Write CI/CD pipeline for speech models service

.github/workflows/gcp_models.yml

@@ -12,6 +12,7 @@ on:
 env:
   SERVICE: models
   REGION: us-central1
+  CONTAINER_NAME: speech-models
 
 jobs:
   deploy:
@@ -40,9 +41,34 @@ jobs:
 
       - name: Build and Push Docker image
         run: |
-          docker build -t gcr.io/${{ vars.GCP_PROJECT_ID }}/${{ env.SERVICE }} -f backend/modal/Dockerfile .
-          docker push gcr.io/${{ vars.GCP_PROJECT_ID }}/${{ env.SERVICE }}
-      
+          docker build -t gcr.io/${{ vars.GCP_PROJECT_ID }}/${{ env.SERVICE }}:${GITHUB_SHA::7} -f backend/modal/Dockerfile .
+          docker push gcr.io/${{ vars.GCP_PROJECT_ID }}/${{ env.SERVICE }}:${GITHUB_SHA::7}
+
+      - name: Create SSH Key
+        run: |
+          mkdir -p ~/.ssh
+          echo "${{ secrets.SPEECH_MODELS_SSH_PRIVATE_KEY }}" > ~/.ssh/id_ed25519
+          chmod 600 ~/.ssh/id_ed25519
+
+      - name: Deploy Docker image
+        run: |
+          ssh -o StrictHostKeyChecking=no \
+          ${{ secrets.SPEECH_MODELS_SSH_USERNAME }}@${{ secrets.SPEECH_MODELS_SSH_HOST }} \
+          'echo "[-] Pull latest Speech Models image..."; \
+          docker pull gcr.io/${{ vars.GCP_PROJECT_ID }}/${{ env.SERVICE }}:${GITHUB_SHA::7}; \
+          echo "[-] Remove current Speech Models container..."; \
+          docker rm -f ${{ vars.CONTAINER_NAME }}; \
+          echo "[-] Start new Speech Models container..."; \
+          docker run -d --name ${{ vars.CONTAINER_NAME }} -p 8080:8080 \
+            --volume /var/lib/nvidia/lib64:/usr/local/nvidia/lib64 \
+            --volume /var/lib/nvidia/bin:/usr/local/nvidia/bin \
+            --device /dev/nvidia0:/dev/nvidia0 \
+            --device /dev/nvidia-uvm:/dev/nvidia-uvm \
+            --device /dev/nvidiactl:/dev/nvidiactl \
+            -e OPENAI_API_KEY=${{ secrets.OPENAI_API_KEY }} \
+            -e HUGGINGFACE_TOKEN=${{ secrets.HUGGINGFACE_TOKEN }} \
+            gcr.io/${{ vars.GCP_PROJECT_ID }}/${{ env.SERVICE }}:${GITHUB_SHA::7}'
+
       # - name: Deploy to Cloud Run
       #   id: deploy
       #   uses: google-github-actions/deploy-cloudrun@v2