TranscriptAI

A full-stack web application for automated summarization of meeting transcripts using Google's Gemini AI model. The system provides RESTful APIs for transcript processing and a React-based frontend for user interaction.

Architecture Overview

The application follows a microservices-inspired architecture with separate backend and frontend components:

Backend Architecture

• Express.js Server: Handles HTTP requests with middleware for security, logging, and rate limiting
• AI Integration: Interfaces with Google Gemini 1.5 Flash API for natural language processing
• Data Processing: Supports text input and DOCX file parsing using Mammoth.js
• Email Service: SMTP-based email delivery using Nodemailer
• Validation Layer: Input sanitization and schema validation using Joi
• Logging System: Structured logging with Winston and daily rotation
• Metrics Collection: Performance monitoring and request tracking

Frontend Architecture

• React Application: Component-based UI with lazy loading for performance
• State Management: Local state with React hooks for form data and API responses
• HTTP Client: Axios with retry logic and error handling
• Styling: Utility-first CSS with Tailwind CSS
• Build System: Vite for fast development and optimized production builds

Features

• AI-powered transcript summarization using Google Gemini 1.5 Flash
• File upload support for DOCX documents
• Customizable summarization prompts with predefined templates
• Real-time summary generation with progress indicators
• Email delivery of generated summaries
• Rate limiting and request throttling
• Comprehensive error handling and validation
• Responsive web interface
• Request logging and metrics collection

Technology Stack

Backend Dependencies

• Node.js >= 18.0.0
• Express 5.1.0
• Axios 1.11.0 (with retry plugin)
• Joi 18.0.1
• Winston 3.17.0
• Nodemailer 7.0.5
• Helmet 8.1.0
• Compression 1.8.1
• Express Rate Limit 8.1.0

Frontend Dependencies

• React 19.1.1
• Vite 7.1.2
• Tailwind CSS 4.1.12
• Axios 1.12.1
• Mammoth 1.10.0

Prerequisites

• Node.js 18+ runtime
• Google Gemini API key
• SMTP email service credentials (optional)

Installation

Repository Setup

git clone <repository-url>
cd TranscriptAI

Backend Configuration

cd backend
npm install

Create environment configuration:

cp .env.example .env

Edit .env with required variables:

AI_API_KEY=your_gemini_api_key_here
PORT=5000
LOG_LEVEL=info
# Email configuration
EMAIL_HOST=smtp.gmail.com
EMAIL_PORT=587
EMAIL_SECURE=false
EMAIL_USER=your_email@gmail.com
EMAIL_PASS=your_app_password

Start the backend server:

npm start

Frontend Configuration

cd ../frontend
npm install
npm run dev

API Reference

POST /v1/summarize

Generates AI-powered summary of provided transcript text.

Request Body:

{
  "transcript": "string (required)",
  "prompt": "string (required)"
}

Response:

{
  "success": true,
  "summary": "Generated summary text"
}

Error Response:

{
  "success": false,
  "error": "Error description"
}

Rate Limiting: 100 requests per 15 minutes per IP

GET /health

Health check endpoint returning server status.

Response:

{
  "status": "ok",
  "timestamp": "2025-09-15T17:28:00.000Z",
  "uptime": 3600
}

GET /metrics

Application metrics and performance data.

Response:

{
  "requests_total": 150,
  "requests_duration_ms": 2500,
  "active_connections": 5
}

Development

Running Tests

cd backend
npm test

Linting

cd frontend
npm run lint

Building for Production

cd frontend
npm run build

Security Considerations

• Input validation and sanitization using Joi schemas
• Rate limiting to prevent API abuse
• Helmet.js for security headers
• CORS configuration for cross-origin requests
• Environment variable management for sensitive data
• Request logging for audit trails

Error Handling

The application implements comprehensive error handling:

• ValidationError: Invalid input data (400)
• ExternalAPIError: AI service failures (502)
• RateLimitError: Request throttling (429)
• InternalServerError: Unexpected errors (500)

All errors are logged with structured data and return consistent JSON responses.