---- This post is a work in progress ----

What is this?

This is a post about exploring vector similarity search in the browser using DuckDB-WASM. It is a work in progress and I will be updating it as I go. I personally want to learn more about using DuckDB-WASM, but I also think it's useful to explore the utility of vector similarity search in the browser for fun projects, but also for more useful applications.

Why DuckDB?

DuckDB is a lightweight, embedded analytical database that is super easy to use. I've written about using DuckDB here. It also has a WebAssembly version that can run in the browser, duckdb-wasm, which means we can run DuckDB in the browser and use (most) all of its features, such as it's vector similarity search extension, vss. Using this, we can perform vector similarity search in the browser!

What's the goal?

The goal is to add a search bar to this blog such that you can search for sections of posts which are related to the input search term. This will be done by embedding DuckDB-WASM in the browser and using the vss extension to search for the most similar text in the blog. I know there are easier ways to do "search" for this blog, but I think this is a fun way to explore DuckDB-WASM and should give some interesting results.