Picture-management-website

Zhejiang University 2025 B/S Course Project

AI Service Configuration

• The backend expects the AI microservice to be reachable via the AI_SERVICE_URL environment variable (defaults to http://ai-service:5000, which is wired up when running with Docker Compose).
• When running the Spring Boot backend locally, export AI_SERVICE_URL to point to your running ai-service instance, e.g. http://localhost:5000, to enable the new AI-assisted endpoints.
• The AI microservice now combines a CLIP-based zero-shot vision model with heuristics to produce multi-type tags (scene, subject, mood, etc.). Tune it via ENABLE_VISION_MODEL, VISION_MODEL_ID, and related env vars inside ai-service.

Conversational MCP Access

• The tools/mcp-image-search folder contains a standalone Model Context Protocol server that exposes the photo search endpoints to LLM clients (e.g., Claude Desktop).
• Configure it with PICTURE_API_BASE_URL, PICTURE_AI_BASE_URL, and PICTURE_API_TOKEN, then register the search_images tool inside your MCP-compatible client to let models retrieve photos through natural dialogue.

Documentation

• See docs/README.md for full backend/frontend/AI-service documentation, setup, and run instructions.