Local model chore: turn a brain-dump into a clean to-do list

The Stack

Steps

1. 1

   Run the chore on a real local model (offline)

   Install Ollama, pull a Gemma 3 model, and run the to-do prompt on your messy notes. With LM Studio you'd do the same in a window; here CI does it headless. After the download this runs with wifi off. CI runs exactly this with the tiny gemma3:270m and checks the model returned a real answer.

   command -v ollama >/dev/null 2>&1 || curl -fsSL https://ollama.com/install.sh | sh
   curl -s http://localhost:11434/api/version >/dev/null 2>&1 || (ollama serve >/tmp/ollama-serve.log 2>&1 &)
   for i in $(seq 1 60); do curl -s http://localhost:11434/api/version >/dev/null 2>&1 && break; sleep 1; done
   ollama pull gemma3:270m >/dev/null 2>&1
   OUT=$(ollama run gemma3:270m "Turn these messy notes into a short, organized to-do list:
   - talked to Sam about the Q3 launch, he will send the deck
   - need to book the dentist
   - buy milk and coffee
   - reply to the landlord about the lease")
   if [ -n "$OUT" ]; then
     echo "chore1 OK: gemma3:270m produced a $(printf '%s' "$OUT" | wc -c | tr -d ' ')-char to-do list offline (quality fenced)"
   else
     echo "EMPTY OUTPUT"; exit 1
   fi

2. 2

   Use the recommended size for real work (the quality step, not checked by CI)

   For day-to-day use, pull gemma3:4b (about 3.3GB, fits 8GB) or gemma3:12b (about 8GB, for 16GB). The bigger model writes a cleaner list. Output quality is non-deterministic, so CI never claims it.