Smart Recipe Nutritionist: A Machine Learning-Powered Intelligent Recipe Nutrition Assistant

Smart Recipe Nutritionist is a machine learning-based Python command-line application. Its core features include predicting the feasibility of ingredient combinations, querying nutritional components, recommending similar recipes, and generating daily menus. This project demonstrates a complete ML-driven CLI workflow, including automated testing and an object-oriented clean code architecture, making it an excellent reference case from prototype to deployable application.

In daily cooking, there are pain points such as confusion about ingredient pairing, inconvenience in nutritional queries, and lack of personalized recommendations; traditional recipe apps only provide fixed searches. This project aims to solve these problems. Initially developed as a programming course project for School21 Russia, it was later refactored into a professional portfolio. Unlike demo projects, it is a runnable CLI application with a complete test suite and documentation, covering the entire workflow from data preprocessing to model deployment.

Supports two modes: 1. Ingredient Prediction Mode: Input ingredients (e.g., chicken, tomato, garlic) to get a combination feasibility score (based on ML model), nutritional components (local database + optional USDA API), and top 3 similar recipe recommendations; 2. Daily Menu Generation Mode: Automatically generates breakfast, lunch, and dinner menus, including nutritional values, scores, and source links.

Adopts a modular object-oriented design. Ingredient prediction workflow: Parse input → Feature vectorization → Model prediction → Nutritional query → Similarity search → Formatted output. Daily menu generation workflow: Recipe type inference → Nutritional query → Score sorting → Three-meal selection → Formatted menu.

The model is trained using the Epicurious recipe dataset (epi_r.csv). Feature engineering converts ingredients into binary vectors; training is done via scikit-learn with cross-validation to select optimal parameters, and the model is serialized and saved as a joblib file. Nutritional data comes from a local standardized database (ingredient_nutrition_daily_values.csv).

Professional Python project structure: src (core code), notebooks (experiment records), data (datasets), models (pre-trained models), tests (Pytest tests), etc. Code features: Object-oriented design, type hints, comprehensive error handling, automated testing, and detailed documentation.

Current limitations: The recipe dataset is mainly Western cuisine, the nutritional database may lack special ingredients, and similarity search is based on simple overlap. Future improvements: Integrate international recipe data sources, introduce user preference learning algorithms, add web/mobile interfaces, and support user feedback loops to optimize the model.

This project is an excellent example of ML application, demonstrating how to integrate ML models into a practical CLI tool, reflecting professional code quality and engineering practices. It has reference value for developers learning ML projects from prototype to deployment, and its open-source implementation provides an extensible foundation for the community.