Santhoshnadella/MESHTRAIN ? reverse-engineered prompt
Reverse engineered prompt
Build me a working prototype of MeshTrain, a decentralized AI compute network that lets people send inference and training jobs to remote peers instead of using a central server.
I want the app to let a user start a node, discover other peers, send a text prompt or image generation request, and get the result back from a remote machine. It should also support sending a small dataset to a worker for LoRA style fine tuning, then return the trained adapter. If possible, include a simple way to split data across multiple peers for swarm training, and a basic local ledger so workers can earn and track credits.
Make the whole thing feel secure and practical, with encrypted peer to peer communication, model verification, and a sandboxed execution flow so unsafe code does not run on the main process. A simple desktop or local web interface would be great for starting nodes, viewing jobs, and seeing network status. Look up current docs online if you need to.