
+        )
+
+    async def vector_search(
+        self,
+        collection_id: str,
+        query_vector: list[float],
+        top_k: int,
+        filters: dict[str, Any] | None = None,
+        search_type: str = 'vector',
+        query_text: str = '',
+        vector_weight: float | None = None,
+    ) -> list[dict[str, Any]]:
+        """Handle VECTOR_SEARCH action."""
+        return await self.ap.vector_db_mgr.search(
+            collection_name=collection_id,
+            query_vector=query_vector,
+            limit=top_k,
+            filter=filters,
+            search_type=search_type,
+            query_text=query_text,
+            vector_weight=vector_weight,
+        )
+
+    async def vector_delete(
+        self, collection_id: str, file_ids: list[str] | None = None, filters: dict[str, Any] | None = None
+    ) -> int:
+        """Handle VECTOR_DELETE action.
+
+        Deletes vectors associated with the given file IDs from the collection.
+        Each file_id corresponds to a document whose vectors will be removed.
+
+        Args:
+            collection_id: The collection to delete from.
+            file_ids: File IDs whose associated vectors should be deleted.
+                Each file_id maps to a set of vectors stored with that file_id
+                in their metadata.
+            filters: Filter-based deletion (not yet supported, will raise).
+        """
+        count = 0
+        if file_ids:
+            await self.ap.vector_db_mgr.delete_by_file_id(collection_name=collection_id, file_ids=file_ids)
+            count = len(file_ids)
+        elif filters:
+            count = await self.ap.vector_db_mgr.delete_by_filter(collection_name=collection_id, filter=filters)
+        return count
+
+    async def vector_list(
+        self,
+        collection_id: str,
+        filters: dict[str, Any] | None = None,
+        limit: int = 20,
+        offset: int = 0,
+    ) -> tuple[list[dict[str, Any]], int]:
+        """Handle VECTOR_LIST action.
+
+        Args:
+            collection_id: The collection to list from.
+            filters: Optional metadata filters.
+            limit: Maximum number of items to return.
+            offset: Number of items to skip.
+
+        Returns:
+            Tuple of (items, total).
+        """
+        return await self.ap.vector_db_mgr.list_by_filter(
+            collection_name=collection_id,
+            filter=filters,
+            limit=limit,
+            offset=offset,
+        )
+
+    async def get_file_stream(self, storage_path: str) -> bytes:
+        """Handle GET_KNOWLEDEGE_FILE_STREAM action.
+
+        Uses the storage manager abstraction to load file content,
+        regardless of the underlying storage provider.
+        """
+        # Validate storage_path to prevent path traversal
+        normalized = posixpath.normpath(storage_path)
+        if normalized.startswith('/') or '..' in normalized.split('/'):
+            raise ValueError('Invalid storage path')
+        content_bytes = await self.ap.storage_mgr.storage_provider.load(normalized)
+        return content_bytes if content_bytes else b''
--- a/src/langbot/pkg/storage/mgr.py
+++ b/src/langbot/pkg/storage/mgr.py
@@ -3,7 +3,7 @@ from __future__ import annotations

 from ..core import app
 from . import provider
-from .providers import localstorage, s3storage
+from .providers import localstorage


 class StorageMgr:
@@ -21,6 +21,8 @@ class StorageMgr:
        storage_type = storage_config.get('use', 'local')

        if storage_type == 's3':
+            from .providers import s3storage
+
            self.storage_provider = s3storage.S3StorageProvider(self.ap)
            self.ap.logger.info('Initialized S3 storage backend.')
        else:
--- a/Show More
+++ b/Show More
				`@@ -0,0 +1 @@`
				`from .runtime import RAGRuntimeService as RAGRuntimeService`