RAG Experiments with LlamaIndex 🚀

Welcome to a hands-on playground for exploring Retrieval-Augmented Generation (RAG) pipelines using LlamaIndex and modern open-source tools!

What is this repo?

This repository demonstrates a progression of RAG workflows, from the simplest to the most advanced, using real-world document parsing, chunking, vector search, and LLM-powered Q&A.

---

Note: Before running any notebook, you must download the sample documents. Run:

bash download_docs.sh

This will fetch the Docling technical report and DSPy paper into the docs/ folder.

🟢 Easy: Quickstart with LlamaIndex

• Notebook: llamaindex.ipynb
• What you'll learn:
  • Basic document ingestion
  • Simple vector search
  • Out-of-the-box LlamaIndex RAG pipeline

---

🟡 Intermediate: Smarter Chunking with Chonkie

• Notebook: llamaindex_chonkie.ipynb
• What you'll learn:
  • Perform Semantic chunking through Chonkie
  • Integrating Chonkie with LlamaIndex

---

🔴 Advanced: Full-Stack RAG with Docling & Custom Pipelines

• Notebook: llamaindex_chonkie_docling.ipynb
• What you'll learn:
  • Parsing PDFs and extracting images/tables with Docling
  • Advanced metadata and provenance tracking
  • End-to-end LLamaIndex RAG with custom chunking (Docling), vector DB (Qdrant), and LLMs (Ollama)

---

🛠️ Technologies Used

• Docling – Document parsing & PDF enrichment
• Chonkie – Chunking
• Qdrant – Vector database
• LlamaIndex – RAG orchestration
• Ollama – Local LLM serving

---

Get Started

1. Clone the repo
2. Install dependencies: pip install -r requirements.txt
3. Start Qdrant (see compose.yaml)
4. Download the sample docs: bash download_docs.sh
5. Open a notebook and start experimenting!

---

Explore, learn, and build your own RAG pipeline!