E-commerce microservices dùng Django/DRF + Docker Compose, bán LAPTOP và CLOTHES.

Các Django projects chính:

• api_gateway (Django templates)
  • Frontend HTML + routing/proxy gọi sang các services.
• customer_service (MySQL)
  • Đăng ký/đăng nhập khách hàng, profile, token auth.
• staff_service (MySQL)
  • Đăng nhập staff + dashboard gọi proxy sang product/cart.
• product_service (PostgreSQL)
  • Catalog thống nhất cho LAPTOP + CLOTHES.
  • Search kết hợp FTS + substring.
• cart_service (PostgreSQL)
  • Cart + checkout + orders.
• interaction_service (PostgreSQL + Neo4j)
  • Ghi event hành vi (view/search/cart/purchase/chat).
  • Sync dữ liệu sang Neo4j KG bằng management command.
• ai_chat_service (FAISS + Gemini + Neo4j)
  • Chatbot RAG (KB markdown) + cá nhân hoá theo segment/hành vi.
  • Vector store FAISS local trong service (không dùng ChromaDB).
  • Embeddings: paraphrase-multilingual-MiniLM-L12-v2 (Sentence-Transformers).

Hạ tầng đi kèm trong docker-compose.yml:

• MySQL: customer_mysql, staff_mysql
• PostgreSQL: product_postgres, cart_postgres, interaction_postgres
• Neo4j: neo4j

• Mỗi service có Dockerfile/requirements riêng.
• DB độc lập; không có cross-service foreign keys (chỉ chia sẻ IDs qua REST).
• Inter-service communication: REST qua requests.
• SQL init scripts tự chạy khi DB containers start (tạo table + sample data).

• API Gateway: http://localhost:8000
• Customer service: http://localhost:8001
• Product service: http://localhost:8002
• Cart service: http://localhost:8004
• Staff service: http://localhost:8005
• AI Chat service: http://localhost:8006
• Interaction service: http://localhost:8007

• Neo4j Browser: http://localhost:7475
• Neo4j Bolt: bolt://localhost:7688

MySQL (container 3306, host map port tuỳ chỉnh):

• Customer MySQL: localhost:3307
• Staff MySQL: localhost:3308

PostgreSQL (container 5432, host map port tuỳ chỉnh):

• Cart Postgres: localhost:5435
• Product Postgres: localhost:5436
• Interaction Postgres: localhost:5437

Prerequisites:

• Docker Desktop (Windows)
• Docker Compose v2 (docker compose)

Tạo file .env ở repo root (để gọi Gemini; không có key vẫn chạy nhưng sẽ fallback rule-based):

GEMINI_API_KEY=your_gemini_api_key_here

Chạy từ repo root:

docker compose up --build

Mở UI:

• http://localhost:8000

Stop:

docker compose down

Reset dữ liệu (drop volumes):

docker compose down -v

• Username: admin
• Password: password123

(Seed bởi staff_service/sql/init.sql.)

Đăng ký tại:

• /customer/register/

• username: demo.customer@example.com

• password: 123456

• Home: /
• Product listing (search + pagination):
  • Laptops: /products/laptops/?search=mac&page=1
  • Clothes: /products/clothes/?search=polo&page=1
• Product detail: /products/<laptops|clothes>/<id>/
• Customer auth: /customer/register/, /customer/login/, /customer/logout/
• Cart: /cart/, /cart/checkout/
• Staff: /staff/login/, /staff/dashboard/

• GET /api/v1/chat/health/
• POST /api/v1/chat/message/ JSON: { "user_id", "message", "context" }

Gateway proxy:

• POST /api/chat/message/

• GET /api/health/
• GET/POST /api/events/ (log hành vi; dùng cho demo + sync Neo4j)

• GET /api/v1/products/?product_type=LAPTOP|CLOTHES&q=<term>&page=1
• CRUD: /api/v1/products/ và /api/v1/products/<id>/

• Cart: GET /api/cart/, POST /api/cart/items/, PATCH/DELETE /api/cart/items/<item_id>/
• Checkout: POST /api/checkout/
• Orders: GET /api/orders/

• POST /api/register/, POST /api/login/, GET /api/profile/

• POST /api/login/, GET /api/profile/
• Proxy endpoints (require staff token):
  • /api/proxy/products/... → product_service
  • /api/proxy/orders/... → cart_service

• KB markdown: ai_chat_service/knowledge_base/
• FAISS index lưu local (bind mount):
  • /app/vector_store/kb.index
  • /app/vector_store/kb_meta.json
  • /app/vector_store/embedder_meta.json

Re-index thủ công (trong container ai_chat_service):

python manage.py init_kb --force

Lưu ý: lần đầu service chạy sẽ tải model embeddings từ HuggingFace (cần internet).

Init scripts:

• customer_service/sql/init.sql
• staff_service/sql/init.sql
• cart_service/sql/init.sql
• interaction_service/sql/init.sql

• Muốn chạy lại SQL init scripts: docker compose down -v rồi up lại.
• Nếu port bị trùng: sửa mapping trong docker-compose.yml.