This work introduces a decentralized AI-based system for individualized meal planning and grocery bill minimization. The system uses user-specific dietary needs, local trends and nutritional requirements to create context-dependent meal calendars. Fundamental personalization and filtering are done locally through light machine learning models and rule-based engines, while natural language meal plan synthesis is done through the use of the LLaMA 3 large language model executed locally using Ollama. For improving specialized cost estimation API also estimates daily cost based on parsed ingredients and live prices. We prioritize privacy, low network reliance, and flexibility with an active user feedback loop in the architecture. Experimental evaluation demonstrates high personalization precision, low rule violation rates, and accurate grocery cost prediction (MAE ₹6.8, MAPE 7.9%), along with fast computation times, making the system effective for scalable, privacy-preserving dietary planning.