DEV Community

Cover image for NutriMate – Local AI Meal Planner for My Roommate's Severe Allergies
Viraj Radadiya
Viraj Radadiya

Posted on

NutriMate – Local AI Meal Planner for My Roommate's Severe Allergies

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built NutriMate for my roommate, Alex, who suffers from severe, multi-ingredient food allergies (peanuts, gluten, and dairy).

Problem: Planning daily meals was a constant stress for Alex. Commercial recipe apps often miss cross-contamination risks or hide allergy filters behind expensive paywalls. Moreover, Alex didn't feel comfortable uploading sensitive health and dietary preferences to public AI cloud platforms.

Solution: NutriMate is a fully offline, local AI-powered meal assistant. It analyzes available pantry items, cross-checks every ingredient against Alex's strict allergy profile, and generates personalized, safe weekly meal plans and grocery lists in seconds.

Demo

Code

{}

How I Built It

NutriMate is powered entirely by open-source AI infrastructure running locally on a standard laptop:

  • LLM / Open-Weights: Llama 3.2 (3B-Instruct) running locally via Ollama for high-speed, zero-latency meal reasoning.
  • Agent Framework: LangChain to enforce strict system guardrails, structured JSON output formats, and deterministic allergy verification.
  • Backend & UI: Python FastAPI for local endpoint management and Streamlit for a clean, user-friendly dashboard.

Why Does Open Innovation Matter?

Using open-source AI wasn't just an alternative—it was the only option that truly solved Alex's core concerns:

  1. Complete Data Privacy: Alex's medical allergy records and daily eating habits remain 100% on their personal laptop. No data ever touches a third-party server.
  2. Offline & Reliability: It runs with zero internet connection. Whether cooking in a basement apartment or traveling without WiFi, NutriMate works instantly.
  3. Zero Operating Cost: Running open-weight models locally eliminates expensive API subscription fees, making it free forever.
  4. Deterministic Behavior: Open-source frameworks allowed us to fine-tune system prompts and temperature settings to ensure the model never hallucinates unsafe ingredients.

User Feedback

Here is what Alex said when I installed NutriMate on their laptop:

"I used to spend hours scanning ingredient lists and overthinking dinner. NutriMate gave me complete peace of mind because I know my data stays on my laptop and every recipe is 100% safe for me to eat. It’s genuinely a lifesaver!"

Prize Categories

  • Main Challenge: Build for a Friend

Top comments (0)