llm-gateway/deploy/deploy.sh

#!/bin/bash
# ============================================================
# LLM Gateway — One-command deploy (run locally on Mac)
#
# Usage:
#   bash deploy/deploy.sh
#   bash deploy/deploy.sh --skip-build     # skip local build
#   bash deploy/deploy.sh --health-only    # just check remote health
# ============================================================
set -euo pipefail

ERIK_HOST="82.165.222.127"
ERIK_USER="root"
REMOTE_DIR="/opt/llm-gateway"
GITEA_BRANCH="main"
HEALTH_URL="http://${ERIK_HOST}:3100/health"
HEALTH_URL_CF="https://llm-gateway.context-x.org/health"

RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m'

info()    { echo -e "${GREEN}[INFO]${NC}  $*"; }
warn()    { echo -e "${YELLOW}[WARN]${NC}  $*"; }
error()   { echo -e "${RED}[ERROR]${NC} $*" >&2; exit 1; }
section() { echo -e "\n${BLUE}>> $*${NC}"; }

# Parse args
SKIP_BUILD=false
HEALTH_ONLY=false
for arg in "$@"; do
  case $arg in
    --skip-build)   SKIP_BUILD=true ;;
    --health-only)  HEALTH_ONLY=true ;;
    *) warn "Unknown argument: $arg" ;;
  esac
done

# -------------------------------------------------------
# Health-only mode
# -------------------------------------------------------
if [[ "$HEALTH_ONLY" == "true" ]]; then
  section "Remote health check"
  STATUS=$(curl -s -o /dev/null -w "%{http_code}" "$HEALTH_URL" 2>/dev/null || echo "000")
  BODY=$(curl -s "$HEALTH_URL" 2>/dev/null || echo "{}")
  echo "  HTTP status: $STATUS"
  echo "  Response:    $BODY"
  [[ "$STATUS" == "200" ]] && info "Gateway is healthy." || warn "Gateway may be unhealthy."
  exit 0
fi

# -------------------------------------------------------
# 0. Pre-deploy checks
# -------------------------------------------------------
section "0. Pre-deploy checks"

command -v npm >/dev/null || error "npm not found."
command -v git >/dev/null || error "git not found."
command -v ssh >/dev/null || error "ssh not found."
command -v curl >/dev/null || error "curl not found."

# Check we're on main (warn only — don't block)
CURRENT_BRANCH=$(git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "unknown")
if [[ "$CURRENT_BRANCH" != "$GITEA_BRANCH" ]]; then
  warn "Current branch is '$CURRENT_BRANCH', not '$GITEA_BRANCH'. Proceeding anyway."
fi

# Check for uncommitted changes
if ! git diff --quiet 2>/dev/null; then
  warn "You have uncommitted changes. They will NOT be deployed."
fi

info "Pre-deploy checks passed."

# -------------------------------------------------------
# 1. Local build
# -------------------------------------------------------
section "1. Local build"

if [[ "$SKIP_BUILD" == "true" ]]; then
  warn "Skipping local build (--skip-build)"
else
  info "Running: npm run build"
  npm run build
  info "Build successful."
fi

# -------------------------------------------------------
# 2. Push to Gitea
# -------------------------------------------------------
section "2. Pushing to Gitea"

COMMIT_SHA=$(git rev-parse --short HEAD 2>/dev/null || echo "unknown")
info "Pushing commit $COMMIT_SHA to Gitea ($GITEA_BRANCH)..."

git push origin "$GITEA_BRANCH"
info "Push complete."

# -------------------------------------------------------
# 3. Deploy on Erik
# -------------------------------------------------------
section "3. Deploying on Erik (${ERIK_HOST})"

info "Connecting via SSH..."

# shellcheck disable=SC2087
ssh -o ConnectTimeout=15 "${ERIK_USER}@${ERIK_HOST}" bash << 'REMOTE_SCRIPT'
set -euo pipefail

REMOTE_DIR="/opt/llm-gateway"
LOG_DIR="/var/log/llm-gateway"

echo "[remote] Pulling latest code..."
cd "$REMOTE_DIR"
git fetch origin
git reset --hard origin/main

echo "[remote] Installing dependencies..."
npm install --prefer-offline 2>/dev/null || npm install

echo "[remote] Building..."
npm run build

echo "[remote] Ensuring log directory..."
mkdir -p "$LOG_DIR"

echo "[remote] Restarting PM2 processes..."
if pm2 list | grep -q "llm-gateway"; then
  pm2 reload llm-gateway llm-learning --update-env
else
  echo "[remote] PM2 processes not found — starting from ecosystem config..."
  pm2 start deploy/ecosystem.config.cjs
fi

pm2 save

echo "[remote] Deploy complete."
REMOTE_SCRIPT

info "Remote deploy finished."

# -------------------------------------------------------
# 4. Post-deploy health check
# -------------------------------------------------------
section "4. Post-deploy health check"

MAX_RETRIES=8
RETRY_DELAY=5

info "Waiting ${RETRY_DELAY}s for restart to complete..."
sleep $RETRY_DELAY

for i in $(seq 1 $MAX_RETRIES); do
  STATUS=$(curl -s -o /dev/null -w "%{http_code}" --max-time 10 "$HEALTH_URL" 2>/dev/null || echo "000")

  if [[ "$STATUS" == "200" ]]; then
    BODY=$(curl -s --max-time 10 "$HEALTH_URL" 2>/dev/null || echo "{}")
    info "Health check PASSED (HTTP 200)"
    echo "  $BODY"
    break
  fi

  if [[ $i -eq $MAX_RETRIES ]]; then
    warn "Health check did not return 200 after ${MAX_RETRIES} attempts."
    warn "Check logs on Erik: ssh root@${ERIK_HOST} 'pm2 logs llm-gateway --lines 50'"
    exit 1
  fi

  info "  Attempt $i/$MAX_RETRIES — HTTP $STATUS. Retrying in ${RETRY_DELAY}s..."
  sleep $RETRY_DELAY
done

# -------------------------------------------------------
# 5. Summary
# -------------------------------------------------------
echo ""
echo -e "${GREEN}Deploy successful!${NC}"
echo ""
echo "  Commit:         $COMMIT_SHA"
echo "  Direct:         $HEALTH_URL"
echo "  Cloudflare:     $HEALTH_URL_CF"
echo "  PM2 status:     ssh root@${ERIK_HOST} 'pm2 status'"
echo "  Logs:           ssh root@${ERIK_HOST} 'pm2 logs llm-gateway'"
echo ""