# To get an Ollama prompt, running on Modal.
$ modal shell --region us-east ollama-modal.py \
    --cmd 'systemctl start ollama && ollama run llama3:instruct'
>>> Send a message (/? for help)

# For one-off things.
modal run ollama-modal.py [--text "Why is the sky blue?"] [--lookup]

# If using --lookup, first deploy.
modal deploy ollama-modal.py

# MODEL=... can be used to control what ollama model we bake into
# underlying images, defaulting to llama3:instruct (8b). See from
# https://ollama.com/library for other options.