Skip to main content

Deploy to Device

The SDK ships with an onboarding script that copies SDK modules to a fresh Dev Edition, installs system dependencies, and runs a smoke test in one shot. There's also a manual path if you prefer to script your own deployment.

python onboarding/onboard_device.py <host> <user> --password <pw>

What it does, in order:

  1. SSH into <host> as <user>.
  2. Create ~/sdk/audio/ and ~/sdk/ssh/.
  3. SFTP-copy audio/audio_sdk.py and ssh/ssh_sdk.py.
  4. sudo apt-get install -y python3-pyaudio (idempotent).
  5. sudo usermod -a -G audio system so the SDK can open the codec without sudo.
  6. Register any bundled OpenClaw skills.
  7. Run health_check(loopback=True) and report.

The script is idempotent — running it twice is safe and will only re-do work that's actually missing.

Arguments​

onboard_device.py <host> <user> [options]

--password TEXT    SSH password (mutually exclusive with --key)
--key      PATH    SSH private key path
--port     INT     SSH port (default 22)
--no-audio         Skip audio SDK install (only deploy ssh_sdk)
--no-ssh           Skip ssh SDK install (only deploy audio_sdk)
--dry-run          Print what would happen, do nothing

Manual path​

If you'd rather script the deploy yourself:

HOST=172.168.20.145
USER=system

ssh "$USER@$HOST" "mkdir -p ~/sdk/audio ~/sdk/ssh"

scp audio/audio_sdk.py "$USER@$HOST:~/sdk/audio/"
scp ssh/ssh_sdk.py     "$USER@$HOST:~/sdk/ssh/"

ssh "$USER@$HOST" "
  sudo apt-get update &&
  sudo apt-get install -y python3-pyaudio &&
  sudo usermod -a -G audio system
"

After usermod, the new group only takes effect on next login. Either log out and back in, or use newgrp audio to apply it to the current shell.

Using the SDK from Python on the device​

Once deployed, you can import the modules:

ssh system@172.168.20.145
python3
>>> import sys
>>> sys.path.insert(0, "/home/system/sdk")
>>> from audio.audio_sdk import AudioSDK
>>> AudioSDK().health_check()

To make the import path permanent, append to ~/.bashrc:

echo 'export PYTHONPATH="$HOME/sdk:$PYTHONPATH"' >> ~/.bashrc

Deploying your own code alongside the SDK​

The same SSH SDK from your workstation works for shipping your own scripts:

from ssh_sdk import InternSSH

with InternSSH("172.168.20.145", "system", key_path="~/.ssh/id_ed25519") as ssh:
    ssh.put_dir("./my_project", "/home/system/my_project")
    result = ssh.run("cd /home/system/my_project && python3 main.py")
    print(result.stdout)

If your project needs extra Python deps, add a requirements.txt and install via SSH:

ssh.put("requirements.txt", "/home/system/my_project/requirements.txt")
ssh.run("pip install -r /home/system/my_project/requirements.txt", check=True)

Running as a systemd service​

Long-running scripts should run under systemd, not in an SSH session.

/etc/systemd/system/my-project.service:

[Unit]
Description=My Intern project
After=network-online.target openclaw.service

[Service]
Type=simple
User=system
WorkingDirectory=/home/system/my_project
ExecStart=/usr/bin/python3 /home/system/my_project/main.py
Restart=on-failure

[Install]
WantedBy=multi-user.target

Install and enable:

sudo systemctl daemon-reload
sudo systemctl enable --now my-project
journalctl -u my-project -f

Watch out for OTA​

bootstrap-server updates intern-server, bootstrap-server, OpenClaw, and the web SPA. It does not touch anything under /home/system/, so your code is safe across updates. But the OS itself is managed by apt unattended-upgrades — if you depend on a specific Python or library version, pin it in your project's venv or pip install it explicitly during deploy.