moorcheh-ai/memanto — reverse-engineered prompt

Build me a Python tool called Memanto that gives AI agents a real long term memory instead of just stuffing old notes into the prompt.

I want a simple CLI where I can create an agent, save memories, recall relevant memories, and ask questions that get answered using the saved memory. The main commands should feel like remember, recall, answer, status, serve, ui, upload, and agent management. Memories should support types like fact, preference, goal, decision, instruction, relationship, event, error, and similar categories, with metadata like confidence, source, timestamps, and version history so old or conflicting information is handled clearly.

It should use a Moorcheh API key for the backend retrieval service, read config from environment variables, and work without needing another LLM key for normal memory answers. Please include examples, tests, Docker setup, and a small optional local REST API plus browser UI for trying it out. Look up current docs online if you need to.