Pawelle: a private pet companion for my friend's cat, powered by a local open model
Description and use
Pawelle is a local‑first pet‑wellbeing app that uses the Gemma 3 1B model to generate daily meal and play plans, keeping all data on the user’s machine.
Problem
A friend needed a simple, daily pet‑wellbeing app that keeps data local and uses an AI to generate meal and play plans.
Outcome
All 24 generated plans in a 24‑run test were valid, with a median generation time of ~2.6 seconds, and the user reported daily usage.
Workflow
- 1. User creates a cat profile (name, photos, allergies).
- 2. Each day user taps through a quick check‑in (mood, appetite, energy, play, litter box).
- 3. App sends a prompt to the local Gemma 3 1B model via Ollama.
- 4. Safety rules filter out dangerous content before the model runs.
- 5. Model returns a short friendly sentence and selects meal times/play ideas from fixed lists.
- 6. App checks the output for allergens, medicine words, fake links, and retries if needed.
- 7. The plan is displayed and stored in a local SQLite database for diary history.
Evidence
Pawelle is a small app that helps my friend Shontu look after her cat, Pinky. Each day Shontu taps through a quick check‑in (mood, appetite, energy, play, litter box), and Pawelle writes a short, friendly plan for food and play.
In a 24-run test with the real model, all 24 plans were valid, with a median time of about 2.6 seconds.
The plan is written by a small open-weight AI running on her own computer. There's no account and no cloud, and Pinky's details and photos stay on the machine.
- Record date: 2026-10-05T01:55:55.222Z