NutriFit RAG is an agentic retrieval-augmented generation (RAG) system designed for the nutrition and fitness domain. Unlike a conventional chatbot, this agent actively decides which sources to consult, how to combine information, and when to ask the user for clarification.
Architecture
- Curated knowledge base: Scientific articles on nutrition, training guides, food composition databases and supplementation protocols
- Multimodal retrieval: Semantic search in nutrition documents + structured retrieval in food databases
- Agentic reasoning: The agent decomposes complex queries (“what diet should I follow if I train in the morning and I’m lactose intolerant?”) into sub-tasks, consults multiple sources and synthesizes a coherent answer
- Conversational memory: Maintains conversation context for progressive suggestions and goal tracking
Capabilities
- Personalized nutrition plans based on goals, restrictions and preferences
- Gym routine recommendations adapted to fitness level and available equipment
- Supplementation analysis with interactions and evidence-based dosages
- Progress tracking and dynamic adjustment of recommendations
- Source citation for every suggestion
Technology
- LLM with reasoning and function calling capabilities
- Vector database for nutrition document embeddings
- Automated ingestion pipeline for new sources
- REST API for integration with mobile and web apps
