This RAG-based virtual assistant chatbot delivers precise answers from a  ChromaDB-powered  knowledge base focused on L&T Construction. Users can toggle chat visibility, erase previous conversations, and switch between light and dark themes. The backend APIs  and  frontend interface  were built seamlessly to ensure smooth user experience. Responses include reference links and are stored in a database for future retrieval.  Advanced Retrieval-Augmented Generation (RAG)  techniques were implemented to  minimize data hallucination  and maintain high accuracy. The system is optimised for minimal token consumption , reducing pay-per-token costs, and ensures long-term agent interactions by  managing context window  efficiently, keeping token-per-minute rates low. Prompt engineering techniques like  zero-shot and chain-of-thought prompting  were used to effectively handle irrelevant queries. Additionally, a logging framework using Azure App Insights was integrated for monitoring and debugging.