Day 1: Vector Search Fundamentals on Qdrant - Vector Database

Points, Vectors and Payloads

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Day 1

Points, Vectors and Payloads

Understanding Qdrant’s core data model is essential for building effective vector search applications. This lesson establishes the precise technical vocabulary and concepts you’ll use throughout the course.

Points: The Core Entity

Points are the central entity that Qdrant operates with. A point is a record consisting of three components:

Unique ID (64-bit unsigned integer or UUID)
Vector (dense, sparse, or multivector)
Optional Payload (metadata)

Distance Metrics

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Day 1

Distance Metrics

After vectors are stored, we can use their spatial properties to perform nearest neighbor searches that retrieve semantically similar items based on how close they are in this space.

The position of a vector in embedding space only reflects meaning as far as the embedding model has learned to encode it. The model and its training objective tell you what “close” means.

Quick rule of thumb

Most users do not need to design a distance metric from scratch:

Text Chunking Strategies

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Day 1

Text Chunking Strategies

So far we’ve talked about points - what they’re made of, and how Qdrant compares them for approximate nearest neighbor search using distance metrics like cosine similarity, dot product, or Euclidean distance.

But none of this matters until we give Qdrant something meaningful to compare. That brings us to the real beginning of the system.

Disclaimer: In this section, we focus on text chunking. Although other types of data (images, videos, audio and code) can also be chunked, we are covering the basics of text chunking as it is the most popular type of data.

Demo: Semantic Movie Search

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Day 1

Demo: Semantic Movie Search

Let’s synthesize everything we’ve learned today into a practical project: a semantic search engine for science fiction movies.

Follow along in Colab:

Project Overview: When Search Understands Meaning

Imagine asking a search engine: “Show me movies about questioning reality and the nature of existence” and getting back The Matrix, Inception, and Ex Machina, but not because these titles contain those exact words, but because the system understands what these films are actually about.

Project: Building a Semantic Search Engine

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Day 1

Project: Building a Semantic Search Engine

Now that you’ve seen how semantic search works with movies, it’s time to build your own. Choose a domain you care about and create a search engine that understands meaning, not just keywords.

Your Mission

Build a semantic search engine for a topic of your choice. You’ll discover how chunking strategy affects search quality in your specific domain.

Estimated Time: 120 minutes