Rohit-Dnath/RAMen — reverse-engineered prompt

Build me a small, fast in memory data store that feels like Redis for normal cache use, but is clearly aimed at AI apps and agents.

I want it to accept the usual Redis style commands so an existing Redis client can point to it on port 6379 and mostly just work for common things like strings, hashes, lists, sets, sorted sets, expiry, and pub sub. On top of that, add native vector storage and nearest match search, a semantic cache that can return a saved AI answer when a new prompt means basically the same thing, and simple agent memory commands for saving and recalling facts by session.

Please also include a built in MCP mode so tools like Claude can use it directly, plus a small live dashboard in the browser that shows keys, memory use, and cache hit rate. Keep it as a single Go binary, easy to run locally or in Docker, with snapshot saves to disk, and be honest that this is for one machine, not clustering or failover. Look up current docs online if you need to.