This article explains the technical differences between traditional Retrieval-Augmented Generation (RAG) and Agentic RAG. It also provides a high-level overview of the Linux directory structure, detailing the specific purposes of key directories like bin, etc, and var.
Key points
Traditional RAG performs a single retrieval step, while Agentic RAG allows iterative information gathering and tool selection.
The quality of standard RAG is highly dependent on the initial retrieval process, whereas Agentic RAG provides a feedback loop to refine evidence.
The Linux root directory (/) serves as the base, with specific subdirectories designated for system configuration (etc), binary files (bin/sbin), and runtime data (var).
Agentic RAG improves task completion by enabling the LLM to adjust its search strategy based on whether collected information is sufficient.