7 search types for your RAG pipelines and LLM apps

Published: August 11th 2025

LLMs are frozen at training time

So to answer even a simple question like

"What is the weather like in my city today",

they need access to external information.

That is where RAG (Retrieval Augmented Generation) comes in.

RAG is not just about vector search and embeddings, like it was widely associated with in 2023

RAG is about getting data to pass to your LLMs context windows so that they can return useful and accurate answers.

That data can come from web sites, pdfs, your database, and the files and folders on your computer. Basically any external source of data.

Here are 7 search methods you can use to retrieve data

→ Web search (e.g google to get current weather info) → File search (e.g coding assistants to read your source code) → Database search (to answer questions about your business data) → Grep search (a fast text based search to find file and folder contents using keyword search) → Directory search (list files and folder in a directory) → Semantic search (search against vector embeddings of chunked data) → Keyword Search (search using keyword matching against text)

Get your LLM the right data, so that it can give you useful results.