Create a vector database

Create a Qdrant database with a default collection, a plan, and 3 to 10 nodes. The database becomes Running after a few minutes.

Before you begin

• A TIR project in Delhi or Chennai, and a role with Create permission on Vector Database.
• On a prepaid account, credit for about 24 hours of the nodes’ hourly price. If the credit falls short, TIR refuses the create and the message says how many credits to add.
• To use a private cluster, one of your private clusters. See Allocate nodes to projects.

On E2E Cluster, compute is billed per node, per hour, from the moment the database is created; see Vector Database billing.

Open the create form

In TIR, open your project. In the side menu, under Storage, select Vector Database. Select Create Database, or Launch Database for a project’s first database.

Name the database

TIR fills Cluster Name with a random name that starts with tir-qdrant. Keep it or type your own. A name holds up to 50 characters, starts and ends with a lowercase letter or digit, and uses only lowercase letters, digits and hyphens. The name must be unique in the project.

Set up the default collection

TIR creates one collection for you when the database is ready, with the values you set here.

• Collection name is default-collection until you change it.
• Distance is how similarity between two vectors is measured. Cosine is selected by default. Euclid measures straight-line distance, Dot uses the dot product, and Manhattan adds up the differences along each dimension. Choose the one your embedding model recommends.
• Vector Dimension is the number of values in each vector, at least 1. Set it to the size your embedding model produces.
• Shard Number is how many partitions the collection is split into, at least 1. Setting it to a multiple of your node count spreads the data evenly.
• Replication Factor is how many copies of each shard the cluster keeps, at least 3.
• Sharding method is set to Auto.

Choose a plan

Under Select Plan, choose where the database runs.

• E2E Cluster runs it on E2E’s own capacity. Choose a plan by its price per hour and the RAM and vCPUs each node gets.
• Private Cluster runs it on your own private cluster. Choose the cluster in Private Cluster, then set the CPU and RAM each node gets. If you have yet to create a private cluster, the form says so.

Set the node count and encryption

Node Count is 3 by default; use the slider or the number field to choose any value from 3 to 10. You can add nodes later, and the node count only ever goes up; see Scale up a vector database.

Each node starts with a 30 GB disk. To make it bigger later, see Resize the disk.

Enable Disk Encryption is off by default. Turn it on to secure the disk with AES-256 encryption managed by TIR. A dialog asks you to confirm with Enable Encryption. Encryption is permanent: once the database is created, it stays on, so decide now.

Below the form, an estimate shows about how many million vectors your plan holds for a vector dimension and replication factor you pick there. On E2E Cluster, the hourly, monthly and total price for your nodes also appear; for the rates, see E2E’s pricing page.

Launch the database

Select Launch. TIR returns to the list, where the database shows Deploying and then Running. Select its row and open Overview to see its endpoint and keys.

What to do next

Connect to the database with its endpoint and API key. For help with a message, see Troubleshooting Vector Database.