Kannaka Memory

About

Kannaka provides AI agents with memory that resonates rather than relying on indexes. Each memory exists as a wavefront within a 10,000 dimensional tensor field, and recall is accomplished through a single tensor product without a vector index. The system forgets via destructive interference, consolidates memories during dream cycles, establishes beliefs as stable phase singularities, and synchronizes across agents using NATS. It is built in Rust, licensed under MIT, and distributed as a single static binary.