CoreOS AI — Multi-LLM Realtime GenAI Search Engine

CoreOS AI is a production-grade Multi-LLM AI Assistant & Search Engine inspired by Perplexity AI and ChatGPT.

It combines:

• Multi-LLM Architecture
• Realtime AI Routing
• Retrieval-Augmented Generation (RAG)
• Smart API-based Realtime Intelligence
• Google OAuth Authentication
• Context-aware AI responses
• Streaming AI Responses
• Realtime Search Intelligence
• API-first Architecture

to deliver fast, intelligent, and realtime AI-powered experiences.

---

🚀 Project Vision

Traditional Search Engines → Links

CoreOS AI → Direct Intelligent Answers

CoreOS AI is designed to:

• understand user intent
• fetch realtime information
• reduce hallucinations
• provide contextual AI responses
• optimize speed and token usage
• support multiple AI models dynamically
• support scalable AI infrastructure
• support realtime API integrations

---

🧠 Core Features

🔥 Multi-LLM Architecture

CoreOS dynamically routes requests across multiple LLM providers.

Integrated Models

Google AI

Gemini 2.5 Flash

Used for:

• fast responses
• large context handling
• realtime chat
• reasoning
• production chat performance

Gemini 2.5 Flash-Lite

Used for:

• lightweight tasks
• query classification
• token optimization
• ultra-fast responses

---

OpenAI

GPT-4o Mini

Used for:

• structured responses
• coding tasks
• debugging
• high-quality formatting
• advanced assistant tasks

---

Groq

Llama 3.1 8B Instant

Used for:

• ultra-fast responses
• low-latency AI
• lightweight conversations
• streaming responses

---

Mistral AI

Mistral Small

Used for:

• reasoning support
• alternative inference routing
• balanced performance

---

⚡ Smart Query Routing

CoreOS intelligently classifies user queries into categories.

Query Types

• General Knowledge
• Realtime Queries
• Web Research
• Coding Queries
• Casual Conversations
• Crypto Queries
• Cricket / IPL Queries

This prevents unnecessary searches and improves:

• speed
• token optimization
• response quality
• realtime accuracy

---

🌐 Realtime Intelligence Layer

CoreOS uses API-first realtime architecture instead of relying only on pretrained AI knowledge.

---

🏏 Cricbuzz Cricket API Integration

Provider

RapidAPI

API

Cricbuzz Cricket API

Features

• Live IPL scores
• Match scorecards
• Recent matches
• Upcoming matches
• Match summaries
• Team statistics
• Live match status

Endpoints Used

/matches/v1/live
/matches/v1/recent
/matches/v1/upcoming
/mcenter/v1/{matchId}/hscard

Headers

x-rapidapi-host
x-rapidapi-key

---

💰 CoinGecko API Integration

Features

• Bitcoin prices
• Ethereum prices
• Solana prices
• Realtime crypto market updates
• Token price tracking

Endpoints Used

/api/v3/simple/price

Supported Coins

• Bitcoin
• Ethereum
• Solana

---

🔎 Tavily Search Integration

Purpose

Used as:

• fallback search engine
• web research system
• internet retrieval layer
• external knowledge fetcher

Features

• realtime web search
• article retrieval
• search summarization
• contextual search results

---

🧠 Retrieval-Augmented Generation (RAG)

CoreOS implements RAG architecture.

Purpose

• retrieve relevant context
• reduce hallucinations
• improve factual accuracy
• support intelligent answers
• support contextual responses

Flow

User Query
   ↓
Retriever
   ↓
Relevant Context
   ↓
LLM Processing
   ↓
Final Response

---

🔐 Authentication System

CoreOS uses Google OAuth Authentication with Passport.js.

Technologies Used

• Passport.js
• passport-google-oauth20
• JWT Authentication
• Express Sessions

Features

• Google Login
• Secure JWT generation
• OAuth callback handling
• Protected authentication flow
• User session management

OAuth Routes

/auth/google
/auth/google/callback

---

💬 Conversational AI Features

CoreOS supports:

• multi-turn conversations
• contextual memory
• dynamic chat routing
• AI streaming responses
• model switching
• realtime conversations

---

⚡ Performance Optimizations

The system is optimized for:

• low latency
• fast responses
• streaming output
• reduced token usage
• scalable AI routing
• efficient API handling
• compact prompts
• minimized search overhead

---

🏗️ Complete Tech Stack

Frontend

• React.js
• Vite
• Tailwind CSS
• JavaScript
• Axios
• React Hooks

---

Backend

• Node.js
• Express.js
• REST APIs
• Middleware Architecture

---

AI & LLM Providers

• Gemini API
• OpenAI API
• Groq API
• Mistral AI

---

AI Frameworks

• LangChain

---

Authentication

• Passport.js
• JWT
• Google OAuth 2.0

---

Database

• MongoDB
• Mongoose

---

Realtime APIs

• Cricbuzz Cricket API
• CoinGecko API
• Tavily Search API

---

📦 Major NPM Packages Used

Backend Packages

npm install express
npm install mongoose
npm install cors
npm install dotenv
npm install axios
npm install jsonwebtoken
npm install bcryptjs
npm install passport
npm install passport-google-oauth20
npm install express-session
npm install cookie-parser
npm install multer
npm install socket.io
npm install @google/generative-ai
npm install openai
npm install groq-sdk
npm install @mistralai/mistralai
npm install langchain
npm install @langchain/core
npm install @langchain/community
npm install tavily

---

Frontend Packages

npm install react
npm install react-dom
npm install react-router-dom
npm install axios
npm install tailwindcss
npm install lucide-react
npm install socket.io-client
npm install framer-motion

---

⚙️ Environment Variables

Create a .env file.

# DATABASE
MONGO_URI=your_mongodb_uri

# JWT
JWT_SECRET=your_jwt_secret

# GOOGLE OAUTH
GOOGLE_CLIENT_ID=your_google_client_id
GOOGLE_CLIENT_SECRET=your_google_client_secret

# AI PROVIDERS
GOOGLE_API_KEY=your_gemini_api_key
OPENAI_API_KEY=your_openai_api_key
GROQ_API_KEY=your_groq_api_key
MISTRAL_API_KEY=your_mistral_api_key

# SEARCH
TAVILY_API_KEY=your_tavily_api_key

# RAPID API
RAPIDAPI_KEY=your_rapidapi_key
RAPIDAPI_HOST=cricbuzz-cricket.p.rapidapi.com

# COINGECKO
COINGECKO_API_KEY=your_coingecko_api_key

---

🏛️ System Architecture

User Query
    ↓
Smart Query Classifier
    ↓
Realtime Detection Layer
    ↓
API / RAG Retrieval
    ↓
Multi-LLM Routing
    ↓
Context Aggregation
    ↓
Streaming Final Response

---

🧠 Multi-LLM Routing Strategy

Query Type	Model Used
Fast General Chat	Gemini 2.5 Flash
Lightweight Queries	Gemini 2.5 Flash-Lite
Coding & Structured Output	GPT-4o Mini
Ultra Fast Responses	Llama 3.1 8B Instant
Reasoning Support	Mistral Small

---

🔄 Realtime Query Flow

Cricket Query Example

User asks:
Current IPL score

↓
Query Detection

↓
Cricbuzz API

↓
Extract Important Data

↓
LLM Response Generation

↓
Final Answer

---

💰 Crypto Query Flow

User asks:
Bitcoin price right now

↓
Crypto Query Detection

↓
CoinGecko API

↓
Realtime Price Extraction

↓
LLM Response

---

🔍 News / Research Flow

User asks:
Latest AI news

↓
Tavily Search

↓
Search Summarization

↓
LLM Context Processing

↓
Final Answer

---

🔐 Google OAuth Flow

User clicks Google Login

↓
/auth/google

↓
Google OAuth Consent

↓
/auth/google/callback

↓
JWT Generation

↓
Authenticated Session

---

📁 Project Structure

CoreOS/
│
├── backend/
│   ├── services/
│   ├── sockets/
│   ├── validators/
│   ├── routes/
│   ├── controllers/
│   ├── middleware/
│   ├── models/
│   ├── app.js
│   ├── server.js
│   └── .env
│
├── frontend/
│   ├── src/
│   │   ├── app/
│   │   ├── auth/
│   │   ├── chat/
│   │   ├── pages/
│   │   ├── service/
│   │   ├── hooks/
│   │   └── assets/
│   │
│   ├── vite.config.js
│   └── package.json
│
└── README.md

---

🛠️ Installation & Setup

1. Clone Repository

git clone https://github.com/afdhruval/perplexity.git
cd perplexity

---

2. Install Backend Dependencies

npm install

---

3. Install Frontend Dependencies

cd frontend
npm install

---

4. Setup Environment Variables

Create .env

Add all required API keys.

---

5. Run Backend

npm run dev

---

6. Run Frontend

cd frontend
npm run dev

---

💡 Example Queries

Realtime Queries

Current IPL score

Bitcoin price right now

Latest AI news

---

Coding Queries

Explain React hooks

Fix this Node.js error

---

Research Queries

Explain Retrieval-Augmented Generation

---

🔥 Advanced Engineering Highlights

• Multi-LLM orchestration
• Realtime AI architecture
• Smart query routing
• API-first realtime intelligence
• RAG-powered retrieval system
• Streaming AI responses
• Token optimization
• Google OAuth integration
• Production-grade backend structure
• Scalable AI architecture
• Context-aware conversations
• Dynamic model switching

---

🧩 Engineering Concepts Used

• Retrieval-Augmented Generation (RAG)
• Multi-LLM orchestration
• Prompt engineering
• API-based realtime systems
• Query classification
• JWT authentication
• OAuth 2.0
• Streaming architecture
• Context aggregation
• AI response optimization
• Low-latency systems
• Realtime retrieval systems

---

🎯 Use Cases

• AI Search Engine
• AI Research Assistant
• Realtime Information Assistant
• AI Chatbot Platform
• Coding Assistant
• Crypto Assistant
• Sports Realtime Assistant
• Knowledge Assistant
• GenAI Demonstration Platform

---

🚀 Future Improvements

• Voice assistant support
• PDF RAG support
• Pinecone vector database integration
• AI agents
• Web scraping pipelines
• Advanced memory systems
• Team collaboration
• Multi-user chat rooms
• AI workflow automation

---

👨‍💻 Author

Dhruval

---

🏁 Final Note

CoreOS AI demonstrates real-world implementation of:

• Multi-LLM Systems
• RAG Architecture
• Realtime AI Engineering
• Smart Query Routing
• Production-grade AI Infrastructure
• Modern AI Application Design
• Scalable AI Backend Systems
• Realtime API Integrations
• AI Search Architecture

It is designed as a scalable foundation for next-generation AI-powered assistant systems.