MANHGG Room Manager

Full-stack room listing and rental consulting app with AI-assisted listing generation and a RAG chatbot.

Technology

• Frontend: React + Vite + Tailwind CSS
• Backend web API: Java 21 + Spring Boot
• AI service: Python + FastAPI
• Database: MongoDB
• Vector database: Qdrant
• Media storage: Cloudinary, with legacy GridFS/filesystem migration support
• LLM integration: Gemini for listing generation and optional RAG chat generation
• Deployment: Docker Compose
• Auth: token-based user/admin login

AI Features

• AI listing generator for Facebook/Zalo room posts.
• Floating AI chatbox for 24/7 room consulting.
• RAG pipeline over public room fields.
• Sensitive internal fields such as real address and internal note are excluded from the AI index.

See:

• AI architecture
• RAG pipeline
• Model evaluation notes

Run for Development

Create a local .env from .env.example, then run:

docker compose -f docker-compose.dev.yml up

Open:

• Frontend: http://localhost:5173
• Backend API: http://localhost:8080
• AI service: http://localhost:8000/health
• Qdrant: http://localhost:6333/dashboard

After rooms already exist in MongoDB, run the admin-only endpoint once to rebuild the vector index:

POST /api/ai/reindex
Authorization: Bearer <token>

New or updated rooms are indexed automatically after save.

Run Production-Like Docker

docker compose up --build

Open http://localhost:5173.

AI Configuration

The project runs without an LLM key by using local hashing embeddings and a local chat fallback:

EMBEDDING_PROVIDER=local_hash

To use Gemini for chat or embeddings:

GEMINI_API_KEY=your_key
GEMINI_CHAT_MODEL=gemini-2.5-flash
EMBEDDING_PROVIDER=gemini
GEMINI_EMBEDDING_MODEL=text-embedding-004

If you change embedding provider or dimension, recreate the Qdrant collection or reset the Docker qdrant-data volume.

Source Structure

backend/      Spring Boot API, auth, rooms, media, AI proxy endpoints
frontend/     React room manager, AI chatbox
ai-service/   Python FastAPI service for embeddings, Qdrant retrieval, RAG chat
docs/         AI architecture and RAG pipeline notes