Repolode
Submit
GiovanniPasq

GiovanniPasq/agentic-rag-for-dummies

A modular Agentic RAG tutorial and project built with LangGraph, featuring hierarchical indexing, conversation memory, and query clarification.

View on GitHub
Stars
4,268
Forks
587
Open issues
0
Last commit
3 days ago

What is it?

Agentic RAG for Dummies is a learning resource and extensible project showing how to build a retrieval-augmented generation system where an agent decides when and how to search documents. It is organized as both an interactive notebook for step-by-step learning and a modular project where the LLM provider, embedding model, PDF converter, and agent workflow can each be swapped out.

The indexing pipeline splits Markdown documents into large parent chunks based on headers and small child chunks for search, so queries match precise passages while answers receive surrounding context. At query time, a LangGraph workflow summarizes conversation history, rewrites ambiguous or multi-part questions, and pauses for user clarification when needed. Parallel agent subgraphs handle each sub-query, with self-correction, context compression, and hard limits on tool calls and iterations.

The runnable app defaults to Ollama for local models and uses Qdrant with hybrid dense and sparse (BM25) retrieval. Examples are included for OpenAI, Anthropic, and Google models, and the project also documents Langfuse observability and RAGAS-based evaluation.

Key features

  • 01Hierarchical indexing that searches small child chunks and returns larger parent chunks for context.
  • 02Query rewriting and clarification, including pausing to ask the user for missing details.
  • 03Parallel agent subgraphs that split complex questions into sub-queries.
  • 04Self-correction that re-queries when initial results are insufficient.
  • 05Context compression and conversation summaries to keep prompts lean.
  • 06Support for Ollama locally, with examples for OpenAI, Anthropic, and Google models.

Use cases

  • Learning how agentic RAG workflows are structured with LangGraph.
  • Prototyping a document question-answering chatbot over PDFs.
  • Starting point for a custom RAG system where the model or retrieval layer needs to be swapped.
  • Comparing retrieval and answer quality with RAGAS evaluation.

Quick start

  1. 1

    Install Ollama from the official site

  2. 2

    Pull the default local model

    ollama pull granite4.1:8b
  3. 3

    Install the Python dependencies for the notebook (run the tutorial cells in order)

    pip install -qU langchain-openai

Things to know

  • The runnable app is Ollama-first and needs a local model; smaller models below 8B may hallucinate or ignore retrieval instructions.
  • Cloud providers require your own API key and their usage costs.
  • The README describes the project as a tutorial, so the modular app may need adaptation for production use.

Topics

#rag#langgraph#ai-agents#llm#vector-search#qdrant#ollama#python#agent#agentic-ai

FAQ

What is agentic-rag-for-dummies?
A modular Agentic RAG tutorial and project built with LangGraph, featuring hierarchical indexing, conversation memory, and query clarification. Agentic RAG for Dummies is a learning resource and extensible project showing how to build a retrieval-augmented generation system where an agent decides when and how to search documents. It is organized as both an interactive notebook for step-by-step learning and a modular project where the LLM provider, embedding model, PDF converter, and agent workflow can each be swapped out.
Is agentic-rag-for-dummies free to use? What license does it have?
agentic-rag-for-dummies is open source under the MIT license. Check the license terms before using it in commercial products.
Is agentic-rag-for-dummies still maintained?
Actively maintained. The last commit to agentic-rag-for-dummies was on Oct 17, 2026.
What can I use agentic-rag-for-dummies for?
Learning how agentic RAG workflows are structured with LangGraph. Prototyping a document question-answering chatbot over PDFs. Starting point for a custom RAG system where the model or retrieval layer needs to be swapped. Comparing retrieval and answer quality with RAGAS evaluation.
How do I install agentic-rag-for-dummies?
Install Ollama from the official site Pull the default local model: ollama pull granite4.1:8b Install the Python dependencies for the notebook (run the tutorial cells in order): pip install -qU langchain-openai
Where can I get agentic-rag-for-dummies?
The source code and installation instructions are on GitHub: https://github.com/GiovanniPasq/agentic-rag-for-dummies

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects