Vistara's permissionless network allows anyone to register their hardware for public use. The network guarantees security, isolated execution, and scalability through a unit called a spacecore, which operates in a secure, virtualized environment. The instructions for distributing hardware resources are outlined in the hardware as code (HAC) file.
hal.toml
[spacecore]
name = "AI Inference Node"
description = "AI inference node."
[hardware] # resource allocation
cpu_cores = 4 # Number of CPU cores
memory_gb = 16 # Memory in GB
storage_gb = 500 # Storage in GB
bandwidth_mbps = 100 # Network bandwidth in Mbps
pcie_devices = [ "NVIDIA GRID vGPU" ] # GPU
[environment]
os = "linux"
runtime = "microvm"
image = "decentralized/llm-node:latest"
[network]
public_ip = true
p2p_port = 30303
api_port = 8080
