aipackage/install.sh

#!/usr/bin/env bash
# ─────────────────────────────────────────────
# Cezen AI Suite — Entry Level Installer
# Usage:
#   sudo bash install.sh           → Phase 1 (drivers + schedules reboot → Phase 2)
#   sudo bash install.sh --phase=2 → Phase 2 (all software, run after reboot)
# ─────────────────────────────────────────────
set -e

TIER="entry"
PHASE="1"
SKIP_ROLES=""
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
ANSIBLE_DIR="$SCRIPT_DIR/ansible"

# Load saved config (written by web setup UI before phase 1)
[ -f /opt/cezen/install.conf ] && source /opt/cezen/install.conf

for arg in "$@"; do
  case $arg in
    --tier=*)  TIER="${arg#*=}" ;;
    --phase=*) PHASE="${arg#*=}" ;;
    --skip=*)  SKIP_ROLES="${arg#*=}" ;;
  esac
done

# ── Preflight ──────────────────────────────────
check_root() {
  if [ "$EUID" -ne 0 ]; then
    echo "ERROR: Run as root: sudo bash install.sh"
    exit 1
  fi
}

check_os() {
  if [ -f /etc/os-release ]; then
    . /etc/os-release
    if [[ "$ID" != "ubuntu" ]]; then
      echo "ERROR: Ubuntu 22.04 required. Detected: $PRETTY_NAME"
      exit 1
    fi
    echo "✓ OS: $PRETTY_NAME"
  fi
}

install_ansible() {
  if ! command -v ansible-playbook &>/dev/null; then
    echo "→ Installing Ansible..."
    apt-get update -qq
    apt-get install -y -qq ansible python3-pip
  fi
  echo "✓ Ansible ready"
}

has_nvidia_pci_gpu() {
  for vendor_file in /sys/bus/pci/devices/*/vendor; do
    [ -f "$vendor_file" ] || continue
    if [ "$(tr '[:upper:]' '[:lower:]' < "$vendor_file")" = "0x10de" ]; then
      return 0
    fi
  done
  return 1
}

has_working_nvidia_driver() {
  command -v nvidia-smi &>/dev/null && nvidia-smi &>/dev/null
}

# ── Phase 1: NVIDIA drivers only ──────────────
run_phase1() {
  echo ""
  echo "╔══════════════════════════════════════════╗"
  echo "║   Cezen AI Suite — Phase 1: NVIDIA       ║"
  echo "╚══════════════════════════════════════════╝"

  if ! has_nvidia_pci_gpu; then
    echo "No NVIDIA GPU found. Continuing with CPU/non-GPU installation path."
    PHASE="2"
    run_phase2
    return
  fi

  ANSIBLE_STDOUT_CALLBACK=yaml \
  ansible-playbook -i localhost, -c local "$ANSIBLE_DIR/phase1_nvidia.yml" \
    -e "tier=$TIER"

  # Register phase 2 as a one-shot systemd service so it runs after reboot
  cat > /etc/systemd/system/cezen-phase2.service << EOF
[Unit]
Description=Cezen AI Suite Phase 2 Installer
After=network-online.target nvidia-persistenced.service
Wants=network-online.target

[Service]
Type=oneshot
ExecStart=/bin/bash -lc 'set -o pipefail; /bin/bash ${SCRIPT_DIR}/install.sh --phase=2 --tier=${TIER} 2>&1 | tee -a /var/log/cezen-install.log'
RemainAfterExit=yes
StandardOutput=journal+console
StandardError=journal+console

[Install]
WantedBy=multi-user.target
EOF

  systemctl daemon-reload
  systemctl enable cezen-phase2.service

  echo ""
  echo "✓ Phase 2 registered — will run automatically after reboot"
  echo "→ Rebooting in 10 seconds..."
  sleep 10
  reboot
}

# ── Phase 2: Full stack ────────────────────────
run_phase2() {
  echo ""
  echo "╔══════════════════════════════════════════╗"
  echo "║   Cezen AI Suite — Phase 2: Stack        ║"
  echo "╚══════════════════════════════════════════╝"

  GPU_AVAILABLE=false
  if ! has_working_nvidia_driver; then
    echo "No working NVIDIA GPU/driver found. Continuing with CPU/non-GPU installation path."
    echo "GPU-only features such as NVIDIA Docker runtime, DCGM metrics, and vLLM serving will be skipped or left inactive."
  else
    GPU_AVAILABLE=true
    echo "✓ NVIDIA driver: $(nvidia-smi --query-gpu=driver_version --format=csv,noheader | head -1)"
  fi

  # Build skip_roles extra var (comma-separated list, empty string = skip nothing)
  EXTRA_VARS="tier=$TIER skip_roles=\"$SKIP_ROLES\" gpu_available=$GPU_AVAILABLE"
  echo "→ Tier: $TIER | Skip: ${SKIP_ROLES:-none}"
  echo "→ GPU available: $GPU_AVAILABLE"

  ANSIBLE_STDOUT_CALLBACK=yaml \
  ansible-playbook -i localhost, -c local "$ANSIBLE_DIR/entry.yml" \
    -e "$EXTRA_VARS"

  # Disable one-shot service so it doesn't run again on next reboot
  systemctl disable cezen-phase2.service 2>/dev/null || true

  echo ""
  echo "╔══════════════════════════════════════════╗"
  echo "║   Cezen AI Suite installation complete!  ║"
  echo "║                                          ║"
  echo "║   JupyterLab  → http://localhost:8888    ║"
  echo "║   Ollama API  → http://localhost:11434   ║"
  echo "║   MLflow      → http://localhost:5000    ║"
  echo "║   MinIO       → http://localhost:9001    ║"
  echo "║   Grafana     → http://localhost:3000    ║"
  echo "╚══════════════════════════════════════════╝"
}

# ── Main ───────────────────────────────────────
check_root
check_os
install_ansible

if [ "$PHASE" = "1" ]; then
  run_phase1
elif [ "$PHASE" = "2" ]; then
  run_phase2
else
  echo "ERROR: Unknown phase '$PHASE'. Use --phase=1 or --phase=2"
  exit 1
fi