carrasquelalex1/hipocampo ? reverse-engineered prompt

Build me a Python based memory server for AI agents that can save and recall useful things across sessions.

I want it to store both technical notes and user profile style memories in PostgreSQL, use vector search for semantic recall, and also do fast keyword style searching so results feel smart and reliable. It should expose the memory features through an MCP server so any compatible agent can connect to it, save memories, search them, and link related items together. I also want a simple browser playground or test page so I can try saving and searching memories without needing a client.

Please make it feel production ready, with a clean local setup, Docker support, an example environment file, and clear docs for running it. If you need to check current MCP or embedding docs online, go ahead. Keep the design focused on long term memory, retrieval quality, and easy use for other agents.