wllama

last updated: Jul 07, 2024

https://github.com/ngxson/wllama

Somebody compiled llama.cpp for webassembly; there's a demo you can play with, and use embeddings to compare vectors, once you download an embedding model.

It would be neat to use this with a small model to see if you could do a reasonable job of doing search on a blog using vector embeddings and cosine similarity, all in the browser.

↑ up