Tastefold

Inspiration

Cooking is a hobby for me, and I originally built Tastefold for my own kitchen. I could not find a recipe app that combined Git-inspired recipe versioning with hands-free cooking controls.

What it does

Tastefold is a local recipe book app. Recipes, shopping lists, cooking history, and settings stay in the browser.

An original recipe can have connected variations. A variation stores only what changed, continues to inherit later improvements from the original, and can apply selected changes back to the original.

Cook mode has resumable sessions, timers, a cooking log, and optional offline voice controls. Tastefold also includes shopping lists, recipe history and comparison, tags and collections.

How we built it

Tastefold uses Svelte 5 and TypeScript. Dexie and IndexedDB store data in the browser, and Vosk provides local speech recognition.

The project existed before Build Week. It already had the local recipe editor, variation model, basic Cook and voice controls, import and export.

During Build Week, I added resumable Cook sessions, reliable timers, cooking logs, named recipe history, comparison and restoration, organization tools, a visual redesign.

I used GPT-5.6 Sol and Codex in ChatGPT Work. I decided what to build and how it should behave, then reviewed every iteration.

Codex generated all the Build Week implementation code. It carried each decision through the screens, local data, imports, translations, and tests. For example, resumable cooking required coordinated changes to Cook mode, timers, the Finish flow, cooking history, backups, and tests. Codex updated those connected parts together, which made development much faster.

Challenges we ran into

Timers had to remain correct after a reload, device sleep, or delayed background updates. Saving a countdown was not enough because time could appear to stop while the page was inactive.

Recipe restoration also had to preserve the existing history instead of quietly rewriting it. Shopping-list transfers needed safe ways to merge or replace data without using a server.

Accomplishments that we're proud of

I am proud that Tastefold is something I actually use in my kitchen, not just a hackathon demo.

What we learned

I learned that reliable timers should store a fixed end time rather than depend on a counter running continuously.

I also learned that restore-as-new makes recipe history easier to understand and safer to use. Local-first software avoids server complexity, but it still requires careful work around saving, transfer, updates, and recovery.

What's next for Tastefold

I plan to keep using Tastefold in my own kitchen and improve it based on real cooking sessions. I want to improve shopping list behavior, make offline voice controls more reliable across languages and make commands more natural, and I'd like to eventually do a complete redesign to make it easier and more intuitive to use.

Built With

• svelte
• typescript