Qdrant

Qdrant is a cutting-edge vector similarity search engine and vector database designed for the next generation of AI applications. It provides a robust and production-ready service with an intuitive API for storing, searching, and managing points—vectors with additional payload. Qdrant is built with Rust 🦀, ensuring high performance and reliability even under heavy load. For benchmarks, visit here.

Qdrant excels at vector similarity search, making it ideal for use cases involving neural-network or semantic-based matching, faceted search, and more. It supports extended filtering, allowing for complex query capabilities.

Built in Rust, Qdrant leverages the language's efficiency and safety features, providing exceptional performance and reliability. It uses modern CPU architectures and asynchronous I/O to maximize throughput.

Qdrant can attach JSON payloads to vectors, enabling both storage and filtering based on payload values. It supports various data types and query conditions, including keyword matching, full-text filtering, numerical ranges, and geo-locations.

Qdrant introduces support for sparse vectors, allowing for efficient keyword-specific searches. Sparse vectors extend traditional BM25 or TF-IDF ranking, enhancing transformer-based neural network capabilities.

Qdrant offers built-in vector quantization to reduce RAM usage by up to 97%. It dynamically balances search speed and precision, making vector search more resource-efficient.

Qdrant supports horizontal scaling through sharding and replication. It ensures zero-downtime rolling updates and seamless dynamic scaling, providing robustness for production environments.

Qdrant is available as a fully managed cloud service, including a free tier, offering convenience and scalability for users without the need for infrastructure management.

To start using Qdrant with Python, install the client:

pip install qdrant-client

Create an in-memory or persistent Qdrant instance:

from qdrant_client import QdrantClient
qdrant = QdrantClient(":memory:")  # For testing
client = QdrantClient(path="path/to/db")  # For persistent storage

Run Qdrant locally using Docker:

docker run -p 6333:6333 qdrant/qdrant

Connect to the Qdrant instance:

qdrant = QdrantClient("http://localhost:6333")

For production deployment, refer to the installation and security guides.

Qdrant offers official and community-maintained client libraries for various languages:

• Official:
  • Go client
  • Rust client
  • JavaScript/TypeScript client
  • Python client
  • .NET/C# client
  • Java client
• Community:
  • Elixir
  • PHP
  • Ruby

Explore the capabilities of Qdrant through various demo projects:

• Semantic Text Search: Deploy a neural search using pre-trained networks.
• Similar Image Search - Food Discovery: Use visual search for food discovery.
• Extreme Classification - E-commerce Product Categorization: Use similarity learning models for product categorization.

For more solutions, visit the Demo Projects section.

Qdrant provides REST and gRPC APIs for integration:

• REST API: OpenAPI Documentation
• gRPC API: gRPC Documentation

Qdrant integrates seamlessly with various tools and platforms:

• Cohere
• DocArray
• Haystack
• LangChain
• LlamaIndex
• OpenAI - ChatGPT retrieval plugin
• Microsoft Semantic Kernel