MeetingAI — MERN Stack Meeting Analyzer

A full-stack AI Meeting Analyzer built with MongoDB, Express, React, and Node.js. Analyzes meeting transcripts to extract action items, unresolved topics, participation dominance, and keyword topic clusters — with zero ML or speech processing.

---

Project Structure

meeting-analyzer/
├── backend/                  # Express + MongoDB API
│   ├── controllers/
│   │   ├── analyzerEngine.js  # Core NLP rule engine
│   │   ├── authController.js
│   │   └── meetingController.js
│   ├── middleware/
│   │   ├── auth.js            # JWT middleware
│   │   └── errorHandler.js
│   ├── models/
│   │   ├── User.js
│   │   └── Meeting.js
│   ├── routes/
│   │   ├── auth.js
│   │   └── meetings.js
│   ├── .env.example
│   ├── package.json
│   └── server.js
│
└── frontend/                 # React SPA
    ├── public/
    │   └── index.html
    ├── src/
    │   ├── components/
    │   │   ├── Layout.js         # Sidebar + Outlet
    │   │   └── AnalysisResult.js # Shared analysis UI (tabs + chart)
    │   ├── context/
    │   │   └── AuthContext.js    # JWT auth state
    │   ├── pages/
    │   │   ├── Login.js
    │   │   ├── Register.js
    │   │   ├── Dashboard.js      # Stats + meeting history
    │   │   ├── Analyzer.js       # Transcript input + live analysis
    │   │   └── MeetingDetail.js  # Saved meeting + toggle actions
    │   ├── services/
    │   │   └── api.js            # Axios instance + API helpers
    │   ├── App.js
    │   ├── index.js
    │   └── index.css
    └── package.json

---

Features

• JWT authentication — register, login, protected routes
• Transcript analysis — rule-based NLP, no ML required
• Participation dominance — word frequency per speaker with bar chart
• Action item detection — modal obligation signal scanning
• Unresolved topic detection — hedging/blocking language patterns
• Keyword topic clustering — 5 semantic domains
• AI insight generation — plain-English dominance summary
• Saved meetings — full CRUD with pagination
• Toggle action/unresolved — mark items done in detail view
• Dashboard stats — aggregate counts across all meetings

---

Setup

Prerequisites

• Node.js v18+
• MongoDB (local or Atlas)

1. Backend

cd backend
cp .env.example .env
# Edit .env: set MONGO_URI and JWT_SECRET
npm install
npm run dev

Backend runs on http://localhost:5000

2. Frontend

cd frontend
cp .env.example .env
npm install
npm start

Frontend runs on http://localhost:3000

---

API Endpoints

Auth

Method	Endpoint	Description
POST	/api/auth/register	Register new user
POST	/api/auth/login	Login
GET	/api/auth/me	Get current user

Meetings (all protected)

Method	Endpoint	Description
POST	/api/meetings/analyze	Analyze without saving
POST	/api/meetings	Analyze and save
GET	/api/meetings	List user's meetings
GET	/api/meetings/:id	Get single meeting
DELETE	/api/meetings/:id	Delete meeting
PATCH	/api/meetings/:id/action/:itemId	Toggle action done
PATCH	/api/meetings/:id/unresolved/:itemId	Toggle resolved
GET	/api/meetings/stats/summary	Dashboard stats

---

Transcript Format

For best results, use labeled speaker turns:

Alice: I think we should move the launch date to Q3.
Bob: Agreed, but we haven't resolved the pricing model.
Alice: Can you own the pricing research by Friday, Bob?

Plain text (no speaker labels) works too — action/topic detection still runs, but dominance analysis is skipped.

---

Analysis Engine

All analysis is rule-based in backend/controllers/analyzerEngine.js:

• Speaker parsing — regex Name: message pattern
• Word frequency — counts tokens per speaker for dominance %
• Action detection — scans for modal signals (can you, need to, by Friday, I'll…)
• Unresolved detection — hedging phrases (still open, not yet, tbd, table it…)
• Topic clustering — 5 keyword clusters: timeline, budget, technical, marketing, team/process
• Insight synthesis — rule-based paragraph from dominance + silence counts