⬡ NeuralChat AI

Real-time AI chat platform with streaming responses, conversation memory, and multi-model support.

Features

• 🔴 Real-time streaming — responses stream token by token via WebSocket
• 🧠 Conversation memory — full message history sent with each request (last 20 messages)
• 🤖 Multi-model support — switch between [Model 1] and [Model 2] mid-conversation
• 💬 Multiple conversations — sidebar with conversation list and history
• ⚡ Rate limiting — 60 requests/minute per IP
• 🔒 Helmet.js security headerS

Tech Stack

Layer	Tech
Frontend	React 18, Socket.io-client
Backend	Node.js, Express, Socket.io
AI	Groq API (streaming)
Real-time	WebSockets

Getting Started

```bash

Clone the repo

git clone https://github.com/jatin/neuralchat-ai cd neuralchat-ai

Install all dependencies

npm run install:all

Set up environment

cp .env.example .env

Add your GROQ_API_KEY to .env

Run server + client (in two terminals)

npm run dev # server on :4000 npm run client # React on :3000 ```

Project Structure

``` neuralchat-ai/ ├── server/ │ └── index.js # Express + Socket.io server, Groq streaming ├── client/ │ └── src/ │ ├── App.jsx # Main chat UI │ └── App.css # Styles ├── .env.example └── package.json ```

How It Works

1. Client connects via WebSocket on load
2. User creates a conversation → server stores it in memory with a UUID
3. User sends a message → server calls Groq with full history
4. Groq streams tokens back → server forwards each delta to client via stream:delta events
5. Client appends deltas in real-time → smooth streaming effect

---

Built by Jatin — 2026