Mymemorysys

Create a sophisticated, containerized semantic memory system for Claude Code. The system needs to use FastAPI as the backend, connecting to a PostgreSQL database with the pgvector extension. Key features include automatically retrieving and incrementally injecting relevant historical memories into the context. It must support real-time saving and management of memories, handling embedding generation using the jina-embeddings-v5-text-small-retrieval-mlx model running on MLX. Implement advanced retrieval logic, such as composite embedding creation (tags + summary + content), multi-factor scoring (similarity, category, recency, importance), and automatic deduplication. The system should also manage memory expiration and support various memory categories like 'identity' and 'code'. Finally, provide deployment scripts to handle Docker Compose setup and hook installation.