The article introduces the concept of agentic RAG, which enhances traditional retrieval-augmented generation (RAG) pipelines by incorporating a reasoning engine into the process. In standard RAG, queries are handled in a linear fashion, leading to potential inaccuracies. Agentic RAG shifts this model into a loop where an AI agent evaluates queries, determining the necessary information and the most suitable tools for retrieval. This approach uses PostgreSQL and n8n to orchestrate the workflow. Setting up an agentic RAG system involves using PostgreSQL for multiple functions, including storing embeddings, managing chat history, and querying document metadata. This simplification avoids the complexity of using multiple databases. The article details the setup of two essential database tables: one for metadata and another for document content. This structure allows for efficient lookups, maintaining connections between chunked vectors and their source files. The workflow begins with a chat trigger or webhook, which initiates the process. The RAG AI agent evaluates user queries and selects the appropriate tools, linking to PostgreSQL for execution. Four essential tools are mentioned: a document selector, a raw content retriever, and a granular lookup tool. This flexibility allows the agent to utilize SQL for straightforward queries while reserving vector searches for more complex semantic retrieval. By integrating these capabilities, the system maximizes retrieval accuracy and reduces unnecessary computations.