FutureUnreal/What-to-eat-today

Implements hybrid retrieval combining graph-based entity relationships with vector similarity search across a Neo4j/Milvus stack, enabling multi-step reasoning for recipe recommendations. The system processes natural language queries through a Flask backend to retrieve relevant cooking instructions and ingredient relationships, then generates personalized guidance via LLM integration. Containerized deployment with Docker Compose provides one-command setup across Windows/Linux/macOS with Nginx reverse proxy, frontend React/Next.js interface, and persistent data stores.

126 stars.