【Shopee】Insight Hub

InsightHub is an internal tool for automatically extracting knowledge from Java projects and building an AI-assisted project knowledge center.

It integrates:

Spring AI + Ollama + LLM for understanding and explaining code.
MyBatis + MySQL for task tracking and persistent storage.
Milvus for storing code embeddings and enabling semantic search.
Redis for caching.

✨ Features

Automatic structure extraction (classes, methods, relationships)
Method-level comment generation via LLM
Embedding generation + Milvus vector storage
Project-level semantic search and intelligent Q&A

🌍 Docker Compose

This project includes a preconfigured Docker Compose setup for local development.

♻ How to start services

docker-compose up -d

This will start:

MySQL (on port 3306)
Redis (on port 6379)
Milvus (on port 19530)
Nebula Graph (optional)
Ollama (on port 11434)

🎧 Pull LLM model for Ollama (e.g. codellama)

curl http://localhost:11434/api/pull -d '{"name": "codellama:7b"}'

⚖️ Configuration

application-local.yaml

Make sure this file exists in src/main/resources/:

spring:
  datasource:
    url: jdbc:mysql://localhost:3306/insighthub
    username: insighthub_user
    password: insighthub_pass
    driver-class-name: com.mysql.cj.jdbc.Driver

  data:
    redis:
      host: localhost
      port: 6379

  ai:
    openai:
      base-url: http://localhost:11434
      api-key: dummy
      chat:
        model: codellama:7b
      embedding:
        model: llama-embedding

milvus:
  host: localhost
  port: 19530
  database: default
  collection: insighthub_vectors

Activating the profile

In application.yaml:

spring:
  profiles:
    active: local

Or via command line:

./mvnw spring-boot:run -Dspring-boot.run.profiles=local

⚙️ Local Dev Workflow

Start all services via Docker Compose:

docker-compose up -d

Start the Spring Boot app:

./mvnw spring-boot:run -Dspring-boot.run.profiles=local

Test endpoint:

curl -X POST http://localhost:8080/api/test/ask \
     -H "Content-Type: text/plain" \
     -d "你是谁"

You should get a natural language answer from the local LLM via Ollama.

💡 Notes

You can configure different LLMs via Ollama (e.g. llama2, mistral, llama3)
Milvus is used to store embeddings for later retrieval (e.g. RAG)
The project is designed to support agent-based architecture