Embeddings e busca vetorial

qdrant

qdrant/qdrant

A vector search service combining embeddings, sparse vectors, and metadata filters for retrieval and recommendation.

Abrir repositório oficial ↗ Página inicial do projeto ↗
★ 34,5KEstrelas
⑂ 2,7KForks
702Problemas em aberto
RustIdioma
Apache-2.0Licença
Q90Pontuação editorial

Visão geral

Qdrant stores vectors with business metadata so semantic matching can be constrained by category, time, or tenant. It is a retrieval layer within an application. Evaluate it with representative data, filtered recall, update behavior, and resource use.

Principais recursos

  • Dense and sparse search
  • Payload filtering
  • Hybrid retrieval
  • Quantization and disk storage
  • Distributed deployment

Requisitos, instalação e início rápido

Start a local Docker instance, install a supported client, and select vector dimensions and distance. Bind test ports locally; configure persistent storage and production authentication.

Uso

Index 100 product descriptions with categories. Compare dense and hybrid retrieval against labeled queries, then verify updates and deletes.

How it works
Encode content, store points in a collection, then query with a compatible vector and payload filters. Dense and sparse search can be combined; quantization and disk storage trade resources against retrieval behavior.

Audience and requirements
Search and RAG backend teams. Service runtime, embedding model, memory, and persistent storage.

Practical use cases
Knowledge retrieval; similar-image search; recommendation; tenant-filtered search.

Limitations and selection
Requires an encoder and application layer. Default demo deployment lacks authentication. Filters require server-side authorization enforcement.

Related projects and selection
langchain-ai/langgraph：Complement: an agent decides when to retrieve and retry.

microsoft/graphrag：Comparison: GraphRAG builds graph-derived context; Qdrant supplies vector storage and search.

Source review
Editorial analysis of upstream sources, without runtime or benchmark testing. Proposed workflows are editorial suggestions.

Compatibilidade do modelo e casos de uso

Stores embeddings rather than running a chat model; ingestion and query encoding must be compatible.

Observações sobre licença e riscos

The repository page identifies Apache-2.0. Read LICENSE; model weights and datasets may have separate terms.

Editorial source review 2026-09-09T05:00:00.950Z. README and live repository page verified; current stars/forks from GitHub HTML. Last-push metadata retained from 2026-09-05 discovery snapshot. No runtime benchmark. Integration proposals are editorial analysis.

Lançamento e manutenção

Reviewed 2026-09-09. Counters come from repository pages; features are based on upstream documentation. See Releases in the source links. Editorial analysis of upstream sources, without runtime or benchmark testing. Proposed workflows are editorial suggestions.