BeyondChats – Full Stack & AI Content System

This repository contains my submission for the Full Stack Engineer / Technical Product Manager (Fresher) role at BeyondChats.

The objective of this assignment was to demonstrate:

• End-to-end system thinking
• Backend–frontend integration
• Practical scraping & API design
• Ability to ship reliable software under time constraints

---

🧩 Project Overview

The system is divided into three logical phases, as requested:

1. Laravel Backend (Core System)

• Scrapes the 5 oldest BeyondChats blog articles
• Normalizes and stores structured content in MySQL
• Exposes RESTful CRUD APIs
• Acts as the single source of truth for content

2. Node.js + AI Pipeline (Minimal Implementation)

• Fetches the latest article from Laravel APIs
• Collects competitor references (mocked / configurable)
• Generates an improved version using an LLM-style workflow
• Publishes generated content back to Laravel as a new article

Phase 2 is intentionally minimal but functional, prioritizing correctness and extensibility over excessive orchestration.

3. React Frontend

• Fetches articles from the Laravel API
• Displays original vs generated articles distinctly
• Deployed as a static frontend for fast delivery

---

🏗️ Tech Stack

Backend

• Laravel 9
• PHP 8
• MySQL
• Guzzle HTTP Client
• Symfony DomCrawler
• Docker (production)

Frontend

• React (Create React App)
• Fetch API
• Responsive CSS

AI / Automation

• Node.js
• LLM-style text generation workflow
• Extensible for OpenAI / Gemini / Claude APIs

---

🔄 System Architecture

[ React (Netlify) ] | v [ Laravel API (Render) ] | v [ MySQL Database (Railway) ]

[ Node.js AI Service ] | v [ Laravel API ]

---

🔄 Data Flow

Phase 1: Scraping & Storage

1. Laravel scrapes BeyondChats blog listing
2. Extracts valid article URLs
3. Scrapes title & content
4. Stores articles with source_type = original

Phase 2: AI Content Enhancement

1. Node.js fetches the latest article
2. Generates an improved version (simulated LLM workflow)
3. Saves new article with:
   • source_type = generated
   • reference_urls

Phase 3: Presentation

1. React frontend calls /api/articles
2. Displays articles with clear labeling
3. Separates original vs generated content visually

---

🗄️ Database Schema

articles

Column	Type
id	bigint
title	string
content	longText
source_url	string
source_type	enum (original, generated)
reference_urls	json (nullable)
created_at	timestamp
updated_at	timestamp

---

🔌 API Endpoints

Scraping

• GET /api/scrape-articles
  Scrapes and stores the 5 oldest BeyondChats articles.

Articles CRUD

• GET /api/articles
• GET /api/articles/{id}
• POST /api/articles
• PUT /api/articles/{id}
• DELETE /api/articles/{id}

---

🖥️ Local Setup

1. Clone Repository

git clone https://github.com/your-username/beyondchats-fullstack-assignment.git
cd beyondchats-fullstack-assignment


2. Backend Setup
cd backend-laravel/beyondchats-backend
composer install
php artisan key:generate
php artisan migrate
php artisan serve


3. Frontend Setup
cd frontend-react/beyondchats-frontend
npm install
npm start


4. AI Service
cd ai-node
npm install
node index.js



🌐 Live Deployment

Frontend (Netlify)
https://fullstack-and-ai-content-sytem.netlify.app/

Backend API (Render)
https://beyondchats-fullstack-assignment.onrender.com/

Database
Railway Cloud MySQL



⚖️ Trade-offs & Design Decisions

Prioritized a stable backend pipeline over complex AI orchestration

Used CRA for predictable frontend behavior

Implemented AI pipeline minimally but correctly

Focused on correctness, clarity, and extensibility

🧪 Evaluation Alignment

This submission emphasizes:

Backend completeness

Clear system design

Production-style deployment

Honest engineering trade-offs



🙋‍♂️ Final Notes

This project reflects how I approach real-world problems:

Build incrementally

Deploy early

Make conscious trade-offs

Optimize for reliability and clarity

Thank you for reviewing my submission.

— Rushikesh Dharme