diff --git a/.gitea/workflows/deploy.yml b/.gitea/workflows/deploy.yml
index f853e74..214388d 100644
--- a/.gitea/workflows/deploy.yml
+++ b/.gitea/workflows/deploy.yml
@@ -38,22 +38,51 @@ jobs:
 
       - name: Ensure API backend is running
         run: |
-          # Start MySQL + Redis via docker compose (try common locations)
-          for dir in /opt/zhixi /root/zhixi /home/*/zhixi; do
-            if [ -f "$dir/docker-compose.yml" ]; then
-              cd "$dir" && docker compose up -d mysql redis 2>/dev/null || true
-              break
-            fi
-          done
+          set -x
+
+          # Clone or pull api-server repo for docker-compose.yml
+          if [ -d /tmp/api-server ]; then
+            cd /tmp/api-server && git pull 2>/dev/null || true
+          else
+            git clone http://localhost:3000/suche-Hermes/api-server.git /tmp/api-server 2>/dev/null || true
+          fi
 
           # Create shared network if missing
           docker network inspect zhixi-net >/dev/null 2>&1 || docker network create zhixi-net
 
-          # Restart API container if not running
-          if ! docker ps --format '{{.Names}}' | grep -q '^zhixi-api$'; then
-            echo "[deploy] zhixi-api is down, attempting restart..."
-            docker start zhixi-api 2>/dev/null || true
-            sleep 5
+          # Start MySQL + Redis via docker compose
+          if [ -f /tmp/api-server/docker-compose.yml ]; then
+            cd /tmp/api-server && docker compose up -d mysql redis 2>&1 || true
+          fi
+
+          # Wait for MySQL to be ready
+          sleep 5
+
+          # Check current state
+          echo "=== Container status ==="
+          docker ps -a --format 'table {{.Names}}\t{{.Status}}' 2>/dev/null | grep -iE 'zhixi|mysql|redis' || true
+
+          # If API container exists but stopped, start it
+          if docker ps -a --format '{{.Names}}' | grep -q '^zhixi-api$'; then
+            if ! docker ps --format '{{.Names}}' | grep -q '^zhixi-api$'; then
+              echo "[deploy] zhixi-api exists but stopped, starting..."
+              docker start zhixi-api 2>&1 || true
+              sleep 8
+            fi
+          else
+            # No container — create one
+            echo "[deploy] zhixi-api not found, building and creating..."
+            cd /tmp/api-server && docker build -t zhixi-api:latest . 2>&1 || true
+            ENV_FILE=""
+            [ -f /etc/zhixi/.env.production ] && ENV_FILE="--env-file /etc/zhixi/.env.production"
+            docker run -d \
+              --name zhixi-api \
+              --network zhixi-net \
+              --restart unless-stopped \
+              -p 3001:3000 \
+              $ENV_FILE \
+              zhixi-api:latest 2>&1 || true
+            sleep 8
           fi
 
           # Health check
@@ -61,9 +90,12 @@ jobs:
             echo "[deploy] Backend health OK"
           else
             echo "[deploy] WARNING: Backend health check failed"
-            docker ps --format 'table {{.Names}}\t{{.Status}}' 2>/dev/null | grep -i zhixi || true
+            echo "=== API container logs (tail 40) ==="
+            docker logs zhixi-api --tail 40 2>&1 || true
           fi
 
+          set +x
+
       - name: Install Nginx config (HTTPS)
         run: |
           if [ -f /etc/letsencrypt/live/longde.cloud/fullchain.pem ]; then