AiNutrify: A Multi-Agent AI System for Image-Based Nutritional Analysis and Recipe Generation Using CrewAI

Abstract

The increasing complexity of meal planning and limited access to instant nutritional information creates significant barriers to healthy eating. This paper presents AiNutrify, an innovative AI-driven platform that leverages multi-agent systems and multi-modal large language models to automate dietary assistance. The system employs the CrewAI framework to orchestrate specialized agents for two core workflows: recipe generation from ingredient images and comprehensive nutritional analysis of prepared dishes. Using Google's Gemini model via LiteLLM for multi-modal understanding and Gradio for user interface, the system processes food images to generate personalized recipes compliant with dietary restrictions or provide detailed nutritional breakdowns. Initial validation demonstrates 92% ingredient recognition accuracy, 100% dietary compliance filtering, and 88% recipe quality approval. The modular architecture ensures scalability and transparency, while Pydantic models enforce structured outputs. For nutritional analysis, it evaluates an image of a prepared dish to estimate calories, macronutrients, micronutrients, and provide a health evaluation. The implementation utilizes a Gradio-based interface and Google's multi-modal Gemini model for image understanding. Testing confirms the system's ability to deliver coherent recipe suggestions and detailed nutritional breakdowns, demonstrating the effectiveness of agentic AI in promoting informed dietary choices.