Build for a Friend: a local-first meal planner on open-weight Gemma

Description and use

A local‑first meal planner that runs entirely on a laptop, using the open‑weight Gemma 3 1B model via Ollama. It generates a 7‑day Indian meal plan with grocery list and budget check, enforces allergy safety with a deterministic guardrail, and operates offline without cloud dependencies.

Problem

Providing a high-protein vegetarian meal plan on a ₹2000/week budget while avoiding peanuts and lactose, without relying on cloud services or API keys.

Outcome

The app delivers a 7‑day meal plan within ~40s on CPU, keeps all health data local, works offline, and is free forever.

Workflow

• 1. Install Ollama and pull gemma3:1b model
• 2. Run python3 app.py
• 3. Open http://localhost:8000 in browser
• 4. Enter friend’s profile (diet, allergies, dislikes, budget)
• 5. Receive streamed meal plan token by token
• 6. Review Allergy Check section for flagged dishes
• 7. Adjust plan if needed

Evidence

My roommate Nagaraj needed high-protein vegetarian food on a ₹2000/week budget — while strictly avoiding peanuts and lactose.

Source

Friend Meal Planner — a web app (vanilla HTML/JS + a zero-dependency Python server, stdlib only) that generates a 7-day Indian meal plan with a grocery list and budget check.

Source

run python3 app.py and open http://localhost:8000 (needs Ollama + gemma3:1b pulled; ~40s per plan on CPU)

Source

Testing caught something important: the 1B model kept suggesting paneer and yogurt to my lactose-allergic friend, despite explicit instructions.

Source

I added a deterministic allergy guardrail — every plan is scanned line-by-line against an allergen keyword map and an ⚠️ Allergy Check section is appended flagging risky dishes for review.

Source