Smart Reply, Enhance & Translate

AI-powered assistant to generate context-aware replies, enhance your writing, and translate text — directly from your browser, web app, or mobile device.

---

Table of Contents

• Overview
• Features
• Architecture
• Tech Stack
• Getting Started
  • Prerequisites
  • Clone the Repository
  • Backend Setup
  • Browser Extension Setup
  • Web Frontend Setup
  • Flutter Mobile App Setup
• API Reference
• Caching System
• Deployment
• Contributing
• License

---

Overview

Smart Reply, Enhance & Translate is a full-stack AI tool suite built to boost productivity for anyone who works with text. Whether you're a support agent handling repetitive messages, a freelancer drafting professional emails, or someone who needs quick translations, this project provides a seamless experience across multiple platforms.

The suite consists of three client interfaces — a browser extension (Chrome/Firefox), a React web application, and a Flutter mobile app — all powered by a shared Node.js backend that communicates with state-of-the-art LLMs via OpenRouter.

Three core modes:

Mode	What it does
Smart Reply	Generates up to 4 context-aware reply suggestions for any message
Smart Enhance	Improves your text across grammar, clarity, conciseness, and structure
Smart Translate	Translates text into your target language with 4 style variations

---

Features

• 4 AI suggestions per request across all three modes (Reply, Enhance, Translate)
• 6 response formats — Professional, Friendly, Casual, Formal, Flirty, Romantic — with tooltip descriptions
• Multi-language translation — English, Spanish, French, German, Chinese, Arabic, Bengali, and more
• Multiple free LLM models — automatically selects the best model per operation type
• LLM-agnostic — compatible with any OpenRouter-supported model (free or paid)
• Custom API key support — bring your own key for full privacy and control
• SHA256-based intelligent caching — prevents duplicate API calls, 5-minute TTL
• Docker support — spin up the backend with a single command
• Browser Extension (Chrome/Firefox) — insert text directly into any field, keyboard shortcut Ctrl+Shift+T, context menu integration, auto-detect selected text
• Web App — animated UI with Framer Motion, auto-resizing textarea, keyboard shortcuts (Ctrl/Cmd + Enter), responsive and accessible
• Mobile App (Flutter) — native Android support, optimized for on-the-go usage

---

Architecture

smart-reply/
├── backend/              # Node.js + Express API server
├── extension/            # Browser extension (Chrome/Firefox, Manifest V3)
├── frontend/             # React web application
└── smart_reply_app/      # Flutter mobile application

All clients communicate with the same backend via three REST endpoints. The backend handles LLM selection, prompt construction, caching, and response formatting.

---

Tech Stack

Layer	Technology
Backend	Node.js, Express
Web Frontend	React, Zustand, Tailwind CSS, Framer Motion, Lucide React
Browser Extension	Manifest V3, Vanilla JS, HTML, CSS
Mobile App	Flutter, Provider, Google Fonts
AI Integration	OpenRouter API
Containerization	Docker
Deployment	Vercel (frontend), Render / Railway / Fly.io (backend)

---

Getting Started

Prerequisites

Before you begin, make sure you have the following installed:

• Node.js v18 or higher
• npm v9 or higher
• Flutter SDK (only for mobile)
• Docker (optional, for containerized backend)
• An OpenRouter API key

---

Clone the Repository

git clone https://github.com/mahmud-r-farhan/smart-reply.git
cd smart-reply

---

Backend Setup

The backend is a single Express server that serves all three clients. You only need to set it up once.

Option A — Local Setup

1. Navigate to the backend directory:

cd backend

2. Create a .env file:

OPENROUTER_API_KEY=your_openrouter_api_key
PORT=5006

3. Install dependencies and start the server:

npm install
npm start

The backend will be running at http://localhost:5006.

Option B — Docker Setup

docker build -t smart-reply-backend ./backend
docker run -d -p 5006:5006 --env OPENROUTER_API_KEY=your_openrouter_api_key smart-reply-backend

Verify it's running by visiting http://localhost:5006/health.

---

Browser Extension Setup

Chrome

1. Open Chrome and navigate to chrome://extensions
2. Enable Developer Mode (toggle in the top-right corner)
3. Click Load unpacked and select the extension/ folder
4. The Smart Reply icon will appear in your toolbar

Firefox

1. Open Firefox and navigate to about:debugging#/runtime/this-firefox
2. Click Load Temporary Add-on...
3. Select the manifest.json file inside the extension/ folder
4. The extension will be loaded for the current browser session

Connect the extension to your backend:

1. Click the extension icon in the toolbar
2. Open Settings
3. Paste your backend base URL (e.g., http://localhost:5006/api)
4. Optionally set default source and target languages for quick translation
5. Click Save

---

Web Frontend Setup

1. Navigate to the frontend directory:

cd frontend

2. Create a .env file:

VITE_API_ENDPOINT=http://localhost:5006/api

3. Install dependencies and start the development server:

npm install
npm run dev

The app will be available at http://localhost:5173.

---

Flutter Mobile App Setup

1. Navigate to the mobile app directory:

cd smart_reply_app

2. Configure the API endpoint in lib/utils/constants.dart:

// For Android Emulator (default):
const String baseUrl = 'http://10.0.2.2:5006/api';

// For a physical device, replace with your local machine's IP:
const String baseUrl = 'http://192.168.1.X:5006/api';

3. Get dependencies and run the app:

flutter pub get
flutter run

---

API Reference

All endpoints accept and return JSON. The base URL is configurable (default: http://localhost:5006).

---

POST /api/suggest-reply

Generates up to 4 context-aware reply suggestions for a given message.

Request body:

{
  "message": "Can we reschedule the meeting?",
  "format": "professional"
}

Response:

{
  "suggestions": [
    "Absolutely, let's find a new time that works for both of us.",
    "Sure, please let me know your availability.",
    "No problem, I can adjust my schedule accordingly.",
    "Of course, happy to reschedule. When would you like to meet?"
  ]
}

---

POST /api/enhance-text

Generates up to 4 improved variations of your input text.

Request body:

{
  "text": "We need to reschedul the meeting because of conflict.",
  "format": "professional"
}

Response:

{
  "enhancements": [
    "We need to reschedule the meeting due to a scheduling conflict.",
    "Due to a conflict, let's reschedule the meeting at your earliest convenience.",
    "I apologize, but a conflict has arisen — could we reschedule the meeting?",
    "There's a scheduling conflict; please suggest alternative times for the meeting."
  ]
}

---

POST /api/translate-text

Translates the given text into the target language with style variations.

Request body:

{
  "text": "Hello, how are you?",
  "format": "friendly",
  "language": "bengali"
}

Response:

{
  "translations": [
    "হ্যালো, আপনি কেমন আছেন?",
    "হাই, তুমি কেমন আছো?",
    "হেলো, তোমার খবর কী?",
    "হ্যালো, আপনার অবস্থা কেমন?"
  ]
}

---

GET /health

Returns a simple health check response.

Response:

{ "status": "ok" }

---

Supported format values for all endpoints:

Value	Description
professional	Clear, formal, business-appropriate
friendly	Warm and approachable
casual	Relaxed, everyday language
formal	Strict and official tone
flating	Flirty compliments with light romantic interest
romantic	Affectionate and emotionally warm

---

Caching System

As of v0.4, the backend uses an intelligent SHA256-based caching layer to avoid redundant API calls and reduce latency.

How it works:

• A unique cache key is generated from the full prompt and model identifier using SHA256 hashing
• Cache entries expire automatically after 5 minutes (TTL)
• Different prompts always produce different cache keys, eliminating false cache hits

Example behavior:

Request 1: "Hello"    → Cache MISS  → Calls OpenRouter → Stores in cache
Request 2: "Hi"       → Cache MISS  → Calls OpenRouter → Stores in cache
Request 1 (repeat)    → Cache HIT   → Returns instantly (no API call)

Performance improvements in v0.4:

Metric	Before	After
Response Size	100%	~32% (65–70% reduction)
API Cache Hits	0%	~60%
Cache Key Collisions	High	0% (SHA256 hashing)
React Re-renders	High	Low (40–50% reduction)
Bundle Size	100%	~87% (12–15% reduction)
Request Cancellation	❌	✅ Prevents zombie requests
Security Headers	❌	✅ XSS/Clickjacking protection

---

Deployment

The backend can be deployed to any Node.js-compatible cloud platform. Set the OPENROUTER_API_KEY environment variable, then update your clients' API endpoint accordingly.

Recommended platforms:

• Render — free tier available, easy GitHub integration
• Railway — fast deploys, simple environment management
• Fly.io — global edge deployment with Docker support
• Any VPS (Ubuntu + PM2 or Docker)

Frontend deployment:

The React web app deploys seamlessly to Vercel. Set VITE_API_ENDPOINT as an environment variable in your Vercel project settings.

---

Contributing

Contributions of all kinds are welcome — bug fixes, new features, documentation improvements, translations, and more.

How to contribute:

1. Fork the repository
2. Create a new branch for your changes:

   git checkout -b feature/your-feature-name

3. Make your changes with clear, descriptive commits
4. Test your changes thoroughly
5. Open a pull request with a summary of what you've changed and why

Guidelines:

• Keep pull requests focused on a single change
• Write clear commit messages (e.g., fix: correct cache key collision on empty input)
• If you're adding a new feature, consider updating this README accordingly
• Report bugs or suggest improvements via GitHub Issues

---

License

This project is open source. See the repository for license details.

---

Built by Mahmud Rahman · Powered by OpenRouter

"Automate your responses. Amplify your productivity."