The latest articles on DEV Community by FlexTeam26 (@flexteam26). https://dev.to/flexteam26 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%2F4232711%2Fc0e2e4a7-0a5d-4325-acca-fb822a0bd1e0.png https://dev.to/flexteam26 en FlexTeam26 Sat, 07 Nov 2026 19:43:08 +0000 https://dev.to/flexteam26/building-cocktailsabc-652-normalized-recipes-and-a-what-can-i-actually-make-engine-5c48 https://dev.to/flexteam26/building-cocktailsabc-652-normalized-recipes-and-a-what-can-i-actually-make-engine-5c48 <p>I kept hitting the same wall with cocktail sites: a 1,400-word story about the Negroni before I find out whether I can actually make one. So I built <a href="https://cocktailsabc.com" rel="noopener noreferrer">cocktailsabc.com</a> around one question — <em>what can I make from the bottles on my shelf right now?</em></p> <h2> The core feature: My Bar </h2> <p>You pick the bottles you own. The site returns every recipe you can make, ranked by how few ingredients you're missing. Two missing items is a shopping list; zero missing is tonight's drink.</p> <p>Sounds trivial until you look at the data underneath it.</p> <h2> The actual problem was ingredient normalization </h2> <p>Recipe sources are a mess:</p> <ul> <li>"sweet vermouth" vs "red vermouth" vs "Italian vermouth" — same thing</li> <li>"dry sherry" vs "fino" — same thing, different granularity</li> <li>garnishes counted as ingredients in some sources, not others</li> <li>"bar spoon", "barspoon", "1 dash" — units that mean nothing to a matching engine</li> </ul> <p>So the pipeline is:</p> <ol> <li> <strong>Collect</strong> recipes from public sources, keep the original text.</li> <li> <strong>Normalize</strong> ingredient strings into canonical IDs with an alias table (the alias table is 80% of the work).</li> <li> <strong>Classify</strong> each line: ingredient, substitution, or garnish. Garnishes don't block a match.</li> <li> <strong>Index</strong> the result as a set of canonical ingredient IDs per recipe.</li> <li> <strong>Match</strong> against the user's inventory as a set-submission test, then rank by <code>|recipe_set - inventory|</code>.</li> </ol> <p>The ranking is boring on purpose — a missing-count sort beats any clever scoring once the ingredient IDs are clean. Garbage IDs make every algorithm downstream look bad.</p> <h2> Public dataset </h2> <p>652 normalized recipes (tagged, with canonical ingredient IDs) are published as a dataset on Zenodo: DOI <code>10.5281/zenodo.24041819</code>. CSV and JSONL. If you want to build on top of it, it's there.</p> <h2> Indexing and AI discoverability </h2> <p>Small site, so no ad budget — the play is being findable:</p> <ul> <li>full sitemap + <code>llms.txt</code> so AI assistants can cite the recipe pages</li> <li>IndexNow submissions on every publish (Bing/Yandex pick up within minutes)</li> <li>one clean URL per recipe, canonical tags, hreflang for the localized hubs</li> </ul> <h2> What I'd genuinely like feedback on </h2> <ul> <li>Is "missing ingredient count" the right ranking, or should it weight <em>how likely you are to already own it</em> (bitters, simple syrup)?</li> <li>Substitution handling: should "replace Campari with Aperol" show as a match or a separate tier?</li> <li>What would make a recipe dataset trustworthy enough for you to build on?</li> </ul> <p>Stack: TypeScript front end, Python crawl/normalize pipeline, served on Kubernetes in my homelab.</p> <p>Try it: <a href="https://cocktailsabc.com" rel="noopener noreferrer">cocktailsabc.com</a> — dataset DOI <code>10.5281/zenodo.24041819</code>.</p> softwaredevelopment webdev