From the course: Vector Databases in Practice: Deep Dive

Recap: Building a vector database

From the course: Vector Databases in Practice: Deep Dive

Recap: Building a vector database

“

- [Instructor] We've done a lot. We've learned how to create structure and add data to a vector database. Let's pause here to recap the steps so far, and also discuss what you can do with the database. First, you started up a new database instance with (indistinct) eight cloud services. This created a server that you or any other user can communicate with, given of course, the right authentication details. And then, we created a blueprint for our data. This meant creating the collections for our data with each collection definition, including the data structure like which properties and references to store. It also included how the data would be managed, like configuring modules for integrated vectorization and retrieval augmented generation. We then imported the data into our database instance according to the collection definition. In other words, building each object according to that blueprint that we defined. That meant providing the property values. And then you reference data for each object according to the collection definition. And this means that now you can perform any other searches you learned about on your own database. So, here's an example. To run queries on your own database, you can simply replace the connection script from using the connected demo DB function to connect to my DB function as I've done here. You'll be then able to run any of the existing queries on the database you've just built. And remember that this is possible because as you imported the data, the vector database stored that data as well as creating what's called indexes to help you look through the data efficiently. What we've built so far is obviously a fairly small database, but excitingly, you now have the tools to potentially build a database with millions or even billions of objects. Given of course, the appropriate hardware resources, vector database can scale to these types of numbers while remaining performant, which is really exciting and a differentiating factor in this data rich age. And the skills to populate those larger databases really don't differ from populating a smaller database. Now, in the next section, let's start to put this in something a little bit more user-friendly. What we'll do is to create a friendly interface for our users by building a simple web app. There, you'll see how to create a web app that's powered by a vector database to really take this journey end-to-end.

Contents