This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
What I Built
Fridge-to-Feast is a zero-effort, local-first recipe visualizer specifically designed for exhausted, busy, or unapologetically lazy home cooks who adhere to a 100% pure vegetarian diet.
Instead of staring into an open fridge with decision fatigue—or giving up on home cooking because washing 5 greasy pots feels like a mountain of labor—Fridge-to-Feast turns random ingredients into an instant, satisfying feast with minimal physical effort:
- Dual-Input Fridge Intake: Snap and drop a quick shelf photo, or type scattered ingredients (
"Potatoes, Tomatoes, Coriander"), or tap 1-click pantry staple chips. - The Sloth Cooking Energy Slider: Select how much effort you can spare:
- ⚡ Microwave Chef (3 mins): Zero pan watching, zero chopping, single mug or bowl.
- 🍳 One-Pan Stir-Fry (15 mins): Quick sizzle, light dice, maximum flavor-to-effort ratio.
- 🍲 Slow Simmer (30 mins): Low-heat stew or curry—walk away and take a nap while it simmers.
- "Minimize Dishes" 1-Vessel Lock: Automatically adapts every recipe to a single mug, skillet, or pot, eliminating cutting boards and prep bowls so cleanup takes 15 seconds.
- "Lazy Substitutions" (Zero-Store Swaps): Missing an ingredient? Don't leave your couch. Tap any pantry alternative (e.g., swapping Paneer with Tofu or Boiled Chickpeas, or Butter with Ghee + Salt). The cooking steps dynamically rephrase so instructions match how the swap actually behaves in the pan.
- 🦥 "Sloth Nap" Hands-Free Cooking Timer: Synced to your recipe duration with soothing Web Speech voice announcements ("Wake up, master... your feast is ready...") and real-time synthesized pentatonic chimes so you can rest your eyes without burning dinner.
Demo
- Live Demo: Deployed on GitHub Pages
- Local Preview: Simply open
index.htmlin any modern browser—no build steps or package installations required!
Visual Walkthrough & Features:
- Dual-Input Dropzone: Visual image scanning alongside quick-tag ingredient chips.
- Strict Pure Veg Shield: Rejects meat, poultry, seafood, eggs, gelatin, and animal broths while intelligently preserving veg terms like eggplant and chickpeas.
- Tactile Sloth Aesthetic: Warm cream (
#faf6f0), roasted brown (#2b2420), and emerald green (#10b981) color palette with light and dark mode toggling. - Offline Sloth Mode: Fully functional with zero API key required through a built-in algorithmic culinary heuristic engine.
Code
The full source code is available in this repository:
- Repository: Fridge-to-Feast on GitHub
- License: MIT License
Clean Modular Architecture (Zero Heavyweight Framework Overhead):
Week 3/ ├── index.html # Semantic HTML5, Tailwind v4 theme, modal drawers & accessible UI ├── style.css # Custom tokens, tactile shadows, cross-browser scrollbars ├── favicon.svg # Scalable vector sloth chef mascot with dark-mode awareness ├── package.json # Lightweight development server scripts ├── .gitignore # Secrets, OS junk, logs, and build artifacts exclusion ├── .vscode/settings.json # Workspace CSS & HTML linting configuration └── js/ ├── config.js # Storage keys, pantry staples, cooking modes, pure veg regex rules ├── audio.js # Web Audio API sound synthesizer (warm pentatonic pop & chime notes) ├── state.js # Reactive state container, localStorage sync & canonical deduplication ├── timer.js # Sloth Nap hands-free timer with Web Speech API voice synthesis ├── substitutions.js # Pure veg zero-store pantry substitutions engine & AI rephraser ├── recipes-data.js # Curated offline pure veg culinary database ├── recipe-engine.js # Recipe synthesis engine & single-vessel constraint enforcer ├── gemini.js # Google AI Studio dynamic model probe & API caller ├── scanner.js # Client-side FileReader fridge photo dropzone & local visual analyzer ├── ui.js # DOM visual bindings, theme toggle, toasts & modals └── app.js # Application bootstrap & lifecycle orchestration How I Built It
Fridge-to-Feast is built around a resilient hybrid AI and local-first architecture:
-
Multimodal Vision & Recipe Synthesis (Google AI Studio / Gemini API):
- Uses
gemini-1.5-flashwith dynamic fallback across Gemini models. - Inspects raw fridge photos to identify edible vegetarian produce and builds customized step-by-step recipes tailored to the user's chosen energy level.
- When substituting ingredients, a dedicated AI rephrasing prompt rewrites instructions so cooking steps accurately reflect the culinary nature of the replacement item.
- Uses
-
The Offline Sloth Engine (Resilient Local Heuristic Inference):
- When offline or when no API key is provided, the app seamlessly engages a local-first heuristic recipe engine.
- Backed by
OFFLINE_RECIPES_DATABASEand procedural single-vessel synthesis with canonical stem deduplication (usinggetCanonicalStem), ensuring ingredients and steps are never awkwardly repeated.
-
Pure Vegetarian Guardrails:
- A multi-tier regex and keyword sanitizer actively screens inputs to ensure 100% vegetarian compliance, preventing accidental non-veg contamination.
-
Web Audio & Web Speech APIs:
- Instead of loading heavy external MP3 audio assets, tactile clicks and soothing chimes are generated entirely in real-time via the browser's
AudioContext. - The Sloth Nap Timer discovers system neural voices (such as Microsoft Aria/Jenny or macOS Samantha) to deliver unhurried, gentle wake-up announcements.
- Instead of loading heavy external MP3 audio assets, tactile clicks and soothing chimes are generated entirely in real-time via the browser's
Why Does Open Innovation Matter?
Kitchen Privacy & Personal Sovereignty:
Your personal fridge photos, dietary habits, and grocery inventory stay where they belong: inside your browser'slocalStorage. Closed commercial apps monetize user grocery data; open innovation guarantees user privacy by default.Resilience Across the Offline Gap:
Cooking happens in real-world kitchens where Wi-Fi drops or data signals lag. By pairing open-source AI with local algorithmic heuristic engines, users can cook hot meals even when completely disconnected from the internet.Zero Paywalls for Daily Sustenance:
Nobody should have to pay a recurring monthly subscription or hit API rate limits just to figure out what to cook with leftover potatoes and rice. Open-source keeps technology accessible to everyone.Community-Tailored Culinary Knowledge:
Proprietary cooking apps frequently overlook global vegetarian traditions. An open-source codebase empowers developers worldwide to contribute regional pantry staples, spices, and authentic vegetarian substitution cheat sheets.
My Agent Session
This project was built pair-programming with the DeepMind Antigravity AI coding agent. Key milestones navigated during the session included:
- Diagnosing and resolving subtle browser TextMate grammar traps (
accept="image/*"comment collisions). - Modernizing Tailwind CSS v4 CSS-first token architectures with class-based
.darkvariants. - Eliminating compound substitution collisions (e.g. preventing
"1 tbsp oil/butter"from duplicating into repeated terms). - Synthesizing tactile Web Audio API pentatonic soundscapes and crafting the custom scalable SVG sloth chef favicon.
The Lazy Payoff
- Decision Fatigue Eliminated: Saves 25–35 minutes of scrolling through complex, ingredient-heavy recipes.
- Dishes Avoided: Reduces kitchen sink pileup by 75–80%, locking meal prep to exactly 1 cooking vessel (0 prep plates, 0 cutting boards).
- Zero Grocery Store Trips: The Lazy Substitutions engine spares you a 40-minute round trip to the supermarket just because you were missing one herb or dairy item.
- Guilt-Free Rest: The Sloth Timer voice announcer lets you lie down on the sofa and nap peacefully while your one-pot stew or mug toss cooks safely.
Prize Categories
- Primary Track: Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
- Partner Track: Google AI Studio / Gemini API (Multimodal Vision & Creative Recipe Synthesis)
Built with ❤️ and maximum kitchen laziness for Hacktoberfest 2026 🦥✨
Happy cooking, and even happier napping!

