Plan lunch without losing track of pantry stock
At a glance
Cheomil uses a local Gemma model to propose meals from measured pantry stock and waits for confirmation before deducting ingredients. This workflow is listed under meal-planning, pantry, local-ai.
- Agents / tools
- Cheomil, Gemma 3 4B, Ollama
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) On Windows, install Node.js 24.12 or newer and Ollama. Clone https://github.com/MasterFloppa/Cheomil.git, enter the Cheomil directory, run `ollama pull gemma3:4b`, and keep Ollama running at http://127.0.0.1:11434. 2) Run `npm start` and open http://127.0.0.1:3000. The production app needs no npm runtime dependency installation. 3) Enter your pantry ingredients with quantities in grams, milliliters, or pieces. Adding an existing ingredient increases its stock. 4) Ask what you can make for lunch. Review the proposed ingredients, quantities, serving count, and cooking instructions; ask follow-up questions if needed. 5) After cooking, identify the dish in chat, review the quantities to deduct, and confirm the inventory change. Record grocery purchases through the same review-and-confirm loop.
Worth knowing
The author added quantity checks after a recipe treated two grams of onion as meaningful. Starting a new chat clears conversation memory while preserving pantry stock.
