Match init-models.sh to the live preseed list

Five models from the production GPU host's current pull set. Picks up
the idempotency-checking loop pattern from the source script so re-runs
print "already present" instead of re-pulling.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

deployments/ai-stack/init-models.sh

@@ -1,12 +1,22 @@
 #!/bin/sh
 # Preseed Ollama with the models the stack should have available at startup.
 # Runs once via the model-init service (see docker-compose.yml). Safe to
-# re-run — `ollama pull` is a no-op for models already present.
+# re-run — already-present models are skipped.
 #
-# Add or remove pulls to taste. The host needs enough disk for everything
-# listed here; check sizes at https://ollama.com/library before adding.
+# Add or remove tags to taste. The host needs enough disk for everything
+# listed; check sizes at https://ollama.com/library before adding.
 
 set -e
 
-ollama pull mistral-nemo:12b
-ollama pull llama3.2:3b
+MODELS="dolphin3:8b llama3.1:8b ministral-3:8b mistral-nemo:12b qwen3.6:latest"
+
+for model in $MODELS; do
+  if ollama list | awk 'NR>1 {print $1}' | grep -qx "$model"; then
+    echo "✓ $model already present"
+  else
+    echo "→ Pulling $model…"
+    ollama pull "$model"
+  fi
+done
+
+echo "Done."