added management to loaded models to prevent overloading GPU and keep things as fast as possible.

setup-ai-stack.sh

@@ -206,6 +206,7 @@ if [[ "$1" != "--update" ]]; then
       docker run -d -p 3000:8080 --gpus all \
         -e OPENAI_API_BASE_URL=http://pipelines:9099 \
         -e OPENAI_API_KEY=0p3n-w3bu! \
+        -e OLLAMA_MAX_LOADED_MODELS=1
         -v ollama:/root/.ollama \
         -v open-webui:/app/backend/data \
         --name open-webui \
@@ -476,6 +477,7 @@ if [[ "$1" != "--update" ]]; then
         -e ENABLE_BACKENDS=llama-cuda,ollama \
         -e INCLUDE_DEFAULT_MODELS=true \
         -e AUTOLOAD_MODELS=true \
+        -e MODEL_IDLE_TIMEOUT=600
         --restart unless-stopped \
         localai/localai:latest-aio-gpu-nvidia-cuda-12