Start Qdrant with Docker and prepare your first vector search

Qdrant is an open source vector database that you can run in a container inside your own network. This tutorial follows the README as far as it goes, which covers the server start and the connection, and marks where it stops.

Niklas Schult4 min read

Illustration: A jar of pebbles holding a compass, with a key and a signpost whose arrows point to similar pebbles.

What you will have and what you need first

At the end you will have a Qdrant server running in a Docker container and a Python client connected to it. That is the base for an internal knowledge search, because the documents and their vectors stay on a machine you control. The Qdrant README describes Qdrant as a vector similarity search engine and database that stores points, meaning vectors with an additional payload.

You need Docker on a machine or virtual server in your network, and Python with pip on the machine where you write code. The README is short on the later steps. It shows how to start the server and how to connect, but it contains no example for creating a collection, inserting vectors or running a search. The fourth section says where to find those steps.

The README also says Qdrant is written in Rust and offers a REST API with an OpenAPI 3.0 specification. For faster production searches it provides a gRPC interface as well. Both interfaces exist next to the client libraries, so a team can choose the access path that fits its existing applications.

Start Qdrant in a container

The README gives one command for a local start. It runs the image qdrant/qdrant and publishes port 6333 to the host, which is the port the Python client uses later.

Shell
docker run -p 6333:6333 qdrant/qdrant

The README warns that this command starts an insecure deployment without authentication, open to all network interfaces. For a first test on a laptop that is acceptable. On a server in the company network, read the security guide that the README links before anyone else can reach the port. The README also points to an installation guide that you should read before production use.

Connect with the Python client

Qdrant offers official clients for Go, Rust, JavaScript and TypeScript, Python, .NET and Java. This tutorial uses Python. The following command installs the client package from PyPI.

Shell
pip install qdrant-client

The next snippet imports the client class and creates a connection to the server you just started. The url parameter points to port 6333 on the local machine. If the server runs elsewhere, replace localhost with its address.

Python
from qdrant_client import QdrantClient

# Connect to the local Qdrant server
client = QdrantClient(url="http://localhost:6333")

Collection, vectors and search: what the README leaves open

The README stops after the connection. It does not show how to create a collection, how to insert points or how to query them, so this tutorial does not invent those calls. The README points to a Quick Start guide, a detailed documentation and a tutorial for building a semantic search engine, and those are the places for the next steps.

It does describe what a search can do. Points carry a JSON payload, and you can filter on it with must, should and must_not clauses. Dense vectors serve semantic similarity, and hybrid queries can combine several vectors. For a knowledge search, the payload is where you would keep details such as the source document, which matters when you later need to cite where an answer came from.

The README lists further features that matter once a prototype grows. Multitenancy partitions data for several users, and sharding with replication lets a deployment scale horizontally. Quantization is described as a way to reduce RAM usage while you tune the balance between speed and precision.

Checking, pitfalls and use in a company

The README names no verification step. A practical check is to confirm that the container is running and that the client call above raises no connection error. The README also mentions a Web UI for exploring collections and managing data, though it gives no address for it here. The main pitfall the documentation names is the open default: no authentication and all network interfaces. It also warns you to read the installation and security guides before production.

In a company, someone has to own the server, its updates and its access rules. Check with your data protection officer which data goes into the vectors, since embeddings of personal or confidential documents remain protected content. Lawyers and tax advisers should also consider section 203 StGB, the German rule on professional secrecy, before indexing client files. Our pages on local AI and security and data protection cover the operating side.

souveraen.ai is a platform for company knowledge that answers from connected sources and cites them, and it runs either hosted in Germany or on-premise. Teams that want to build their own components can use the OpenAI-compatible developer API. A self-built Qdrant setup suits those who want to own the whole pipeline, including the knowledge management layer.

Sources

More articles

All articles

How safe are AI API routes via resellers and routers?

Security firm Dragos sent about two million requests with fake credentials to AI providers and found compromised services and resellers reading customer prompts. Companies that use AI through third parties need to know who sits on the path of their data.

Data protection4 min

Run Text Generation Inference with Docker on Your Server

Text Generation Inference (TGI) serves open language models from a single Docker container on your own hardware. This tutorial starts it, sends requests over HTTP and keeps prompts inside your network.

Tutorial4 min

Cloud outage liability: who answers when Microsoft fails

On 12 November 2026 a software fault took Exchange Online and Outlook offline for about 36 hours. A legal analysis published shortly afterwards argues that responsibility for such outages often lies with the operator rather than the provider.

Regulation4 min

Test it with your own documents.

Free for 3 days, then from €24.90 per user per month. Cancellable monthly.

Free for 3 days, then from €24.90 per user per month.Free for 3 days
Try free for 3 days