Building a Vector Database from Scratch

Building a Vector Database from Scratch I have built a vector database with Rust from scratch to learn how such a database work under the hood. The database is not production-ready, though functional. Building it let me understand the basics of vector database, and I would like to keep it with written words. Introduction Recently LLM-powered tools make a huge difference on how people work, and many of them are based on RAG. As a developer, I want to build something customized and useful with LLM, but struggled with complex concepts like embeddings, RAG, aligning, etc. Building something fundamental of RAG systems, or semantic search can be a good starting point. Retrieval Augmented Generation (RAG) is a context providing layer above LLM, aiming at more precise answers based on information provided by the user. Instead of simply passing the query to the LLM, RAG system will search the related information in the database, combine them with the query, and provide it to the LLM. Searching