What Are Vector Databases?

Regular Databases = Rows and Columns

Imagine a spreadsheet that keeps track of all the student clubs at Champlain College:

id	name	category	room
1	Anime Club	Media	C-102
2	Art Club	Creative	C-103
3	Robotics Team	Tech	C-108
4	Champlain Music Society	Creative	C-109

With this kind of relational database, you can basically ask:

• “Show me all Creative clubs.”
• “Which club is in room C-108?”
• “List all clubs with ‘Art’ in the name.”

But if you ask:

“Which club is most similar to the Robotics Team?”

A regular database would shrug. It doesn’t understand similarity unless you explicitly define it.

---

Vector Databases = Grouping by Meaning

Let’s now imagine that each student club has its own room on the main floor of the college. Here’s what happens:

• All students who love drawing and painting go to the Art Club room.
• All students who are into robots and programming gather in the Robotics room.
• All who enjoy anime and manga meet in the Anime Club room.

So if you walk into the Art Club, you’ll find similar people,those with shared interests.
They’re not just grouped by a label, they’re physically close in the space of the college.

Now think of vector databases the same way.

Instead of rooms, vector databases have a multi-dimensional space (think of a map with hundreds of directions). Each item, like a word, image, product, or document, is converted into a list of numbers called a vector. That vector is like coordinates that say:

“Put this word here in space.”

So:

• “Food” gets placed in a space near “Lettuce”, “Burger”, and “Restaurant”
• But “Food” is far away from “Foot”, even though they sound alike

Why? Because they don’t mean the same thing.

Image credit:
Source: “An Introduction to Vector Databases for Beginners”
© Xomnia — Used for educational purposes only.

---

How Does This Help?

When you search a vector database, it looks at the “location” of your query in space and finds things that are nearby, just like asking:

“I’m interested in robotics. Which room should I go to?”

The system says: “Go to Room C-108 with the Robotics Team. Oh, and the Tech Club and AI Club are just next door.”

So, you’re not just matching words, you’re matching meaning.

---

Summary

• Relational databases are like spreadsheets: great for exact info and filters
• Vector databases are like a campus map: they organize data by meaning and group similar things near each other
• Just like how Champlain has rooms for each club interest, vector databases have spaces for each concept
• This lets AI answer fuzzy questions like: “Show me stuff like this” or “Find things with similar vibes”