Towards Data Science is a community-powered publication that showcases work in data science, machine learning and artificial intelligence. Every day newcomers, seasoned researchers and industry practitioners publish tutorials, research notes and real-world case studies that help the field move forward.

Towards Data Science

A practical guide to building an Agentic RAG system on top of traditional SQL databases containing long-text documents, without requiring schema changes or data migration. The architecture uses a ReAct agent that orchestrates two specialized tools: a SQL tool for structured queries and aggregations, and a vector search tool (FAISS) for semantic retrieval. The agent intelligently routes queries through SQL-only, vector-only, or hybrid pathways depending on query type. Key design considerations include proper use of tool docstrings vs. system prompts for agent guidance, and a critical distinction between pre-filtering and post-filtering in vector databases — FAISS performs post-filtering which can cause missed results when top-k global search excludes relevant documents before metadata filtering is applied.

Building Cost-Efficient Agentic RAG on Long-Text Documents in SQL Tables