中文 | English

一个智能的浏览器端智能体，能够实时捕获、转录并帮助你处理来自浏览器标签页和麦克风的音频。让你在会议或学习过程中不再错过重要的想法！

受 Notion AI Meeting Notes 功能的启发，本项目解决了以下痛点：

1. 🎨 个性化 Prompt - 不同于固定的 AI 方案，你可以根据需求调整提示词
2. 🌐 Web 使用 - 直接在浏览器中使用，无需安装桌面应用
3. 🔒 隐私安全 - 所有音频处理都在本地进行，敏感会议数据不会上传到云端

在繁忙的会议或 YouTube 学习过程中，灵光一现的想法转瞬即逝。记笔记会打断你的专注，而等到会议结束，那些洞察往往已经遗忘。

环境智能体是你的 Always-On AI 伴侣，它能够：

• 持续监听浏览器音频和麦克风输入
• 以最小的干扰帮助捕捉稍纵即逝的想法
• 基于上下文生成智能摘要
• 回答你在会议中没来得及问的问题

• 抓拍笔记 - 一键记录重要观点
• 抓拍待办 - 即时创建行动事项
• 抓拍问题 - 记录问题以便后续探索

AI 生成的会议纪要，理解上下文和你的标注

一个创新功能，解决未提出的问题。AI 会自动：

• 从你的笔记中识别关键问题
• 基于会议上下文提供深思熟虑的答案
• 帮助你轻松完成会后反思

• 标签页音频捕获 - 使用 WebRTC getDisplayMedia API 捕获浏览器标签页音频
• 麦克风支持 - 捕获你的声音用于提问和评论
• 实时转录 - 由火山引擎流式 ASR 服务提供支持
• 智能处理 - 豆包大模型用于智能摘要和问答
• 现代化 UI - 简洁、无干扰的界面，带有浮动控件

1. 克隆仓库

   git clone https://github.com/celineycn/ambience-agent.git
   cd ambient-agent

2. 安装依赖

   npm install

3. 配置环境变量 在根目录创建 .env 文件：

   PORT=3000
   VOLCENGINE_APP_ID=your_app_id_here
   VOLCENGINE_ACCESS_KEY=your_access_key_here
   ARK_API_KEY=your_ark_api_key_here

4. 启动开发服务器

   npm run dev

5. 打开浏览器 访问 http://localhost:3000

支持：

• ✅ Chrome/Chromium 内核浏览器（推荐）
• ✅ Edge
• ✅ Brave

所需 API：

• WebRTC 和 getDisplayMedia
• MediaRecorder API
• Web Audio API
• WebSocket API

1. macOS 限制：由于浏览器安全限制，macOS 浏览器无法捕获桌面应用程序（如 Zoom 桌面版）的音频。仅支持浏览器标签页音频和麦克风输入。

2. 需要 HTTPS：出于安全原因，标签页音频捕获需要 HTTPS 或 localhost。

3. 权限要求：应用会请求显示捕获和麦克风访问权限。

我们欢迎各种形式的贡献！无论是：

• 🐛 修复 Bug
• ✨ 新功能
• 📝 文档改进
• 🌍 翻译
• 💡 想法和建议

请随时提出 Issue 或提交 Pull Request。

MIT 许可证 - 可自由用于个人或商业用途。

An intelligent browser-based agent that captures, transcribes, and helps you process audio from browser tabs and microphone in real-time. Never miss important ideas during meetings or learning sessions again!

Inspired by Notion's AI Meeting Notes capability, this project addresses several pain points:

1. 🎨 Customizable Prompts - Unlike fixed AI solutions, you can adjust prompts to fit your needs
2. 🌐 Web-First - Works directly in your browser, no desktop app installation required
3. 🔒 Privacy-First - All audio processing happens locally, no sensitive meeting data uploaded to the cloud

During busy meetings or YouTube learning sessions, brilliant ideas come and go in seconds. Taking notes disrupts your focus, and by the time the meeting ends, those insights are often lost forever.

Ambient Agent is your always-on AI companion that:

• Continuously listens to browser audio and microphone input
• Helps capture fleeting thoughts with minimal interruption
• Generates intelligent summaries based on context
• Answers questions you didn't get to ask during the meeting

• Quick Note - Capture important points with one click
• Quick Todo - Create action items instantly
• Quick Question - Record questions to explore later

AI-generated meeting minutes that understand context and your annotations

An innovative feature that addresses unasked questions. The AI automatically:

• Identifies key questions from your notes
• Provides thoughtful answers based on meeting context
• Helps you complete post-meeting reflection effortlessly

• Tab Audio Capture - WebRTC getDisplayMedia API for browser tab audio
• Microphone Support - Capture your voice for questions and comments
• Real-time Transcription - Powered by Volcengine's streaming ASR service
• Smart Processing - Doubao LLM for intelligent summarization and Q&A
• Modern UI - Clean, distraction-free interface with floating controls

1. Clone the repository

   git clone https://github.com/celineycn/ambience-agent.git
   cd ambient-agent

2. Install dependencies

   npm install

3. Configure environment variables Create a .env file in the root directory:

   PORT=3000
   VOLCENGINE_APP_ID=your_app_id_here
   VOLCENGINE_ACCESS_KEY=your_access_key_here
   ARK_API_KEY=your_ark_api_key_here

4. Start the development server

   npm run dev

5. Open your browser Navigate to http://localhost:3000

Supported:

• ✅ Chrome/Chromium-based browsers (recommended)
• ✅ Edge
• ✅ Brave

Required APIs:

• WebRTC and getDisplayMedia
• MediaRecorder API
• Web Audio API
• WebSocket API

1. macOS Limitation: Due to browser security restrictions, macOS browsers cannot capture audio from desktop applications (e.g., Zoom desktop app). Only browser tab audio and microphone input are supported.

2. HTTPS Required: For security reasons, tab audio capture requires HTTPS or localhost.

3. Permissions: The app will request permissions for display capture and microphone access.

We welcome contributions! Whether it's:

• 🐛 Bug fixes
• ✨ New features
• 📝 Documentation improvements
• 🌍 Translations
• 💡 Ideas and suggestions

Please feel free to open issues or submit pull requests.

MIT License - feel free to use this project for personal or commercial purposes.

Made with ❤️ by vibecoders who forget to ask questions in meetings