Simmer: From Save to Serve

Share recipes to Simmer and turn them into cooked meals with instant grocery lists and pantry-aware matching.

Comment

🍳 Simmer: From Save to Serve

Inspiration 💡

We don’t have a recipe discovery problem. We have an execution problem.

People save recipes constantly — from YouTube, blogs, TikTok, Instagram, and screenshots — but most of them never get cooked.

The friction isn’t inspiration. It’s the questions that follow:

  • Do I already have the ingredients? 🧂
  • What do I still need? 🛒
  • Is this safe for my dietary restrictions? 🥗
  • Can I realistically make this tonight? ⏳

At the moment motivation meets uncertainty, friction wins.

Simmer was built to close that gap — the space between saving a recipe and actually cooking it.

Our goal was simple:

Turn “I should make this” into “Dinner is ready.” 🍽️

What We Built 🚀

Simmer is an AI-powered execution engine for saved recipes.

Instead of building another recipe library, Simmer works with the content users already consume and transforms it into a cook-ready plan in seconds.

The Core Loop 🔄

  1. Import any recipe
  • Share a link directly to the app 📤 and it gets saved
  • Paste a YouTube or blog link (TikTok, Instagram are incoming next 😉)
  • Upload a screenshot
  • Snap a photo of a cookbook 📖
  • Upload a PDF
  1. AI parses it instantly 🤖
  • Structured ingredients
  • Clear step-by-step instructions
  • Prep and cook times
  1. Pantry-aware intelligence 🧺
  • “What can I cook now” status — based on available ingredients
  • Smart substitutions — healthy swaps, allergen alternatives, etc.
  1. Instant grocery list 🛍️
  • One-tap add
  • Pantry reconciliation
  1. Focused cooking mode 🔥
  • Step-by-step interface
  • Smart timers
  • Ingredient scaling

In early use, most users imported a recipe and generated a cook-ready grocery list within their first session — directly converting saved inspiration into an actionable meal plan.

Simmer doesn’t optimize for browsing. It optimizes for completion. ✅

The Chef Bot — Context-Aware Execution Assistant 🤖✨

Simmer includes a conversational AI companion that is deeply integrated into the app’s logic.

Unlike generic chatbots, the Chef Bot operates with structured awareness of:

  • The user’s pantry inventory 🧂
  • Saved recipe collection 📚
  • Dietary preferences 🥦
  • Allergen restrictions ⚠️

Powered by Gemini, the bot can:

  • Navigate users to specific screens based on questions
  • Generate new recipes (just name whatever you are craving for and the bot will curate a recipe 😉🌶️)
  • Open saved recipes
  • Suggest substitutions

This enables queries like:

  • “What are some healthy alternatives for the recipe?”
  • “What is the nutritional value?”
  • “Find a spicy pasta recipe I can cook tonight.”

The Chef Bot doesn’t just reply with text — it performs actions inside the app.

It turns AI from passive assistant into active execution partner.

Technical Architecture 🏗️

Simmer is built as a production-ready, cross-platform application.

Frontend

  • Flutter (iOS & Android)
  • Riverpod for reactive state management
  • Hive for offline-first local storage
  • Optimistic UI updates
  • Edge-to-edge immersive design

Backend

  • Firebase Authentication
  • Firestore database
  • Node.js
  • Async background cloud sync

All core data (recipes, pantry, shopping lists) is stored locally first, then synced asynchronously. Perceived Latency = ~0, even under unstable network conditions.

AI Pipeline 🧠

Simmer leverages Gemini 2.5 Flash for ultra-low latency multimodal reasoning.

Supported Inputs

  • URLs (YouTube, Web)
  • Screenshots
  • Cookbook photos 📷
  • PDFs

Monetization Strategy (RevenueCat-Powered) 💰

Simmer uses a freemium subscription model designed for fast activation and sustainable scaling.

Free Tier

  • Basic pantry management
  • 3 AI imports to start
  • Basic shopping list

Simmer Pro (Entitlement: Simmer Pro) 💎

  • Unlimited AI imports
  • Unlimited Pantry Camera Scans
  • Full “What Can I Cook?” engine

Challenges We Faced 🧩

1. Recipe Parsing Variability

Recipes differ wildly in structure. We refined prompts and enforced strict schema validation to normalize outputs.

2. Pantry Matching Complexity

Ingredient naming inconsistencies required fuzzy matching and canonical mapping to avoid false negatives.

3. Balancing Power with Simplicity

Early versions were feature-heavy. We refocused around one hero flow:

Import → Pantry Check → Grocery List → Cook 🍳

Everything else supports that loop.

What We Learned 📚

  • AI must feel invisible.
  • Monetization should shape architecture early.
  • Friction, not inspiration, is the real cooking bottleneck.
  • Completion is more valuable than discovery.

Built With

Share this project:

Updates

Submission history