fix(compose): image-gen default off on non-GPU platforms + dreamforge network convention

dream-server/docker-compose.amd.yml

@@ -70,6 +70,7 @@ services:
   open-webui:
     environment:
       - OPENAI_API_KEY=no-key
+      - ENABLE_IMAGE_GENERATION=${ENABLE_IMAGE_GENERATION:-true}
 
   dashboard-api:
     environment:

dream-server/docker-compose.base.yml

@@ -82,7 +82,11 @@ services:
       ENABLE_SEARCH_QUERY_GENERATION: "true"
       WEB_SEARCH_RESULT_COUNT: "5"
       # ── ComfyUI Image Generation (SDXL Lightning 4-step) ──
-      ENABLE_IMAGE_GENERATION: "${ENABLE_IMAGE_GENERATION:-true}"
+      # Default OFF in base: ComfyUI ships only on GPU backends (amd/nvidia per
+      # comfyui/manifest.yaml). The amd and nvidia overlays re-assert this as
+      # ${...:-true} so image-gen is on for GPU users by default. CPU/Apple
+      # users would otherwise hit a missing-engine error from Open WebUI.
+      ENABLE_IMAGE_GENERATION: "${ENABLE_IMAGE_GENERATION:-false}"
       IMAGE_GENERATION_ENGINE: "comfyui"
       COMFYUI_BASE_URL: "http://comfyui:8188"
       IMAGE_SIZE: "1024x1024"

dream-server/docker-compose.nvidia.yml

@@ -29,3 +29,4 @@ services:
   open-webui:
     environment:
       AUDIO_STT_MODEL: "deepdml/faster-whisper-large-v3-turbo-ct2"
+      ENABLE_IMAGE_GENERATION: "${ENABLE_IMAGE_GENERATION:-true}"

dream-server/extensions/services/dreamforge/compose.yaml

@@ -54,9 +54,3 @@ services:
       options:
         max-size: "10m"
         max-file: "3"
-    networks:
-      - dream-network
-
-networks:
-  dream-network:
-    external: true