Qdrant Documentation

What Qdrant is

Qdrant is a vector database: it stores embeddings (lists of numbers that capture the meaning of text, images or other data) and finds the ones closest to a query vector. Each point has a vector and a JSON payload, and you can filter on the payload while searching, for example "documents similar to this question, but only from course CS101 and only in English". It is written in Rust and released under Apache 2.0.

What the docs cover

• Quickstart: running Qdrant locally in Docker, creating a collection, inserting points and making a search.
• Concepts: collections, points, payloads, distance metrics (cosine, dot product, Euclidean) and indexing.
• Filtering: payload conditions that combine with vector search, one of the things a dedicated vector database does well.
• Hybrid and multi-vector search: combining dense and sparse vectors, and fusion of several result lists.
• Quantisation: compressing vectors to save memory, with the accuracy trade-off explained.
• Tutorials: semantic search, RAG with popular frameworks, and recommendation examples.

Is it worth learning?

If you are building AI features beyond a prototype, learning one dedicated vector database is useful, and Qdrant's docs are clear about the trade-offs. But be honest about need: if your data already lives in PostgreSQL and you have up to a few million vectors, pgvector is often enough and saves you a second system. Reach for Qdrant when filtering, hybrid search or memory use become the bottleneck, or when the vector store is a service of its own.

Prerequisites and time

Python or JavaScript, and a basic idea of embeddings from any RAG course. A weekend is enough to index a few thousand documents and build a search endpoint.

Free or paid?

The open source engine is free to self-host. As of October 2026 Qdrant Cloud has a free-forever single-node cluster for testing and prototypes (check the pricing page for the current size); larger managed clusters are paid.

Best for: developers building semantic search, RAG retrieval or recommendations who have outgrown a quick prototype.