Skip to main content

// Section 14.3 · Operate

Install and Bootstrap

1 min14.3Operate
// 2 of 11 · node cli

One line. Node, Ollama, model, CLI.

// 14.3 · install and bootstrap · macos + linux · windows via wsl2

Key claim

The node key never leaves the machine: a separate, low-value secp256k1 key that signs liveness and results only, never the staking wallet's key.

Distribution is a one-line install script at https://parallelix.io/install.sh. On a fresh Linux GPU box it goes from nothing to ready to register in one command: Node.js, the Ollama inference runtime (GPU auto-detected), the default model llama3.2, and the parallelix-node CLI itself. macOS is supported the same way; native Windows is not, so Windows operators run everything inside WSL2. All state lives under ~/.parallelix/.

What the one-liner installs

// node · ollama · model · cli · init

CURL THE ONE-LINER

Run `curl -fsSL https://parallelix.io/install.sh | sh`. On a fresh Linux GPU box this bootstraps the whole operator stack in one command.

NODE.JS

Installs Node.js LTS if it is missing or older than 18. The CLI is a small Node program; this is its only language runtime.

OLLAMA

Installs Ollama via its official installer (which auto-detects the GPU) and starts it. Ollama is the open-source LLM runtime that actually executes inference.

DEFAULT MODEL

Pulls the default model llama3.2 (~2 GB) so the node can serve from first start. Override with PARALLELIX_MODEL, or skip with PARALLELIX_SKIP_MODEL=1.

PARALLELIX-NODE

Downloads parallelix-node from https://parallelix.io/cli/parallelix-node.mjs, installs its crypto deps, and places a shim at /usr/local/bin/parallelix-node.

INIT NODE KEY

Operator runs `parallelix-node init`. A separate, low-value secp256k1 node key (never the staking wallet's key) is generated at ~/.parallelix/node.key (mode 0600), and config is written to ~/.parallelix/config.json. It prints the nodeKeyHash to register on-chain in the Console.

Sample install session

// bash · two commands · ready to stake in the Console and attach

// install + init · bash session
# Install · Node + Ollama + llama3.2 + parallelix-node, in one line
$ curl -fsSL https://parallelix.io/install.sh | sh
✓ Node v22.x present
✓ Ollama running
✓ model llama3.2 ready
✓ parallelix-node installed

# Initialise · generates a node key, prints the nodeKeyHash
$ parallelix-node init
✓ node key generated at ~/.parallelix/node.key (mode 0600)
✓ config written to ~/.parallelix/config.json
// register this node on-chain (in the Console), using:
   tier         3 (GPU · high)
   nodeKeyHash  0x9c27a4...

Supported targets

// macos + linux live · windows via wsl2 only

SUPPORTED

Linux

The one-liner bootstraps Node, Ollama, the model, and the CLI on a fresh box. NVIDIA, AMD (ROCm), and CPU fallback are detected.

SUPPORTED

macOS (Apple Silicon + Intel)

Metal GPU via Ollama on Apple Silicon; unified memory caps the tier at high. Intel runs CPU mode unless a supported GPU is present.

WSL2 ONLY

Windows

Native Windows is not supported. Install WSL2, open Ubuntu, and run the one-liner and every command inside it.