This article guides readers through building a practical Retrieval Augmented Generation (RAG) pipeline. It begins by detailing the necessary environment setup, including Python 3.12, Docker, and obtaining a Google Gemini API key. The project is then initialized with a virtual environment and necessary Python packages.To store vector embeddings, the tutorial instructs on running pgvector within a Docker container, configuring environmental variables in a .env file. A clear directory structure is presented, outlining five Python files to be created sequentially. The first script, 01_setup_db.py, establishes a PostgreSQL table with a vector column and enables the pgvector extension, explaining the choice of 768 dimensions for compatibility.Script 02_create_index.py focuses on creating an HNSW index for efficient vector search, with adjustable parameters for different use cases. The ingestion process is covered in 03_ingest.py, where documents are embedded using Gemini and stored in the database, emphasizing the distinct task_type for documents versus queries.Step 4, detailed in 04_search.py, demonstrates how to convert a user query into a vector embedding and perform a semantic search using pgvector to retrieve similar documents. Finally, 05_rag.py integrates these components into a full RAG pipeline. This script retrieves relevant documents based on a user's question, constructs a prompt with this context, and then uses Gemini to generate a grounded answer. Verification steps are provided to confirm the database setup, stored documents, and embedding dimensions.