qdrant: A vector similarity search engine and vector database for AI applications.
What problem it solves
AI applications need to store, search, and manage embeddings together with the metadata needed to refine results. They also require flexible filtering, hybrid retrieval, and deployment options for different application environments.
Key features
Flexible vector search — Search dense vectors, sparse vectors, and multivectors, including objects represented by multiple embeddings or late-interaction models.
Source: README
Payload filtering — Attach JSON payloads to vectors and combine keyword, full-text, numeric, and geographic conditions with should, must, and must_not clauses.
Source: README
Hybrid search — Combine multiple vectors in one query and merge results through configurable fusion strategies, including Reciprocal Rank Fusion and Distribution-Based Score Fusion.
Source: README
Distributed deployment — Run a horizontally scalable deployment with sharding, replication, and collection updates or resizing that do not require downtime.
Source: README
Edge operation — Use the lightweight in-process edition on edge devices and in resource-constrained environments, with local storage and optional synchronization with a Qdrant server.
Source: README
Quick start
Copied verbatim from the project's README.
docker run -p 6333:6333 qdrant/qdrantfrom qdrant_client import QdrantClient
client = QdrantClient(url="http://localhost:6333")Who it may suit
Inference: may suit developers building semantic search, recommendation, recommendation-by-example, faceted search, or other vector-based AI applications. It may also suit teams that need a client-server deployment or a locally integrated edge deployment.
Topics to discuss
Which search and fusion setup works best for a particular AI application?
How should payload filters and indexes be designed for complex queries?
When should a team choose a client-server deployment, Qdrant Edge, or a managed deployment?
Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.
Read the original README https://github.com/qdrant/qdrant#readme
ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.
Comments
…