The latest articles on DEV Community by Ryan (@ryancodedev). https://dev.to/ryancodedev https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4157159%2Fae924bb8-c3a0-45c0-ba05-4c34ba150a8f.png https://dev.to/ryancodedev en Ryan Sun, 04 Oct 2026 15:19:20 +0000 https://dev.to/ryancodedev/gluten-gumshoe-37f3 https://dev.to/ryancodedev/gluten-gumshoe-37f3 <p><em>This is a submission for the <a href="https://dev.to/challenges/hacktoberfest-weekend-2026-10-01">Hacktoberfest Weekend Challenge: Build for a Friend</a></em></p> <h2> What I Built </h2> <p>My spouse has some sensitivities to gluten, so we're constantly reading ingredient labels to see what foods might contain gluten. It can be a time-consuming process, especially for foods with long ingredient lists.</p> <p>We previously would snap a photo of an ingredient list and send it to ChatGPT to ask if a product contained gluten, but photo processing can quickly cause you to hit your AI usage limits. It was important to me to develop a web app that runs client-side on the user's device and can be run offline, so usage limits and data connectivity are not a concern.</p> <p><strong>Gluten Gumshoe</strong> is a web app with a client-side AI OCR engine that can "read" the ingredients from a label and run them through a set of client-side rules to quickly determine if an ingredient may contain gluten.</p> <h2> Demo </h2> <p>Try out the alpha version at: <a href="https://gluten-gumshoe.pages.dev" rel="noopener noreferrer">https://gluten-gumshoe.pages.dev</a></p> <h2> How I Built It </h2> <p>For "reading" the ingredient labels, I used <strong>PP-OCRv5</strong>, which is an AI-based OCR system that can run locally on a smartphone, tablet or laptop. This provides several advantages to a simple OCR engine as ingredient labels can contain 10–30 separate lines or text regions. The current PP-OCRv5 mobile recognizer is good at reading distorted/rotated text, which one often encounters when trying to take a picture of a product's label. With the help of Codex, I implemented the PP-OCRv5 engine for images obtained from a live camera capture or a pre-existing photo.</p> <p>The ingredients list is then run through a set of client-side rules to assess if an ingredient may contain gluten or not, using data from the US FDA data for packaged foods.</p> <h2> Why Does Open Innovation Matter? </h2> <p>When you're concerned if eating something can make you sick, you don't want to think about chatbot usage limits or if you have good cellular service. You'd rather be enjoying the food than staring at an app's loading spinner. This solution can run client-side and use local AI to read an ingredient label and provide insights in seconds. This is also privacy-focused as all data stays on your device, so you don't need to worry about a chatbot accessing your photo's location data or analyzing something in the background of your image.</p> devchallenge weekendchallenge hf26challenge