# Let ollama to save models into our home directory
export OLLAMA_MODELS=/path/to/our/ml-models/ollama/

# Installing models (if needed):
path/to/ollama pull llava3 # Or llama3.2-vision, gemma3, etc

# Start ollama in a sandbox with a custom sandbox name "ollama"
firejail --noprofile --net=none --name=ollama path/to/ollama serve

# Join an existing sandbox and sending a command to ollama
firejail --noprofile --net=none --join=ollama \
  curl -X POST http://localhost:11434/api/chat \
  -H "Content-Type: application/json" \
  -d '{...}'