🧠 MeetingMind

MeetingMind is an open-source, fully-automated AI Meeting Assistant. It acts as your personal invisible notetaker: automatically joining your Google Meets using a headless bot, recording the session, generating word-level transcripts with speaker diarization, and giving you an interactive "Spotify-like" dashboard to playback and query your meetings using AI.

Think of it as an open-source alternative to Fathom, Fireflies.ai, or Otter.ai.

---

🚀 Core Features

• 🤖 Automated Bot Worker: A headless Puppeteer-based worker that silently joins your Google Meets when invited via the Chrome Extension.
• 🎙️ High-Fidelity Recording: Direct audio/video stream capturing uploaded securely and reliably to AWS S3.
• 📝 Blazing Fast Transcripts: Powered by Deepgram AI, delivering highly accurate speaker-diarized text with exact word-level timestamping.
• 🎵 Synced Video Playback UI: A premium dashboard featuring a custom video player. Click any word in the transcript to jump the video to that exact moment.
• 💬 Context-Aware AI Chat: Ask questions, generate summaries, or pull action items from your meeting transcript directly inside the dashboard using Gemini/OpenAI.
• 🔗 1-Click Chrome Extension: A custom extension injected into your Google Meet UI allows you to invite the bot with a single button press.

---

🏗️ Architecture

MeetingMind is composed of four distinct microservices, communicating via PostgreSQL and Redis for robust distributed queuing.

1. /frontend (Next.js 14, React, Tailwind CSS): The beautiful, premium user dashboard. Features the complex synced video player, shareable links, and AI Chat UI.
2. /backend (Node.js, Express, Drizzle ORM): The central nervous system. Manages authentication, database operations, securely signs URLs for AWS S3 fetching, and hosts the REST API.
3. /bot-worker (Node.js, Puppeteer): The heavy lifter. A scalable consumer that reads from a Redis queue. It opens Google Chrome headlessly, joins the Meet, records the DOM/Canvas, processes the .webm file, and talks to Deepgram for transcripts.
4. /meetingmind-extension (Manifest v3, Vanilla JS): A lightweight browser extension that reads your active Google Meet URL and sends a payload to the backend to add a job to the Redis queue.

---

🛠️ Quick Start & Local Development

Prerequisites

• Node.js: v18 or higher
• PostgreSQL: A running instance (local or cloud like Supabase/Neon).
• Redis: A running instance for BullMQ queues (local or cloud like Upstash).
• AWS S3: An S3 bucket with proper IAM write credentials configured.
• Deepgram API Key: For transcription.
• Google OAuth Credentials: For the user Sign-In system.

1. Database & Backend Setup

Navigate to the /backend folder.

cd backend
npm install

Create a .env file in the /backend directory:

PORT=5000
DATABASE_URL="postgresql://user:pass@localhost:5432/meetingmind"
FRONTEND_URL="http://localhost:3000"
JWT_SECRET="your_secret_string"

AWS_ACCESS_KEY_ID="your_aws_key"
AWS_SECRET_ACCESS_KEY="your_aws_secret"
AWS_REGION="us-east-1"
AWS_BUCKET_NAME="your-bucket-name"

REDIS_HOST="127.0.0.1"
REDIS_PORT=6379

GOOGLE_CLIENT_ID="your_google_id.apps.googleusercontent.com"

Push the database schema using Drizzle and start the server:

npx drizzle-kit push --config=drizzle.config.js
npm run dev

2. Frontend Setup

Navigate to the /frontend folder.

cd frontend
npm install

Create a .env.local file in the /frontend directory:

NEXT_PUBLIC_API_URL="http://localhost:5000"
NEXT_PUBLIC_GOOGLE_CLIENT_ID="your_google_id.apps.googleusercontent.com"

Start the Next.js application:

npm run dev

The dashboard is now running on http://localhost:3000.

3. Bot Worker Setup

Navigate to the /bot-worker folder.

cd bot-worker
npm install

Create a .env file in the /bot-worker directory:

DATABASE_URL="postgresql://user:pass@localhost:5432/meetingmind"

AWS_ACCESS_KEY_ID="your_aws_key"
AWS_SECRET_ACCESS_KEY="your_aws_secret"
AWS_REGION="us-east-1"
AWS_BUCKET_NAME="your-bucket-name"

REDIS_HOST="127.0.0.1"
REDIS_PORT=6379

DEEPGRAM_API_KEY="your_deepgram_key"
GEMINI_API_KEY="your_gemini_key"

Start the headless bot worker:

npm run dev

4. Chrome Extension Installation

To easily invite the bot to your meetings:

1. Open Google Chrome and navigate to chrome://extensions/.
2. Turn on Developer mode in the top right corner.
3. Click Load unpacked in the top left.
4. Select the /meetingmind-extension folder.
5. Make sure the config.js file inside the extension points to your local backend (http://localhost:5000).

---

🚢 Deploying the Bot Worker

Deploying the frontend (Vercel) and backend (Render/Heroku) is standard. However, deploying the Bot Worker requires specific dependencies because it runs a full headless Chrome browser instance via Puppeteer.

If you are deploying the bot worker to a Linux VPS (like DigitalOcean, AWS EC2, or Hetzner), you must install the necessary OS-level dependencies for Chrome.

Ubuntu/Debian Deployment Instructions for Bot:

1. SSH into your server.
2. Install Node.js, Redis, and clone the repository.
3. Install the required Chrome dependencies:

sudo apt-get update
sudo apt-get install -yq \
    gconf-service libasound2 libatk1.0-0 libc6 libcairo2 libcups2 \
    libdbus-1-3 libexpat1 libfontconfig1 libgcc1 libgconf-2-4 \
    libgdk-pixbuf2.0-0 libglib2.0-0 libgtk-3-0 libnspr4 \
    libpango-1.0-0 libpangocairo-1.0-0 libstdc++6 libx11-6 \
    libx11-xcb1 libxcb1 libxcomposite1 libxcursor1 libxdamage1 \
    libxext6 libxfixes3 libxi6 libxrandr2 libxrender1 libxss1 \
    libxtst6 ca-certificates fonts-liberation libnss3 lsb-release \
    xdg-utils wget

4. Navigate to /bot-worker, run npm install, populate your .env file, and start the process using a manager like pm2:

npm install -g pm2
pm2 start src/index.js --name "meetingmind-bot"

Alternatively, you can Dockerize the bot worker using a base image that already contains Puppeteer dependencies (e.g., ghcr.io/puppeteer/puppeteer:latest).

---

🤝 Contributing

Contributions are welcome! If you find a bug, want to add styling updates, or want to expand the AI integration (e.g., adding Anthropic support), feel free to open a Pull Request.

📄 License

This project is open-source and available under the MIT License.