Get spun around by our new vector quantization algorithm that utilizes the power of random rotations to improve the speed-quality tradeoff of vector search with Weaviate.
1.32 adds collection aliases for no-downtime collection migrations, efficient & powerful rotational quantization (RQ), GA replica movement, memory reduction with compressed HNSW connections and more!
Design for speed, build for experience.
Master the art of scaling a vector database like Weaviate.
Learn how Weaviate and NVIDIA enable fast, scalable vector search for agentic AI.
Weaviate `1.31` implements the MUVERA encoding algorithm for multi-vector embeddings. In this blog, we dive the algorithm in detail, including what MUVERA is, how it works, and whether it might make sense for you.
1.31 adds MUVERA for multi-vector embeddings, new BM25 operators, the ability to add new object vectors, and more!
Read about BlockMax WAND and multi-vector embeddings in GA, API-based user management, RAG improvements, xAI model support, and more!
How Weaviate achieved 10x Faster Keyword Search and 90% index compression
Read about multi-vector embedding support, improved keyword/hybrid searches, role-based access control and async replication going GA, new nvidia modules, and more.
Read about a preview of role-based access control, improved indexing, Weaviate Embeddings, new Voyage Multimodal model and a Japanese BM25 tokenizer!
Learn about the challenges of filtered vector search and how Weaviate tackles them with ACORN.
