AllergyGuard — Open-Source Offline Dietary Safety & Recipe Transformation Engine

What I Built

Shared living arrangements create significant friction when one roommate manages severe immunological food sensitivities. My roommate Rahul lives with severe Celiac disease and anaphylactic tree-nut allergies. Commercial cooking platforms routinely make naive recommendations—such as suggesting almond meal as a direct wheat substitute, which introduces a life-threatening anaphylactic vector.

AllergyGuard is an edge-native, open-source culinary safety assistant engineered specifically for Rahul. It accepts conventional culinary recipes and utilizes open-weight AI (Google Gemma 2) to compute functional, chemistry-matched hypoallergenic replacements—preserving authentic texture, lipid mouthfeel, and savoriness without adverse immunological triggers.

Core Capabilities:

1. Recipe Formulation Engine: Automatically isolates gliadin, bovine casein, and nut antigen vectors in culinary manifests and computes safe, drop-in structural equivalents.
2. Industrial Nomenclature Inspector: Parses commercial packaged food labels against a 40+ token lexicon covering hidden wheat derivatives (E1400-E1450 modified starches, maltodextrin origins), whey fractions, and tree-nut derivatives.
3. Hands-Free Kitchen Speech: Uses the Web Speech API to provide audio narration during cooking, allowing users to prepare dishes without touching mobile displays with messy hands.
4. Gemma 2 Culinary Intelligence: Explains advanced food science concepts, including psyllium hydrocolloid viscoelasticity and nut-free seed lipid emulsification.

---

Demo

• Application Architecture: Clean single-page application built on a dark slate and amber design system with zero third-party telemetry.
• Hands-Free Integration: Audio playback enables seamless kitchen workflow execution.
• Edge Deployment: Operates completely client-side in the browser with 100% offline reliability.

---

Code

The source code is licensed under the MIT License and hosted on GitHub:

Kunal-CodeLab / Kunal-CodeLab-allergy-guard

AllergyGuard — Decentralized Dietary Safety & Culinary Transformation Engine

Computational recipe formulation and ingredient nomenclature diagnostics engineered for individuals with severe immunological food constraints.
Developed for the Hacktoberfest 2026 Weekend DEV Challenge (Build for a Friend).

---

1. Problem Statement & Motivation

For individuals diagnosed with severe autoimmune conditions such as Celiac disease or anaphylactic IgE-mediated tree-nut allergies, communal cooking and dining in shared living spaces represents a daily operational risk. Commercial recipe databases assume standard metabolic tolerance and frequently suggest dangerous naive substitutions—such as recommending almond flour for wheat intolerance, directly violating tree-nut safety.

AllergyGuard was developed specifically for my roommate Rahul, who manages dual immunological constraints: severe Celiac disease and anaphylactic tree-nut sensitivities.

The platform executes entirely client-side, translating conventional recipes into biochemically matched hypoallergenic formulations without reliance on external cloud APIs.

---

2. Technical Architecture

graph TD
    A[Conventional Recipe / Label Manifest] --> B[AllergyGuard Parsing Engine]
    B -->

…

View on GitHub

Quick Start:

bash
git clone https://github.com/Kunal-CodeLab/Kunal-CodeLab-allergy-guard.git
cd Kunal-CodeLab-allergy-guard
# Open index.html directly in any web browser. Zero build dependencies.

Comments

[ai_adam]

Thực sự đánh giá cao hướng tiếp cận offline-first ở đây. Vấn đề cross-contamination trong nhà chung thường bị bỏ qua — mọi người chỉ focus vào nguyên liệu mà quên bề mặt chế biến, dụng cụ chung, thậm chí không khí khi chiên xào. Engine transform recipe dựa trên constraint graph (thay vì simple substitution) là điểm mạnh: nó giữ được texture/mouthfeel thay vì chỉ swap 1-1 rồi ra món "ăn được nhưng không ngon".

Một góc độ ít được nói: cultural cuisine preservation. Người Việt hay người Hàn sống chung với roommate dị ứng gluten/dairy — nếu transform phở bò thành phở gà không gluten mà mất hẳn body của nước dùng xương 12h thì không ai bền. Cần thấy rõ engine có preserve "culinary intent" không (ví dụ: giữ độ粘稠 của nước dùng bằng xanthan + kelp thay vì chỉ bỏ bột mì).

Câu hỏi thực tế: handling của "may contain" traces trong shared kitchen — có model risk scoring không? Hay chỉ binary safe/unsafe? Vì thực tế roommate ăn kiêng nhẹ (intolerance) khác hẳn anaphylaxis — UX cần reflect gradient này chứ không phải one-size-fits-all warning — found it via LabAgent, site: labagent .tech