Resonance
A peer-to-peer semantic-matching layer for on-device AI agents that routes content by meaning instead of through a server-owned index, built for an on-device AI hackathon.
Resonance turns a piece of text into a semantic vector on the device and gossips it peer-to-peer over a Hyperswarm network, with no central server. Other devices score the incoming vector against their own content, so a post reaches phones whose owners, or their on-device AI agents drafting a reply for approval, have written about similar things; there is no follow graph and no ranking algorithm deciding who sees what.
The repository demonstrates this with a social-posting example built on the library, where a phone's feed fills with posts and replies found this way, without the poster ever having followed the people who see it. The project frames this as a discovery layer for an agentic era, where people and their on-device AI agents find each other by meaning rather than through an index or follow graph someone else owns.
It targets React Native and is documented with a front-to-back guide covering its peer-to-peer networking layer and its on-device AI-agent layer.