Skip to main content

Cross-Platform Install

Install teale-node on Linux, Windows, or Android. teale-node is a single Rust binary (6.2 MB) that brings full Teale network participation to non-Apple platforms.


Supported hardware

teale-node supports a wide range of GPU and CPU backends:

BackendPlatformsNotes
NVIDIA (CUDA)Linux, WindowsRequires CUDA 12.0+ and driver 525+
AMD (ROCm)LinuxRequires ROCm 5.7+
Intel (SYCL)Linux, WindowsRequires oneAPI runtime
VulkanLinux, Windows, AndroidBroad GPU support via Vulkan 1.2+
CPU-onlyAll platformsWorks everywhere, slower inference

Install

Cargo (from source)

cargo install teale-node

Binary download

Download the latest binary for your platform from GitHub Releases:

  • Linux x86_64 (CUDA, ROCm, Vulkan, CPU)
  • Linux aarch64 (Vulkan, CPU)
  • Windows x86_64 (CUDA, Vulkan, CPU)
  • Android aarch64 (Vulkan, CPU)

Docker

CPU variant:

docker run -d --name teale teale/node

CUDA variant (requires NVIDIA Container Toolkit):

docker run -d --gpus all --name teale teale/node:cuda

Configuration

On first run, teale-node generates an Ed25519 identity keypair and writes a default configuration file. The config lives at ~/.teale/config.toml (or %APPDATA%\Teale\config.toml on Windows).

teale-node uses the same relay protocol as the macOS app, so cross-platform nodes participate in the same network as Mac and iPhone users.

# Start the node
teale-node up

# Check status
teale-node status

# View or edit configuration
teale-node config show

Key configuration options

[network]
relay = "auto" # Auto-discover relays, or set a specific relay address
max_connections = 32

[inference]
backend = "auto" # auto, cuda, rocm, sycl, vulkan, cpu
max_memory_gb = 8 # Maximum VRAM/RAM to use for models

[identity]
# Auto-generated on first run. Do not edit.
# keypair = "~/.teale/identity.key"

Verify

teale-node status

You should see the node status, detected GPU backend, and network connectivity.

Uninstall

cargo uninstall teale-node
rm -rf ~/.teale

Or remove the binary and data directory manually.


Next steps