#!/usr/bin/env bash
# Deploy Submatter / Echodo production stack on a dedicated host (CT 116).
#
# Run on the CT after git sync, or via SSH from CI:
#   ssh root@192.168.68.116 'cd /opt/submatter && bash scripts/deploy-prod.sh'
#
# Modes:
#   build   — serialized image build + compose up (default; needs ~8 GB RAM)
#   up      — compose up only (images already built)
#   pull    — docker compose pull + up (when using a registry; future)
set -euo pipefail

repo_root="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
cd "$repo_root"

mode="${1:-build}"
compose_files=(-f docker/docker-compose.prod.yml -f docker/docker-compose.prod.standalone.yml)

# Optional Phase B: export COMPOSE_PROFILES=agent-server before deploy
if [[ -n "${COMPOSE_PROFILES:-}" ]]; then
  echo "[deploy-prod] profiles=${COMPOSE_PROFILES}"
fi

if [[ ! -f .env ]]; then
  echo "[deploy-prod] ERROR: missing ${repo_root}/.env — copy from .env.example and fill production values." >&2
  exit 1
fi

export DOCKER_BUILDKIT=1

compose() {
  docker compose --project-directory "$repo_root" "${compose_files[@]}" "$@"
}

# Preserve separately managed services (such as opencode-v2); orphan cleanup
# requires an explicit operator decision, not every application deployment.
case "$mode" in
  build)
    bash scripts/build-prod.sh
    compose up -d
    ;;
  up)
    compose up -d
    ;;
  pull)
    compose pull
    compose up -d
    ;;
  *)
    echo "Usage: $0 [build|up|pull]" >&2
    exit 1
    ;;
esac

echo "[deploy-prod] waiting for web health..."
deadline=$((SECONDS + 120))
while (( SECONDS < deadline )); do
  if compose ps web 2>/dev/null | grep -q "(healthy)"; then
    echo "[deploy-prod] web is healthy"
    compose ps
    exit 0
  fi
  sleep 5
done

echo "[deploy-prod] WARN: web not healthy within 120s — check logs:" >&2
compose logs web --tail 40
exit 1