🧠 QuickBite AI — Smart Meal Planning Assistant

✨ Project Story

QuickBite AI was born out of a simple frustration — the endless cycle of "What should I eat today?" combined with a busy schedule, limited budget, and the desire to eat healthier. I wanted to create something that could think like a nutritionist, act like a meal planner, and adapt like a personal assistant.

The project started as a small experiment in AI-assisted nutrition and quickly grew into a full interactive ecosystem combining:

AI meal generation (via Anthropic Claude)
Real-time ingredient availability checks
Nutrition estimation for manual meals
Daily and weekly stats tracking
Persistent setup and inventory management

💡 Inspiration

As a student juggling long study hours, part-time work, and random meal times, I noticed how hard it was to maintain consistency in nutrition. Most of my peers either skipped meals or relied heavily on fast food. I wanted to change that — with technology.

The inspiration was clear:

"What if your meal planner could understand what you have in your kitchen and help you cook something healthy instantly?"

That question became the foundation for QuickBite AI.

⚙️ How It Works

QuickBite AI is built using Python and Streamlit for the UI, with Claude AI acting as the reasoning engine for recipes and nutrition.

Key Modules

setup.py — Handles user setup (age, budget, goals, dietary needs)
inventory.py — Manages real-time ingredient availability and shopping lists
stats.py — Calculates daily/weekly nutrient totals
meal_history.py — Tracks previously generated meals
shopping_list.py — Summarizes what ingredients users need to buy
app.py — The main control hub, connecting everything together

Tech Stack

Frontend/UI: Streamlit
Backend: Python (JSON-based state management)
AI Integration: Anthropic Claude (structured JSON recipe generation)
Data Source: USDA FoodData Central JSON (for fallback nutrient estimation)
Storage: Local JSON persistence (data/state.json)

Core Features

🧠 AI-Generated Recipes: Structured recipes with time, cost, and nutrients.
🍽️ Manual Meal Analysis: Claude estimates calories and macros for custom meals.
🔢 Dynamic Nutrition Tracking: Auto-updates stats as meals are added or deleted.
🧵 Inventory Awareness: Checks what ingredients you have and highlights missing ones.
📊 Weekly Meal Planner: Organize meals by weekday.
🛒 Auto Shopping List: Instantly compiles what you need to buy.

🤔 What I Learned

Structured AI Outputs Matter: Working with AI models taught me how important it is to define strict JSON schemas. Freeform AI output is nearly impossible to integrate into code without structure.
State Management in Streamlit: Streamlit's st.session_state allowed persistent, multi-page interactions, which taught me how to handle reactive state safely in a UI-driven environment.
Human-AI Collaboration: Claude didn’t just generate text — it became a partner in reasoning about inventory, portions, and ingredient mapping.
UX Simplicity: I learned that the less you make users think, the better the experience. The dashboard layout became clean, modular, and easy to navigate.

🚀 Challenges

AI Reliability: Getting Claude to output pure JSON required a custom cleaning and error-repair layer.
Ingredient Fuzzy Matching: Building an algorithm that can recognize "olive oil" and "oil" as the same thing.
Real-Time Nutrition Sync: Updating daily totals instantly without breaking session state.
Streamlit Limitations: Designing modal popups and interactive dashboards in a linear framework.