Vespa

Vespa, developed by Vespa.ai, is a platform that integrates data and AI capabilities for online applications. According to the vendor, it offers a wide array of query capabilities, a robust computation engine, seamless operability, efficient data management, and comprehensive application development support. The product is designed to cater to businesses of all sizes in various industries, including e-commerce, media and publishing, financial services, telecommunications, and travel and hospitality.

Key Features

Query capabilities in Vespa: According to the vendor, Vespa supports querying by vectors, structured data, and text. It enables nearest neighbor, approximate (ANN) or exact querying with various distance metrics for vectors. For structured data, Vespa allows exact, substring, and regular expression querying, numerical range querying, geo-distance querying, and predicates. Additionally, Vespa supports full-text querying with tokenization, stemming, and positional information.

Computation engine: The vendor states that Vespa's computation engine empowers users to define mathematical expressions over scalars and tensors. Computation can be written manually or imported from machine learning tools such as TensorFlow, LightGBM, XGBoost, or ONNX-compatible tools. It can be performed on document fields, query values, application package data, or built-in features within Vespa. Vespa supports both sparse and dense tensors, enabling a wide range of computation on collections of numbers. Multi-phase inference allows for allocating more computational power to the best candidate items by utilizing a simpler model for initial selection and a more advanced model for the top candidates.