Once your documents are turned into embeddings, you need somewhere to keep them and a fast way to find the ones closest to a question. That is the job of a vector database. Instead of matching exact values like a traditional database, it returns the vectors that are most similar to a query vector.
The lessons start with a short explainer on what a vector database is and how embeddings and indexes fit together. Two videos then explain the key trade-off behind fast search: exact versus approximate nearest neighbor search and how HNSW indexes find close matches without checking every vector. After that you will get hands-on with ChromaDB, a beginner-friendly option, and a longer tutorial compares it with Pinecone and Weaviate.
Watch the lessons in order. The HNSW deep dive in lesson 3 is optional, so feel free to skip it for now and come back once you have built something with a vector database.
A short explainer on what a vector database is: how embeddings represent data, why a normal database is a poor fit for similarity search, and how indexes make it fast.
Video 4minThe trade-off at the heart of vector search: checking every vector for the exact closest matches versus using an approximate HNSW index that is much faster and usually good enough.
Video 6minAn optional deeper look at how HNSW works and why it is a faster alternative to a brute-force k-nearest-neighbor search.
Video 7minA hands-on start with ChromaDB: installing it, creating a collection, adding data, and querying it with semantic search.
Video 11minA longer tutorial that works with three popular vector databases, so you can compare a local option with hosted ones and see how the same ideas carry across tools.
Video 1h 55min