esengine/HiveMind ? reverse-engineered prompt

Build me a Python project that lets each user fine tune their own personal AI assistant locally, then send only the trained adapter up to a server so everyone can benefit from the shared improvements without sharing raw data.

I want a simple command line flow for creating sample training data, training a local model with LoRA, chatting with the personalized model, starting a small web server, and uploading or downloading adapters. The server should store adapters, show basic status, list uploads, let me trigger aggregation, and serve the latest combined version. Use a Qwen based model by default, and make it practical for GPU training, with a lighter setup if memory is limited.

Please include clear defaults, helpful error messages, and a few tests. If you need current library details, look them up online and keep the setup easy to run from the README.