Vespa

Vespa is an open-source, large-scale vector database and AI search platform originally developed at Yahoo to handle demanding real-time search, recommendation, and retrieval-augmented generation (RAG) workloads. Unlike traditional vector databases that focus solely on semantic similarity, Vespa unifies full-text lexical search, vector similarity, structured filtering, and tensor computations into a single distributed engine. It allows engineering teams to execute complex hybrid queries and run machine learning inference models (such as ONNX or XGBoost and cross-encoders) directly on-node during the query pipeline. Featuring continuous real-time data ingestion and multi-phase ranking, Vespa provides the low-latency infrastructure required by high-scale applications like Perplexity and Spotify to serve billions of documents and process high-throughput queries.