# Sophisticated Controllable Agent for Complex RAG Tasks

An advanced retrieval-augmented generation agent that uses a deterministic graph to answer complex questions over a user's own data that simple similarity search cannot solve.

Controllable-RAG-Agent implements a retrieval-augmented generation solution built around a deterministic graph that acts as the reasoning core of a highly controllable, autonomous agent, aimed at complex questions that plain semantic-similarity retrieval fails on. The graph structure lets the agent's behavior be inspected and steered rather than left as an opaque chain of model calls, so it can answer non-trivial questions grounded in a user's own documents.

The repository is a Jupyter Notebook project and is distributed alongside the author's separate paid book and course on retrieval-augmented generation techniques and production AI systems. It is licensed under Apache 2.0 and depends on LangChain and LangGraph for its underlying graph and retrieval components.

Website: https://github.com/NirDiamant/Controllable-RAG-Agent