// Section 14.11 · Operate

Sample Sessions, End to End

1 min14.11Operate

// 6 of 11 · node cli

Sample sessions. End to end.

// 14.6 · six sessions · v2.0.0 · output shape is illustrative

Caveat·Output shape disclaimer

Command output shown below illustrates the shape of each interaction. Exact formatting may differ across platforms and releases. The command names and flags are canonical for v2.0.0; do not script against the output format.

1 SETUP 2 START · DASHBOARD 3 START · DAEMON 4 SYSTEMD UNIT 5 STATUS 6 DRAIN + DETACH

Six sessions

// bash + ini · each session is atomic

// 14.6.1 · session

Setup (fresh machine)

Run parallelix-node setup on a freshly installed machine. Five steps: hardware detect, Ollama check, model pull, node key generation, Console staking link. After the operator stakes in the Console, setup auto-detects the nodeId and saves it to config.

// session 1 · setup · v2.0.0 · bash

$ parallelix-node setup
[1/5] hardware   NVIDIA RTX 4090 · 24 GB VRAM · tier 3 (GPU · high)
[2/5] ollama     running
[3/5] model      recommend: qwen2.5:14b (11 GB, fits 90% VRAM)
      pulling qwen2.5:14b ...                             ✓
[4/5] node key   generated at ~/.parallelix/node.key (0600)
                 nodeKeyHash  0x9c27a4f1...e83b
[5/5] stake now  https://app.parallelix.io/operate/stake?tier=3&keyHash=0x9c27a4...
      waiting for on-chain registration ...
      detected nodeId 42 · saved to config
      install as service? [y/N]: y
      ✓ installed + started as systemd service 'parallelix-node'

Re-running setup is safe: it resumes from where it left off. The manual path (probe, then init, then start --node-id) still works as an alternative.

// 14.6.2 · session

Start (interactive dashboard)

Start the daemon in the foreground to see the live dashboard. Status, installed models, lifetime earnings, and a per-request feed with SHA-256 PoE ticks stream to the terminal. Use --plain to force line output under systemd or pipes.

// session 2 · start · v2.0.0 · interactive dashboard

$ parallelix-node start
// parallelix-node v2.0.0  · node #42
● status      active · heartbeats 142 · served 37
  models      qwen2.5:14b  llama3.2
  earnings    312.4 PRLX (lifetime)
  req feed    req_a3f1  sha256:4f9c…  47 ms  · ai-inference
              req_a3f5  sha256:8b22…  91 ms  · ai-inference
              heartbeat ok · seq 143

Reward accrues on-chain against the nodeId; the operator claims it in the Console with claim(42). The daemon shows the lifetime earnings for convenience but never holds or moves funds.

// 14.6.3 · session

Start (daemon mode)

Run the node 24/7. The service command installs a systemd unit (Linux) or launchd agent (macOS) that survives reboot and auto-restarts on failure. Setup offers this; you can also run it manually.

// session 3 · service · run 24/7

$ parallelix-node service --gpu
✓ installed + started as systemd service 'parallelix-node'
· logs: journalctl -u parallelix-node -f
· stop: sudo systemctl stop parallelix-node

// 14.6.4 · session

The systemd unit it writes

parallelix-node service writes this unit for you and enables it. Shown here so you know what is running; you do not write it by hand.

// session 4 · /etc/systemd/system/parallelix-node.service · ini

[Unit]
Description=ParalleliX Node
After=network-online.target ollama.service
Wants=network-online.target

[Service]
Type=simple
ExecStart=/usr/local/bin/parallelix-node start --gpu
Restart=on-failure
RestartSec=5

[Install]
WantedBy=multi-user.target

// 14.6.5 · session

Status (snapshot)

Point-in-time read of node state. Useful for cron checks or external monitoring.

// session 5 · status · bash

$ parallelix-node status
node_id           42
state             EXECUTING
in-flight         2 requests
accrued (24h)     312.4 PRLX  (claim in Console)
uptime (30d)      99.8 %
reputation        n/a (at launch)
last error        none
coord latency     43 ms

// 14.6.6 · session

Draining and detaching

Console action

Unstaking is an on-chain action in the Console. When the operator calls requestUnstake(42) there, the daemon receives an in_cooldown signal from the coordinator and exits cleanly; after the 7-day cooldown the operator calls withdrawStake(42) in the Console.

// session 6 · drain → detach · bash

# operator called requestUnstake(42) in the Console
$ parallelix-node status
node_id           42
state             DRAINING
in-flight         1 request    (no new dispatch; finishing current work)
cooldown ends     2026-06-05 14:03 UTC
note              withdraw in the Console after cooldown

# after withdrawStake(42) in the Console, the daemon detaches
$ parallelix-node status
node_id           42
state             DETACHED
note              stake returned in full (no slashing); re-stake to rejoin

The principal is always returned in full: there is no slashing and no deductions. To rejoin, stake again in the Console and attach the machine.

// Reading path

Node CLI · parent hub Operator onboarding sequence end to end Staking with no slashing