Update to compose file for my usage

docker-compose-gpu.yml

@@ -9,13 +9,14 @@ services:
     env_file:
       - .env
     environment:
+      # This internal communication uses Docker's network, not the host's localhost, so it's fine
       - ORPHEUS_API_URL=http://llama-cpp-server:5006/v1/completions
     deploy:
       resources:
         reservations:
           devices:
             - driver: nvidia
-              count: all
+              device_ids: ['0']
               capabilities: [gpu]
     restart: unless-stopped
     depends_on:
@@ -25,7 +26,8 @@ services:
   llama-cpp-server:
     image: ghcr.io/ggml-org/llama.cpp:server-cuda
     ports:
-      - "5006:5006"
+      # Changed from "5006:5006" to bind only to localhost on the host
+      - "127.0.0.1:5006:5006"
     volumes:
       - ./models:/models
     env_file:
@@ -38,7 +40,7 @@ services:
         reservations:
           devices:
             - driver: nvidia
-              count: all
+              device_ids: ['0']
               capabilities: [gpu]
     restart: unless-stopped
     command: >
@@ -65,4 +67,3 @@ services:
         echo "Model file already exists"
       fi'
     restart: "no"
-