#!/usr/bin/env bash
# gpu-lara.sh — Spin up/down Ollama on gaming PC (192.168.1.10)
# Usage: gpu-lara up | down | status | run "prompt"

GAMING_PC="192.168.1.10"
OLLAMA_PORT="11434"

up() {
  echo "🎮 Starting Ollama on gaming PC..."
  if ! ping -c1 -W2 "$GAMING_PC" &>/dev/null; then
    echo "❌ Gaming PC offline ($GAMING_PC)"
    return 1
  fi
  ssh "$GAMING_PC" "sudo systemctl start ollama" 2>/dev/null
  # Wait for ready
  for i in {1..10}; do
    if ssh "$GAMING_PC" "curl -sf http://localhost:$OLLAMA_PORT/api/tags" &>/dev/null; then
      echo "✅ Ollama ready (${i}s)"
      # Pre-load model
      ssh "$GAMING_PC" "ollama run glm-4.7-flash '/no_think ping' 2>/dev/null" &>/dev/null &
      echo "🎮 GLM-4.7-Flash loading..."
      return 0
    fi
    sleep 1
  done
  echo "❌ Ollama failed to start"
  return 1
}

down() {
  echo "🎮 Stopping Ollama on gaming PC..."
  ssh "$GAMING_PC" "sudo systemctl stop ollama" 2>/dev/null
  echo "✅ GPU freed — game on"
}

status() {
  if ! ping -c1 -W2 "$GAMING_PC" &>/dev/null; then
    echo "❌ Gaming PC offline"
    return 1
  fi
  local ps
  ps=$(ssh "$GAMING_PC" "ollama ps 2>/dev/null")
  if echo "$ps" | grep -q "glm-4.7-flash"; then
    echo "🎮 Lara online — GLM-4.7-Flash loaded"
    echo "$ps"
  elif ssh "$GAMING_PC" "systemctl is-active ollama" &>/dev/null; then
    echo "🎮 Ollama running (no model loaded)"
  else
    echo "💤 Ollama stopped — GPU free"
  fi
}

run_prompt() {
  # Auto-start if needed
  if ! ssh "$GAMING_PC" "curl -sf http://localhost:$OLLAMA_PORT/api/tags" &>/dev/null; then
    up || return 1
    sleep 3
  fi
  ssh "$GAMING_PC" "ollama run glm-4.7-flash '$1'"
}

case "${1:-status}" in
  up)     up ;;
  down)   down ;;
  status) status ;;
  run)    run_prompt "$2" ;;
  *)      echo "Usage: gpu-lara up|down|status|run \"prompt\"" ;;
esac