D&D Rules RAG Pipeline

A Retrieval-Augmented Generation (RAG) pipeline using LangChain 1.0 to answer questions about D&D rules from markdown documents processed via olmOCR.

Features Covered

• Document Indexing: Load markdown files, split into chunks, and store embeddings in Qdrant.
• RAG Agent: Flexible multi-step queries using create_agent with a retrieval tool.
• RAG Chain: Fast single-call Q&A using middleware for context injection.
• LangSmith Tracing: Built-in observability for debugging and monitoring.

Pre-Reqs

You'll need to ensure you have uv installed before proceeding.

# Install dependencies
uv sync

You will also need an OpenAI API Key (OPENAI_API_KEY) and optionally a LangSmith API Key (LANGSMITH_API_KEY) for tracing.

Preparing Documents with olmOCR

Before running the RAG pipeline, convert your PDF rulebooks to markdown using olmOCR:

# Pull the Docker image (large, includes the model, ~30GB)
docker pull alleninstituteforai/olmocr:latest-with-model

# Convert PDFs to markdown
docker run --gpus all \
  -v "$(pwd)":/workspace \
  alleninstituteforai/olmocr:latest-with-model \
  -c "python -m olmocr.pipeline /workspace/output --markdown --pdfs /workspace/PDFs/*.pdf"

Place the generated markdown files in the PDFs/ directory.

Running the Example

1. Launch the Notebook:

   uv run jupyter notebook RAG_Pipeline.ipynb

2. Run the Cells:

   • Set up your environment and API keys.
   • Index your D&D rulebook markdown files into Qdrant.
   • Test the RAG Agent for flexible, multi-step queries.
   • Test the RAG Chain for fast, single-call Q&A.
   • Use the interactive demo to ask your own questions!

Project Structure

• RAG_Pipeline.ipynb: The main interactive RAG pipeline notebook.
• PDFs/: Directory containing markdown files (converted from PDFs via olmOCR).
• pyproject.toml: Dependency management.

Key Components

Component	Purpose
DirectoryLoader	Load markdown files from disk
RecursiveCharacterTextSplitter	Split documents into retrievable chunks
QdrantVectorStore	Store and search embeddings
@tool decorator	Create retrieval tool for agent
create_agent	Build LangChain 1.0 agent
AgentMiddleware	Inject context for RAG chain

Credits

Built using LangChain, LangGraph, and olmOCR.