Backend Engineering Assignment

Overview

This project is a Spring Boot microservice that uses PostgreSQL as the source of truth for posts and comments, and Redis as the stateless guardrail layer for concurrency control, cooldowns, virality scoring, and notification batching.

Stack

• Java 17
• Spring Boot 3.3
• Spring Data JPA
• Spring Data Redis
• PostgreSQL 16
• Redis 7
• Docker Compose
• JUnit 5 / Mockito

Implemented Requirements

• POST /api/posts to create a post
• POST /api/posts/{postId}/comments to create a comment
• POST /api/posts/{postId}/like to like a post
• Redis virality score updates in real time
• Horizontal bot reply cap of 100 per post
• Vertical thread depth cap of 20
• Bot-to-human cooldown of 10 minutes
• Notification throttling with pending Redis list batching
• Scheduled sweeper every 5 minutes

Data Model

PostgreSQL entities

• User: id, username, is_premium
• Bot: id, name, persona_description
• Post: id, author_id, author_type, content, created_at
• Comment: id, post_id, parent_comment_id, author_id, author_type, content, depth_level, created_at

author_type was added alongside author_id so the service can safely distinguish whether an actor is a human or a bot.

Guardrail Design

Virality score

Redis key: post:{id}:virality_score

• Bot reply: +1
• Human like: +20
• Human comment: +50

Horizontal cap

Redis key: post:{id}:bot_count

• Every bot reply first reserves a slot using atomic Redis INCR
• If the incremented value is greater than 100, the request is rejected with 429 Too Many Requests
• The slot is immediately released with DECR
• This ensures that even under concurrent request races, only the first 100 bot replies are allowed through

Transaction integrity

• Redis guardrails are checked before the database write is considered successful
• If a database transaction fails after a bot slot is reserved, the reserved slot is released during rollback handling
• This prevents Redis counters from drifting away from persisted database state

Vertical cap

• Nested comments compute depthLevel = parent.depthLevel + 1
• Any comment request producing depthLevel > 20 is rejected

Cooldown cap

Redis key: cooldown:bot_{botId}:human_{humanId}

• TTL: 10 minutes
• If the key already exists, the bot interaction is rejected with 429 Too Many Requests

Notification Engine

When a bot interacts with a user's post:

• If the user has not received a notification in the last 15 minutes, the app logs:
  • Push Notification Sent to User
• A notification cooldown key is then set in Redis
• If the user is still inside the cooldown window, the message is pushed into:
  • user:{id}:pending_notifs
• The user id is also registered in:
  • pending_notif_users

The scheduled sweeper runs every 5 minutes and:

• scans all users with pending notifications
• drains their Redis list
• logs a summarized message such as:
  • Summarized Push Notification: Bot X and [N] others interacted with your posts.

Running Locally

1. Start PostgreSQL and Redis

docker compose up -d

2. Run the application

mvn spring-boot:run

3. Run tests

mvn test

Default Configuration

• PostgreSQL URL: jdbc:postgresql://localhost:5432/assignment_db
• PostgreSQL username: postgres
• PostgreSQL password: postgres
• Redis host: localhost
• Redis port: 6379

Supported environment overrides:

• DB_URL
• DB_USERNAME
• DB_PASSWORD
• REDIS_HOST
• REDIS_PORT

Seed Data

The app starts with sample seed data from src/main/resources/data.sql:

• Users:
  • 1 -> alice
  • 2 -> bob
• Bots:
  • 1 -> Nova
  • 2 -> Atlas

This makes the Postman collection usable immediately after startup.

API Examples

Create post

POST /api/posts

{
  "authorType": "USER",
  "authorId": 1,
  "content": "Hello world"
}

Create comment

POST /api/posts/{postId}/comments

{
  "authorType": "BOT",
  "authorId": 1,
  "parentCommentId": null,
  "content": "Automated reply"
}

Like post

POST /api/posts/{postId}/like

{
  "userId": 1
}

Testing Coverage

Automated tests cover:

• the 200 concurrent bot request spam case, asserting exactly 100 successful bot replies
• rejection when comment depth exceeds 20
• rejection when a bot hits the same human again within the 10 minute cooldown
• notification batching behavior while the user notification cooldown is active

Submission Notes

• The repository includes source code, docker-compose.yml, README, and Postman collection
• Redis is used for all counters, cooldowns, and pending notification batching
• PostgreSQL remains the source of truth for persisted content
• The application is stateless with no in-memory production counters or static state