Start a local
model service.
By Alvin
Use a Mac running macOS 14 or newer for the Ollama route here. Apple M-series Macs use their processor and graphics chip; Intel Macs use the processor. Python preparation and the pack’s Terminal folder are on page 5. [33] [40]
Install the runner
Download Ollama from its official site, open the disk image and drag Ollama to Applications. Open it and follow its setup for the ollama Terminal command. Allow space for model downloads; our two small installed models occupied about 3.3 GB and 2.0 GB. [32] [33]
Check whether its local service is running
ollama list
A table of installed models means the service answered; an empty table means you can download one next. A connection error means the runner is not serving requests yet. Open Ollama, or use a separate Terminal window for the following local-only service command. Leave that window running. [41]
OLLAMA_NO_CLOUD=1 ollama serve
Use a second Terminal window for the workshop commands. If the service is already running, keep it and use that second window. We ran a task-owned loopback service with cloud features disabled and reused the two models already installed. This walkthrough did not repeat the downloads.