Building a RAG application from PDFs with Docling

Description

In this tutorial, we’ll build a local Retrieval-Augmented Generation (RAG) application using IBM Granite 4 Micro, running entirely on Ollama, and powered by Docling for document parsing. You’ll learn how to: ✅ Run Granite 4 Micro locally — no cloud, no API keys ✅ Use Docling to process PDFs, DOCX, and HTML into structured data ✅ Build a LangChain-powered RAG pipeline with your own knowledge base ✅ Create a simple Streamlit interface to query your documents ✅ Keep your data private, fast, and fully under your control By the end, you’ll understand how to combine these open technologies to create your own secure, explainable, and affordable AI applications. 💡 Why this matters: Granite 4 represents a new class of small, efficient, and trustworthy AI models that can run locally while delivering enterprise-grade performance. When paired with Docling and modern dev frameworks, it enables a new generation of private, production-ready AI workflows.