feat：Pr/add docker compose

[hikariming]

This PR adds Docker Compose configuration to provide a one-click startup experience similar to Dify, significantly reducing the environment setup complexity.

Motivation

UltraRAG requires configuring a vector database (Milvus) and object storage (MinIO) to function properly. Setting up these dependencies manually can be tedious and error-prone for new users. This PR aims to provide an out-of-the-box Docker Compose solution that allows users to get started quickly, while still allowing them to customize or opt-out of any component as needed.

Changes

Added Files

• docker/docker-compose.yml (GPU setup): UI + Milvus + vLLM (LLM + Embedding)
• docker/docker-compose.cpu.yml (CPU-only): UI + Milvus (without vLLM)
• docker/.env.example: Environment variables template for GPU setup
• docker/.env.cpu.example: Environment variables template for CPU setup
• docker/config/kb_config.json: Pre-configured knowledge base settings (Milvus URI)
• docker/README.md: Comprehensive setup and usage documentation

Modified Files

• ui/backend/app.py: Added support for default API configuration via environment variables
• ui/frontend/main.js: Added UI pre-fill for default LLM/Embedding endpoints
• .gitignore: Added docker/volumes/ to exclude persistent data from version control

Usage

Quick Start (GPU):
cd docker
cp .env.example .env

Edit .env to set model paths

docker compose up -d --build

Quick Start (CPU):
cd docker
cp .env.cpu.example .env
docker compose -f docker-compose.cpu.yml up -d --build

Then open http://localhost:5050

Design Philosophy

• Optional: Users can still choose to run services separately or use external API endpoints
• Flexible: CPU mode allows using hosted LLM/Embedding APIs instead of local vLLM
• Isolated: All persistent data is stored under docker/volumes/ (gitignored)

See docker/README.md for detailed documentation.