Noemaai-labs Charts Noema-atlas To Connect And Swap AI Files Safely
Noema-atlas is a new release that provides a peer-to-peer network for sharing local language model weight files directly between machines. The system downloads models from peers around the world while using Hugging Face and mirrors as fallbacks. Every byte transferred is checked against a cryptographic fingerprint to ensure the files are safe and untampered.
Developer Noemaai-labs created this tool to address frustrations with relying on a single central source for downloading models. They built the software using Rust to bring peers together without using traditional trackers. The team designed it to verify content and let users control exactly what gets broadcast.
Peer-to-peer model sharing
- Serves models straight to other machines.
- Stores identical model files only once.
- Checks every byte against content hash.
- Rescues models taken down from hubs.
- Runs native and lightweight on desktop.
This software is for people who want more control over how their local language models are stored and distributed. Users can rescue and reshare weight files that disappeared from public hosting sites. Those running scripts or working over remote connections can use the included command line interface to manage their downloads.
Development and security notes
The developers note that catching a single poisoned chunk mid-download before the rest arrives is on the roadmap. Pickle-class and executable file types are refused by default to keep the system secure. Both desktop applications are still a work in progress and the team actively welcomes public collaboration to fix bugs.
"Noema Atlas is a peer-to-peer network software using Iroh for local LLM weights, free and open source (Apache-2.0)." Source: Reddit