deploy/deploy.sh

@@ -61,7 +61,9 @@ fi
 # Force recreation so updated port mappings, env vars, and container config always apply on deploy.
 compose up -d --force-recreate --remove-orphans backend frontend
 if [ "$DEPLOY_BUILD_AI_SERVICE" = "true" ]; then
-  compose up -d --force-recreate ai-service ollama
+  # Ollama is opt-in (compose "bundled-ollama" profile). Deploys reuse an
+  # existing/shared Ollama via OLLAMA_BASE_URL instead of starting a duplicate.
+  compose up -d --force-recreate ai-service
 fi
 
 if [ -n "${OLLAMA_MODEL:-}" ]; then

docker-compose.yml

@@ -75,12 +75,14 @@ services:
       context: ./tools/summarizer
       dockerfile: Dockerfile
     environment:
+      # Point at an existing/shared Ollama by setting OLLAMA_BASE_URL in .env
+      # (e.g. http://<host-ip>:11435). The in-compose ollama service below is
+      # opt-in via the "bundled-ollama" profile, so it is NOT started by default
+      # and no duplicate Ollama container is created.
       - OLLAMA_BASE_URL=${OLLAMA_BASE_URL:-http://ollama:11434}
       - OLLAMA_MODEL=${OLLAMA_MODEL:-qwen2.5:7b}
     ports:
       - "8001:8001"
-    depends_on:
-      - ollama
     networks:
       - default
       - shared_services
@@ -91,7 +93,11 @@ services:
       timeout: 10s
       retries: 3
 
+  # Opt-in only: start with `docker compose --profile bundled-ollama up`.
+  # Left out of the default set so deploys reuse an existing/shared Ollama
+  # (configured via OLLAMA_BASE_URL) instead of spinning up a duplicate.
   ollama:
+    profiles: ["bundled-ollama"]
     image: ollama/ollama:latest
     ports:
       - "11434:11434"