BASILISK / artifact virtual
artifact-opensource on GitHub

A loop, not a pipeline.

Basilisk

A model trains against Hektor's vector store. Artifact Engine serves it as GGUF. Symbiote turns that inference into an agent that can act. Edge meshes every machine willing to run one. And Spore puts an agent on each of them — every idle device a host, every host tasked with one market or one slice of the world. Enough of them running at once and the swarm starts sensing and deciding on its own, feeding what it learns back into Hektor. The system doesn't need to be operated. It runs itself.

01Hektor Vector StudioTraining memory02Artifact EngineGGUF inferencePRIVATE03SymbioteAgentic runtime04Artifact EdgeMesh network05SporeSwarm host
training memory in, swarm signal back — a closed loopclick a plate to open it