About
A practical AI cooking system, not a recipe content farm.
My Royal Chef AI pairs a third-party culinary LLM (via OpenRouter) with our own on-device AI/ML engine. Ask anything — ingredients you actually have, a photo of your fridge, a technique that went wrong — and get clear, practical recipes with quantities, steps, and timing. Mechanical requests (scaling, conversions, known swaps, common fixes) never leave your device.
Platform
What this is.
Core idea
Context-aware cooking help.
It is not a static recipe search page. It understands ingredients, constraints, time, and skill level so the answer changes with the situation.
Built by
Aporia Systems
Created by Yunus Emre Vurgun and Yusuf Ceylan. The focus is practical software with clear systems thinking, not decorative product noise.
Every open-ended chat turn is answered by a third-party culinary LLM (via OpenRouter), integrated into our server-side pipeline, with live web search when freshness matters. In the browser chat, mechanical requests — resizing a recipe to a serving count, unit and temperature conversions, known substitutions, curated fixes for common mistakes — are interpreted and answered instantly on your device first, at zero tokens, by our own AI/ML engine (C++ compiled to WebAssembly, plus JavaScript). The chat modes (Portion, Swap, Fix) feed the engine the context it needs, so terse commands like “2 ppl” or “x2” just work. If the server is unreachable, the on-device engine — fuzzy matching, scoring, and structured cooking knowledge over 342 tested recipes — takes over instantly, in the browser and in the Android app.
Architecture
How a request moves through the system.
1. Input
Ingredients or a question
Type what you have — ingredient names, amounts, time, or a dish idea.
2. Chef LLM
A real culinary brain
Your message is sent from our server to a third-party culinary LLM (OpenRouter) tuned for cooking: exact measurements, temperatures, and timings.
3. Fresh context
Web search + images
When a question needs current facts the server searches the web and cites sources. One photo per message is understood by a vision model.
4. Structured output
Clear, usable results
Clean markdown — ingredient tables, numbered steps, timing guides — rendered properly on web and Android, exportable as PDF.
5. On-device semantic engine
Never dead air
In the browser chat, mechanical requests — resizing a recipe, unit conversions, known substitutions, curated fixes — are interpreted and answered instantly on your device at zero tokens. If the server hiccups, the same local engine answers from 342 tested recipes and a cooking knowledge base — no spinning wheel of death.
6. Your history
Sync when you want it
Guests get one free chef answer, then the on-device engine keeps helping. A free account gets 15 chef messages a day and syncs history and saved recipes; Premium raises the daily limit to 90.
Platforms
Same chef, two clients.
Web
Browser chat + tools
The browser chat runs our own on-device AI/ML engine — C++ compiled to WebAssembly — for instant zero-token answers: recipe resizing, unit and temperature conversions, substitutions, exact recipe lookups, curated cooking facts and mistake fixes. Only what needs real culinary judgment goes through our server to a third-party LLM (OpenRouter), with compact pre-digested context. The site also hosts the kitchen tools — Substitution Smart, Flavor Pairing, Recipe Cost — and the Pasta Perfetta 3D cooking simulator. Voice input uses your browser's own speech recognition; one image per message is supported.
Android
Native app (APK)
The Android app is a native Kotlin client — not a web wrapper — with the same chef, account, and daily quotas. Its own on-device engines cover the fallback, plus app-only tools: Fridge Rescue (on-device pantry), Cook Tonight (structured dinner planning), and rewarded ads that earn bonus messages for free accounts. Distributed as a signed APK; a Play Store build is in preparation.
Principles
Product rules that shape the interface.
Speed first
Answers should feel immediate: compact prompts, cached repeat questions, and a local engine that never leaves you waiting on a spinner.
One chef, every mode
Recipe, shopping, and planning requests run through the same chef, so answers stay consistent instead of living in separate tools.
No content bloat
No forced reading, no engagement bait, and no extra narrative before the useful answer.
Honest freemium
15 free chef messages a day for everyone. Premium is one clear price — $12.99/month — and pays for the LLM that answers you.
Team
Who is building it.
Architecture & ML
Yunus Emre Vurgun
Architecture, custom ML, third-party LLM integration, backend, networking, and security.
Android
Yusuf Ceylan
Android native client, mobile optimizations, and distribution.
Contact
Need to reach the team?
Use the contact page for product questions, feedback, and partnership inquiries.