Skip to content

kelindar/search

Embedded vector search for Go projects

Visit official site

What kelindar/search is

The library is described as a Go package for embedded vector search and semantic embeddings, built on llama.cpp and suited to small and medium projects.

What you can do with it

  • Create a vector index from computed embeddings
  • Save the index to disk and load it later
  • Call shared C libraries from Go through purego

Getting started

  • Download a prebuilt binary or build from source
  • Initialise a model from a GGUF file
  • Generate embeddings for input text
  • Add vectors to a new index

Pros & cons

✓ What we like

  • No cgo simplifies deployment and cross compilation
  • An index can be saved to disk and loaded later

! What to watch out for

  • BERT models must use the GGUF format
  • The card targets datasets under 100,000 entries

FAQ

Which model format is supported?

The material states BERT models are supported when they use the GGUF format.

Is GPU acceleration available?

The feature list names prebuilt binaries compiled with Vulkan GPU support.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review