Baby Ledger AIThe AllerSee™ scanner is the allergen label scanner built into Baby Ledger AI. The design decision I care about most in it isn't a feature you can see in a screenshot. It's what the scanner does when it has no data, because for a safety tool, "what do you do when the data is empty" is the whole game.
The silent failure most scanners have
Most barcode allergen scanners do the efficient thing: look up the product, read the ingredient list, and check it against the allergens you care about. That's fine when the product is in the database. But scan an imported product the database has never seen, and a lot of scanners just return "no allergens found" and stop. That's a silent failure. "No data" renders on screen exactly like "safe," when the truth is the tool has no idea. For an allergy parent, the dangerous moment isn't when the database is full - it's when it's empty.
The decision: get more careful when the data is thin
So the scanner is built to get more careful, not less, exactly when the data is thin. An empty or unreadable record is treated as a reason to look harder - including reading label text across several languages - never as an all-clear. It leans toward telling you "double-check this" over "looks fine." It can over-flag, and honestly I'd rather it do that than stay quiet. "No data" is never shown as "no allergens" - if it can't read something, it says so.
The scan that proved the point out: an imported product returned zero ingredient data from the open product database we pull from (Open Food Facts, over 3 million products). Instead of a green check, the scanner flagged peanut against a profile that had peanut set as an allergen, and threw a hard "do not feed" block. To be honest, the product name literally said peanut butter, so a careful human catches this one too - the point isn't that it's clever, it's that it refused to fake confidence on missing data instead of returning "no allergens found."
Two lessons from building it solo
1. The problem that looks hardest is usually the one you can buy. "Read foreign-language label text" sounds like the hard part. It isn't - that's a solved problem off the shelf. The genuinely hard part was the logic and the ontology: what counts as a match, what counts as "safe," and what the app does when the data is incomplete. That's where the months went, and it's where most of the actual product lives.
2. Build the safe failure mode first. The most important behavior in the whole app is what happens when something goes wrong - sparse data, a timeout, an unreadable label. If you build the happy path first and bolt on error handling later, your "safety" feature is decorative.
How it's built
For the builders who find their way here: it's an Expo / React Native app on Supabase (Postgres with row-level security, plus realtime for co-parent sync), with the AI calls server-proxied through an edge function so no API key ever ships in the client bundle. Open Food Facts is the barcode database, and in-app purchases run on StoreKit 2. Happy to talk architecture - the allergen ontology, the empty-data design, or the server-proxy-vs-client-key question - if you reach out.
The honest limits
To be clear about the limits, because they matter for an allergy tool: false negatives are still possible, AllerSee is an awareness and label-reading aid, not a medical device, and it never replaces reading the full label or talking to your pediatrician. When the product database has nothing, it gets more cautious instead of waving the product through - but "more cautious" is not the same as "catches everything," and it never will be.
The safety layer - the AllerSee allergen cross-check, barcode scanning, and the daily FDA recall check - is free and unlimited on every plan. See how the AllerSee scanner works, or read our allergen guides.
Frequently asked questions
What does the scanner do when a product isn't in the database?
It gets more cautious, not less. An empty or unreadable record is treated as a reason to look harder, never as an all-clear, and "no data" is never shown as "no allergens found." It leans toward telling you to double-check rather than waving a product through.
Does the scanner guarantee a food is safe for my child?
No. AllerSee is an awareness and label-reading aid, not a medical device. False negatives are possible, so it never replaces reading the full label yourself or talking to your pediatrician.
Can it read foreign-language labels?
It reads label text across several languages, including Japanese, Chinese, Korean, German, and Cyrillic alongside English (the app interface is in English). When it cannot read something, it says so rather than passing the product.