nixos/hosts/fw/modules/openclaw-vm.nix

# OpenClaw VM — Ubuntu 24.04 QEMU guest for interactive `openclaw onboard` setup.
# Runs alongside the Podman container (openclaw.nix on .97.60) on a separate IP (.97.61).
# The container serves the gateway/webchat; this VM is for the daemon/onboarding workflow.
{ config, pkgs, lib, ... }:

let
  vmName = "openclaw-vm";
  vmStateDir = "/var/lib/${vmName}";
  vmMac = "02:00:00:00:03:01";
  vmIp = "${config.networkPrefix}.97.61";
  gateway = "${config.networkPrefix}.97.1";
  tapDevice = "vm-openclaw";

  sshAuthorizedKeys = import ../utils/ssh-keys.nix;

  gitRepoUrl = "https://git.cloonar.com/openclawd/config.git";

  # Cloud-init user-data
  userData = pkgs.writeText "user-data" ''
    #cloud-config
    users:
      - name: openclaw
        shell: /bin/bash
        sudo: ALL=(ALL) NOPASSWD:ALL  # needed for openclaw onboard --install-daemon; VM is internal-only, SSH-key-only
        groups: [sudo]
        ssh_authorized_keys:
          ${lib.concatMapStringsSep "\n      " (k: "- ${k}") sshAuthorizedKeys}

    package_update: true
    package_upgrade: true
    packages:
      - git
      - curl
      - build-essential

    runcmd:
      - [bash, /var/lib/cloud/scripts/setup-openclaw.sh]

    write_files:
      - path: /etc/cloud/cloud.cfg.d/99-disable-network-config.cfg
        content: |
          network: {config: disabled}
      - path: /var/lib/cloud/scripts/setup-openclaw.sh
        permissions: "0755"
        content: |
          #!/bin/bash
          set -euxo pipefail
          # Wait for network to be fully ready
          for i in $(seq 1 30); do
            if curl -fsS --max-time 5 https://nodejs.org/ >/dev/null 2>&1; then
              break
            fi
            echo "Waiting for network... ($i/30)"
            sleep 2
          done
          # Install Node.js 22 from official binary tarball
          NODE_MAJOR=22
          NODE_VERSION=$(curl -fsSL "https://nodejs.org/dist/latest-v''${NODE_MAJOR}.x/" | grep -oP 'node-v\K[0-9.]+' | head -1)
          curl -fsSL "https://nodejs.org/dist/v''${NODE_VERSION}/node-v''${NODE_VERSION}-linux-x64.tar.xz" \
            | tar -xJf - -C /usr/local --strip-components=1
          # Verify node and npm are available
          node --version
          npm --version
          # Install OpenClaw globally
          npm install -g openclaw@latest
          # Clone the repo as openclaw user
          mkdir -p /home/openclaw/.openclaw
          su - openclaw -c 'git clone ${gitRepoUrl} /home/openclaw/.openclaw/workspace'
  '';

  # Cloud-init meta-data
  metaData = pkgs.writeText "meta-data" ''
    instance-id: ${vmName}
    local-hostname: ${vmName}
  '';

  # Cloud-init network-config (netplan v2, MAC-based match)
  networkConfig = pkgs.writeText "network-config" ''
    network:
      version: 2
      ethernets:
        id0:
          match:
            macaddress: "${vmMac}"
          addresses: [${vmIp}/24]
          routes:
            - to: default
              via: ${gateway}
          nameservers:
            addresses: [${gateway}]
  '';

  ubuntuImageUrl = "https://cloud-images.ubuntu.com/noble/current/noble-server-cloudimg-amd64.img";
  ubuntuImageName = "noble-server-cloudimg-amd64.img";
in
{
  # Ensure KVM is available
  boot.kernelModules = [ "kvm-intel" ];

  # State directory
  systemd.tmpfiles.rules = [
    "d ${vmStateDir} 0755 root root - -"
  ];

  # Init service: download cloud image, create disk, generate seed ISO
  systemd.services."${vmName}-init" = {
    description = "Initialize ${vmName} disk and cloud-init seed";
    wantedBy = [ "multi-user.target" ];
    before = [ "${vmName}.service" ];
    requiredBy = [ "${vmName}.service" ];

    path = with pkgs; [ curl qemu_kvm cdrkit ];

    serviceConfig = {
      Type = "oneshot";
      RemainAfterExit = true;
    };

    script = ''
      set -euo pipefail

      mkdir -p "${vmStateDir}"

      # Download Ubuntu cloud image if not present
      if [ ! -f "${vmStateDir}/${ubuntuImageName}" ]; then
        echo "Downloading Ubuntu 24.04 cloud image..."
        curl -fSL -o "${vmStateDir}/${ubuntuImageName}" "${ubuntuImageUrl}"
      fi

      # Create qcow2 disk from cloud image if not present
      if [ ! -f "${vmStateDir}/disk.qcow2" ]; then
        echo "Creating qcow2 disk from cloud image..."
        qemu-img convert -f qcow2 -O qcow2 "${vmStateDir}/${ubuntuImageName}" "${vmStateDir}/disk.qcow2"
        qemu-img resize "${vmStateDir}/disk.qcow2" 20G
      fi

      # Always regenerate seed ISO (picks up config changes)
      # Cloud-init expects files named user-data, meta-data, network-config
      echo "Generating cloud-init seed ISO..."
      tmpdir=$(mktemp -d)
      cp ${userData} "$tmpdir/user-data"
      cp ${metaData} "$tmpdir/meta-data"
      cp ${networkConfig} "$tmpdir/network-config"

      genisoimage -output "${vmStateDir}/seed.iso" \
        -volid cidata -joliet -rock \
        "$tmpdir/user-data" "$tmpdir/meta-data" "$tmpdir/network-config"

      rm -rf "$tmpdir"
    '';
  };

  # QEMU VM service
  systemd.services."${vmName}" = {
    description = "OpenClaw QEMU VM";
    wantedBy = [ "multi-user.target" ];
    after = [ "${vmName}-init.service" "network-online.target" ];
    requires = [ "${vmName}-init.service" ];
    wants = [ "network-online.target" ];

    path = with pkgs; [ iproute2 qemu_kvm ];

    serviceConfig = {
      Type = "simple";
      Restart = "on-failure";
      RestartSec = 10;

      ExecStartPre = pkgs.writeShellScript "${vmName}-tap-setup" ''
        set -euo pipefail
        # Clean up stale TAP device if it exists
        if ip link show ${tapDevice} &>/dev/null; then
          ip link set ${tapDevice} down || true
          ip link delete ${tapDevice} || true
        fi
        # Create TAP device and attach to server bridge
        ip tuntap add dev ${tapDevice} mode tap
        ip link set ${tapDevice} master server
        ip link set ${tapDevice} up
      '';

      ExecStart = lib.concatStringsSep " " [
        "${pkgs.qemu_kvm}/bin/qemu-system-x86_64"
        "-machine type=q35,accel=kvm"
        "-cpu host"
        "-smp 2"
        "-m 4096"
        "-drive file=${vmStateDir}/disk.qcow2,format=qcow2,if=virtio"
        "-drive file=${vmStateDir}/seed.iso,format=raw,if=virtio,media=cdrom"
        "-netdev tap,id=net0,ifname=${tapDevice},script=no,downscript=no"
        "-device virtio-net-pci,netdev=net0,mac=${vmMac}"
        "-nographic"
        "-serial mon:stdio"
      ];

      ExecStopPost = pkgs.writeShellScript "${vmName}-tap-cleanup" ''
        if ip link show ${tapDevice} &>/dev/null; then
          ip link set ${tapDevice} down || true
          ip link delete ${tapDevice} || true
        fi
      '';
    };
  };
}