Agentic RAG for Dummies
A modular Agentic RAG tutorial and project template built with LangGraph, demonstrating hierarchical retrieval, conversation memory, query clarification and multi-agent orchestration for retrieval-augmented generation.
Agentic RAG for Dummies demonstrates how to build a modular, agent-driven retrieval-augmented-generation system with LangGraph, aiming to bridge the gap between basic RAG tutorials and an extensible architecture. It uses hierarchical indexing, searching small child chunks for precision while retrieving larger parent chunks for context, and adds conversation memory, query clarification for ambiguous questions, LangGraph-coordinated agent orchestration, a multi-agent map-reduce step for decomposing complex queries, self-correcting re-queries, and context compression to keep working memory manageable.
It can be used either as an interactive notebook for learning the core concepts, or as a modular project whose LLM provider, embedding model, PDF converter and agent workflow can each be swapped independently. The runnable application defaults to Ollama and can be adapted to any chat model supported by LangChain, with examples for Anthropic, OpenAI and Google, plus optional observability through Langfuse and evaluation through RAGAS metrics.