This thesis project concerns the creation of an AI-powered chatbot for answering customer questions. The project is based on a Retrieval-Augmented Generation (RAG) architecture, leveraging language models (LLMs) to deliver accurate responses enriched by relevant context retrieved from external, user-provided knowledge sources. The RAG chatbot utilizes a history-aware retriever with an LLM to reformulate user queries, thus ensuring meaningful and accurate retrieval from the context database, which includes scraped documents and pre-processed knowledge chunks. The current implementation employs modern techniques in LLMs, such as Langchain, Groq, and embedding models from HuggingFace for efficient vector representation. A dynamic document retrieval chain is combined with a conversational interface to deliver responses tailored to user input. Evaluation of the RAG chatbot's LLMs was conducted using the RAGAS library, focusing on various relevant metrics provided by the framework. The results demonstrate that the proposed RAG-based chatbot can effectively balance retrieval precision and generative capabilities, providing a robust solution for customer question answering.