# ChipiAI Chat — Realtime AI‑Powered Chat Platform

ChipiAI Chat is a modular, production‑ready chat system built with Go, gRPC, WebSockets, MongoDB, Kafka, and Nuxt. It supports real‑time messaging, AI responses via callback gRPC, persistent chat history, and multi‑room architecture.

This project demonstrates a complete modern backend pipeline:
WebSocket → gRPC → AI Service → Callback → MongoDB → WebSocket.

---

## 🚀 Features

- **Go/Echo**
- **Realtime WebSocket chat**
- **AI message processing** via gRPC (ContextEnhancementService)
- **Callback server** for AI responses
- **MongoDB message storage**
- **Room creation and message history**
- **JWT authentication**
- **Kafka consumer integration**
- Clean, modular Go architecture

---

## 📦 Project Structure

---

## 🔧 Technologies Used

| Component | Technology |
|-----------|------------|
| Backend | Go 1.22 |
| Realtime | WebSocket (gorilla/websocket) |
| AI Processing | gRPC |
| Storage | MongoDB |
| Authentication | JWT |
| Frontend | Nuxt 3 |
| Messaging | Kafka |
| Deployment | Docker Compose |

---

## ⚙️ Setup & Installation

### 1. Clone the repository

```bash
git clone https://github.com/pasaz/ChipiAiChat.git
cd ChipiAiChat

2. Start infrastructure (Mongo, Kafka, Postgres)

docker-compose up -d

3. Run the backend

go run main.go

Backend starts on:

• HTTP: http://localhost:8080, http://localhost:7070
• gRPC: :50051, :50052, :50053

---

🔐 Authentication

The chat endpoint requires a valid JWT:

Authorization: Bearer <token>

Token is validated using:

JWT_KEY_CHAT

from .env file.

---

💬 WebSocket API

Connect

ws://localhost:7070/chat?roomId=<roomId>

Send message

Client sends plain text:

"Hello AI"

Receive AI response

Callback server pushes AI messages back through the same WebSocket.

---

🧠 gRPC Services

MessageService (client → AI)

message NewMessageContent {
  string username = 1;
  string content  = 2;
  string roomId   = 3;
}

service MessageService {
  rpc Message(NewMessageContent) returns (Empty);
}

AiCallback (AI → backend)

message AiMessage {
  string username = 1;
  string content  = 2;
  string roomId   = 3;
}

service AiCallback {
  rpc SendAiMessage(AiMessage) returns (google.protobuf.Empty);
}

---

🗄️ MongoDB Models

Message

type Message struct {
    ID        primitive.ObjectID `bson:"_id,omitempty"`
    RoomID    string             `bson:"roomId"`
    Username  string             `bson:"username"`
    Content   string             `bson:"content"`
    CreatedAt time.Time          `bson:"createdAt"`
}

Room

type Room struct {
    ID        primitive.ObjectID `bson:"_id,omitempty"`
    Name      string             `bson:"name"`
    CreatedAt time.Time          `bson:"createdAt"`
}

---

📡 REST API

Create room

POST /create/room

Request body:

{
  "name": "General"
}

Response:

{
  "roomId": "6778c3f2e4b0a3c1d2f9a123"
}

Get room messages

GET /get/room/:id

Response:

[
  {
    "roomId": "6778c3f2e4b0a3c1d2f9a123",
    "username": "Sava",
    "content": "Hello",
    "createdAt": "2026-01-04T14:30:00Z"
  }
]

---

🧪 Kafka Consumer

The consumer listens for user messages and processes them asynchronously.

go consumer.GetMessageUser(&wg)

---

🧩 Callback Logic

AI sends a gRPC callback:

ws := chat.Connections[req.Username]
ws.WriteMessage(1, []byte(req.Content))

database.SaveMessage(req.RoomId, req.Username, req.Content)