M365Con 2027 – free online conference · 11–22 Jan 2027 · 120 sessions m365con.net

Qdrant

Product

Qdrant is an open-source vector database and similarity search engine written in Rust, used to store embeddings with payload metadata for filtered semantic search and retrieval-augmented generation.

◆ PRODUCTQdrantM365.FM
Title graphic: Qdrant (Product)

Also discussed

The episode about Qdrant also covers Cybersecurity Maturity Model Certification (CMMC), Llama, Ollama, Graphics processing unit (GPU), ISO/IEC 27001 and Vector store.

Description

Qdrant is an open-source vector database and similarity search engine written in Rust. It stores embeddings, which are numeric vector representations of content such as text, together with payload metadata, and it returns the items most similar to a query vector. Typical uses are semantic search and retrieval-augmented generation (RAG), where an application looks up relevant source material and passes it to a language model as context.

In a RAG pipeline, Qdrant acts as the vector store. Documents are ingested, converted into embeddings and written to Qdrant. At query time, the question is embedded in the same way and the closest matches are retrieved. A Microsoft Learn page on building RAG pipelines with Qdrant and Azure Files describes this setup with Azure Files as the document source. Orchestration frameworks such as LangChain, LlamaIndex and Haystack handle ingestion and retrieval at query time, so Qdrant itself does not need to be called directly by every part of the application.

A central feature is the payload. Qdrant can store arbitrary JSON-serializable metadata alongside each vector and filter on it when a query runs. According to the Microsoft Learn page, this filtering happens during HNSW graph traversal instead of as a post-filter, so search performance is not reduced by the filter. This matters when a search must be restricted, for example to documents with certain attributes, while still being ranked by semantic similarity. Collections are created automatically on the first insert, with a fitting vector dimension and distance metric.

Deployment options are a decision point. Qdrant can be self-hosted, for example on Azure Kubernetes Service together with Azure Files private endpoints, which keeps the pipeline inside the tenant's network boundary. Alternatively, Qdrant Cloud offers a managed service with the same API as the self-hosted version, and according to the page a free tier suffices for development and testing. Qdrant is also available through the Azure Marketplace, which allows billing through an Azure subscription. The documentation page referenced here does not state licensing terms, capacity limits or performance figures, so those points need to be checked separately when planning a project.

In the Microsoft ecosystem, Qdrant appears next to services such as Azure Files, Azure Kubernetes Service and the Azure Marketplace, and it can be combined with .NET-based applications that use semantic search. The M365.FM episode "Build AI Applications with .NET 10" (2025-11-03) places Qdrant, together with Weaviate, in the infrastructure layer of an AI application architecture under "Semantic Search & Vector Databases". In that layer it sits alongside data management with Entity Framework Core and Cosmos DB and CI/CD plus MLOps with GitHub Actions. The episode treats it as one building block of a larger stack rather than a standalone product, and it describes AI features such as smart suggestions and voice input as part of the application's intelligence layer.

For teams choosing a vector store, the practical questions are where the data may be hosted, whether metadata filtering is needed alongside similarity ranking, and whether a managed or self-operated service fits operational capacity. The sources describe Qdrant as covering both hosting models with the same API, which allows development on a managed free tier and a later move to a self-hosted deployment within a private network, though the sources do not describe a migration process.

Overview compiled with AI from M365.FM episodes and the facts on this page. You can improve it with the edit button below.

TypeProduct

Episodes about Qdrant

Episodes in which Qdrant is discussed:

Frequently asked questions

Questions marked with AI were written with AI from the facts on this page.

What is Qdrant used for? (AI)

Qdrant is a vector database and similarity search engine used to store embeddings with payload metadata for filtered semantic search and retrieval-augmented generation. In a RAG pipeline it serves as the vector store that returns the most similar content for a query.

How does Qdrant handle metadata filtering? (AI)

Qdrant stores arbitrary JSON-serializable metadata as a payload and can filter on it when a query runs. According to the Microsoft Learn page, filtering happens during HNSW graph traversal instead of as a post-filter, so search performance is not reduced.

Can Qdrant be run on Azure? (AI)

Yes. It can be self-hosted, for example on Azure Kubernetes Service with Azure Files private endpoints, or used as the managed Qdrant Cloud service. It is also available through the Azure Marketplace, which allows billing through an Azure subscription.

Glossary

Community wiki

Short info above, community knowledge below: add practical tips, examples or corrections. New contributions are reviewed first; trusted members publish directly.

Experts and companies for this topic

Nobody has listed this topic yet. Add it to your profile.

Sources

Maintained by Mirko Peters (Microsoft MVP) · Last updated: