Vespa

Vespa is an open-source big data serving engine and hybrid search platform built for massive-scale real-time computation. Engineered with a C++ search and tensor computation core coupled with a Java container layer, Vespa executes multi-phase ranking pipelines combining BM25 lexical search, vector similarity (HNSW), GBDT, and ONNX neural network models directly at query time. It scales across thousands of nodes to serve billions of documents with sub-10 millisecond latencies, supporting real-time mutations, rich tensor expressions, and distributed data grouping.