What's installed on your Qdrant server?
- Qdrant 1.20 with its REST API and the browser dashboard at /dashboard
- An API key, so every request without it is refused
- HTTPS through Caddy on port 443, at an sslip.io address based on your IP
- A persistent Docker volume, so collections survive restarts
- Telemetry turned off
- Firewall with only ports 22, 80 and 443 open; ports 6333 and 6334 (gRPC) stay closed to the outside
What size do you need for Qdrant?
Qdrant's capacity guide shows the sum: raw vector size is the number of vectors x dimensions x 4 bytes, then you add the index and 20 percent headroom. Its worked example, one million 768-dimension vectors with two copies, needs about 7.3 GB of RAM and 10.8 GB of disk, so half of that for a single copy. With fewer vectors, start at 1 or 2 GB. Around a million, look at 4 GB or more. Quantization cuts the RAM you need. There's no graphics card: search and indexing run on the CPU, which matters most while you load data.
| Use | Size | Price | Choose |
|---|---|---|---|
| Starter | 1 GB RAM · 1 vCore · 25 GB NVMe | $2.58/mo | |
| Medium | 2 GB RAM · 1 vCore · 40 GB NVMe | $4.01/mo | |
| Powerful | 4 GB RAM · 2 vCore · 60 GB NVMe | $6.89/mo |
Who is it for?
- Developers building RAG applications or document search
- Product teams working on recommendations or similarity search
- Anyone who wants vector data kept on their own server
- Applications that already run their embedding model as a separate service
What happens after you order?
- Set the size and complete your payment.
- Your server starts and Qdrant installs itself. This took about 3 minutes in our real-server tests.
- Open your server from My servers in the panel. The Overview tab shows the Site address, Admin address and Admin key.
- Open the Admin address and paste the key into the API key field. Create your first collection from the dashboard or through the API.
- In your application, use the Site address with port 443 and send the key in the api-key header.
With every server
- Daily backups included in the price
- Unlimited traffic
- Full root access
- Scale RAM and CPU up or down later, and grow the disk
- Location: Istanbul · IST-1.
Good to know
- Qdrant doesn't create vectors. Your application or provider runs the embedding model, and Qdrant stores and searches what it receives.
- The gRPC port 6334 is closed to the outside. Switch gRPC clients to REST, or use your root access to open the port and the firewall yourself.
- The key in your panel has full access. Keep it out of browser code and use it only in your server-side application.
- This install is a single node with no replication. For high availability in production, Qdrant's documentation recommends at least three nodes.
- Updates are your responsibility. Take a snapshot and read the release notes before you upgrade.
Qdrant questions and answers
How much RAM does Qdrant need?
It depends on how many vectors you have and their dimensions. Going by Qdrant's own example, one million 768-dimension vectors with a single copy need about 3.7 GB of RAM and 5.4 GB of disk. With fewer vectors you can start at 1 or 2 GB, and quantization lowers the RAM you need.
How do I connect my application to a Qdrant server?
Use the value in your panel's Site address row with port 443, and send the Admin key in the api-key header. Some client libraries try port 6333 by default. That port is closed to the outside in this install, so set the port explicitly.
Does Qdrant need a GPU?
No. Qdrant searches and indexes on the CPU, and GPU support is optional. Seturack servers have no graphics card. While you load a large collection, indexing keeps the CPU busy for a while, and searches can feel it during that time.
How do I open the Qdrant dashboard?
Open the link in the Admin address row of your panel and paste the Admin key value into the API key field. From there you can see your collections and run test queries. The Username and Password rows on the same page are for SSH.
How many vectors fit on a Qdrant server?
There's no fixed number. Dimensions, data type, payloads and quantization all change the answer. Start from number of vectors x dimensions x 4 bytes, then add the index and 20 percent headroom. If it isn't enough, you grow the RAM and disk from your panel.
How do I back up my Qdrant data?
Your server is backed up every night, the last 7 days are kept, and that's included in the price. For a copy of a single collection, use Qdrant's snapshot API, and move the resulting file wherever you like.