Skip to content

iamdanielkitchen/yonder

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

Yonder

List the dishes you eat on repeat; get one new recipe tuned to how far you're willing to stretch. Every pick is new to you but a plausible hit — grounded in a 2,476-recipe corpus of tested dishes, not improvised.

Live: yonderrecipes.com

Yonder — list the dishes you eat on repeat, set how far from them to go, and get one new dish

The problem

Most recipe apps optimize for fit — they show you more of what you already eat. The good ones bury you in options and turn discovery into a research project. Neither helps the person who likes to cook, is bored of the rotation, but won't gamble on a recipe that misses.

What you get back

You list what you actually eat on repeat. You get one destination — a single new dish, tuned to the distance you set:

  • A bridge — two or three sentences on why the leap is worth taking, anchored to what you already cook.
  • Kept vs. new — which axes stay familiar (flavor, method, format, texture) and which one changes.
  • A real recipe — from the corpus: servings, time, ingredients, steps, and the short shopping delta for what's new.
  • A souvenir — one technique or ingredient you carry forward.
  • Swaps — substitutions for anything hard to find.

React with one tap — would I make this? — and the next pick tunes itself.

How the engine works

Never gamble on a recipe that misses; never stop growing your range. The engine optimizes for calibrated distance, not broad fit.

  • A distance dial (1–5) controls leap size — from barely a stretch to the deep end.
  • Mood (Wholesome / Neutral / Indulgent) and focus (steer toward a subset of your rotation) shape the pick without touching the distance.
  • Hard constraints are absolute — allergies are never served, diet is a hard filter, equipment is a can-you-cook-it gate, and the leap stays inside the meal slot.

Picks come from the 2,476-recipe corpus. The model serves an improvised sketch only when nothing in the corpus fits the calibrated distance — and labels it as one.

Stack

  • Frontend — vanilla JavaScript, zero build. One page.
  • Backend — a single Cloudflare Worker; the API key stays server-side.
  • Model — Anthropic claude-sonnet-4-6, one canonical system prompt.
  • Data — Cloudflare KV for reactions, D1 (SQLite) for accounts.
  • Cost — about a cent per destination.

Deployed serverless on Cloudflare.


This repo is a public showcase. The application source lives in a private repository.

About

Yonder — a recipe app that breaks your food rut one dish at a time. Calibrated novelty on Cloudflare Workers + Claude (claude-sonnet-4-6), grounded in a 2,476-recipe corpus. Public showcase; source is private. Live at yonderrecipes.com.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors