Struggling to decide what to cook with the ingredients you already have? We've got you covered!
WhatToCookToday is an AI-powered recipe recommendation system that helps you discover delicious recipes based on the ingredients you have in your kitchen. Simply tell us what's available, and we'll instantly suggest personalized recipes you can make right now.
Food Waste & Meal Planning Challenge:
- 36% of food waste occurs at the household level due to poor meal planning
- Users struggle to utilize existing ingredients effectively
- Traditional recipe search requires exact ingredient matching
Our Solution:
- Semantic ingredient understanding that recognizes ingredient relationships and substitutions
- Context-aware recommendations based on what users actually have available
- Reduces food waste by maximizing utilization of existing ingredients
- Simplifies meal planning with instant, personalized recipe suggestions
Food.com Recipe Dataset:
- Source: Food.com Recipes and Interactions Dataset from Kaggle
- Size: 180K+ real-world recipes with comprehensive metadata
- Content: Recipe titles, ingredients, instructions, ratings, nutritional information, and user interactions
- Processing: Extensive data cleaning, normalization, and LangChain-compatible document conversion
- Quality Assurance: Filtered for completeness, removed duplicates, and standardized formatting
Technical Performance:
- Search Accuracy: Semantic embeddings enable context-aware matching beyond keyword limitations
- Response Time: Sub-second recipe retrieval through optimized vector search
- Resource Efficiency: CPU-compatible model optimized for low-resource environments
- Scalability: Efficiently handles large-scale dataset with ChromaDB vector operations# 🍳 WhatToCookToday
Struggling to decide what to cook with the ingredients you already have? We've got you covered!
WhatToCookToday is an intelligent recipe recommendation system that tackles food waste and meal planning challenges by helping users discover recipes based on their available ingredients. Built with Retrieval-Augmented Generation (RAG) and powered by 180K+ real-world recipes from the Food.com dataset, this system delivers personalized cooking suggestions through advanced semantic search and natural language processing.
- Smart Ingredient Recognition: Advanced semantic understanding of natural language ingredient descriptions
- Massive Recipe Database: 180K+ real-world recipes from Food.com dataset
- Context-Aware Search: Dense vector search using semantic embeddings, eliminating keyword matching limitations
- RAG Implementation: Retrieval-Augmented Generation with CPU-optimized models for low-resource environments
- Real-time Processing: ChromaDB-powered vector database for instant recipe matching
- Beginner-Friendly Output: Clear, step-by-step cooking guidance with markdown formatting
- Responsive Design: Gradio-based interactive UI optimized for seamless user experience
- Food Waste Reduction: Helps minimize waste by suggesting recipes for existing ingredients
Try the live application: WhatToCookToday on Hugging Face Spaces
- Frontend: Gradio with custom CSS styling for interactive UI
- AI Model: OpenAI GPT-4o-mini optimized for CPU-compatible, low-resource environments
- RAG Architecture: Retrieval-Augmented Generation for context-aware recipe recommendations
- Vector Database: ChromaDB with dense vector search capabilities
- Embeddings: HuggingFace sentence-transformers/all-MiniLM-L6-v2 for semantic understanding
- Data Processing: LangChain for document conversion and data preprocessing
- Dataset: 180K+ recipes from Food.com with comprehensive data cleaning and normalization
- Database Storage: Google Drive integration for ChromaDB persistence
- Deployment: Hugging Face Spaces for scalable cloud hosting
-
Clone the repository:
git clone https://github.com/magnifiques/WhatToCookToday.git cd WhatToCookToday -
Install dependencies:
pip install -r requirements.txt
-
Set up environment variables: Create a
.envfile in the root directory:OPENAI_API_KEY=your_openai_api_key_here DRIVE_FILE_ID=your_google_drive_file_id_here
-
Run the application:
python app.py
The app will be available at
http://localhost:7860
OPENAI_API_KEY: OpenAI API key for recipe generationDRIVE_FILE_ID: Google Drive file ID containing the ChromaDB database
The application automatically downloads and extracts the ChromaDB from Google Drive on first run. The database contains pre-indexed recipe data for fast retrieval.
Simple ingredients:
"I have tomato and pasta, what should I cook?"
Multiple ingredients:
"I have chicken, potatoes, and onions, what can I make?"
Pantry staples:
"I have rice and beans, any recipe ideas?"
User Input → Semantic Processing → Dense Vector Search (ChromaDB) →
Recipe Retrieval (180K+ Food.com recipes) → RAG Processing (GPT-4o-mini) →
Formatted Recipe Output
RAG (Retrieval-Augmented Generation) Pipeline:
- Data Preprocessing: 180K+ Food.com recipes cleaned, normalized, and converted to LangChain-compatible documents
- Semantic Embeddings: HuggingFace transformers create dense vector representations of recipes
- Vector Storage: ChromaDB stores and indexes recipe embeddings for efficient similarity search
- Retrieval: Context-aware search matches user ingredients to relevant recipes without keyword limitations
- Generation: CPU-optimized GPT-4o-mini formats retrieved recipes into beginner-friendly instructions
Key Technical Advantages:
- Semantic Understanding: Goes beyond keyword matching to understand ingredient relationships
- Scalability: Efficiently handles large-scale dataset with optimized vector operations
- Resource Efficiency: CPU-compatible implementation suitable for various deployment environments
- Real-time Performance: ChromaDB enables sub-second recipe retrieval and matching
app.py: Gradio-based interactive UI with custom styling and natural language input supportgenerate_results.py: RAG implementation with OpenAI integration for context-aware recipe formattingchroma_utils.py: ChromaDB operations, vector retrieval, and semantic search functionality- Database: Pre-processed and indexed 180K+ Food.com recipes with semantic embeddings
- Data Pipeline: LangChain-compatible document processing and normalization system
- Retrieval-Augmented Generation combines the best of both search and generation
- Context-aware processing ensures recipes match available ingredients precisely
- CPU-optimized models enable deployment in resource-constrained environments
- Semantic search understands ingredient relationships and cooking contexts
- Large-scale dataset handling with 180K+ Food.com recipes
- LangChain integration for seamless document processing and conversion
- Data normalization ensures consistent recipe formatting and quality
- Vector embedding optimization for efficient similarity search operations
- Natural language processing supports conversational ingredient input
- Real-time streaming responses with progress indicators
- Mobile-responsive design optimized for all device types
- Intuitive interface designed for seamless cooking workflow integration
Contributions are welcome! Please feel free to submit a Pull Request.
- Fork the repository
- Create your feature branch (
git checkout -b feature/AmazingFeature) - Commit your changes (
git commit -m 'Add some AmazingFeature') - Push to the branch (
git push origin feature/AmazingFeature) - Open a Pull Request
This project is licensed under the MIT License - see the LICENSE file for details.
- OpenAI for the GPT-4o-mini model
- Hugging Face for embeddings and hosting
- ChromaDB for vector database capabilities
- Gradio for the intuitive web interface
Created with 🤎 (and a mixture of mathematics, statistics, and tons of calculations 👩🏽🔬) by Arpit Vaghela
- GitHub: @magnifiques
- Feel free to reach out for questions or collaborations!