Email Assistant — AI-Powered Email Triage

(React • Node.js • LangChain.js • Gmail API • OAuth 2.0 • Gemini LLM)

An AI assistant that connects securely to Gmail, fetches emails, and auto-prioritizes them into High / Medium / Low categories using Google Gemini via LangChain.js. Built with a React client for the interface and a Node.js server to handle Gmail OAuth, email retrieval, and LLM classification.

---

✨ Features

• Secure Gmail OAuth 2.0 login for user-specific email access.
• Email priority classification (High / Medium / Low) powered by Google Gemini LLM + LangChain.js.
• React client with a simple, intuitive UI for viewing categorized emails.
• Node.js server with APIs for authentication, fetching emails, and classification.
• Lightweight design — no external DB/storage required; real-time classification only.

---

🧱 Project Structure

Email_Assistant/
├─ client/                # React app – UI
│  ├─ src/
│  │  ├─ components/      # Reusable UI parts
│  │  ├─ pages/           # Login, Inbox, Priority views
│  │  ├─ lib/             # API client
│  │  └─ main.jsx|tsx
│  └─ package.json
│
├─ server/                # Node.js + Express + LangChain.js
│  ├─ routes/             # /auth, /emails, /classify
│  ├─ utils/              # helper functions
│  └─ server.js           # entry point
│  └─ package.json
│
└─ README.md

---

🛠️ Tech Stack

• Frontend: React, Axios/Fetch, React Router
• Backend: Node.js, Express, LangChain.js
• Auth & Mail: Gmail API, OAuth 2.0
• LLM: Google Gemini (API key) via LangChain.js

---

🔐 Environment Variables

Create server/.env:

# Gmail API
GMAIL_SCOPES=https://www.googleapis.com/auth/gmail.readonly

# Gemini LLM
GEMINI_API_KEY=your_gemini_api_key

⚠️ The client does not require .env — it only consumes the backend APIs.

---

🚀 Local Development

1) Server

cd server
npm install
npm run dev

2) Client

cd client
npm install
npm run dev

• Server runs on → http://localhost:4000
• Client runs on → http://localhost:5173

---

🔗 Core API Endpoints

Method	Endpoint	Description
GET	/api/auth/google	Start Google OAuth flow
GET	/api/auth/google/callback	OAuth redirect handler
GET	/api/emails	Fetch user emails
POST	/api/classify	Classify emails via Gemini LLM

---

🧠 Email Classification Logic

• Emails are fetched via Gmail API.
• Each email snippet is passed to LangChain.js → Gemini LLM.
• The LLM responds with a priority label (High / Medium / Low) based on context.

Example response:

{
  "emailId": "1783ac9...",
  "subject": "Interview scheduled for Monday",
  "priority": "High"
}

---

📌 Future Enhancements

• Support for multi-account connections.
• Smart summaries per category.
• Custom rules engine (user-defined filters).

---

⚡ Built to make your inbox actionable, not overwhelming.