caiss

Why we included this project

Caiss is a compact similarity search engine you can embed directly in an application when you need fast nearest-neighbor lookups over high-dimensional vectors and don't want to run a separate database service. It exposes a C-style SDK with Python, Java, and C# bindings, so it works across MacOS, Linux, and Windows without locking you into one stack. The bundled training routine adjusts ANN parameters toward a target accuracy, which spares you the usual hand-tuning, and it supports both built-in and custom distance functions. It also covers the full CRUD lifecycle with basic SQL-style operations, plus caching, multithreading, and batch queries, which makes it a practical fit for semantic search over words, sentences, and other embedded content. A Docker image and a demo web page make it easy to try before committing.