Axon Cluster

A private local network for delegating AI inference from lower-power devices to a stronger machine without exposing the cluster as a public service.

Rust · libp2p · Ollama · Local AI

Question

Can low-power machines securely delegate local AI inference to a stronger machine without exposing the cluster as a public service?

Method

Use a private libp2p network with a pre-shared key, Noise-encrypted transport, mDNS discovery, and a leader/subordinate request protocol backed by Ollama.

Evidence

The Rust implementation includes leader and subordinate modes, protocol framing, Ollama integration, an HTTP bridge, a web interface, and a recorded demo.

Current limit

The current design targets trusted devices on one local network; it does not claim Byzantine fault tolerance, public discovery, or multi-leader scheduling.

Next step

Test failure and reconnect behavior across multiple local devices before expanding the trust or scheduling model.