IntellectLight

A Spring Boot authentication and knowledge management service with RAG (Retrieval-Augmented Generation) capabilities, social networking features, and comprehensive content management.

Tech Stack

Component	Technology
Runtime	Java 21, Spring Boot 3.2.4
Database	MySQL 8.0+ (MyBatis)
Cache	Redis + Caffeine (two-tier)
Search	Elasticsearch 9.2.1
AI/RAG	Spring AI (OpenAI, DeepSeek)
Auth	JWT (RSA 2048-bit) + Spring Security
Messaging	Kafka + Canal CDC
Storage	Aliyun OSS
Monitoring	Spring Actuator

Features

• Authentication — JWT-based auth with access/refresh tokens, verification codes (SMS/email), password reset, and login audit logging
• Knowledge Posts — Full lifecycle management (draft → pending → published) with OSS-backed content storage and integrity checks (ETag, SHA-256)
• RAG Q&A — AI-powered question answering over post content via Elasticsearch vector store and streaming SSE responses
• Social Graph — Follow/unfollow with bidirectional relationship tracking, paginated follower/following lists
• Search — Full-text search with Elasticsearch and autocomplete suggestions via completion suggester
• Counters — Compact SDS-format counters for likes, favorites, followers, following, and post counts with Kafka-driven aggregation
• Hot Key Detection — Automatic cache TTL extension based on access frequency with configurable heat levels
• CDC Sync — Canal captures MySQL binlog changes via the transactional outbox pattern, publishing to Kafka for Elasticsearch indexing

Project Structure

src/main/java/com/tongji/
├── auth/           # Authentication, JWT, verification codes
├── cache/          # Two-tier caching (Redis + Caffeine), hot key detection
├── common/         # Error codes, exceptions, global handler, utilities
├── config/         # Elasticsearch, Redisson, thread pool configuration
├── counter/        # SDS-format counters, Kafka aggregation
├── knowpost/       # Knowledge post CRUD, RAG, AI description
├── profile/        # User profile management
├── relation/       # Social graph (follow/follower)
├── search/         # Elasticsearch search, Canal CDC consumers
├── storage/        # Aliyun OSS presigned URL generation
└── user/           # User service

API Endpoints

Authentication — /api/v1/auth

Method	Path	Description
POST	/send-code	Send verification code
POST	/register	Register new user
POST	/login	Login (password or code)
POST	/token/refresh	Refresh token pair
POST	/logout	Logout and revoke token
POST	/password/reset	Reset password
GET	/me	Get current user info

Knowledge Posts — /api/v1/knowposts

Method	Path	Description
POST	/drafts	Create new draft
POST	/{id}/content/confirm	Confirm content upload
PATCH	/{id}	Update metadata
POST	/{id}/publish	Publish post
PATCH	/{id}/top	Pin/unpin post
PATCH	/{id}/visibility	Update visibility
DELETE	/{id}	Soft delete
GET	/feed	Public feed (paginated)
GET	/mine	User's published posts
GET	/detail/{id}	Post details
GET	/{id}/qa/stream	RAG Q&A (SSE streaming)
POST	/{id}/rag/reindex	Trigger RAG reindex
POST	/description/suggest	AI-generated description

Search — /api/v1/search

Method	Path	Description
GET	/	Keyword search with pagination
GET	/suggest	Autocomplete suggestions

Relations — /api/v1/relation

Method	Path	Description
POST	/follow	Follow user
POST	/unfollow	Unfollow user
GET	/status	Relation status query
GET	/following	Following list (paginated)
GET	/followers	Followers list (paginated)
GET	/counter	User dimension counters

Actions & Counters — /api/v1/action, /api/v1/counter

Method	Path	Description
POST	/like, /unlike	Like/unlike entity
POST	/favorite, /unfavorite	Favorite/unfavorite entity
GET	/{etype}/{eid}	Get aggregated counts

Storage — /api/v1/storage

Method	Path	Description
POST	/presign	Get presigned upload URL

Prerequisites

• Java 21
• Maven 3+
• MySQL 8.0+
• Redis
• Elasticsearch 9.2.1
• Kafka
• Canal (for CDC)
• Aliyun OSS account

Getting Started

1. Set up the database

   mysql -u root -p < db/schema.sql

2. Configure the application

   Create or update src/main/resources/application.yml with your environment-specific settings for MySQL, Redis, Elasticsearch, Kafka, Canal, and Aliyun OSS.

3. Build

   mvn clean package

4. Run

   java -jar target/IntellectLight-1.0-SNAPSHOT.jar

   The application starts on the default Spring Boot port. Health check is available at /actuator/health.

Testing

mvn test

Architecture Highlights

• Two-tier caching — Caffeine (local, ~10-30s TTL) backed by Redis (distributed) with automatic hot key TTL extension
• Transactional outbox — Guarantees at-least-once event delivery from MySQL to Kafka via Canal CDC
• SDS counters — Compact 20-byte big-endian format packing 5 counter dimensions into a single Redis value
• Distributed locking — Redisson for coordinating counter rebuilds and concurrent operations
• Stateless auth — RSA-signed JWT tokens with no server-side session storage