diff --git a/.gitea/workflows/deploy.yml b/.gitea/workflows/deploy.yml
index 8c8a0fd..17c92b1 100644
--- a/.gitea/workflows/deploy.yml
+++ b/.gitea/workflows/deploy.yml
@@ -148,75 +148,10 @@ jobs:
             sleep 10
           done
 
-      # ═══ Prediction (FastAPI → redis-211) ═══
-      - name: Deploy prediction via SSH
-        env:
-          DEPLOY_KEY: ${{ secrets.DEPLOY_SSH_KEY }}
-          PRED_HOST: 192.168.1.18
-          PRED_PORT: 32023
-        run: |
-          mkdir -p ~/.ssh
-          printf '%s\n' "$DEPLOY_KEY" > ~/.ssh/id_deploy
-          chmod 600 ~/.ssh/id_deploy
-
-          SSH_OPTS="-i ~/.ssh/id_deploy -o StrictHostKeyChecking=no -o ConnectTimeout=10 -o ServerAliveInterval=15 -p $PRED_PORT"
-          SCP_OPTS="-i ~/.ssh/id_deploy -o StrictHostKeyChecking=no -P $PRED_PORT"
-
-          REMOTE_DIR=/home/apps/kcg-prediction
-
-          # 코드 전송 (rsync 대체: tar + scp)
-          tar czf /tmp/prediction.tar.gz -C prediction --exclude='__pycache__' --exclude='venv' --exclude='.env' .
-          for attempt in 1 2 3; do
-            echo "SCP prediction attempt $attempt/3..."
-            if scp $SCP_OPTS /tmp/prediction.tar.gz root@$PRED_HOST:/tmp/prediction.tar.gz; then break; fi
-            [ "$attempt" -eq 3 ] && { echo "ERROR: SCP failed"; exit 1; }
-            sleep 10
-          done
-
-          # systemd 서비스 파일 전송
-          scp $SCP_OPTS deploy/kcg-prediction.service root@$PRED_HOST:/tmp/kcg-prediction.service
-
-          # 원격 설치 + 재시작 (단일 SSH — tar.gz는 SCP에서 이미 전송됨)
-          ssh $SSH_OPTS root@$PRED_HOST bash -s << 'SCRIPT'
-            set -e
-            REMOTE_DIR=/home/apps/kcg-prediction
-            mkdir -p $REMOTE_DIR
-            cd $REMOTE_DIR
-
-            # 코드 배포
-            tar xzf /tmp/prediction.tar.gz -C $REMOTE_DIR
-
-            # venv + 의존성
-            python3 -m venv venv 2>/dev/null || true
-            venv/bin/pip install -r requirements.txt -q
-
-            # SELinux 컨텍스트 (Rocky Linux)
-            chcon -R -t bin_t venv/bin/ 2>/dev/null || true
-
-            # systemd 서비스 갱신
-            if ! diff -q /tmp/kcg-prediction.service /etc/systemd/system/kcg-prediction.service >/dev/null 2>&1; then
-              cp /tmp/kcg-prediction.service /etc/systemd/system/kcg-prediction.service
-              systemctl daemon-reload
-              systemctl enable kcg-prediction
-            fi
-
-            # 재시작
-            systemctl restart kcg-prediction
-
-            # health 확인 (60초 — 초기 로드에 ~30초 소요)
-            for i in $(seq 1 12); do
-              if curl -sf http://localhost:8001/health > /dev/null 2>&1; then
-                echo "Prediction healthy (attempt ${i})"
-                rm -f /tmp/prediction.tar.gz /tmp/kcg-prediction.service
-                exit 0
-              fi
-              sleep 5
-            done
-            echo "WARNING: Prediction health timeout (서비스는 시작됨, 초기 로드 진행 중)"
-            systemctl is-active kcg-prediction && echo "Service is active"
-            rm -f /tmp/prediction.tar.gz /tmp/kcg-prediction.service
-          SCRIPT
-          echo "Prediction deployment completed"
+      # ═══ Prediction (FastAPI) — CI/CD 제외, 수동 배포 ═══
+      # ssh redis-211 에서 수동 배포:
+      #   scp prediction/*.py redis-211:/home/apps/kcg-prediction/
+      #   ssh redis-211 "sudo systemctl restart kcg-prediction"
 
       - name: Cleanup
         if: always()