he goal was to develop an intelligent chatbot that could efficiently handle client inquiries, reducing support dependency while ensuring quick and accurate responses.

The chatbot needed to manage a wide range of queries effectively, improve response accuracy over time, and maintain security while handling large datasets with high performance and scalability.

PDF documents were processed to extract FAQs, which were vectorized and stored in ChromaDB. The chatbot followed the Retrieval Augmented Generation (RAG) framework using Langchain for precise responses. Initially powered by OpenAI’s GPT-3.5-turbo, it was later upgraded to GPT-4o-mini for better performance and lower latency. It was deployed as a FastAPI-based API, hosted on Amazon EC2 in Docker containers, and secured using JWT authentication. Query logs were stored in MySQL for continuous improvement.

The chatbot significantly reduced support queries, enabling instant resolution and freeing up resources. Users reported improved satisfaction, and the system's efficiency continued to grow with daily interactions, further reducing reliance on human support.