Feder

This paper introduces the HNSW algorithm, which is widely adopted in vector databases and search engines for its efficient and robust performance on high-dimensional data. HNSW is foundational in powering modern vector search systems.

Computer vision interface for vector search with native integrations for Qdrant, Pinecone, LanceDB, and Milvus. Enables natural language search, configurable vector database backends, and visualization of search matches across billions of images.

AI-ready data visualization platform for massive datasets of embeddings. Atlas enables interactive exploration of millions of vectors in your web browser, with automatic dimensionality reduction and semantic clustering.

Rust-based PostgreSQL extension accelerating vector similarity search with diskless HNSW (20x faster than pgvector), DiskANN indexes, and hybrid SQL for ANN on embeddings. Enables FP16/INT8/binary vectors with ACID compliance; perfect for high-throughput RAG in DB, real-time analytics. Superior speed/resource efficiency vs dedicated vector stores like Weaviate, all within Postgres.

Timescale extension for pgvector introducing StreamingDiskANN for disk-optimized, high-recall ANN search (28x lower p95 latency vs Pinecone), with hybrid SQL+vector capabilities. Supports binary quantization, filtered queries; scales RAG/analytics to billions on existing Postgres without sharding.

Hybrid billion-scale vector search method combining HNSW with inverted file indexes, enabling cost-efficient search by keeping centroids in memory while storing vectors on disk.