Retrieval-Augmented Generation (RAG) pipelines are essential for Large Language Models (LLMs) to access information beyond their training data, improving accuracy and reducing hallucinations by fetching relevant external documents. Evaluating RAG performance is complex, as it's challenging to determine whether issues arise from retrieval or generation. Traditional metrics like BLEU and ROUGE are inadequate for RAG, leading to the development of specialized tools like RAGAS, which assess aspects such as faithfulness and context relevance. This tutorial guides users through setting up a RAG pipeline using LangChain for orchestration and FAISS for vector storage, employing the databricks/dolly-15k dataset for benchmarking. Automated evaluation is integrated using CircleCI, allowing continuous quality assurance by triggering performance checks with each code change. This setup includes using API-based LLMs and embedding models from TogetherAI, with environment variables securely managed in the CI environment. The comprehensive approach ensures RAG systems maintain reliability and performance over time, supporting applications like enterprise document search and critical domains such as healthcare and legal assistance.