Feast, Milvus, and Docling: A Quickstart for RAG

This project demonstrates building a Retrieval-Augmented Generation (RAG) application using Feast. It expands on a basic RAG demo, showcasing how to transform PDFs into LLM-ready text data with Docling, use Milvus as a vector database for embedding storage and retrieval, and perform PDF transformations with Docling during ingestion. Key features demonstrated include online feature retrieval, declarative feature definitions, vector search, handling structured and unstructured context, and versioning/reusability. The project includes sample data, a Python file defining Feast feature views and entities, a YAML file configuring offline and online stores, and two main notebooks: one for PDF text extraction and Parquet storage using Docling, and another for ingesting and managing data with Feast.